Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A web-based intervention to support self-management of patients with type 2 diabetes mellitus: effect on self-efficacy, self-care and diabetes distress

A web-based intervention to support self-management of patients with type 2 diabetes mellitus:... Background: Management of diabetes mellitus is complex and involves controlling multiple risk factors that may lead to complications. Given that patients provide most of their own diabetes care, patient self-management training is an important strategy for improving quality of care. Web-based interventions have the potential to bridge gaps in diabetes self-care and self-management. The objective of this study was to determine the effect of a web-based patient self-management intervention on psychological (self-efficacy, quality of life, self-care) and clinical (blood pressure, cholesterol, glycemic control, weight) outcomes. Methods: For this cohort study we used repeated-measures modelling and qualitative individual interviews. We invited patients with type 2 diabetes to use a self-management website and asked them to complete questionnaires assessing self-efficacy (primary outcome) every three weeks for nine months before and nine months after they received access to the website. We collected clinical outcomes at three-month intervals over the same period. We conducted in-depth interviews at study conclusion to explore acceptability, strengths and weaknesses, and mediators of use of the website. We analyzed the data using a qualitative descriptive approach and inductive thematic analysis. Results: Eighty-one participants (mean age 57.2 years, standard deviation 12) were included in the analysis. The self-efficacy score did not improve significantly more than expected after nine months (absolute change 0.12; 95% confidence interval −0.028, 0.263; p = 0.11), nor did clinical outcomes. Website usage was limited (average 0.7 logins/ month). Analysis of the interviews (n = 21) revealed four themes: 1) mediators of website use; 2) patterns of website use, including role of the blog in driving site traffic; 3) feedback on website; and 4) potential mechanisms for website effect. Conclusions: A self-management website for patients with type 2 diabetes did not improve self-efficacy. Website use was limited. Although its perceived reliability, availability of a blog and emailed reminders drew people to the website, participants’ struggles with type 2 diabetes, competing priorities in their lives, and website accessibility were barriers to its use. Future interventions should aim to integrate the intervention seamlessly into the daily routine of end users such that it is not seen as yet another chore. Keywords: Diabetes mellitus, Online systems, Patient self-management, Self-efficacy, Repeated measures modelling, Qualitative methods * Correspondence: yuca@smh.ca Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Postal address: 30 Bond St, Toronto, ON M5B 1W8, Canada Department of Medicine, University of Toronto, Toronto, ON, Canada Full list of author information is available at the end of the article © 2014 Yu et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 2 of 14 http://www.biomedcentral.com/1472-6947/14/117 Background Methods Management of diabetes mellitus is complex, and in- Study overview volves controlling multiple risk factors that may lead to This study consisted of five phases: 1) development of the complications. However, care gaps exist: the Behavioral intervention, 2) feasibility testing, 3) usability testing; 4) Risk Factor Surveillance System has estimated that only refinement of the intervention, and 5) evaluation of the 68% of patients with type 1 or type 2 diabetes had intervention using a cohort study and individual inter- HbA1c measured at least twice in the previous year [1], views. The study protocol and results of the first four despite a recommendation from the American Diabetes phases are reported elsewhere [15,16]. We report here the Association that it be measured at least two to four results of Phase 5. times per year [2]. Given that patients provide most of their own diabetes care, patient self-management train- Diabetes online companion: a web-based self- ing is an important strategy for improving quality of care management intervention [3], particularly in the current era of patient-centred out- The Diabetes Online Companion is a self-contained dia- comes and comparative clinical effectiveness research betes self-management website that was systematically [4]. Patient self-management interventions have demon- developed according to self-efficacy theory. Self-efficacy strated benefits in terms of both quality of life [5] and refers to “beliefs in one’s capabilities to organize and glycemic control [6], but participation is low [7], effect- execute the courses of action required to produce given iveness wanes over time [6], and access to trained pro- attainments” [17]. Randomized controlled trials have fessionals to support self-management is limited [8]. shown that diabetes self-management education pro- Web-based self-management interventions are promis- grams incorporating principles of self-efficacy are associ- ing because they offer ease of access for patients who ated with improvements in knowledge [18], health are computer-literate, and they can be scaled up with lit- behaviours [18,19], self-efficacy [18-20], HbA1c [18-21], tle cost [9]. Web-based media have improved patient weight [18], and microvascular complications [19]. knowledge, the extent of behaviour change, and clinical Our intervention incorporated evidence-based content outcomes for a range of conditions [10]. However, prin- and behaviour-change strategies and followed the princi- ciples of effective education, self-management support, ples of user-centred design [15]. The website had four and behaviour change have not been incorporated into main components: 1) general information (static), 2) tai- current diabetes-related websites [11-13]. Reviews of lored information (interactive), 3) self-monitoring logs existing diabetes websites showed that they presented (interactive), and 4) a blog (interactive) (see Additional file didactic information of variable quality, they required 1 for sample screenshots). We posted a total of 53 blog advanced reading levels, and they followed a static, posts over the intervention period, initially at a frequency newspaper-format display, rather than harnessing the in- of one per week. After four weeks of limited user activity, herent advantages of websites, such as interactive tech- we increased the frequency of blog posts to two per week nology, social support, and problem-solving assistance and added email prompts with each new posting. The [11,13]. A systematic review of electronic diabetes- topics, which covered medical content, diabetes-related related tools found that they had moderate but incon- news items, and practical issues, were selected on the basis sistent effects on a variety of psychological and clinical of our feasibility and usability testing [15]. In addition, par- outcomes, including HbA1c and weight; tools that were ticipants received weekly email reminders to visit the site more interactive tools were associated with continued or complete their self-management trackers, as well as no- website use and greater clinical improvement [10]. In tices of any new content [15]. addition, greater website use was correlated with greater clinical improvements: regular website users had greater reductions in HbA1c compared with intermittent users. Cohort study Although this finding could be a consequence of the Participants healthy user effect [14], addressing usability issues to in- We conducted a single-arm pre-post cohort study. Con- crease the proportion of regular users may increase the secutive series of individuals with diabetes were recruited effectiveness of interventions. from two family practice units and two endocrinology In a previous study, we developed an approach to clinics in Toronto (one each from two academic health address many of these limitations of existing web-based science centres). Those eligible for inclusion were aged ≥ interventions [15]. In the current study, we tested the 25 years with at least one of HbA1c > 7.0% (53 mmol/ impact of this approach on self-efficacy, quality of life, mol), systolic blood pressure > 130 mmHg, low-density- self-care, blood pressure, cholesterol, glycemic control, lipoprotein cholesterol (LDL-C) > 2.0 mmol/L, or body and exercise promotion amongst people with type 2 mass index (BMI) > 25 kg/m . We excluded those who diabetes. had Canadian Cardiovascular Society class 3 or 4 angina, Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 3 of 14 http://www.biomedcentral.com/1472-6947/14/117 did not speak English, were not available for follow-up, or recorded secular events that might have affected our out- had no regular access to the telephone and internet. comes (such as diabetes-related news reports). Sample size calculation Outcomes Using a range of correlations from 0.2 to 0.8, a signifi- Website usage: We analyzed logs for the web server to cance level of 0.05, and a power of 80%, we calculated assess the frequency and duration of specific compo- that a sample of at most 52 participants was required to nents of the intervention [16]. Specifically, we collected detect a change of 0.5 units in self-efficacy score after data for the following variables: duration of use by indi- the intervention (relative to the score before implemen- vidual users, frequency of use, site penetration, most fre- tation). Differences of 0.1 to 0.5 in self-efficacy score quently accessed tools and pages, and patterns of use have been correlated with metabolic control, eating be- over time. haviour, exercise behaviour, and other self-management Patient-centred outcomes: We assessed self-efficacy, our behaviours [22,23]. A formula for paired mean compari- primary outcome, with the Modified Grossman Self- sons was applied [30], and the longitudinal nature of the efficacy for Diabetes Scale, which has moderate to high re- study increased its power [31]. A previous analysis re- liability (Cronbach’s alpha = 0.51 to 0.86; Additional file 1) ported a dropout rate of 20%–51% in studies of self- [22,23]. We selected self-efficacy because not only has it management [32]; we further adjusted the sample size to been validated in predicting and promoting patient behav- account for an expected dropout rate of 40%. iour change, but it also has been demonstrated to improve clinical outcomes [18,20,24,25]. We assessed self-care be- Data analysis haviour with the Summary of Diabetes Self-Care Activities Linear mixed models were used to examine the effect of Measure – Revised [26] and diabetes-specific quality of life the intervention and time (intervention × time inter- with the Diabetes Distress Scale [27]. These patient-based action) on self-efficacy, self-care, and diabetes distress. outcomes were selected because they are relevant mea- We selected these models to accommodate the complex- sures of knowledge use by patients. ities of typical longitudinal data sets for continuous out- Clinical outcomes: We collected data on HbA1c, sys- comes; specifically, they allowed us to account properly tolic and diastolic blood pressure, LDL-C, and weight for both within- and between-participant variability every three months. These outcomes were chosen to in- [33,34] and have been used in previous studies for simi- form the sample size calculations in future trials. lar analyses [35-37]. The models were also adjusted for age, sex, ethnicity, income (above or below Can $30, 000), Data collection education, employment, and health literacy, as each of We obtained data for age, sex, ethnicity, education, self- these variables could affect the study outcomes [21,38,39]. reported health literacy, employment, duration of diabetes, The model examining the self-care outcome was also complications, smoking status, medications, HbA1c, sys- adjusted for interaction terms of the aforementioned tolic blood pressure, LDL-C, weight, current use of and variables with time. No additional interaction terms with comfort with a computer and the internet, self-care score, time were included for other outcome models, because all self-efficacy score, and quality-of-life score at baseline. Out- additional interaction terms examined were non-significant. come data were collected by means of patient-completed To avoid inflation of R , all variables were specified a priori, questionnaires. For the pre- and post-implementation and all interactions were tested simultaneously using a phases, aggregates of patient-completed questionnaires cut-off value of 0.30 [40]. Models were assessed by means were obtained every three weeks for nine months through of residual plots. web-based surveys, resulting in 12 data points for each To assess the potential effect of missing income data phase. Health literacy was measured by a three-item vali- for three of the participants, a sensitivity analysis (imput- dated questionnaire completed by the patients [28,29]. ing income as both high and low) was performed. Miss- HbA1c and LDL-C were collected from medical records ing health literacy data for 15 of the participants were via chart audit. Systolic and diastolic blood pressures were imputed using the mode of the distribution, because measured by the research coordinator and were recorded 95% of the remaining participants were health literate. as the average of three readings. Weight was also measured Linear mixed models were also used to examine the effect by the research coordinator. At the end of the study, each of the intervention and time (intervention × time inter- participant was asked to disclose whether he or she had action) on secondary outcomes. These models were ad- used other web-based interventions and if so, whether justed for age, self-efficacy score, income, ethnicity, and those interventions employed text- or image-based didactic insulin use (for HbA1c and weight only). We also com- materials, interactive technology, or behavioural strategies. pared the effect of the intervention between users and non- To assess for threats to validity from historical effects, we users of the website. Finally, we used descriptive statistics to Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 4 of 14 http://www.biomedcentral.com/1472-6947/14/117 analyze website usage. R software version 2.1.15 was used baseline self-efficacy, self-care, and diabetes distress are for all analyses [41]. reported in Table 1 (Demographic characteristics and baseline values of observational cohort and qualitative Interviews study). Individual interviews were conducted 2 to 21 weeks after completion of quantitative data collection. We used a pur- Website use posive sampling strategy to recruit participants with a The mean number of days on which users logged in dur- range of experiences and characteristics [42] (sex, age, eth- ing the study period was 8.2 days (standard deviation nicity, duration of diabetes, educational attainment, in- 13); the median was three days. The average frequency come) from the broader pool of cohort study participants. of use was 0.7 logins/month, or one visit every 5.8 weeks, We developed a semi-structured interview guide to elicit distributed as follows: non-user: 11 participants (14%); participants’ views regarding the following website features: infrequent user (<2 times/month): 61 participants (75%); acceptability, usability, strengths and weaknesses of the frequent user (>2 times/month): seven participants (9%); intervention, facilitators and barriers to its use, user satis- heavy user (>1 time/week): two participants (2%). Web- faction, and sustainability of use (Additional file 1). We site usage across all users ranged from 4 to 50 logins/ made the website available during each interview, in case week (median 14.5/week), with peaks of 50 logins in the interviewee wanted to show the interviewer something week 10 and 37 logins in week 27. Increased use of the on the website. website during those weeks appeared to be driven by the All interviews were audiotaped and transcribed verba- blog. In general, website use appeared to parallel blog tim [43]. Transcripts were inductively analyzed to iden- use, with users visiting the blog repeatedly during the tify emergent categories and themes using a constant same login or visit (Figure 1). The most-accessed pages comparative approach [44]. Coding was conducted inde- during week 10 were the blog (regarding medication log, pendently by three team members with expertise in supplements, and insulin) (34% of hits) and the blood qualitative research methods (CHY, JAP, SH) [44]. After pressure (8%) and medication (9%) logs. For week 27, coding an initial subset of interviews, a preliminary cod- the most-accessed pages were the blog (regarding foot ing framework was developed on the basis of the emer- care) (44% of hits), “My blood glucose log” (32%), and “7 ging analysis, with discussion and consensus amongst steps to take care of your feet” (3%). the analysts [45]; the framework was then iteratively Overall, the most frequently accessed tools, for both tested and refined with subsequent interviews [44]. The- first-time and return users, were the blog, followed by matic saturation was attained with 21 interviews [42]. “My blood glucose log”, “My medication log”,and “My ac- NVivo software (version 9) was used to assist with data tivity log”. Regarding site penetration, users viewed 6.6 management and retrieval. Techniques to ensure analytic pages per session, spending an average of 5 minutes 43 sec- rigour included use of multiple analysts, negative case ana- onds on the site, and 1 minute 39 seconds per page. lysis, and triangulation of the qualitative findings with the quantitative results [42,44,46]. Triangulation consisted of Blog use 1) examining the interview data through the lens of “effect Within the blog section of the website, there were a total on self-efficacy”, 2) corroborating qualitative findings with of 569 page views by 35 participants over the study period, quantitative data, and 3) interrogating how the Diabetes with peaks at week 10 (54 views), week 27 (43 views), and Online Companion affected self-efficacy [46]. week 30 (53 views), corresponding to blog entries about the medication log, supplements and insulin, and foot and Research ethics kidney care, respectively. A total of 13 comments respond- The study was approved by the Research Ethics Boards ing to the blog postings were submitted by five partici- of St. Michael’s Hospital (reference number 09–091) and pants. These comments took the following forms: 1) Sunnybrook Health Sciences Centre (reference number responding to the blog (agreement or disagreement); 2) 177–2009). All participants gave written and verbal in- requesting help with or providing feedback on the website; formed consent. 3) requesting help with self-management; 4) offering assist- ance, empowerment, and their own solutions (including Results food recipes); 5) self-reporting behaviour change; 6) shar- Cohort study ing responses to medication; and 7) warning others about Of the 98 participants recruited, 81 had complete data col- interactions with health care providers. lection for at least two time points (one before and one after the intervention was implemented) and were in- Use of interactive and static tools cluded in the analysis. The questionnaire response rate for Overall, 47 (67%), 63 (90%), and 43 (60%) of 70 users vis- these 81 participants was 83%. Patients’ characteristics and ited static, interactive, and log pages, respectively, at least Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 5 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 1 Demographic characteristics and baseline values of observational cohort and qualitative study Participants Observational cohort (%, n = 81) Qualitative study (%, n = 21) Sex Male 44 (54%) 9 (43%) Female 37 (46%) 12 (57%) Age (years) 20–39 7 (9%) 2 (10%) 40–59 37 (46%) 7 (33%) 60–79 36 (44%) 12 (57%) > 80 1 (1%) 0 Ethnicity White 50 (62%) 17 (81%) Asian 24 (30%) 4 (19%) African American 6 (7%) 0 Hispanic 1 (1%) 0 Duration of diabetes mellitus (years) < 5 25 (31%) 6 (29%) 5–9 16 (20%) 5 (24%) 10–14 19 (23%) 3 (14%) 15–20 16 (20%) 5 (24%) > 20 years 5 (6%) 2 (10%) Education < High school 1 (1%) (0) High school 11 (14%) 1 (5%) College 21 (26%) 5 (24%) University 48 (59%) 15 (71%) Employment status Employed 45 (56%) 11 (52%) Retired 24 (30%) 7 (33%) Unemployed 7 (9%) 1 (5%) Disability 2 (2%) (0) Student 3 (4%) 2 (10%) Annual income (Can$) <15 000 17 (21%) 3 (14%) 15 000 to 29 999 8 (10%) 4 (19%) 30 000 to 59 999 22 (27%) 7 (33%) 60 000 to 89 999 23 (28%) 6 (29%) >90 000 11 (14%) 1 (5%) Insulin use Yes 48 (59%) 7 (33%) No 33 (41%) 14 (67%) Purpose of computer use* Business 2 (2%) 1 (5%) Personal 24 (30%) 5 (24%) Both 54 (68%) 15 (71%) Frequency of computer use* < 1 time/week 3 (4%) 0% 1–2 times/week 5 (6%) 0% 3–6 times/week 9 (11%) 0% ≥ 1 time/day 63 (79%) 21(100% ) Comfort with computer use Somewhat uncomfortable 2 (2%) 0% Neutral 5 (6%) 0% Somewhat comfortable 26 (32%) 8 (38%) Very comfortable 46 (57%) 13 (62%) Did not respond 2 (2%) 0% Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 6 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 1 Demographic characteristics and baseline values of observational cohort and qualitative study (Continued) Frequency of Internet use for diabetes < 1 time/week 66 (84%) 18 (85%) 1–2 times/week 7 (9%) 1 (5%) 3–6 times/week 5 (6%) 2 (10%) ≥ 1 time/day 1 (1%) 0% Comfort with Internet use Somewhat uncomfortable 2 (2%) 0% Neutral 3 (4%) 0% Somewhat comfortable 29 (36%) 8 (38%) Very comfortable 47 (58%) 13 (62%) Did not respond 0% 0% Self-efficacy; mean (SD) 4.61 (0.58) 5.11 (0.52) Self-care; mean (SD) 3.35 (1.12) 3.31 (0.84) Diabetes distress; mean (SD) 40.75 (16.21) 37.33 (14.74) HbA1c; mean (SD) 7.64% (1.29) 7.17% (0.98) 60.0 mmol/L (14.1) 54.9 mmol/L (10.7) Systolic blood pressure (mm Hg); mean (SD) 129.24 (13.84) 124.10 (8.73) Diastolic blood pressure (mm Hg); mean (SD) 76.03 (8.51) 74.71 (8.87) LDL-C (mmol/L); mean (SD) 2.11 (0.78) 2.20 (0.73) Weight (kg); mean (SD) 90.65 (21.76) 84.47 (15.13) *Data missing for one participant. Based on Statistics Canada data for low income cut-off [43], we selected $30 000 as a minimal level of income comfortable for activities of daily living and self-management capability for our analysis. Data missing for two participants. Abbreviation: LDL-C Low-density lipoprotein cholesterol. once. These users had a mean of 3.4, 4.5, and 9.3 visits/user implementation trajectory (effect: 0.12; 95% CI: −−0.028, to each of these page types, respectively. 0.263; p = 0.11; Figure 2 and Table 2). Self-care: The self-care score improved by 0.44 (95% Patient-centred outcomes CI: 0.23, 0.63; p < 0.0001) beyond what was expected at Self-efficacy: Despite a significant short-term increase in nine months (Figure 2 and Table 2). self-efficacy score immediately after implementation of the Diabetes distress: Despite an immediate short-term intervention (0.13; 95% confidence interval [CI]: 0.06, 0.20; decrease in diabetes distress score (−−2.29; 95% CI: −3.76, − p < 0.0004), by nine months, this outcome had not 0.81; p = 0.002), by nine months, this outcome had not increased significantly more than expected from its pre- decreased significantly over what was expected (effect: 1.84; 95% CI: −4.81, 1.12; p = 0.22; Figure 2 and Table 2). There was no interaction with insulin use by time or intervention for any of these outcomes. Self-care scores were positively correlated with age (0.04/year, 95% CI: 0.02, 0.06), p <0.001). Diabetes distress varied with age and sex: younger fe- male participants had greater diabetes distress. When we conducted the sensitivity analysis with missing values for income assumed to be below $15,000 and miss- ing values for health literacy assumed to be the mode, there were no changes in results for self-efficacy, self care, or diabetes distress. Clinical outcomes Seventy-three of the participants were included in the Figure 1 Website login and blog use by week. Black bar: Number analysis of clinical outcomes. The other eight partici- of logins per week. Grey bar: Number of blog views per week. pants were excluded because of missing data for HbA1c, Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 7 of 14 http://www.biomedcentral.com/1472-6947/14/117 the exception of diastolic blood pressure (users: +3.27 mm Hg; non-users: −1.6 mm Hg; p = 0.014; Additional file 1). Interviews Twenty-one individuals (Table 1) participated in an inter- view. The sample consisted of White and Asian men and women of various ages, duration of diabetes, educational attainment, and employment status, who used computers frequently and were comfortable with using the internet. Analysis of the interviews revealed numerous themes, four of which were most relevant to interpretation of the co- hort study’s negative results, in particular, exploration of why participants used the website to only a limited extent. Additional themes will be the focus of future publications. The following four themes are considered here: 1) barriers and facilitators of website use; 2) patterns of website use, including the role of the blog in driving site traffic; 3) gen- eral feedback on website characteristics; and 4) potential mechanisms for the effect of the website on self-efficacy, behaviour change, and diabetes distress. Representative quotes for each theme appear in Table 3. 1) Barriers and facilitators to use: Participants stated that they struggled with competing health and life con- cerns. They reported that it was “not just diabetes” that they dealt with (Table 3; 1a) and that they had to man- age other concurrent medical conditions (Table 3; 1b). They spoke about their attempts to balance illness work with everyday life work; they found that after completing the latter, “there wasn’t a lot” of time or energy left for self-management of their disease, much less to use the website (Table 3; 1b). Some participants identified lack of motivation as “a me thing, as opposed to a site thing”, while others commented that “laziness” (Table 3; 1c) was Figure 2 Self-efficacy, self-care, and diabetes distress nine months a barrier to use. before and nine months after intervention implementation. Participants’ attitudes toward diabetes also coloured Reference categories used in the plot were as follows: female, mean their approach to self-management and thus their use of age 57.34 years, employed, university education, income > Can the site. In particular, participants reported feeling frus- $30,000, adequate health literacy, white. trated with the uncontrolled nature of their disease, and the collection of self-monitoring information that showed blood pressure, LDL-C, or weight within 90 days of the a lack of metabolic control exacerbated this frustration self-efficacy data or because no data were obtained after (Table 3; 1d). Similarly, some participants said that they implementation of the intervention. The intervention had were sometimes overcome with a sense of futility. They no effect on HbA1c, blood pressure, LDL-C, or weight in perceived that regardless of their actions, some outcomes either the unadjusted or the adjusted models (Table 2). such as dialysis were inevitable (Table 3; 1e); hence, they Comparison of users and non-users: saw no value in learning about the disease or in trying to A total of 70 participants (86%) used the website at least self-manage the disease or use the website. once, whereas 11 (14%) did not use the site at all. At the Eleven (52%) of the 21 interview participants said the re- nine-month follow-up after implementation of the inter- quirement for a login and password prevented them from vention, there was no difference between users and non- using the website because they often forgot their password users in terms of self-efficacy (0.15 vs. 0.13, p = 0.35) or (Table 3; 1f). Others were limited by poor computer or self-care (0.18 vs. 0.13, p = 0.21). Users had a greater internet access and said they would prefer a mobile solution reduction in diabetes distress than did non-users (−4.7 (Table 3; 1g). Finally, some participants noted that the oner- vs. -0.9, p < 0.0001). There was no difference in the effect ous process for correcting error in log entries discouraged of using the intervention on any secondary outcomes, with them from using the self-management tools (Table 3; 1h). Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 8 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 2 Summary statistics of psychological and clinical outcomes (with 95% confidence interval), at implementation of intervention and 9 months later Outcome Immediate change p-value Change in outcome at study end, p-value (at time of intervention) over expected value Self-efficacy 0.13 (0.06, 0.20) 0.0004 0.12 (−0.028, 0.263) 0.11 Self-care 0.26 (−0.17, 0.68) 0.24 0.44 (0.23, 0.63) <0.0001 Diabetes distress −2.29 (−3.76, −0.81) 0.002 −1.84 (−4.81, 1.12) 0.22 HbA1c 0.37 (−0.11, 0.85) 0.13 −0.055 (−1.13, 1.02) 0.93 Weight (kg) 0.37 (−3.00, 3.74) 0.83 −1.10 (−9.31, 7.12) 0.77 Blood pressure (mm Hg) Systolic 1.79 (−5.87, 9.26) 0.66 −5.89 (−18.26,12.37) 0.53 Diastolic 0.43 (−4.13, 4.98) 0.85 −8.22 (−19.18, 2.74) 0.14 LDL-C (mmol/L) −0.0006 (−0.26, 0.26) 0.996 0.14 (−0.55, 0.82) 0.72 Abbreviation: LDL-C Low-density lipoprotein cholesterol. In contrast, other website characteristics appeared to en- the blog afforded them the opportunity to obtain “expert” courage users to visit and return. The perceived reliability medical answers anonymously (Table 3; 2c,i). The provision of the website’s information and the perception that it was of “fellow patient” advice was characterized as promoting a an “authoritative source” drew users (Table 3; 1i). In sense of community that some participants felt might com- addition, email reminders prompted them to return; such bat their sense of isolation (Table 3; 2c,ii). However, this prompts seemed well-suited to what users characterized was not a universal sentiment, and some participants felt as a “fast-paced world” and served as effective reminders uncomfortable with and disconnected from the blog to make diabetes self-care a high priority (Table 3; 1j). (Table 3; 2c,ii). In addition, there was a tension between a Similarly, routinization of the online experience appeared desire for online community and a fear of “looking foolish”. to routinize use of the internet for certain aspects of For example, some participants said they were afraid of health care. For example, participants reported that in- “putting [up] a stupid question” or surmised that others creasing their use of the internet and expanding their were “shy about how they write” (Table 3; 2c,iii). Other rea- scope of internet activities created a new norm for internet sons offered for not contributing to the blog included par- usage, such that it became more commonplace for them ticipants’ perceptions that they had nothing to offer, that to “look up whatever is interesting to me”, “track stuff on- the blog was not their preferred mode of communication, line”, and embrace “going paperless” (Table 3; 1 k). and that they preferred face-to-face communication 2) Patterns of website use, including role of the blog: (Table 3; 2c,iv). There were mixed views regarding the reli- Participants said that use of the website was driven by ability of the blog content: one user commented that a blog their individual context and circumstances. Rather than represented “the blind leading the blind”,but others said browsing at random, users said they were goal-directed: they were reassured by the fact it was moderated by an ex- when they had a specific concern, they focused on that pert (Table 3; 2c,iv). area of the website (Table 3; 2a). For example, one par- 3) General feedback on website characteristics: Partici- ticipant was initially motivated to visit (and subsequently pants commented on their impressions of the website continued to visit) the foot care section of the website overall and provided feedback concerning general features because of her foot symptoms (Table 3; 2b). Participants that were appreciated. Participants perceived that the web- also commented that they used the website to gauge the site was accurate, comprehensive (Table 3; 3a), and easy to urgency of their concerns and to try to obtain immediate navigate (Table 3; 3b). While they appreciated the website’s answers to their questions (Table 3; 2b). provision of “evidence-based medical content”,they re- We explored potential reasons for the unexpected find- ported a desire for more practical solutions and “real life ing that the blog was the most frequently accessed tool answers” (Table 3; 3c) and spoke about finding a balance and appeared to drive website usage. Participants identi- between these two characteristics. fied dual roles of the blog in providing a forum for both 4) Mechanisms of effect of the website on self-efficacy, “expert” and “fellow patient” advice (Table 3; 2c). “Fellow behaviour change, and diabetes distress: Despite appar- patient” advice consisted of submitting suggested content ently limited use of the website, the intervention appeared or commenting on blog posts (i.e. participating in a dis- to have an effect on self-efficacy, behaviour change, and cussion thread). For patients who were otherwise uncom- diabetes distress. Deeper exploration of the data regarding fortable with asking questions of a health care provider, patterns of use and website features uncovered factors that Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 9 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 3 Themes identified and representative quotes from in-depth interviews Theme/subtheme Representative quotes 1) Barriers and facilitators of website use: Barriers to website use Patient-related barriers a) Competing health concerns “Being 4 cancers in my life”. [3B53, 78-year old woman] “it’s not just diabetes that I deal with”. [2B16, 37-year old woman] b) Competing life concerns “I think for me, health issues surround migraines and fatigue. By the time I was home and had any free time, there wasn’t a lot there. And then life style and other stresses the business and other stresses the business…” [2B16, 37-year old woman] c) Lack of motivation “It’s a me thing, as opposed to a site thing”. [2B09, 47-year old woman] “Laziness”. [3B52, 65-year old woman] d) Frustration with diabetes “Part of it is, when you see the blood sugar is really high, I already know it’s high. I’m not taking the medication. So to log the fact that they are high, ends up making you more frustrated. So why do that”. [2B09, 47-year old woman] e) Futility regarding diabetes Sometimes I think no matter what I do it’s not going to matter because you read about this… they all end up on dialysis… they all, you know…” [2B09, 47-year old woman] Website-related barriers f) Login and password requirement “I keep forgetting my password”. [3B27, 58-year old woman] “The ability to change the password is good as well… Because I think that helps in returning, otherwise I would always go back and find your original email”. [2B16, 37-year old woman] g) Limited computer and/or internet “I didn’t have too much time to look at it ’cause I don’t have a computer at home access [3B39, 67-year old man] “I wish I can have a mobile like app”. [3A10, 52-year old woman] “on the road for work and limited web access”. [1A13, 47-year old man] h) Onerous nature of data entry to use “Um, but cause I did go in and I did try and do the tracking and I think cause I thought that was on logs an ongoing basis was the most useful part of it. But it was kind of a pain in the neck to use it… and kind of a pain in the ass getting where I wanted to go. I put some information and I wanted to delete it and I don’t know if I ever succeeded in getting rid of it”. [2B01, 48-year old man] Facilitators of website use i) Perceived reliability of information “It has to be really, really tuned in or connected to the latest developments either at [hospital name] or the world… With tons of information by real authorities. If I had a diabetic questions, I thought you guys might answer it, I would consider this to be an authoritative source”. [2B01, 48-year old man] j) Reminders, prompts of new content “Actually love getting the reminders, I really do. In this day in age everybody is busy, but I’m really good about it. The day I get the reminder, I find time to go in, I read the blogs… As I said, if you told me there was something new under a certain heading, I would definitely go in. I just like the reminders, in this fast paced world you don’t always have time for things and the last thing you think about is yourself. So I think the reminders are a great thing, and I always look”. [3B09, 61-year old woman] k) Increasing comfort with internet use “Slowly I got used to [the internet] and since that time I utilize the possibility to look after whatever is interesting to me”. [3B53, 78-year old woman] “I starting writing a lot of emails to lawyers and the public trustee so I’m kind of used to no, more so than before, tracking stuff online. Kind of going paperless is neat”. [2B01, 48-year old man] 2) Patterns of use of the website, including role of blog in driving site traffic a) Goal-directed use “And that’s kind of how I use the tool. It’s not as much a “Do I go in daily and read a new section each day?” No, it’s when I have a specific concern that I hone in on it to a degree”. [2B09, 47-year old woman] “Right, so there are a lot of places [in the website to which] you can go. But if those places don’t impact you… For example, if you don’t have high blood pressure, if it doesn’t impact you, you are less likely to go back to it. Then when it impacts you, then you know the resource is there”. [2B09, 47-year old woman] b) Use in context of current concern “Because the bottom of my toes is not sensitive and the side of my foot is not sensitive. So I looked up this website and tried to find out what could I do and is there any kind of help that I could find”. [3B52, 65-year old woman] Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 10 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 3 Themes identified and representative quotes from in-depth interviews (Continued) -Motivation for repeated use “The foot complication is the one I usually look up”. [3B52, 65-year old woman] -To gage urgency “I would look something to see if I thought if it was urgent”. [3B13, 54-year old woman] “Now, I have to say yesterday in desperation I actually tried to look up some medication information on this website…” [2B09, 47-year old woman] c) Role of blog in driving site traffic: “When new blogs coming up then I read it. Not always at once, but still eventually”. [3B53, 78-year old woman] Dual roles of blog “‘Ask the Expert’ should remain open all the time… And maybe it’s not just ‘Ask the Expert’; maybe it’s ‘Ask someone’. So you need an ‘Ask the Expert’ section and ‘Ask the Fellow Patient’ section”. [2B09, 47-year old woman] i) Opportunity for anonymous request “It provides an opportunity where people however anonymous they want to be and it might be for information something someone could be shy asking a heath care practitioner meanwhile go ahead and ask on something like that. So in that sense, I think that’s useful”. [2B16, 37-year old woman] ii) Fostering a sense of community “It’s interesting because people will come with a question through [the blog, such as] “Well, this worked for me”,or “This will work for everyone.” It is a sort of communal sense and conversation”. [3B53, 78-year old woman] “So, I think learning to develop your support systems is extremely important for a diabetic. And that having a forum where even if you don’t have a lot of people in your life that you can talk to about this, but having a forum where maybe you can go on and have an online community can be very helpful. Provided that they understand that any information that they read has to be verified”. [3B13, 58 year-old woman] -Social support “For example, if you are feeling yourself alone and loneliness is not a very good thing health-wise. It leads to depression and everything else and if you feel that you have to communicate, and then communicate here”. [3B49, 46-year old man] -Though this was not a universal sentiment: “Sometimes with a blog it can be communal and it can be conversation, but it can also make people others felt a lack of connection feel disconnected because: how comfortable are they in conversation?” [2B16, 37-year old woman] iii) However, tension between desire for “Because we are afraid that we are putting a stupid question”. [3B53, 78-year old woman] online community and fear of ridicule, “I found at times, responding to other posts, that [I felt] shy about responding. But I did go ahead and timidity do that even though I thought at times that’s not necessarily my usual style”. [2B16 37-year old woman] “It might take time to catch on, because people who are feeling shy about it may need to see what other people do with it for quite a while”. [2B16, 37-year old woman] “I think some people are shy about how they write.” [2B16, 37-year old woman] “I’ve just never been interactive…” [3B09, 61-year old woman] iv) Other reasons for not contributing to the blog: -Nothing new to contribute “But I’m not a poster. I read, I learn, I just don’t put my 2 cents in. [I don’t post because] I would say the same thing – why would I bother?” [3B09, 61-year old woman] -Reliability of content: Balance between NO “I find for the most part it’s the blind leading the blind. I guess this one is being moderated but by evidence anecdote, trusted source and large you have a bunch of people who don’t know anything kind of spewing forth”. versus patient centred [3B13, 54-year old woman] YES “However, the way this one is designed essentially it goes and it’s approved or reviewed before being posted in the first place. Which I think is a good style”. [2B16, 37-year old woman] -Perceived relevance to specific “I don’t know what the ages are but I’m thinking a younger person would be more familiar with demographic population communication in a way that is not face to face, more so”. [2B16, 37-year old woman] 3) General feedback on website characteristics: a) Accurate and comprehensive In fact, what I noticed is that their seemed to be more than what I would have anticipated… [2B01, 48-year old man] So there are lots of things on the website that really did help. So things like how to control the blood pressure and how to do all those other… there is lots of things that help. [2B09, 47-year old woman] b) Easy to navigate “The website itself is easy to navigate. And I think that list is very good and I’ve found that every time I have looked, [the answer] could typically could fall into one of those categories. The answer might not be there, but I know where to start to look”. [2B09, 47-year old woman] c) But could be more relevant: -Want practical solutions Maybe too many people are inundated with expert answers and not enough with real life answers. [3B09, 61-year old woman] Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 11 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 3 Themes identified and representative quotes from in-depth interviews (Continued) But that when you need it the most, that’s when I found it didn’t have everything I needed. It was still missing… the next piece; the next piece is what happens when you are in that mess? [3B49, 46-year old man] Yeah, it combines that very practical app, ok so here is an endocrinologist talking about blah, blah, blah, new scientific in road into diabetes… ok, that’s great on an intellectual level… What does this mean on a practical level? For those that want to and can handle the scientific information, perfect. And those of us who can’t or don’t want to then there’s other things on the site. [3B13, 54-year old female] What do you value about each type of information, the practical versus medical evidence based? 70 or 80 [percent of it practical] versus 30 and 20 for the medical. [3B53, 78-year old woman] But nowhere does it tell you how to deal with it. So, what I’m looking for and what I was looking for in that peer support was other people in the same position who have found solutions to issues that typical websites don’t tell you [2B09, 47-year old woman] 4) Potential mechanism of impact of website on self-efficacy, behavior change and diabetes distress a) Unanticipated pattern of use of DOC “I actually I haven’t done it lately but when I first got this I searched a lot of stuff, I was interested so I did search. I printed it out I actually have some in a binder, which I have referred to on occasion. It’s -Print out nice to have something black and white that you can refer to. That’s why I like certain websites, because you can print it off and read it. I can search other references and I’ve done that. And I keep a binder and I refer to it when…” [3B09, 61-year old woman] b) Unanticipated impact of reminder “There were the reminders that you hadn’t visited the site in a while. [When I got those reminders], emails I said, I will answer it later… This companion is a good reminder, I heard about it before but then let’s go back and check to make sure that I’m understanding properly…” [3B27, 58-year old woman] “Constantly reminding me about the things that we need to be aware of. Most of us know but the Online Companion was a good reminder and got me thinking of things that I need to constantly do (some that we conveniently forget)”. [3B55, 58-year old woman] c) Unanticipated role of DOC “Because of the recording of activity… I started… it guilted me into starting to record my activity and in “kickstarting” self-management behaviors once I started recording it because also coming off of the Metformin, I started recording it there and then I switched to recording it in books and now I simply record it on the calendar”. [1A06, 62-year old woman] Abbreviations: St. Mike’s St. Michael’s Hospital, Toronto, ON, Canada. App application. DOC Diabetes Online Companion. might account for these quantitative findings. For ex- distress than those who did not use it. Despite a user- ample, rather than returning to the site to revisit and re- centred design process and an increase in the frequency view items, some participants reported that they printed of blog posting from weekly to twice weekly, use of the items of interest from the website and subsequently re- website (as ascertained by login records) was limited. ferred to these paper copies (Table 3; 4a). The use of re- Our interviews revealed that both patient-related factors minder emails also had an effect. Participants reported (e.g. competing health and life concerns, a sense of futil- that these emails not only prompted them to return to ity) and website-related factors (e.g. requirement for and log into the website, but also encouraged them in login, limited computer or internet access) limited use of their own self-management (Table 3; 4b). Finally, one par- the website. We also found that participants were moti- ticipant noted that the website had kick-started her self- vated to access the website on the basis of their current management behaviours, by initially “guilting” her into needs and concerns, as well as new blog postings, with recording them online. Although she did not subsequently the blog fulfilling a need for both “expert” medical con- login to the website to record these behaviours, she did tent and peer support and a sense of community. continue to record them on paper. These qualitative findings have confirmed the import- ance of website features such as the reliability and authori- Discussion tativeness of information [47], as well as the use of blogs We found that a self-management website for patients [15] and reminders [48] for continued engagement of with type 2 diabetes led to no improvement in self- users. Our findings also emphasize the need to provide a efficacy, diabetes distress, or clinical outcomes over the greater proportion of “practical” patient-centred content. study period. However, there was an improvement in A recent qualitative study analyzing 3005 diabetes-related self-care (a secondary outcome), and the group that used blog posts showed that the most influential blogs were the website experienced significantly lower diabetes those written by patients and that only 10% of blogs cited Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 12 of 14 http://www.biomedcentral.com/1472-6947/14/117 biomedical literature [49], highlighting the dual needs for disease. For example, a systematic review of the benefits reliable, evidence-based content and engaging patient- and limitations of social media in the context of chronic based content. disease identified benefits (increased interaction and social Our data also suggest that mobile devices are a potential support, tailored and accessible information) and limita- avenue through which to improve accessibility and use of tions (quality concerns and lack of reliability, confidential- a self-management site. A Cochrane review of computer- ity, and privacy) [59] to those we identified. Similarly, our based diabetes self-management interventions identified finding of a reduction in diabetes distress in conjunction 16 randomized controlled trials, which showed a small ef- with no improvements in clinical outcomes echoes find- fect on glycemic control (−2.3 mmol/mol or −0.2%, p = ings from intervention strategies targeting other chronic 0.009), with the mobile phone subgroup experiencing a diseases. For example, another systematic review examin- greater effect (−5.5 mmol/mol or −0.5%, p < 0.00001) [50]. ing the effect of social media on psychological and physical Given the increasing preference for mobile devices over outcomes in chronic disease found a relatively large body desktop computers [51] for health information resources of evidence demonstrating psychological benefit (19 iden- [52], mobile technology may overcome the barriers to tified studies) but limited evidence for physical outcomes website access and use that we identified, through greater (4 identified studies) [60]. integration with patients’ existing routine, such that self- This study was limited by its non-randomized design. management is no longer seen as additional “illness work”. However, we employed a repeated-measures design that However, as with web-based technology, a systematic permitted reliable assessment of baseline self-efficacy. Al- approach to development, testing, implementation, and though our primary outcome (self-efficacy) was a non- evaluation of mobile health technology is warranted. Al- clinical outcome, it is a validated predictor of patient though such technologies are proliferating, with over 736 behaviour change and clinical outcomes [18,20,24,25]. The applications related to diabetes alone, their usability and infrequency of website use likely limited the effect of this clinical effectiveness are variable [53], and concerns exist intervention, but we obtained valuable insights regarding regarding their effectiveness and safety, as well as the mediators of website use through our individual interviews. security of personal health information [54]. Our findings The qualitative evaluation was conducted by individuals regarding user engagement with web-based technology who were also involved in developing the intervention, echo those for mobile technology: an evaluation of 10 which created a potential for bias; however, we guarded mobile diabetes applications emphasized the importance against this bias by including individuals who were not in- of user-centred design, an engaging interface, and context- volved in designing the website as members of the qualita- driven use [55]. tive analytic team and by having three coders. As such, we Competing health concerns were identified as a barrier were able to obtain and report critical feedback that par- to web-based self-management. Patients’ adherence to ticipants openly shared. Study strengths include the use of diabetes care is affected by multimorbidity (e.g. depres- multiple repeated measures, the use of validated out- sion), which in turn directly affects self-management comes, dual coding of all transcripts, and triangulation of ability and competes for time and attention [56]. For ex- the qualitative findings with the quantitative results ample, patients with a greater number of comorbidities [42,44,46]. placed a lower priority on diabetes and had worse dia- betes self-management ability [57]. Future interventions should consider strategies, such as shared decision- Conclusion making and priority-setting, to empower patients with mul- Increasing use of the World Wide Web by consumers tiple comorbidities to optimize their self-care [58]. For for health information and ongoing revolutions in so- example, a patient may identify mood management as a cial media are strong indicators that consumers are priority, which is key to subsequent self-care. Thus, greater welcoming and demanding a new era of technology in integration with the person’s cognitive, emotional, and health care. However, full potential of this technology health information-seeking preferences, daily living routine, is hindered by limited uptake and high attrition rates. and health context through the use of patient-based con- Use of the Diabetes Online Companion may be opti- tent, mobile devices, and individualized decision-making, mized by integrating a mobile interface, emphasizing may be further strategies to maximize website use and re- “practical” patient-centred content such as a patient- duce intervention attrition. led blog, and including a “prioritization” feature to Finally, our results may be extrapolated to other help users with competing concerns. Our research chronic diseases. In particular, our finding of the need findings have shed light on these limitations by identifying for tailored content and peer support, balanced with characteristics associated with website use and attrition concerns regarding information reliability and confidenti- and suggesting strategies to reduce website attrition as a ality, is applicable to other strategies for managing chronic way to potentially optimize clinical outcomes. Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 13 of 14 http://www.biomedcentral.com/1472-6947/14/117 Additional file 5. Cochrane J, Conn VS: Meta-analysis of quality of life outcomes following self-management training. Diabetes Educ 2008, 34:815–823. 6. Minet L, Møller S, Vach W, Wagner L, Henriksen JE: Mediating the effect of Additional file 1: Appendix A: Representative Screenshots of the self-care management intervention in type 2 diabetes: a meta-analysis Diabetes Online Companion. Appendix B: Description of primary of 47 randomised controlled trials. Patient Educ Couns 2010, 80:29–41. outcome scales. Appendix C: Semistructured interview guide. 7. Ruppert K, Uhler A, Siminerio L: Examining patient risk factors, comorbid Appendix D: Effect of Intervention comparing users and non-users conditions, participation, and physician referrals to a rural diabetes (after adjustment for time). Appendix E: References. self-management education program. Diabetes Educ 2010, 36:603–612. 8. Shaw K, Killeen M, Sullivan E, Bowman P: Disparities in diabetes self-management education for uninsured and underinsured adults. Competing interests Diabetes Educ 2011, 37:813–819. The authors declare that they have no competing interests. 9. Shekelle PG, Morton SC, Keeler EB: Costs and Benefits of Health Information Technology. Rockville: Agency for Healthcare Research and Quality (US); 2006. Authors’ contributions 10. Yu CH, Bahniwal R, Laupacis A, Leung E, Orr MS, Straus SE: Systematic CHY conceived of the study, collected and analyzed the data, and wrote the review and evaluation of web-accessible tools for management of manuscript. JAP contributed to the study design, analyzed the data, diabetes and related cardiovascular risk factors by patients and contributed to the interpretation of the data, and critically revised the healthcare providers. J Am Med Inform Assoc 2012, 19:514–522. manuscript for important intellectual content. MM, BRS, AL, and SES 11. Bull SS, Gaglio B, McKay HG, Glasgow RE: Harnessing the potential of the contributed to the study design and to interpretation of the data and internet to promote chronic illness self-management: diabetes as an critically revised the manuscript for important intellectual content. GL example of how well we are doing. Chronic Illness 2005, 1:143–155. analyzed the data and wrote part of the manuscript. SH conducted the 12. van Vugt M, de Wit M, Cleijne WH, Snoek FJ: Use of behavioral change interviews and qualitative analyses and drafted portions of the manuscript. techniques in web-based self-management programs for type 2 diabetes DN conducted the computer programming to develop and refine the patients: systematic review. J Med Internet Res 2013, 15:e279. website, collected data on web usage and critically revised the manuscript 13. Seidman JJ, Steinwachs D, Rubin HR: Design and testing of a tool for for important intellectual content. OB contributed to interpretation of the evaluating the quality of diabetes consumer-information Web sites. data and critically revised the manuscript for important intellectual content. J Med Internet Res 2003, 5:e30. CHY is guarantor for this article. All authors have given final approval of the 14. Shrank WH, Patrick AR, Brookhart MA: Healthy user and related biases in version to be published, and all agree to be accountable for all aspects of observational studies of preventive interventions: a primer for the work in ensuring that questions related to the accuracy or integrity of physicians. J Gen Intern Med 2011, 26:546–550. any part of the work are appropriately investigated and resolved. 15. Yu CH, Parsons J, Hall S, Newton D, Jovicic A, Lottridge D, Shah BR, Straus SE: User-centered design of a web-based self-management site for individuals Acknowledgements with type 2 diabetes – providing a sense of control and community. We thank Jovita Sundaramoorthy and Carolyn Gall-Casey of the Canadian BMC Med Inform Decis Mak 2014, 14:60. Diabetes Association for their input and feedback related to project design. 16. Yu C, Parsons J, Mamdani M, Lebovic G, Shah BR, Bhattacharyya O, Laupacis We also thank Kevin Thorpe for his invaluable input into the choice of A, Straus SE: Designing and evaluating a web-based selfmanagement site methods for the statistical analysis. for patients with type 2 diabetes - systematic website development and study protocol. BMC Med Inform Decis Mak 2012, 12:57. Funding 17. Bandura A: Self-Efficacy: the Exercise of Control. New York: W.H. Freeman; 1997. The study was funded by the Canadian Institutes of Health Research (CIHR) 18. Glasgow RE, Toobert DJ, Hampson SE, Brown JE, Lewinsohn PM, Donnelly J: Knowledge to Action Operating Grant (funding reference number KAL Improving self-care among older patients with type II diabetes: the 94484). Dr. SE Straus is supported by a Tier 1 Canada Research Chair. “sixty something…” study. Patient Educ Couns 1992, 19:61–74. 19. Reichard P, Toomingas B, Rosenqvist U: Changes in conceptions and Author details attitudes during five years of intensified conventional insulin treatment Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Postal address: 30 in the stockholm diabetes intervention study (SDIS). Diabetes Educ 1994, Bond St, Toronto, ON M5B 1W8, Canada. Department of Medicine, 20:503–508. University of Toronto, Toronto, ON, Canada. Dhalla Lana School of Public 20. Anderson RM, Funnell MM, Butler PM, Arnold MS, Fitzgerald JT, Feste CC: Health, University of Toronto, Toronto, ON, Canada. Applied Health Research Patient empowerment: results of a randomized controlled trial. Centre, St. Michael’s Hospital, Toronto, ON, Canada. Department of Physical Diabetes Care 1995, 18:943–949. Therapy, University of Toronto, Toronto, ON, Canada. Institute of Health 21. Trief PM, Teresi JA, Eimicke JP, Shea S, Weinstock RS: Improvement in diabetes Policy, Management and Evaluation, University of Toronto, Toronto, ON, self-efficacy and glycaemic control using telemedicine in a sample of older, Canada. Institute for Clinical Evaluative Sciences, G1 06, 2075 Bayview ethnically diverse individuals who have diabetes: the IDEATel project. Avenue, Toronto, ON M4N 3M5, Canada. Sunnybrook Research Institute, Age Ageing 2009, 38:219–225. Sunnybrook Health Sciences Centre, Toronto, ON, Canada. 22. Aljasem LI, Peyrot M, Wissow L, Rubin RR: The impact of barriers and self-efficacy on self-care behaviors in type 2 diabetes. Diabetes Educ 2001, Received: 22 July 2014 Accepted: 27 November 2014 27:393–404. 23. Grossman HY, Brink S, Hauser S: Self-efficacy in adolescent girls and boys with insulin-dependent diabetes mellitus. Diabetes Care 1987, 10:324–329. References 24. Nakahara R, Yoshiuchi K, Kumano H, Hara Y, Suematsu H, Kuboki T: Prospective 1. Centers for Disease Control and Prevention: Diabetes Report Card 2012. study on influence of psychosocial factors on glycemic control in Japanese Atlanta, GA: Centers for Disease Control and Prevention, US Department of patients with type 2 diabetes. Psychosomatics 2006, 47:240–245. Health and Human Services; 2012. 25. Lorig KR, Ritter PL, Jacquez A: Outcomes of border health Spanish/English 2. American Diabetes Association: Standards of medical care in diabetes chronic disease self-management programs. Diabetes Educ 2005, 31:401–409. 2013. Diabetes Care 2013, 36:S11–S66. 26. Toobert DJ, Hampson SE, Glasgow RE: The summary of diabetes self-care 3. Haas L, Maryniuk M, Beck J, Cox CE, Duker P, Edwards L, Fisher EB, activities measure: results from 7 studies and a revised scale. Diabetes Care Hanson L, Kent D, Kolb L, McLaughlin S, Orzeck E, Piette JD, Rhinehart AS, 2000, 23:943–950. Rothman R, Sklaroff S, Tomky D, Youssef G: National standards for diabetes 27. Polonsky WH, Fisher L, Earles J, Dudl RJ, Lees J, Mullan J, Jackson RA: self-management education and supportdiabetes. Diabetes Care 2014, Assessing psychosocial distress in diabetes: development of the Supplement 1:S144–S153. Diabetes Distress Scale. Diabetes Care 2005, 28:626–631. 4. Selby JV, Beal AC, Frank L: The Patient-Centered Outcomes Research Institute (PCORI) national priorities for research and initial research 28. Chew LD, Bradley KA, Boyko EJ: Brief questions to identify patients with agenda. J Am Med Assoc 2012, 307:1583–1584. inadequate health literacy. Fam Med 2004, 36:588–594. Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 14 of 14 http://www.biomedcentral.com/1472-6947/14/117 29. Chew LD, Griffin JM, Partin MR: Validation of screening questions for 57. Kerr E, Heisler M, Krein S, Kabeto M, Langa K, Weir D, Piette J: Beyond limited health literacy in a large VA outpatient population. J Gen Intern comorbidity counts: how do comorbidity type and severity influence Med 2008, 23:561–566. diabetes patients’ treatment priorities and self-management? J Gen Intern 30. Zar JH: Biostatistical Analysis. Prentice Hall: Upper Saddle River, N.J; 1996. Med 2007, 22:1635–1640. 31. Liu G, Liang KY: Sample size calculations for studies with correlated 58. Joosten EA, DeFuentes-Merillas L, de Weert GH, Sensky T, van der Staak CP, observations. Biometrics 1997, 53:937–947. de Jong CA: Systematic review of the effects of shared decision-making 32. Norris SL, Engelgau MM, Narayan KM: Effectiveness of self-management on patient satisfaction, treatment adherence and health status. training in type 2 diabetes: a systematic review of randomized Psychother Psychosom 2008, 77:219–226. controlled trials. Diabetes Care 2001, 24:561–587. 59. Moorhead SA, Hazlett DE, Harrison L, Carroll JK, Irwin A, Hoving C: A new 33. Diggle P, Heagerty P, Liang KY, Zeger S: Analysis of Longitudinal Data. dimension of health care: systematic review of the uses, benefits, and Oxford, New York: Oxford University Press; 2002. limitations of social media for health communication. J Med Internet Res 2013, 15:e85. 34. Verbeke G, Molenberghs G: Linear Mixed Models for Longitudinal Data. 60. Merolli M, Kathleen G, Martin-Sanchez F: Health outcomes and related effects Verlag, New York: Springer; 2009. of using social media in chronic disease management: A literature review 35. Laird NM, Donnelly C, Ware JH: Review papers: longitudinal studies with and analysis of affordances. J Biomed Inform 2013, 46:957–969. continuous responses. Stat Methods Med Res 1992, 1:225–247. 36. Cnaan A, Laird NM, Slasor P: Tutorial in biostatistics: using the general linear mixed model to analyse unbalanced repeated measures and doi:10.1186/s12911-014-0117-3 Cite this article as: Yu et al.: A web-based intervention to support self- longitudinal data. Stat Med 1997, 16:2349–2380. management of patients with type 2 diabetes mellitus: effect on self- 37. Sherrill DL, Holberg CJ, Enright PL, Lebowitz MD, Burrows B: Longitudinal efficacy, self-care and diabetes distress. BMC Medical Informatics and analysis of the effects of smoking onset and cessation on pulmonary Decision Making 2014 14:117. function. Am J Respir Crit Care Med 1994, 149:591–597. 38. Figaro MK, Elasy T, BeLue R: Exploring socioeconomic variations in diabetes control strategies: impact of outcome expectations. J Natl Med Assoc 2009, 101:18–23. 39. Osborn CY, Cavanaugh K, Wallston KA, Rothman RL: Self-efficacy links health literacy and numeracy to glycemic control. J Health Commun 2010, 15:146–158. 40. Harrell FE, Lee KL, Mark DB: Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996, 15:361–387. 41. R Team: R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing; 2006. 42. Kuzel AJ: Sampling in Qualitative Inquiry. In Doing Qualitative Research. Edited by Crabtree BF, Miller W. Thousand Oaks: Sage Publications; 1999:33–45. 43. Poland BP: Transcription quality as an aspect of rigor in qualitative research. Qual Inq 1995, 1:290–310. 44. Kvale S: InterViews: An Introduction to Qualitative Research Interviewing. Thousand Oaks: Sage Publications; 1996. 45. Pope C, Ziebland S, Mays N: Qualitative research in health care: analysing qualitative data. BMJ 2000, 320:114–116. 46. Bryman A: Quantity and Quality. In Social Research. London: Routledge; 1988. 47. Lemire M, Paré G, Sicotte C, Harvey C: Determinants of internet use as a preferred source of information on personal health. Int J Med Inform 2008, 77:723–734. 48. Nijland N, van Gemert-Pijnen JE, Kelders SM, Brandenburg BJ, Seydel ER: Factors influencing the use of a web-based application for supporting the self-care of patients with type 2 diabetes: a longitudinal study. J Med Internet Res 2011, 13:e71. 49. Gruzd A, Black FA, Le Yen TN, Amos K: Investigating biomedical research literature in the blogosphere: a case study of diabetes and glycated hemoglobin (HbA1c). J Med Libr Assoc 2012, 100:34–42. 50. Pal K, Eastwood SV, Michie S, Farmer AJ, Barnard ML, Peacock R, Wood B, Inniss JD, Murray E: Computer-based diabetes self-management interventions for adults with type 2 diabetes mellitus. Cochrane Database Syst Rev 2010, 3. 51. Smith A: A: 35% of American Adults Own a Smartphone: One Quarter of Smartphone Owners Use their Phone for Most of their Online Browsing. Washington: Pew Research Center; 2011. 52. Manhattan Research: U.S. mobile health audience jumps to 95 million adults. Submit your next manuscript to BioMed Central Cybercitizen Health® U.S. 2013. Available at http://manhattanresearch.com/ and take full advantage of: News-and-Events/Press-Releases/mobile-health-95-million. 53. Demidowich AP, Lu K, Tamler R, Bloomgarden Z: An evaluation of diabetes • Convenient online submission self-management applications for Android smartphones. J Telemed • Thorough peer review Telecare 2012, 18:235–238. 54. Eng DS: The promise and peril of mobile health applications for diabetes • No space constraints or color figure charges and endocrinology. Pediatr Diabetes 2013, 14:231–238. • Immediate publication on acceptance 55. Årsand E, Frøisland DH, Skrøvseth SO, Chomutare T, Tatara N, Hartvigsen G, Tufano JT: Mobile health applications to assist patients with diabetes: lessons • Inclusion in PubMed, CAS, Scopus and Google Scholar learned and design implications. J Diabetes Sci Technol 2012, 6:1197–1206. • Research which is freely available for redistribution 56. Ciechanowski PS, Katon WJ, Russo JE: Depression and diabetes: impact of depressive symptoms on adherence, function, and costs. Arch Intern Med Submit your manuscript at 2000, 160:3278–3285. www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Medical Informatics and Decision Making Springer Journals

A web-based intervention to support self-management of patients with type 2 diabetes mellitus: effect on self-efficacy, self-care and diabetes distress

Loading next page...
 
/lp/springer-journals/a-web-based-intervention-to-support-self-management-of-patients-with-URMU8v4qAP

References (71)

Publisher
Springer Journals
Copyright
Copyright © 2014 by Yu et al.; licensee BioMed Central Ltd.
Subject
Medicine & Public Health; Health Informatics; Information Systems and Communication Service; Management of Computing and Information Systems
eISSN
1472-6947
DOI
10.1186/s12911-014-0117-3
pmid
25495847
Publisher site
See Article on Publisher Site

Abstract

Background: Management of diabetes mellitus is complex and involves controlling multiple risk factors that may lead to complications. Given that patients provide most of their own diabetes care, patient self-management training is an important strategy for improving quality of care. Web-based interventions have the potential to bridge gaps in diabetes self-care and self-management. The objective of this study was to determine the effect of a web-based patient self-management intervention on psychological (self-efficacy, quality of life, self-care) and clinical (blood pressure, cholesterol, glycemic control, weight) outcomes. Methods: For this cohort study we used repeated-measures modelling and qualitative individual interviews. We invited patients with type 2 diabetes to use a self-management website and asked them to complete questionnaires assessing self-efficacy (primary outcome) every three weeks for nine months before and nine months after they received access to the website. We collected clinical outcomes at three-month intervals over the same period. We conducted in-depth interviews at study conclusion to explore acceptability, strengths and weaknesses, and mediators of use of the website. We analyzed the data using a qualitative descriptive approach and inductive thematic analysis. Results: Eighty-one participants (mean age 57.2 years, standard deviation 12) were included in the analysis. The self-efficacy score did not improve significantly more than expected after nine months (absolute change 0.12; 95% confidence interval −0.028, 0.263; p = 0.11), nor did clinical outcomes. Website usage was limited (average 0.7 logins/ month). Analysis of the interviews (n = 21) revealed four themes: 1) mediators of website use; 2) patterns of website use, including role of the blog in driving site traffic; 3) feedback on website; and 4) potential mechanisms for website effect. Conclusions: A self-management website for patients with type 2 diabetes did not improve self-efficacy. Website use was limited. Although its perceived reliability, availability of a blog and emailed reminders drew people to the website, participants’ struggles with type 2 diabetes, competing priorities in their lives, and website accessibility were barriers to its use. Future interventions should aim to integrate the intervention seamlessly into the daily routine of end users such that it is not seen as yet another chore. Keywords: Diabetes mellitus, Online systems, Patient self-management, Self-efficacy, Repeated measures modelling, Qualitative methods * Correspondence: yuca@smh.ca Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Postal address: 30 Bond St, Toronto, ON M5B 1W8, Canada Department of Medicine, University of Toronto, Toronto, ON, Canada Full list of author information is available at the end of the article © 2014 Yu et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 2 of 14 http://www.biomedcentral.com/1472-6947/14/117 Background Methods Management of diabetes mellitus is complex, and in- Study overview volves controlling multiple risk factors that may lead to This study consisted of five phases: 1) development of the complications. However, care gaps exist: the Behavioral intervention, 2) feasibility testing, 3) usability testing; 4) Risk Factor Surveillance System has estimated that only refinement of the intervention, and 5) evaluation of the 68% of patients with type 1 or type 2 diabetes had intervention using a cohort study and individual inter- HbA1c measured at least twice in the previous year [1], views. The study protocol and results of the first four despite a recommendation from the American Diabetes phases are reported elsewhere [15,16]. We report here the Association that it be measured at least two to four results of Phase 5. times per year [2]. Given that patients provide most of their own diabetes care, patient self-management train- Diabetes online companion: a web-based self- ing is an important strategy for improving quality of care management intervention [3], particularly in the current era of patient-centred out- The Diabetes Online Companion is a self-contained dia- comes and comparative clinical effectiveness research betes self-management website that was systematically [4]. Patient self-management interventions have demon- developed according to self-efficacy theory. Self-efficacy strated benefits in terms of both quality of life [5] and refers to “beliefs in one’s capabilities to organize and glycemic control [6], but participation is low [7], effect- execute the courses of action required to produce given iveness wanes over time [6], and access to trained pro- attainments” [17]. Randomized controlled trials have fessionals to support self-management is limited [8]. shown that diabetes self-management education pro- Web-based self-management interventions are promis- grams incorporating principles of self-efficacy are associ- ing because they offer ease of access for patients who ated with improvements in knowledge [18], health are computer-literate, and they can be scaled up with lit- behaviours [18,19], self-efficacy [18-20], HbA1c [18-21], tle cost [9]. Web-based media have improved patient weight [18], and microvascular complications [19]. knowledge, the extent of behaviour change, and clinical Our intervention incorporated evidence-based content outcomes for a range of conditions [10]. However, prin- and behaviour-change strategies and followed the princi- ciples of effective education, self-management support, ples of user-centred design [15]. The website had four and behaviour change have not been incorporated into main components: 1) general information (static), 2) tai- current diabetes-related websites [11-13]. Reviews of lored information (interactive), 3) self-monitoring logs existing diabetes websites showed that they presented (interactive), and 4) a blog (interactive) (see Additional file didactic information of variable quality, they required 1 for sample screenshots). We posted a total of 53 blog advanced reading levels, and they followed a static, posts over the intervention period, initially at a frequency newspaper-format display, rather than harnessing the in- of one per week. After four weeks of limited user activity, herent advantages of websites, such as interactive tech- we increased the frequency of blog posts to two per week nology, social support, and problem-solving assistance and added email prompts with each new posting. The [11,13]. A systematic review of electronic diabetes- topics, which covered medical content, diabetes-related related tools found that they had moderate but incon- news items, and practical issues, were selected on the basis sistent effects on a variety of psychological and clinical of our feasibility and usability testing [15]. In addition, par- outcomes, including HbA1c and weight; tools that were ticipants received weekly email reminders to visit the site more interactive tools were associated with continued or complete their self-management trackers, as well as no- website use and greater clinical improvement [10]. In tices of any new content [15]. addition, greater website use was correlated with greater clinical improvements: regular website users had greater reductions in HbA1c compared with intermittent users. Cohort study Although this finding could be a consequence of the Participants healthy user effect [14], addressing usability issues to in- We conducted a single-arm pre-post cohort study. Con- crease the proportion of regular users may increase the secutive series of individuals with diabetes were recruited effectiveness of interventions. from two family practice units and two endocrinology In a previous study, we developed an approach to clinics in Toronto (one each from two academic health address many of these limitations of existing web-based science centres). Those eligible for inclusion were aged ≥ interventions [15]. In the current study, we tested the 25 years with at least one of HbA1c > 7.0% (53 mmol/ impact of this approach on self-efficacy, quality of life, mol), systolic blood pressure > 130 mmHg, low-density- self-care, blood pressure, cholesterol, glycemic control, lipoprotein cholesterol (LDL-C) > 2.0 mmol/L, or body and exercise promotion amongst people with type 2 mass index (BMI) > 25 kg/m . We excluded those who diabetes. had Canadian Cardiovascular Society class 3 or 4 angina, Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 3 of 14 http://www.biomedcentral.com/1472-6947/14/117 did not speak English, were not available for follow-up, or recorded secular events that might have affected our out- had no regular access to the telephone and internet. comes (such as diabetes-related news reports). Sample size calculation Outcomes Using a range of correlations from 0.2 to 0.8, a signifi- Website usage: We analyzed logs for the web server to cance level of 0.05, and a power of 80%, we calculated assess the frequency and duration of specific compo- that a sample of at most 52 participants was required to nents of the intervention [16]. Specifically, we collected detect a change of 0.5 units in self-efficacy score after data for the following variables: duration of use by indi- the intervention (relative to the score before implemen- vidual users, frequency of use, site penetration, most fre- tation). Differences of 0.1 to 0.5 in self-efficacy score quently accessed tools and pages, and patterns of use have been correlated with metabolic control, eating be- over time. haviour, exercise behaviour, and other self-management Patient-centred outcomes: We assessed self-efficacy, our behaviours [22,23]. A formula for paired mean compari- primary outcome, with the Modified Grossman Self- sons was applied [30], and the longitudinal nature of the efficacy for Diabetes Scale, which has moderate to high re- study increased its power [31]. A previous analysis re- liability (Cronbach’s alpha = 0.51 to 0.86; Additional file 1) ported a dropout rate of 20%–51% in studies of self- [22,23]. We selected self-efficacy because not only has it management [32]; we further adjusted the sample size to been validated in predicting and promoting patient behav- account for an expected dropout rate of 40%. iour change, but it also has been demonstrated to improve clinical outcomes [18,20,24,25]. We assessed self-care be- Data analysis haviour with the Summary of Diabetes Self-Care Activities Linear mixed models were used to examine the effect of Measure – Revised [26] and diabetes-specific quality of life the intervention and time (intervention × time inter- with the Diabetes Distress Scale [27]. These patient-based action) on self-efficacy, self-care, and diabetes distress. outcomes were selected because they are relevant mea- We selected these models to accommodate the complex- sures of knowledge use by patients. ities of typical longitudinal data sets for continuous out- Clinical outcomes: We collected data on HbA1c, sys- comes; specifically, they allowed us to account properly tolic and diastolic blood pressure, LDL-C, and weight for both within- and between-participant variability every three months. These outcomes were chosen to in- [33,34] and have been used in previous studies for simi- form the sample size calculations in future trials. lar analyses [35-37]. The models were also adjusted for age, sex, ethnicity, income (above or below Can $30, 000), Data collection education, employment, and health literacy, as each of We obtained data for age, sex, ethnicity, education, self- these variables could affect the study outcomes [21,38,39]. reported health literacy, employment, duration of diabetes, The model examining the self-care outcome was also complications, smoking status, medications, HbA1c, sys- adjusted for interaction terms of the aforementioned tolic blood pressure, LDL-C, weight, current use of and variables with time. No additional interaction terms with comfort with a computer and the internet, self-care score, time were included for other outcome models, because all self-efficacy score, and quality-of-life score at baseline. Out- additional interaction terms examined were non-significant. come data were collected by means of patient-completed To avoid inflation of R , all variables were specified a priori, questionnaires. For the pre- and post-implementation and all interactions were tested simultaneously using a phases, aggregates of patient-completed questionnaires cut-off value of 0.30 [40]. Models were assessed by means were obtained every three weeks for nine months through of residual plots. web-based surveys, resulting in 12 data points for each To assess the potential effect of missing income data phase. Health literacy was measured by a three-item vali- for three of the participants, a sensitivity analysis (imput- dated questionnaire completed by the patients [28,29]. ing income as both high and low) was performed. Miss- HbA1c and LDL-C were collected from medical records ing health literacy data for 15 of the participants were via chart audit. Systolic and diastolic blood pressures were imputed using the mode of the distribution, because measured by the research coordinator and were recorded 95% of the remaining participants were health literate. as the average of three readings. Weight was also measured Linear mixed models were also used to examine the effect by the research coordinator. At the end of the study, each of the intervention and time (intervention × time inter- participant was asked to disclose whether he or she had action) on secondary outcomes. These models were ad- used other web-based interventions and if so, whether justed for age, self-efficacy score, income, ethnicity, and those interventions employed text- or image-based didactic insulin use (for HbA1c and weight only). We also com- materials, interactive technology, or behavioural strategies. pared the effect of the intervention between users and non- To assess for threats to validity from historical effects, we users of the website. Finally, we used descriptive statistics to Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 4 of 14 http://www.biomedcentral.com/1472-6947/14/117 analyze website usage. R software version 2.1.15 was used baseline self-efficacy, self-care, and diabetes distress are for all analyses [41]. reported in Table 1 (Demographic characteristics and baseline values of observational cohort and qualitative Interviews study). Individual interviews were conducted 2 to 21 weeks after completion of quantitative data collection. We used a pur- Website use posive sampling strategy to recruit participants with a The mean number of days on which users logged in dur- range of experiences and characteristics [42] (sex, age, eth- ing the study period was 8.2 days (standard deviation nicity, duration of diabetes, educational attainment, in- 13); the median was three days. The average frequency come) from the broader pool of cohort study participants. of use was 0.7 logins/month, or one visit every 5.8 weeks, We developed a semi-structured interview guide to elicit distributed as follows: non-user: 11 participants (14%); participants’ views regarding the following website features: infrequent user (<2 times/month): 61 participants (75%); acceptability, usability, strengths and weaknesses of the frequent user (>2 times/month): seven participants (9%); intervention, facilitators and barriers to its use, user satis- heavy user (>1 time/week): two participants (2%). Web- faction, and sustainability of use (Additional file 1). We site usage across all users ranged from 4 to 50 logins/ made the website available during each interview, in case week (median 14.5/week), with peaks of 50 logins in the interviewee wanted to show the interviewer something week 10 and 37 logins in week 27. Increased use of the on the website. website during those weeks appeared to be driven by the All interviews were audiotaped and transcribed verba- blog. In general, website use appeared to parallel blog tim [43]. Transcripts were inductively analyzed to iden- use, with users visiting the blog repeatedly during the tify emergent categories and themes using a constant same login or visit (Figure 1). The most-accessed pages comparative approach [44]. Coding was conducted inde- during week 10 were the blog (regarding medication log, pendently by three team members with expertise in supplements, and insulin) (34% of hits) and the blood qualitative research methods (CHY, JAP, SH) [44]. After pressure (8%) and medication (9%) logs. For week 27, coding an initial subset of interviews, a preliminary cod- the most-accessed pages were the blog (regarding foot ing framework was developed on the basis of the emer- care) (44% of hits), “My blood glucose log” (32%), and “7 ging analysis, with discussion and consensus amongst steps to take care of your feet” (3%). the analysts [45]; the framework was then iteratively Overall, the most frequently accessed tools, for both tested and refined with subsequent interviews [44]. The- first-time and return users, were the blog, followed by matic saturation was attained with 21 interviews [42]. “My blood glucose log”, “My medication log”,and “My ac- NVivo software (version 9) was used to assist with data tivity log”. Regarding site penetration, users viewed 6.6 management and retrieval. Techniques to ensure analytic pages per session, spending an average of 5 minutes 43 sec- rigour included use of multiple analysts, negative case ana- onds on the site, and 1 minute 39 seconds per page. lysis, and triangulation of the qualitative findings with the quantitative results [42,44,46]. Triangulation consisted of Blog use 1) examining the interview data through the lens of “effect Within the blog section of the website, there were a total on self-efficacy”, 2) corroborating qualitative findings with of 569 page views by 35 participants over the study period, quantitative data, and 3) interrogating how the Diabetes with peaks at week 10 (54 views), week 27 (43 views), and Online Companion affected self-efficacy [46]. week 30 (53 views), corresponding to blog entries about the medication log, supplements and insulin, and foot and Research ethics kidney care, respectively. A total of 13 comments respond- The study was approved by the Research Ethics Boards ing to the blog postings were submitted by five partici- of St. Michael’s Hospital (reference number 09–091) and pants. These comments took the following forms: 1) Sunnybrook Health Sciences Centre (reference number responding to the blog (agreement or disagreement); 2) 177–2009). All participants gave written and verbal in- requesting help with or providing feedback on the website; formed consent. 3) requesting help with self-management; 4) offering assist- ance, empowerment, and their own solutions (including Results food recipes); 5) self-reporting behaviour change; 6) shar- Cohort study ing responses to medication; and 7) warning others about Of the 98 participants recruited, 81 had complete data col- interactions with health care providers. lection for at least two time points (one before and one after the intervention was implemented) and were in- Use of interactive and static tools cluded in the analysis. The questionnaire response rate for Overall, 47 (67%), 63 (90%), and 43 (60%) of 70 users vis- these 81 participants was 83%. Patients’ characteristics and ited static, interactive, and log pages, respectively, at least Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 5 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 1 Demographic characteristics and baseline values of observational cohort and qualitative study Participants Observational cohort (%, n = 81) Qualitative study (%, n = 21) Sex Male 44 (54%) 9 (43%) Female 37 (46%) 12 (57%) Age (years) 20–39 7 (9%) 2 (10%) 40–59 37 (46%) 7 (33%) 60–79 36 (44%) 12 (57%) > 80 1 (1%) 0 Ethnicity White 50 (62%) 17 (81%) Asian 24 (30%) 4 (19%) African American 6 (7%) 0 Hispanic 1 (1%) 0 Duration of diabetes mellitus (years) < 5 25 (31%) 6 (29%) 5–9 16 (20%) 5 (24%) 10–14 19 (23%) 3 (14%) 15–20 16 (20%) 5 (24%) > 20 years 5 (6%) 2 (10%) Education < High school 1 (1%) (0) High school 11 (14%) 1 (5%) College 21 (26%) 5 (24%) University 48 (59%) 15 (71%) Employment status Employed 45 (56%) 11 (52%) Retired 24 (30%) 7 (33%) Unemployed 7 (9%) 1 (5%) Disability 2 (2%) (0) Student 3 (4%) 2 (10%) Annual income (Can$) <15 000 17 (21%) 3 (14%) 15 000 to 29 999 8 (10%) 4 (19%) 30 000 to 59 999 22 (27%) 7 (33%) 60 000 to 89 999 23 (28%) 6 (29%) >90 000 11 (14%) 1 (5%) Insulin use Yes 48 (59%) 7 (33%) No 33 (41%) 14 (67%) Purpose of computer use* Business 2 (2%) 1 (5%) Personal 24 (30%) 5 (24%) Both 54 (68%) 15 (71%) Frequency of computer use* < 1 time/week 3 (4%) 0% 1–2 times/week 5 (6%) 0% 3–6 times/week 9 (11%) 0% ≥ 1 time/day 63 (79%) 21(100% ) Comfort with computer use Somewhat uncomfortable 2 (2%) 0% Neutral 5 (6%) 0% Somewhat comfortable 26 (32%) 8 (38%) Very comfortable 46 (57%) 13 (62%) Did not respond 2 (2%) 0% Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 6 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 1 Demographic characteristics and baseline values of observational cohort and qualitative study (Continued) Frequency of Internet use for diabetes < 1 time/week 66 (84%) 18 (85%) 1–2 times/week 7 (9%) 1 (5%) 3–6 times/week 5 (6%) 2 (10%) ≥ 1 time/day 1 (1%) 0% Comfort with Internet use Somewhat uncomfortable 2 (2%) 0% Neutral 3 (4%) 0% Somewhat comfortable 29 (36%) 8 (38%) Very comfortable 47 (58%) 13 (62%) Did not respond 0% 0% Self-efficacy; mean (SD) 4.61 (0.58) 5.11 (0.52) Self-care; mean (SD) 3.35 (1.12) 3.31 (0.84) Diabetes distress; mean (SD) 40.75 (16.21) 37.33 (14.74) HbA1c; mean (SD) 7.64% (1.29) 7.17% (0.98) 60.0 mmol/L (14.1) 54.9 mmol/L (10.7) Systolic blood pressure (mm Hg); mean (SD) 129.24 (13.84) 124.10 (8.73) Diastolic blood pressure (mm Hg); mean (SD) 76.03 (8.51) 74.71 (8.87) LDL-C (mmol/L); mean (SD) 2.11 (0.78) 2.20 (0.73) Weight (kg); mean (SD) 90.65 (21.76) 84.47 (15.13) *Data missing for one participant. Based on Statistics Canada data for low income cut-off [43], we selected $30 000 as a minimal level of income comfortable for activities of daily living and self-management capability for our analysis. Data missing for two participants. Abbreviation: LDL-C Low-density lipoprotein cholesterol. once. These users had a mean of 3.4, 4.5, and 9.3 visits/user implementation trajectory (effect: 0.12; 95% CI: −−0.028, to each of these page types, respectively. 0.263; p = 0.11; Figure 2 and Table 2). Self-care: The self-care score improved by 0.44 (95% Patient-centred outcomes CI: 0.23, 0.63; p < 0.0001) beyond what was expected at Self-efficacy: Despite a significant short-term increase in nine months (Figure 2 and Table 2). self-efficacy score immediately after implementation of the Diabetes distress: Despite an immediate short-term intervention (0.13; 95% confidence interval [CI]: 0.06, 0.20; decrease in diabetes distress score (−−2.29; 95% CI: −3.76, − p < 0.0004), by nine months, this outcome had not 0.81; p = 0.002), by nine months, this outcome had not increased significantly more than expected from its pre- decreased significantly over what was expected (effect: 1.84; 95% CI: −4.81, 1.12; p = 0.22; Figure 2 and Table 2). There was no interaction with insulin use by time or intervention for any of these outcomes. Self-care scores were positively correlated with age (0.04/year, 95% CI: 0.02, 0.06), p <0.001). Diabetes distress varied with age and sex: younger fe- male participants had greater diabetes distress. When we conducted the sensitivity analysis with missing values for income assumed to be below $15,000 and miss- ing values for health literacy assumed to be the mode, there were no changes in results for self-efficacy, self care, or diabetes distress. Clinical outcomes Seventy-three of the participants were included in the Figure 1 Website login and blog use by week. Black bar: Number analysis of clinical outcomes. The other eight partici- of logins per week. Grey bar: Number of blog views per week. pants were excluded because of missing data for HbA1c, Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 7 of 14 http://www.biomedcentral.com/1472-6947/14/117 the exception of diastolic blood pressure (users: +3.27 mm Hg; non-users: −1.6 mm Hg; p = 0.014; Additional file 1). Interviews Twenty-one individuals (Table 1) participated in an inter- view. The sample consisted of White and Asian men and women of various ages, duration of diabetes, educational attainment, and employment status, who used computers frequently and were comfortable with using the internet. Analysis of the interviews revealed numerous themes, four of which were most relevant to interpretation of the co- hort study’s negative results, in particular, exploration of why participants used the website to only a limited extent. Additional themes will be the focus of future publications. The following four themes are considered here: 1) barriers and facilitators of website use; 2) patterns of website use, including the role of the blog in driving site traffic; 3) gen- eral feedback on website characteristics; and 4) potential mechanisms for the effect of the website on self-efficacy, behaviour change, and diabetes distress. Representative quotes for each theme appear in Table 3. 1) Barriers and facilitators to use: Participants stated that they struggled with competing health and life con- cerns. They reported that it was “not just diabetes” that they dealt with (Table 3; 1a) and that they had to man- age other concurrent medical conditions (Table 3; 1b). They spoke about their attempts to balance illness work with everyday life work; they found that after completing the latter, “there wasn’t a lot” of time or energy left for self-management of their disease, much less to use the website (Table 3; 1b). Some participants identified lack of motivation as “a me thing, as opposed to a site thing”, while others commented that “laziness” (Table 3; 1c) was Figure 2 Self-efficacy, self-care, and diabetes distress nine months a barrier to use. before and nine months after intervention implementation. Participants’ attitudes toward diabetes also coloured Reference categories used in the plot were as follows: female, mean their approach to self-management and thus their use of age 57.34 years, employed, university education, income > Can the site. In particular, participants reported feeling frus- $30,000, adequate health literacy, white. trated with the uncontrolled nature of their disease, and the collection of self-monitoring information that showed blood pressure, LDL-C, or weight within 90 days of the a lack of metabolic control exacerbated this frustration self-efficacy data or because no data were obtained after (Table 3; 1d). Similarly, some participants said that they implementation of the intervention. The intervention had were sometimes overcome with a sense of futility. They no effect on HbA1c, blood pressure, LDL-C, or weight in perceived that regardless of their actions, some outcomes either the unadjusted or the adjusted models (Table 2). such as dialysis were inevitable (Table 3; 1e); hence, they Comparison of users and non-users: saw no value in learning about the disease or in trying to A total of 70 participants (86%) used the website at least self-manage the disease or use the website. once, whereas 11 (14%) did not use the site at all. At the Eleven (52%) of the 21 interview participants said the re- nine-month follow-up after implementation of the inter- quirement for a login and password prevented them from vention, there was no difference between users and non- using the website because they often forgot their password users in terms of self-efficacy (0.15 vs. 0.13, p = 0.35) or (Table 3; 1f). Others were limited by poor computer or self-care (0.18 vs. 0.13, p = 0.21). Users had a greater internet access and said they would prefer a mobile solution reduction in diabetes distress than did non-users (−4.7 (Table 3; 1g). Finally, some participants noted that the oner- vs. -0.9, p < 0.0001). There was no difference in the effect ous process for correcting error in log entries discouraged of using the intervention on any secondary outcomes, with them from using the self-management tools (Table 3; 1h). Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 8 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 2 Summary statistics of psychological and clinical outcomes (with 95% confidence interval), at implementation of intervention and 9 months later Outcome Immediate change p-value Change in outcome at study end, p-value (at time of intervention) over expected value Self-efficacy 0.13 (0.06, 0.20) 0.0004 0.12 (−0.028, 0.263) 0.11 Self-care 0.26 (−0.17, 0.68) 0.24 0.44 (0.23, 0.63) <0.0001 Diabetes distress −2.29 (−3.76, −0.81) 0.002 −1.84 (−4.81, 1.12) 0.22 HbA1c 0.37 (−0.11, 0.85) 0.13 −0.055 (−1.13, 1.02) 0.93 Weight (kg) 0.37 (−3.00, 3.74) 0.83 −1.10 (−9.31, 7.12) 0.77 Blood pressure (mm Hg) Systolic 1.79 (−5.87, 9.26) 0.66 −5.89 (−18.26,12.37) 0.53 Diastolic 0.43 (−4.13, 4.98) 0.85 −8.22 (−19.18, 2.74) 0.14 LDL-C (mmol/L) −0.0006 (−0.26, 0.26) 0.996 0.14 (−0.55, 0.82) 0.72 Abbreviation: LDL-C Low-density lipoprotein cholesterol. In contrast, other website characteristics appeared to en- the blog afforded them the opportunity to obtain “expert” courage users to visit and return. The perceived reliability medical answers anonymously (Table 3; 2c,i). The provision of the website’s information and the perception that it was of “fellow patient” advice was characterized as promoting a an “authoritative source” drew users (Table 3; 1i). In sense of community that some participants felt might com- addition, email reminders prompted them to return; such bat their sense of isolation (Table 3; 2c,ii). However, this prompts seemed well-suited to what users characterized was not a universal sentiment, and some participants felt as a “fast-paced world” and served as effective reminders uncomfortable with and disconnected from the blog to make diabetes self-care a high priority (Table 3; 1j). (Table 3; 2c,ii). In addition, there was a tension between a Similarly, routinization of the online experience appeared desire for online community and a fear of “looking foolish”. to routinize use of the internet for certain aspects of For example, some participants said they were afraid of health care. For example, participants reported that in- “putting [up] a stupid question” or surmised that others creasing their use of the internet and expanding their were “shy about how they write” (Table 3; 2c,iii). Other rea- scope of internet activities created a new norm for internet sons offered for not contributing to the blog included par- usage, such that it became more commonplace for them ticipants’ perceptions that they had nothing to offer, that to “look up whatever is interesting to me”, “track stuff on- the blog was not their preferred mode of communication, line”, and embrace “going paperless” (Table 3; 1 k). and that they preferred face-to-face communication 2) Patterns of website use, including role of the blog: (Table 3; 2c,iv). There were mixed views regarding the reli- Participants said that use of the website was driven by ability of the blog content: one user commented that a blog their individual context and circumstances. Rather than represented “the blind leading the blind”,but others said browsing at random, users said they were goal-directed: they were reassured by the fact it was moderated by an ex- when they had a specific concern, they focused on that pert (Table 3; 2c,iv). area of the website (Table 3; 2a). For example, one par- 3) General feedback on website characteristics: Partici- ticipant was initially motivated to visit (and subsequently pants commented on their impressions of the website continued to visit) the foot care section of the website overall and provided feedback concerning general features because of her foot symptoms (Table 3; 2b). Participants that were appreciated. Participants perceived that the web- also commented that they used the website to gauge the site was accurate, comprehensive (Table 3; 3a), and easy to urgency of their concerns and to try to obtain immediate navigate (Table 3; 3b). While they appreciated the website’s answers to their questions (Table 3; 2b). provision of “evidence-based medical content”,they re- We explored potential reasons for the unexpected find- ported a desire for more practical solutions and “real life ing that the blog was the most frequently accessed tool answers” (Table 3; 3c) and spoke about finding a balance and appeared to drive website usage. Participants identi- between these two characteristics. fied dual roles of the blog in providing a forum for both 4) Mechanisms of effect of the website on self-efficacy, “expert” and “fellow patient” advice (Table 3; 2c). “Fellow behaviour change, and diabetes distress: Despite appar- patient” advice consisted of submitting suggested content ently limited use of the website, the intervention appeared or commenting on blog posts (i.e. participating in a dis- to have an effect on self-efficacy, behaviour change, and cussion thread). For patients who were otherwise uncom- diabetes distress. Deeper exploration of the data regarding fortable with asking questions of a health care provider, patterns of use and website features uncovered factors that Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 9 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 3 Themes identified and representative quotes from in-depth interviews Theme/subtheme Representative quotes 1) Barriers and facilitators of website use: Barriers to website use Patient-related barriers a) Competing health concerns “Being 4 cancers in my life”. [3B53, 78-year old woman] “it’s not just diabetes that I deal with”. [2B16, 37-year old woman] b) Competing life concerns “I think for me, health issues surround migraines and fatigue. By the time I was home and had any free time, there wasn’t a lot there. And then life style and other stresses the business and other stresses the business…” [2B16, 37-year old woman] c) Lack of motivation “It’s a me thing, as opposed to a site thing”. [2B09, 47-year old woman] “Laziness”. [3B52, 65-year old woman] d) Frustration with diabetes “Part of it is, when you see the blood sugar is really high, I already know it’s high. I’m not taking the medication. So to log the fact that they are high, ends up making you more frustrated. So why do that”. [2B09, 47-year old woman] e) Futility regarding diabetes Sometimes I think no matter what I do it’s not going to matter because you read about this… they all end up on dialysis… they all, you know…” [2B09, 47-year old woman] Website-related barriers f) Login and password requirement “I keep forgetting my password”. [3B27, 58-year old woman] “The ability to change the password is good as well… Because I think that helps in returning, otherwise I would always go back and find your original email”. [2B16, 37-year old woman] g) Limited computer and/or internet “I didn’t have too much time to look at it ’cause I don’t have a computer at home access [3B39, 67-year old man] “I wish I can have a mobile like app”. [3A10, 52-year old woman] “on the road for work and limited web access”. [1A13, 47-year old man] h) Onerous nature of data entry to use “Um, but cause I did go in and I did try and do the tracking and I think cause I thought that was on logs an ongoing basis was the most useful part of it. But it was kind of a pain in the neck to use it… and kind of a pain in the ass getting where I wanted to go. I put some information and I wanted to delete it and I don’t know if I ever succeeded in getting rid of it”. [2B01, 48-year old man] Facilitators of website use i) Perceived reliability of information “It has to be really, really tuned in or connected to the latest developments either at [hospital name] or the world… With tons of information by real authorities. If I had a diabetic questions, I thought you guys might answer it, I would consider this to be an authoritative source”. [2B01, 48-year old man] j) Reminders, prompts of new content “Actually love getting the reminders, I really do. In this day in age everybody is busy, but I’m really good about it. The day I get the reminder, I find time to go in, I read the blogs… As I said, if you told me there was something new under a certain heading, I would definitely go in. I just like the reminders, in this fast paced world you don’t always have time for things and the last thing you think about is yourself. So I think the reminders are a great thing, and I always look”. [3B09, 61-year old woman] k) Increasing comfort with internet use “Slowly I got used to [the internet] and since that time I utilize the possibility to look after whatever is interesting to me”. [3B53, 78-year old woman] “I starting writing a lot of emails to lawyers and the public trustee so I’m kind of used to no, more so than before, tracking stuff online. Kind of going paperless is neat”. [2B01, 48-year old man] 2) Patterns of use of the website, including role of blog in driving site traffic a) Goal-directed use “And that’s kind of how I use the tool. It’s not as much a “Do I go in daily and read a new section each day?” No, it’s when I have a specific concern that I hone in on it to a degree”. [2B09, 47-year old woman] “Right, so there are a lot of places [in the website to which] you can go. But if those places don’t impact you… For example, if you don’t have high blood pressure, if it doesn’t impact you, you are less likely to go back to it. Then when it impacts you, then you know the resource is there”. [2B09, 47-year old woman] b) Use in context of current concern “Because the bottom of my toes is not sensitive and the side of my foot is not sensitive. So I looked up this website and tried to find out what could I do and is there any kind of help that I could find”. [3B52, 65-year old woman] Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 10 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 3 Themes identified and representative quotes from in-depth interviews (Continued) -Motivation for repeated use “The foot complication is the one I usually look up”. [3B52, 65-year old woman] -To gage urgency “I would look something to see if I thought if it was urgent”. [3B13, 54-year old woman] “Now, I have to say yesterday in desperation I actually tried to look up some medication information on this website…” [2B09, 47-year old woman] c) Role of blog in driving site traffic: “When new blogs coming up then I read it. Not always at once, but still eventually”. [3B53, 78-year old woman] Dual roles of blog “‘Ask the Expert’ should remain open all the time… And maybe it’s not just ‘Ask the Expert’; maybe it’s ‘Ask someone’. So you need an ‘Ask the Expert’ section and ‘Ask the Fellow Patient’ section”. [2B09, 47-year old woman] i) Opportunity for anonymous request “It provides an opportunity where people however anonymous they want to be and it might be for information something someone could be shy asking a heath care practitioner meanwhile go ahead and ask on something like that. So in that sense, I think that’s useful”. [2B16, 37-year old woman] ii) Fostering a sense of community “It’s interesting because people will come with a question through [the blog, such as] “Well, this worked for me”,or “This will work for everyone.” It is a sort of communal sense and conversation”. [3B53, 78-year old woman] “So, I think learning to develop your support systems is extremely important for a diabetic. And that having a forum where even if you don’t have a lot of people in your life that you can talk to about this, but having a forum where maybe you can go on and have an online community can be very helpful. Provided that they understand that any information that they read has to be verified”. [3B13, 58 year-old woman] -Social support “For example, if you are feeling yourself alone and loneliness is not a very good thing health-wise. It leads to depression and everything else and if you feel that you have to communicate, and then communicate here”. [3B49, 46-year old man] -Though this was not a universal sentiment: “Sometimes with a blog it can be communal and it can be conversation, but it can also make people others felt a lack of connection feel disconnected because: how comfortable are they in conversation?” [2B16, 37-year old woman] iii) However, tension between desire for “Because we are afraid that we are putting a stupid question”. [3B53, 78-year old woman] online community and fear of ridicule, “I found at times, responding to other posts, that [I felt] shy about responding. But I did go ahead and timidity do that even though I thought at times that’s not necessarily my usual style”. [2B16 37-year old woman] “It might take time to catch on, because people who are feeling shy about it may need to see what other people do with it for quite a while”. [2B16, 37-year old woman] “I think some people are shy about how they write.” [2B16, 37-year old woman] “I’ve just never been interactive…” [3B09, 61-year old woman] iv) Other reasons for not contributing to the blog: -Nothing new to contribute “But I’m not a poster. I read, I learn, I just don’t put my 2 cents in. [I don’t post because] I would say the same thing – why would I bother?” [3B09, 61-year old woman] -Reliability of content: Balance between NO “I find for the most part it’s the blind leading the blind. I guess this one is being moderated but by evidence anecdote, trusted source and large you have a bunch of people who don’t know anything kind of spewing forth”. versus patient centred [3B13, 54-year old woman] YES “However, the way this one is designed essentially it goes and it’s approved or reviewed before being posted in the first place. Which I think is a good style”. [2B16, 37-year old woman] -Perceived relevance to specific “I don’t know what the ages are but I’m thinking a younger person would be more familiar with demographic population communication in a way that is not face to face, more so”. [2B16, 37-year old woman] 3) General feedback on website characteristics: a) Accurate and comprehensive In fact, what I noticed is that their seemed to be more than what I would have anticipated… [2B01, 48-year old man] So there are lots of things on the website that really did help. So things like how to control the blood pressure and how to do all those other… there is lots of things that help. [2B09, 47-year old woman] b) Easy to navigate “The website itself is easy to navigate. And I think that list is very good and I’ve found that every time I have looked, [the answer] could typically could fall into one of those categories. The answer might not be there, but I know where to start to look”. [2B09, 47-year old woman] c) But could be more relevant: -Want practical solutions Maybe too many people are inundated with expert answers and not enough with real life answers. [3B09, 61-year old woman] Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 11 of 14 http://www.biomedcentral.com/1472-6947/14/117 Table 3 Themes identified and representative quotes from in-depth interviews (Continued) But that when you need it the most, that’s when I found it didn’t have everything I needed. It was still missing… the next piece; the next piece is what happens when you are in that mess? [3B49, 46-year old man] Yeah, it combines that very practical app, ok so here is an endocrinologist talking about blah, blah, blah, new scientific in road into diabetes… ok, that’s great on an intellectual level… What does this mean on a practical level? For those that want to and can handle the scientific information, perfect. And those of us who can’t or don’t want to then there’s other things on the site. [3B13, 54-year old female] What do you value about each type of information, the practical versus medical evidence based? 70 or 80 [percent of it practical] versus 30 and 20 for the medical. [3B53, 78-year old woman] But nowhere does it tell you how to deal with it. So, what I’m looking for and what I was looking for in that peer support was other people in the same position who have found solutions to issues that typical websites don’t tell you [2B09, 47-year old woman] 4) Potential mechanism of impact of website on self-efficacy, behavior change and diabetes distress a) Unanticipated pattern of use of DOC “I actually I haven’t done it lately but when I first got this I searched a lot of stuff, I was interested so I did search. I printed it out I actually have some in a binder, which I have referred to on occasion. It’s -Print out nice to have something black and white that you can refer to. That’s why I like certain websites, because you can print it off and read it. I can search other references and I’ve done that. And I keep a binder and I refer to it when…” [3B09, 61-year old woman] b) Unanticipated impact of reminder “There were the reminders that you hadn’t visited the site in a while. [When I got those reminders], emails I said, I will answer it later… This companion is a good reminder, I heard about it before but then let’s go back and check to make sure that I’m understanding properly…” [3B27, 58-year old woman] “Constantly reminding me about the things that we need to be aware of. Most of us know but the Online Companion was a good reminder and got me thinking of things that I need to constantly do (some that we conveniently forget)”. [3B55, 58-year old woman] c) Unanticipated role of DOC “Because of the recording of activity… I started… it guilted me into starting to record my activity and in “kickstarting” self-management behaviors once I started recording it because also coming off of the Metformin, I started recording it there and then I switched to recording it in books and now I simply record it on the calendar”. [1A06, 62-year old woman] Abbreviations: St. Mike’s St. Michael’s Hospital, Toronto, ON, Canada. App application. DOC Diabetes Online Companion. might account for these quantitative findings. For ex- distress than those who did not use it. Despite a user- ample, rather than returning to the site to revisit and re- centred design process and an increase in the frequency view items, some participants reported that they printed of blog posting from weekly to twice weekly, use of the items of interest from the website and subsequently re- website (as ascertained by login records) was limited. ferred to these paper copies (Table 3; 4a). The use of re- Our interviews revealed that both patient-related factors minder emails also had an effect. Participants reported (e.g. competing health and life concerns, a sense of futil- that these emails not only prompted them to return to ity) and website-related factors (e.g. requirement for and log into the website, but also encouraged them in login, limited computer or internet access) limited use of their own self-management (Table 3; 4b). Finally, one par- the website. We also found that participants were moti- ticipant noted that the website had kick-started her self- vated to access the website on the basis of their current management behaviours, by initially “guilting” her into needs and concerns, as well as new blog postings, with recording them online. Although she did not subsequently the blog fulfilling a need for both “expert” medical con- login to the website to record these behaviours, she did tent and peer support and a sense of community. continue to record them on paper. These qualitative findings have confirmed the import- ance of website features such as the reliability and authori- Discussion tativeness of information [47], as well as the use of blogs We found that a self-management website for patients [15] and reminders [48] for continued engagement of with type 2 diabetes led to no improvement in self- users. Our findings also emphasize the need to provide a efficacy, diabetes distress, or clinical outcomes over the greater proportion of “practical” patient-centred content. study period. However, there was an improvement in A recent qualitative study analyzing 3005 diabetes-related self-care (a secondary outcome), and the group that used blog posts showed that the most influential blogs were the website experienced significantly lower diabetes those written by patients and that only 10% of blogs cited Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 12 of 14 http://www.biomedcentral.com/1472-6947/14/117 biomedical literature [49], highlighting the dual needs for disease. For example, a systematic review of the benefits reliable, evidence-based content and engaging patient- and limitations of social media in the context of chronic based content. disease identified benefits (increased interaction and social Our data also suggest that mobile devices are a potential support, tailored and accessible information) and limita- avenue through which to improve accessibility and use of tions (quality concerns and lack of reliability, confidential- a self-management site. A Cochrane review of computer- ity, and privacy) [59] to those we identified. Similarly, our based diabetes self-management interventions identified finding of a reduction in diabetes distress in conjunction 16 randomized controlled trials, which showed a small ef- with no improvements in clinical outcomes echoes find- fect on glycemic control (−2.3 mmol/mol or −0.2%, p = ings from intervention strategies targeting other chronic 0.009), with the mobile phone subgroup experiencing a diseases. For example, another systematic review examin- greater effect (−5.5 mmol/mol or −0.5%, p < 0.00001) [50]. ing the effect of social media on psychological and physical Given the increasing preference for mobile devices over outcomes in chronic disease found a relatively large body desktop computers [51] for health information resources of evidence demonstrating psychological benefit (19 iden- [52], mobile technology may overcome the barriers to tified studies) but limited evidence for physical outcomes website access and use that we identified, through greater (4 identified studies) [60]. integration with patients’ existing routine, such that self- This study was limited by its non-randomized design. management is no longer seen as additional “illness work”. However, we employed a repeated-measures design that However, as with web-based technology, a systematic permitted reliable assessment of baseline self-efficacy. Al- approach to development, testing, implementation, and though our primary outcome (self-efficacy) was a non- evaluation of mobile health technology is warranted. Al- clinical outcome, it is a validated predictor of patient though such technologies are proliferating, with over 736 behaviour change and clinical outcomes [18,20,24,25]. The applications related to diabetes alone, their usability and infrequency of website use likely limited the effect of this clinical effectiveness are variable [53], and concerns exist intervention, but we obtained valuable insights regarding regarding their effectiveness and safety, as well as the mediators of website use through our individual interviews. security of personal health information [54]. Our findings The qualitative evaluation was conducted by individuals regarding user engagement with web-based technology who were also involved in developing the intervention, echo those for mobile technology: an evaluation of 10 which created a potential for bias; however, we guarded mobile diabetes applications emphasized the importance against this bias by including individuals who were not in- of user-centred design, an engaging interface, and context- volved in designing the website as members of the qualita- driven use [55]. tive analytic team and by having three coders. As such, we Competing health concerns were identified as a barrier were able to obtain and report critical feedback that par- to web-based self-management. Patients’ adherence to ticipants openly shared. Study strengths include the use of diabetes care is affected by multimorbidity (e.g. depres- multiple repeated measures, the use of validated out- sion), which in turn directly affects self-management comes, dual coding of all transcripts, and triangulation of ability and competes for time and attention [56]. For ex- the qualitative findings with the quantitative results ample, patients with a greater number of comorbidities [42,44,46]. placed a lower priority on diabetes and had worse dia- betes self-management ability [57]. Future interventions should consider strategies, such as shared decision- Conclusion making and priority-setting, to empower patients with mul- Increasing use of the World Wide Web by consumers tiple comorbidities to optimize their self-care [58]. For for health information and ongoing revolutions in so- example, a patient may identify mood management as a cial media are strong indicators that consumers are priority, which is key to subsequent self-care. Thus, greater welcoming and demanding a new era of technology in integration with the person’s cognitive, emotional, and health care. However, full potential of this technology health information-seeking preferences, daily living routine, is hindered by limited uptake and high attrition rates. and health context through the use of patient-based con- Use of the Diabetes Online Companion may be opti- tent, mobile devices, and individualized decision-making, mized by integrating a mobile interface, emphasizing may be further strategies to maximize website use and re- “practical” patient-centred content such as a patient- duce intervention attrition. led blog, and including a “prioritization” feature to Finally, our results may be extrapolated to other help users with competing concerns. Our research chronic diseases. In particular, our finding of the need findings have shed light on these limitations by identifying for tailored content and peer support, balanced with characteristics associated with website use and attrition concerns regarding information reliability and confidenti- and suggesting strategies to reduce website attrition as a ality, is applicable to other strategies for managing chronic way to potentially optimize clinical outcomes. Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 13 of 14 http://www.biomedcentral.com/1472-6947/14/117 Additional file 5. Cochrane J, Conn VS: Meta-analysis of quality of life outcomes following self-management training. Diabetes Educ 2008, 34:815–823. 6. Minet L, Møller S, Vach W, Wagner L, Henriksen JE: Mediating the effect of Additional file 1: Appendix A: Representative Screenshots of the self-care management intervention in type 2 diabetes: a meta-analysis Diabetes Online Companion. Appendix B: Description of primary of 47 randomised controlled trials. Patient Educ Couns 2010, 80:29–41. outcome scales. Appendix C: Semistructured interview guide. 7. Ruppert K, Uhler A, Siminerio L: Examining patient risk factors, comorbid Appendix D: Effect of Intervention comparing users and non-users conditions, participation, and physician referrals to a rural diabetes (after adjustment for time). Appendix E: References. self-management education program. Diabetes Educ 2010, 36:603–612. 8. Shaw K, Killeen M, Sullivan E, Bowman P: Disparities in diabetes self-management education for uninsured and underinsured adults. Competing interests Diabetes Educ 2011, 37:813–819. The authors declare that they have no competing interests. 9. Shekelle PG, Morton SC, Keeler EB: Costs and Benefits of Health Information Technology. Rockville: Agency for Healthcare Research and Quality (US); 2006. Authors’ contributions 10. Yu CH, Bahniwal R, Laupacis A, Leung E, Orr MS, Straus SE: Systematic CHY conceived of the study, collected and analyzed the data, and wrote the review and evaluation of web-accessible tools for management of manuscript. JAP contributed to the study design, analyzed the data, diabetes and related cardiovascular risk factors by patients and contributed to the interpretation of the data, and critically revised the healthcare providers. J Am Med Inform Assoc 2012, 19:514–522. manuscript for important intellectual content. MM, BRS, AL, and SES 11. Bull SS, Gaglio B, McKay HG, Glasgow RE: Harnessing the potential of the contributed to the study design and to interpretation of the data and internet to promote chronic illness self-management: diabetes as an critically revised the manuscript for important intellectual content. GL example of how well we are doing. Chronic Illness 2005, 1:143–155. analyzed the data and wrote part of the manuscript. SH conducted the 12. van Vugt M, de Wit M, Cleijne WH, Snoek FJ: Use of behavioral change interviews and qualitative analyses and drafted portions of the manuscript. techniques in web-based self-management programs for type 2 diabetes DN conducted the computer programming to develop and refine the patients: systematic review. J Med Internet Res 2013, 15:e279. website, collected data on web usage and critically revised the manuscript 13. Seidman JJ, Steinwachs D, Rubin HR: Design and testing of a tool for for important intellectual content. OB contributed to interpretation of the evaluating the quality of diabetes consumer-information Web sites. data and critically revised the manuscript for important intellectual content. J Med Internet Res 2003, 5:e30. CHY is guarantor for this article. All authors have given final approval of the 14. Shrank WH, Patrick AR, Brookhart MA: Healthy user and related biases in version to be published, and all agree to be accountable for all aspects of observational studies of preventive interventions: a primer for the work in ensuring that questions related to the accuracy or integrity of physicians. J Gen Intern Med 2011, 26:546–550. any part of the work are appropriately investigated and resolved. 15. Yu CH, Parsons J, Hall S, Newton D, Jovicic A, Lottridge D, Shah BR, Straus SE: User-centered design of a web-based self-management site for individuals Acknowledgements with type 2 diabetes – providing a sense of control and community. We thank Jovita Sundaramoorthy and Carolyn Gall-Casey of the Canadian BMC Med Inform Decis Mak 2014, 14:60. Diabetes Association for their input and feedback related to project design. 16. Yu C, Parsons J, Mamdani M, Lebovic G, Shah BR, Bhattacharyya O, Laupacis We also thank Kevin Thorpe for his invaluable input into the choice of A, Straus SE: Designing and evaluating a web-based selfmanagement site methods for the statistical analysis. for patients with type 2 diabetes - systematic website development and study protocol. BMC Med Inform Decis Mak 2012, 12:57. Funding 17. Bandura A: Self-Efficacy: the Exercise of Control. New York: W.H. Freeman; 1997. The study was funded by the Canadian Institutes of Health Research (CIHR) 18. Glasgow RE, Toobert DJ, Hampson SE, Brown JE, Lewinsohn PM, Donnelly J: Knowledge to Action Operating Grant (funding reference number KAL Improving self-care among older patients with type II diabetes: the 94484). Dr. SE Straus is supported by a Tier 1 Canada Research Chair. “sixty something…” study. Patient Educ Couns 1992, 19:61–74. 19. Reichard P, Toomingas B, Rosenqvist U: Changes in conceptions and Author details attitudes during five years of intensified conventional insulin treatment Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Postal address: 30 in the stockholm diabetes intervention study (SDIS). Diabetes Educ 1994, Bond St, Toronto, ON M5B 1W8, Canada. Department of Medicine, 20:503–508. University of Toronto, Toronto, ON, Canada. Dhalla Lana School of Public 20. Anderson RM, Funnell MM, Butler PM, Arnold MS, Fitzgerald JT, Feste CC: Health, University of Toronto, Toronto, ON, Canada. Applied Health Research Patient empowerment: results of a randomized controlled trial. Centre, St. Michael’s Hospital, Toronto, ON, Canada. Department of Physical Diabetes Care 1995, 18:943–949. Therapy, University of Toronto, Toronto, ON, Canada. Institute of Health 21. Trief PM, Teresi JA, Eimicke JP, Shea S, Weinstock RS: Improvement in diabetes Policy, Management and Evaluation, University of Toronto, Toronto, ON, self-efficacy and glycaemic control using telemedicine in a sample of older, Canada. Institute for Clinical Evaluative Sciences, G1 06, 2075 Bayview ethnically diverse individuals who have diabetes: the IDEATel project. Avenue, Toronto, ON M4N 3M5, Canada. Sunnybrook Research Institute, Age Ageing 2009, 38:219–225. Sunnybrook Health Sciences Centre, Toronto, ON, Canada. 22. Aljasem LI, Peyrot M, Wissow L, Rubin RR: The impact of barriers and self-efficacy on self-care behaviors in type 2 diabetes. Diabetes Educ 2001, Received: 22 July 2014 Accepted: 27 November 2014 27:393–404. 23. Grossman HY, Brink S, Hauser S: Self-efficacy in adolescent girls and boys with insulin-dependent diabetes mellitus. Diabetes Care 1987, 10:324–329. References 24. Nakahara R, Yoshiuchi K, Kumano H, Hara Y, Suematsu H, Kuboki T: Prospective 1. Centers for Disease Control and Prevention: Diabetes Report Card 2012. study on influence of psychosocial factors on glycemic control in Japanese Atlanta, GA: Centers for Disease Control and Prevention, US Department of patients with type 2 diabetes. Psychosomatics 2006, 47:240–245. Health and Human Services; 2012. 25. Lorig KR, Ritter PL, Jacquez A: Outcomes of border health Spanish/English 2. American Diabetes Association: Standards of medical care in diabetes chronic disease self-management programs. Diabetes Educ 2005, 31:401–409. 2013. Diabetes Care 2013, 36:S11–S66. 26. Toobert DJ, Hampson SE, Glasgow RE: The summary of diabetes self-care 3. Haas L, Maryniuk M, Beck J, Cox CE, Duker P, Edwards L, Fisher EB, activities measure: results from 7 studies and a revised scale. Diabetes Care Hanson L, Kent D, Kolb L, McLaughlin S, Orzeck E, Piette JD, Rhinehart AS, 2000, 23:943–950. Rothman R, Sklaroff S, Tomky D, Youssef G: National standards for diabetes 27. Polonsky WH, Fisher L, Earles J, Dudl RJ, Lees J, Mullan J, Jackson RA: self-management education and supportdiabetes. Diabetes Care 2014, Assessing psychosocial distress in diabetes: development of the Supplement 1:S144–S153. Diabetes Distress Scale. Diabetes Care 2005, 28:626–631. 4. Selby JV, Beal AC, Frank L: The Patient-Centered Outcomes Research Institute (PCORI) national priorities for research and initial research 28. Chew LD, Bradley KA, Boyko EJ: Brief questions to identify patients with agenda. J Am Med Assoc 2012, 307:1583–1584. inadequate health literacy. Fam Med 2004, 36:588–594. Yu et al. BMC Medical Informatics and Decision Making 2014, 14:117 Page 14 of 14 http://www.biomedcentral.com/1472-6947/14/117 29. Chew LD, Griffin JM, Partin MR: Validation of screening questions for 57. Kerr E, Heisler M, Krein S, Kabeto M, Langa K, Weir D, Piette J: Beyond limited health literacy in a large VA outpatient population. J Gen Intern comorbidity counts: how do comorbidity type and severity influence Med 2008, 23:561–566. diabetes patients’ treatment priorities and self-management? J Gen Intern 30. Zar JH: Biostatistical Analysis. Prentice Hall: Upper Saddle River, N.J; 1996. Med 2007, 22:1635–1640. 31. Liu G, Liang KY: Sample size calculations for studies with correlated 58. Joosten EA, DeFuentes-Merillas L, de Weert GH, Sensky T, van der Staak CP, observations. Biometrics 1997, 53:937–947. de Jong CA: Systematic review of the effects of shared decision-making 32. Norris SL, Engelgau MM, Narayan KM: Effectiveness of self-management on patient satisfaction, treatment adherence and health status. training in type 2 diabetes: a systematic review of randomized Psychother Psychosom 2008, 77:219–226. controlled trials. Diabetes Care 2001, 24:561–587. 59. Moorhead SA, Hazlett DE, Harrison L, Carroll JK, Irwin A, Hoving C: A new 33. Diggle P, Heagerty P, Liang KY, Zeger S: Analysis of Longitudinal Data. dimension of health care: systematic review of the uses, benefits, and Oxford, New York: Oxford University Press; 2002. limitations of social media for health communication. J Med Internet Res 2013, 15:e85. 34. Verbeke G, Molenberghs G: Linear Mixed Models for Longitudinal Data. 60. Merolli M, Kathleen G, Martin-Sanchez F: Health outcomes and related effects Verlag, New York: Springer; 2009. of using social media in chronic disease management: A literature review 35. Laird NM, Donnelly C, Ware JH: Review papers: longitudinal studies with and analysis of affordances. J Biomed Inform 2013, 46:957–969. continuous responses. Stat Methods Med Res 1992, 1:225–247. 36. Cnaan A, Laird NM, Slasor P: Tutorial in biostatistics: using the general linear mixed model to analyse unbalanced repeated measures and doi:10.1186/s12911-014-0117-3 Cite this article as: Yu et al.: A web-based intervention to support self- longitudinal data. Stat Med 1997, 16:2349–2380. management of patients with type 2 diabetes mellitus: effect on self- 37. Sherrill DL, Holberg CJ, Enright PL, Lebowitz MD, Burrows B: Longitudinal efficacy, self-care and diabetes distress. BMC Medical Informatics and analysis of the effects of smoking onset and cessation on pulmonary Decision Making 2014 14:117. function. Am J Respir Crit Care Med 1994, 149:591–597. 38. Figaro MK, Elasy T, BeLue R: Exploring socioeconomic variations in diabetes control strategies: impact of outcome expectations. J Natl Med Assoc 2009, 101:18–23. 39. Osborn CY, Cavanaugh K, Wallston KA, Rothman RL: Self-efficacy links health literacy and numeracy to glycemic control. J Health Commun 2010, 15:146–158. 40. Harrell FE, Lee KL, Mark DB: Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996, 15:361–387. 41. R Team: R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing; 2006. 42. Kuzel AJ: Sampling in Qualitative Inquiry. In Doing Qualitative Research. Edited by Crabtree BF, Miller W. Thousand Oaks: Sage Publications; 1999:33–45. 43. Poland BP: Transcription quality as an aspect of rigor in qualitative research. Qual Inq 1995, 1:290–310. 44. Kvale S: InterViews: An Introduction to Qualitative Research Interviewing. Thousand Oaks: Sage Publications; 1996. 45. Pope C, Ziebland S, Mays N: Qualitative research in health care: analysing qualitative data. BMJ 2000, 320:114–116. 46. Bryman A: Quantity and Quality. In Social Research. London: Routledge; 1988. 47. Lemire M, Paré G, Sicotte C, Harvey C: Determinants of internet use as a preferred source of information on personal health. Int J Med Inform 2008, 77:723–734. 48. Nijland N, van Gemert-Pijnen JE, Kelders SM, Brandenburg BJ, Seydel ER: Factors influencing the use of a web-based application for supporting the self-care of patients with type 2 diabetes: a longitudinal study. J Med Internet Res 2011, 13:e71. 49. Gruzd A, Black FA, Le Yen TN, Amos K: Investigating biomedical research literature in the blogosphere: a case study of diabetes and glycated hemoglobin (HbA1c). J Med Libr Assoc 2012, 100:34–42. 50. Pal K, Eastwood SV, Michie S, Farmer AJ, Barnard ML, Peacock R, Wood B, Inniss JD, Murray E: Computer-based diabetes self-management interventions for adults with type 2 diabetes mellitus. Cochrane Database Syst Rev 2010, 3. 51. Smith A: A: 35% of American Adults Own a Smartphone: One Quarter of Smartphone Owners Use their Phone for Most of their Online Browsing. Washington: Pew Research Center; 2011. 52. Manhattan Research: U.S. mobile health audience jumps to 95 million adults. Submit your next manuscript to BioMed Central Cybercitizen Health® U.S. 2013. Available at http://manhattanresearch.com/ and take full advantage of: News-and-Events/Press-Releases/mobile-health-95-million. 53. Demidowich AP, Lu K, Tamler R, Bloomgarden Z: An evaluation of diabetes • Convenient online submission self-management applications for Android smartphones. J Telemed • Thorough peer review Telecare 2012, 18:235–238. 54. Eng DS: The promise and peril of mobile health applications for diabetes • No space constraints or color figure charges and endocrinology. Pediatr Diabetes 2013, 14:231–238. • Immediate publication on acceptance 55. Årsand E, Frøisland DH, Skrøvseth SO, Chomutare T, Tatara N, Hartvigsen G, Tufano JT: Mobile health applications to assist patients with diabetes: lessons • Inclusion in PubMed, CAS, Scopus and Google Scholar learned and design implications. J Diabetes Sci Technol 2012, 6:1197–1206. • Research which is freely available for redistribution 56. Ciechanowski PS, Katon WJ, Russo JE: Depression and diabetes: impact of depressive symptoms on adherence, function, and costs. Arch Intern Med Submit your manuscript at 2000, 160:3278–3285. www.biomedcentral.com/submit

Journal

BMC Medical Informatics and Decision MakingSpringer Journals

Published: Dec 14, 2014

There are no references for this article.