Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Measuring financial toxicity as a clinically relevant patient‐reported outcome: The validation of the COmprehensive Score for financial Toxicity (COST)

Measuring financial toxicity as a clinically relevant patient‐reported outcome: The validation of... Original Article Measuring Financial Toxicity as a Clinically Relevant Patient-Reported Outcome: The Validation of the COmprehensive Score for financial Toxicity (COST) 1 1 2 3 Jonas A. de Souza, MD, MBA ; Bonnie J. Yap, MS ; Kristen Wroblewski, MS ; Victoria Blinder, MD, MSc ; 4 1 5 1 Fabiana S. Araujo,  PhD ; Fay J. Hlubocky, PhD ; Lauren H. Nicholas, PhD ; Jeremy M. O’Connor, MD ; 6 1 1 7 Bruce Brockstein, MD ; Mark J. Ratain, MD ; Christopher K. Daugherty, MD ; and David Cella, PhD BACKGROUND: Cancer and its treatment lead to increased financial distress for patients. To the authors’ knowledge, to date, no stan- dardized patient-reported outcome measure has been validated to assess this distress. METHODS: Patients with AJCC Stage IV solid tumors receiving chemotherapy for at least 2 months were recruited. Financial toxicity was measured by the COmprehensive Score for financial Toxicity (COST) measure. The authors collected data regarding patient characteristics, clinical trial participation, health care use, willingness to discuss costs, psychological distress (Brief Profile of Mood States [POMS]), and health-related quality of life (HRQOL) as measured by the Functional Assessment of Cancer Therapy: General (FACT-G) and the European Organization for Re- search and Treatment of Cancer (EORTC) QOL questionnaires. Test-retest reliability, internal consistency, and validity of the COST measure were assessed using standard-scale construction techniques. Associations between the resulting factors and other variables were assessed using multivariable analyses. RESULTS: A total of 375 patients with advanced cancer were approached, 233 of whom (62.1%) agreed to participate. The COST measure demonstrated high internal consistency and test-retest reliability. Factor analyses revealed a coherent, single, latent variable (financial toxicity). COST values were found to be correlated with income (correlation coef- ficient [r]5 0.28; P<.001), psychosocial distress (r5 -0.26; P<.001), and HRQOL, as measured by the FACT-G (r5 0.42; P<.001) and by the EORTC QOL instruments (r5 0.33; P<.001). Independent factors found to be associated with financial toxicity were race (P5.04), employment status (P<.001), income (P5.003), number of inpatient admissions (P5.01), and psychological distress (P5.003). Will- ingness to discuss costs was not found to be associated with the degree of financial distress (P5.49). CONCLUSIONS: The COST measure demonstrated reliability and validity in measuring financial toxicity. Its correlation with HRQOL indicates that financial toxici- ty is a clinically relevant patient-centered outcome. Cancer 2017;123:476-84. V 2016 The Authors. Cancer published by Wiley Periodi- cals, Inc. on behalf of American Cancer Society. This is an open access article under the terms of the Creative Commons Attribution- NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. KEYWORDS: cost of cancer, financial burden, financial toxicity, patient-reported outcome (PRO). INTRODUCTION Patients with cancer often are confronted with the financial consequences of cancer treatment, which may include signifi- 1,2 cant out-of-pocket costs, loss of income, and caregiver burden. The objective financial consequences of cancer, as well 3,4 as the subjective financial concerns, have been broadly termed “financial toxicity.” This burden has since been linked 5 6 with several clinically relevant patient outcomes, including health-related quality of life (HRQOL) ; symptom burden ; 7 8 compliance ; and, most recently, survival. Within the context of current policies and the Patient Protection and Affordable Care Act (ACA), many newly in- sured patients are expected to have a higher cost share because of preferential enrollment into high-deductible plans offer- 9-11 ing greater upfront affordability. Combined with escalating cancer costs, these high-deductible plans may increase financial distress and further exacerbate disparities in cancer care. To abrogate the impact of financial distress and Corresponding author: Jonas A. de Souza, MD, MBA, Department of Medicine, The University of Chicago Medicine, 5841 S. Maryland Ave, Chicago, IL 60637; Fax: (773) 702-3163; jdesouza@chicagobooth.edu 1 2 Department of Medicine, The University of Chicago Medicine, Chicago, Illinois; Department of Public Health Sciences, The University of Chicago Medicine, Chica- 3 4 go, Illinois; Center for Health Policy and Outcomes, Memorial Sloan Kettering Cancer Center, New York, New York; Institute of Psychology, Illinois Institute of 5 6 Technology, Chicago, IL; Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland; Department of Medicine, NorthShore University HealthSystem, Evanston, Illinois; Department of Medical Social Sciences, Northwestern University, Chicago, Illinois We thank Jocelyn Herrera, Kristen Kipping-Johnson, and Nancy Haefling for their support. Additional supporting information may be found in the online version of this article DOI: 10.1002/cncr.30369, Received: June 22, 2016; Revised: August 9, 2016; Accepted: September 8, 2016, Published online October 7, 2016 in Wiley Online Library (wileyonlinelibrary.com) 476 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al minimize its potential to augment disparities, there is an Measures Sociodemographic characteristics, health care use, urgent need for policy makers, researchers, and clinicians and willingness to discuss costs to accurately measure financial toxicity. However, to the Data related to age, sex, education, work status, marital sta- best of our knowledge, this patient-centered toxicity is tus, race, ethnicity, ECOG PS, and cancer type were col- rarely assessed in clinical practice or research. lected. We asked patients for their income from all sources The COmprehensive Score for financial Toxicity within the previous year. Household income was measured (COST) patient-reported outcome measure (PROM) was as a function of the federal poverty level (FPL).The FPL is previously developed by de Souza et al to assess financial used for Health Insurance Marketplace cost assistance un- toxicity in patients with cancer. In a multistep process, der the ACA, in which tax credit eligibility for health insur- 155 patients with advanced cancer were interviewed to de- ance ranges from 138% to 400% of the FPL in states that velop the 11-item COST measure. In response to the ur- decided to expand Medicaid. Financial toxicity was mea- gent need for such a tool, the instrument was adopted into sured by the COST measure (Fig. 1). Lower COST values clinical practice and research before the establishment of indicate more financial toxicity. General mood disturbance its psychometric properties. The current study examines or psychological distress was measured by the Brief Profile the COST measure with respect to its psychometric prop- of Mood States (Brief-POMS). Emergency department erties, or how well it measures the construct of interest. visits and inpatient admissions from 1 year before the inter- The importance of knowing these properties was empha- view date were collected as measures of health resource use. sized by the US Food and Drug Administration guidance To control for different communication preferences related on PROMs, in which it was recommended that an instru- to costs and thus any resulting bias when responding to the ment’s measurement properties should be well established 14 questionnaires, patients were asked whether they were will- before its use. Finally, to assess its relevance for clinical ing to discuss costs with their medical team (“I would like practice as well as patient-centered research, we also aimed to talk about my out-of-pocket health care costs when a test to evaluate whether patient-reported financial toxicity was or treatment is recommend”). This question is based on correlated with HRQOL. the seminal work of Alexander et al for the assessment of communication preferences. Concurrent participation in MATERIALS AND METHODS a clinical trial at the time of the interview also was recorded because patients taking part in clinical trials may not be re- Sample sponsible for some of their treatment costs. Patients were eligible if they were aged 18 years with a diagnosis of AJCC stage IV cancer, regardless of prior dis- HRQOL assessments ease. Patients receiving chemotherapy (oral, intravenous, HRQOL was assessed by 2 widely validated instruments: or both) for at least 2 months at the time of the interview the Functional Assessment of Cancer Therapy-General and with an Eastern Cooperative Oncology Group perfor- (FACT-G) and the European Organization for Research mance status (ECOG PS) of 0 to 2 were approached for and Treatment of Cancer’s Quality of Life Questionnaire- participation. Because the goal of the current study was to global health status (EORTC-QOL). We hypothesized determine whether this instrument was valid in measuring that higher financial toxicity would have mild to moderate financial concerns, we chose patients with advanced dis- correlation with worse HRQOL. ease who had been receiving therapy for at least 2 months as a sample most likely to have received health care bills Psychometric analyses and to have experienced financial issues. Patients were Guidelines from the International Society for Pharmacoe- 19,20 recruited at 2 separate cancer centers: The University of conomics and Outcomes Research (ISPOR) and the Chicago Medicine and The NorthShore University COSMIN study (COnsensus-based Standards for the se- HealthSystem. The study protocol was approved by local lection of health Measurement INstruments) for Institutional Review Boards, and all patients provided in- PROM development were followed. The methods used to formed consent. Patients were recruited from May 2013 assess the factor structure and its analysis are available in to February 2015. Potential participants were told that the Supporting Information. the general goal of the study was “to learn about factors that may affect your experience as a cancer patient,” but Reliability and Validity they were not prospectively told the specific objectives of The internal consistency of the COST measure, or the de- the current study. gree to which individual items that comprise the scale Cancer February 1, 2017 477 Original Article Figure 1. . COmprehensive Score for financial Toxicity (COST)-Functional Assessment of Chronic Illness Therapy (FACIT). Items 2, 3, 4, 5, 8, 9, and 10 were reverse scored. The lower the score, the worse the financial toxicity. measure the same latent variable, was assessed by the pothesized that the COST measure would have a mild to Cronbach a. Values> .90 were considered excellent. In moderate, statistically significant correlation with income. addition, for a sample of patients, the COST measure was The rationale for the mild and moderate correlations is readministered within a 7-day interval (test-retest). The that strong correlations would conclude that the COST test-retest reliability was assessed by the intraclass correla- PROM would actually be measuring psychological tion coefficient (ICC) using a 1-way random effects distress, HRQOL, or income, rather than financial toxici- model. ty. Divergent validity was assessed using correlations be- Convergent validity was assessed by calculating the tween the Marlowe-Crowne Social Desirability Scale Pearson correlation between the COST and the Brief- (MCSDS) and the COST measure. The MCSDS is a POMS measures. We hypothesized that the COST mea- widely used measure of self-reported social desirability. sure (financial toxicity) would have a mild to moderate For validation purposes, it measures the degree to which and statistically significant correlation with the Brief- individuals attempt to present themselves in a favorable POMS measure (psychological distress). Similarly, we hy- light. We hypothesized that financial toxicity would not 478 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al have statistically significant correlations with social desir- univariable analysis (P<.10), and those in which we had a ability. All the instruments are described in Supporting substantial interest (clinical trial participation and willing- Information Table 1. ness to discuss costs). Interactions were tested, and if they were found to be statistically significant, they were included Factors Associated With Financial Toxicity in the final model. Missing data were addressed by each Multivariable linear regression analyses were performed to instrument’s guidelines. All analyses were performed with compare the average COST values obtained from groups Stata statistical software (version 13; StataCorp LP, College of patients that were expected to differ with respect to the Station, Tex). construct (“known groups” validity). The dependent vari- able was the COST value. Independent variables consid- RESULTS ered included: patient sociodemographic characteristics, Of the 375 patients who were approached, 236 (62.9%) cancer type, length of original diagnosis (from the time of agreed to participate. Reasons for nonparticipation were the original cancer diagnosis to the interview date), health “not interested in any research” (103 patients) and “not care use, psychological distress, communication preferen- feeling well” (36 patients). In addition, 3 patients started ces, and whether the patient was taking part in a clinical the survey and withdrew consent due to its financial con- trial. Considering the financial domain of this PROM, we notation, leaving a total of 233 evaluable patients (partici- included household income and use of health care resour- pation rate of 62.1%). Respondents were more likely to ces (with potential implications for cost-sharing and loss be younger on average (58.42 years vs 63.21 years; of income) as additional factors to be investigated. We P<.001) and married (73.3% vs 61.3%; P5 .02) com- thus hypothesized that individuals in lower FPL groups pared with nonrespondents. Sex (P5 .08), insurance sta- would have worse financial toxicity, as would patients tus (P5 .07), race (P5 .07), and household income with higher health care use (as measured by emergency de- distribution (P5 .07) were found to demonstrate a trend partment visits or inpatient admissions), independent of toward a difference but did not reach statistical signifi- the other variables. cance at the .05 level. Data regarding the 142 patients (37.9%) who declined participation are shown in Online Statistical Analyses Supporting Information Table 2. Differences between participants and nonparticipants with Participants had a median age of 59 years (range, regard to baseline characteristics were tested using Student t 27-88 years; mean6 standard deviation [SD], 58.426 tests or chi-square tests, as appropriate. The Pearson corre- 11.47 years), and 58% of patients were female. All had lation coefficient was used to assess the relation between the health insurance coverage, which was mostly private or COST measure and HRQOL. In addition, partial correla- purchased by their employer (62%), followed by Medi- tions with HRQOL were calculated, adjusting for age, care with or without supplementation (31%). No patients ECOG PS, and variables that were found to be signifi- had acquired insurance through the ACA Marketplace. cantly associated with financial toxicity on multivariable The median length of time from the first cancer diagnosis analyses. Correlations were defined as mild if between 0.20 was 485 days (range, 56-9294 days). The median house- and 0.39, moderate if between 0.40 and 0.59, strong if be- hold income was 376% of the FPL (range, 0%-7964%). tween 0.60 and 0.79, and very strong if between 0.80 and In addition, 47% of patients had completed college or 1.0. A sample size of 233 patients provided 80% power had achieved higher education, whereas 15% had less with which to detect a correlation coefficient >0.18 be- than a college education, with 5.1% of patients having less tween the COST measure and the HRQOL instruments than a high school education. The median COST value using a 2-sided significance level of .05. For the test-retest was 23 (range, 0-44; mean6 SD, 22.236 11.89). Table analysis, a sample size of 20 patients who were assessed 1 summarizes the characteristics of the patients included twice within a 7-day period provided> 80% power to in the analyses, as well as their COST values. Online Sup- demonstrate excellent reliability (ICC of 0.75, assuming porting Information Table 3 describes their primary tu- an ICC under the null hypothesis of 0.35) with a signifi- mor types. Online Supplementary Information Table 4 cance level of .05. In differential item functioning analyses, shows the factor analysis results. a Bonferroni correction was used to control for multiple comparisons, with a P value threshold set at <.0045. On Reliability and Validity the multivariable analyses, variables included in the final The COST measure demonstrated excellent internal con- model were those approaching statistical significance on sistency, with a Cronbach a of .92. The Cronbach alphas Cancer February 1, 2017 479 Original Article TABLE 1. Patient Characteristics and COST Values Characteristic N5 233 COST (Mean6SD) Univariable P Institution The University of Chicago 199 (85.4%) 22.346 11.79 .74 NorthShore University HealthSystem 34 (14.6%) 21.606 12.65 Median age (range), y 59 (27-88) 50 51 (21.9%) 20.566 11.71 .05 51-64 110 (47.2%) 20.886 11.84 65-75 56 (24.0%) 25.516 10.98 75 16 (6.9%) 25.376 14.01 Sex Female 136 (58.4%) 20.626 11.57 .01 Male 97 (41.6%) 24.506 12.02 Marital status Married 170 (73.0%) 23.276 11.86 .02 Divorced/separated/widowed 38 (16.3%) 20.996 11.86 Never married 24 (10.3%) 16.406 10.67 Race/ethnicity White, non-Hispanic 154 (66.1%) 23.906 12.21 .03 African American 54 (23.2%) 18.926 10.71 Hispanic 14 (6.0%) 18.466 9.73 Other (Asian and Native American) 10 (4.3%) 21.546 11.40 Education level <College 36 (15.5%) 22.206 12.04 .11 Some college or technical training 87 (37.3%) 20.286 11.18 Completed college 55 (23.6%) 22.296 11.92 Graduate or professional degree 55 (23.6%) 25.296 12.51 Insurance type Private or employer-based 144 (61.8%) 21.896 11.75 .04 Medicare (with or without supplementation) 73 (31.3%) 24.476 11.99 Medicaid 13 (5.6%) 15.316 8.14 COBRA continuation coverage 3 (1.3%) 14.206 19.17 Employment status Working (full time or part time) 78 (33.5%) 25.086 11.85 <.001 Unemployed 21 (9.0%) 12.376 10.72 Retired 76 (32.6%) 25.846 10.81 On short-term or long-term disability 45 (19.3%) 14.986 8.82 Others (student or homemaker) 12 (5.1%) 26.696 9.49 ECOG performance status 0 116 (49.8%) 21.616 11.89 .26 1 76 (32.6%) 23.686 11.27 2 5 (2.1%) 16.426 15.57 Median length of cancer diagnosis (range) 485 d (56-9294 d) 1 y 90 (38.6%) 21.266 11.90 .32 > 1 y 143 (61.4%) 22.846 11.89 Median household income (range) of poverty level 376.6% (0%-7964%) 200% of poverty level 27 (11.6%) 15.446 10.03 <.001 200%-400% of FPL 83 (35.6%) 20.986 10.77 400%-600% of FPL 50 (21.5%) 20.956 13.01 600%-800% of FPL 21 (9.0%) 27.576 11.28 >800% of FPL 24 (10.3%) 31.506 10.01 Median no. of inpatient admissions (range) 1 (0-12) 2 156 (66.9%) 23.396 11.46 .009 3 41 (17.6%) 17.996 12.26 Median no. of emergency room visits (range) 0 (0-7) 2 187 (80.3%) 22.216 11.76 .77 3 10 (4.3%) 23.326 13.22 Median Brief-POMS (range) 14 (1-50) <.001 Abbreviations: COBRA, Consolidated Omnibus Budget Reconciliation Act; COST, COmprehensive Score for financial Toxicity; ECOG, Eastern Cooperative On- cology Group; FPL, federal poverty level; POMS, Profile of Mood States; SD, standard deviation. Some percentages do not add to 100% due to missing data. Univariate P values were derived from linear regression models. Lower COST values indicate higher toxicity. The category for less than a college education included 12 patients (5.1%) who completed junior high or middle school and 24 patients (10.3%) who com- pleted high school. 480 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al for males and females were comparable at .92 and .91, re- Factors Associated With Financial Toxicity As shown in Table 1, we found several factors that were as- spectively. The test-retest analysis revealed an ICC of 0.80 sociated with financial toxicity on univariable analyses. (95% confidence interval, 0.57-0.92). The Pearson corre- Communication preferences and clinical trial status were lation for the COST measure and the Brief-POMS was - not found to be significantly associated with financial tox- 0.26 (P<.001), indicating that worse financial toxicity icity (Table 2). However, given their theoretical impor- was correlated with higher psychological distress. Not sur- tance when assessing financial toxicity, they were included prisingly, financial toxicity was found to be correlated as potential confounders in the multivariable analyses. In with household income at 0.28 (P<.001), indicating fi- the final multivariable model, employment status nancial toxicity (lower COST values) among patients in (P<.001), race (P5 .04), household income (P5 .003), the lower FPL groups. With regard to divergent validity, psychological distress (P5 .003), and the number of in- the association between the COST measure and social de- patient admissions (P5 .01) were found to be significant- sirability (MCSDS) was near zero (0.11) and not statisti- ly associated with financial toxicity when controlling for cally significant (P5 .11), as initially hypothesized. age, sex, marital status, insurance type, clinical trial partic- TABLE 2. COST Values, Willingness to Discuss ipation, and communication preferences, as shown in Costs, and Clinical Trial Status Table 3. It is important to note that no statistically signifi- COST cant interactions were found, including the interaction be- N5 233 (Mean6 SD) Univariable P tween employment and income (P5 .32). Nonwhite individuals tended to have lower COST scores (worse fi- Willing to discuss costs with care team Yes 106 (45.5%) 22.546 12.41 .49 nancial toxicity) compared with white individuals. Simi- No or unsure 105 (45.1%) 21.426 11.28 larly, unemployed patients had lower COST scores Clinical trial status compared with other groups, reaching statistical signifi- Yes 65 (27.9%) 23.266 11.92 .41 No 168 (72.1%) 21.846 11.89 cance for all groups except those on disability. Higher psy- chological distress values were associated with lower Abbreviations: COST, COmprehensive Score for financial Toxicity; SD, stan- dard deviation. COST scores. Those patients with 3 inpatient admis- Some percentages do not add to 100% due to missing data. Univariate P sions had lower COST scores (indicating worse financial values were derived from linear regression models. Lower COST values in- dicate higher toxicity. toxicity), by nearly 6 points on average, compared with a b TABLE 3. Factors Associated With COST on Multivariate Analyses Factor Coefficient (95% CI) Adjusted P Race/ethnicity .04 White, non-Hispanic (Base) Hispanic 22.41 (28.66 to 3.83) African American 25.14 (29.60 to 20.67) Other (Asian and Native American) 29.85 (219.82 to 0.12) Household income .003 200% of FPL (Base) >200%-400% of FPL 2.55 (22.85 to 7.95) >400%-600% of FPL 3.61 (22.12 to 9.34) >600%-800% of FPL 9.39 (0.80 to 17.98) >800% of FPL 11.68 (4.93 to 18.44) Employment status <.001 Unemployed (Base) On short-term or long-term disability 2.30 (24.04 to 8.65) Working (full or part time) 9.58 (3.03 to 16.14) Retired 10.69 (3.58 to 17.81) Others (student or homemaker) 12.61 (3.71 to 21.50) No. of inpatient admissions .01 2 (Base) 3 25.52 (29.87 to 21.16) Psychological distress-Brief-POMS (per 1-point increase) 20.34 (20.56 to 20.12) .003 Abbreviations: 95% CI, 95% confidence interval; COST, COmprehensive Score for financial Toxicity; FPL, federal poverty level; POMS, Profile of Mood States. Lower COST values indicate higher toxicity. The multivariate model included those variables with a P<.1 on univariate analysis in addition to communication preferences and clinical trial participation (only the variables that remained significant on the multivariate model are presented in the table). Cancer February 1, 2017 481 Original Article those with fewer admissions. In addition, a higher house- admissions also impact the financial toxicity of individual hold income as a percentage of the FPL was found to be patients. The current study has several strengths. First, we an- associated with higher COST scores (less financial toxici- alyzed data regarding individuals who declined participa- ty), as those at >800% of the FPL had a nearly 12-point tion, with only 3 patients withdrawing consent because of higher mean COST score compared with those at this being a study assessing financial issues. A major 200% of the FPL. strength also was assessing potential confounders when measuring financial toxicity. We controlled for psycho- Financial Toxicity and HRQOL logical distress and for participation in clinical trials, and The median FACT-G value was 79 (range, 23-108; the results were found to be independent of these factors. mean6 SD, 77.016 17.21). The median EORTC- In addition, although communication preferences were QOL value was 66 (range, 0-100; mean6 SD, 61.516 not the focus of the current study, there was the theoreti- 22.41). The Pearson correlation for the COST measure cal concern that a patient’s willingness to discuss financial and the FACT-G was 0.42 (P<.001), whereas its correla- issues could potentially introduce bias into the patient’s tion with the EORTC-QOL was 0.33 (P<.001). When perspective about the topic and, thus, the self-reported we controlled for age, ECOG PS, income, psychological COST assessment. However, the willingness to discuss fi- distress, inpatient admissions, employment status, and nancial issues was not found to be significantly associated ethnicity, these correlations remained statistically signifi- with COST values. Finally, an instrument is of little rele- cant (FACT-G partial correlation of 0.31 [P<.001]; and vance if it is not correlated with clinically meaningful out- EORTC-QOL partial correlation of 0.20 [P<.001]), comes. Herein, we demonstrated that financial toxicity, as thereby confirming our hypothesis that financial toxicity measured by the COST measure, is correlated with was correlated with worse HRQOL. HRQOL as measured by 2 validated quantitative HRQOL instruments (FACT-G and EORTC-QOL), DISCUSSION when adjusting for potentially confounding factors. 19-21 Using a hypothesis-based approach, the results of the A limitation of the current study is that the study current study validate the COST measure as a measure of sample was drawn from tertiary referral health care cen- financial toxicity specifically developed for patients with ters, and all participants had some form of insurance cov- cancer. We also demonstrated its statistically significant erage. However, it is well known that even insured correlation with HRQOL, thus establishing it as a clini- patients can face significant financial burdens, especially if cally relevant patient-centered measure. their insurance plan has high deductibles or if patients are The reported correlations (Brief-POMS, HRQOL, at risk of exceeding lifetime limits. In addition, because and income) were mild but statistically significant, as hy- it was not feasible to extensively and repeatedly interview pothesized. These correlations add to the instrument’s va- patients with a poor ECOG PS, we limited our sample to lidity because it demonstrates that the COST-PROM is those patients with an ECOG PS <3. Also, given our not measuring psychosocial distress, HRQOL, or income, cross-sectional design, we did not assess whether financial as we would potentially observe with strong or very strong toxicity was related to out-of-pocket costs, loss of produc- correlations. The identification of factors associated with tivity, or other factors. These limitations should be con- financial toxicity is equally important. To the best of our sidered from the perspective that the goal of the current knowledge, the current study is the first to report on the study was to develop a financial toxicity instrument and relationship between financial toxicity and the use of not to identify all the populations at risk or the drivers of health care resources. We found that higher numbers of financial toxicity within this validation cohort. We aimed inpatient admissions were associated with higher financial to demonstrate that this instrument measures what it was toxicity, even when adjusted for potentially confounding designed to measure. Therefore, the objective of the cur- factors. The importance of inpatient admissions in total rent study was to empirically demonstrate the association health care costs has been demonstrated previously. Chas- between financial toxicity, as measured by this instru- tek et al demonstrated that 55% of total health care costs ment, with well-known social factors that in theory would within the last 6 months of life are related to inpatient play a role in it, demonstrating the validity of the instru- admissions. However, to our knowledge, the study by ment. In this regard, race, income, inpatient admissions, Chastek et al did not address patient’s financial toxicity. and employment status were found to be associated with In the current study, we demonstrated that inpatient financial toxicity in the current study population, as 482 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al equity as a cofounder of PrescriptIQ; personal fees from multiple measured by the COST instrument, independent of psy- drug companies for acting as an expert witness in patent litigation; chological distress, the willingness to discuss costs, and personal fees from AbbVie; personal fees and stock options from other factors. Underserved groups within the current Biscayne Pharmaceuticals; grants from Dicerna Pharmaceuticals; study sample, such as those with low income, as well as Af- and personal fees from Cantex Pharmaceuticals, Genentech, Kinex rican American individuals, were found to have worse Pharmaceuticals, Onconova Therapeutics, Shionogi, Xspray, Agios Pharmaceuticals, Cyclacel, Drais Pharmaceuticals, Circle Pharma, COST values; it is likely that the findings of the current Portola Pharmaceuticals, and Venaxis for work performed outside study would be replicated in uninsured or underinsured of the current study. In addition, Dr. Ratain has a patent related to samples. This hypothesis should be validated and con- UGT1A1 genotyping and irinotecan use with royalties paid to the firmed in future studies with a more diverse insurance case Mayo Clinic and a patent related to a genomic prescribing system mix and specifically designed to identify populations at licensed to PrescriptIQ. risk, such as those with less than a high school education, which was also a population relatively underrepresented AUTHOR CONTRIBUTIONS in the current study sample (5.1%). Similarly, the differ- Jonas A. de Souza participated in planning, conducting, interpret- ing and reporting and is the guarantor of the study. Bonnie J. Yap, ences with regard to the drivers of financial toxicity among Kristen Wroblewski, Christopher K. Daugherty, and David patients with different cancer types should be examined Cella participated in planning, conducting, and reporting the because these drivers may vary by the type of disease or by study. Victoria Blinder, Fabiana S. Ara ujo, Fay J. Hlubocky, the type of treatment. Although toxicity thresholds were Lauren H. Nicholas, Jeremy M. O’Connor, Bruce Brockstein, not analyzed within this validation sample, the quantita- and Mark J. Ratain participated in interpretation and reporting. tive nature of the COST instrument will allow further prospective studies to determine the exact impact of finan- REFERENCES cial toxicity on HRQOL in comparison with other symp- 1. Bestvina CM, Zulling LL, Rushing C, et al. Patient-oncologist cost communication, financial distress, and medication adherence. J Oncol toms and on other outcomes, such as survival. These Pract. 2014;10:162-167. prospective studies also are needed to determine respon- 2. Northouse L, Williams Al, Given B, McCorkle R. Psychosocial care for family caregivers of patients with cancer. J Clin Oncol. 2012;30: siveness to change, as well as floor and ceiling effects of the 1227-1234. COST-PROM. Finally, because the objective of the cur- 3. Pollack, A. For profit, industry seeks cancer drugs. The New York rent study was to validate whether the COST-PROM cap- Times. September 1st, 2009. http://www.nytimes.com/2009/09/02/ health/research/02cancerdrug.html. Accessed September 20, 2016. tured financial toxicity, we needed patients who had 4. de Souza JA, Yap BJ, Hlubocky FJ, et al. The development of a fi- actually had experience with health care bills. Therefore, nancial toxicity patient-reported outcome in cancer: the COST mea- sure. Cancer. 2014;120:3245-3253. we assessed patients with stage IV disease who were receiv- 5. Zafar SY, McNeil RB, Thomas CM, Lathan CS, Ayanian JZ, ing treatment for at least 2 months with chemotherapy, Provenzale D. Population-based assessment of cancer survivors’ fi- and we were unable to conclude that newly diagnosed nancial burden and quality of life: a prospective cohort study. J Oncol Pract. 2015;11:145-150. patients would provide similar results. 6. Lathan CS, Cronin A, Tucker-Seeley R, Zafar SY, Ayanian JZ, As society increasingly considers the costs incurred Schrag D. Association of financial strain with symptom burden and by patients with cancer as a side effect of treatment, instru- quality of life for patients with lung or colorectal cancer. J Clin Oncol. 2016;34:1732-1740. ments to measure financial toxicity should be patient- 7. Neugut AI, Subar M, Wilde ET, et al. Association between prescrip- centered, scientifically derived, and clinically relevant. tion co-payment amount and compliance with adjuvant hormonal therapy in women with early-stage breast cancer. J Clin Oncol. 2011; The correlation with HRQOL establishes financial toxici- 29:2534-2542. ty as a clinically relevant, patient-centered measure. The 8. Ramsey SD, Bansal A, Fedorenko CR, et al. Financial insolvency as incorporation of financial toxicity assessments into obser- a risk factor for early mortality among patients with cancer. J Clin Oncol. 2016;34:980-986. vational research and clinical trials will ensure a patient- 9. The Henry J. Kaiser Family Foundation. What the actuarial values centered foundation in the evaluation of financial distress. in the Affordable Care Act mean. http://www.kff.org/healthreform/ upload/8177.pdf. Accessed September 20, 2016. 10. Ungar L, O’Donnel J. Dilemma over deductibles: costs crippling mid- FUNDING SUPPORT dle class. http://www.usatoday.com/story/news/nation/2015/01/01/ middle-class-workers-struggle-to-pay-for-care-despite-insurance/19841 Supported by grant University of Chicago Institute for Translation- 235/. Accessed September 20, 2016. al Medicine-Clinical Translational Science Award UL1 TR000430. 11. High-deductible health plans, gamble for some, on the rise [tran- script]. All Things Considered. National Public Radio. July 28, 2013. http://www.npr.org/2013/07/28/206370593/high-deductible-health- CONFLICT OF INTEREST DISCLOSURES plans-gamble-for-some-on-the-rise. Accessed September 20, 2016. Kristen Wroblewski received a grant from the National Institutes of 12. Zafar SY, Peppercorn JM, Schrag D, et al. The financial toxicity of Health/National Cancer Institute (P30 CA14599) for work per- cancer treatment: a pilot study assessing out-of-pocket expenses and formed as part of the current study. Mark J. Ratain has received the insured cancer patient’s experience. Oncologist. 2013;18:381-390. Cancer February 1, 2017 483 Original Article 13. Huntington SF, Weiss BM, Vogl DT, et al. Financial toxicity in in- 20. Patrick DL, Burke LB, Gwaltney CJ, et al. Content validity–estab- sured patients with multiple myeloma: a cross-sectional pilot study. lishing and reporting the evidence in newly developed patient- Lancet Haematol. 2015;2:e408-e416. reported outcomes (PRO) instruments for medical product evalua- 14. U.S. Department of Health and Human Services FDA Center for tion: ISPOR PRO Good Research Practices Task Force report: part Drug Evaluation and Research; U.S. Department of Health and Hu- 2-assessing respondent understanding. Value Health. 2011;14:978- man Services FDA Center for Biologics Evaluation and Research; U.S. Department of Health and Human Services FDA Center for 21. Mokkink LB, Terwee CB, Patrick DL, et al. The COSMIN check- Devices and Radiological Health. Guidance for industry: patient- list for assessing the methodological quality of studies on measure- reported outcome measures: use in medical product development to ment properties of health status measurement instruments: an support labeling claims: draft guidance. Health Qual Life Outcomes. international Delphi study. Qual Life Res. 2010;19:539-549. 2006;4:79. 22. Crowne DP, Marlowe D. A new scale of social desirability indepen- 15. Cella DF, Jacobsen PB, Orav EJ, Holland JC, Silberfarb PM, Rafla dent of psychopathology. J Consult Psychol. 1960;24:349-354. S. A brief POMS measure of distress for cancer patients. J Chronic 23. Park BW, Lee S, Lee AR, Lee KH, Hwang SY. Quality of life differ- Dis. 1987;40:939-942. ences between younger and older breast cancer patients. J Breast 16. Alexander GC, Casalino LP, Meltzer DO. Patient-physician commu- Cancer. 2011;14:112-118. nication about out-of-pocket costs. JAMA. 2003;290:953-958. 17. Cella DF, Tulsky DS, Gray G, et al. The Functional Assessment of 24. Faller H, Brahler E, Harter M, et al. Performance status and depres- Cancer Therapy scale: development and validation of the general sive symptoms as predictors of quality of life in cancer patients. A measure. J Clin Oncol. 1993;11:570-579. structural equation modeling analysis. Psychooncology. 2015;24: 18. Aaronson NK, Ahmedzai S, Bergman B, et al. The European Orga- 1456-1462. nization for Research and Treatment of Cancer QLQ-C30: a 25. Streiner DL, Norman GR. Health Measurement Scales: A Practical quality-of-life instrument for use in international clinical trials in on- Guide to Their Development and Use. 2nd ed. New York: Oxford cology. J Natl Cancer Inst. 1993;85:365-376. University Press; 1995. 19. Patrick DL, Burke LB, Gwaltney CJ, et al. Content validity–estab- 26. StataCorp LP. Stata Statistical Software: Release 13. College Station, lishing and reporting the evidence in newly developed patient- TX: StataCorp LP; 2013. reported outcomes (PRO) instruments for medical product evalua- 27. Chastek B, Harley C, Kallich J, Newcomer L, Paoli CJ, Teitelbaum tion: ISPOR PRO Good Research Practices Task Force report: part AH. Health care costs for patients with cancer at the end of life. 1–eliciting concepts for a new PRO instrument. Value Health. 2011; 14:967-977. J Oncol Pract. 2012;8:75s-80s. 484 Cancer February 1, 2017 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Cancer Pubmed Central

Measuring financial toxicity as a clinically relevant patient‐reported outcome: The validation of the COmprehensive Score for financial Toxicity (COST)

Loading next page...
 
/lp/pubmed-central/measuring-financial-toxicity-as-a-clinically-relevant-patient-reported-WMsmgY70P6

References (31)

Publisher
Pubmed Central
Copyright
© 2016 The Authors. Cancer published by Wiley Periodicals, Inc. on behalf of American Cancer Society.
ISSN
0008-543X
eISSN
1097-0142
DOI
10.1002/cncr.30369
Publisher site
See Article on Publisher Site

Abstract

Original Article Measuring Financial Toxicity as a Clinically Relevant Patient-Reported Outcome: The Validation of the COmprehensive Score for financial Toxicity (COST) 1 1 2 3 Jonas A. de Souza, MD, MBA ; Bonnie J. Yap, MS ; Kristen Wroblewski, MS ; Victoria Blinder, MD, MSc ; 4 1 5 1 Fabiana S. Araujo,  PhD ; Fay J. Hlubocky, PhD ; Lauren H. Nicholas, PhD ; Jeremy M. O’Connor, MD ; 6 1 1 7 Bruce Brockstein, MD ; Mark J. Ratain, MD ; Christopher K. Daugherty, MD ; and David Cella, PhD BACKGROUND: Cancer and its treatment lead to increased financial distress for patients. To the authors’ knowledge, to date, no stan- dardized patient-reported outcome measure has been validated to assess this distress. METHODS: Patients with AJCC Stage IV solid tumors receiving chemotherapy for at least 2 months were recruited. Financial toxicity was measured by the COmprehensive Score for financial Toxicity (COST) measure. The authors collected data regarding patient characteristics, clinical trial participation, health care use, willingness to discuss costs, psychological distress (Brief Profile of Mood States [POMS]), and health-related quality of life (HRQOL) as measured by the Functional Assessment of Cancer Therapy: General (FACT-G) and the European Organization for Re- search and Treatment of Cancer (EORTC) QOL questionnaires. Test-retest reliability, internal consistency, and validity of the COST measure were assessed using standard-scale construction techniques. Associations between the resulting factors and other variables were assessed using multivariable analyses. RESULTS: A total of 375 patients with advanced cancer were approached, 233 of whom (62.1%) agreed to participate. The COST measure demonstrated high internal consistency and test-retest reliability. Factor analyses revealed a coherent, single, latent variable (financial toxicity). COST values were found to be correlated with income (correlation coef- ficient [r]5 0.28; P<.001), psychosocial distress (r5 -0.26; P<.001), and HRQOL, as measured by the FACT-G (r5 0.42; P<.001) and by the EORTC QOL instruments (r5 0.33; P<.001). Independent factors found to be associated with financial toxicity were race (P5.04), employment status (P<.001), income (P5.003), number of inpatient admissions (P5.01), and psychological distress (P5.003). Will- ingness to discuss costs was not found to be associated with the degree of financial distress (P5.49). CONCLUSIONS: The COST measure demonstrated reliability and validity in measuring financial toxicity. Its correlation with HRQOL indicates that financial toxici- ty is a clinically relevant patient-centered outcome. Cancer 2017;123:476-84. V 2016 The Authors. Cancer published by Wiley Periodi- cals, Inc. on behalf of American Cancer Society. This is an open access article under the terms of the Creative Commons Attribution- NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. KEYWORDS: cost of cancer, financial burden, financial toxicity, patient-reported outcome (PRO). INTRODUCTION Patients with cancer often are confronted with the financial consequences of cancer treatment, which may include signifi- 1,2 cant out-of-pocket costs, loss of income, and caregiver burden. The objective financial consequences of cancer, as well 3,4 as the subjective financial concerns, have been broadly termed “financial toxicity.” This burden has since been linked 5 6 with several clinically relevant patient outcomes, including health-related quality of life (HRQOL) ; symptom burden ; 7 8 compliance ; and, most recently, survival. Within the context of current policies and the Patient Protection and Affordable Care Act (ACA), many newly in- sured patients are expected to have a higher cost share because of preferential enrollment into high-deductible plans offer- 9-11 ing greater upfront affordability. Combined with escalating cancer costs, these high-deductible plans may increase financial distress and further exacerbate disparities in cancer care. To abrogate the impact of financial distress and Corresponding author: Jonas A. de Souza, MD, MBA, Department of Medicine, The University of Chicago Medicine, 5841 S. Maryland Ave, Chicago, IL 60637; Fax: (773) 702-3163; jdesouza@chicagobooth.edu 1 2 Department of Medicine, The University of Chicago Medicine, Chicago, Illinois; Department of Public Health Sciences, The University of Chicago Medicine, Chica- 3 4 go, Illinois; Center for Health Policy and Outcomes, Memorial Sloan Kettering Cancer Center, New York, New York; Institute of Psychology, Illinois Institute of 5 6 Technology, Chicago, IL; Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland; Department of Medicine, NorthShore University HealthSystem, Evanston, Illinois; Department of Medical Social Sciences, Northwestern University, Chicago, Illinois We thank Jocelyn Herrera, Kristen Kipping-Johnson, and Nancy Haefling for their support. Additional supporting information may be found in the online version of this article DOI: 10.1002/cncr.30369, Received: June 22, 2016; Revised: August 9, 2016; Accepted: September 8, 2016, Published online October 7, 2016 in Wiley Online Library (wileyonlinelibrary.com) 476 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al minimize its potential to augment disparities, there is an Measures Sociodemographic characteristics, health care use, urgent need for policy makers, researchers, and clinicians and willingness to discuss costs to accurately measure financial toxicity. However, to the Data related to age, sex, education, work status, marital sta- best of our knowledge, this patient-centered toxicity is tus, race, ethnicity, ECOG PS, and cancer type were col- rarely assessed in clinical practice or research. lected. We asked patients for their income from all sources The COmprehensive Score for financial Toxicity within the previous year. Household income was measured (COST) patient-reported outcome measure (PROM) was as a function of the federal poverty level (FPL).The FPL is previously developed by de Souza et al to assess financial used for Health Insurance Marketplace cost assistance un- toxicity in patients with cancer. In a multistep process, der the ACA, in which tax credit eligibility for health insur- 155 patients with advanced cancer were interviewed to de- ance ranges from 138% to 400% of the FPL in states that velop the 11-item COST measure. In response to the ur- decided to expand Medicaid. Financial toxicity was mea- gent need for such a tool, the instrument was adopted into sured by the COST measure (Fig. 1). Lower COST values clinical practice and research before the establishment of indicate more financial toxicity. General mood disturbance its psychometric properties. The current study examines or psychological distress was measured by the Brief Profile the COST measure with respect to its psychometric prop- of Mood States (Brief-POMS). Emergency department erties, or how well it measures the construct of interest. visits and inpatient admissions from 1 year before the inter- The importance of knowing these properties was empha- view date were collected as measures of health resource use. sized by the US Food and Drug Administration guidance To control for different communication preferences related on PROMs, in which it was recommended that an instru- to costs and thus any resulting bias when responding to the ment’s measurement properties should be well established 14 questionnaires, patients were asked whether they were will- before its use. Finally, to assess its relevance for clinical ing to discuss costs with their medical team (“I would like practice as well as patient-centered research, we also aimed to talk about my out-of-pocket health care costs when a test to evaluate whether patient-reported financial toxicity was or treatment is recommend”). This question is based on correlated with HRQOL. the seminal work of Alexander et al for the assessment of communication preferences. Concurrent participation in MATERIALS AND METHODS a clinical trial at the time of the interview also was recorded because patients taking part in clinical trials may not be re- Sample sponsible for some of their treatment costs. Patients were eligible if they were aged 18 years with a diagnosis of AJCC stage IV cancer, regardless of prior dis- HRQOL assessments ease. Patients receiving chemotherapy (oral, intravenous, HRQOL was assessed by 2 widely validated instruments: or both) for at least 2 months at the time of the interview the Functional Assessment of Cancer Therapy-General and with an Eastern Cooperative Oncology Group perfor- (FACT-G) and the European Organization for Research mance status (ECOG PS) of 0 to 2 were approached for and Treatment of Cancer’s Quality of Life Questionnaire- participation. Because the goal of the current study was to global health status (EORTC-QOL). We hypothesized determine whether this instrument was valid in measuring that higher financial toxicity would have mild to moderate financial concerns, we chose patients with advanced dis- correlation with worse HRQOL. ease who had been receiving therapy for at least 2 months as a sample most likely to have received health care bills Psychometric analyses and to have experienced financial issues. Patients were Guidelines from the International Society for Pharmacoe- 19,20 recruited at 2 separate cancer centers: The University of conomics and Outcomes Research (ISPOR) and the Chicago Medicine and The NorthShore University COSMIN study (COnsensus-based Standards for the se- HealthSystem. The study protocol was approved by local lection of health Measurement INstruments) for Institutional Review Boards, and all patients provided in- PROM development were followed. The methods used to formed consent. Patients were recruited from May 2013 assess the factor structure and its analysis are available in to February 2015. Potential participants were told that the Supporting Information. the general goal of the study was “to learn about factors that may affect your experience as a cancer patient,” but Reliability and Validity they were not prospectively told the specific objectives of The internal consistency of the COST measure, or the de- the current study. gree to which individual items that comprise the scale Cancer February 1, 2017 477 Original Article Figure 1. . COmprehensive Score for financial Toxicity (COST)-Functional Assessment of Chronic Illness Therapy (FACIT). Items 2, 3, 4, 5, 8, 9, and 10 were reverse scored. The lower the score, the worse the financial toxicity. measure the same latent variable, was assessed by the pothesized that the COST measure would have a mild to Cronbach a. Values> .90 were considered excellent. In moderate, statistically significant correlation with income. addition, for a sample of patients, the COST measure was The rationale for the mild and moderate correlations is readministered within a 7-day interval (test-retest). The that strong correlations would conclude that the COST test-retest reliability was assessed by the intraclass correla- PROM would actually be measuring psychological tion coefficient (ICC) using a 1-way random effects distress, HRQOL, or income, rather than financial toxici- model. ty. Divergent validity was assessed using correlations be- Convergent validity was assessed by calculating the tween the Marlowe-Crowne Social Desirability Scale Pearson correlation between the COST and the Brief- (MCSDS) and the COST measure. The MCSDS is a POMS measures. We hypothesized that the COST mea- widely used measure of self-reported social desirability. sure (financial toxicity) would have a mild to moderate For validation purposes, it measures the degree to which and statistically significant correlation with the Brief- individuals attempt to present themselves in a favorable POMS measure (psychological distress). Similarly, we hy- light. We hypothesized that financial toxicity would not 478 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al have statistically significant correlations with social desir- univariable analysis (P<.10), and those in which we had a ability. All the instruments are described in Supporting substantial interest (clinical trial participation and willing- Information Table 1. ness to discuss costs). Interactions were tested, and if they were found to be statistically significant, they were included Factors Associated With Financial Toxicity in the final model. Missing data were addressed by each Multivariable linear regression analyses were performed to instrument’s guidelines. All analyses were performed with compare the average COST values obtained from groups Stata statistical software (version 13; StataCorp LP, College of patients that were expected to differ with respect to the Station, Tex). construct (“known groups” validity). The dependent vari- able was the COST value. Independent variables consid- RESULTS ered included: patient sociodemographic characteristics, Of the 375 patients who were approached, 236 (62.9%) cancer type, length of original diagnosis (from the time of agreed to participate. Reasons for nonparticipation were the original cancer diagnosis to the interview date), health “not interested in any research” (103 patients) and “not care use, psychological distress, communication preferen- feeling well” (36 patients). In addition, 3 patients started ces, and whether the patient was taking part in a clinical the survey and withdrew consent due to its financial con- trial. Considering the financial domain of this PROM, we notation, leaving a total of 233 evaluable patients (partici- included household income and use of health care resour- pation rate of 62.1%). Respondents were more likely to ces (with potential implications for cost-sharing and loss be younger on average (58.42 years vs 63.21 years; of income) as additional factors to be investigated. We P<.001) and married (73.3% vs 61.3%; P5 .02) com- thus hypothesized that individuals in lower FPL groups pared with nonrespondents. Sex (P5 .08), insurance sta- would have worse financial toxicity, as would patients tus (P5 .07), race (P5 .07), and household income with higher health care use (as measured by emergency de- distribution (P5 .07) were found to demonstrate a trend partment visits or inpatient admissions), independent of toward a difference but did not reach statistical signifi- the other variables. cance at the .05 level. Data regarding the 142 patients (37.9%) who declined participation are shown in Online Statistical Analyses Supporting Information Table 2. Differences between participants and nonparticipants with Participants had a median age of 59 years (range, regard to baseline characteristics were tested using Student t 27-88 years; mean6 standard deviation [SD], 58.426 tests or chi-square tests, as appropriate. The Pearson corre- 11.47 years), and 58% of patients were female. All had lation coefficient was used to assess the relation between the health insurance coverage, which was mostly private or COST measure and HRQOL. In addition, partial correla- purchased by their employer (62%), followed by Medi- tions with HRQOL were calculated, adjusting for age, care with or without supplementation (31%). No patients ECOG PS, and variables that were found to be signifi- had acquired insurance through the ACA Marketplace. cantly associated with financial toxicity on multivariable The median length of time from the first cancer diagnosis analyses. Correlations were defined as mild if between 0.20 was 485 days (range, 56-9294 days). The median house- and 0.39, moderate if between 0.40 and 0.59, strong if be- hold income was 376% of the FPL (range, 0%-7964%). tween 0.60 and 0.79, and very strong if between 0.80 and In addition, 47% of patients had completed college or 1.0. A sample size of 233 patients provided 80% power had achieved higher education, whereas 15% had less with which to detect a correlation coefficient >0.18 be- than a college education, with 5.1% of patients having less tween the COST measure and the HRQOL instruments than a high school education. The median COST value using a 2-sided significance level of .05. For the test-retest was 23 (range, 0-44; mean6 SD, 22.236 11.89). Table analysis, a sample size of 20 patients who were assessed 1 summarizes the characteristics of the patients included twice within a 7-day period provided> 80% power to in the analyses, as well as their COST values. Online Sup- demonstrate excellent reliability (ICC of 0.75, assuming porting Information Table 3 describes their primary tu- an ICC under the null hypothesis of 0.35) with a signifi- mor types. Online Supplementary Information Table 4 cance level of .05. In differential item functioning analyses, shows the factor analysis results. a Bonferroni correction was used to control for multiple comparisons, with a P value threshold set at <.0045. On Reliability and Validity the multivariable analyses, variables included in the final The COST measure demonstrated excellent internal con- model were those approaching statistical significance on sistency, with a Cronbach a of .92. The Cronbach alphas Cancer February 1, 2017 479 Original Article TABLE 1. Patient Characteristics and COST Values Characteristic N5 233 COST (Mean6SD) Univariable P Institution The University of Chicago 199 (85.4%) 22.346 11.79 .74 NorthShore University HealthSystem 34 (14.6%) 21.606 12.65 Median age (range), y 59 (27-88) 50 51 (21.9%) 20.566 11.71 .05 51-64 110 (47.2%) 20.886 11.84 65-75 56 (24.0%) 25.516 10.98 75 16 (6.9%) 25.376 14.01 Sex Female 136 (58.4%) 20.626 11.57 .01 Male 97 (41.6%) 24.506 12.02 Marital status Married 170 (73.0%) 23.276 11.86 .02 Divorced/separated/widowed 38 (16.3%) 20.996 11.86 Never married 24 (10.3%) 16.406 10.67 Race/ethnicity White, non-Hispanic 154 (66.1%) 23.906 12.21 .03 African American 54 (23.2%) 18.926 10.71 Hispanic 14 (6.0%) 18.466 9.73 Other (Asian and Native American) 10 (4.3%) 21.546 11.40 Education level <College 36 (15.5%) 22.206 12.04 .11 Some college or technical training 87 (37.3%) 20.286 11.18 Completed college 55 (23.6%) 22.296 11.92 Graduate or professional degree 55 (23.6%) 25.296 12.51 Insurance type Private or employer-based 144 (61.8%) 21.896 11.75 .04 Medicare (with or without supplementation) 73 (31.3%) 24.476 11.99 Medicaid 13 (5.6%) 15.316 8.14 COBRA continuation coverage 3 (1.3%) 14.206 19.17 Employment status Working (full time or part time) 78 (33.5%) 25.086 11.85 <.001 Unemployed 21 (9.0%) 12.376 10.72 Retired 76 (32.6%) 25.846 10.81 On short-term or long-term disability 45 (19.3%) 14.986 8.82 Others (student or homemaker) 12 (5.1%) 26.696 9.49 ECOG performance status 0 116 (49.8%) 21.616 11.89 .26 1 76 (32.6%) 23.686 11.27 2 5 (2.1%) 16.426 15.57 Median length of cancer diagnosis (range) 485 d (56-9294 d) 1 y 90 (38.6%) 21.266 11.90 .32 > 1 y 143 (61.4%) 22.846 11.89 Median household income (range) of poverty level 376.6% (0%-7964%) 200% of poverty level 27 (11.6%) 15.446 10.03 <.001 200%-400% of FPL 83 (35.6%) 20.986 10.77 400%-600% of FPL 50 (21.5%) 20.956 13.01 600%-800% of FPL 21 (9.0%) 27.576 11.28 >800% of FPL 24 (10.3%) 31.506 10.01 Median no. of inpatient admissions (range) 1 (0-12) 2 156 (66.9%) 23.396 11.46 .009 3 41 (17.6%) 17.996 12.26 Median no. of emergency room visits (range) 0 (0-7) 2 187 (80.3%) 22.216 11.76 .77 3 10 (4.3%) 23.326 13.22 Median Brief-POMS (range) 14 (1-50) <.001 Abbreviations: COBRA, Consolidated Omnibus Budget Reconciliation Act; COST, COmprehensive Score for financial Toxicity; ECOG, Eastern Cooperative On- cology Group; FPL, federal poverty level; POMS, Profile of Mood States; SD, standard deviation. Some percentages do not add to 100% due to missing data. Univariate P values were derived from linear regression models. Lower COST values indicate higher toxicity. The category for less than a college education included 12 patients (5.1%) who completed junior high or middle school and 24 patients (10.3%) who com- pleted high school. 480 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al for males and females were comparable at .92 and .91, re- Factors Associated With Financial Toxicity As shown in Table 1, we found several factors that were as- spectively. The test-retest analysis revealed an ICC of 0.80 sociated with financial toxicity on univariable analyses. (95% confidence interval, 0.57-0.92). The Pearson corre- Communication preferences and clinical trial status were lation for the COST measure and the Brief-POMS was - not found to be significantly associated with financial tox- 0.26 (P<.001), indicating that worse financial toxicity icity (Table 2). However, given their theoretical impor- was correlated with higher psychological distress. Not sur- tance when assessing financial toxicity, they were included prisingly, financial toxicity was found to be correlated as potential confounders in the multivariable analyses. In with household income at 0.28 (P<.001), indicating fi- the final multivariable model, employment status nancial toxicity (lower COST values) among patients in (P<.001), race (P5 .04), household income (P5 .003), the lower FPL groups. With regard to divergent validity, psychological distress (P5 .003), and the number of in- the association between the COST measure and social de- patient admissions (P5 .01) were found to be significant- sirability (MCSDS) was near zero (0.11) and not statisti- ly associated with financial toxicity when controlling for cally significant (P5 .11), as initially hypothesized. age, sex, marital status, insurance type, clinical trial partic- TABLE 2. COST Values, Willingness to Discuss ipation, and communication preferences, as shown in Costs, and Clinical Trial Status Table 3. It is important to note that no statistically signifi- COST cant interactions were found, including the interaction be- N5 233 (Mean6 SD) Univariable P tween employment and income (P5 .32). Nonwhite individuals tended to have lower COST scores (worse fi- Willing to discuss costs with care team Yes 106 (45.5%) 22.546 12.41 .49 nancial toxicity) compared with white individuals. Simi- No or unsure 105 (45.1%) 21.426 11.28 larly, unemployed patients had lower COST scores Clinical trial status compared with other groups, reaching statistical signifi- Yes 65 (27.9%) 23.266 11.92 .41 No 168 (72.1%) 21.846 11.89 cance for all groups except those on disability. Higher psy- chological distress values were associated with lower Abbreviations: COST, COmprehensive Score for financial Toxicity; SD, stan- dard deviation. COST scores. Those patients with 3 inpatient admis- Some percentages do not add to 100% due to missing data. Univariate P sions had lower COST scores (indicating worse financial values were derived from linear regression models. Lower COST values in- dicate higher toxicity. toxicity), by nearly 6 points on average, compared with a b TABLE 3. Factors Associated With COST on Multivariate Analyses Factor Coefficient (95% CI) Adjusted P Race/ethnicity .04 White, non-Hispanic (Base) Hispanic 22.41 (28.66 to 3.83) African American 25.14 (29.60 to 20.67) Other (Asian and Native American) 29.85 (219.82 to 0.12) Household income .003 200% of FPL (Base) >200%-400% of FPL 2.55 (22.85 to 7.95) >400%-600% of FPL 3.61 (22.12 to 9.34) >600%-800% of FPL 9.39 (0.80 to 17.98) >800% of FPL 11.68 (4.93 to 18.44) Employment status <.001 Unemployed (Base) On short-term or long-term disability 2.30 (24.04 to 8.65) Working (full or part time) 9.58 (3.03 to 16.14) Retired 10.69 (3.58 to 17.81) Others (student or homemaker) 12.61 (3.71 to 21.50) No. of inpatient admissions .01 2 (Base) 3 25.52 (29.87 to 21.16) Psychological distress-Brief-POMS (per 1-point increase) 20.34 (20.56 to 20.12) .003 Abbreviations: 95% CI, 95% confidence interval; COST, COmprehensive Score for financial Toxicity; FPL, federal poverty level; POMS, Profile of Mood States. Lower COST values indicate higher toxicity. The multivariate model included those variables with a P<.1 on univariate analysis in addition to communication preferences and clinical trial participation (only the variables that remained significant on the multivariate model are presented in the table). Cancer February 1, 2017 481 Original Article those with fewer admissions. In addition, a higher house- admissions also impact the financial toxicity of individual hold income as a percentage of the FPL was found to be patients. The current study has several strengths. First, we an- associated with higher COST scores (less financial toxici- alyzed data regarding individuals who declined participa- ty), as those at >800% of the FPL had a nearly 12-point tion, with only 3 patients withdrawing consent because of higher mean COST score compared with those at this being a study assessing financial issues. A major 200% of the FPL. strength also was assessing potential confounders when measuring financial toxicity. We controlled for psycho- Financial Toxicity and HRQOL logical distress and for participation in clinical trials, and The median FACT-G value was 79 (range, 23-108; the results were found to be independent of these factors. mean6 SD, 77.016 17.21). The median EORTC- In addition, although communication preferences were QOL value was 66 (range, 0-100; mean6 SD, 61.516 not the focus of the current study, there was the theoreti- 22.41). The Pearson correlation for the COST measure cal concern that a patient’s willingness to discuss financial and the FACT-G was 0.42 (P<.001), whereas its correla- issues could potentially introduce bias into the patient’s tion with the EORTC-QOL was 0.33 (P<.001). When perspective about the topic and, thus, the self-reported we controlled for age, ECOG PS, income, psychological COST assessment. However, the willingness to discuss fi- distress, inpatient admissions, employment status, and nancial issues was not found to be significantly associated ethnicity, these correlations remained statistically signifi- with COST values. Finally, an instrument is of little rele- cant (FACT-G partial correlation of 0.31 [P<.001]; and vance if it is not correlated with clinically meaningful out- EORTC-QOL partial correlation of 0.20 [P<.001]), comes. Herein, we demonstrated that financial toxicity, as thereby confirming our hypothesis that financial toxicity measured by the COST measure, is correlated with was correlated with worse HRQOL. HRQOL as measured by 2 validated quantitative HRQOL instruments (FACT-G and EORTC-QOL), DISCUSSION when adjusting for potentially confounding factors. 19-21 Using a hypothesis-based approach, the results of the A limitation of the current study is that the study current study validate the COST measure as a measure of sample was drawn from tertiary referral health care cen- financial toxicity specifically developed for patients with ters, and all participants had some form of insurance cov- cancer. We also demonstrated its statistically significant erage. However, it is well known that even insured correlation with HRQOL, thus establishing it as a clini- patients can face significant financial burdens, especially if cally relevant patient-centered measure. their insurance plan has high deductibles or if patients are The reported correlations (Brief-POMS, HRQOL, at risk of exceeding lifetime limits. In addition, because and income) were mild but statistically significant, as hy- it was not feasible to extensively and repeatedly interview pothesized. These correlations add to the instrument’s va- patients with a poor ECOG PS, we limited our sample to lidity because it demonstrates that the COST-PROM is those patients with an ECOG PS <3. Also, given our not measuring psychosocial distress, HRQOL, or income, cross-sectional design, we did not assess whether financial as we would potentially observe with strong or very strong toxicity was related to out-of-pocket costs, loss of produc- correlations. The identification of factors associated with tivity, or other factors. These limitations should be con- financial toxicity is equally important. To the best of our sidered from the perspective that the goal of the current knowledge, the current study is the first to report on the study was to develop a financial toxicity instrument and relationship between financial toxicity and the use of not to identify all the populations at risk or the drivers of health care resources. We found that higher numbers of financial toxicity within this validation cohort. We aimed inpatient admissions were associated with higher financial to demonstrate that this instrument measures what it was toxicity, even when adjusted for potentially confounding designed to measure. Therefore, the objective of the cur- factors. The importance of inpatient admissions in total rent study was to empirically demonstrate the association health care costs has been demonstrated previously. Chas- between financial toxicity, as measured by this instru- tek et al demonstrated that 55% of total health care costs ment, with well-known social factors that in theory would within the last 6 months of life are related to inpatient play a role in it, demonstrating the validity of the instru- admissions. However, to our knowledge, the study by ment. In this regard, race, income, inpatient admissions, Chastek et al did not address patient’s financial toxicity. and employment status were found to be associated with In the current study, we demonstrated that inpatient financial toxicity in the current study population, as 482 Cancer February 1, 2017 Measuring Financial Toxicity/de Souza et al equity as a cofounder of PrescriptIQ; personal fees from multiple measured by the COST instrument, independent of psy- drug companies for acting as an expert witness in patent litigation; chological distress, the willingness to discuss costs, and personal fees from AbbVie; personal fees and stock options from other factors. Underserved groups within the current Biscayne Pharmaceuticals; grants from Dicerna Pharmaceuticals; study sample, such as those with low income, as well as Af- and personal fees from Cantex Pharmaceuticals, Genentech, Kinex rican American individuals, were found to have worse Pharmaceuticals, Onconova Therapeutics, Shionogi, Xspray, Agios Pharmaceuticals, Cyclacel, Drais Pharmaceuticals, Circle Pharma, COST values; it is likely that the findings of the current Portola Pharmaceuticals, and Venaxis for work performed outside study would be replicated in uninsured or underinsured of the current study. In addition, Dr. Ratain has a patent related to samples. This hypothesis should be validated and con- UGT1A1 genotyping and irinotecan use with royalties paid to the firmed in future studies with a more diverse insurance case Mayo Clinic and a patent related to a genomic prescribing system mix and specifically designed to identify populations at licensed to PrescriptIQ. risk, such as those with less than a high school education, which was also a population relatively underrepresented AUTHOR CONTRIBUTIONS in the current study sample (5.1%). Similarly, the differ- Jonas A. de Souza participated in planning, conducting, interpret- ing and reporting and is the guarantor of the study. Bonnie J. Yap, ences with regard to the drivers of financial toxicity among Kristen Wroblewski, Christopher K. Daugherty, and David patients with different cancer types should be examined Cella participated in planning, conducting, and reporting the because these drivers may vary by the type of disease or by study. Victoria Blinder, Fabiana S. Ara ujo, Fay J. Hlubocky, the type of treatment. Although toxicity thresholds were Lauren H. Nicholas, Jeremy M. O’Connor, Bruce Brockstein, not analyzed within this validation sample, the quantita- and Mark J. Ratain participated in interpretation and reporting. tive nature of the COST instrument will allow further prospective studies to determine the exact impact of finan- REFERENCES cial toxicity on HRQOL in comparison with other symp- 1. Bestvina CM, Zulling LL, Rushing C, et al. Patient-oncologist cost communication, financial distress, and medication adherence. J Oncol toms and on other outcomes, such as survival. These Pract. 2014;10:162-167. prospective studies also are needed to determine respon- 2. Northouse L, Williams Al, Given B, McCorkle R. Psychosocial care for family caregivers of patients with cancer. J Clin Oncol. 2012;30: siveness to change, as well as floor and ceiling effects of the 1227-1234. COST-PROM. Finally, because the objective of the cur- 3. Pollack, A. For profit, industry seeks cancer drugs. The New York rent study was to validate whether the COST-PROM cap- Times. September 1st, 2009. http://www.nytimes.com/2009/09/02/ health/research/02cancerdrug.html. Accessed September 20, 2016. tured financial toxicity, we needed patients who had 4. de Souza JA, Yap BJ, Hlubocky FJ, et al. The development of a fi- actually had experience with health care bills. Therefore, nancial toxicity patient-reported outcome in cancer: the COST mea- sure. Cancer. 2014;120:3245-3253. we assessed patients with stage IV disease who were receiv- 5. Zafar SY, McNeil RB, Thomas CM, Lathan CS, Ayanian JZ, ing treatment for at least 2 months with chemotherapy, Provenzale D. Population-based assessment of cancer survivors’ fi- and we were unable to conclude that newly diagnosed nancial burden and quality of life: a prospective cohort study. J Oncol Pract. 2015;11:145-150. patients would provide similar results. 6. Lathan CS, Cronin A, Tucker-Seeley R, Zafar SY, Ayanian JZ, As society increasingly considers the costs incurred Schrag D. Association of financial strain with symptom burden and by patients with cancer as a side effect of treatment, instru- quality of life for patients with lung or colorectal cancer. J Clin Oncol. 2016;34:1732-1740. ments to measure financial toxicity should be patient- 7. Neugut AI, Subar M, Wilde ET, et al. Association between prescrip- centered, scientifically derived, and clinically relevant. tion co-payment amount and compliance with adjuvant hormonal therapy in women with early-stage breast cancer. J Clin Oncol. 2011; The correlation with HRQOL establishes financial toxici- 29:2534-2542. ty as a clinically relevant, patient-centered measure. The 8. Ramsey SD, Bansal A, Fedorenko CR, et al. Financial insolvency as incorporation of financial toxicity assessments into obser- a risk factor for early mortality among patients with cancer. J Clin Oncol. 2016;34:980-986. vational research and clinical trials will ensure a patient- 9. The Henry J. Kaiser Family Foundation. What the actuarial values centered foundation in the evaluation of financial distress. in the Affordable Care Act mean. http://www.kff.org/healthreform/ upload/8177.pdf. Accessed September 20, 2016. 10. Ungar L, O’Donnel J. Dilemma over deductibles: costs crippling mid- FUNDING SUPPORT dle class. http://www.usatoday.com/story/news/nation/2015/01/01/ middle-class-workers-struggle-to-pay-for-care-despite-insurance/19841 Supported by grant University of Chicago Institute for Translation- 235/. Accessed September 20, 2016. al Medicine-Clinical Translational Science Award UL1 TR000430. 11. High-deductible health plans, gamble for some, on the rise [tran- script]. All Things Considered. National Public Radio. July 28, 2013. http://www.npr.org/2013/07/28/206370593/high-deductible-health- CONFLICT OF INTEREST DISCLOSURES plans-gamble-for-some-on-the-rise. Accessed September 20, 2016. Kristen Wroblewski received a grant from the National Institutes of 12. Zafar SY, Peppercorn JM, Schrag D, et al. The financial toxicity of Health/National Cancer Institute (P30 CA14599) for work per- cancer treatment: a pilot study assessing out-of-pocket expenses and formed as part of the current study. Mark J. Ratain has received the insured cancer patient’s experience. Oncologist. 2013;18:381-390. Cancer February 1, 2017 483 Original Article 13. Huntington SF, Weiss BM, Vogl DT, et al. Financial toxicity in in- 20. Patrick DL, Burke LB, Gwaltney CJ, et al. Content validity–estab- sured patients with multiple myeloma: a cross-sectional pilot study. lishing and reporting the evidence in newly developed patient- Lancet Haematol. 2015;2:e408-e416. reported outcomes (PRO) instruments for medical product evalua- 14. U.S. Department of Health and Human Services FDA Center for tion: ISPOR PRO Good Research Practices Task Force report: part Drug Evaluation and Research; U.S. Department of Health and Hu- 2-assessing respondent understanding. Value Health. 2011;14:978- man Services FDA Center for Biologics Evaluation and Research; U.S. Department of Health and Human Services FDA Center for 21. Mokkink LB, Terwee CB, Patrick DL, et al. The COSMIN check- Devices and Radiological Health. Guidance for industry: patient- list for assessing the methodological quality of studies on measure- reported outcome measures: use in medical product development to ment properties of health status measurement instruments: an support labeling claims: draft guidance. Health Qual Life Outcomes. international Delphi study. Qual Life Res. 2010;19:539-549. 2006;4:79. 22. Crowne DP, Marlowe D. A new scale of social desirability indepen- 15. Cella DF, Jacobsen PB, Orav EJ, Holland JC, Silberfarb PM, Rafla dent of psychopathology. J Consult Psychol. 1960;24:349-354. S. A brief POMS measure of distress for cancer patients. J Chronic 23. Park BW, Lee S, Lee AR, Lee KH, Hwang SY. Quality of life differ- Dis. 1987;40:939-942. ences between younger and older breast cancer patients. J Breast 16. Alexander GC, Casalino LP, Meltzer DO. Patient-physician commu- Cancer. 2011;14:112-118. nication about out-of-pocket costs. JAMA. 2003;290:953-958. 17. Cella DF, Tulsky DS, Gray G, et al. The Functional Assessment of 24. Faller H, Brahler E, Harter M, et al. Performance status and depres- Cancer Therapy scale: development and validation of the general sive symptoms as predictors of quality of life in cancer patients. A measure. J Clin Oncol. 1993;11:570-579. structural equation modeling analysis. Psychooncology. 2015;24: 18. Aaronson NK, Ahmedzai S, Bergman B, et al. The European Orga- 1456-1462. nization for Research and Treatment of Cancer QLQ-C30: a 25. Streiner DL, Norman GR. Health Measurement Scales: A Practical quality-of-life instrument for use in international clinical trials in on- Guide to Their Development and Use. 2nd ed. New York: Oxford cology. J Natl Cancer Inst. 1993;85:365-376. University Press; 1995. 19. Patrick DL, Burke LB, Gwaltney CJ, et al. Content validity–estab- 26. StataCorp LP. Stata Statistical Software: Release 13. College Station, lishing and reporting the evidence in newly developed patient- TX: StataCorp LP; 2013. reported outcomes (PRO) instruments for medical product evalua- 27. Chastek B, Harley C, Kallich J, Newcomer L, Paoli CJ, Teitelbaum tion: ISPOR PRO Good Research Practices Task Force report: part AH. Health care costs for patients with cancer at the end of life. 1–eliciting concepts for a new PRO instrument. Value Health. 2011; 14:967-977. J Oncol Pract. 2012;8:75s-80s. 484 Cancer February 1, 2017

Journal

CancerPubmed Central

Published: Oct 7, 2016

There are no references for this article.