Financial Toxicity in Adults With Cancer: Adverse Outcomes and Noncompliance

Thomas G. Knight; Allison M. Deal; Stacie B. Dusetzina; Hyman B. Muss; Seul Ki Choi; Jeannette T. Bensen; Grant R. Williams


J Oncol Pract. 2018;14(11) 

In This Article


We conducted a retrospective, cross-sectional analysis of data collected for a large hospital-based observational cancer cohort. The University of North Carolina (UNC) at Chapel Hill Health Registry/Cancer Survivorship Cohort is an institutional review board–approved (No. 09–0605) incident/prevalent cohort of patients with cancer that integrates a comprehensive database of clinical, epidemiologic, and interview data with repositories of biologic specimens and tumor tissue. Participants were recruited by research assistants embedded into a variety of UNC Health Care oncology outpatient clinics at the North Carolina Cancer Hospital with the following eligibility criteria: age ≥ 18 years, North Carolina mailing address, and English or Spanish language proficiency. All patients meeting these criteria in these clinics were approached by research staff, and 52% of approached individuals were successfully enrolled. Participants who were unable to provide informed consent or participate in interview questionnaires were excluded. Patients were enrolled in the Health Registry between May 2010 and October 2015, and these patients completed baseline questionnaires between January 2011 and November 2015. For inclusion in our analysis, we restricted data to those from participants ≥ 18 years old who had a recorded cancer diagnosis and had completed the survey ≥ 90 days from their diagnosis. We chose this time frame to theoretically capture participants who had already been receiving bills for their treatment and were able to gauge the impact of these bills on their life.

Within 2 weeks of enrollment, interviews were conducted by trained staff using a computer-assisted telephone interview. The interview included numerous structured and validated questionnaires with topics ranging from medical and social histories to cancer-specific health assessments, including barriers to care. The focus of our analyses was the Patient Satisfaction Questionnaire (PSQ-18). The PSQ-18 is an 18-item scale created for patients to evaluate their doctor; it specifically measures general satisfaction, technical quality, interpersonal manner, communication, financial aspects, time spent with doctor, and accessibility and convenience.[22,23] It has also been previously used in the assessment of financial toxicity.[24] Our analyses also included in the Functional Assessment of Cancer Therapy–General Population (FACT-GP) and an internally developed access to health care questionnaire. The FACT-GP, version 4, is a 21-item scale that measures health-related quality of life using four subscales: physical, functional, emotional, and social well-being.[25] The access to health care questionnaire used in this study has not been validated but was constructed using questions adapted from existing validated questionnaires in addition to questions developed by the Health Registry team (Data Supplement).

In addition to survey responses, demographics (eg, age, race) were confirmed with the patient and clinical information was obtained from the UNC Health Care medical record, including cancer diagnosis and stage. Of note, there were only data available in the UNC Health record, because we did not have access to nonaffiliated practices or hospitals. Data received for this project were linked using an honest broker model (the independent UNC Health Registry serving as the honest broker with a de-identified analytic file containing the medical record, demographic, and survey data were provided to the investigators for this analysis).

Our primary outcome was patient-reported financial toxicity using the PSQ-18 statement: "You have to pay for more medical care than you can afford." In the questionnaire, participants were asked to respond to this statement in one of five categories (strongly agree, agree, uncertain, disagree, and strongly disagree). Participants were dichotomized as exhibiting financial toxicity (strongly agree or agree) or not exhibiting financial toxicity (all other responses).

Descriptive statistics were used to describe the baseline characteristics of the sample. Patients' characteristics between the two groups (ie, financial toxicity or not exhibiting financial toxicity) were evaluated using t test for continuous variables and the x2 test for categorical variables. Adjusted risk ratios and 95% CIs for the association between patient clinical and sociodemographic characteristics and the presence of subjective financial toxicity were estimated using Poisson regression with a robust variance estimate.[26] Variables were selected for inclusion in multivariable models, using a combination of clinical relevance and statistical selection criterion in unadjusted analyses (P <.10). All analyses were completed using SAS, version 9.3 (SAS Institute, Cary, NC).