Sample & recruitment
In January 2015, in collaboration with CP, a random sample of 8122 CP guarantors was created from CP’s national database. The sample included guarantors from CP’s hospital clients, age 18 or older, who were paying off CP debt as of January 1, 2015. The principal reason for focusing on guarantors (who are usually also the CP patients), and not the patients per se, is that patients may be under the legal age of consent or be legally unable to consent to study participation due to diminished mental capacity. Guarantors are by definition of legal age and capacity to consent to participate. Some hospital clients (and their guarantors) were excluded from the sample due to restrictions in place within their Business Associates Agreements (BAA) with CP. CP provided only the names, addresses and telephone numbers of potential participants; other demographic information on the whole sample was not available. All members of the sample were sent an initial recruitment letter by mail which described the study’s aims and alerted potential participants to the study procedures. Pacific Market Research (PMR), an independent market research firm located in Washington State, was given a subcontract to conduct the telephone surveys.
After the recruitment letters were sent out, PMR made up to four calls to each potential participant over the course of approximately three weeks. The call center verified that the potential participant was making payments to CP, was of legal age to participate and could communicate in English. Guarantors who could not communicate in English were excluded. The desired sample size for the project (1000 respondents) was based on a power calculation designed to represent the approximately 40,600 guarantors who met inclusion criteria. The desired sample size drove the recruitment strategy. PMR made telephone calls to each potential participant until the 1000 respondent threshold was met; in the end 8075 potential participants were called at least once. Due to legal concerns related to the Telephone Consumer Protection Act, no voicemail messages were left for potential participants.
Instrument
Once the potential participant consented to the research, the interviewer posed a series of questions about their medical debt and potential outcomes that could be attributed to medical debt, including both financial challenges (e.g., taking on credit card debt) and access to care (e.g., skipping doses of prescription medicine due to cost). The majority of the questions were drawn from the 2012 Biennial Health Insurance Survey (BHIS), and others were drawn from Kaiser Family Foundation surveys on health insurance and the Affordable Care Act. The survey also contained a number of demographic items designed to capture the profiles of guarantors, and questions about satisfaction with CP drawn from a previous satisfaction survey conducted by a CP consultant. At the conclusion of the survey, participants were given the option of receiving a $10 check via mail as a thank you for participation.
Data analysis
A raw data file was created by PMR with the responses to each question. The file did not contain any identifying information about individual respondents (e.g., name, address, telephone number) to comply with HIPAA regulations. CP provided the total amount of debt the guarantor was currently paying off (high balance); this variable was added to the dataset using a unique identifier that linked the data file to CP’s database. The data analysis included simple descriptive statistics (frequencies, means) for all variables and cross-tabulations to compare categorical variables. Tables present sample n’s and valid percentages; respondents omitting answers to a particular question were excluded item-by-item from analyses.
Given the lack of comparison group in this study, the analyses also explored whether individuals using CP for the first time and individuals with other (non-CP) medical debt reported more negative outcomes compared to other guarantors. These analyses assumed that if CP participation is associated with reductions in negative outcomes associated with medical debt in the literature, then an argument can be made that CP influences these outcomes in a positive way. Logistic regression models were run with each of seventeen measured negative outcomes (e.g., taking on credit card debt, skipping needed medical treatments) as an outcome variable (yes/no) and indicator variables for these attributes as predictors. All seventeen models controlled for total amount of medical debt, which included a combination of CP debt and any other medical debt currently being paid off over time (categorical: <$2000; $2000–3999; $4000–7999; $8000–9999; $10,000 or more), along with household income (categorical: <$20,000; $20,000–39,999; $40,000–59,999; $60,000–79,999; $80,000 or more).
We received a Waiver of Prior Authorization under HIPAA in order to contact guarantors, and all study procedures were reviewed and approved by the Institutional Review Board of Arcadia University (Federal-wide assurance #00000449). While the researchers collaborated with CP in order to access information about guarantors (e.g., phone numbers, names, amount of debt), the study was conducted independently, such that CP staff were not involved in data collection, analysis or report writing. The study was funded by the W.K. Kellogg Foundation.