Estimating Patient Out-of-Pocket and Time Costs With MEPS Data

We estimated out-of-pocket and time costs among cancer survivors and adults without a cancer history (as the comparison group) from the MEPS data stratified by age group (18-64 years and ≥65 years). Cancer survivors were identified from a question asking if a doctor or other health professional had ever told the person they had cancer or a malignancy of any kind. Respondents were asked about their age(s) at each cancer diagnosis, and the time since first cancer diagnosis was calculated as the difference between age at first diagnosis and age at the survey interview and categorized as less than 2 years, 2-5 years, 6-10 years, and longer than 10 years, or unknown. Other characteristics included sex, race and ethnicity, marital status, educational attainment, health insurance coverage, and MEPS priority conditions (arthritis, asthma, diabetes, emphysema, heart disease [angina, coronary heart disease, heart attack, other heart condition or disease], high cholesterol, hypertension, and stroke), which were classified by the total number of conditions.

Annual Out-of-Pocket Spending Statistical Analyses. Annual out-of-pocket medical spending measured in the MEPS included patient out-of-pocket payments for hospital inpatient stays, emergency room visits, provider and outpatient visits, prescription drugs, and other medical services not covered by health insurance. Net out-of-pocket spending associated with cancer was calculated as the difference between cancer survivors and adults without a cancer history by age group. All spending was adjusted to 2019 US dollars. To preserve sample weights and nationally representativeness of our estimates, we did not match adults without a cancer history to cancer survivors. Instead we used multivariable 2-part models to estimate out-of-pocket costs adjusted for characteristics that vary between adults with and without a cancer history, including age, sex, educational attainment, and number of comorbid conditions. In the 2-part model, the first part is a logistic model for the probability of having any spending, followed by a generalized linear model with a gamma distribution and a log link among individuals with any spending. This approach is commonly used with health-care spending data because of the many individuals with zero spending and the skewness of the distribution among individuals with any spending (21,22,32,33). P less than .05 was considered statistically significant, and all tests of statistical significance were 2-sided. All estimates were weighted to account for the MEPS complex survey design and survey nonresponse.

Annual Patient Time Costs. Patient time costs include round-trip travel to care, waiting for care, and receiving care and were estimated by calculating annual medical service frequencies, applying service-specific time estimates, summarizing annual patient time, and multiplying by the hourly value of patient time, as has been done elsewhere (15–17). Medical service categories were identified from the MEPS visit files and consolidated files (29) and included overnight hospitalizations, emergency room visits, ambulatory surgery, provider office-based or hospital outpatient visits, chemotherapy, and radiation therapy. The MEPS stopped collecting information separately about chemotherapy and radiation therapy in 2013; estimates of service frequencies for chemotherapy and radiation therapy are based on data from 2008-2012 only. Annual service frequency was calculated for each service category. The annual hospital length of stay was a summary of inpatient days from all hospitalizations for the year.

Estimates of patient time associated with round-trip travel to care, waiting for care, and receiving care were calculated separately for each service category using national data sources from previously published studies (15–17). For example, the average time spent with a physician during an office visit in these earlier studies was calculated from the National Ambulatory Medical Care Survey. Patient time for emergency room visits was calculated as the difference between arrival time and discharge time from the National Hospital Ambulatory Medical Care Survey Emergency Department Patient Record. Patient time in the hospital (in days) was measured as the difference between admission and discharge dates and multiplied by 16 hours, an estimate of waking hours that could alternatively be spent pursuing usual activities, including work and leisure. Round-trip travel time to usual source of medical care was estimated from responses to a question from the MEPS about how long it takes to get to the usual medical provider and was added to all service time estimates. Waiting time was added to office-based or hospital outpatient visits, chemotherapy, and radiation therapy estimates. Time estimates for emergency room visits, hospitalizations, and ambulatory surgeries were based on the difference between admission and discharge time, so waiting time was not added to these estimates separately. All patient time estimates were estimated separately by metropolitan statistical area and nonmetropolitan statistical area status to reflect any differences in urban and rural travel, wait time, or practice patterns. As in previous studies (15–17), we used the median US wage ($19.14/h in 2019) to value patient time in our primary analyses of all services as well as for service-specific estimates. Another approach for valuing patient time based on age- and sex-specific wages, also known as the “human capital” approach (17,18), differentially values time for people not in the workforce or who have lower-paying jobs than for people with higher-paid work. In this study, we chose to value patient time equally with the median wage to avoid these inequities.

Annual Patient Time Cost Statistical Analyses. Estimates of annual service frequencies, patient time, and patient time costs for cancer survivors and adults without a cancer history used separate multivariable analyses to control for age, sex, educational attainment, and the number of comorbid conditions. We present adjusted predicted marginals from the multivariable regression analyses, which directly standardize the outcome of each group to the covariate distribution of the overall population (34). These standardized results can be compared like percentages. Net patient time cost associated with cancer was calculated as the difference in time costs between cancer survivors and adults without a cancer history by age group. P less than .05 was considered statistically significant, and all tests of statistical significance were 2-sided. All estimates were weighted to account for the MEPS complex survey design and survey nonresponse.