Kristin N. Ray, MD, MS; Amalavoyal V. Chari, PhD; John Engberg, PhD; Marnie Bertolet, PhD; and Ateev Mehrotra, MD, MPH
Objectives: The typical focus in discussions of healthcare spending is on direct medical costs such as physician reimbursement. The indirect costs of healthcare—patient opportunity costs associated with seeking care, for example—have not been adequately quantified. We aimed to quantify the opportunity costs for adults seeking medical care for themselves or others.
Study Design: Secondary analysis of the 2003-2010 American Time Use Survey (ATUS).
Methods: We used the nationally representative 2003-2010 ATUS to estimate opportunity costs associated with ambulatory medical visits. We estimated opportunity costs for employed adults using self-reported hourly wages and for unemployed adults using a Heckman selection model. We used the Medical Expenditure Panel Survey to compare opportunity costs with direct costs (ie, patient out-of-pocket, provider reimbursement) in 2010.
Results: Average total time per visit was 121 minutes (95% CI, 118-124), with 37 minutes (95% CI, 36-39) of travel time and 84 minutes (95% CI, 81-86) of clinic time. The average opportunity cost per visit was $43, which exceeds the average patient’s out-of-pocket payment. Total opportunity costs per year for all physician visits in the United States were $52 billion in 2010. For every dollar spent in visit reimbursement, an additional 15 cents were spent in opportunity costs.
Conclusions: In the United States, opportunity costs associated with ambulatory medical care are substantial. Accounting for patient opportunity costs is important for examining US healthcare system efficiency and for evaluating methods to improve the efficient delivery of patient-centered care.
Am J Manag Care. 2015;21(8):567-574
Time spent seeking healthcare represents a burden to patients, lost productivity to employers and society, and a potential inefficiency within healthcare systems. The Institute of Medicine has identified improving timeliness of care, including reducing waiting time, as 1 of the 6 key quality goals in the US healthcare system.1
Patient time burden (measured in minutes) and patient time costs (measured in dollars) are 2 methods of measuring the time spent by patients traveling to, waiting for, and receiving medical care. While guidelines recommend that patient time costs should be included in economic evaluations,2
these time costs are rarely addressed, often due to lack of perceived importance or unavailable data.3
Opportunity costs, which value patient time based on the value of foregone activities, are 1 method of estimating patient time costs. Opportunity costs are increasingly relevant given the increasing emphasis on patient-centered care,1
recognition that some physician visits may not require face-to-face care,4,5
and innovation in healthcare delivery options that may reduce time burden (eg, telemedicine).
To date, there exist no rigorous national estimates of opportunity costs associated with ambulatory medical care. One prior study provided important estimates of the time burden associated with adult ambulatory visits,6
but these time estimates did not include time costs, which are needed to incorporate these time burdens into economic assessments. Other prior studies offered estimates of time spent during only specific portions of ambulatory encounters such as travel time7
and face-to-face physician time.8
Indirect costs of specific illnesses or procedures are often estimated, but these estimates only rarely include patient time costs.9-13
Using nationally representative surveys, we estimated opportunity costs for adults seeking medical care for themselves or loved ones. We examined both per visit opportunity costs for ambulatory medical visits and aggregate opportunity costs across all physician visits in the United States.
We used 3 nationally representative data sources. We assessed opportunity costs associated with seeking ambulatory medical care using the 2003-2010 American Time Use Survey (ATUS). To further contextualize our results, we also determined time spent face-to-face with providers using the 2003-2010 National Ambulatory Medical Care Survey (NAMCS), and annual number of ambulatory physician visits and per visit direct medical costs using the 2010 Medical Expenditure Panel Survey (MEPS).Description of Surveys
The ATUS, administered by the Bureau of Labor Statistics, estimates time spent by noninstitutionalized civilians within the US population by surveying individuals randomly selected from households that completed the Current Population Survey.14
ATUS respondents are interviewed on a randomly selected day (including weekends). Via telephone interview, respondents recount time spent from 4 am the prior day until 4 am on the interview day; these time diaries are then coded by activity. The response rate has ranged from 53% to 58% from 2003 to 2010, with survey fatigue being the primary reason for nonresponse.15
Sampling weights and successive difference replicate weights allow for nationally representative estimates. ATUS includes detailed demographic, employment, and income data. The 2003-2010 data files include 106,657 respondents aged ≥18 years.
The NAMCS, administered by the National Center for Health Statistics, characterizes visits to office-based physicians.16
NAMCS uses multistage sampling: first sampling physicians within primary sampling units, then sampling patient visits from among sampled physicians’ visits. The 2003-2010 NAMCS includes 223,516 physician visits with sampling weights for national estimates.
The MEPS, administered by the Agency for Healthcare Research and Quality, was used to estimate costs and counts of physician visits.17
Expenditures are collected through household and medical provider interviews regarding actual payments by patients, public and private insurance, other public programs, and any other sources. We estimated the annual number of physician visits nationally, out-of-pocket costs, and total expenditures per physician visit. The 2010 MEPS Household Component sampled 32,846 respondents with sampling weights for national estimates, and includes 19,053 respondents reporting 1 or more physician visits.Measuring Time Components of Visits
The ATUS codes distinguish between time spent seeking medical care for oneself, for another adult, or for a child. While ATUS has separate codes for time obtaining medical care and time waiting for medical care, these codes do not adequately differentiate between time actually spent receiving clinical care and other time in the clinic. For example, ATUS categorizes the act of paying for care as “obtaining care” and includes “waiting while doctor examines child” as “waiting for care.” Because these delineations do not appear to adequately distinguish between times where medical care was and was not being received, we aggregated obtaining and waiting time into “clinic time,” which represents time spent obtaining or waiting for care. Additionally, ATUS specifically codes time spent traveling for medical care for oneself, and also codes travel related to caring for another adult or child. We included time spent traveling for medical care for oneself or time traveling related to caring for another adult or a child as “travel time” only when the person also reported “clinic time” on that day. “Total time” was the sum of travel time and clinic time. We excluded a small number of extreme outliers (>6 hours clinic time, n = 74) to focus our analysis on ambulatory encounters. This resulted in 3927 respondents reporting clinic time for themselves, other adults, or children.
Given the limitations of ATUS categories for distinguishing between time obtaining care and time waiting for care, we used NAMCS to estimate average face-to-face provider time. For each sampled physician visit (adult [n = 185,412] and pediatric [n = 38,104]), the physician or nurse working with the physician is asked to record “time spent with physician.” NAMCS data were not used to estimate total time or opportunity costs. Instead, these estimates were determined to contextualize the time estimates obtained from the ATUS.Estimating Opportunity Costs
We estimated opportunity costs for all ambulatory medical visits and also for the subset of visits by employed individuals through methods used previously to determine opportunity costs of informal elder care.18
For employed ATUS respondents (n = 1925, 49% of all respondents with a visit), we estimated opportunity costs using self-reported wages. In typical labor economic theory, hourly wage is considered a valid measure of the value of one’s time during both working and nonworking hours.19
For respondents reporting employment but not reporting wages (n = 305), we imputed wages through a linear regression model using age, sex, race/ethnicity, education, year, and state. For each respondent, hourly wages were multiplied by total time reported within the ATUS to determine a total opportunity cost inclusive of both travel and clinic time. In sensitivity analysis, we determined opportunity costs only for those reporting wages; wages were adjusted to 2010 dollars using the Consumer Price Index.20
For unemployed ATUS respondents, we valued each individual’s time by imputing wages. Because our previously estimated linear regression model only described the relationship between wages and socioeconomic characteristics for the sample of working individuals, imputing wages for nonworking individuals using this estimated relationship would result in biased estimates of opportunity cost. For this reason, we adopted the approach of Heckman21
in order to treat this bias as an omitted-variable problem and to correct for it using a 2-step approach. Specifically, we first estimated a regression that predicted work-participation, and we then used the estimated coefficients to construct the omitted variable (the inverse Milles ratio), which was then included in the regression predicting wages. Our models of wages and workforce participation both included age, sex, race/ethnicity, education, year, and state. Additionally, the 2-step procedure is strengthened by the specification of a variable that predicts workforce participation but not wages; the presence of household children under 6 years old was included for this purpose.
PDF is available on the last page.