Aging out of dependent coverage and the effects on the use of inpatient medical care

In terms of methodology, this paper is also similar to few recent papers that focus on the pre-ACA period. Before the ACA’s dependent coverage mandate was enforced, many private health insurance contracts covered dependents through age 19 or 23 if they are a full time student. Using a RD design, few studies find that health insurance coverage rates among young adults exhibited a discrete change at these cutoff ages (Anderson et al. 2012, 2014; Cardella and Depew 2014; Yörük 2016, 2017).

We cannot use the waves before 2011 since the ACA was introduced on September, 2010. Until 2012, the MEPS reported beginning and ending date of each hospital stay (including day, month, and year). Based on this information, we were able to compute the exact number of days in each month that an individual had stayed in a hospital. Starting from 2013, the MEPS only reports the month and year of inpatient stays’ beginning date and ending date. Without information on the exact day, we do not have enough information to calculate the length of stay and expenditure associated with it. Thus, our analysis on inpatient stays uses data from 2011 and 2012.

Since information on the exact birth date is not available, it is not possible to determine the exact date of turning 26 for each respondent. Therefore, it is impossible to determine the treatment status of a respondent for the month that she turns 26. In order to address this problem, we exclude the month that each respondent turns 26 from the sample.

Using the event files of the HC of the MEPS, we calculate the exact number of days in each month that an individual stayed in a hospital. If the individual reports multiple inpatient stays at a given month or if the stay spans to multiple months, we calculate the total charge, payment, and out of pocket costs based on the total number of stays in a given month. For instance, if a patient reports two separate stays within the same month, we add up the spending from each stay in order to find the total spending in that particular month. Alternatively, if the stay spans multiple months, we calculate the spending for each month based on the number of days of stay for each particular month.

It is also possible to estimate similar non-parametric RD models. The results from these models are comparable to those from the parametric models. The results from the non-parametric models are not reported due to space constraints but are available from the authors upon request.

Total charge, total payment and out-of-pocket cost tend to be skewed and results might be sensitive to functional form assumption. To address this possibility, we estimate alternative RD models for the log transformation of these variables (we add 1 to each variable before transformation). The results remain statistically insignificant for total charge and total payments. We find a relatively small and marginal significant impact for out-of-pocket costs, but this effect disappears with relatively smaller age bandwidths. We thank an anonymous referee for pointing this suggestion.