## Thursday, February 9, 2017

### Some statistical tests pertaining to health expenditures in 2013 and 2014

Some statistical tests pertaining to health expenditures in 2013 and 2014

Question:   Consider the population of people with 12 months of private insurance coverage in 2013 and 12 months of private coverage in 2014.   What happened to the mean and median in per-capita out-of-pocket health expenditures for fully insured people after the opening of state exchanges?

Note:   The purpose here is to examine how the ACA may have impacted expenditures for people with private health insurance.  Hence, everyone in both samples has coverage for all 12 months.

Data and Methodology:  The data used in this study was obtained from the 2013 and 2014 MEPS surveys.   The variables examined were TOTEXP14, TOTEXP13, TOTSLF14, and TOTSLF13.

The tests presented here are the standard classical t-test on the difference between two means, the non-parametric test of differences in medians and the Wilcoxon rank sum test.

The test on medians works by calculating a common median for the variables in the two years and looking at the percent above and below the common median.

The RANKSUM test combines the two years and compares the sum of the ranks for each year to the expected sum of the ranks for each year.

STATA was used to calculate the test statistics.  Here is an informative link to information on using STATA for median and RANKSUM comparisons.

The statistical tests presented here do NOT employ survey weights.

Results of Statistical Tests:

 Comparison of Means for total out-of-pocket expenses and total health expenses TOTSLF TOTEXP 2013 633.7 4068.857 2014 589.8 4330.368 Diff. -43.9 261.5 % Diff. -6.9% 6.4% Two-Tailed p-value for classical t-test 0.0302 0.0935

 Comparison of Medians for total out-of-pocket expenses and total health expenses TOTSLF TOTEXP 2013 186 972 2014 170.5 951 Diff. 15.5 21 % Diff. 8.3% 2.2% p-value for Test on Equality of Medians 0.006 0.419 p-value for rank sum test 0.0007 0.4814

Discussion of Statistical Tests:

The per-capita mean out-of-pocket expenditure went down by 6.9%.   This change is significant at the 0.05 level.

The mean total health expenditures went up 2.2%.   This change is significant at the 0.10 level.

The tests reveal a 8.3% decrease in median out-of-pocket expenses.  The difference in median out-of-pocket expenses is highly significant.

The change in per-capita median out-of-pocket expenditures is not significant.

Comments on Results:

It appears as though the ACA had a small impact on reducing out-of-pocket health expenses even though total per-capita health expenditures grew and innovations like higher deductible health plans continue to grow in popularity.

The sample sizes used for these tests are around 14,000 fully insured people per year.   This is not large enough for me to be really confident of results because extremely expensive cases have a disproportionate impact on both total and out-of-pocket health expenditures.    A general survey of the entire population does not allow us to understand what is happening among really expensive health care cases.   The only way to get a handle on the expensive cases is through a separate sample of really expensive cases.  The more precise information about expensive cases would then be combined with results from the general survey in order to get a more accurate estimate of average expenditures.