Hi I used 2018-2021 data for those age 65+ and found that only about 45% had Medicare Part D coverage. The Centers fof Medicare and Medicaid Services, 75% of the age 65+ group has Medicare Part D. Does anyone know about why there is such discrepancy? Percentages with Medicare Advantage program and private insurance coverage are consistent with the known perentages. On the data set, the variable name is: MCAREPARTD. I looked up NHIS questionnaire to see if it only refers to stand-alone Medicare Part D (that is, not including Part D through Medicare Advantage programs). If that is the case, the ~45% in each year is about right. However, the questionnaire skip patterns does not specifically mention to exclude Medicare Advantage people. Would anyone be able to shed light on this? Thanks so much.
I want to note that we have seen this request and I am looking into it, but don’t have a definitive answer at this time (but hope to have more information next week). My first suspicion is that some amount of the discrepancy is related to the target population of the two data collections. NHIS focuses on the community-dwelling population (e.g., it excludes individuals in nursing homes, which may be highly relevant for this question). Additionally, while the universe of MCAREPARTD should be restricted to those on Medicare (IPUMS NHIS variable HIMCAREE), I wonder about the focal population of CMS data versus NHIS data is also a contributing factor. I assume that there is another layer beyond this and hope to have an update next week. Thanks for your ongoing patience!
Thanks so much for your response. I will wait for your update next week. I was looking up NHIS questonnaires, and I wonder the difference is also becasue NHIS Part D question did not include those in Part C or MA-PD?
I have another question about discrepancy on Medicare Advantage (Part C). In fact, the NHIS Part C percantage is lower than the national average of Part C. The NHIS Part C question is specifically about those who have Part B and thoese who refused/did not know if they had Part B. Would you be able to shed light on this, too?
Thanks for your patience. To avoid burying the lede, I will begin with a comment shared by a colleague who is much more familiar with healthy policy literature and available data resources. They indicated that in general, self-reported survey data tends to undercount public program enrollment when compared to administrative enrollment counts. If you are interested in actual enrollment numbers they recommend administrative data (MCBS data may be a good place to start). If you are interested in correlates of the fee-for-service traditional Part D benefit, the NHIS could be very helpful.
Here is a quick tabulation of MCAREPARTD for persons age 65+ in 2091-2021 using IPUMS NHIS.
Compared to this Kaiser Family Foundation report, the NHIS numbers approximately map onto the blue half of the stacked-bars in Figure 1 (even moreso with consideration for the figure not imposing age restrictions and the NHIS not including institutionalized populations). In line with your hypothesis, the blue bars in the KFF figure are only those enrolled in the Medicare Prescription Drug Plan and excludes those with Part D enrollment in Medicare Advantage. While the skip patterns don’t indicate an exclusion of Medicare Advantage enrollees, the MCAREPARTD survey text explicitly states “also known as the Medicare Prescription Drug Plan”. Because not all Part C managed plans include a Part D drug benefit, replicating this total likely isn’t as simple as adding the Part C and Part D numbers.
I am not sure I understand your second question about Part C. If you are asking for clarification of the universe for MCADVNTG, your interpretation is correct: only persons who reported that they had Medicare Part B coverage (or were not sure if they did) were asked about their Part C coverage. I did a quick tabulation of enrollment in 2019-2021 and see values ranging from 15.9 million in 2019 to 21.2 million in 2021. I see the discrepancy you describe, but am not aware of additional information beyond the general comment above about enrollment counts versus correlates of enrollment.
I hope this helps. I am sure you were hoping for a more definitive response.
Thanks so much for your helpful response. I will take all into account when I prepare my report (a correaltional study).