A Descriptive Study of HIV Patients Highly Adherent to Antiretroviral

Abstract

HIV medication adherence is a topic of major public health concern in the United States. Adherent patients may be less likely to experience treatment failure, AIDS presentations and extreme medical costs. We evaluate a cohort of highly adherent Medicare beneficiaries to establish if the out of pocket costs of HIV medications are an inherent barrier to adherence. We analyzed a 100% sample of Medicare Part-D prescription medications. The drug and out ofpocket costs for HIV and non-HIV medications of highly adherent cohort were extracted and analyzed. The average gross drug cost per beneficiary was $34,029for HIV medications and $11,439for non-HIV medications. Average out of pocket costs per beneficiary was $454for HIV medications and $129 for non-HIV medications. Out of pocket costs do not reasonably appear to be a barrier to adherence for Part-D beneficiaries.

Introduction

HIV is a major infectious disease and was implicated in at least 16,350 deaths in the United States in 2017.¹ In 2020, the 'National HIV/AIDS Strategy for the United States', a Federal plan to end the HIV epidemic, was updated to include (medication) treatment adherence as a core goal and an underlying requirement of expanding treatment to HIV+ Americans.² Medication adherence for HIV+ patients refers to the consistency with which patients are dispensed, receive and consume HIV medications.³ HIV medications are understood to prevent AIDS presentations among HIV+ patients.^4-6 While some debate remains internationally on when to start patients on retroviral drugs there is consensus that once treatment begins interruption (non-adherence) can produce viral resistance, treatment failure, AIDS and AIDS related mortality.^7-10 While HIV+ patients are dispensed medications other than 'HIV medications' where treatment adherence could influence their clinical outcomes (diabetes, heart disease, cancer) adherence among this population has historically meant dispensation of anti-retroviral drugs alone.^3,11

Cost, clinical complications, toxicity and social determinants are common concerns for the management of HIV treatment adherence. The extreme financial cost of medications in the United States of America (US) are well documented.¹² The clinical complications of HIV disease have also been shown to influence adherence, especially HIV related dementia.^13,14 Medication toxicity and drug interactions have been a major barrier to adherence as HIV treatment emerged and may still be a barrier in specific cases.^15,16 The social determinates of treatment adherence include incarceration, homelessness, poverty, drug addiction and care deserts (rural areas). These socially marginalized experiences (should) make consistent access to any treatment, let alone HIV treatment even more difficult than cost alone.^17-19

Adherence to HIV treatment is a major focus of clinical research because of the cost, consequences and possible explanatory value of HIV medication non-adherence on the outcomes of HIV+ Americans.^20-22 HIV medications are high cost medications and may be out of reach for some patients in the US. Several subsidy programs for HIV medications including Health Resources and Services Administration's Ryan White Part-B, AIDS Drug Assistance Programs (ADAP), Medicaid and manufacturer specific rebates and programs are used as core sources of HIV medications by HIV+ Americans.^23,24 These programs provide medications at lower or no cost, where available.

Use of large observational databases (including claim-based) in retrospective HIV research is less frequent compared to other clinical domains. In this project, our motivation is to demonstrate strength and weaknesses of Center for Medicare and Medicaid Services (CMS) claim-based data for HIV research. This work at the same time a part of a HIV-focused project that identifies and analyzes Common Data Elements (CDE) available in routine healthcare data or in completed HIV clinical trials. Here, we aim to observe if highly adherent patients have high out of pocket costs for HIV and non-HIV medications. We also consider if program specific features of Medicare influence out of pocket costs among a population of highly adherent HIV+ patients.

Methods

Data Source: We use a 100% sample of Medicare encounter level claims from 1999 through 2017. Data was acquired through the Centers for Medicare and Medicaid Services (CMS) Virtual Research Data Centers' Chronic Condition Warehouse (VRDC CCW). VRDC contains both Medicare and Medicaid data. VRDC provides encounter level claims as well as pre-processing features that support analysis. These pre-processing features include a unique patient identifier (Beneficiary ID) which follows individuals over time, across programs (parts A, B, C, D or Medicaid) and pre-calculates coverage and expenditure at the beneficiary level. VRDC pharmacy dispensation data for Medicare Part-D uses 'National Drug Code' (NDC) codes. The data availability delay (non-availability of recent data) is longer for Medicaid; as of fall 2019 Medicare 2016 and Medicaid 2012 encounter level claims were available. Medicaid 2013 and 2014 claims were also available but only for a sub-set of States. This data delay determined our analysis as Medicare centric; our study does not consider Medicaid claims. We used calendar year 2016 since this was the most recent data available in VRDC to us as of October 2019.

Cohort Definitions: We analyzed two cohorts, an 'HIV Cohort' and an 'HIV Adherent Cohort'. The HIV Cohort used the following inclusion criteria: (1) two HIV-1 viral load tests in calendar year 2016 billed to an outpatient (B -Carrier Claim, identified by CPT code 87536) and (2) at least one pharmacy claim in 2016. The HIV Adherent Cohort had the following additional criterion, (3) 80% proportion of days in 2016 covered by a dispensation that included an HIV medication ingredient (patients switching different treatment regiments were this way properly classified as adherent).

Our cohort criteria were created to mitigate possible incompleteness of claims data. For example, the requirement for two viral load tests was used to ensure that Medicare insurance is being utilized and billed for HIV monitoring care. Similarly, the criterion for at least one reimbursed pharmacy claim was used to demonstrate (by presence of a claim) that Part-D medication coverage is accessible to the analyzed patients.

Data Analysis: Our analysis describes the demography, overall costs, clinical drug level costs (HIV and non-HIV separately) and costs per member per month. We aim to discover if highly adherent patients have high out of pocket expenses which could potentially explain non-adherent patient medication avoidance behavior, if it exists. As a secondary outcome we aim to discover if medication seeking behavior differs within our highly adherent cohort by medication type. Our cohort and highly adherent cohort should not be confused for the Medicare or Medicaid HIV burden in the US in 2016.

Adherence was determined at the ingredient level per individual patient. Each medication dispensation had the presence of HIV ingredients evaluated and if present, the date of dispensation and the count of days of supply was extracted. Dates of dispensation were plotted in a SAS array loop to distribute the days of supply from the corresponding date of HIV dispensation to a 365-day calendar view spanning January 1 st to December 31st. For each dispensation the array loops over each Beneficiary ID calendar day cell marking 1 from the beginning of the dispensation date until the days of supply are exhausted. Then the non-zero calendar days are aggregated and divided by 365 to produce the percent of days in the calendar year covered by HIV medications. Days covered by multiple HIV medications (multiple regiment) and combination pills (or other dose forms) with more than one HIV ingredient apply only to the given day (are not counted twice) as days evaluate to either 0 (none) or 1 (present).

Medication Mapping and Value Sets: VRDC raw dispensation data uses NDC codes to identify dispensed medication. We created a custom list of NDCs that contained HIV medication ingredient string tokens. Our list was validated against RxNorm Ingredients by mapping our NDCs to RxNorm Clinical Drugs. Our list of 32 HIV tokens captured 32 RxNorm HIV ingredients. Since multiple NDC codes represent identical clinical drugs, we transformed NDC-coded dispensation data into RxNorm Clinical Drugs. 2918 2016 Part-D NDC codes converted into 138 clinical drug codes. For example, NDCs: 00081010855 (Zidovudine 100 MG Oral Capsule [Retrovir]) and 00081010856 (Zidovudine 100 MG Oral Capsule [Retrovir]) both converted into Clinical Drug 1710616 Zidovudine 100 MG Oral Capsule. In addition to reducing the number of distinct items to analyze, it also allowed us to use RxNorm relationships of clinical drugs to ingredients. We used OMOP Vocabulary as a source of RxNorm mappings and relationships. For example, the top five most frequently used HIV ingredients from our list were: Emtricitabine, Tenofovir Disoproxil, Lamivudine, Ritonavir and Dolutegravir.

Part-D Medication Plan Considerations: Part-D is a highly complex prescription drug program which contains multiple eligible plans and phase specific rules for reimbursement.^25,26 These rules are Part-D specific. Eligibility in Part-D requires that the patient be enrolled in Medicare and eligible for a member plan. Member plans can be low cost and for qualifying beneficiaries, specific low income plans are available.^27,28 All member plans are subject to program level phases which determine reimbursement. Reimbursement amount per dispensation is determined by the amount billed to date by the individual patient and the individual dollars within a dispensation's phase. The amount due by the patient reflects this phased payment structure and changes depending on how much a patient has spent, dollar by dollar and the expenditure thresholds of specific phases. The more a patient pays the further through the Part-D phases they progress until they reach 'catastrophic' personal financial contributions to their prescription drug care threshold. Phase specific financial thresholds vary depending on the calendar year; in 2016 patients reach 'catastrophic' coverage (and enjoy reduced out of pocket expenditure) when their out of pocket spending exceeds $7,062.50 on eligible medications. Patients whose medication contributions have exceeded this financial threshold are reach a 'catastrophic' status, this means that they are no longer affected by possible high co-pays for HIV (and non-HIV) drugs. Catastrophic is the last, and most expensive (for Medicare as a payor) Part-D phase. Once patient payment contributions reach the catastrophic threshold, patients pay a radically reduced share of their medication costs with the remainder of the cost being paid by Medicare and the individual member plan itself.

For example, a patient with a monthly prescription expenditure across multiple prescriptions totaling $1,000 a month would never reach the threshold of $7,062.50 within the calendar year despite $12,000 in annual expenses. In month one, the patient would begin by paying 100% of expenses until the ' deductible phase' is satisfied. Then the patient would pay 25% of expenses with the remainder paid by their plan until the 'coverage gap phase' is satisfied. Similarly, the patient pays 25% of their medication cost with the remainder being paid by the manufacturer until the final 'catastrophic phase' is satisfied. Once a patient spent the $7062.50 and reached the final catastrophic phase, the patient pays 5% of their medication costs (or a flat cash payment of $3 (for generic drugs) to $8 (for brand drugs) dollars per dispensation whichever is smaller).

Expenses covered by the plan and the manufacturer do not count towards the catastrophic patient contribution threshold. In turn, 75% of expenditures do not count towards the catastrophic limit. If the deductible on their plan is $500, they would pay $500 + 25% of $1,000 a month through the end of the year, or $500 + $2,875(25% of $11,500) which is less than $7,062.50. In 2016, a patient's monthly medication expenses would need to exceed 2,500 a month in gross drug costs to reach the catastrophic contribution threshold and enter the final Part-D phase, assuming a deductible of $500. The patient would only be paying 25% of the qualifying amount in most intermediary phases, limiting their contribution towards the $7,062.50 threshold.

Multiple non-patient payers can contribute to patient contribution medication costs (before and after the catastrophic phase). In the case of HIV medications, Ryan White Part-B and Medicaid may contribute to medication costs attributed to the patient through reimbursement, rebates direct supply or additional mechanisms. These contributions may count towards the individual patient's catastrophic threshold. Part-D low income subsidies may also contribute to patient medication expenses which in some non-patient payments would count towards the catastrophic threshold. This study was specifically interested in the Out of Pocket Costs paid by the patient (OOP), Gross Drug Cost (GDC), month of the dispensation and if the dispensation included an HIV anti-retroviral ingredient. OOP in this study does not consider True Out Of Pocket (TrOOP) dollars which are dollars not paid by the insurer, but not necessarily paid by the patient. The patients' payment phase was not considered specifically but out of pocket expenses per-member per-month was used to assume 'catastrophic' phase in population level terms where an aggressive decrease in out of pocket spending would indicate that catastrophic spending was generally reached. In this study OOP is learned from the VRDC variable 'patient total payment amount'.²⁹

Results

Table 1 shows descriptive characteristics of the two cohorts we analyzed. We found 43,859 patients fulfilling criteria for our HIV cohort. 72% (31,784 patients) of such patients are at the same time highly adherent to HIV therapy (fulfilling criteria of the HIV adherent cohort). The demography parameters are mostly comparable between the HIV cohort and the HIV adherent cohort. Blacks are disproportionately represented in both the HIV cohort and the HIV adherent cohort relative to US census data. This reflects the disproportionate burden of HIV among Blacks living in the US.³⁰ The proportion of females is consistent with 20-25% national prevalence across all ages (especially older adults). Medicare is commonly remembered as health plan for seniors aged 65+. The Table 1 demography data, however, show that thanks to other Medicare eligibility mechanism (at least in 2016), many HIV+ patients under the age of 65 use Medicare and Medicare Part D as their health plan that pays for their care. Table 2 shows that 27% of patients in HIV Adherent cohort are 65+ while the remaining 73% are younger than 65 (e.g., 30% are aged 45-54). Age was calculated for each patient as of December 31, 2016.

Table 1. Descriptive characteristic of HIV and HIV Adherent Cohorts *.

	HIV Cohort	HIV Adherent Cohort
Population	Count	Count (proportion)
Distinct Population	43,859	31,784 (0.72)
Median Age	56	57
Gender	Count (Proportion)	Count (Proportion)
Male	32,540 (0.74)	24,028 (0.76)
Female	10,953 (0.25)	7,499 (0.23)
Race	Count (Proportion)	Count (Proportion)
Unknown	344 (0.00)	249 (0.00)
White	20,851 (0.47)	16,140 (0.51)
Black	18,636 (0.42)	12,385 (0.39)
Other	508 (0.01)	400 (0.01)
Asian	359 (0.00)	287 (0.00)
Hispanic	2,594 (0.05)	1,919 (0.06)
N. American Native	201 (0.00)	147 (0.00)
Age	Count (Proportion)	Count (Proportion)
18-24	100 (0.00)	50 (0.00)
25-34	1,383 (0.03)	701 (0.02)
35-44	3,866 (0.09)	2,354 (0.07)
45-54	12,675 (0.31)	9,085 (0.30)
55-64	12,008 (0.29)	9,242 (0.31)
65+	10,504 (0.25)	8,007 (0.27)

^*Demography is reported where consistent across calendar years within VRDC. Inconsistent demography includes patients who report more than one distinct race, gender or whose year of birth is reported as at least two different values on a Medicare encounter claim over our 18 CMS observation years.

Table 2. Cohort level Gross Drug Cost of medications for HIV adherent cohort in calendar year 2016.

	HIV RX	Non-HIV RX	Total
# of Dispensations	782,410	1,971,941	2,754,351
Days of Supply	24,033,279	62,165,378	86,198,657
Out of Pocket Cost	$14,448,956	$4,064,004	$18,512,960
Gross Drug Cost	$1,081,596,649(75 %)	$357,881,840 (25%)	$1,439,478,489 (100%)

Table 2 shows on cohort level the cost and other characteristics of HIV and non-HIV medications in the HIV adherent cohort. HIV medication costs represent 75% of the total medication cost. The top five most dispensed HIV drugs (on RxNorm clinical drug level) were: 1. Emtricitabine 200 MG / Tenofovir disoproxil fumarate 300 MG Oral Tablet, 2. Ritonavir 100 MG Oral Tablet, 3. Raltegravir 400 MG Oral Tablet, 4. Dolutegravir 50 MG Oral Tablet, 5. Efavirenz 600 MG / emtricitabine 200 MG / Tenofovir disoproxil fumarate 300 MG Oral Tablet. The top five most dispensed non-HIV drugs were: 1. Gabapentin 300 MG Oral Capsule, 2. Omeprazole 20 MG Delayed Release Oral Capsule, 3. Zolpidem tartrate 10 MG Oral Tablet, 4. Sulfamethoxazole 800 MG / Trimethoprim 160 MG Oral Tablet, 5. Amlodipine 10 MG Oral Tablet.

Table 3 shows annual person level drug cost for HIV adherent cohort, such as average annual total gross drug cost and out of pocket drug cost. Per person annual gross drug costs are $34,030 for HIV medications and $11,439 for non-HIV medications (GDC for HIV medications are 2.97 higher). Per person annual OOP costs are $18 per dispensation for HIV medications and $5 per dispensation for non-HIV medications (OOP for HIV medications are 3.6 times higher). Table 3 also shows similar comparison for per dispensation GDC cost.

Table 3. Annual person level drug cost and additional drug cost measures (HIV adherent cohort).

		HIV RX	Non-HIV RX
Averages of Gross Drug Cost (GDC) and Out of Pocket (OOP)
	Average GDC Per Person	$34,029	$11,439
	Average OOP Per Person	$454	$129
	Average OOP Per Dispensation	$18	$5
	Average GDC per Day of Supply	$45	$14
	Average OOP Per Day of Supply	$0	$0
GDC per Dispensation
	Mean	$1,382	$181
	Median	$1,363	$12
	Max	$10,511	$80,486
OOP Per Dispensation
	Mean	$18	$2
	Median	$0	$0
	Max	$3,125	$4,174

Graph A shows the volume of dispensations is much higher for some HIV medications relative to the size of the population taking them at the clinical drug level. Graph B also suggests OOP relative to GDC is much higher for HIV rather than non-HIV medications. Two non-HIV medications appear to be in OOP to GDC class with HIV medications.

Graphs C-F show temporal trend over time (x axis shows 12 months of Jan 2016 through Dec of 2016). Graph C considers over time, the count of distinct patients seeking a dispensation in each calendar month. It shows some monthly variation over time for both HIV and non-HIV medications which appear symmetrical. Graph D plots the count of dispensations by month showing generally much higher volume of non-HIV drug dispensations compared to HIV drug dispensations. Comparing Graphs C and D demonstrates that highly adherent patients are taking HIV medications more consistently than non-HIV medications, but those patients taking non-HIV medications receive more dispensations for non-HIV medications than HIV medications. HIV dispensations never exceed 70 thousand a month, and non-HIV dispensations never drops below 150 thousand dispensations per month. Graphs E and F both show OOP costs (E on per patient bases and F as sum for whole cohort). Both graphs show higher OOP costs early in the year and much lower amounts from April till December.

Figure 2 shows drug cost on individual drug level. To make the graph more readable, we only include frequent HIV clinical drugs that were taken by at least 200 highly adherent cohort patients and had non-zero GDC and OOP costs. Each horizontal bracket represents a distinct clinical drug that contained an HIV ingredient and drugs are listed in decreasing annual cost from left to right. The top three most expensive HIV clinical drugs and their averaged gross drug cost per patient (dark blue line) are: (1) efavirenz 600 MG / emtricitabine 200 MG / Tenofovir disoproxil fumarate 300 MG Oral Tablet ($24,200); (2) abacavir 600 MG / dolutegravir 50 MG / Lamivudine 300 MG Oral Tablet ($23,092); and cobicistat 150 MG / elvitegravir 150 MG / emtricitabine 200 MG / Tenofovir disoproxil fumarate 300 MG Oral Tablet ($22,393). The clinical drugs taken by most patients (see gray bars) are emtricitabine 200 MG / Tenofovir disoproxil fumarate 300 MG Oral Tablet (9,653 patients), Ritonavir 100 MG Oral Tablet (8,888 patients) and raltegravir 400 MG Oral Tablet (6,594 patients). Typical annual supply for a single HIV medication would exceed the 2016 catastrophic threshold ($7,062.50, light blue line) in 26 out of 39 clinical drugs shown in Figure 2. In most cases, HIV+ patients take more than one such HIV clinical drug. If their regimens require more than one clinical drug dispensation (HIV or non-HIV medication) the rate of reaching catastrophic phase accelerate and this reduces the annual, per drug OOP cost faster. If patients experience regiment transition within the highly adherent cohort the GDC per person by clinical drug would not reflect a full 12-month cost but perhaps less than 12 months. GDC per person is not a 'list price' but an average of what was charged for a clinical drug.

Figure 2. Graph of selected HIV clinical drugs (X axis) and their Gross Drug Costs Per-Patient (left Y axis, dark blue line) and count of distinct HIV adherent cohort patients taking that clinical drug (Right Y axis, gray bars).

Discussion

Our study shows that 2016 Medicare data can be used to study a large cohort of HIV+ patients that are highly adherent to ART therapy (34,784 patients in HIV adherent cohort). The total size of HIV+ patients in Medicare data is much larger and we fully acknowledge that missing data (events not reimbursed by Medicare medical plan or Part-D pharmacy plan) is a biasing factor. However, despite this data heterogeneity, there is still a large cohort of highly adherent patients where we observe evidence of regular disease monitoring and treatment. Two similar adherence studies (on other data sources) had sample sizes 2,030 and 23,343 patients (see 'Relationship to prior studies' section below). Within CMS VRDC system, thanks to a shared unique patient identifier across Medicare and Medicaid datasets, it would be theoretically possible to study data overlap (what care and medications are billed to various plans) for patients dually enrolled in Medicare and Medicaid. However, large data delay for Medicaid data prevents such an analysis for recent years (Medicaid is 4 years behind Medicare in terms of data being up to date; in the fall of 2019, Medicaid data from 2012 are the most recent; considering majority of US states). An alternative source, the IBM Watson MarketScan Medicaid database, has a much shorter data delay. Our study is the first to focus on drug out of pocket expenses on a large cohort of HIV patients highly adherent to ART therapy.

The speed at which patients become catastrophic within the calendar year is remarkable. These thresholds were most likely intended to be reached in rare circumstances; however most HIV patients reach them routinely due to high cost of HIV medicines. While Medicare Part-D only provides HIV care to a minority of HIV+ Americans, in-year-adherence and relatively low OOP costs per person could suggest Part-D as a worthy model for providing high costs drugs to patients impacted by the HIV epidemic in the US. Our study also reveals interesting data about the volume of non-HIV medications taken by patients in the HIV Adherent cohort. This is significant because it impacts controlled total out of pocket expenses. Of particular note, in terms of non-HIV medications are drugs used to treat Hepatitis C (especially the eradication therapy for Hepatitis C).

Relationship to prior studies: Several prior retrospective observational studies analyzed HIV adherence using claims data but, to our knowledge, none have targeted out of pocket costs. Kangethe et al analyzed adherent and non-adherent HIV cohorts and total cost of care (medications cost + outpatient/inpatient visit cost) using California Medicaid data from 2007-2016.²⁰ They did observe differences in medication cost (adherent compared to non-adherent) but the total annual cost of care was not significantly different (their non-adherent HIV group had lower medication cost but it was compensated by higher visit costs). Their study did not detect enough 'highly adherent' patients for follow-up-analysis when using a region specific, private network of healthcare claims billed to Medicaid. Our adherent cohort sample size (n: 31,784) is 15.6 times larger compared to Kangethe's study (n: 2,030).

A second relevant study by Youn et al. analyzed Medicaid claims from 2001 to 2010. It found 52.4% 2-year adherence to ART in 2010 (using a threshold of >90% days covered to indicate adherence and considered whole regimen)³¹. This adherence threshold is very high for a single payer program view. Private companies also offer reimbursement, rebates and mail order HIV medications in the US. Compared to Youn's study, we used a lower threshold (>=80% days) and applied it at the ingredient level (rather than a complete HIV regiment). Youn et al. studied a sample of 23,343 patients from years (2001-2010). Our cohorts are not generally comparable as ART initiation was not a requirement in our study but was for theirs. Our monthly pattern of dispensations may indicate that even this lower 80% level may be too high considering that patients may be stockpiling some drugs at the end of calendar year while their co-pays are lower due to catastrophic status.

Limitations: Our study has several limitations. First, it only analyzed HIV patients with insurance coverage and relies on reimbursed pharmacy claims for complete medication history. Patients could have been using additional medication. For example, additional over-the-counter non-HIV medication. Patients who were deemed non-adherent may actually be adherent as medication dispensation not reimbursed by Medicare Part-D is unobserved in our data set. Second, the NDC mapping to RxNorm clinical drug (extracted from OMOP Vocabulary) does not cover all NDC codes appearing in the VRDC pharmacy claims. However, a cursory analysis of the non-mapped NDC codes indicate that the non-mapped NDC codes do not represent any significant HIV or non-HIV clinical drug. Third, our list of HIV ingredients and HIV clinical drug contain drugs that are also used as anti-retrovirals for other infectious diseases (Hepatitis C and B). Using manual review, we excluded some such drugs where review of label showed clear use for non-HIV condition (for example, some clinical drugs containing ritonavir). Moreover, hepatitis C eradication treatment typically only lasts 12 weeks. Fourth, use of HIV drugs in pre-exposure prophylaxis (PrEP) may interfere with our analysis. Patients could qualify for our cohort if they were highly adherent consumers of HIV medication ingredients; and PrEP patients would be HIV- and consuming Truvada. In the HIV cohort, A total of 17 patients had no documented diagnosis for HIV disease across VRDC years. This could be due to payor dynamics, or the 17 patients may be true HIV- cases that consume HIV medications. Fifth, the eligibility criteria for HIV+ patients to quality for Medicare are changing over time. While our 2016 data showed demography indicating wide range of HIV patients by age, later analysis for years after 2016 (where criterial for obtaining Medicare under age 65 were made stricter) may show demography shifted more heavily towards age 65+.

Conclusions

Our study shows that HIV medications have significant cost (HIV medication cost were 75% of total drug cost in HIV adherent cohort). However, our analysis of OOP cost on annual basis for highly adherent patients (at $454 per year per patient) indicates that high OOP does not appear to be a barrier to ART therapy adherence for Part-D Medicare beneficiaries. With carefully constructed cohort inclusion criteria, Medicare data can be used to study some special populations of HIV+ patients, namely those where Medicare claims data do indeed contain evidence of active and regular HIV-related care. Moreover, because of national scale of Medicare, the overall sample size for such special cohort is high compared to studies using other claims-based databases.

Figures & Table

Figure 1. Patients by dispensations (A), GDC by OOP (B) and patients (C), dispensations (D), OOP per member per month (E) and total out of pocket expenditure per-member per-month (F).