Lipira and colleagues recently published this comprehensive review article framing a research agenda to evaluate the impact of the Affordable Care Act (commonly called the “ACA” or “Obamacare”) on HIV care, outcomes, prevention, and disparities. Lipira and colleagues identified several challenges to comprehensive HIV care that could be remedied with informed programming and policy. First, the ACA does not increase access to drugs or care for people living with HIV due to a change in the source of coverage for these patients and a shift in the distribution of costs among the healthcare payers. Second, some wrap-around care services that provide food, housing, and transportation could be lost to people with HIV who go from uninsured to insured, which can place them in a vulnerable position without healthcare coverage.
The researchers pose several important but challenging questions for us to answer, including:
- Do changes in access to and/or quality of care after the ACA lead to improvements in health for patients with HIV?
- What are the drivers of total costs and cost growth reduction in HIV care and prevention?
- How is the incidence of HIV changing afer the ACA and to what extent are changes in HIV incidence attributable to the ACA?
- Which subpopulations are benefiting the most/least from changes in access/quality/spending?
How can we move from association to causal inference?
Crude statistical methods can lead to bias conclusions and misinform policy decision makers. Rigorous study design with appropriate statistical analysis is necessary to provide unbiased empirical evidence of a policy’s effectiveness. Several quasi-experimental approaches are available (e.g., regression discontinuity design, interrupted times series analysis, and propensity score matching); however, they are limited by the type of scenario and data that are available. Certain assumptions are necessary to generate causal interpretation from these methods, without which, biased conclusions can still emerge from a lack of internal validity. Stata 15 has new features for extended estimating equations that we could try. What methods do you think are needed for causal inference?
What data do we need?
I am intrigued by these policy questions; however, a lack of a single dataset (that I know about) from commercial claims or the government that has all the necessary variables prevent me from designing a study that would result in causal interpretation. We need to pull information from lots of credible sources to assemble a dataset to support that study design. I want the usual cascade of care numbers (i.e., cases diagnosed, linked to care, on antiretroviral therapy, and virally suppressed), sexually transmitted infections, costs by payer type (including Ryan White HIV/AIDS Program, Centers for Medicare/Medicaid Services, private insurance, patient out-of-pocket) and all by state and over time (at least years but preferably a smaller increment). Do we need patient-level data or would aggregates by state be sufficient? It is possible that models based on one set of sources (Medicaid and CDC Surveillance) could be validated with comparison to the observations from another source (Truven Marketscan)?
I leave you with Lipira’s conclusion:
In order to make a meaningful impact on relevant health care policy, swift and judicious contribution to this body of literature is imperative.
Let’s get to work.
Source: Lipira L, Williams E, Hutcheson R, Katz A. Evaluating the Impact of the Affordable Care Act on HIV Care, Outcomes, Prevention, and Disparities: A Critical Research Agenda. Journal of Health Care for the Poor and Underserved, Volume 28, Number 4, November 2017, pp. 1254-1275.
Acknowledgments: Drs. Mark Bounthavong and Nathaniel Hendrix contributed to this post.