2027 Regression Discontinuity in Health Services Research

Imed Bouchrika, PhD

by Imed Bouchrika, PhD

Co-Founder and Chief Data Scientist

What is regression discontinuity in health services research?

Regression discontinuity (RD) is a quasi-experimental design for evaluating an intervention assigned using a measurable threshold. The measure is called the running variable; the threshold is the cutoff. Researchers compare outcomes immediately on either side, where patients or organizations may be similar except for the change in treatment eligibility.

For example, a hospital may offer extra discharge support when a risk score crosses a specified value. An RD study could examine whether readmissions change at that cutoff. It would estimate an effect for patients near the threshold, not for all patients receiving discharge support. The method can inform operational decisions, while the choice between degree choices for healthcare operations and health information careers depends on whether your goal is to act on findings or build the data systems used to produce them.

Table of contents

How does regression discontinuity identify causal effects in healthcare studies?

RD looks for an abrupt change in outcomes at the same point where a policy changes access to treatment. Its causal logic is that, without the policy, expected outcomes would have changed smoothly across the cutoff. A jump in outcomes can then be attributed to the policy only if competing explanations for that jump are credible to rule out.

Consider an age-based coverage rule. Researchers may compare outcomes just below and above the eligibility age rather than comparing all younger and older adults. That narrows differences associated with aging, but it does not solve every problem: insurance options, retirement decisions, or care-seeking behavior may also change at that age. For administrators completing undergraduate preparation before taking on evaluation work, healthcare administration degree completion through accelerated study is a separate decision from acquiring the statistical training needed to make a causal claim.

When should researchers use regression discontinuity designs in healthcare administration?

Use RD when a real, consistently applied rule assigns eligibility at a known cutoff and there are sufficient observations close to it. Potential applications include risk-score-based care management, income-threshold assistance, and facility performance policies. The outcome should be measured comparably on both sides, and the cutoff should be relevant to the decision the administrator needs to make.

RD is less suitable when staff routinely override the rule, the cutoff changed during the study period, or another intervention starts at the same threshold. A health system should also ask whether a local result will answer its operational question: evidence about patients near a risk-score cutoff may support adjusting that cutoff, but it may say little about extending the program to much lower-risk patients. For professionals choosing an initial credential to manage such programs, a healthcare administration degree affordability guide addresses a different question from whether a particular dataset supports RD.

       

How should researchers evaluate regression discontinuity evidence in healthcare studies?

Begin with the decision the study is meant to inform. A statistically credible effect near a cutoff may still be too small, too uncertain, or too narrowly applicable to justify changing a systemwide policy. Conversely, a local result may be valuable when administrators are deciding whether to move that very cutoff.

A practical review should follow these steps:

  1. Confirm the documented assignment rule, cutoff, study period, and whether the analysis concerns eligibility or actual treatment.
  2. Inspect plots and sample sizes near the cutoff, then check whether the estimated effect persists under reasonable bandwidth choices.
  3. Look for evidence of score manipulation, abrupt changes in baseline characteristics, and other policies beginning at the same threshold.
  4. Match the estimated population, outcome, and uncertainty to the proposed operational decision; state what the study cannot establish.

Professionals who intend to lead independent methodological research may wish to compare shorter doctoral pathways in healthcare administration, while verifying that the degree's research training fits that aim. For leaders who mainly need to use evidence responsibly, collaboration with a qualified analyst and targeted methods coursework may be the more direct investment.

Other Things You Should Know About Healthcare Administration

Do healthcare administrators need to write statistical code?

Not always. Leaders who commission evaluations should be able to define the policy question, understand the study's limits, and work with analysts. Jobs that conduct RD analyses generally require hands-on data and coding skills.

Can patient privacy rules prevent a health services research project?

They can affect access and procedures. Before requesting patient-level data, confirm the organization's privacy, data-use, and institutional review requirements; a promising research design does not itself authorize data access.

Is a healthcare administration doctorate required to publish research?

No. Publication depends on the quality of the work and the journal's standards, not possession of a particular degree. Advanced training and experienced collaborators can nevertheless help with study design and review.

Should an administrator learn one statistical method or broader evaluation skills?

Build broader skills first: framing questions, understanding data quality, assessing comparison groups, and interpreting uncertainty. Learn RD in depth when threshold-based policies are relevant to your work.