Guide to recent advances in difference-in-differences methodology for population health studies

Area of Science:

• Epidemiology
• Econometrics
• Health Services Research

Background:

• Difference-in-differences (DiD) is a standard method in population health research.
• Classic DiD assumes two groups (exposed/unexposed) and two periods (before/after).
• Real-world scenarios often involve multiple groups, periods, or varied exposure timing, complicating standard DiD.

Purpose of the Study:

• To explain the challenges of applying difference-in-differences methodology in complex, real-world settings.
• To summarize alternative methods addressing deviations from classic DiD assumptions.
• To provide guidance on selecting appropriate estimators for heterogeneous treatment effects.

Main Methods:

• Review and explanation of challenges in applying difference-in-differences in non-standard situations.
• Summary of proposed alternative statistical methods for complex DiD scenarios.
• Discussion of considerations for choosing between different difference-in-differences estimators.

Main Results:

• Standard difference-in-differences (DiD) reasoning is insufficient when treatment effects vary across groups or over time (heterogeneous treatment effects).
• Implicit assumptions in typical DiD methods for aggregating multiple comparisons can be misleading.
• Alternative methods explicitly address the aggregation of heterogeneous treatment effects.

Conclusions:

• Real-world difference-in-differences (DiD) analyses must account for time-varying and group-specific exposure effects.
• Assessing overall impact requires explicit weighting of heterogeneous effects.
• The choice of DiD approach depends on exposure assignment and data availability before and after exposure.