Causal inference methods to study nonrandomized, preexisting development interventions.

Area of Science:

• Global Health
• Development Economics
• Epidemiology

Background:

• Community-level interventions are vital for global health and development.
• Community-randomized trials are the gold standard but face challenges in external validity and sustainability.
• There is a need for rigorous observational methods to assess the impact of widespread, nonrandomized interventions.

Purpose of the Study:

• To introduce and illustrate a matched cohort design for estimating the impact and sustainability of nonrandomized, preexisting community interventions.
• To apply this design to a sanitation, water, and hygiene intervention in rural India.
• To demonstrate the utility of matched cohort designs in development efforts.

Main Methods:

• Utilized the potential outcomes model for causal inference to motivate the matched cohort design.
• Applied the design to a sanitation mobilization, water supply, and hygiene intervention in rural India.
• Enrolled 1,284 children under 5 years old in a matched sample of 25 villages, measuring outcomes over 12 months.

Main Results:

• Observed a significant 33 percentage point increase in new toilet construction (95% CI = 28%, 39%).
• Found no statistically significant impact on height-for-age Z scores (adjusted difference = 0.01, 95% CI = -0.15, 0.19).
• Found no statistically significant impact on diarrhea prevalence (adjusted longitudinal prevalence difference = 0.003, 95% CI = -0.001, 0.008).

Conclusions:

• Matched cohort designs can effectively estimate impacts from widely used, nonrandomized interventions in development settings.
• While demonstrating increased sanitation infrastructure, the intervention showed no measurable impact on child nutritional status or diarrhea.
• Causal interpretation of impacts from matched cohort designs requires stronger assumptions compared to prospective, randomized studies.