Methods/Statistics
The impact of vision center establishment on patient visits and cataract surgical volume: a multi-country staggered adoption difference-in-differences analysis Brittany Peterson* Brittany Peterson Peterson Peterson Peterson Peterson Francis I. Proctor Foundation, University of California, San Francisco, USA
Visual impairment impacts millions globally, with a disproportionate burden in low- and middle-income countries (LMICs). Vision centers (VCs) have been established to expand access to eye care for individuals living in remote communities, yet rigorous evidence on their impact is limited. Evaluating such interventions is complicated by staggered implementation across hospitals. We used staggered difference-in-differences methods to estimate the impact of VC establishment on patient visits and cataract surgical volume across eye hospitals in 10 LMICs.
We analyzed hospital-level quarterly data from 41 hospitals between January 2020 and April 2024. As a benchmark, we estimated two-way fixed effects models and used the Goodman–Bacon decomposition to assess potential bias arising from staggered adoption. Our primary analysis employed the Callaway and Sant’Anna (C&S) group-time difference-in-differences estimator, which is robust to staggered adoption and treatment effect heterogeneity. Parallel trends were assessed, and comparisons were restricted to pre-treatment periods that did not exhibit differential trends. We additionally examined heterogeneity by geographic region.
Using the C&S estimator, VC establishment was associated with a modest, non-significant increase in average number of cataract surgeries per quarter (ATT=31.9; 95% CI -32.4, 96.2; p=0.31) and a significant increase in average number of patients per quarter (ATT=723.1; 95% CI 272.2, 1,173.5; p=0.029). Region-stratified analyses suggested larger effects in hospitals located in India and South/Central America.
In this staggered adoption setting, VC establishment was associated with increased patient volume. Results suggested a potential effect on cataract surgical volume as well, although the same size was not large enough to detect this. These findings underscore the importance of applying modern difference-in-differences methods when evaluating health system interventions implemented at different times.