Table 2.

Best performance scenarios versus best cost-effective scenarios under different prevalence

Prevalence	Scenarios	Sen.	Spe.	PPV	NPV	Increase in FP	Decrease in FN	Cost in 251,535 population (million US$)	Incre. Cost in 251,535 population (million US$)	Effect (QALYs in 251,535 population)	Incre. Effect (QALYs in 251,535 population)
4%	Status quo	0.933	0.877	0.240	0.997	-	-	1487	-	2,321,412	-
	BCES	0.958	0.824	0.185	0.998	12,756	250	1498	11	2,322,073	662
8%	Status quo	0.933	0.877	0.397	0.993	-	-	1575	-	2,303,843	-
	BCES	0.963	0.804	0.299	0.996	16,909	604	1590	14.7	2,304,685	842
12%	Status quo	0.933	0.877	0.508	0.990	-	-	1664	-	2,286,274	-
	BCES	0.966	0.792	0.387	0.994	18,900	1,028	1680	16	2,287,248	974

The status quo represents the scenario based on the best model performance with the highest area under the curve. Costs and effects were estimated for the population of the Lifeline Express. The willingness-to-pay level was determined as 3 times per-capita GDP (US$ 30,828). Under each prevalence of referable diabetic retinopathy, the status quo was cost-effective. Compared to the status quo, BCES required for higher sensitivity (i.e., lower specificity), showed higher NPV but lower PPV, leading to increased FP cases and decreased FN cases, gaining an extra 662-974 QALYs with an additional 11–16 million costs in the population of Lifeline Express.

Sen. sensitivity, Spe. specificity, PPV positive predictive value, NPV negative predictive value, FP false positive, FN false negative, QALY quality-adjusted life-year, BCES best cost-effective scenario, GDP gross domestic product.