Table 3.

Performance of ZERO-G Estimator on 99 simulated datasets.

	Reported incidence	ZERO-G estimated incidence	% improvement
NRMSE	1.14 (1.032–1.216)	0.778 (0.621–1.02)	31.1% (16.1–36.8%)
Correlation with true incidence rates (Spearman’s ρ)	0.426 (0.35–0.488)	0.657 (0.448–0.746)	44.2% (29.1–54%)
Geographic Bias	0.863 (0.774–0.931)	0.512 (0.315–0.671)	40.5% (23.5–58.2%)
Financial Bias	0.557 (0.237–2.336)	0.156 (0.056–781)	66.8% (33.4–84.6%)

Geographic bias is the absolute correlation between $d_{\mathit{ij}}$ and the incidence rates, relative to the bias observed in the true data (0 signifies no bias). Financial bias is the ratio of incidence rates in zones receiving fee reimbursement to incidence rates in zones not benefiting from fee reimbursement, relative to the ratio observed in the true data (0 signifies no bias). Median values are reported with IQR in parentheses.