Comparison of alternative risk adjustment measures for predictive modeling: high risk patient case finding using Taiwan's National Health Insurance claims

Table 6 Overlap of top predicted groups by different risk adjustment models

	Model 1: Demographics	Model 2: ACGs & Demographics	Model 3: ADGs, Sel. EDCs & Demographics	Model 4: 2002 Expenditures & Demographics	Model 5: ADGs, Sel. EDCs, 2002 Expenditures & Demographics
2003 Top 0.5% Predicted User
Demographics	0.49%*	0.10%	0.03%	0.03%	0.02%
ACGs & Demographics		0.50%	0.07%	0.08%	0.06%
ADGs, 19 EDCs & Demographics			0.50%	0.19%	0.29%
2002 Expenditures & Demographics				0.50%	0.36%
ADGs, 19 EDCs, 2002 Expenditures & Demographics					0.50%
2003 Top 1% Predicted User
Demographics	1.00%	0.21%	0.11%	0.15%	0.11%
ACGs & Demographics		1.00%	0.30%	0.30%	0.32%
ADGs, 19 EDCs & Demographics			1.00%	0.40%	0.65%
2002 Expenditures & Demographics				1.00%	0.69%
ADGs, 19 EDCs, 2002 Expenditures & Demographics					1.00%
2003 Top 5% Predicted User
Demographics	4.96%*	2.43%	2.09%	2.46%	2.01%
ACGs & Demographics		5.00%	3.43%	3.02%	3.21%
ADGs, 19 EDCs & Demographics			5.00%	3.09%	4.06%
2002 Expenditures & Demographics				5.00%	3.73%
ADGs, 19 EDCs, 2002 Expenditures & Demographics					5.00%

*: There were several observations with exactly the same predicted value at the threshold.

ISSN: 1472-6963