Comparison of Rx-defined morbidity groups and diagnosis- based risk adjusters for predicting healthcare costs in Taiwan

Table 4 Predictive models for medication cost

Morbidity Index	Source of morbidity	Predictors	Model performance - prediction of medication cost
			Concurrent (year 2006)			Prospective (year 2007)
			R²	MAPE	MAPE* (%)	R²	MAPE	MAPE* (%)
(none)		Age + gender	0.151	155.2	118.9	0.153	166.0	119.1
Deyo's CCI	Diagnosis	Age + gender + CCIs	0.426	113.5	87.0	0.366	129.6	93.0
Elixhauser's Index	Diagnosis	Age + gender + E. Index	0.514	97.4	74.7	0.434	115.5	82.9
ADG	Diagnosis	Age + gender + ADGs	0.431	114.4	87.7	0.360	131.2	94.2
Rx-MG	Medication	Age + gender + Rx-MGs	0.615	89.6	68.6	0.485	110.3	79.1
ADG + Rx-MG	Diagnosis & medication	Age + gender + ADGs + Rx-MGs	0.638	85.6	65.6	0.505	106.2	76.2
ADG + Rx-MG	Diagnosis & medication	Age + gender + ADGs + Rx-MGs + prior medication cost				0.684	73.8	53.0
ADG + Rx-MG	Diagnosis & medication	Age + gender + ADGs + Rx-MGs + prior medication cost + (prior medication cost)²				0.684	73.7	52.9

ISSN: 1472-6963