Multimorbidity as a predictor of health service utilization in primary care: a registry-based study of the Catalan population

Table 3 Contributions of multimorbidity measurements on predictive modelling of use of healthcare resources in Primary Care

	A: Frequent attendants			B: Home care needs			C: Social worker needs			D: Polypharmacy patients
	AIC	R²	AUC	AIC	R²	AUC	AIC	R²	AUC	AIC	R²	AUC
Baseline model	3.08	21.5%	0.771	1.38	33.4%	0.862	1.01	20.6%	0.809	3.27	29.9%	0.835
Charlson index	2.83	28.7%	0.808	1.30	37.6%	0.878	0.96	24.8%	0.841	2.78	41.4%	0.880
Number of chronic diseases	2.60	35.3%	0.840	1.27	38.9%	0.886	0.93	27.6%	0.862	2.44	49.1%	0.906
Clinical Risks Groups (CRG)	2.70	32.5%	0.830	1.30	37.5%	0.883	0.95	25.4%	0.851	2.38	50.9%	0.912
Adjusted Morbidity Groups (GMA)	2.49	38.4%	0.853	1.25	40.0%	0.890	0.91	29.3%	0.872	2.41	50.1%	0.910

The table reports the statistics indicating model fitting of the multiple regression analyses carried out to estimate each of the outcome variables (A to D). The first row describes absolute values of the three statistics: AIC: Akaike Information Criterion (in millions); R²: deviance-based R-squared measure; and AUC: Area Under the ROC Curve for predictive models including as covariates: age group, sex, socioeconomic status and all the first order interactions between these variables, but not multimorbidity measurements (Baseline model)
The subsequent rows correspond to the contributions of the four multimorbidity measures to model fitting for each outcome variable (A to D), namely: i) Charlson index; ii) Number of chronic diseases; iii) Clinical Risks Groups (CRG), and, iv) Adjusted Morbidity Groups (GMA)

ISSN: 2731-4553