TABLE 1.

Overview of the three most frequently identified risk stratification models with their characteristics and diagnostic properties for different outcomes

	First author, year	Adjusted Clinical Group (ACG)	Charlson Comorbidity Index (CCI)	Hierarchical Condition Categories (HCC)
Categories		ACG categories (1‐93), Resource Utilization Bands (RUBs), Expended Diagnosis Clusters (EDC) count	Six categories based on chronic condition count	Score based on aggregated conditions (70 categories)
Total number of studies in which the model was applied		n = 23	n = 19	n = 7
Diagnostic properties for different outcomes:
Hospitalization	Haas 4	C = 0.73	C = 0.68	C = 0.67
	Lemke 12	AUC = 0.80	AUC = 0.78
(Number of hospitalizations)	Shadmi 16	R 2 = .24	R 2 = .11
(Unplanned hospitalizations)	Maltenfort 11	AUC = 0.82
	Inouye 20		C = 0.72
	Ou 21		C = 0.61
	Mosley 25			AUC = 0.64
Emergency department visits	Haas 4	C = 0.67	C = 0.59	C = 0.58
	Ou 21		C = 0.63
	Wallace 22		C = 0.58
Costs
(Top 10% total costs)	Haas 4	C = 0.76	C = 0.70	C = 0.70
(Total costs)	Brilleman 14	R 2 = .41	R 2 = .34
(Pharmaceutical costs)	Aguado 10	R 2 = .39
(Total costs)	Sicras‐Mainar 13	R 2 = .37
(Total costs)	Charlson 18		R 2 = .22
(Total costs)	Charlson 19		R 2 = .20
(High total costs)	Ou 21		C = 0.64
Utilization of different healthcare services
(GP visits)	Brilleman 15	R 2 = .37	R 2 = .26
(Primary care visits)	Shadmi 16	R 2 = .54	R 2 = .18
(Specialist visits)	Shadmi 16	R 2 = .45	R 2 = .13
(Number of diagnostic imaging tests)	Shadmi 16	R 2 = .37	R 2 = .15
(Visits)	Sicras‐Mainar 13	R 2 = .42
(Number of diagnoses/reasons for visit)	Sicras‐Mainar 13	R 2 = .77
(High outpatient visits)	Ou 21		C = 0.63
Input data for the model		Age, gender, diagnostic codes, pharmaceutical information, healthcare costs	Presence or absence of chronic conditions based on diagnosis codes; weighted	ICD‐9 of ICD‐10 diagnosis codes

Abbreviations: AUC, area under the ROC curve; C, C‐statistic; R², R‐square.