Table 2.

Model Fit Statistics

Model	Bayesian Information Criteriaa	Log-Likelihoodb	Number of Parametersc	Entropyd	Patients per Class
Edinburgh cohort
					1	2	3	4	5	6	7
1-class	14 578.4	−7200.3	29	1	458	…	…	…	…	…	…
2-class	14 167.1	−6902.8	59	0.8	307	151	…	…	…	…	…
3-class	13 968.5	−6711.6	89	0.8	173	152	133	…	…	…	…
4-class	13 901.7	−6586.3	119	0.8	162	128	127	41	…	…	…
5-class	13 829.1	−6458.1	149	0.9	147	132	103	39	37	…	…
6-class	13 869.1	−6383.2	149	0.9	130	123	83	42	42	38	…
7-class	13 980.5	−6350.0	209	0.9	113	103	65	57	42	41	37
ARREST cohort
					1	2	3	4	5	6	7
1-class	19 279.6	−10 157.3	26	1	758	…	…	…	…	…	…
2-class	18 636.4	−9741.5	53	0.8	576	182	…	…	…	…	…
3-class	18 360.2	−9515.8	80	0.8	410	232	116	…	…	…	…
4-class	18 284.9	−9374.1	107	0.8	284	197	162	115	…	…	–
5-class	18 269.8	−9259.3	134	0.8	276	139	121	115	107	…	…
6-class	19 409.7	−9171.1	161	0.8	171	153	119	116	106	93	…
7-class	19 462.4	−9107.9	188	0.8	172	131	117	112	88	78	60
SAFO cohort
					1	2	3	4	5	6	7
1-class	2065.0	−976.1	21	1	214	…	…	…	…	…	…
2-class	1963.3	−866.3	43	0.8	125	89	…	…	…	…	…
3-class	1945.9	−798.6	65	0.9	125	67	22	…	…	…	…
4-class	1981.6	−757.4	87	0.8	79	71	44	20	…	…	…
5-class	2035.2	−725.1	109	0.9	59	54	45	36	20	…	…
6-class	2121.1	−709.1	131	0.9	58	53	38	34	18	13	…
7-class	2174.1	−676.5	153	0.9	51	40	36	37	23	20	7

Model fit statistics for latent class models from 1 to 7 classes in the Edinburgh, ARREST, and SAFO cohorts.

^aDefined in Methods section.

^bMeasures the fit of the model to the data.

^cMeasure of model complexity.

^dMeasure for class separation; it ranges from zero to 1; values ≥0.8 indicate good separation of the different classes.