Table 1

Goodness of fit statistics for models of pain trajectory from latent class growth analysis.

Model	AICLL	BICLL	CAICLL	Test of improvement in model fit compared to model with 1 less clustera
Back pain
3 Cluster	7033	7090	7101	<0.001
4 Cluster	6947	7025	7040	<0.001
5 Cluster	6917	7016	7035	<0.001
6 Cluster	6905	7025	7048	<0.001
7 Cluster	6900	7040	7067	0.01
8 Cluster	6900	7061	7092	0.19
Facial pain
2 Cluster	4314	4351	4358	<0.001
3 Cluster	4277	4334	4345	<0.001
4 Cluster	4251	4329	4344	<0.001
5 Cluster	4246	4345	4364	0.008
6 Cluster	4252	4372	4395	0.58
Headache
2 Cluster	7616	7652	7659	<0.001
3 Cluster	7340	7397	7408	<0.001
4 Cluster	7288	7366	7381	<0.001
5 Cluster	7273	7372	7391	<0.001
6 Cluster	7256	7376	7399	<0.001
Stomach pain
2 Cluster	7003	7039	7046	<0.001
3 Cluster	6885	6943	6954	<0.001
4 Cluster	6834	6912	6927	<0.001
5 Cluster	6822	6921	6940	<0.001
6 Cluster	6822	6942	6965	0.17

AIC, Akaike’s information criterion; BIC, Bayes’ information criteria; CAIC, consistent Akaike’s information criterion; LL, log-likelihood. Optimal models based on goodness of fit statistics shown in bold.