Factors that can predict pain with walking, 12 months after total knee arthroplasty

Figures & data

Table 1. Fit indices for interference of pain with walking GMM solutions over 5 assessments

GMM	LL	AIC	BIC	Entropy	VLMR
Pain-related interference with walking
1-Class^a	−2,136	4,290	4,319	n/a	n/a
2-Class^b	−2,105	4,236	4,279	0.84	61^c
3-Class	−2,099	4,232	4,288	0.63	12^d

AIC: Akaike information criterion; BIC: Bayesian information criterion; CFI: comparative fit index; GMM: growth mixture model; LL: log likelihood; n/a: not applicable; RMSEA: root mean square error of approximation; SRMR: standardized root mean square residual; VLMR: Vuong-Lo-Mendell-Rubin likelihood ratio test for K-1 (H0) vs. K classes.

a Latent growth curve with linear and quadratic components; chi² = 49.37, 11 df, p < 0.00005, CFI =0.69, RMSEA =0.13, SRMR =0.13.

b 2-class model was selected. The BIC was smallest for the 2-class model and the VLMR was consistent in indicating that the 2-class solution fitted the data better than the 1-class or 3-class solutions.

c p < 0.001.

d Not significant.

Table 2. GMM parameter estimates for predicted growth mixture model latent classes from 5 assessments of pain-related interference with walking

	Recurrent Interference	Continuous Improvement
Parameter estimates	(n^a = 45) mean (SE)	(n^a = 157) mean (SE)
Intercept	6.96 (0.26)^f	5.63 (0.14)^f
Linear slope	−0.14 (0.05)^e	−0.16 (0.03)^f
Quadratic slope	0.003 (0.001)^e	0.002 (< 0.0005^d)^f
Variances^b
Intercept^c	0.97 (0.19)^e	0.97 (0.19)^e

GMM: growth mixture model; SE: standard error.

a Predicted class sizes based on their most likely class membership.

b Slope variances were fixed at zero to assist in estimation.

c Intercept variances were set equal for the classes to aid in estimation, given the small sample size.

d Standard error too small to be displayed with only 3 decimals in the Mplus output.

e p < 0.01

f p < 0.001

Trajectories of observed and estimated scores for the latent classes of pain-related interference with walking.

Table 5 Differences in symptom-related and psychological characteristics between the Recurrent Interference and Continuous Improvement classes based on an evaluation of pain-related interference with walking

	Recurrent	Continuous
	Interference^a	Improvement^a	Statistics		Effect size
	(n = 45)	(n = 157)	p-value	95% CI	Cohen’s d
Symptoms
Fatigue severity (LFS)	3.4 (2.1)	2.6 (2.1)	0.02	−1.6 to −1.4	0.37
Fatigue interference (FSS7)	4.6 (1.6)	3.8 (1.5)	0.005	−1.3 to −0.2	0.52
Depression (HADS)	4.6 (3.9)	3.2 (2.8)	0.03	−2.7 to −0.1	0.44
Anxiety (HADS)	5.4 (4.1)	4.4 (3.5)	0.1	−2.2 to 0.2	0.28
Psychological characteristics from the brief illness perception questionnaire
Consequences	6.8 (1.6)	6.2 (1.8)	0.04	−1.2 to −0.03	0.34
Personal control	5.9 (2.4)	5.2 (2.4)	0.06	−1.6 to 0.02	0.30
Identity	7.0 (1.6)	6.5 (1.7)	1.0	−1.0 to 0.1	0.30
Concern	6.2 (2.3)	4.7 (2.6)	< 0.001	−2.4 to −0.7	0.76
Emotional response	5.3 (2.5)	4.3 (2.6)	0.02	−1.9 to −0.2	0.38

a Values are mean (SD)

Cohen’s d effect sizes: Small ≥0.2; Medium ≥0.5; Large ≥0.8.

FSS7: fatigue severity scale; HADS: hospital and anxiety scale; LFS: Lee fatigue scale.