Markov modelling of viral load adjusting for CD4 orthogonal variable and multivariate conditional autoregressive mapping of the HIV immunological outcomes among ART patients in Zimbabwe

Table 2 Model selection process for the various model fitted for both viral load and CD4 cell counts multistate models

Covariate	−2*LL^a	AIC^b	Models compared	LRT ^c statistic	Df^d	P-value
Viral load multistate model
Model 1- No covariates	7074.605		–	–	–	–
Model 2- CD4 effects	7014.406	7062.99	Model 1 vs Model 2	59.61457	12	2.65e-08^**
Model 3- CD4 time-varying effects	7007.724	7055.72	Model 1 vs Model 3	66.88163	12	1.22e-09^**
Model 4- CD4 effects+ CD4 time-varying effects	6930.714	7002.71	Model 2 vs Model 4	83.69604	12	8.10e-13^**
Model 4- CD4 effects+ CD4 time-varying effects	6930.714	7002.71	Model 3 vs Model 4	77.01372	12	1.53e-11^**
CD4 cell counts multistate model
Model 1- No covariates	7334.08		–	–	–	–
Model 2- VL effects	7298.644	7342.64	Model 1 vs Model 2	35.36379	11	2.16e-04^**
Model 3- VL time-varying effects	7256.166	7344.17	Model 1 vs Model 3	77.84109	33	1.74e-05^**
Model 4- VL effects + VLtime-varying effects	7227.461	7337.46	Model 2 vs Model 4	71.18325	33	1.28e-04^**
Model 4- VL effects + VLtime-varying effects	7227.461	7337.46	Model 3 vs Model 4	28.70595	11	2.52e-03^**

^aLog-likelihood ^bAkaike’s information criterion ^cLikelihood ratio test ^ddegrees of freedom ^**Significant at 5%

ISSN: 1742-4682