Modeling factors associated with dialysis adequacy using longitudinal data analysis: Generalized estimating equation versus quadratic inference function

In hemodialysis patients, changes in dialysis adequacy are recorded at regular intervals and studied longitudinally. The aim of this study was to determine the factors affecting dialysis adequacy using the generalized estimating equation (GEE) and to compare them with the quadratic inference function (QIF).

This longitudinal study examined the records of 153 end-stage renal diseases (ESRD) patients. Longitudinal data on the dialysis adequacy index and demographic and clinical characteristics were obtained from the patient files. The first-order GEE (GEE1), second-order GEE (GEE2), and QIF models were fitted with different correlation structures, and then the best correlation structure was selected using the quasi-likelihood information criterion (QIC), Akaike information criterion (AIC), and Bayesian information criterion (BIC). Then, the selected models were compared based on the relative efficiency of the estimated regression coefficients.

The majority of patients (59.5%) had unfavorable dialysis adequacy (KT/V<1.2). Women had more favorable dialysis adequacy than men, and patients <60 years had more favorable dialysis adequacy than older. In the GEE1, GEE2, and QIF models, the coefficients of dialysis history, dialysis duration, weight, gender, and age showed a significant relationship with dialysis adequacy (p<0.05) The relative efficiencies of GEE2 versus GEE1,  and QIF versus GEE1  and GEE2 were 1.163, 1.13, and 1.028, respectively.

Dialysis adequacy is not optimal in most hemodialysis patients. The different models yield quite similar coefficient estimates, but GEE2 with unstructured correlation is more efficient than GEE1, and QIF is more efficient than both GEE1 and GEE2.