Categorical state sequence analysis and regression tree to identify determinants of care trajectory in chronic disease: Example of end-stage renal disease

Abstract

Background

Patients with chronic diseases, like patients with end-stage renal disease (ESRD), have long history of care driven by multiple determinants (medical, social, economic, etc.). Although in most epidemiological studies, analyses of health care determinants are computed on single health care events using classical multivariate statistical regression methods. Only few studies have integrated the concept of treatment trajectories as a whole and studied their determinants.

Methods

All 18- to 80-year-old incident ESRD patients who started dialysis in Ile-de-France or Bretagne between 2006 and 2009 and could be followed for a period of 48 months after initiation of a renal replacement therapy were included (n = 5568). Their care trajectories were defined as categorical state sequences. Associations between patients’ characteristics and care trajectories were assessed using a regression tree model together with a discrepancy analysis.

Results

On average, each patient experienced 1.56 different renal replacement therapies (min = 1; max = 5) during the 48 months of follow-up. About 55% of patients never changed treatment and only 1% tried three or more renal replacement therapy modalities. Twelve homogeneous care trajectory groups were identified. Covariates explained 12% of the discrepancy between groups, particularly age, regions and initiation of hemodialysis with a catheter.

Conclusions

Regression tree analysis of categorical state sequence highlighted geographical disparities in the care trajectory of French patients with ESRD that cannot be observed when focusing on a single outcome, such as survival. This method is an original tool to visualize and characterize care trajectories, notably in the context of chronic condition like ESRD.