Recovery Patterns: Longitudinal Cluster Analysis of Physical Function Following Abdominal Surgery.

Abstract

A rise in the proportion of day surgery has seen a concomitant increase in the proportion of patients recovering at home. Blended eHealth is well situated to provide this group with medical support and supervision. However, a data-driven description of the heterogeneity is missing. To identify clinically meaningful patterns of functional recovery following abdominal surgery and describe how the emergent patient characteristics differ between them. This was a secondary data analysis of two datasets collected through two previously conducted RCTs. We used k-medoids clustering and Growth Mixture Modelling on the longitudinal patient reported outcome measurement information system (PROMIS) physical function (PF) t-scores of 649 patients. Differences in patient characteristics between the resultant clusters were identified through statistical tests. Three clusters - fast, intermediate and uneven recovery - were identified regardless of the dataset or statistical technique. A fourth cluster - relapse - was identified by both statistical techniques but only in the presence of heavy surgery. The fifth and sixth clusters - low gain and high gain - were identified for both light and heavy surgery, but only through k-medoids clustering. Trajectories of physical function following abdominal surgery are heterogenous but distinct clinically meaningful patterns can be extracted. This classification may facilitate shared-decision making during pre-operative care and future research may utilize them as targets for prediction.