Using large administrative data for mining patients' trajectories for risk stratification: An example from urological...

Area of Science:

• Urology
• Health Services Research
• Biostatistics

Background:

• Urological patients exhibit diverse hospitalisation patterns.
• Understanding these trajectories is crucial for improving patient outcomes and resource allocation.

Purpose of the Study:

• To identify distinct clusters of urological patient hospitalisation trajectories.
• To associate these clusters with baseline risk factors and quality indicators.

Main Methods:

• Utilised the Victorian Admitted Episodes Dataset (2009-2019) for 98,782 urological patients.
• Employed Latent Class Trajectory Modelling (LCTM) to define hospitalisation clusters.
• Used logistic regression to identify predictive baseline factors, including comorbidities and diagnoses.

Main Results:

• Identified five distinct hospitalisation trajectory clusters.
• Higher hospitalisation clusters correlated with increased length of stay, readmissions, and complications.
• Comorbidities (renal disease, diabetes) and specific diagnoses (urological cancers, benign prostatic hyperplasia) were associated with higher-risk trajectories.

Conclusions:

• A novel LCTM approach effectively clusters urological patient hospitalisation risks.
• Baseline comorbidities and diagnoses are significant predictors of higher hospitalisation rates and adverse outcomes.
• This methodology enables proactive identification of high-risk patients for targeted interventions.