COMPARATIVE EVALUATION OF THE PRENTICE WILLIAMS PETERSON METHOD AGAINST COX PH IN RECURRENT STROKE SURVIVAL MODELLING

Atika Nur Royyanah; Jerry Dwi Trijoyo Purnomo; Shofi Andari

doi:10.53806/jmscowa.v7i1.1550

Authors

Atika Nur Royyanah Department of Statistics, Institut Tekonologi Sepuluh Nopember, Surabaya, 60111, Indonesia
Jerry Dwi Trijoyo Purnomo Department of Statistics, Institut Tekonologi Sepuluh Nopember, Surabaya, 60111, Indonesia
Shofi Andari Department of Statistics, Institut Tekonologi Sepuluh Nopember, Surabaya, 60111, Indonesia

DOI:

https://doi.org/10.53806/jmscowa.v7i1.1550

Keywords:

Cumulative Hazard Curve; PWP Model; Recurrent Events Stroke.

Abstract

The global burden of recurrent stroke, reflected in increased mortality, long-term disability, and healthcare costs, demands accurate risk prediction. This study compares the effectiveness of the Prentice Williams-Peterson (PWP) models with the standard Cox Proportional Hazards model in analyzing recurrent stroke events, using 10-month data from 353 patients with a total of 429 recurrent events at a hospital in Surabaya. The objective is to identify which framework better captures the risk of sequential events based on information criteria. The PWP-Total Time (PWP-TT) model proved to be the most optimal approach, yielding an AIC of 4187.686 (compared to 5042.718 for Cox PH and 4252.759 for PWP-GT) and a BIC of 4219.691 (compared to 4915.213 for Cox PH and 4285.292 for PWP-GT). Based on the PWPTT as the best-fitting model, the smoking variable was found to influence the response variable, namely the time until a patient experiences a stroke.