A Markov multistate analysis of the relationship between performance status and death among an ambulatory population of cancer patients

Background — The relationship between performance status and death among cancer patients has been of increasing interest over the past years. However, few studies have implemented statistical models that adequately capture the longitudinal nature of performance status assessments collected under intermittent observation.

Aim — The main research aims were to examine the association between performance status and death and to determine the probability of deterioration in performance status over time.

Design — This was a population-based longitudinal study among adult outpatients diagnosed with cancer. Throughout their observation period, all patients were repeatedly assessed for performance status using an 11-point scale with a score of 100 being the best, 10 being the worst and 0 representing death. A Markov multistate model accounting for intermittent observation was implemented in which each score represented a distinct state in the model.

Results — The cohort consisted of 27,739 patients with over 157,000 assessments. The rate of transition to death increases with a quadratic trend as performance status declines. The 1-month and 3-month probability of deterioration also increases with a quadratic trend as performance status declines. The relative rate of transition to death decreases as we compare lower scores (relative rate = 2.20 for comparing scores 90 vs 100 and relative rate = 1.23 for comparing scores 10 vs 20).

Conclusion — There is a significant relationship between performance status and rate of transition to death. The Markov multistate model provides a comprehensive understanding of the shape of this relationship, which facilitates the interpretation of performance status and provides strength in its use as a prognostic tool in a clinical setting.