Context: Although prior studies show the importance of self-reported symptom scores as predictors of cancer survival, most are based on scores recorded at a single point in time.
Objectives: To show that information on repeated assessments of symptom severity improves predictions for risk of death and to use updated symptom information for determining whether worsening of symptom scores is associated with a higher hazard of death.
Methods: This was a province-based longitudinal study of adult outpatients who had a cancer diagnosis and had assessments of symptom severity. We implemented a time-to-death Cox model with a time-varying covariate for each symptom to account for changing symptom scores over time. This model was compared with that using only a time-fixed (baseline) covariate for each symptom. The regression coefficients of each model were derived based on a randomly selected 60% of patients, and then, the predictive performance of each model was assessed via concordance probabilities when applied to the remaining 40% of patients.
Results: This study had 66,112 patients diagnosed with cancer and more than 310,000 assessments of symptoms. The use of repeated assessments of symptom scores improved predictions for risk of death compared with using only baseline symptom scores. Increased pain and fatigue and reduced appetite were the strongest predictors for death.
Conclusion: If available, researchers should consider including changing information on symptom scores, as opposed to only baseline information on symptom scores, when examining hazard of death among patients with cancer. Worsening of pain, fatigue, and appetite may be a flag for impending death.
Keywords: Cox model; Symptom scores; concordance probability; longitudinal data; time-varying covariates.
Copyright © 2014 American Academy of Hospice and Palliative Medicine. Published by Elsevier Inc. All rights reserved.