External validation of A Web-Based Prognostic Tool for Predicting Survival for Patients in Hospice Care. Issue 3 (September 2013)

• Record Type: Journal Article
• Authors: Miladinovic, Branko; Mhaskar, Rahul; Kumar, Ambuj; Kim, Sehwan; Schonwetter, Ronald; Djulbegovic, Benjamin
• Is Part Of: Journal of palliative care. Volume 29:Issue 3(2013)
• Page Start: 140
• Page End: 146
• Publication Date: 2013-09
• Subjects: Prognostat -- survival prognostication -- Palliative Performance Scale -- external validation -- Royston-Parmar survival models; Terminal care -- Periodicals; 362.17505
• DOI: 10.1177/082585971302900302
• Languages: English
• ISSNs: 0825-8597

Abstract

Prognostat is an interactive Web-based prognostic tool for estimating hospice patient survival based on a patient's Palliative Performance Scale (PPS) score, age, gender, and cancer status. The tool was developed using data from 5, 893 palliative care patients, which was collected at the Victoria Hospice in Victoria, British Columbia, Canada, beginning in 1994. This study externally validates Prognostat with a retrospective cohort of 590 hospice patients at LifePath Hospice and Palliative Care in Florida, USA. The criteria used to evaluate the prognostic performance were the Brier score, area under the receiver operating curve, discrimination slope, and Hosmer-Lemeshow good-ness-of-fit test. Though the Kaplan-Meier curves show each PPS level to be distinct and significantly different, the findings reveal low agreement between observed survival in our cohort of patients and survival predicted by the prognostic tool. Before developing a new prognostic model, researchers are encouraged to update survival estimates obtained using Prognostat with the information from their cohort of patients. If it is to be useful to patients and clinicians, Prognostat needs to explicitly report patient risk scores and estimates of baseline survival.