Inter-rater reliability in performance status assessment among healthcare professionals: an updated systematic review and meta-analysis

Survival prediction for patients with incurable malignancies is invaluable information during end-of-life discussions, as it helps the healthcare team to appropriately recommend treatment options and consider hospice enrolment. Assessment of performance status may differ between different healthcare professionals (HCPs), which could have implications in predicting prognosis. The aim of this systematic review and meta-analysis is to update a prior systematic review with recent articles, as well as conduct a meta-analysis to quantitatively compare performance status scores.

A literature search was carried out in Ovid MEDLINE, Embase, and Cochrane Central Register of Controlled Trials, from the earliest date until the first week of August 2019. Studies were included if they reported on (1) Karnofsky Performance Status (KPS), Eastern Cooperative Oncology Group (ECOG) Performance Status, and/or Palliative Performance Scale (PPS) and (2) assessment of performance status by multiple HCPs for the same patient sets. The concordance statistics (Kappa, Krippendorff’s alpha, Kendall correlation, Spearman rank correlation, Pearson correlation) were extracted into a summary table for narrative review, and Pearson correlation coefficients were calculated for each study and meta-analyzed with a random effects analysis model. Analyses were conducted using Comprehensive Meta-Analysis (Version 3) by Biostat.

Fourteen articles were included, with a cumulative sample size of 2808 patients. The Pearson correlation coefficient was 0.787 (95% CI: 0.661, 0.870) for KPS, 0.749 (95% CI: 0.716, 0.779) for PPS, and 0.705 (95% CI: 0.536, 0.819) for ECOG. Four studies compared different tools head-to-head; KPS was favored in three studies. The quality of evidence was moderate, as determined by the GRADE tool.

The meta-analysis’s Pearson correlation coefficient ranged from 0.705 to 0.787; there is notable correlation of performance status scores, with no one tool statistically superior to others. KPS is, however, descriptively better and favored in head-to-head trials. Future studies could now examine the accuracy of KPS assessment in prognostication and focus on model-building around KPS.