Karnofsky performance status revisited: reliability, validity, and guidelines.

TLDR

The Karnofsky Performance Status (KPS) scale lacks documented reliability and validity, and no empirical guidelines exist for its use. The study surveyed 293 cancer patients, had physicians rate their KPS, evaluated interrater reliability with 75 patients, and used regression to derive seven behaviorally based questions that guide more accurate KPS ratings. The KPS demonstrated good reliability and validity, and the proposed seven‑variable guideline improves rating accuracy, enhancing reliability and validity for clinical and research use.

Abstract

Little research has been conducted documenting the reliability and validity of the Karnofsky Performance Status (KPS) scale, and guidelines based on empirical data do not exist to govern its use. Two hundred ninety-three cancer patients completed a questionnaire that assesses their physical and psychosocial difficulties. Physicians rated patients on the KPS and a subsample of 75 patients was used to evaluate interrater reliability. Analyses were conducted to evaluate the interrater reliability and construct validity of the KPS. The KPS was shown to have good reliability and validity. Detailed examination of the reliability data suggested areas in which physicians err in their judgments. Multiple regression techniques were used to empirically identify seven behaviorally based questions that would be helpful in predicting KPS scores. The seven variables included weight loss, weight gain, reduced energy, difficulty walking, driving, grooming, and working part time. An interview approach with behaviorally based guidelines is presented using these variables to obtain relevant data and make more accurate KPS ratings. With the approach suggested and the guidelines presented, oncologists may train themselves to use the KPS in a standard way, which should increase reliability and validity of the KPS and has implications for patients and research studies that use KPS as a stratifying variable.