Reliability of pre-admission patient-reported outcome measures postoperatively assessed via proxies: a prospective, multicenter observational study

Abstract

Pre-admission status obtained through patient-reported outcome measures is an essential metric in both clinical and research settings for prognostication and treatment decisions. It is frequently collected by proxies, although its reliability has yet to be thoroughly investigated. The objective was to determine the reliability of proxy assessments regarding pre-ICU admission status via patient-reported outcome measures and to explore the impact of the ICU setting on these assessments. Prospective multicentre observational study in two tertiary care university hospitals in Germany, including surgical adult patients able to independently answer the patient-reported outcome measures (SF-36, EQ-5D-5L, WHODAS 2.0, IADL, and Barthel Index) with a proxy available. Patients were interviewed pre-operatively, while proxies were interviewed post-operatively in the ICU or normal ward, depending on the patient's location. The reliability of patient-reported outcome measures was analyzed using Bland–Altman plots and Cohen’s kappa. Of 204 patient-proxy pairs, 102 were admitted to an ICU. The median patient and proxy age were 69 and 64 years, with 41% and 68% female, respectively. Bland–Altman plots demonstrated insufficient reliability of proxy ratings, as the 95% limits of agreement fell outside the minimal clinically important difference (MCID) for all questionnaires. However, a significant bias was evident only among ICU patients, showing worse ratings from proxies for the WHODAS 2.0, IADL, and subsets of the SF-36. For the EQ-5D, bias appeared in both the ICU and non-ICU cohorts. The dichotomous analysis of the within-pairs-difference supported the findings, revealing a high proportion of pairs with differences outside the MCID (n (%)—SF-36 PCS normal ward: 47 (46%), ICU: 58 (57%); SF-36 MCS normal ward: 59 (58%), ICU: 62 (61%); WHODAS 2.0 normal ward: 58 (57%), ICU: 78 (77%); EQ-5D-5L normal ward: 40 (39%), ICU: 46 (45%); Barthel-Index normal ward: 22 (22%), ICU: 21 (21%)). Cohen’s kappa indicated moderate reliability for the IADL. The reliability of proxy assessment with the instruments used was insufficient, exhibiting a significant bias in the pre-admission status of ICU patients; therefore, it should be applied with caution. Trial registration: ClinicalTrials.gov (NCT03785444—28th of December 2018).