The Magnitude and Variability of Stroke Patient-Proxy Disagreement across Multiple Health Domains

Abstract Objectives To quantify the extent and variability of bias introduced when caregivers, or proxies, complete patient-reported outcome measures (PROM) on behalf of stroke patients. Design Cross-sectional survey study conducted between July 2018 and November 2019. Setting Ambulatory clinic of a cerebrovascular center or rehabilitation unit. Participants A consecutive sample of stroke patients (N=200) and their proxies who were able and willing to complete PROMs. Proxies completed PROMs as they believed the patient would answer. Interventions Not applicable. Main Outcome Measures PROMs included Neuro-QoL cognitive function, PROMIS physical function, social role satisfaction, anxiety, fatigue, pain interference, sleep disturbance, Patient Health Questionnaire-9 translated to PROMIS Depression, and PROMIS Global Health. Results The study included 200 stroke patients (age, 62.2±13.3; 41.5% women) and their proxies (age 56.5±13.9; 70% women, 72% spouses). Proxies reported worse functioning and more symptoms across all PROM domains compared with patients (average difference, 0.3-3.0 T score points). Reliability between dyad responses was moderate across all domains (intraclass correlation coefficients (2,1), 0.49-0.76) and effect sizes were small (d=0.04-0.35). Cognitive function, anxiety, and depression had the lowest agreement, whereas physical function, pain, and sleep had the highest agreement based on the Bland-Altman method. At the individual level, a large proportion of dyads had meaningfully different scores across domains (range, 40%-57%; dyads differed >5 T score points). Few predictors of disagreement were identified through multinomial regression models. Conclusions At the aggregate level, small differences were detected between stroke patient-proxy pairs, with lower agreement on more subjective domains. At the individual level, a large proportion of dyads reported meaningfully different scores on all domains, affecting the interpretability of proxy responses on PROMs in a clinical setting.