OBJECTIVES: Patient reported quality of life prQoL is rarely measured in the clinical setting. Proxy assessments of QoL are vulnerable to clinician bias, but when patient reports are difficult to obtain, proxies may be sought to inform economic analyses. This study compares clinician estimates with patient reported QoL and examines the impact of experience, profession, and how long the patient was known to the clinician.

METHODS: Seventy-five advanced heart failure patients undergoing frailty assessment, completed the EQ5D-5L. 39 clinicians (11 medical, 23 nursing and 5 allied health) completed the proxy version (V1) within 72 hours of patient assessment- producing 194 pairs. QoL responses were scored using Australian utility values. Bland Altman analyses examined bias and confidence intervals, and Spearman rank coefficients were used to assess correlation over utility, VAS and domain responses. The absolute difference between responses were compared across clinician variables using Mann Whitney, and Kruskal-Wallis tests. Minimum important difference was taken at 0.05 and 7 on the utility and VAS scales respectively.

RESULTS: Clinician EQ5D scores were moderately correlated (r=0.38), but QoL was considered higher than patient report at 0.60 vs 0.54 (p=0.008), with the same bias apparent in 3/5 of the domains and wide confidence intervals exceeding minimum important difference. No significant improvements in concordance was found with years of experience , profession, nor how long the patient was known to the clinician – however individual clinicians did perform better than others with significant heterogeneity apparent amongst the 39 participants.

CONCLUSIONS: Discordance between clinician estimates and patient reported QoL suggests that caution is required when considering proxy QoL assessments in this population. Factors identifying excellent proxies might be better elicited in qualitative investigation of factors influencing clinician responses.