Reliability of Proxy-reported Patient-reported Outcomes Measurement Information System Physical Function and Pain...

Patient-reported Outcomes Measurement Information System (PROMIS) instruments are useful to evaluate health status, but its use can be challenging for some vulnerable elderly patients, requiring aid from their proxies. Whether the proxies could be accurate informants is unknown. The goal of this study was to compare elderly patients' and their proxies' answers with PROMIS physical function (PF) and pain interference (PI) computer adaptive test for the evaluation of patients' outcomes after musculoskeletal injury. In addition, to correlate patients' reported PF with the Timed Up and Go (TUG) test.

This prospective cohort study, from February to September 2016, in the Orthopaedic trauma clinic of two level I Trauma centers, included 273 patients aged 65 years or older, ambulatory, cognitively intact, with a discernible proxy. PROMIS PF and PI, TUG, and the "FRAIL" Questionnaire screening tool were performed. The correlation of PROMIS scores between patients and proxies, and also with the TUG score, was assessed using Spearman rank correlation. The Bland-Altman analysis served to check agreement and bias. Subgroup comparison was tested using probit transformations.

The mean age of patients was 75.7 years, SD 7.5 (62.2 years; SD, 13.8 for proxies), 66.7% women, 57.1% married, and 34% with femoral fractures. A significant correlation and agreement of PROMIS PF and PI scores were found between patients and proxies (Spearman rho for both, PF and PI = 0.73), although proxies tended to overestimate the interference of pain on patient's performance (median difference, -1.7; P < 0.001). The correlation was markedly stronger in nonfrail patients and in those with faster TUG scores. There was also a correlation between patients' PROMIS PF and TUG test (Spearman rho = - 0.58).

Proxies are good informants of the PF of ambulatory, cognitively intact elderly patients, as evaluated by the PROMIS PF instrument, after musculoskeletal injury, although they tend to slightly overestimate PI. The use of proxy-reported PROs might better characterize functional impairment and pain in a vulnerable patient population, and it could decrease selection bias in outcomes research.

Diagnostic level II.