DOI: 10.1097/ao9.0000000000000027 ISSN: 3068-4129

Preoperative Patient-Reported Outcomes Measurement Information System Physical Function Score and Postoperative Outcomes in Noncardiac Surgery

Heather L. Lander, Judith F. Baumhauer, David Cella, Marjorie S. Gloff, Yang Gu, Isaac Y. Wu, Julie A. Wyrobek, David J. Mitten, Elizabeth Levatino, Kathleen Fear, Laurent G. Glance
Show PDF Cite

Background:

Preoperative risk assessment is central to achieving optimal surgical outcomes but is resource-intensive. We hypothesized that Patient-Reported Outcomes Measurement Information System Physical Function (PROMIS-PF) scores could serve as a low-resource, patient-completed tool to help identify high-risk surgical patients before surgery.

Methods:

We conducted a retrospective study of adults undergoing noncardiac surgery at a large academic medical center between 2015 and 2023. Multivariable logistic regression was used to examine the association between PROMIS-PF scores and the composite outcome of 30-day mortality and major complications. We evaluated the predictive performance of the PROMIS-PF score alone and compared it with a baseline clinical model, the baseline model plus PROMIS-PF score, and the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) risk calculator using optimism-corrected measures of discrimination and calibration.

Results:

Among 3,399 patients (mean ± SD of age: 57.3 ± 16.1 years; 1,719 [50.6%] were women and 1,680 [49.4%] were men), 152 (4.5%) had death or major complication; 12 died (0.35%); 687 (20.2%) had PROMIS-PF scores of 15 to 30; 2,498 (73.5%) had scores of 31 to 55; and 214 (6.3%) had scores of 56 to 80. Compared with PROMIS-PF scores of 56 to 80, those with scores of 15 to 30 had significantly higher odds of adverse outcomes (adjusted odds ratio [aOR], 3.13; 95% CI: 1.24 to 7.94; P = 0.02), whereas those with scores of 31 to 55 did not differ significantly (aOR, 1.92; 95% CI: 0.84 to 4.38; P = 0.11). The PROMIS-PF score alone demonstrated poor discrimination (C statistic, 0.52) and poor calibration (calibration slope, 0.64; calibration intercept, −1.1). The parsimonious clinical model (without PROMIS-PF score) showed very good discrimination (C statistic, 0.82) and acceptable calibration (calibration slope, 0.93; calibration intercept, −0.19); the addition of PROMIS-PF score to the clinical model did not improve the performance of the clinical model. The ACS NSQIP risk calculator exhibited worse discrimination (C statistic, 0.78) but better calibration (calibration slope: 0.99; calibration intercept: −0.03) than the baseline clinical model. In our secondary analysis, we found that a PROMIS-PF score of 30 or lesser was strongly associated with the ACS-NSQIP risk of 30-day mortality. Compared with patients with a predicted risk of mortality of 0.1% or lesser, patients with a predicted mortality of 5.0 to 10.0% (aOR, 20.6; 95% CI: 7.20 to 58.7; P < 0.001) and greater than 10.0% (aOR, 66.9; 95% CI: 26.8 to 167.1; P < 0.001) had higher odds of PROMIS-PF score of 30 or lesser.

Conclusion:

In this single-center exploratory study, lower preoperative PROMIS-PF scores were associated with higher 30-day mortality and major complications among patients undergoing noncardiac surgery. Although the PROMIS-PF score demonstrated limited predictive value for the composite outcome, it was strongly associated with the ACS NSQIP-predicted risk of 30-day mortality. Further evaluation of low PROMIS-PF scores as a predictor of 30-day mortality in a larger multicenter cohort is warranted.

More from our Archive