Improving equitable collection and analysis of PROMIS Global Health data over time following spine surgery: characterizing survey nonresponse and missing data.

BACKGROUND CONTEXT: Patient-reported outcome measures (PROMs), the gold standard for outcome assessment in spine surgery, exhibit variability over time. Incomplete PROM collection, however, introduces nonresponse bias and limits the generalizability of time-based analyses of outcomes.

PURPOSE: This study compared PROM-respondents and nonrespondents in spine surgery to characterize survey nonresponse and improve equitable patient representation in time series PROM analyses.

STUDY DESIGN: Retrospective study.

PATIENT SAMPLE: Patients undergoing surgery at a large, tertiary care center in the United States between July 2009 and February 2023 for lumbar spinal stenosis without spondylolisthesis (LSS), lumbar spinal stenosis with spondylolisthesis (LSP), or cervical spondylotic myelopathy (CSM).

OUTCOME MEASURES: The primary outcome was completeness of available PROM records, which was defined as having Patient-Reported Outcomes Measurement Information System (PROMIS)-Global Health scores once within 2 years preoperatively and twice within 2 years postoperatively.

METHODS: Demographic variables of age, sex, race, marital status, employment, insurance, body mass index (BMI), smoking, and Area Deprivation Index (ADI) were obtained from the electronic medical record. These characteristics were compared by PROM-completeness within each pathology group. Comparative analyses between the PROM-complete and PROM-incomplete patients within each pathology group were conducted using the Satterthwaite t-test for continuous variables, Pearson's chi-square test for categorical variables, and Mann-Whitney U test for ordinal variables. Among patients with complete PROMs, availability of PROMIS-Global Health within 2 years pre- and postoperatively was plotted in bins of 84- and 168-days width to characterize the distribution of time points represented in PROM data for these patients. To visualize geographic variation in likelihood of representation in time series PROMs analyses, census block-level heatmaps were generated for each pathology group showing predicted probability of PROM-completeness by logistic regression with age, sex, race, marital status, employment status, insurance category, BMI, and smoking status as predictor variables.

RESULTS: About 4,938 patients (1,751 LSS, 1,711 LSP, 1,476 CSM) were analyzed. PROM-complete patients varied significantly from PROM-incomplete patients in demographic distributions. PROM-complete patients were more likely of White race, married, retired, and less likely to be current smokers. LSS and CSM PROM-complete patients were more likely to have Medicare insurance than PROM-incomplete patients.

CONCLUSION: Patients completing PROMs for spine surgery may differ from those who do not, with greater representation of White race, being married, retiree status, and Medicare insurance among those with complete PROMs. As PROMs are further incorporated into physician evaluation, value-based reimbursement, and predictive analytics for surgical outcomes, understanding survey nonresponse will be critical for generating equitable, individualized, and informed applications to support healthcare decisions.