Materials and Methods
Study Design, Setting, and Subjects
A retrospective observational study was conducted of patients aged 18+ who received spinal surgery between November 1, 2018 and August 1, 2019 within a tertiary care institution; institutional IRB approval was obtained for all use of PHI. Patients with a self-reported visual analog scale (VAS) pain score and PROMIS global physical health (GPH) and global mental health (GMH) scores at the time of presentation were included. To maintain sufficient power, the five most common intracohort spinal procedures were retained: 1–2 segment cervical decompression/fusion (d/f), cervical or thoracic laminotomy, lumbar laminotomy, lumbar interbody fusions (LIF), and arthrodesis for spinal deformity of less than or equal to eight segments.
Primary (GPH score) and secondary (GMH, VAS scores) outcome variables were calculated at presentation, postoperative day (POD) 90 (range: 75–106), and 12 months (M) postoperatively (range: 10–13 months). GPH and GMH are summary scores generated from a 10-question self-administered PROMIS Global Health survey, in which patients rank attributes of their wellbeing on ordinal scales of 1 (poor) to 5 (excellent). GPH is the sum of scores from questions relating to physical health/functioning, pain intensity, and fatigue; GMH comprises questions relating to overall quality of life (QOL), mental health, social satisfaction, and emotional problems. Higher scores indicate better health-related QOL. VAS is a graphical representation of a patient's subjective pain experience; patients denote pain intensity on a 100 mm line connecting pain extremes ("no pain" to "unbearable pain"), which is converted to a 1 to 10 scale.
Patient-specific data included preoperative health (Charleson comorbidity score) and demographic characteristics (age at surgery, sex [male or female], and self-identified race). Due to underrepresentation of some races, race was categorized as White, Black/African American (AA), and "Other/Non-White Minorities."
Patient Community Characteristics Potentially Contributing to Physical and Mental Health
Patient addresses were linked to census tracts, zip-codes, and counties using the Geographic Information Systems (ArcGIS) mapping program. The behavioral risk factor surveillance system—a publicly available survey of US adults regarding health-related risk behaviors, chronic conditions, and preventative service utilization—was used to calculate the percentage of a patient's county that reported: (1) regular physical activity; (2) physical activity meeting CDC recommendations; (3) low general health; (4) chronic physical distress; (5) chronic mental distress; and (6) chronic infrastructural barriers to physical activity. The ADI, which indicates census tract-level disadvantage, was constructed using 17 area-level US Census indicators of poverty, education, employment, and physical environment from American Community Survey (ACS) data; scores were ranked into deciles based on national distributions.[23,24] ACS data was used to deduce regional 5-year means (2014–2019) of median household incomes/values, local unemployment, proportion of families under the federal poverty line (FPL), and percentages of families with six figure household income.
Chi-square analysis, one-way analysis of variance (ANOVA) followed by Bonferroni-corrected post-hoc t tests, and Kruskal–Wallis analyses followed by Dunn post-hoc nonparametric test for multiple comparisons were used to determine if race, community SES, and procedure affected the degree of physical impairment, mental distress, or overall pain, at presentation. To assess if differences in PROs were clinically relevant, minimum clinically important differences (MCIDs) were calculated for each PRO as 1/2 of the sample standard deviation based on presenting means and stratified by procedure type (Supplemental Table 1, https://links.lww.com/BRS/B857). Univariate and multivariate linear regression models with stepwise selection (α = 0.05) were constructed of GPH on demographic, socioeconomic, and clinical characteristics. These models adjusted for potential confounding by first incorporating the effect of spinal pathology on presenting PROMIS scores. For all regressions, variance inflation factors (VIF) were calculated to assess multicollinearity; predictors with VIF is less than 10 were retained. Statistical significance is defined by two-sided test with a P < 0.05. Analyses were performed using SAS (SAS 9.4, SAS Institute Inc., Cary, NC) and GraphPad Prism (v8.4.2) (GraphPad Software, La Jolla, CA, www.graphpad.com).
Spine. 2022;47(11):781-791. © 2022 Lippincott Williams & Wilkins