Letter to the Editor Regarding “Social Vulnerability Index Is Associated With Increased Emergency Department Utilization and Persistent Opioid Use after Elective Lumbar Decompression Surgery”

Dear Editor,

Plantz et al. draw welcome attention to the relationship between social vulnerability and health care use after lumbar decompression. ¹ Their finding of more opioid prescriptions at 6 months and 1 year is clinically relevant because patient-reported improvement was otherwise similar between groups. We were less certain, however, about what “persistent opioid use” represents in this analysis. The outcome was defined as any new opioid prescription at the follow-up point of interest, but the observation window around 180 or 365 days was not specified, and it is unclear whether a single prescription was sufficient for classification. A prescription near 1 year could represent continued treatment, an intermittent refill, or treatment for an unrelated condition. As reported, the outcome appears to capture prescribing rather than dispensing or actual consumption. Schultz et al restricted their cohort to opioid-naive patients and defined new persistent use as at least 1 prescription fill between 90 and 180 days after surgery. ² Bologheanu et al excluded patients who filled an opioid prescription within 3 months before surgery and required at least 1 fill during both the first 90 postoperative days and days 90 to 180. ³ Prospective work has also assessed patient-reported use and reasons for continued use after spine surgery. ⁴ We therefore wonder whether the authors could clarify the ascertainment window, the number of prescriptions required, and whether the records represented orders, fills, or reported use. These details would help readers distinguish a late recorded prescription from sustained consumption or opioid dependence.

We also noted an apparent inconsistency in the reported effect measure. The Methods describe multivariable logistic regression, whereas the corresponding tables label the estimates as risk ratios. If standard binary logistic regression was performed as reported, exponentiating the coefficients would ordinarily yield odds ratios rather than risk ratios. ⁵ This distinction matters because the 1-year outcome occurred in 25.9% of socially vulnerable patients; the rare-outcome approximation therefore cannot be assumed, and an estimate of 3.315 should not be read directly as a 3.315-fold risk unless a model that estimates risk ratios was fitted. Clarifying the model and link function would indicate how the estimates should be interpreted. Adjusted predicted probabilities and an absolute risk difference may convey the clinical magnitude more directly. If a risk ratio was the intended estimand, a log-binomial model or modified Poisson model with robust variance could be considered as a sensitivity analysis. ⁵ Given the 31 outcome events and the number of reported predictors, describing the candidate covariates and variable-selection procedure would also help readers assess the stability of the adjusted estimate. Clarifying these points would not detract from the reported association and would help readers interpret both the opioid outcome and its association with social vulnerability more accurately.

ORCID iD

Shun Yang https://orcid.org/0009-0008-7322-3914