Lin, RT, Ramanathan, R, Dalton, JF, Como CJ, Lee I, Gonzalez C, Tang MY, Oyekan AA, Chang AY, Spitnale M, Shaw JD, Lee JY, & Wawrose, RA. Prior year hospital admission predicts 30-day hospital readmission after spine surgery. North American Spine Society Journal (NASSJ) 2025; DOI: 10.1016/j.xnsj.2025.100750.

This retrospective, single-center cohort study by Lin et al sought to identify predictors of 30-day hospital readmission following spine surgery, with particular hypothesis regarding the importance on prior hospital admission, and to develop a machine learning (ML)-based readmission risk score. While previous literature has examined predictors such as surgical indication, length of stay, and postoperative complications, the contribution of prior preoperative hospitalization had not been clearly established in the spine population. Furthermore, existing predictive scoring systems have relied on relatively limited patient factors, leading to the goal of a more comprehensive approach.

The authors reviewed the electronic medical record of 653 patients who underwent decompression with or without fusion for degenerative or traumatic indications between June 2020 and June 2021 at a single academic institution. Patients were excluded if they lacked at least one year of available preoperative admission data. Readmission was defined as any unplanned hospital admission within 30 days following the index surgery, excluding planned admissions including wound checks or physical therapy. Demographic, comorbidity, surgical, and perioperative data were abstracted, including Charlson Comorbidity Index (CCI), body mass index, surgical approach, interbody use, estimated blood loss, length of stay, and in-hospital complications.

Initial univariate analyses identified several variables to be associated with readmission, which were then evaluated in Cox proportional hazards models adjusted for sex, indication, and surgical level. Predictors of 30-day readmission included higher age-adjusted CCI (HR 1.4), traumatic indications (HR 2.6), thoracic procedures (HR 2.0), postoperative complications before discharge (HR 4.6), postoperative complications requiring reoperation (HR 7.5), and longer length of stay (HR 2.6). Importantly, hospital admission within one year prior to surgery also independently predicted readmission (HR 1.5), supporting the authors’ central hypothesis.

Based on these factors, the authors constructed a spine-specific risk calculator with a maximum score of 30 points. Patients were stratified into four categories: low risk (0–4 points, 2.2% readmission), moderate risk (5–9, 5.6%), high risk (10–14, 11.7%), and extreme risk (≥15, 39.5%). The model demonstrated strong discriminatory ability (C-index 0.79, Somers’ D 0.56, AIC 516.8), outperforming the institution’s existing standardized hospital readmission score (C-index 0.71, Somers’ D 0.41, AIC 546.7). These findings suggested that incorporating spine-specific surgical and perioperative variables meaningfully improved prediction of postoperative readmission compared to generalized models.

Several aspects of this study stand out. First, the inclusion of prior hospital admission was important as it may have served as a proxy for overall frailty, uncontrolled comorbidities, or health system dependence. This is a simple, easily identifiable risk factor that could readily be applied at the point of care.

Second, the study highlighted that while preoperative admission was significant, perioperative and postoperative factors—particularly in-hospital complications (HR 4.6) and complications requiring reoperation (HR 7.5)—conferred greater risk for readmission. This distinction is important when counseling patients, as it emphasizes the multifactorial nature of readmission and the need for both optimization before surgery and vigilant perioperative care.

Finally, the development of a spine-specific, ML-based tool was a potentially impactful addition to the movement towards precision risk stratification in surgery overall. This may help institutions categorize patients into distinct groups of readmission risk, but most specifically highlights the factors to be taken into account with such predictions.

Nonetheless, study limitations should be acknowledged. The single-institution study design and cohort size limit external generalizability, particularly as only 44 readmissions were observed. The model was also validated without external testing, raising concern for potential overfitting. Additionally, unmeasured confounders such as socioeconomic status, smoking, or social support, which are all factors known to affect readmission risk, were not available in the dataset.

In summary, Lin et al provide a thoughtful analysis of predictors of hospital readmission after spine surgery, with novel emphasis on prior preoperative admission as a marker of increased odds of 30-day readmission. Their ML-based risk calculator, incorporating both preoperative and perioperative variables, outperformed a standard institutional model and demonstrated potential clinical utility for patient counseling and targeted postoperative surveillance. While further validation in larger, multi-institutional cohorts will be important, this reviewed study advances efforts to refine risk prediction in spine surgery and highlights an accessible, clinically meaningful variable—recent hospitalization—that warrants consideration in both research and practice.