To the editor,
We would like to express our sincere admiration to Roy et al. [
1] for their insightful study, “A Predictive Model of Failure to Rescue After Thoracolumbar Fusion.” This important work underscores the significant role of frailty in determining surgical outcomes and contributes meaningfully to the ongoing discussion on optimizing patient care in spine surgery. By leveraging the risk analysis index within a large, high-quality dataset, the authors offer valuable insights into the predictive relationship between frailty and failure to rescue (FTR) outcomes. Their findings align with an expanding body of evidence that highlights the critical impact of frailty on surgical risk stratification and resource allocation in high-risk patient populations [
2-
5].
While the model introduced in the study demonstrates excellent predictive performance, reflected in a high area under receiver operating characteristic curve of 0.918, we would like to offer feedback regarding the selection of predictors in the final model. Specifically, the inclusion of postoperative variables such as cardiac arrest raises questions about the model’s real-world clinical applicability and utility. As cardiac arrest occurs in close proximity to mortality, its predictive value in facilitating timely and actionable intervention and rescue efforts is inherently limited. Although incorporating such variables may enhance the model’s statistical performance, it may also reduce its practical utility, as predictions are most beneficial when they provide clinicians with actionable insights at or near the onset of a complication rather than at a stage nearing mortality [
6-
8]. Moreover, this including factors such as cardiac arrest may divert attention from preoperative factors that are more temporally relevant and universally applicable and could offer more actionable insights for clinicians in varied healthcare settings. Therefore, a reevaluation of the selected predictors, emphasizing those with consistent relevance across different patient demographics, would enhance the model’s utility and ensure that it better serves clinical decision-making processes.
To enhance the model’s practical value, it may be beneficial to first develop a predictive framework focusing on the variables available at the time of the index complication. A subsequent step could then assess the incremental impact of including later-occurring predictors. This approach would provide a clearer understanding of the added value of such variables while ensuring the model remains clinically valid and applicable in guiding early risk assessment, identification, and intervention efforts.
Refining the model to focus on predictors that are available earlier in the postoperative course—ideally at the onset of the first recorded complication—would enable clinicians to identify patients with a greater likelihood of benefitting from rescue interventions. Such an approach would enhance the model’s ability to enhance timely clinical decision-making and improve patient outcomes. For example, in the ACS-NSQIP (American College of Surgeons National Surgery Quality Improvement Program) databases, which was utilized in this study, variables including comorbid existing conditions, the type of the surgery carried out and length of operation might be more helpful. It should also be noted that including variables such as cardiac arrest might heavily impact the more clinically useful variables and influence their relative contribution to the model due to the fact that a significant portion of the effect of a variable such as type of surgery be mediated through occurrence of cardiac death and as a results by including the cardiac death in the multivariable analysis, the relative contribution of type of surgery gets significantly smaller.
In summary, we commend the authors for their valuable contribution to the field and appreciate their work in advancing the understanding of FTR in spine surgery. Future iterations of the model that emphasize early, actionable predictors could further improve its clinical applicability, ensuring its relevance as a tool for guiding timely and effective interventions.
NOTES
-
Conflict of Interest
The authors have nothing to disclose.
-
Acknowledgments
The authors would like to express their gratitude to Dr. Ahmadi for their contributions to the initial version of the manuscript.
REFERENCES
- 1. Roy JM, Segura AC, Rumalla K, et al. A predictive model of failure to rescue after thoracolumbar fusion. Neurospine 2023;20:1337-45.
- 2. Baek W, Park SY, Kim Y. Impact of frailty on the outcomes of patients undergoing degenerative spine surgery: a systematic review and meta-analysis. BMC Geriatrics 2023;23:771.
- 3. Bowers CA, Varela S, Conlon M, et al. Comparison of the Risk Analysis Index and the modified 5-factor frailty index in predicting 30-day morbidity and mortality after spine surgery. J Neurosurg Spine 2023;39:136-45.
- 4. Gupta NK, Prvulovic ST, Zoghi S, et al. Complementary effects of postoperative delirium and frailty on 30-day outcomes in spine surgery. Spine J 2025;25:966-73.
- 5. Thommen R, Bowers CA, Segura AC, et al. Baseline frailty measured by the risk analysis index and 30-day mortality after surgery for spinal malignancy: analysis of a prospective registry (2011–2020). Neurospine 2024;21:404-13.
- 6. Lee YH, Bang H, Kim DJ. How to establish clinical prediction models. Endocrinol Metab (Seoul) 2016;31:38-44.
- 7. Steyerberg EW. Clinical prediction models: a practical approach to development, validation, and updating. Berlin (Germany): Springer Nature; 2009.
- 8. Harrell FE Jr. Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. Berlin (Germany): Springer Nature; 2001.
Citations
Citations to this article as recorded by
