Can the American College of Surgeons Risk Calculator Predict 30-Day Complications After Spine Surgery?

Michael H. McCarthy, MD, MPH; Partik Singh, BA; Rusheel Nayak, BA; Joseph P. Maslak, MD; Tyler J. Jenkins, MD; Alpesh A. Patel, MD; Wellington K. Hsu, MD


Spine. 2020;45(9):621-628. 

In This Article

Materials and Methods

The Risk Calculator includes patient demographics and comorbidities from >40 states and a variety of health care settings ranging from large academic centers to small community hospitals. It draws on datasets from >3.8 million operations from 740 participating hospitals from 2012 to2016.[13] Participating hospitals employ clinical reviewers which collect and verify data further augmenting the accurate and robust nature of this information; furthermore, this data collection has shown satisfactory inter-rater reliability.[14] The Risk Calculator became publicly available in 2013 and generates risk profiles by entering the CPT code for the planned procedure and a variety of patient specific variables (Figure 1 A). The input variables provide a comprehensive clinical profile allowing the Risk Calculator to return risk estimates for the 11 specific complications, predicted length of stay, and requirements of skilled nursing home or rehabilitation facility (Figure 1 B).

Figure 1.

(A) The ACS Risk Calculator allows patients and providers to input the CPT code for the planned procedure and patient specific variables. (B) The ACS Risk Calculator provides an estimated risk for 11 complications, predicted length of stay, and need for skilled nursing or rehab facility.

Our patient cohort was selected via a retrospective review of 641 patients who underwent elective spine surgery at a single institution (Northwestern Memorial Hospital; Chicago, IL), from 2009 to 2015 by querying the Northwestern Medicine Enterprise Data Warehouse (NMEDW) for both cervical and lumbar fusion CPT codes: lumbar arthrodesis (22558, 22585, 22600, 22612 22614, 22630, 22632–22634) and cervical arthrodesis (22551,22554, 22558, 22600, 22614). Exclusion criteria consisted of patients with incomplete or absent preoperative documentation of demographics and/or medical comorbidities, without 30 days of postoperative follow-up, and a history of neoplasm or infection as primary indications for surgery. The selected cohort was composed of 237 patients who underwent primary lumbar fusion and 404 patients undergoing primary cervical fusion. Incidence of complications within the 30-day postoperative period were identified via institutional chart review and subsequently compared to individual patient preoperative Risk Calculator outcomes to assess accuracy.

Descriptive statistics were calculated for the overall sample with further delineation of anterior and posterior cervical fusion groups, whereas the lumbar data were analyzed as one fusion cohort. In contrast to the cervical cohort, further delineation between anterior and posterior fusions was unable to be accurately assessed due to variable interbody coding within the NMEDW; thus, a single lumbar fusion cohort was utilized. Logistic regression models were fit with actual complication occurrence (yes vs. no) as the dependent variable in each model and the Risk Calculator estimated risk as the independent variable. Odds ratio (OR) with 95% confidence intervals (CI), P values, and c-statistic values were calculated for each complication and outcome category for all patients and individual subgroups. C-statistic was used as the measure of concordance for each model. Receiver-operating charateristic (ROC) curves were plotted for each complication type to visually depict the predictive ability of the estimated risks. All analyses were conducted for the overall sample with further stratification of the lumbar fusions and anterior versus posterior fusion cohort for the cervical portion. Additional analyses were stratified by length of fusion (one-level vs. multilevel). All analyses were conducted using SAS v9.4 (SAS Institute, Cary, NC).