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


Measures for quality improvement within surgery have incentivized utilization of perioperative metrics such as the use of surgical risk calculator in numerous fields.[15–17] These preoperative assessments enable providers to identify potential risk and provide meaningful perioperative information to patients. Recent legislative efforts have potentiated the importance of "big data" as health care information moves to a publicly available format impacting reimbursement structures. Risk calculators are helpful in recognizing patients predisposed to complications and initiating preventative measure in the preoperative time period. Schenker et al[18] found that risk calculators provide surgeons with improved preoperative morbidity and mortality estimates, thus improving the informed consent process. The attention on these assessments is a result of the use complication occurrence as a proxy for the quality of care within public reporting efforts.[19]

Although the ACS Risk Calculator has been validated in recent literature, these findings are for select surgical populations and may not apply to all.[8] As far as spine surgery patients are concerned, information regarding the applicability of the Risk Calculator is limited. Veeravagu et al performed a prospective study comparing their novel spine Risk Assessment Tool to CCI and the ACS NSQIP Risk Calculator. The authors stratified predicted outcomes to low, medium and high rates and found the ACS NSQIP calculator to consistently produce complication predictions that underestimated actual occurrence, most notably for the medium-risk group (5.9% predicted, 34.5% observed).[12] Similarly, Wang et al performed a retrospective review assessing geriatric patients (age >60 years) undergoing decompression for spinal stenosis. The authors noted observed complications to be 43.8%, whereas predicted complications were 13.7%; however, they did find the Risk Calculator to accurately predict death in addition to demonstrating a high sensitivity and specificity for predicting renal failure and readmission.[19]

The outcomes of our study demonstrated that the Risk Calculator was not effective for predicting complications in patients undergoing lumbar surgery and only marginal benefit after cervical procedures. One potential explanation is that one-level fusions were more common among the lumbar cohort, whereas two-level fusions were most common among the cervical cohort. These differences could have explained the risk calculator's ability to predict more complications among the cervical cohort. It should be acknowledged that the predictions of "any complication" and "discharge to skilled nursing facility" should not be trivialized and although other complications were not accurately predicted, these specific risk profiles alone can be clinically impactful for providers and patients. Inpatient posthospital facilities may improve post-discharge patient care and play a role in reducing hospital readmissions.[20] Additionally, cost-prediction and postoperative expectations can play a significant role in both patient care and overall burden on the healthcare system. A recent study found that 30- and 90-day costs of care for discharge to postoperative rehabilitation centers were significantly higher than the corresponding costs of care for discharge to home.[21] One recent study assessing predictors for non-home discharge after spine surgery demonstrated patients discharged to a destination other than home had a 3.74 greater odds of hospital length of stay ≥5 days; furthermore, these authors associated the delay in discharge to bed availability at rehabilitation centers.[22] Costs associated with delayed discharge due to unanticipated postoperative rehabilitation requirements, which has been shown to be as high as $15,000, could be potentially mitigated with use of the risk calculator.

The limitations of the Risk Calculator as it applies to spine patients were highlighted with this study. One important conclusion from this study is that a universal risk calculator does not have the capabilities to accurately and precisely assess patients' risk profiles regardless of the planned procedure. The variety of each patients' health profile and their individual responses to procedures posed a difficult equation for calculators that are built to apply population-based statistics on individual level. Our study is limited by its inability to evaluate differences among patients and disease process complexity between our two cohorts.

Another limitation of our study is the homogenous nature of our fusion cohort, which can limit the discriminatory capacity of the Risk Calculator. A recent study by Sebastian et al assessed the Risk Calculator's prediction for patients undergoing one-level posterior lumbar fusions, and similar to our study, these authors found the Risk Calculator to inadequately capture risk profiles for their patient cohort.[23] However, these authors caution against broad interpretations of the Risk Calculator poor predictive performance in light of homogenous study cohorts. We acknowledge this potential limitation, but highlight the addition of both cervical and lumbar fusion cohorts in our study with further stratification of anterior and posterior fusion groups among each cohort. Regardless, it is the opinion of the authors that surgical risk profile tools should have capabilities to account for variations among patients and provide reproducible and consistent results.

Furthermore, the retrospective nature of this cohort study potentially precludes accurate collection via selection bias. The small sample size used to assess predictive value of the Risk Calculator model for specific complications might not be acceptable or generalizable due to the single-institution population. The results of our high-volume, academic, urban medical center have the potential to significantly differ from other hospital settings. Lastly, institutional policies regarding patient selection could effectively influence risk profiles; therefore, one could conclude the preoperative "patient selection" could limit the number of "at risk" patients leading to a population that is institutionally risk stratified.

Risk calculators are important tools that will have increasing roles within our quality- and metric-driven health care system. The number of risk calculators available to providers continues to expand and as utilization of these tools increase their clinical application must be proven. The ACS Risk Calculator is a validated tool providing robust and significant risk estimates for specific procedures; however, in its current state, it lacks generalizability and is not an accurate risk estimate tool for patients undergoing spine surgery. In our study, the Risk Calculator only predicted complications in the categories of "any complication" (P < 0.0001) and "discharge to skilled nursing facility" (P < 0.001) for our cervical cohort and failed to demonstrate any appreciable prediction for our lumbar cohort. The Risk Calculator is a useful tool in risk mitigation within general surgery but its inaccuracy, lack of granularity and finite application in musculoskeletal procedures limit its application within spine surgery.