Predicting Postoperative Complications and Mortality After Acetabular Surgery in the Elderly

A Comparison of Risk Stratification Models

Chang-Yeon Kim, MD; Nikunj N. Trivedi, MD; Lakshmanan Sivasundaram, MD; George Ochenjele, MD; Raymond W. Liu, MD; Heather Vallier, MD


Curr Orthop Pract. 2020;31(2):162-167. 

In This Article


As the elderly population continues to grow, the number of geriatric fractures also will continue to rise. There are currently no guidelines for the treatment of acetabular fractures in elderly patients. In addition to injury characteristics (e.g. articular congruity and hip stability, etc.), orthopaedic surgeons must take into consideration the medical comorbidities of these patients. Risk stratification models have been adapted to assist surgeons in perioperative planning. The purpose of this study was to determine if the ECM, CCI, and CCS are accurate predictors of adverse events after acetabular fracture surgery in elderly patients.

The results of this study suggest that both the CCI and ECM are adequate predictors of postoperative complications, length of stay, nonroutine discharge, and mortality after surgical management of acetabular fractures in geriatric patients, with the ECM demonstrating superior predictive discrimination. Combining the two measures, forming the CCS, led to improvements in predictive discrimination. Schneeweiss et al.[14] demonstrated that even small differences in C-statistics were associated with significant differences in confounding bias. As the addition of the comorbidities found in the CCS is minimal extra work for the surgeon and can provide significant improvements in predictive discrimination, we believe that it is an index worthy of adoption in orthopaedic surgery.

Our findings are consistent with past studies comparing the ECM and CCI. In a retrospective evaluation of approximately 14,000,000 orthopaedic patients, Menendez et al.[8] showed that both the ECM and CCI were excellent predictors of inpatient mortality after orthopaedic surgery, but the ECM was superior. In another study evaluating the two indices in the context of proximal humeral fractures, the authors similarly found that, although both indices were good predictors of inpatient mortality, the ECM outperformed the CCI.[15] Currently, there is a paucity of data on the CCS. Gagne et al.[9] reported that, in a subset of Medicare patients 65 yr or older, the CCS outperformed the CCI and ECM in predicting short-term and long-term mortality.[9] However, to date, no studies have evaluated the index in the setting of orthopaedic intervention.

Contrary to previous studies,[6,9,16] our investigation shows that calculating a simple comorbidity score (both weighted and unweighted) to predict outcomes did not perform as well as using each comorbidity variable separately in a multiple regression model. Although we expected the use of each comorbidity variable separately would be superior based on mathematics, we were impressed by the magnitude of the improvement. We believe this occurred because the population of geriatric patients who sustain acetabular fractures is different than the populations originally used to develop the ECM and CCI. The best clinical application of these comorbidity indices may require a computerized model that incorporates a full logistic regression model, not just the score. Such software is becoming more commonly used due to advances in computing power.[17]

For cardiac, renal, and pulmonary complications, the ECM with the combined measures had good predictive discrimination and also in predicting postoperative intubation and tranfusions. However, it had poor predictive discrimination for complications after surgery. This may be because acute complications after surgery are more dependent on injury characteristics, such as soft-tissue injury, and surgeon technique and less on patient comorbidities. In addition the predictive value for extended length of stay and discharge to a skilled nursing facility was only moderate (AUC 0.6–0.7), which is similar to the the results of Ondeck et al.[7] They also showed the ECM to be a poor predictor (C-statistic of 0.647) of discharge to a facility after total hip arthroplasty.[7] It could be that their outcomes may have been dependent on extrinsic factors (i.e. insurance type), which were not included in our models.

Our study had several limitations worth noting. Because our data were derived from an administrative database, our findings are limited by coding bias, sampling bias, under-reporting of comorbidities, and potential under-reporting of complications.[18–21] Furthermore, we were unable to evaluate the comorbidity indices for complications after discharge.[22] Last, differences in the coding algorithms of the two indices led to small variations in incidence of some comorbidities (prevalence of uncomplicated diabetes was 17.8% in ECM versus 18.0% in CCI). However, this study has a large sample size and a cohort derived from different hospitals across the country, which reduces possible institutional and geographic bias. Although the validity of the ECM and CCI in predicting inpatient and postoperative outcomes has already been reported,[5–8] this is the first study to investigate acetabular fractures in a geriatric population; it also is the first orthopaedic study to include the recently developed CCS.