J. D. Bohnen1,2, G. A. Anderson1,3, R. T. Spence2,4, K. Ladha5, D. Chang1,2 1Massachusetts General Hospital,General Surgery,Boston, MA, USA 2Massachusetts General Hospital,Codman Center For Clinical Effectiveness In Surgery,Boston, MA, USA 3Harvard School Of Medicine,Program For Global Surgery And Social Change,Brookline, MA, USA 4University Of Cape Town,General Surgery,Cape Town, WESTERN CAPE, South Africa 5University Of Toronto, Toronto General Hospital,Department Of Anesthesia,Toronto, Ontario, Canada
Introduction:
The focus of many data collection efforts in the U.S. centers on creating more granular data. The assumption is that more complex data collection is equal to more accurate data and therefore allows better predictions of outcomes. We hypothesized that data is often needlessly complex and that complexity can be a burden to those collecting and analyzing the data. Moreover, it is a barrier to data collection in Low and Middle Income Countries (LMIC’s). We sought to demonstrate this concept by examination of the American Society of Anesthesiologists (ASA) physical classification system.
Methods:
We created every possible two, three and four category combinations of the current five category ASA score. This resulted in 14 combinations of simplified ASA with 2, 3 or 4 categories. We then compared the predictive ability of these simplified scores for post-operative outcomes on all patients in the NSQIP database from 2006-2012 (2.3 million patients). We created unadjusted and adjusted logistic regression models using these 14 different combinations of simplified ASA scores as the predictor variable. Individual model performance was assessed by comparing Receiver Operator Characteristic (ROC) curves for each model with the standard ASA model using the outcomes of in-hospital and 30-day mortality and any morbidity.
Results:
Two of our 4-category models (ASA 1&2, 3, 4,5 and ASA 1, 2, 3, 4&5) and one of our 3-category models (ASA 1&2, 3, 4&5) had ability to predict all outcomes equivalent to standard ASA. Two of the 2-category variables (ASA 1&2&3, 4&5 and ASA 1&2, 3&4&5) provided good estimates that were only slightly worse than standard ASA. These results held for all outcomes and on all subgroups tested. The performance of the 3 best simplified ASA scores were also equivalent to the standard ASA score in the univariate analysis and multivariate analysis.
Conclusion:
Currently there is a desire to strive for the most granular data and use the largest number of variables for risk-adjusted predictions. This complexity is often at the expense of utility. We have used the single best predictor in surgical outcomes research to show this is not necessarily the case. In this example of the ASA scoring system our data show that one can simplify ASA into a 3-category variable without losing any ability to predict patient outcomes. Further research is needed to determine whether other commonly used scoring systems can be simplified without compromising their discrimination ability. When working in LMIC’s simple systems that operate just as well as more complex ones will help to facilitate the spread of surgical data collection and thereby lead to improvements in patient care.
