Reliability of the ASA Physical Status Classification System in Predicting Surgical Morbidity: a Retrospective Analysis

Gen Li, Jeremy P. Walco, Dorothee A. Mueller, Jonathan P. Wanderer, Robert E. FreundlichJ. Medical Syst 2021 · DOI

TL;DR

This study investigates the reliability of the American Society of Anesthesiologists (ASA) Physical Status Classification System in predicting surgical outcomes. Researchers found that while the ASA system shows a correlation with patient comorbidities and demographics, a composite model incorporating these factors significantly outperforms the ASA score alone in predicting short-term postoperative mortality.

What Problem Does This Paper Address?

The ASA Physical Status Classification System has been a cornerstone in anesthesiology for over 60 years, used to assess pre-anesthesia patient health and predict surgical risk. However, its subjective nature has led to inconsistent assignments among clinicians, raising questions about its reliability. While the ASA system is widely used for risk adjustment and billing, it's explicitly recommended not to be used in isolation for mortality prediction. This paper aims to bridge this gap by assessing how well the ASA system correlates with more objective measures of patient illness (comorbidities) and to evaluate its predictive power for postoperative mortality compared to more comprehensive models.

Methods

The researchers employed a retrospective cohort study design, analyzing data from adult patients who underwent surgery at a single medical center. Their primary analysis utilized a multivariable ordinal logistic regression model to examine the relationship between ASA physical status and 31 Elixhauser comorbidities, while controlling for demographic and clinical factors like age, gender, weight, and ASA emergency status. A Least Absolute Shrinkage and Selection Operator (LASSO) approach was used to identify significant covariates for the regression model. For model validation, a five-fold cross-validation was performed. A secondary analysis focused on evaluating the predictive capability of different models for 30-day postoperative mortality using Receiver Operating Characteristic (ROC) curves and c-statistics.

Datasets

The study utilized a dataset of 56,820 surgical cases extracted from the Vanderbilt University Medical Center Perioperative Data Warehouse, covering the period from November 2, 2017, to April 22, 2020. This dataset included patient demographics, ASA physical status, ASA emergency status, 31 Elixhauser comorbidities, and postoperative mortality data (30, 60, and 90 days).

Key Insights and Ideas

  • ASA System is Not Perfect, But Useful: While criticized for subjectivity, the ASA system does correlate with objective patient characteristics and comorbidities, suggesting it captures some essential aspects of patient health.
  • Comorbidities Matter Significantly: A substantial number of Elixhauser comorbidities (27 out of 31) were independently associated with ASA physical status, highlighting the complex interplay between specific health conditions and how anesthesiologists perceive overall patient risk.
  • Demographics Play a Role: Factors like older age, male gender, higher body weight, and ASA emergency status were also independently associated with higher ASA physical status assignments, indicating these are considered by clinicians.
  • Composite Models Reign Supreme for Prediction: The most crucial finding is that a model combining ASA physical status with Elixhauser comorbidities and other patient characteristics significantly outperformed the ASA score alone and other composite comorbidity indices in predicting short-term postoperative mortality. This suggests a more nuanced approach is needed for accurate risk stratification.
  • Potential for Bias: The study noted that some comorbidities were associated with a lower ASA status, suggesting potential bias where providers might overemphasize certain conditions (e.g., neurological, pulmonary) while downplaying others.

Key Results

  • 27 out of 31 Elixhauser comorbidities were independently associated with ASA physical status.
  • Older age (OR 1.39 per 10 years), male gender (OR 1.24), higher body weight (OR 1.08 per 10 kg), and ASA emergency status (OR 2.11) were independently associated with higher ASA physical status.
  • The composite model derived from the primary analysis demonstrated superior predictive power for 30-day postoperative mortality, with a c-statistic of 0.82 (95% CI 0.81-0.83), compared to the ASA physical status alone (c=0.79) and composite comorbidity indices (c=0.80).

Limitations and Future Work

The authors acknowledge several limitations:

  • Data Availability: The composite model requires data available at the end of hospital admission, whereas ASA status is assigned preoperatively. This temporal difference might affect the accuracy of capturing comorbidities present on the day of surgery.
  • Retrospective Nature: The study's retrospective design inherently limits the ability to establish causality.
  • Single-Center Study: The findings are based on data from one institution, and a multicenter analysis would be beneficial for broader validation.
  • Overfitting: The authors note the potential for overfitting, especially when predicting rare events like short-term postoperative mortality.

Future work could involve refining the ASA system with more objective data, developing institutional-specific examples to improve consistency, and conducting multicenter studies to validate the composite model across diverse patient populations and surgical subspecialties. Further examination of surgery-specific subpopulations is also suggested.

Summary updated

BibTeX
@article{li2021reliability,
  author = {Gen Li and Jeremy P. Walco and Dorothee A. Mueller and Jonathan P. Wanderer and Robert E. Freundlich},
  title = {{Reliability of the ASA Physical Status Classification System in Predicting Surgical Morbidity: a Retrospective Analysis}},
  journal = {J. Medical Syst.},
  year = {2021},
  doi = {10.1007/s10916-021-01758-z},
}
Figure