ASA-PS

ASA-PS predicts ASA Physical Status classification from preoperative data to provide an electronically derived measure of patient fitness for surgery using multinomial regression models trained on the 2014 American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) Participant Use Data File.

Key Features:

• Predictive Modeling: Multinomial regression models were developed from the 2014 ACS NSQIP Participant Use Data File using a 21-variable model derived from over half a million elective surgical procedures.
• Cross-Validation: The model underwent internal cross-validation to evaluate performance across different patient cohorts.
• Accuracy Assessment: Model performance was quantified using C-Statistics with achieved accuracy C = 0.77 +/- 0.0025 for the 21-variable model and assessed with calibration plots.
• Concurrent and Predictive Validity: Validity was assessed via associations with preoperative indicators (including comorbidities and BMI) and with postoperative complications, with model-derived ASA-PS showing stronger concurrent validity than traditional observational methods.

Scientific Applications:

• Clinical Decision Support: Aids clinicians in assessing patient fitness for surgery and informing decisions about surgical candidacy and perioperative management.
• Research Utility: Provides a standardized electronic ASA-PS measure to support research on surgical outcomes, quality measurement, and healthcare analytics.
• Quality Improvement: Enables predictive analyses that can inform patient safety protocols and optimize resource allocation.
• Inpatient Admission Justification: Model predictions can be used to support justifications for inpatient admissions based on predicted surgical risk profiles.

Methodology:

Analysis of preoperative variables from the 2014 ACS NSQIP Participant Use Data File using multinomial regression (21-variable model), internal cross-validation, evaluation by C-Statistics (C = 0.77 +/- 0.0025), and assessment with calibration plots.