Purpose: The purpose of this study is to develop a decision aid tool using "real-world" data within the Australian health system to predict weight loss after bariatric surgery and non-surgical care.
Materials and methods: We analyzed patient record data (aged 16+years) from initial review between 2015 and 2020 with 6-month (n=219) and 9-/12-month (n=153) follow-ups at eight clinical obesity services. Primary outcome was percentage total weight loss (%TWL) at 6 months and 9/12 months. Predictors were selected by statistical evidence (p<0.20), effect size (±2%), and clinical judgment. Multiple linear regression and bariatric surgery were used to create simple predictive models. Accuracy was measured using percentage of predictions within 5% of the observed value, and sensitivity and specificity for predicting target weight loss of 5% (non-surgical care) and 15% (bariatric surgery).
Results: Observed %TWL with bariatric surgery vs. non-surgical care was 19% vs. 5% at 6 months and 22% vs. 5% at 9/12 months. Predictors at 6 months with intercept (non-surgical care) of 6% include bariatric surgery (+11%), BMI>60 (-3%), depression (-2%), anxiety (-2%), and eating disorder (-2%). Accuracy, sensitivity, and specificity were 58%, 69%, and 56%. Predictors at 9/12 months with intercept of 5% include bariatric surgery (+15%), type 2 diabetes (+5%), eating disorder (+4%), fatty liver (+2%), atrial fibrillation (-4%), osteoarthritis (-3%), sleep/mental disorders (-2-3%), and ≥10 alcohol drinks/week (-2%). Accuracy, sensitivity, and specificity were 55%, 86%, and 53%.
Conclusion: Clinicians may use DACOS to discuss potential weight loss predictors with patients after surgery or non-surgical care.
Keywords: Decision support model; Management; Obesity; Weight loss.
© 2024. The Author(s).