Predicting and modeling follow‐up clinic visits (persistence) after bariatric surgery

Background Bariatric surgery is known as one of the most effective interventions to manage weight for individuals with obesity. While mean weight loss is typically strong, there exists high variability in these results; not all patients achieve successful weight loss post‐surgery. Follow‐up clinical visits are a known predictor of patient success however little is known about the factors that influence follow‐up persistence. Methods We developed a class of one dimensional 2‐parameter linear discrete dynamical systems to predict the percent of patients that attend each follow up visit. These parameters biologically represent the half‐life in the drop off rate observed in the percent of patients attending follow up visits and their persistence plateau. To test whether the half‐life of persistence differs between surgeries, separate models were developed using data from 7 different types of bariatric surgery performed by the Bariatric Medicine Institute in Utah. Model parameters were determined by regressing between the percent of patients attending follow up visit number n, denoted by the variable, p n,− against the previous follow up visit, (p n−1 ). We then compared the discrete model simulations against the actual persistence data graphically to evaluate the quality of model calibration to observed data and computed the half‐life. Results There were seven types of surgeries (loop to duodenal switch, duodenal switch, sleeve, bypass, duodenal switch to bypass, bandication, and band) where the recursive relationship was found to modeled well by a linear discrete dynamical system p n =ap n‐1 +b. Parameter values specifically the half‐life differed based on surgery type ( Table 1). The optimal half‐life was found in band and bandication while the highest drop in persistence was found in bypass surgeries. Conclusions To date, all models that examine persistence of clinic follow‐up visits post‐surgery and predictions of bariatric surgery success are statistical. Our deterministic system provides new mechanistic insights by characterizing persistence by two parameters. Our findings suggest that while gastric bypass has strong mean weight loss results, persistence for clinical follow up visits drops off early in comparison to gastric band. Furthermore, clinicians can estimate our discrete model parameters based on individual demographics to inform optimal surgery types and predict potential patient drop off pre‐surgery. 1 Half‐life of persistence derived from each surgery typeSurgery Type Half‐LifeLoop DS 0.86587 DS 0.88248 Sleeve 0.80547 Bypass 0.79028 DS to Bypass 0.86417 Bandication 0.97439 Band 0.94337