Pancreatic Surgery - Optimal Caseload Thresholds and Predictive Accuracy

Overview

The main objective of the study is to identify the optimal annual number of cases in a hospital with regard to minimising hospital mortality in pancreatic surgery. In particular, the prognostic value of such case numbers will be analysed.

Main research questions: * Can specific intervention case numbers be identified that are suitable as thresholds for annual minimum volumes and are associated with significantly low hospital mortality? * Almost all previous studies on case number effects have only shown a descriptive association between the number of cases in a given year and the quality of outcomes in the same year. The aim of this study is to investigate whether the correlations described can be demonstrated when using the previous year's procedure volume as a predictor. The study seeks to answer whether the procedure caseload has predictive value, specifically the number of cases in one year and in-hospital mortality in the following year. Background: Numerous studies have demonstrated a correlation between the number of cases and the quality of outcomes for various surgical procedures. For instance, patients who underwent surgery in high-volume hospitals (HVH) had lower mortality rates, longer survival rates, lower complication rates, and lower reoperation rates than patients who underwent surgery in low-volume hospitals (LVH). To subdivide into HVHs and LVHs, either concrete case numbers or quartile or quintile limits with an equal number of operations or clinics per group wer used. The aim of the study is to objectively determine these limits using a spline-modeled caseload term, avoiding arbitrary decisions. One limitation of the previous findings is that they may not be generalisable due to the use of a limited number of cases and outcome quality from the same year. However, it is important to note that the volume from the previous year is crucial in determining the predictive importance of caseload for future outcome quality. A recent study (in press) reported, that there are significant fluctuations in the quality of outcomes among HVHs, even between different years. Therefore, it was hypothesized that using the number of cases as a predictor of high-quality outcomes may lead to overestimation. Methods: The nationwide hospital billing data for Germany (DRG statistics) for the period 2010 to 2019 will be analysed. The risk-adjusted mortality rates are determined. For this purpose, logistic regression models are calculated that adjust the mortality risk for the following variables Sex, age, emergency of admission, year of resection, diagnosis (malign neoplasm vs. benign neoplasm vs. neoplasm of unclear dignity vs. acute pancreatitis vs. chronic pancreatitis vs. other pancreatic diseases), additional procedures (venous resections/ multivisceral resections/ arterial resections/ splenectomy/ cholecystectomy/ biliary drainage/ dialysis procedures) and selected comorbidities. To classify additional procedures in order to reflect extent of surgery and technical difficulty, a slight modification of the classification system as described in Mihaljevic et al, 2021 will be used (PMID: 33386130). The Elixhauser definitions are used for the comorbidities as described in Quan et al, 2005 (PMID: 16224307). The selection of comorbidities to be considered is based on the publication by Hunger et al, 2022 (PMID: 35525416). The case number effect is modelled using natural cubic splines. The 10th, 20th, 40th, 60th, 80th and 90th case number percentiles are used as node points. The adjusted hospital mortality as a function of the number of cases is determined using Estimated Marginal Means. Local extremes (maxima and minima) in the splines are determined using 1st and 2nd graph derivate. Various regression models are calculated using either the number of cases from the current year of operation or the previous year. The predictive accuracy of the models is determined using the established measures from signal detection theory (AUC, sensitivity, specificity, positive predictive value, negative predictive value). Subgroup analyses for individual resection procedures will be performed.

Conditions

Interventions

Eligibility

Outcomes