American Surgical Association

Back to 2019 Abstracts


Can We Predict Incisional Hernia? - Development of a Surgery-Specific Decision-Support Interface
John P. Fischer*, Geoffrey M. Kozak*, Marten N. Basta*, Robyn B. Broach*, Charles Messa*, Irfan Rhemtulla*, Ron P. DeMatteo, Joseph M Serletti
Hospital of the University of Pennsylvania, Philadelphia, PA

Objective(s):
Incisional hernia (IH) is a common and morbid complication after abdominal surgery (AS). However, there is a paucity of readily available, surgeon-facing tools capable of predicting IH occurrence. We aim to identify procedure-specific risk factors independently associated with IH and build a decision-support interface surgeons can utilized at the point-of-care.
Methods:
Patients(N=29,739) undergoing AS from 2005-2016 were retrospectively identified within
inpatient and ambulatory databases at our institution. Surgically-treated IH, complications, and
costs were assessed. Procedure-specific predictive models were generated using regression analysis and corroborated using a validation group.
Results:
Operative IH occurred in 3.8%(N=1,127) of patients at an average follow-up of 57.9 months. Combined cost of care for patients experiencing IH was $62 million. All variables were weighted according to -coefficients generating 8 surgery-specific predictive models for IH occurrence, all of which demonstrated excellent risk discrimination(C-statistic=0.760.89). IH occurred most frequently after colorectal (7.7%) and vascular (5.2%) surgery and the most common occurring risk factors increasing the likelihood of developing IH were history of AS and smoking. An integrated, surgeon-facing risk prediction instrument was created in an App for pre-operative estimation of hernia after AS.
Conclusions:
Using a bioinformatics approach, we identified risk factors predictive of IH and created eight unique surgery-specific models. The models have led to a fully designed and unifying point-of-care risk calculator App generated from a multi-year, longitudinal, multi-hospital dataset.

* By Invitation


Back to 2019 Abstracts