Improving surgical outcomes through optimized hernia prediction

点击翻译按钮获取中文摘要

PROJECT SUMMARY Incisional hernia (IH) is a common, overlooked surgical health problem spanning a broad range of patients and stakeholders. In the U.S., over 153,000 IHs are repaired per year with expenditures exceeding $7 billion. Evidence-based interventions, including preoperative optimization, surgical techniques, and prophylactic mesh, can reduce risk; however, multi-level factors impede clinical translation. One critical barrier is the need for accurate, generalizable risk prediction to link risk recognition, behavior change, and outcomes. Pre-operative risk assessment enables providers to leverage risk information to guide decision-making, surgical planning, and informed consent. Current limitations of IH prediction have created barriers to IH prevention. Our proposal addresses the need for patient-specific, clearly presented risk information to enhance health care, enable individualized risk assessment, and close the gap between optimal practice and actual clinical care in hernia prevention. Our preliminary research has defined the clinical and economic burden of IH, characterized inefficiencies in treatment-oriented paradigms, defined key patient populations for prevention, and demonstrated effective risk reductive surgical techniques. We also show the benefit of using electronic health record-based prediction over administrative claims datasets and the power of machine learning to maximize model performance. Most recently, we created a pilot, portable, clinical decision support-mobile user interface for prediction, setting the stage for this proposal. Our approach is hallmarked by use of a unique multi-source database, innovative applications of machine learning, stake-holder engagement, and inter-disciplinary collaboration. In this proposal, we will identify and discover factors associated with IH using data from >130,000 patients with longitudinal follow-up and characterize intra-operative risk factors using natural language processing. Machine learning will enable improved predictive performance (Aim 1). Models will be tested on a geo-temporally diverse data source and end-user input will guide and prioritize features, format, and functionality, leading to creation of a provider-adapted Hernia Calc housing the predictive models (Aim 2). Hernia Calc will be evaluated in real-world practice to assess contextual determinants and to create a stakeholder-driven implementation protocol to identify strategies to support widespread dissemination (Aim 3). Our approach addresses barriers to IH prevention through development of optimized, validated, specialty- specific IH risk models integrated within a provider-informed interface and implementation strategies for clinical use. This work will lead to a broad, significant, and sustained impact on the field, catalyzing a major pivot towards hernia prevention, enabling precise risk prediction for abdominal surgery patients. Completion of our aims will augment knowledge of hernia and improve health outcomes in surgery allowing a pivot in practice towards prevention and aligning our proposal with Core Missions of the NIH.