The overarching goal of this research is to improve the quality of care of Veterans with skin and soft tissue infections (SSTIs). We propose to use VA electronic health record data to perform a large-scale, nation-wide analysis of management of SSTIs. Aims: This project consists of the following Specific Aims: 1. Differentiate purulent from non-purulent SSTIs treated in the ambulatory VA setting by utilizing Natural Language Processing (NLP). 2. Examine practice variation in the management of SSTIs, including I&D and antibiotic prescribing for purulent and non- purulent SSTIs. Significance: This project is significant because it addresses key, currently unanswered questions regarding best treatment practices for SSTIs, which is a significant problem in the Veteran population as indicated by our preliminary data. This study will provide an understanding of the current management of SSTIs in the VA, which will inform a subsequent Investigator Initiated Research (IIR) study to compare the effectiveness on SSTI outcomes of the various treatment strategies identified. Ultimately these IIR results will reshape the evidence base around best practices for SSTIs management. By incorporating advanced health informatics methods, this proposal will allow for nationwide quantification of inappropriate antibiotic use in SSTIs and will form the basis for ongoing prospective evaluations of antibiotic use by the VA Antimicrobial Stewardship Task Force (ASTF). This proposal will also provide a basis for targeted AS interventions by elucidating clinical factors associated with inappropriate antibiotic use. HSR&D Priority Addressed: This proposal addresses the HSR&D Priority Domain of Healthcare Informatics through its analysis and application of VINCI data to the SSTI population, and its NLP methods development and application in sub-classifying SSTIs. Innovation: The innovation of this project is two-fold. It is the first known attempt to utilize NLP methods to phenotypically characterize SSTI cases. It will also provide a rare first-hand understanding of the practice variation in VA SSTI treatment. Methods: Our approach for Aim 1 will be to use ICD-9 and -10 codes as a screening tool to identify SSTI cases. We then apply NLP of these clinical notes in order to accurately distinguish purulent vs. non- purulent SSTIs. For Aim 2, SSTI treatment will be classified into groups including whether I&D was performed for purulent SSTIs and the type of antibiotics prescribed for purulent and non-purulent SSTIs. We will then examine patient, provider and setting characteristics to determine factors associated with the variation in the treatment of SSTIs. Expected Results include descriptive statistics of antibiotic class and duration for purulent and non- purulent SSTIs as well as I&D rates for purulent SSTIs. Three separate multilevel logistic regression models will provide an understanding of clinical characteristics associated with variation in I&D performance and community-acquired methicillin-resistant Staphylococcus aureus coverage for purulent SSTIs and anti- streptococcal coverage for non-purulent SSTIs. Next Steps: These results will lay the groundwork for a subsequent IIR application to compare the effectiveness on SSTI outcomes of the various treatment strategies identified. This pilot will yield an NLP approach that can be operationalized to facilitate the VA ASTF evaluation of SSTI treatment. It will also provide a greater understanding of the clinical factors associated with VA practice variation, which will allow the ASTF to focus its educational efforts for VA clinicians.
External Links for this Project
Grant Number: I21HX002416-01A1
- Rhoads JLW, Willson TM, Sutton JD, Spivak ES, Samore MH, Stevens VW. Epidemiology, Disposition, and Treatment of Ambulatory Veterans With Skin and Soft Tissue Infections. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2021 Feb 16; 72(4):675-681. [view]
TRL - Applied/Translational, Treatment - Observational
Best Practices, Natural Language Processing, Quality of Care
None at this time.