These tools will no longer be maintained as of December 31, 2024. Archived website can be found here. PubMed4Hh GitHub repository can be found here. Contact NLM Customer Service if you have questions.


Pubmed for Handhelds

PUBMED FOR HANDHELDS

Search MEDLINE/PubMed

  • Title: Development of a machine learning algorithm for prediction of failure of nonoperative management in spinal epidural abscess.
    Author: Shah AA, Karhade AV, Bono CM, Harris MB, Nelson SB, Schwab JH.
    Journal: Spine J; 2019 Oct; 19(10):1657-1665. PubMed ID: 31059819.
    Abstract:
    BACKGROUND CONTEXT: Data regarding risk of failure of nonoperative management in spinal epidural abscess (SEA) are limited. Given the potential for deterioration with treatment failure, a tool that predicts the probability of failure would be of great clinical utility. PURPOSE: We primarily aim to build a machine learning model using independent predictors of nonoperative management failure. Secondarily, we aim to develop an open-access web-based application that provides a patient-specific probability of treatment failure. STUDY DESIGN/SETTING: Retrospective, case-control study. PATIENT SAMPLE: Patients 18 years or older diagnosed with SEA at 2 academic medical centers and 3 community hospitals. OUTCOME MEASURES: Failure of nonoperative management. METHODS: This is a retrospective cohort study of 367 patients with SEA initially managed nonoperatively between 1993 and 2016. The primary outcome was failure of nonoperative management defined as neurologic deterioration, worsened back and/or radicular pain, or persistent symptoms despite initiation of antibiotic therapy. Five machine learning algorithms were developed and assessed by discrimination, calibration, and overall performance. RESULTS: Ninety-nine (27%) patients failed nonoperative management. Factors determined for prediction of nonoperative management were: motor deficit, diabetes, ventral component of abscess relative to thecal sac, history of compression or pathologic vertebral fracture, sensory deficit, active malignancy, and involvement of 3 or more vertebral levels. The elastic-net penalized logistic regression model was chosen as the final model given its superior discrimination, calibration, and overall model performance. This model was incorporated into an open access web application. CONCLUSION: By building a discriminative and well-calibrated model in a user-friendly and open-access digital interface, we hope to provide a prognostic tool that can be used to inform clinical decision-making in real-time.
    [Abstract] [Full Text] [Related] [New Search]