No abstract

Objective Heatlhcare institutions are establishing frameworks to govern and promote the implementation of accurate, actionable, and reliable machine learning models that integrate with clinical workflow. Such governance frameworks require an accompanying technical framework to deploy models in a resource efficient, safe and high-quality manner. Here we present DEPLOYR, a technical framework for enabling real-time deployment and monitoring of researcher-created models into a widely used electronic medical record system. Materials and Methods We discuss core functionality and design decisions, including mechanisms to trigger inference based on actions within electronic medical record software, modules that collect real-time data to make inferences, mechanisms that close-the-loop by displaying inferences back to end-users within their workflow, monitoring modules that track performance of deployed models over time, silent deployment capabilities, and mechanisms to prospectively evaluate a deployed model’s impact. Results We demonstrate the use of DEPLOYR by silently deploying and prospectively evaluating 12 machine learning models trained using electronic medical record data that predict laboratory diagnostic results, triggered by clinician button-clicks in Stanford Health Care’s electronic medical record. Discussion Our study highlights the need and feasibility for such silent deployment, because prospectively measured performance varies from retrospective estimates. When possible, we recommend using prospectively estimated performance measures during silent trials to make final go decisions for model deployment. Conclusion Machine learning applications in healthcare are extensively researched, but successful translations to the bedside are rare. By describing DEPLOYR, we aim to inform machine learning deployment best practices and help bridge the model implementation gap.

Introduction: Sputum non-conversion is smear positive tuberculosis despite anti-tubercular therapy. Various factors may lead to sputum non-conversion including resistance to anti-tubercular drugs, age, gender, disease severity, non-compliance, drugs unavailability etc. Little is known and studied about the contribution of these individual factors. Our study sought to determine the prevalence of sputum smear non-conversion in patients at the end of intensive phase of tuberculosis treatment visiting a tertiary-level institution in Nepal. Methods: A descriptive cross-sectional study was conducted among recorded data of patients undergoing sputum Acid Fast Bacilli staining at the end of intensive phase at National Tuberculosis Control Center from April 2018 to April 2020. The study was approved by Nepal Health Research Council (Registration no: 76012020 P). The convenient sampling method was adopted. The data were analyzed using Microsoft Excel. Point estimate at 95% Confidence Interval was calculated along with frequency and proportion for binary data. Results: Our study found that out of 830 samples that were tested by Acid Fast Bacilli stain at the end of intensive phase, 40 (4.82%) (3.37-6.28 at 95% Confidence Interval) were sputum smear non-converters. The mean age of sputum non-converters was 41.25±15.543 years. Conclusions: The study shows that a significant proportion of patients remain acid-fast stain positive despite the treatment. However, the proportion is low compared to other similar studies around the globe. This study provides program managers with evidence to support the development of more tailored tuberculosis care and need to conduct more intensive studies about various factors that may lead to non-conversion.

No abstract