Presumptive Diagnostic Algorithms are an emerging technology inHealthcareIT. This demo will show the simplicity, agility, and security for deploying a Diagnostic Machine Learning Algorithm that can be used by frontline workers using theMicrosofttechnology ecosystem. Acute Inflammation of Urinary Bladder and Acute Nephritis of Renal Pelvis are separate diagnoses which often have overlapping symptoms. The University of California Irvine provides an Open Source dataset cataloguing assessments of patients presenting with symptoms related to these diagnoses. This Use Case will providetwo demos,one for non-technical users and another that is highly technical using real Healthcare data with a real Machine Learning Predictive Model. A Power App will be used to enter symptoms and return a diagnosis fromAzure ML.
Business Challenge
Developing and Deploying Presumptive Diagnostic Algorithms has proven to be challenging for the Healthcare Industry. Data Science professionals are in high demand and very costly. IntegratingMachine LearningPredictive Algorithms into tools for frontline workers often comes with Infrastructure, Security, and Scalability challenges. There is great Value for a Healthcare organization to train Predictive Algorithms with data representative of their patient population, understand the factors contributing to predictions, securely control who is able to use the predictive algorithms, and thoroughly document usage of the App. Realizing Value can be challenging, but Microsoft offers a suite of tools that lower the barrier to success. This Use Case is not intended to be used in the real world for diagnosis of Acute Inflammation of Urinary Bladder and Acute Nephritis of Renal Pelvis (even though the data is real), but is for the purposes of demonstrating the ease by which these types of Presumptive Diagnostic Algorithms can be deployed to achieve Realized Value.
Business Outcomes
The Business Outcomes for different groups of stakeholders are as follows:
Frontline Workers
Easily access an App to get presumptive diagnoses by entering symptoms
Quick and secure results
Time savings versus manually referencing and cross-checking lists of symptoms
IT Stakeholders
Zero on-prem infrastructure to manage
Low-Cost relative to alternatives
Secure Infrastructure with usage telemetry
Low-effort for integration of front and back-end tools
Healthcare Quality
Track usage of the algorithm, compare predictions to final diagnosis to assess the accuracy
Diagnostic aid for frontline workers
Potential for faster time to treatment if these types of solutions accelerate diagnoses
Financial
Low cost of infrastructure
Reduced cost of Data Science resources
Lower total cost of ownership versus alternatives
Solution Overview
This is a demo of how Machine Learning and predictive modeling can be used in healthcare scenarios to use variables to predict if a patient has a high likely hood of a specific disease. This uses Azure for ML,Power Apps,Power Automateand Power Bi.
