Healthcare Shorts: Predictive Diagnosis
Published Jul 06 2022 01:08 PM 899 Views
Microsoft

Story Background

 

Presumptive Diagnostic Algorithms are an emerging technology in Healthcare IT. This demo will show the simplicity, agility, and security for deploying a Diagnostic Machine Learning Algorithm that can be used by frontline workers using the Microsoft technology 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 provide two 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 from Azure 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. Integrating Machine Learning Predictive 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 Automate and Power Bi. 

 

 

Thanks for reading, Shelly Avery |EmailLinkedIn 

If you are interested in more content like this, then follow  https://aka.ms/HealthcareShorts 

 

Please follow  aka.ms/HLSBlog for all this great content.

Co-Authors
Version history
Last update:
‎Jul 06 2022 01:08 PM
Updated by: