Start Your Search

Karen Hay



Author of

  • +

    PS01 - AI and Smart Technology in Patient Safety Management (ID 3)

    • Event: e-Health 2019 Virtual Meeting
    • Type: Panel Session
    • Track:
    • Presentations: 1
    • Coordinates: 5/27/2019, 10:30 AM - 11:30 AM, Area 1
    • +

      PS01.02 - Using Artificial Intelligence to Flag Laboratory Findings for Public Health (ID 374)

      Karen Hay, Digital Health Solutions and Innovation Branch, Ministry of Health and Long-Term Care; Toronto/CA

      • Abstract

      Purpose/Objectives:
      Ontario laboratories that identify a specimen that is positive for one of 68 communicable diseases of public health significance are legally mandated to report that finding to their local public health unit to enable monitoring of the health of the community and to provide the basis for preventive action. Reporting is typically done through paper-based fax or mail, an archaic and labor-intensive process for laboratories and public health units. At the same time, most laboratories submit all results to the provincial laboratory result repository, the Ontario Laboratory Information System (OLIS). OLIS results encompass more than 95% of the province’s laboratory results originating from hospital, community, and public health laboratories; however, this system is not currently being used for reporting purposes. The project team has undertaken a project to investigate the potential to replace the current method with a digital approach that applies artificial intelligence to OLIS records to identify reportable findings and report the results to public health on behalf of laboratory operators.


      Methodology/Approach:
      Using the provincial laboratory information system (OLIS), public health information system (iPHIS) and surveillance case definitions for diseases of public health significance, the team has created a system which uses logic and machine learning, or artificial intelligence (AI), to distinguish reportable laboratory findings from large volumes of negative results. This process involves cleansing of unstandardized free text data and training of artificial neural networks. 1.png 2.png


      Finding/Results:
      AI and Boolean algorithms have been prototyped for 7 diseases of public health significance. Challenges include the data quality of unstructured, free-text laboratory results, labelling the dataset, and future changes in the laboratory process. Highly accurate identification of diseases of public health significance is anticipated, with results validated by laboratory professionals.


      Conclusion/Implications/Recommendations:
      The current system for identifying and reporting diseases of public health significance is well-positioned for an upgrade with “disruptive technology”. State-of-the-art machine learning and natural language processing technology have the potential to enable a novel approach. Further work is required to determine the applicability within Ontario’s provincial digital health ecosystem.


      140 Character Summary:
      Ontario is investigating an artificial intelligence solution for identification and reporting of diseases of public health significance.