Start Your Search

Thodoros Topaloglou



Author of

  • +

    OS06 - Innovations to Process Non Digital Data (ID 24)

    • Event: e-Health 2019 Virtual Meeting
    • Type: Oral Session
    • Track:
    • Presentations: 1
    • Coordinates: 5/27/2019, 03:45 PM - 04:45 PM, Area 2
    • +

      OS06.01 - Using Natural Language Processing for Improving Coded Data (ID 269)

      Thodoros Topaloglou, Scarborough and Rouge Hospitals; Scarborough/CA

      • Abstract

      Purpose/Objectives:
      Inpatient coding is a complex and tedious process that has not changed in the past thirty years. Health records departments are under constant pressure to meet tight timelines and compete for a shrinking pool of expert coders. Furthermore, the introduction of quality-based funding models increased the pressure on hospitals to improve quality of their data. Missed diagnoses are costly to hospitals. A possible solution is to leverage digital data: computational coding employs tools to “read” clinical documents to recognize evidence and make recommendations on coding of diagnoses and procedures at a higher level of specificity. Scarborough and Rouge Hospital (SRH) embarked on a journey with 3M in 2017 to introduce Computer-Assisted Coding (CAC) to improve coding productivity and data quality. A prerequisite for CAC is availability and access to clinical documentation and other data feeds in electronic and computer readable form. Coders are trained to use the evidence and recommendations made by the computational tools to select appropriate codes. A year later, we conducted a study to understand the impact of this tool on data quality. The study’s objective was to measure the accuracy of the codes captured and whether using this tool influenced weighted cases.


      Methodology/Approach:
      The study was conducted jointly by SRH and 3M, as a before and after intervention comparison of the number of diagnoses and procedures coded by coder and their impact on weighted cases. Multi-variable regression analysis used to measure differences in weighted cases based on variables: diagnosis/procedure count and coder. Charts were randomly assigned to coders pre- and post-implementation; length of stay and weighted cases of charts coded were similar across coders, before and after CAC implementation.


      Finding/Results:
      Number of diagnoses coded post-CAC implementation was significantly higher, as was both, the Comorbidity Levels and Resource Intensity Levels of charts coded. More importantly, there was a statistically significant increase in Resource Intensity Weights (RIW) and Health Based Allocation Model Inpatient Grouper (HIG) weighted cases. Impact on inpatient data (excl. Newborns, Pediatrics and Obstetrics) 6 months post-CAC implementation within 2017/18 FY # of Diagnoses Coded 7% Increase Comorbidity Level 3% shift from Level 0 to Level 2 Resource Intensity Level 4% shift from Level 1 to Level 2 Average RIW 2% Increase (up to 9% for one coder) Average HIG Weighted Case 5% Increase (up to 13% for one coder)


      Conclusion/Implications/Recommendations:
      Results showed clear increase in weighted cases through use of CAC; increase was significant from hospital funding perspective. We anticipate further improvements in coding efficiency; the limiting factor is quality of underlying documentation. The next phase of our journey is to embark on a clinical documentation improvement (CDI) initiative to bring these tools closer to physicians and provide evidence based and data driven tools to improve accuracy and completeness of documentation. Another exciting opportunity in our CAC roadmap is the benefit from advances in natural language processing (NLP) and artificial intelligence (AI) that are incorporated in CAC.


      140 Character Summary:
      Scarborough and Rouge Hospital implemented 3M’s Computer-Assisted Coding tool using NLP engine. Results shows increase in coding quality and weighted cases.