Health Data Analytics and clinical terminology systems like SNOMED CT by Prof. Supten Sarbadhikari, @supten

Healthcare, with its inherent complexity, deals with large volumes of data coming in. Well designed and used EMRs (Electronic Medical Records) can collect huge amounts of data. However, neither the volume nor the velocity of data in traditional modern healthcare may qualify as big data now. Only a small fraction of the tables in an EMR database may be relevant to the current practice of medicine and its corresponding analytics use cases.



Certainly there will be variety in the data, but most of the EMR systems collect very similar data objects and models. However, new use cases supporting genomics and Internet of Medical Things (IoMT) will certainly require a big data approach.

Healthcare (data) analytics describes healthcare analysis activities that can be undertaken as a result of data collected from four areas within healthcare:

  • Claims and cost data
  • Pharmaceutical and research and development (R&D) data
  • Clinical data (collected from electronic medical records (EMRs))
  • Patient behavior and sentiment data (patient behaviors and preferences)


In health information management (HIM) — and in coding, specifically — the HIM professional must understand the importance of their role in interpreting and abstracting the data to be collected and analyzed. In other words, (health) data literacy is essential. While this data is used primarily in reimbursement and claims activities, it also plays a much larger role in clinical data analysis performed in facilities for quality of care reporting, disease management, and best care practices.

HIM Professionals are implementing coding data analytics to continually monitor their coding teams and cost-justify ongoing educational investments. Coding data analytics is a long-term commitment to improve coding performance for productivity and accuracy.

Elements that impact coding productivity data include: the type of electronic health record (EHR) used, the number of systems accessed during the coding process, clinical documentation improvement (CDI) initiatives, turnaround time (TAT) for physician queries, and the volume of non-coding tasks assigned to coding teams.

Accuracy should never be compromised for productivity. That may lead to denied claims, payer scrutiny, reimbursement issues, and other negative financial impacts. Instead, a careful balance between coding productivity and accuracy is considered best practice. Both data sets must be assessed simultaneously. The most common way to collect coding accuracy data is through coding audits and a thorough analysis of coding denials.

The fully-electronic 11th edition of the International Statistical Classification of Diseases (ICD-11) from World Health Organization (WHO) contains (epidemiological or morbidity and mortality causes) 55,000 codes, compared to the 14,400 in ICD-10.

SNOMED CT (India is a member of SNOMED CT and therefore it is available for use by anyone in India, free of cost) from SNOMED International contains 311,000 clinical concepts (including anatomical sites, disease diagnosis and procedures), with their descriptions, and more importantly, poly-hierarchical relationships.

To quote SNOMED CT:
“SNOMED CT is a clinically validated, semantically rich, controlled terminology designed to enable effective representation of clinical information. SNOMED CT is widely recognized as the leading global clinical terminology for use in Electronic Health Records (EHRs). SNOMED CT enables the full benefits of EHRs to be achieved by supporting both clinical data capture, and the effective retrieval and reuse of clinical information.

The term ‘analytics’ is used to describe the discovery of meaningful information from healthcare data. Analytics may be used to describe, predict or improve clinical and business performance, and to recommend action or guide decision making.

Using SNOMED CT to support analytics services can enable a range of benefits, including:

  • Enhancing the care of individual patients by supporting:
    • Retrieval of appropriate information for clinical care
    • Guideline and decision support integration
    • Retrospective searches for patterns requiring follow-up
  • Enhancing the care of populations by supporting:
    • Epidemiology monitoring and reporting
    • Research into the causes and management of diseases
    • Identification of patient groups for clinical research or specialized healthcare programs
  • Providing cost-effective delivery of care by supporting:
    • Guidelines to minimize risk of costly errors
    • Reducing duplication of investigations and interventions
    • Auditing the delivery of clinical services
    • Planning service delivery based on emerging health trends


SNOMED CT has a number of features, which makes it uniquely capable of supporting a range of powerful analytics functions. These features enable clinical records to be queried by:

  • Grouping detailed clinical concepts together into broader categories (at various levels of detail);
  • Using the formal meaning of the clinical information recorded;
  • Testing for membership of predefined subsets of clinical concepts; and
  • Using terms from the clinician’s local dialect.


SNOMED CT also enables:

  • Clinical queries over heterogeneous data (using SNOMED CT as a common reference terminology to which different code systems can be mapped);
  • Analysis of patient records containing no original SNOMED CT content (e.g. free text);
  • Powerful logic-based inferencing using Description Logic reasoners;
  • Linking clinical concepts recorded in a health record to clinical guidelines and rules for clinical decision support; and
  • Mapping to classifications, such as ICD-9 or ICD-10, to utilize the additional features that these provide.


Analytics tasks, which may be enabled or enhanced by the use of SNOMED CT techniques, can be considered in three broad categories:

  1. Point-of-care analytics, which benefits individual patients and clinicians. This includes historical summaries, decision support and reporting.
  2. Population-based analytics, which benefits populations. This includes trend analysis, public health surveillance, pharmacovigilance, care delivery audits and healthcare service planning, and
  3. Clinical research, which is used to improve clinical assessment and treatment guidelines. This includes identification of clinical trial candidates, predictive medicine and semantic searching of clinical knowledge. 


While the use of SNOMED CT for analytics does not dictate a particular data architecture, there are a few key options to consider, including:

  • Analytics directly over patient records;
  • Analytics over data exported to a data warehouse;
  • Analytics over a Virtual Health Record (VHR);
  • Analytics using distributed storage and processing; and
  • A combination of the above approaches.


Practically all analytical processes are driven by database queries. To get the most benefit from using SNOMED CT in patient records, record-based queries and terminology-based queries must work together to perform integrated queries over SNOMED CT enabled data. To this end, SNOMED International is developing a consistent family of languages to support a variety of ways in which SNOMED CT is used. Clinical user interfaces can also be designed to harness the capabilities of SNOMED CT, and to make powerful clinical querying more accessible. Innovative data visualization and analysis tools are becoming more widespread as the capabilities of SNOMED CT content are increasingly utilized.”

Therefore, now it is possible to make every kind of analytics and reporting results much more detailed than it used to be. Billing and coding companies that have adopted predictive analysis tools have received a considerably higher value return from mining their data.

HIM professionals must encourage the administration and policymakers to adopt SNOMED-CT enabled systems to get better informed and analyzed outcomes.

References:

1. https://www.healthcatalyst.com/big-data-in-healthcare-made-simple
2. https://www.healthcareittoday.com/2017/11/15/opening-the-door-to-data-analytics-in-medical-coding-him-scene/
3. https://www.osplabs.com/insights/data-mining-in-medical-coding-and-billing/
4. https://bok.ahima.org/doc?oid=302591#.Xj-2xvkzbDc
5. https://confluence.ihtsdotools.org/display/DOCANLYT/Data+Analytics+with+SNOMED+CT
6. https://confluence.ihtsdotools.org/display/DOC
7. https://jbcr.net.in/JBCR-VOL-6-issue-1-2019-20/current-issues-volume-VI-issue-1-1.html

AUTHOR PROFILE:

Prof. Supten is a Health Informatics Educationist and Independent Consultant for Digital Health Standards, with basic training as a medical doctor and a PhD in Biomedical Engineering from IIT, BHU. Formerly Dean (Academics and Student Affairs) and Professor (Health Informatics) in IIHMR, Delhi. Formerly: Project Director, CHI of National Health Portal. He has been the Founder and Director of Supten Institute and the Founding Director of CAL2CAL Institute. Prof. Supten’s areas of expertise are in Requirements analysis and management in Healthcare information technology, e-learning; medical journalism and clinical decision support systems.