Clinical Data Disorder: Symptoms, Prognosis and Treatment
President and Chief Strategy Officer
Background and Symptoms
Clinical Data Disorder is a chronic but treatable condition acquired through integrating multiple sources of data without addressing an underlying pathology: unstructured and non-standard data. There are over 4 million clinicians documenting care in the United States. Many use different EHRs. Many others use the same EHR differently. And most learned to practice medicine before EHRs even existed.
Do the math. This creates enormous variation in the underlying source data.
When clinical data from various sources are combined without addressing variations in the source clinical documentation, clinical data disorder results. This disorder spans all major domains of clinical data including allergies, care plans, encounters, immunizations, medications, problems, procedures, results, social history and vital signs. For example, a platelet count may be documented with all the following units of measure:
While none of these are actually the preferred unit of measure according to standards (i.e. “10*3/uL”).
The most common symptoms of clinical data disorder include:
- Limited ability to query clinical data
- Low productivity and lethargy of data users
- Slow response time to business needs
- Hair loss among clinicians and analysts
- Executive dysfunction
Without intervention, clinical data disorder is progressive. As more data sources are added, each business function adds new workarounds and patches to deal with the non-standard data. The resulting clutter and complexity increases until maintenance requires additional personnel. While these employees want to perform analytics that improve care outcomes, they spend most of their time wrangling dirty data.
The long-term prognosis for untreated clinical data disorder is reduced operational efficiency and poor care quality. While groups of highly caffeinated individuals can cover-up symptoms for a short-period of time, manual efforts of data mapping and database clean-up will result in long-term fatigue and consequent errors. With over 2 million codes in various terminologies (SNOMED, ICD-9, ICD-10, CPT, LOINC, RxNorm, NDF-RT, CVX), only automated solutions are able to deliver sustainable treatment for clinical data disorder. Moreover, mapping semi-structured text and varying interoperability formats are best done by technology – not by hand.
The treatment for clinical data disorder includes industry leading technology. Effective treatments scale well with additional data sources. At Diameter Health, our clients normalize hundreds of thousands to millions of data elements daily to reverse the pathology of non-standard data. We recommend three steps to treat clinical data disorder:
- Transform and Normalize the data
- Parsing technology that works across 20+ formats of clinical extracts from certified EHRs
- Automation to bring multiple terminologies to common standards
- Targeted Natural Language Processing (NLP) to assign standard codes from unstructured or semi-structured text
- Organize and De-duplicate
- Regroup inbound data into appropriate clinical domains
- Unify like concepts from multiple terminologies
- De-duplicate data to eliminate redundancy. Intelligent de-duplication improves comprehension and enables a consolidated longitudinal view of a patient’s history
- Tag data with logical properties that streamline data query and presentation
- Infer missing data such as laboratory interpretations, drug classes and diagnostic grouping of problems and procedures
- Enriched data provides the basis for advanced data quality scoring
The result is reliable, robust substrate data ready for query, analytics, quality measurement and population health.
While the ICD-10 code for clinical data disorder is yet to be assigned, it’s much more prevalent than other conditions for which ICD-10 codes have been approved (for example, see “V91.07XD” burn due to water-skis on fire, subsequent encounter).