Saturday, April 3, 2010

HIT: Challenges from the Data Perspective

Patient health and medical information generated in a healthcare organization can be categorized by various features. For example, it can be

Transactional, related to a separate event, such as an ADT, exam, order, procedure and so on
Summary, describing the development and status of a condition, family and social history, providing a problem and medication list, etc.

From the vocabulary interoperability standpoint, each piece of information is either coded or uncoded. Coded entries, e.g., diagnoses, always contain the code value from a coding system and the name of the system, as in the following excerpt from a CCD:

<observation classCode="COND" moodCode="EVN">
  <id root="abbbdd09-ae9f-4879-aa80-3893736973d1" />
  <code code="346.2" displayName="Migraine Headache, Variant" codeSystem="2.16.840.1.113883.6.103" codeSystemName="ICD9CM" />
</observation>

Uncoded data is usually a narrative in a free text form, for instance:

DISCHARGE DIAGNOSES
Extruded disk, L4-5, with right L4 and L5 radiculopathy.

From the semantic interoperability and automatic processing point of view, each record can be structured or unstructured. HL7, CCR and CCD-formatted data are good examples of structured records. Images, scanned and text documents represent unstructured ones. Be aware, though, that structured records may, and often do include unstructured pieces. For instance, an HL7 message can contain a free-text report or note.

The rationale behind the HITECH Act and Meaningful Use was not as much about lack of patient information, as about its availability (with privacy and security protection, of course) and usability. Computer programs are very good at categorizing, grouping and sorting electronic records as long as the logical model of the data includes standard codes for diagnoses, medications, problems, etc., and links labs, reports and clinical documents to respective encounters. The challenge though, is to get all these attributes populated.

The Stage 1 Criteria for Meaningful Use make a serious effort towards improved usability by requiring recording patient demographics, problem lists, current and active diagnoses, medications and medication allergy lists and lab test results as structured and largely coded entries. It seems relatively straightforward with regard to patient demographics, diagnoses, problems lists and electronic prescriptions, even though, for example, very few if any EHR systems use the CDC Race and Ethnicity Code Set. The HIT Certification programs together with the Meaningful Use Criteria will eventually encourage EHR system vendors and users to implement standard vocabularies, especially, since the Unified Medical Language System (UMLS) is now available at no charge in the U.S. Interestingly enough, there are no similar requirements for lab systems. The hope is that healthcare providers trying to become meaningful users will gradually steer clinical labs in the same direction. But for now we can expect a lot of activity building and customizing integration interfaces to get lab test results structured, UCUM-normalized and standardized.

The move towards structured and coded EHR will be a huge deal even just going forward. There is an enormous volume of historic data, some of which has not yet been digitized at all. It has to be dealt with in a smart fashion, based on record holders priorities and resources. But this topic deserves a separate discussion.

No comments:

Post a Comment