Managing Healthcare Data

As you read in McConigle, there is a continuum as we move from data to information to knowledge to wisdom. With the incorporation of computers and informatics, we need to be able to manage this continuum in such a way that they can be shared. Otherwise, the data / information / knowledge / wisdom are trapped within a single clinical information system (CIS) or within a single organization. This is essentially the "interoperability" problem that we hear about so much these days.

While interoperability is usually framed in the context of sharing data between organizations or between electronic health records, the problem exists at multiple levels:

1. Clinicians from different departments within a single medical center may not be able to effectively share data due to different information systems, protocols/policies, and even terminology/vernacular.
2. Within a single organization such as Sutter Health or Kaiser, there are many different deployments/instances of an information system (e.g. EHR) with customizations and modifications that make interoperability difficult.
3. Given the structure of the healthcare industry, there is often an incentive to prevent the sharing of data even if it were technologically possible.

The solution is a combination of data standards and data management approaches that generally allow informaticists to:

1. Manage the transition of data to information to knowledge to wisdom
2. Share data/information/knowledge/wisdom between information systems and organizations.

Much of our discussion will revolve around the technology behind data management and interoperability but keep in mind that the solution will likely include changes in bureaucracy, policies, and other socio-political issues.

Grading

This assignment is worth 10 points or 10% of your overall class grade.

Due Date

Friday, April 10, 11:59:59 PM

Late assignments will not be accepted.

Reading

McConigle, pp. 135-137 (previously assigned in Assignment 1)

Lim et al., Chapter 1 - Computational Knowledge and Ontology - PDF available here

Benson, Chapter 2 - Why Interoperability is Hard - PDF available here

Cimino, Desiderata - PDF available here

Lab

For this assignment, you will be working both individually and as a team.

First, as a group, choose a domain topic for which you have previously or would like to implement some sort of clinical decision support tool. This CDS tool can be any type of decision support whether it's to help decrease length of stay or readmissions, to assist at the bedside, or to help with any other clinical decision-making. The key requirements are that the CDS tool uses existing data from a clinical information system as input. The CIS can be any system ranging from an hospital-wide EHR to a department specific system. Ideally, this CDS tool is one with which everyone on your team is familiar.

Second, individually, create a basic terminology/ontology for the above domain. Initially, create a list of terms and concepts that represent the data that your CDS tool requires. For example, if creating a bedside arrhythmia detection tool, you would want to add those terms/concepts that capture specific data your CDS tool would need. For example, the existence of certain prior diagnoses within the patient's medical history may lead the CDS tool to determine that a particular rhythm is a false positive, thus ignoring it.

• Specific cardiac medications (e.g. Digoxin)
• Specific cardiac diagnoses (e.g. congestive heart failure)
• Arrhythmias (e.g. V-Fib, SVT, etc.)

Third, reconvene as a team and compare and discuss your various lists:

• Which concepts occurred in more than one list?
  • Did you use the same terms/words/acronyms? For example, did someone write down "myocardial infarction" vs. "MI" vs. "heart attack"?
• Which concepts only occurred on one person's list?
  • For those who included it, why?
  • For those who did not include it, why?
  • Oversight?
  • Low prevalence/significance?

Fourth, take all of your lists, determine which concepts are the same, similar, or directly related to one another:

• For those that are the same, the mappings are simple - you essentially have a list of synonyms
• For those that are similar, you will need to create mappings between the terms
  • Are the terms similar enough that there is a one-to-one mapping? In other words, you can use them interchangeably and the effect on any downstream analytics is negligible.
  • Is one term a super set of the other? For example, "myocardial infarction" and "cardiac diseases" are related such that MI's are a subset of cardiac diseases

Last, discuss among your team members your experience in mapping the terms:

• What difficulties did you encounter?
• Did you find the need to discuss any details of the terms/concepts during the mapping phase?
• Did this exercise influence your perspective on how to capture, store, or manage healthcare data?

Deliverables

For this lab, you may submit your lab in any format - outlines, paragraphs, diagrams, etc. as a PDF uploaded to Canvas.

During the intensive, we will share and discuss your deliverables between the groups. Please do not discuss or share between groups. I find it very interesting to see the many different approaches people take when trying to capture knowledge and expertise that has been developed over years of training and practice.