I’m prototyping a number of electronic clinical quality measures (eCQM) in Atlas to get a directional sense of how our hospital system is doing and how they are likely to score on upcoming CMS measures.
I know the Atlas rendition will never be perfect, but this can give us leading measures - often well in advance of when our trusted eCQM vendor partners can implement the logic themselves. Additionally, it lets us do additional analyses on the data not always available from those vendors (such as filter/aggregate/trend the cohort and metric (ratio of numerator/denominator) by hospital, provider, time, demographics, etc.)
Attached is a draft cohort definition of the Hospital Harm - Hyperglycemia measure (HH-Hyper) - CMS871v3. It is the Numerator measure, but also includes all of the logic for the Denominator.
HH-Hyper (eCQM CMS871v3) Numerator.json.txt (446.2 KB)
Several questions for the community:
- Are others prototyping electronic clinical measures? If so, what is a good environment for sharing those efforts (e.g. having people submit proposed cohorts so that they can be peer reviewed & vetted)?
- Best practices and documentation on representing sequential measurement in Atlas? The logic requires looking for two days with glucose >= 200 mg/dL followed by a day with no glucose measurement (but still an inpatient). I took a stab at this, but may have encoded it incorrectly. If there aren’t examples like this in the Book of OHDSI, that might be a nice addition.
- Working with datetimes. The logic (and that of multiple new eCQM) requires calculation of time from admission in hours. For example, there is an exclusion for very high glucose in hour before or 6 hours after admission. The logic also uses rolling 24 hour periods as days, starting from the actual time of admission. Are others finding similar need for datetime logic? I’ll submit a feature request to Atlas to support this and reference it in a follow-up post.