Clinical analytics
A clinical dashboard is only as honest as its definitions
When two reports disagree on the same measure, the cause is rarely the chart. It is usually a definition nobody wrote down.
Take a measure that sounds simple: the 30-day readmission rate. Before anyone can calculate it, several decisions have to be made.
- Does a transfer between hospitals count as a discharge?
- Are planned readmissions, such as a scheduled procedure, included?
- Do patients who died during the first stay stay in the denominator?
- Does the 30-day period start on the day of discharge or the day after?
Each answer can be reasonable, and each changes the number. Two teams can both be correct and still bring different figures to the same meeting.
Write the measure down
Give every measure one specification: its numerator, its denominator, its exclusions, its time window and its data sources. Where an external program already defines the measure, follow that definition and record any local differences. Show the definition next to the number so that readers can check it themselves.
Standardize the codes underneath
The same concept can arrive in several forms from different sources. Mapping to shared vocabularies, such as ICD-10 for diagnoses, LOINC for laboratory tests and SNOMED CT for clinical terms, makes a measure comparable. Keep the original value too, because mappings can be wrong and need correcting.
Make every number traceable
A clinician who doubts a figure will ask which patients are behind it. Where their role permits, let them follow a number down to the underlying records. Keep the lineage from source to measure, and show when the data last refreshed, so nobody treats last week's data as today's.
Put a person in charge of each measure
Definitions change as clinical guidance and coding rules change. Each measure needs an owner who approves changes and versions them, so a shift in a trend line can be explained as either a real change in care or a change in the calculation.