Last week, I joined an amazing group of colleagues at the National Quality Forum's Healthcare Information Technology Expert Panel to work on a next generation quality data set. They key breakthrough was the development of a universal terminology for the design of quality measures which captures process and outcome data from electronic systems.
Elements which are captured include:
Datatype (e.g., medication order)
Data (e.g., aspirin)
Attributes (e.g., date/time)
Data Source (e.g., physician, patient, lab)
Data Recorder (e.g., physician, lab, monitor)
Data Setting (e.g., home, hospital, rehab facility)
Health Record Field (e.g., problem list, med list, allergy)
In the original HITEP work last year, 35 datatypes were defined such as encounter, diagnosis, diagnostic study, laboratory, device, intervention, medication, symptom etc. Each datatype can have subtypes describing specific events. Here's an example of the subtypes of the medication datatype
medication administered
medication adverse event
medication allergy
medication discontinued
medication dispensed
medication intolerance
medication order
medication prescribed
medication offered
medication refused
A traditional measure of quality might be
"Was Aspirin administered within 5 minutes of ED arrival in diagnosis of acute MI?"
If an EHR transmits datatypes for encounter, diagnosis, and medication to a quality data warehouse, we could capture the following data:
Datatype - encounter
Data - ED arrival
Attribute - date/time of arrival
Source - registration system
Recorder - ED ward clerk
Setting - ED
Health record field - ED arrival date/time
Datatype - diagnosis
Data - MI
Codelist - SNOMED code 12345
Attribute - date/time of diagnosis
Source - physician
Recorder - physician
Setting - ED
Health record field - encounter diagnosis
Datatype - medication administered
Data - ASA
Codelist - RxNorm code 123456
Attribute - date/time of administration
Source - nurse
Recorder - nurse
Setting - ED
Health record field - medication administered
then the quality measure could be defined as
Diagnosis="SNOMED 12345" AND (medication administered="RxNorm 123456" date/time - ED arrival encounter date/time) < 5 minutes
Such an approach makes quality measures more clearly defined, more directly related to data elements in EHRs, and more easily maintained.
The next steps for NQF include review of their existing 500 quality measures to determine which could be placed into such a framework. If there are gaps or revisions needed, the NQF will work with quality measure development organizations.
Meaningful use of EHRs will likely include quality measurement. Having a framework for recording quality data and computing measures is foundational.
Subscribe to:
Post Comments (Atom)
3 comments:
Interesting information. It would appear that a similar structure could be used for HIE.
What is the timeframe for reviewing the ~500 quality measures?
The elements and sub-elements seem like reasonable tools for extraction, but at the grass roots this type of discrete data regimentation would cover over significant data either missed by category or because of free text. With most care delivered by physicians in small practices with a variety of documentation styles, that might cause the law of unintended consequences to enter the playing field.
I would like to see an attribute of certainty or reliability... something which addresses how reliable the info is... something like Possible/Probable/Definite or some such.
Wondering what folks think of that.
Post a Comment