AHRQ’s CDS “Artifacts”

William HymanWilliam A. Hyman
Professor Emeritus, Biomedical Engineering
Texas A&M University, w-hyman@tamu.edu
Read other articles by this author

The use of the term “artifacts” in the context of software and Clinical Decision Support (CDS) is new to me, but AHRQ uses it to describe a group of CDS projects that have converted practice guidelines into computer code. This terminology seems somewhat at odds with more general definitions of software artifacts which imply byproducts of the software development process rather than the software itself. Also, this usage may or may not be consistent with the more general meaning of artifact as some curious object from times past.

These AHRQ artifacts are collected at CDS Connect where there are 11 examples in 5 groups: Opioids and Pain Management (2), Childhood Obesity (1), Diabetes Management (1), Cardiovascular Disease Prevention (6), and Immunization (1). Nine of these are listed as being a “draft” and two as “active” which is described as ready for use. There is one of four “Knowledge Levels” described and identified for each artifact. These are Narrative (no examples), Semi-Structured (3), Structured (7), and Executable (1). The underlying practice guidelines are in some cases copyrighted. Converting a written flow chart or set of questions in a practice guide to code is a relatively simple exercise. The hard part is making it useful, and the even harder part is proving that it actually is useful in terms of actual outcomes improvement. The latter is something that many developers don’t want to pay much attention to.

Other divisions of the artifacts categorize the type of rule being implemented such as Event-Condition-Action (8), Risk Assessment (1), and Multimodal (2). The publishers of these artifacts include the CDC (2), The MITRE Corporation (7), Children’s Hospital of Philadelphia, and HLN Consulting LLC. Which of these artifacts has been implemented beyond a test phase, to what degree, and with what net effect is not clear.

One of the artifacts references an interesting set of value scoring definitions from the US Preventive Services Task Force which describes itself as an independent, volunteer panel of national experts, These scores are letter grades from A to D, plus I for insufficient evidence. An A means there is high certainty that the net benefit is substantial. B is a high certainty that the net benefit is moderate or there is moderate certainty that the net benefit is moderate to substantial. C is at least moderate certainty that the net benefit is small. D is harsher, meaning there is moderate or high certainty that the service has no net benefit or that the harms outweigh the benefits. One of the artifacts reports that its EHR partner won’t deploy a CDS unless it is A or B.

It is interesting that the artifacts come with a collective disclaimer, including: “While CDS identifies and describes generally recommended courses of intervention, it is not intended to substitute for the advice of a physician or other knowledgeable health care professional or provider. Individual patients may require different treatments from those specified in CDS”. There is also a general liability disclaimer. Herein lies one of the fundamental CDS conundrums. Are you or are you not meant to follow what it tells you?

This type of CDS is also the subject of a current AHRQ funding opportunity. The intent of such funding is to support “innovative research on disseminating evidence into practice through shared, interoperable clinical decision support (CDS) resources”. The criteria for funding includes the use of CQL language to develop “computable guideline recommendations”, ie to convert already available text or charts to code that can be operationalized in EHRs in a useful way. In addition, usability considerations and user-centered design are to be considered. Here it is important to remember that simply including a feature in an EHR does not automatically make it readily useable let alone helpful.

Proposals must include the “5 rights” of clinical information: delivering the right information, to the right person(s), using the right format, in the right channel, and at the right time during workflow. This is an adaptation of the 5 rights of medication use which are right patient, drug, dose, route, and time. There are also 6, 7 and 10 rights versions for medication, reminding us that reducing complex processes to short lists is almost always an oversimplification. Another 5 rights adaptation is in assigning nursing care duties where it takes the form right person, task, circumstance, directions and communication, and supervision and evaluation. This one seems to cheat on the counting a bit. And there are 5 rights of patient placement: level of care, service, nursing unit, bed and time. None of these include my rule that all short lists should include “other” to remind us that such a list has always left things out.

CDS continues to be touted for its potential value, and you may remember that CDS implementation, with some changes, has been a part of Meaningful Use since the initial roll-out. Often CDS hype skips over the “potential” proviso. Yet saying something is good is not the same as it actually being good, no matter how many times you say it.