Imaging Appropriateness Criteria - Brief Proposal
1. Proposed Workitem: Imaging Appropriateness Criteria Profile
- Proposal Editor: Keith Dreyer/Kevin O'Donnell
- Editor: Mike Mardini
- Domain: Radiology, Cardiology
2. The Problem
Research [1] has indicated that it is not unusual for referring physicians to order radiology studies that are either inappropriate, or less appropriate than other studies. Inappropriate studies increase costs, in the form of resources that could be better used elsewhere, reduce quality, in the form of providing sub-optimal information for clinical decisions, and potentially impact safety due to the potential risks of procedures that are invasive or use ionizing radiation.
The root of the problem is that there are hundreds of imaging procedures spanning a dozen modalities, each with their own strengths and weaknesses. Technology development continually adds to the list of procedures, and clinical research continually adds to our understanding of the strengths and weaknesses of each.
Over the years, groups such as ACR and ACC have developed appropriateness criteria and/or guidelines for the appropriate use of imaging. Such criteria distill the knowledge of panels of imaging experts about what procedures are most likely to provide clinically useful information for the least necessary risk for a wide range of clinical indications.
However, actually putting such criteria into effect faces significant hurdles. Traditional approaches involve making documents available for individual physicians to read, understand, remember and apply in future orders. Problems with this include the limited available bandwidth of the physicians, the fact that guidelines are regularly being updated, and challenges distributing guidelines.
Pilot projects [2] have demonstrated that implementing guidelines in the form of regularly updated decision support provided to the referring physician at the time of ordering can have a significant impact on the quality, cost and safety of care.
3. Key Use Case
After seeing a patient, the referring physician orders a radiology procedure which is submitted for performance. Unfortunately, their procedure selection is not based on specialist knowledge in radiology or the most recent information on best practices.
In some cases, the radiology staff is too overloaded to review all the ordered procedures so they are simply performed. In other cases, the radiology staff spends time reviewing the orders, observes an inappropriate procedure order and spends more time tracking down the ordering physician, and spends time discussing the merits of a better procedure. Making all orders appropriate in this fashion is not optimal use of everyones time. In other cases, faced with significant numbers of inappropriate orders, the payer burdens the ordering and performing physicians with a pre-authorization or other type of order review process. Again, sub-optimal.
ACR and other pilot projects have implemented a better approach:
- The referring physician orders radiology procedures.
- The Order Placer sends the ordered procedure and patient details (indications, etc) to the Guideline Server.
- The server scores the ordered procedure (based on the patient details) and assembles a list of "better" alternative procedures
- The server returns the list to the Order Placer.
- Based on internal logic, the Order Placer places the order (e.g. in the case of a good score), or else presents the referring physician with a list of higher ranked alternative procedures along with links to the supporting evidence or the option to contact a radiologist.
- The referring physician modifies the order as appropriate.
4. Standards and Systems
The solution would likely involve the Order Placer, a Guideline Server and other EMR systems.
HL7 has messaging for Admission, Ordering and Patient Record details including coding of indications and procedures.
Proof of concept exists in the form of several projects that have prototyped such a system of servers and Order Placer features.
ACR has prototyped an (open) web API for communication between a Guideline Server and an Order Placer. ACR has set up a server that uses the web API to provide relevant pieces of the ACR Appropriateness Criteria in XML format as a fee-based web service.
5. Discussion
An IHE Profile could accelerate adoption by ordering systems and facilitate testing of both clients and servers. A common interface would allow easy incorporation of various guidelines (ACR, ACC, etc).
The profile would likely involve a query transaction from the Order Placer actor to a Guidelines Server actor, and possibly information feeds (admitting diagnosis, indications, patient allergies, history, etc) to the Order Placer from relevant source actors.