Introduction to Authoring Clinical Practice Guidelines (CPG)

Introduction to Authoring a Clinical Practice Guideline (CPG) as a BPM+ Model

Clinical practice guidelines (CPGs) are a common way of documenting medical knowledge and best practices. They can be quite diverse in scope and content. Implementing a CPG involves a sequence of tasks that need to be completed before the model is ready for clinical use.

Step 1: Choose wisely.

No two guidelines are the same. CPGs are created to meet a clinical goal and are not designed to be easily implemented in a model. Some are well-defined and focused. Others (for example, the 2018 AHA/ACC guideline on congenital heart disease in adults) cover a large number of topics which makes modeling difficult.

Step 2: Clearly state what your goals are.

Are you trying to improve outcomes? Improve adherence? Reduce costs? These goals will later be the benchmarks that will be used to determine the success of your implementation.

Step 3: Be prepared for gaps and traps.

Many CPGs are generated top-down. In a large CPG this can lead to gaps as the content expands. Sometimes terms are ambiguous. A model requires explicit use of tasks and terms. Omissions in the documentation may not become apparent until an attempt is made to generate the model.

Step 4: Identify the DMN models that are required.

DMN models can be used to represent tables, lists, equations or rules. These can be identified while reading through the guideline carefully.

Step 5: Identify the processes involved.

Once you are familiar with the guideline then it is necessary to identify the various processes involved. Often a specialty will show a standard pattern. For example, processes in surgery include determination of eligibility, prehabilitation, preoperative preparation, the surgery itself, early postoperative recovery, late postoperative recovery, discharge and follow-up.

Step 6: Identify the key decision points that require clinician input.

While it is tempting to completely automate a model or to make prescriptive decisions, these require FDA certification of the model before it can be used clinically. If the clinician makes the key decisions, then regulations are more lenient. It is therefore necessary to identify key tasks where the clinician approves, rejects or modifies a decision.

Step 7: Design artifacts that are modular and reusable.

Many CPGs show patterns or require the same DMN models. Making models modular makes them reusable, which speeds development and improves user comprehension.

Step 8: Assembling the separate processes.

After you have designed all of your models then you need to decide how to bring them together. There are 3 options when faced with multiple models. One is to keep them separate. A second is to make them subprocesses to an overarching BPMN model. The third is coordinate them in a CMMN model.

Step 9: Optimize data use to simplify implementation.

The goal is to keep manual data entry to a minimum. This can be achieved by interfacing the model to the electronic health record (EHR) with FIHR. Some clinical inputs are more FIHR-friendly than others, and it may be necessary to revise models to simplify the interface.

Step 10: Prepare test data sets

Once the models are prepared it is necessary to generate sets of test data to use when testing. These should be sufficient to demonstrate how well the model functions. These should include data at the extremes. It should also include data that should not be accepted.

Step 11: Test the model, measure outcomes and retain records.

Testing is done initially, after modification or with significant software changes. The results of testing as well as administrative review should be retained long-term for regulatory compliance.

Step 12: Develop a maintenance and revision schedule.

A schedule should be developed for when to review and update the model. Criteria should be developed for when to retire the model.