Definitions

Clinical decision support tools (or more broadly clinical decision support systems) provide information or direction to help clinicians at at the point of care. They come in several forms:

• Decision Support Alerts: notifications that provide information intended to influence a decision or action. These can include medication prescribing information (e.g. drug-drug interactions), laboratory information (e.g. displaying eGFR when ordering nephrotoxic medications), educational information (e.g. references or links to guidelines), or based on a care plan (e.g. a patient in levothyroxine who has not had a TSH level checked in a year might display a notification.) These alerts can be passive, meaning they are displayed to the end-user, but do not require action, or they may be active or interruptive, meaning that they require acknowledgement, often in the form of a pop-up alert.
• Order Sets: collections of orders to facilitate a standard or best evidence care practice. An example may be a Pneumonia Order Set which includes standard orders for patients admitted with community acquired pneumonia. At MCW, the use of such order sets decreases length of stay.
• Healthcare Plans: Recommendations for health care services, such as health maintenance prompt clinicians to offer recommended services when they are next due. They can track completions from other data in the EHR, such as mammographic services, colonoscopies, or other routine lab testing. Advanced cases can prompt recommended services for patients fitting specific criteria, such as a screening ultrasound for patients with hepatic cirrhosis.
• Test Interpretation: Computers can be programmed to perform complex calculations. An example of this includes calculators for Acid-Base status. More contemporary examples may include computer-aided detection for mammography or stroke detection.

Designing Effective CDS

Knowledge Management

A solid process for managing decision support alerts can help maintain the integrity and usefullness of systems. So before jumping into designing alerts, designing a process for reviewing and maintaining knowledge is important. Plan ahead for when the evidence changes, or a new medication is recommended. Ideally, an inventory of decision support should be available.

Hard coding logic into EHRs can be challenging to maintain. Using components, such as criteria blocks that can be used for multiple alerts, can promote efficiency when it comes to reviewing and updating knowledge.

Review periods, set commonly at two years (not hard and fast, and dependent on the use case) for order sets and decision support alerts should be incorporated into the CDS process. These reviews trigger content experts to review the evidence and the CDS, and changes can be made to reflect current practice. Of course, updates to CDS can be made in between review periods based upon new practice-changing information, guideline updates, or other changes in clinical practice. As alluded to earlier, content experts should be reviewing these areas for changes.

External knowledge vendors can provide products that can be incorporated into electronic health records. Commonly purchased tools include databases for drug-drug interactions, as an example. These vendors should have a similar knowledgement management process for maintaining their content. Organizations should understand what the vendor's process is, and then they can delegate that knowledge maintenance to the vendor.

Implementation Best Practices

The various stakeholders within a clinical care organization will often seek to influence decisions being made. For example, well-intentioned administrators may want to recommend less expensive medications be used over more expensive medications to keep costs down. Leaders managing insurance contracts may wish to ensure that cliicians appropriately document comorbidities such that they may receive higher risk-adjusted payments for their care. But ultimately, we strongly encourage anyone implementing CDS to design and implement the tool in such a way that clinicians find it helpful in providing the best care to their patients.

A consequence of over-alerting is alert fatigue, the concept where clinicians become numb to seeing decision support alerts. No matter how well-intended, alerts may be ignored, bypassed, and brushed aside so that the clinician may complete their intended task(s). Effective CDS systems minimize alert fatigue by ensuring that alerts are appropriately targeted, configured, and monitored.

Alert Targeting

It likley doesn't make much sense to alert an intensive care physician that a patient is due for a colonoscopy. Nor is it helpful for geriatricians to view weight-based pediatric dosing for antibiotics. It may, though, be useful to have geriatric dosing and the recent eGFR. Context really matters in designing effective CDS components.

When designing CDS, think about

• Who should see this alert? The physician/APP? The nursing staff? The revenue cycle staff?
• Which department(s) should see this alert?
• Inpatient or Outpatient?
• Does this need to interrupt the user? Or will a passive alert be satisfactory?

The 10 Commandments.

View Bates et al

1. Speed is everything
2. Anticipate Needs and Deliver in Real Time
3. Fit into the User’s Workflow
4. Little Things Can Make a Big Difference
5. Recognize that Physicians Will Strongly Resist Stopping
6. Changing Direction Is Easier than Stopping
7. Simple Interventions Work Best
8. Ask for Additional Information Only When You Need It
9. Monitor Impact, Get Feedback, and Respond
10. Manage and Maintain Your Knowledge-based Systems

Monitoring and System Feedback

How will you know if your CDS is effectively working? Further, how will you know if it ever stops working? And are you providing value, or noise? Monitoring CDS systems is an essential part of the implementation and maintenance.

One recommendation is to run CDS in silent mode for a few days or weeks before your launch. This will enable you to track the number of times it is firing in the production system. Is the number of triggers expected? Are there departments or situations where the alert is not appropriately being triggered?

After go-live, monitoring of your CDS triggers can also prompt if there are any failures in the CDS that might go undetected.

Vignette

In one hospital, a clinical decision support alerted pediatricians when lead screening was to be due. It functioned by looking at the age of the patient and then looking for lead screening results in laboratory studies, and would prompt a notification if a lead screen had not been completed when it was due. A laboratory information system upgrade changed the coding assigned to the lead testing. When the CDS was looking for lab samples, it no longer properly identified them. The result was that clinicians were not prompted to order lead screening, and patients went without this recommended test. [!img http://informatics.bidmc.org/files/dcifellows/files/amia_2015_poster5.pdf]

Impact of CDS on Clinical Care

CDS has been shown to make important contributions to clinical care.

• Medication error reduction
• Quality improvement

Watch this brief segment for an example of CDS benefits when it comes to medications👇:

Class Exercise

Mammographic Screening

• You have noticed that some of your patients at your organization may not have received their screening mammograms. You used reporting tools within the EHR to note that of patients 40-50 25% of patients had screening mammograms, and of patients 50-70 45% of patients had a screening mammogram at least once every two years. Further, your colleague who manages insurance contracts notes that you are not performing well in your quality contracts.
• You untertake an effort to improve these screening rates. How would you design the decision support function? How would you propose to monitor for effectiveness?

In your plan, please consider

1. What is your goal, and how will you measure it?
2. Who is the audience for your CDS?
3. How would you define your logic?
4. How will you monitor your CDS program?
5. What is your plan for knowledge management?

Stop the Bleeding

• Your organization received word from the Red Cross that your transfusion costs were increasing. When your analysts looked into the matter, they discovered that the transfusion rate per 100 patient-days was higher than expected benchmarks.
• When doing some preliminary invesgitation, you identified a few reasons that may explain the discrepancy, including your organization's role as a regional cancer care center and the only region's level I trauma center.
• After undertaking a chart review, you identified that some patients had transfusions that were not evidence-based. For example, patients with a stable hemocrit of 24% were sometimes transfused. A clear reason for transfusion was not immediately discernable in the chart.
• You are seeking to add CDS to transfusion orders.

In your plan, please consider

1. What is your goal, and how will you measure it?
2. Who is the audience for your CDS?
3. How would you define your logic?
4. How will you monitor your CDS program?
5. What is your plan for knowledge management?