A major problem in all health systems is that care often falls short of evidence based good practice. For example, studies have estimated that diagnostic error exists in 10–15 percent of cases and a study of prescribing errors found 52 errors per 100 admissions.^{1, 2} Technologies which aid clinical decision making and workflow offer substantial opportunities to reduce variation in care while improving the accuracy and speed of decision making. In the future this support will not only come from access to clinical guidance and prompts but the automated interpretation of clinical data, including genomics.

There is strong evidence that decision support tools can improve the quality of clinical decision making and some evidence it can lower costs.^{3, 4}

However, clinical decision support systems (CDSS) also bring challenges and may be met with suspicion by clinical staff who fear erosion of their autonomy. CDSS range from very passive electronic aids, such as hyperlinks to guidelines, to extremely pro-active one-click flow mechanisms. It is likely that over the coming years there will be a gradual trend of moving from advice for specific, discrete clinical scenarios to much broader-looking, proactive advice to health workers.

You come in with a stroke, the admitting physician with one click of the computer now creates the entire workflow from everything from the [real logic] to the laboratory to the diet to the physical therapy and medication.

— Robert Pearl Kaiser Permanente, US

CDSS are often combined with computerized physician (or provider) order entry systems (CPOE), through which medications, tests and procedures are ordered. When replacing hand-written, physically conveyed systems, integrated CPOE and CDSS systems can offer significant time savings and safety gains. The literature suggests a number of valuable benefits:

• between 23 and 92 percent reductionin laboratory turn-around-times⁵
• 48 percent reduction in medicationerrors⁶
• reduced need for ancillary staff.⁷

Decision support is not just for doctors, it can provide a powerful resource for other staff, particularly in settings where they may not have easy access to other clinical staff and patients.

CDSS can be further developed to build protocols into customized workflows, effectively standardizing an entire care pathway. Clinicians treating patients on this pathway therefore have unambiguous processes to follow and tasks to complete, ensuring consistently high standards of care. While this sits at the very proactive end of the spectrum, clinicians should still deviate where needed to meet individual patient needs. Indeed, at Intermountain Healthcare, where standardized clinical workflows have been developed, clinicians are encouraged to tailor the pathway to individual patients, improving system learning.

Standardized workflows can be taken a step further through “one-click flows” or “one-click ordering”. Information is pushed to a “workflow engine” to initiate the process which then sets in motion all the tasks and processes required for that specific workflow. While the productivity and care coordination benefits of such a system are obvious, “one-click flows” can be very difficult to implement. The process must be agreed across multiple teams, and even organizations, and followed up with a significant nurse training effort.

• Avoiding alert fatigue: Some hospitals average 20,000 physician alerts per month. Where there are a large number of alerts, they may be over-ridden and ignored by clinicians.⁸ There are a number of ways to address this problem, including reducing the number of alerts, making essential alerts stand out from others (through sound and colors) and requiring a response from clinicians to a particular alert (such as answering a question).
• Don’t rush into an automated technological solution: When Intermountain thinks about standardizing clinical workflows, they work on paper first. This gives them a chance to see how it works in practice and consider improvements — before integrating it into the technology.
• Use protocols as a tool for learning: No protocol will ever be right for every patient. Clinicians will need to adapt and use their own judgement. One unexpected hazard might be over compliance: Intermountain experienced problems where clinicians were too ready to adopt the default prescribing choice, leading to inappropriate care in some cases.

• Step 1: Choose a high priority clinical process and lay out an evidence-based, best-practice guideline for that process. It doesn’t have to be perfect at this stage.
• Step 2: Integrate it into clinical workflows via the EHR laying out the clinical pathway for a condition once a clear diagnosis has been made.
• Step 3: Capture data on where clinicians vary from protocol; short and long-term clinical outcomes; cost; and patient satisfaction. Intermountain put a lot of resource into analyzing this data and learning from it.
• Step 4: Tell clinicians that no protocol perfectly fits any patient, so they need to ensure they adjust the protocol based on patient need. The idea is to hold on to variation across patients and limit variation across clinicians.
• Step 5: Build in a feedback loop to constantly improve the processes.