Technology optimization: refining clinical decision support

By | November 18, 2019

Physicians on the front lines of healthcare delivery welcome the assistance of clinical decision support technology – as long as it works well, gives proper suggestions and does not get in their way.

Perhaps more than other healthcare information technologies, clinical decision support systems really require CIOs, CMIOs and other health IT and clinical leaders to make sure the tech is highly optimized for ideal performance. Not to mention optimized to meet the specific needs of the organization.

Here, five experts in clinical decision support offer their best practices for healthcare provider organizations to optimize the technology so that physicians and other clinicians get the most out of this key health IT investment.

What is the aim of the technology?

An early best practice is to understand what the organization and, more specifically, the physicians are trying to accomplish with the clinical decision support technology, said David Lareau, CEO of Medicomp Systems, a vendor of physician-driven, point-of-care technology that optimizes EHRs and other health IT.

“For example, the clinicians want tools to help them treat patients with complex medical conditions or chronic diseases and guide them to order the right tests or prescribe appropriate therapies,” he explained.

“CIOs would need to incorporate the appropriate clinical decision support tools where physicians need them within their clinical workflow, whether it’s during the intake process, at the point of care or after the patient encounter.”

David Lareau, Medicomp Systems

“CIOs would need to incorporate the appropriate clinical decision support tools where physicians need them within their clinical workflow, whether it’s during the intake process, at the point of care or after the patient encounter.”

Consider the treatment of patients with juvenile idiopathic arthritis (JIA), an autoinflammatory disease, he offered.

“This complex condition requires physicians to regularly capture details on 71 individual joints,” Lareau explained. “By making clinical decision support tools available in the documentation process, clinicians have point-of-care access to information that can help assess disease progression and identify gaps in care.”

CIOs would need a different strategy if physicians did not want or need detailed clinical guidelines about how to diagnose and treat patients with certain problems, but instead required tools to support clinicians’ own decision-making processes, he said.

“For example, physicians may want clinical decision support tools that provide ready access to a specific patient with a specific problem so they can efficiently and effectively evaluate a patient’s current status and prescribe the most appropriate therapy,” said Laureau.

“To optimize physician workflows, CIOs would need to implement the clinical decision support technology that provides instant views of pertinent lab results, medications and other relevant data without requiring clinicians to hunt through each patient’s complete medical record.”

In summary, the key to deploying clinical decision support tools that enhance decision-making and are embraced by physicians is to start by understanding what users want and need, he added.

Do not interrupt clinician end-users

On a related note, CIOs must be mindful to keep the needs of their clinician end-users in mind so that the clinical decision support tools enhance – and do not interrupt – clinical decision making, Lareau said.

“For example, if a physician has already ordered specific tests or medications for a patient’s current problem, he or she will find it disruptive to have to read through clinical guidelines recommending the same therapies,” he said.

“If the clinical decision support tools don’t consider a patient’s complete health profile,” he added, “including current medications and relevant test results, a physician may spend countless hours filtering through extraneous information to determine a proper treatment plan.”

Ideally, clinical decision support systems should consider medicine’s many complexities, such as possible interactions between an existing medication and a newly prescribed medication, or all the reasons a certain lab results might be abnormal, Lareau said.

“Clinical decision support tools should provide the physician with well-organized data that is relevant to a patient’s current problem, while filtering out information that does not improve clinical decision making,” he said. “With the right clinical decision support tools, physicians can efficiently and effectively make clinical decisions to improve patient care and support quality outcomes.”

A ‘patient-first’ approach: more background data

FDB, also known as First Databank, is a vendor of drug and medical device information for clinical decision making. It engages with CMIOs and other healthcare professionals on a regular basis, and what its executives hear is that the healthcare pros are seeking to improve and optimize available clinical decision support systems while trying to stay abreast of emerging technologies and evolving medical practice.

“The exponential growth of EHR implementations has led to a massive amount of medication information being electronically directed at healthcare providers,” said Dr. Charles Tuchinda, president of FDB and Zynx Health and executive vice president and deputy group head of Hearst Health. “While its relevance continues to be debated, the sheer volume of this content has contributed to alert fatigue.”

“The exponential growth of EHR implementations has led to a massive amount of medication information being electronically directed at healthcare providers.”

Dr. Charles Tuchinda, FDB

Analysis of clinical workflows and medication decision support provide many examples of over-alerting, under-alerting and insufficient information for decision making, he explained. This has increased the probability that providers will ignore potentially important alerts and simultaneously be presented with irrelevant information presented at the wrong time, he added.

“The next step is a ‘patient-first’ approach that leverages additional patient inputs – such as lab results, genomic markers and clinical scoring – to present meaningful medication decision support targeted at a specific patient at the most appropriate time and place in the clinical workflow,” he suggested.

“In this best practice scenario, alerts will only trigger when the additional data inputs confirm that it is pertinent to the patient. Since the guidance is delivered at the right point in the provider’s workflow, it can be evaluated and acted upon to drive an improvement in health.”

For example, when considering the risk of hyperkalemia, relevant drug interaction alerts will only appear if the potassium level exceeds 5.3, and not when the interaction is just within the wider realm of possibility, Tuchinda explained.

“In this example, certain alerts can be automatically suppressed when additional patient information demonstrates that the risk is not present,” he said. “Ultimately, only the most pertinent information is visible for providers to act.”

Identifying the value

Moving on to another clinical decision support optimization best practice, the value of clinical decision support can be measured in terms of positive predictive value, which describes the relevance of decision support, versus its irrelevance, said Aslan Brooke, senior director of technology at Zynx Health.

“Studies have shown that feeding additional inputs such as a lab value to an advanced decision support engine has increased positive predictive value by a factor of three,” he noted. “Ultimately, alert fatigue decreases and clinicians are more likely to respond because patient-specific data demonstrates that a risk is more likely to be present.”

“Studies have shown that feeding additional inputs such as a lab value to an advanced decision support engine has increased positive predictive value by a factor of three.”

Aslan Brooke, Zynx Health

The potential to improve health is much greater when leveraging more than just one piece of patient data, he advised. CIOs should consider the win-win opportunity when enabling advanced clinical decision support that is triggered at the right place in the workflow, fed with meaningful data inputs, and capable of describing its logic when recommending an easy action, he added.

“CIOs should look for a long track record of success across a diverse set of clients,” Brooke continued. “Since healthcare practices vary dramatically, robust solutions leverage a combination of approaches that build upon evidence-based curation and AI-driven insights from real-world activity, where gaps in evidence exist.”

A single approach to clinical decision support is unlikely to catch all of one’s clinicians’ practice patterns and, therefore, achieve less than optimal health outcomes, he explained. A platform to analyze one’s clinical decision support configuration, utilization and performance is critical to success, he said.

Integrated technology addressing workflow needs

Elsewhere on optimization, hospitalized patients with diabetes, for instance, are at a greater risk for a number of adverse outcomes, including infections, complications, length of stay leading to higher overall treatment costs, and, unfortunately, mortality, said Doug Cusick, president and CEO of TransformativeMed, a vendor of technology and services that optimize EHRs and clinical decision support.

“A best practice is when hospitals leverage clinical decision support tools to help minimize these negative outcomes and reduce costs,” he said. “Integrated technology that addresses workflow needs can make clinical decision support more efficient and achieve a number of the health system’s goals, including scaling standardized approaches to blood glucose control, expanding the reach and impact of diabetes educators and endocrinologists, and, of course, improving the outcomes of patients with diabetes.”

“Integrated technology that addresses workflow needs can make clinical decision support more efficient and achieve a number of the health system’s goals.”

Doug Cusick, TransformativeMed

Rather than force clinicians to leaf through paper copies or simply provide a link to static guidelines, technology should be both dynamic and personalized to the patient, allowing better clinical decision support using the technology the hospital already has to constantly scan a patient’s results and recommend updates based on the latest data, he added.

Integrated, actionable and easy to consume

Quite often, managing diabetes requires the assessment and interpretation of disparate laboratory tests, and EHRs present those results in a fragmented or siloed manner, Cusick stated. Another best practice for clinical decision support optimization, he added, is that it should ensure the information provided to clinicians is presented in a manner that is integrated, actionable and easy to consume.

“Clinical decision support is most impactful when this information is provided to clinicians in a workflow-centric design, with notifications of critical values or relevant alerts conveyed directly to mobile devices,” Cusick suggested. “By asking ‘What will the clinician need to do next’ and ensuring that it can be achieved as easily as possible, adherence to recommendations becomes significantly more likely.”

By providing clinical decision support, health IT can reduce adverse events while patients are in the hospital, help them to leave the hospital setting sooner, and set them on a track for better diabetes management when they’re back at home, he added.

Fine-tuning medication alerts

A growing number of healthcare institutions, in the face of the exponential growth of information overload in today’s standard ‘drug-first’ approach to medication clinical decision support, are achieving great success by addressing alert management through customization, said Tuchinda of FDB.

“With the goal of improving medication-related decision-making during order entry, healthcare professionals should customize and fine-tune alerts based on their organization’s local circumstances and best practices through an interface that provides deep, detailed access into the knowledge base of the decision support system,” he advised.

“We have seen evidence demonstrating that customizing medication alerts based on local practice improves their overall positive predictive value,” he said.

The single biggest challenge for healthcare provider organizations about to implement a medication alert customization program is knowing where to begin, he stated. Since the sheer volume of data can be overwhelming, the analytic capabilities of an alert performance management tool are critical; these tools help to indicate which alerts fire most in an organization and what actions should be taken, he added.

“A dashboard powered by a powerful analytics engine helps decision makers quickly identify the top medication alerts by volume,” he noted. “A detailed view into the data gives the ability to drill down into the specific content at the source of the alert and adjust if desired. Once an alert customization program has begun, users can upload their alert firing data to understand alert burden metrics over time.”

Healthcare organizations that have added analytics to their alert customization program achieve greater success in tackling alert fatigue and ultimately realize the many efficiency, patient care and safety benefits associated with physician order entry, he added.

Remote patient monitoring and digital therapeutics

Kuldeep Singh Rajput, CEO and founder of Biofourmis, offers another clinical decision support technology optimization best practice.

“Reliable data is the most essential ingredient in a safe and effective clinical decision, but that shouldn’t only include static data from existing medical records – ongoing monitoring in near real time has become increasingly important,” he advised. “Remote patient monitoring combined with digital therapeutics is a powerful tool in the clinician’s toolbox for more informed clinical decision making and earlier interventions.”

“Reliable data is the most essential ingredient in a safe and effective clinical decision, but that shouldn’t only include static data from existing medical records – ongoing monitoring in near real time has become increasingly important.”

Kuldeep Singh Rajput, Biofourmis

This approach involves collecting active and passive data from patients with remote patient monitoring in their homes and assessing that data via artificial intelligence- and machine learning-driven predictive analytics for more informed clinical decision making and earlier interventions, he said.

“Digital therapeutics are most effective when multiple physiologic parameters are collected from clinical-grade wearable sensors and combined with information that patients provide via a consumer app,” he added. “When this data is analyzed by an AI- and machine learning-driven analytics engine, predictive insights are gleaned that can help physicians identify health decompensation and ideally prevent an adverse event days or even weeks before it would have otherwise occurred.”

Clinical decision support for heart failure

Heart failure is an ideal condition for leveraging remote monitoring and digital therapeutics for clinical decision support, Rajput used as an example.

“Heart failure is a life-threatening, progressive disease with debilitating symptoms,” he said. “It’s the top global cause of hospitalizations for patients older than 65 years and costs $ 108 billion a year to manage worldwide. About 85% of hospitalized patients suffer from an acute heart failure event at least once, and 43% of patients are admitted to the hospital at least four times.”

Yet recent data from the CHAMP-HF registry demonstrates that fewer than one in four patients with heart failure with reduced ejection fraction are on Guideline Directed Medication Therapy, and only 1% are receiving target doses of medications, he noted.

“To reduce those costs and improve outcomes, patients can be enrolled in a remote monitoring program either after a hospital admission or in an outpatient setting,” he suggested. “The patient is monitored with clinical-grade sensors that collect data on vital signs and monitor physical activity while patient-reported information is entered via a patient-facing smartphone app.”

That data is fed into an analytics engine for insights based on the patient’s medication and dosage, he continued.

“This best practice employs digital therapeutics to ensure that Guideline Directed Medical Therapy and target medication doses are achieved based on analysis of the patient’s physiologic data, medical history and published evidence,” he said. “By leveraging this clinical decision support approach, physicians can target the optimal heart failure medication and dosage for each patient, to encourage adherence and to achieve better outcomes.”

Clinical decision support in the ER

There are ways to optimize clinical decision support technology so that it helps best in the emergency department. Bay and bed space in emergency departments and within hospitals often is at a premium. And under value-based care payment models, increased hospitalizations can reduce a health system’s revenue.

“Often, patients are hospitalized because they need monitoring in order for physicians to make clinical and therapeutic decisions; however, a remote patient monitoring program combined with digital therapeutics for clinical decision support can enable many of these patients to be monitored from home,” Rajput suggested. “This approach can free up valuable bay and bed space at the hospital while lowering costs and improving patient satisfaction.”

When patients with certain acute and/or chronic conditions visit the ER, the clinician can make the determination if they would be a good candidate for a remote monitoring program that essentially creates a hospital room in their home, he said.

“If they are, they can be enrolled in the program rather than being admitted to the hospital – and the clinician will receive near-real-time data and insights based on predictive analytics, which can be used for clinical decision support,” he advised.

This type of program includes visits from a physician and a nurse, the delivery of needed equipment and acute medications, and even imaging studies, he explained.

“The patient is monitored around the clock via clinical-grade sensors that collect data on vital signs, physical activity, sleep patterns and other data,” he said. “The patient also answers subjective questions through an app on their mobile device. As the patient progresses, he or she transitions to fewer in-home visits from clinicians, while still being monitored and providing patient-reported data through the app.”

An AI- and machine learning-driven digital therapeutics platform employs predictive analytics to drive clinical decision making and the patient’s ongoing care plan, remotely, he said. In addition, since any signs of decompensation are flagged early, interventions can be delivered as promptly as if the patient were in a hospital facility, he said.

Twitter: @SiwickiHealthIT
Email the writer:
Healthcare IT News is a HIMSS Media publication.

News from