Dhaval Shah
The time has come to realize the potential of clinical decision support systems and tools
February 28, 2023
By Dhaval Shah
We live in a historic and wonderful time, one in which the volume of medical knowledge doubles, by some estimates, every 73 days - not 50 years as it did in the 1950s. Simultaneously, patient-centric delivery models, from personalized medicine to value-based care, require clinicians to address the myriad factors that directly and indirectly impact outcomes for individual patients – an endeavor that requires continual access to actionable data.
In the face of this reality, even the most astute caregivers struggle to make decisions that reflect accepted standards of care, benchmarks that notably continue to rapidly change as new innovations and procedures make many protocols obsolete. Simultaneously, clinical teams are stretched thin, with physician and nurse burnout now a significant issue worldwide.
It is a problem that shows no signs of abating with only 47% of nurses saying they plan to remain in their current job over the next two years. Moreover, nine out of ten nurses are reporting that high patient loads negatively impact patient care and 75% of respondents in the same survey said that their patient loads are unsafe. The situation for physicians is not much better with the overall prevalence of burnout among doctors exceeding 60% in 2021.
Ironically, despite being created to help address many of the issues that increase burnout - among them, the demands of patient care - today’s clinical decision support (CDS) tools often exacerbate the problem. Legacy CDSS systems were created with the expectation that they would make physicians’ lives easier by recommending treatment plans and the steps that comprise them, diagnostic tests, therapies, medicines and doses, and of course, steps to take in the event of problematic developments that cause concern. Unfortunately, few standalone CDSS deliver on this promise.
Many standalone systems draw on isolated databases and data sets. Inaccurate information and algorithms that do not reflect new standards of care are but one cause of adoption issues, with many physicians giving up on such systems after using them just a few times. Alert fatigue is also a significant, yet necessary hindrance in a time when many CDSS are used for dosing error checks.
Most importantly, mistakes that an effective decision support system could theoretically prevent are being made with painful regularity. A recent study found nearly one-in-four patients experienced at least one adverse event while hospitalized. And nearly one-third of these resulted in serious harm that required a prolonged recovery.
Today’s physicians and nurses are being asked to provide more care and process more information than ever before and need support. The time has come to realize the potential of clinical decision support systems and tools.
The era of AI-powered predictive CDSS is here
With the dramatic gains made in artificial intelligence, deep learning, analytics and even highly flexible platforms that enable non-technical users to create powerful machine learning applications, predictive CDS is now not only a reality but one that continues to grow even more effective with use as algorithms are refined and the data that informs them grows in volume, scope and specificity.
Such systems are light years ahead of the standalone systems they replace. Whereas most legacy CDSS rely on established knowledge bases and defined, linear and often rigid rules to lookup recommended treatments for specific conditions, drug and allergy indications and other important factors in patient care, today’s AI-powered systems are predictive and proactive.
Tightly integrated with the electronic health record (EHR), these systems ensure complete access not only to the individual patient’s existing information, but also additions to it from external diagnostic tests or even insights from experts and specialists consulted from across the globe. Most importantly, these next-gen CDS tools provide optimized workflow alerts and actionable insights physicians, nurses and other clinicians can use that reflect detailed analysis of evidence-based clinical information.
The use cases associated with these systems are many and varied. Some of many examples include:
● Diagnostic Imaging: Incorporated within the CDSS, AI-powered imaging analysis can help radiologists detect abnormalities that require treatments faster and more accurately, particularly as imaging technology results in scans that are increasingly complex, multi-layered and more difficult for the human eye to discern. The technology shows great promise, with a researcher at Tulane University recently finding that AI can accurately identify and diagnose colorectal cancer as well or better than pathologists.
● Chronic Disease Management: Predictive CDS tools help identify patients who are at high risk of adverse events as a result of comorbidities, while also providing the recommendations for an intelligent and evidence-based intervention. And because the system is integrated directly with the EHR and appropriate external stakeholders – for example pharmacies and labs – clinicians can also be alerted when patients fail to follow recommended treatment plans. Notably, such systems can also be integrated with feeds that deliver information on social determinants of health and other data to influence patients’ outcomes as well as value-based care and other risk-adjusted care models.
● Precision Medicine: Physicians can be actively supplied with clinical guidelines that reflect not only accepted standards of care for common conditions, but also those that incorporate the patient’s genetic makeup, allergies, family history and other considerations that are crucial when providing precision medicine. Notably, advanced CDSS can even identify applicable clinical trials and alert clinicians when a patient exhibits the required, highly-specific conditions needed to apply.
These use cases and the myriad transformative capabilities of today’s integrated, AI-powered CDSS require diligence. Organizations should consider the following before implementing such solutions.
Today’s CDSS must be viewed through a holistic lens
Advanced CDSS that draw on the advanced AI, machine learning and analytics capabilities are powerful, but they are not solutions that can be simply turned on and left with a “set it and forget it” mindset. Some of the many considers all clinical operations should consider before implementing them include:
● Ensure that the EHR and other core systems the CDSS will integrate with are robust: This includes making sure that all APIs are up-to-date and satisfy the latest interoperability standards, including HL7 and FHIR. This can help ensure that the CDS is not only accurate, but also that false positives and erroneous alerts are kept to a minimum.
● Embrace the cloud: A properly designed cloud, designed specifically for health care and offering the most stringent security protections, not only enables organizations to effectively and cost efficiently consolidate its data in one location for analytics and to inform algorithms, but it also is crucial to ensure disaster recovery capabilities.
● Create a culture that protects patient data: This includes not only making sure that all data, including medical images and patient information, is encrypted in transit and at rest, but also that staff members are regularly trained to recognize phishing schemes and report any anomalies. Notably, in some cases synthetic “patient” data can also be used for specific use cases to provide an extra level of protection.
● Facilitate maintenance and monitoring: The CDSS must be monitored for accuracy and performance. Notably, this also includes ensuring that staff members do not fail to follow established clinical guidelines and procedures because they received trusted recommendations from the CDSS.
Not surprisingly, the use of advanced AI-based CDSS will only increase in light of these capabilities and benefits. Some estimates point to the market reaching $12.4 billion by 2030, with a compound annual growth rate of 9.45% in the years in between.
The brave new world that advanced AI-based CDSS offer is one that will help health care organizations accelerate and thrive in the move to patient-focused care, help reduce clinician burnout, provide physicians and nurses with a trusted and consultative partner, reduce costly and harmful errors, improve outcomes and dramatically lower costs. But it is a transformation that must be taken with care and diligence. As with all powerful technologies, the full potential of CDSS can only be achieved when it is deployed on a strong and thoughtfully constructed foundation.
About the author: Dhaval Shah, is the executive vice president at CitiusTech. He has more than two decades of experience in health care IT, including senior-level roles in engineering, research, software development, and management roles serving pharmaceutical companies, physicians’ practices, and health insurance companies.
We live in a historic and wonderful time, one in which the volume of medical knowledge doubles, by some estimates, every 73 days - not 50 years as it did in the 1950s. Simultaneously, patient-centric delivery models, from personalized medicine to value-based care, require clinicians to address the myriad factors that directly and indirectly impact outcomes for individual patients – an endeavor that requires continual access to actionable data.
In the face of this reality, even the most astute caregivers struggle to make decisions that reflect accepted standards of care, benchmarks that notably continue to rapidly change as new innovations and procedures make many protocols obsolete. Simultaneously, clinical teams are stretched thin, with physician and nurse burnout now a significant issue worldwide.
It is a problem that shows no signs of abating with only 47% of nurses saying they plan to remain in their current job over the next two years. Moreover, nine out of ten nurses are reporting that high patient loads negatively impact patient care and 75% of respondents in the same survey said that their patient loads are unsafe. The situation for physicians is not much better with the overall prevalence of burnout among doctors exceeding 60% in 2021.
Ironically, despite being created to help address many of the issues that increase burnout - among them, the demands of patient care - today’s clinical decision support (CDS) tools often exacerbate the problem. Legacy CDSS systems were created with the expectation that they would make physicians’ lives easier by recommending treatment plans and the steps that comprise them, diagnostic tests, therapies, medicines and doses, and of course, steps to take in the event of problematic developments that cause concern. Unfortunately, few standalone CDSS deliver on this promise.
Many standalone systems draw on isolated databases and data sets. Inaccurate information and algorithms that do not reflect new standards of care are but one cause of adoption issues, with many physicians giving up on such systems after using them just a few times. Alert fatigue is also a significant, yet necessary hindrance in a time when many CDSS are used for dosing error checks.
Most importantly, mistakes that an effective decision support system could theoretically prevent are being made with painful regularity. A recent study found nearly one-in-four patients experienced at least one adverse event while hospitalized. And nearly one-third of these resulted in serious harm that required a prolonged recovery.
Today’s physicians and nurses are being asked to provide more care and process more information than ever before and need support. The time has come to realize the potential of clinical decision support systems and tools.
The era of AI-powered predictive CDSS is here
With the dramatic gains made in artificial intelligence, deep learning, analytics and even highly flexible platforms that enable non-technical users to create powerful machine learning applications, predictive CDS is now not only a reality but one that continues to grow even more effective with use as algorithms are refined and the data that informs them grows in volume, scope and specificity.
Such systems are light years ahead of the standalone systems they replace. Whereas most legacy CDSS rely on established knowledge bases and defined, linear and often rigid rules to lookup recommended treatments for specific conditions, drug and allergy indications and other important factors in patient care, today’s AI-powered systems are predictive and proactive.
Tightly integrated with the electronic health record (EHR), these systems ensure complete access not only to the individual patient’s existing information, but also additions to it from external diagnostic tests or even insights from experts and specialists consulted from across the globe. Most importantly, these next-gen CDS tools provide optimized workflow alerts and actionable insights physicians, nurses and other clinicians can use that reflect detailed analysis of evidence-based clinical information.
The use cases associated with these systems are many and varied. Some of many examples include:
● Diagnostic Imaging: Incorporated within the CDSS, AI-powered imaging analysis can help radiologists detect abnormalities that require treatments faster and more accurately, particularly as imaging technology results in scans that are increasingly complex, multi-layered and more difficult for the human eye to discern. The technology shows great promise, with a researcher at Tulane University recently finding that AI can accurately identify and diagnose colorectal cancer as well or better than pathologists.
● Chronic Disease Management: Predictive CDS tools help identify patients who are at high risk of adverse events as a result of comorbidities, while also providing the recommendations for an intelligent and evidence-based intervention. And because the system is integrated directly with the EHR and appropriate external stakeholders – for example pharmacies and labs – clinicians can also be alerted when patients fail to follow recommended treatment plans. Notably, such systems can also be integrated with feeds that deliver information on social determinants of health and other data to influence patients’ outcomes as well as value-based care and other risk-adjusted care models.
● Precision Medicine: Physicians can be actively supplied with clinical guidelines that reflect not only accepted standards of care for common conditions, but also those that incorporate the patient’s genetic makeup, allergies, family history and other considerations that are crucial when providing precision medicine. Notably, advanced CDSS can even identify applicable clinical trials and alert clinicians when a patient exhibits the required, highly-specific conditions needed to apply.
These use cases and the myriad transformative capabilities of today’s integrated, AI-powered CDSS require diligence. Organizations should consider the following before implementing such solutions.
Today’s CDSS must be viewed through a holistic lens
Advanced CDSS that draw on the advanced AI, machine learning and analytics capabilities are powerful, but they are not solutions that can be simply turned on and left with a “set it and forget it” mindset. Some of the many considers all clinical operations should consider before implementing them include:
● Ensure that the EHR and other core systems the CDSS will integrate with are robust: This includes making sure that all APIs are up-to-date and satisfy the latest interoperability standards, including HL7 and FHIR. This can help ensure that the CDS is not only accurate, but also that false positives and erroneous alerts are kept to a minimum.
● Embrace the cloud: A properly designed cloud, designed specifically for health care and offering the most stringent security protections, not only enables organizations to effectively and cost efficiently consolidate its data in one location for analytics and to inform algorithms, but it also is crucial to ensure disaster recovery capabilities.
● Create a culture that protects patient data: This includes not only making sure that all data, including medical images and patient information, is encrypted in transit and at rest, but also that staff members are regularly trained to recognize phishing schemes and report any anomalies. Notably, in some cases synthetic “patient” data can also be used for specific use cases to provide an extra level of protection.
● Facilitate maintenance and monitoring: The CDSS must be monitored for accuracy and performance. Notably, this also includes ensuring that staff members do not fail to follow established clinical guidelines and procedures because they received trusted recommendations from the CDSS.
Not surprisingly, the use of advanced AI-based CDSS will only increase in light of these capabilities and benefits. Some estimates point to the market reaching $12.4 billion by 2030, with a compound annual growth rate of 9.45% in the years in between.
The brave new world that advanced AI-based CDSS offer is one that will help health care organizations accelerate and thrive in the move to patient-focused care, help reduce clinician burnout, provide physicians and nurses with a trusted and consultative partner, reduce costly and harmful errors, improve outcomes and dramatically lower costs. But it is a transformation that must be taken with care and diligence. As with all powerful technologies, the full potential of CDSS can only be achieved when it is deployed on a strong and thoughtfully constructed foundation.
About the author: Dhaval Shah, is the executive vice president at CitiusTech. He has more than two decades of experience in health care IT, including senior-level roles in engineering, research, software development, and management roles serving pharmaceutical companies, physicians’ practices, and health insurance companies.