Benefits and Challenges of Integrating AI and Machine Learning into EHR Systems

Although we have talked a lot about them, AI and machine learning are the new shiny toys of healthcare — this comes with pros and cons. With how versatile these technologies are, we are constantly finding new and exciting ways and areas to apply them in healthcare. However, since the technology can be applied in such vast directions it is largely still untested – both in the sense of the new areas we are applying it and in the sense of their long-lasting impacts in the areas we have been using it.

Today we are going to narrow our focus on these exciting and vast technologies down to their impact on EHR systems. We reached out to our brilliant Healthcare IT Today Community to ask — what are the potential benefits and challenges of integrating AI and machine learning into EHR systems? Below are their answers.

Maxim Abramsky, AVP Product Management at Edifecs
Benefits: AI improves decision-making by analyzing patient data to support accurate diagnoses and personalized treatment. It automates routine tasks, reduces administrative burdens, and uses predictive analytics to anticipate patient outcomes and improve resource allocation. By streamlining workflows and identifying public health trends, AI fosters better population health management.

Challenges: Adoption faces hurdles like data privacy concerns, biased or incomplete datasets leading to inaccuracies, and difficulty integrating AI with existing EHR systems. High implementation costs and resistance from clinicians, who may distrust AI-generated insights, add complexity. Ensuring compliance with healthcare regulations further slows adoption. AI in Healthcare is often a balance between challenges and benefits.

Dr. Michael Sherling, Cofounder and Chief Medical and Strategy Officer at ModMed
AI integration is the key to unlocking a future with less administrative burden and more streamlined clinical workflows, particularly documentation. Ambient listening built directly within an EHR has the ability to capture a doctor-patient conversation and convert it into more than just words, but valuable structured information, preventing physicians from spending a significant amount of time after hours on notes and enabling them to give patients their undivided attention during visits.

Dr. Mimi Winsberg, Co-Founder and Chief Medical Officer at Brightside Health
As the volume of healthcare data continues to rapidly expand, AI offers a promising way to make EHR data more actionable for clinicians. In mental health care, AI can digest vast amounts of patient information and then detect patterns and summarize treatment to date, thus improving diagnostic accuracy. AI can suggest treatment plans, proactive check-ins, and real-time treatment adjustments to providers, and even predict potential crises before they occur. Technology such as this can lead to more timely and personalized interventions for patients.

Additionally, AI eases the administrative burden on clinicians, saving them valuable time, reducing burnout, and enabling more meaningful patient interactions. Finally, AI can generate to-do lists and action items for providers and serve as a measure of quality assurance. We continue to see AI as a tool for clinicians and health systems, not a replacement for them. Overall, AI can support clinical decision-making and save clinicians hours of evaluation by summarizing and informing timely, right-sized care decisions for patients.

Sandra Johnson, Senior Vice President, Client Services at CliniComp
Integrating AI and machine learning into EHR solutions transforms these systems into dynamic, intelligent tools that enhance clinical decision-making and streamline workflows, ultimately improving patient care and operational efficiency. Additionally, this innovation must address challenges such as data accuracy, patient privacy, and interoperability. By fostering a culture of continuous innovation and prioritizing user feedback, we can ensure EHR systems evolve to meet the ever-changing needs of modern healthcare.

John Hataway, Sr Director, Continuous Improvement & Automation at Savista
Effective integration of AI and machine learning technologies into EHR systems can offer substantial benefits. These technologies can enhance continuity of care, improve error detection, and reduce administrative burdens, allowing healthcare providers to focus more on patients and improve the overall patient experience (and outcomes). AI/ML can support adherence to best practices and expand clinical thought bandwidth by providing insights into potential clinical pathway improvements based on observed outcomes and upstream case characteristics. Similarly, these tools enable better predictive analytics for patient outcomes and risk stratification, facilitating more personalized healthcare delivery while enhancing decision support systems for clinicians.

However, these benefits come with significant challenges. AI inaccuracies (or hallucinations) can pose serious risks, and potential over-reliance on AI/ML may undermine clinical judgment and outcomes, and when built within the structure of EHRs these risks can be amplified. Complexities in data standardization and process integration present substantial hurdles from a cost and effectiveness perspective. Additionally, concerns about patient data privacy and security are always front of mind due to potential vulnerabilities introduced by AI integration. Effective utilization of these tools requires training for both the technology and its users, ensuring ongoing adherence to regulatory and security guidelines. Lastly, the potential benefits of technologies and systems implemented to enhance healthcare delivery must always be weighed against the potential cost/risk of dehumanization of that delivery.

Rob Helton, SVP of Product at WebPT
To make AI truly transformative in healthcare, we need to tackle three core challenges: data quality, interoperability, and cost. Without clean, standardized data, AI can’t deliver reliable insights. Without seamless integration between systems, we can’t achieve holistic coordinated patient care. Without cost-effective solutions or incentives, innovation in a market with shrinking reimbursement margins will stall.

Todd Doze, Chief Executive Officer at Janus Health
AI in the revenue cycle has tremendous potential to reduce health systems’ cost to collect. But integrating AI into EHRs may run the risk of further dividing the AI haves and have-nots. AI solutions can easily cost $500K or more to get started with an EHR-based AI project. Plus there are additional investments in talent, infrastructure, and maintenance. Many providers simply don’t have the resources to go down a do-it-yourself path. Prioritizing AI partnerships over DIY is often more cost-effective for health systems, especially if they don’t already have significant IT infrastructure and staff. Consider partners with EHR-integrated AI solutions to realize transformative changes in labor costs, denial reduction, write-off improvement, and the employee and patient experience.

Tarken Friske, Senior Director of Consulting at Full Spectrum
In today’s world, where timely access to physicians is increasingly difficult, AI and machine learning technologies present several opportunities to benefit patients and providers alike, ultimately driving improvement in patient outcomes. Key benefits include…

• Improved Efficiency: AI is well suited to automating certain non-critical provider tasks like note-taking and data entry, which can save considerable time better spent focusing on patient care
• Enhanced Clinical Decision Support: AI can analyze large volumes of patient data helping clinicians to make more informed decisions; ML algorithms can enhance clinical decision support by analyzing patient histories, lab results, and other data, promoting evidence-driven decisions and potentially reducing diagnostic errors
• Personalized Medicine: AI’s ability to see patterns in large and disparate data sets enables consideration of patient genetics, lifestyle information, and other non-traditional variables to recommend personalized treatments, leading to more precise and effective care

While the potential benefits are obvious, there will be challenges to overcome in integrating these technologies into EHR systems…

• Data Privacy and Security: EHR systems contain sensitive patient information, and integrating AI introduces additional cybersecurity risks; compliance with HIPAA and other privacy regulations becomes increasingly complex with the inclusion of AI access to extensive patient data
• Data Quality and Bias: To function effectively, AI requires high-quality data; EHRs may contain incomplete, outdated, or biased data, which can lead to inaccurate predictions or reinforce existing health disparities
• Interoperability Issues: Different provider networks use different EHR systems; the lack of standardization makes it difficult to integrate AI across platforms, potentially limiting AI’s access to comprehensive datasets
• Provider Reluctance: Healthcare providers may be resistant to new technology due to the learning curve and potential disruptions to workflows; training and change management are necessary to ensure successful AI adoption

Shyam Rajagopalan, Co-Founder and CTO at Infinitus Systems
Historically, integrating AI and machine learning into EHR systems has been challenging due to the difficulty of moving data in and out of these platforms, which were often closed and proprietary. The introduction of standards like SMART on FHIR and FHIR has significantly improved data interoperability, allowing for easier extraction of data and better integration with external applications, especially in modern EHRs that support these standards. This has also enabled the linking of login credentials between EHRs and external tools, simplifying the user experience. However, while extracting data has become easier, re-entering structured data into EHRs remains difficult, with many systems still only supporting basic or unstructured data inputs, such as flat documents or text blurbs.

This progress has made it possible to develop AI and ML applications that focus on administrative tasks like scheduling and benefits management, as these typically rely on pulling existing data from the EHR and are amenable to unstructured entry back. However, the barriers to inputting structured clinical data back into the EHR mean that clinical AI applications—those that could assist in diagnostics, treatment planning, or patient monitoring, are earlier in their development. Overall we’re moving towards a lot more enablement of ML applications in EHRs.

Michael Meucci, President and CEO at Arcadia
One of the challenges of integrating AI into EHR systems is enabling providers to act on abundance. Today, scaled AI deployment is restricted by the lack of workflows. Think about it: Powerful technology and adequately trained models paired with high-quality, aggregated data can generate endless recommendations, new activities, and signals. We need to deliver information to care teams at the right time, place, and format where they can act on it to avoid further burdening providers. In other words, a human must be able to effectively manage the abundance that AI can create. Until we can properly integrate AI output into existing EHR workflows at scale, we cannot fully realize the benefits of technology to make the workforce faster. When we effectively democratize the power of generative AI, we unlock massive benefits.

One specific example is personalized precision medicine, which delivers the best care to every individual patient. Bringing AI-powered evidence-generation capabilities into the EHR can accelerate appropriate clinical decision-making to drive high-value, low-cost patient care based on their unique physiology. I’ve seen providers use point-of-care tools to rapidly generate novel peer-reviewed evidence and use that information to make data-driven decisions on optimal clinical guidelines, care pathways, and the best interventions. In this scenario, the provider wins by using technology as a co-pilot to save time on research that would otherwise require manual effort; the patient wins by receiving the latest evidence-based care that’s optimized to their individual needs for the best possible outcome; and the healthcare provider and payer succeed in value-based care and other quality initiatives.

Srdjan Prodanovich, MD FAAD, Founder and Chief Medical Officer at EZDERM
Artificial intelligence’s real benefit in medicine comes from its ability to analyze complex clinical scenarios, where a patient’s genetic makeup, medical history, laboratory and imaging results, and medications all interplay. Medical providers often face overwhelming amounts of data, making identifying the most relevant information challenging. By utilizing vast medical knowledge, AI can evaluate these factors holistically, generating invaluable clinical decision support that transcends its current role as a scribe. This advancement enhances clinical decision-making and improves the accuracy of diagnoses and the effectiveness of treatment planning. AI can even anticipate potential complications or dangerous drug interactions that might be overlooked. As AI presents this critical information to physicians, it empowers faster, data-driven decisions that ultimately enhance patient care. The result is a more efficient healthcare system, with improved outcomes and reduced burnout for healthcare providers.

So many great points to think about here! Huge thank you to Maxim Abramsky, AVP Product Management at Edifecs, Dr. Michael Sherling, Cofounder and Chief Medical and Strategy Officer at ModMed, Dr. Mimi Winsberg, Co-Founder and Chief Medical Officer at Brightside Health, Sandra Johnson, Senior Vice President, Client Services at CliniComp, John Hataway, Sr Director, Continuous Improvement & Automation at Savista, Rob Helton, SVP of Product at WebPT, Todd Doze, Chief Executive Officer at Janus Health, Tarken Friske, Senior Director of Consulting at Full Spectrum, Shyam Rajagopalan, Co-Founder and CTO at Infinitus Systems, Michael Meucci, President and CEO at Arcadia, and Srdjan Prodanovich, MD FAAD, Founder and Chief Medical Officer at EZDERM for taking the time out of your day to submit a quote to us! And thank you to all of you for taking the time out of your day to read this article! We could not do this without all of your support.

What do you think are the potential benefits and challenges of integrating AI and machine learning into EHR systems? Let us know either in the comments down below or over on social media. We’d love to hear from all of you!