re:Invent 2024

Healthcare's Digital Evolution: From Manual Charts to Generative AI Solutions

Generative AI & LLMOps

Learn how Generative AI is poised to transform healthcare by addressing technological challenges, reducing administrative burdens, enhancing clinical decision-making, and creating more personalized, efficient patient care experiences.

As a healthcare provider, I've witnessed healthcare technology's evolution from multiple vantage points. Starting in pharmaceutical sales, I saw the transformation from handwritten prescriptions to e-prescribing systems – a change that significantly reduced errors and allowed for more meaningful provider interactions.

Later, as a hospital nurse, I experienced the challenging transition period between paper and digital records. The hybrid system required faxing doctors' paper orders to the hospital’s pharmacy for manual digital entry, highlighting the urgent need for complete digitization. The countless hours spent deciphering handwriting, making clarification calls, and managing fragmented systems demonstrated how these inefficiencies not only burdened providers but also impacted patient care.

These firsthand experiences with disconnected systems drove me to pursue a career in healthcare technology development. Drawing from my unique perspective in both sales and nursing, I now focus on creating integrated solutions that enhance provider workflow and patient outcomes.

Healthcare analytics isn't just about data – it's about uncovering patient stories that guide better care decisions. When technology is thoughtfully integrated, it amplifies rather than diminishes the human element in healthcare delivery.

We are on the cusp of a significant shift with the emergence of Generative AI (GenAI). This technology holds the potential to propel healthcare forward, addressing many of the shortcomings of previous innovations while opening new avenues for improved patient care.

The introduction of Electronic Health Records (EHRs) and other integrated digital processes, while necessary, has often failed to live up to expectations. Instead of streamlining our work, they've frequently added to the burden on an already strained healthcare system. However, innovative solutions are emerging to address these challenges. For instance, Sentara Healthcare is implementing an AI-driven Virtual Nursing solution that harnesses generative AI to unlock insights from electronic medical records, biometric devices, and other data sources, enabling automated administrative tasks and real-time intelligence delivery to care teams through fixed in-room devices integrated with their EHR system. This technology automates critical workflows like admission planning, discharge coordination, and patient education, significantly reducing administrative burden while improving care delivery. Such innovations represent healthcare's next evolution, addressing legacy system limitations while creating new possibilities for enhanced patient care by capitalizing on GenAI.

Before we explore further how GenAI can transform healthcare delivery and solve staffing challenges, let's look more closely at the fundamental problems facing our healthcare system.

The EHR Dilemma: A Cautionary Tale

The implementation of EHRs was meant to digitize and streamline patient records, improve communication between providers, and enhance the overall quality of care. However, the reality has often been quite different:

  1. Increased Administrative Burden: Many clinicians, myself included being a nurse, find ourselves spending more time entering data into EHRs than interacting with patients. A study published in the Annals of Internal Medicine found that physicians spend nearly two hours on EHR tasks for every hour of direct patient care.
  2. User Interface Issues: Poorly designed interfaces have led to inefficient workflows and increased chances of errors. The American Medical Association has identified EHR usability as a key contributor to physician burnout.
  3. Alert Fatigue: The overabundance of alerts and notifications in many EHR systems has led to alert fatigue, where clinicians may miss crucial information amidst the noise of less important alerts.
  4. Interoperability Challenges: Despite promises of seamless data sharing, many EHR systems still struggle with interoperability due to varying data standards, proprietary systems, and inconsistent formatting across healthcare providers, making it difficult to seamlessly share and interpret patient information across different platforms and organizations.

These issues have not only frustrated healthcare providers but have also potentially compromised patient care, serving as a stark reminder that technology implementation requires a deep understanding of the healthcare environment. Yet, there's reason for optimism. The healthcare interoperability landscape is undergoing a significant transformation, with over 96% of hospitals now using certified EHRs and adopting standardized FHIR APIs. According to HHS leadership, we're at a pivotal moment where the primary barriers to health information sharing are no longer technical but behavioral. As regulatory initiatives like the information blocking rules drive new imperatives for data exchange, solutions leveraging AWS HealthLake and similar technologies are emerging as powerful tools to address these longstanding challenges, promising to revolutionize how healthcare organizations share and utilize patient data.

The Time Burden: Quantifying the EHR Impact

Recent research has revealed the staggering impact of EHR systems on healthcare providers' time management across various clinical settings. In emergency departments, nurses find themselves dedicating over a quarter of their time to EHR tasks, while only 25% is spent on direct patient care. The constant interruptions are particularly concerning, with nurses facing an average of 14 EHR task disruptions every hour. These interruptions, primarily coming from nursing colleagues and patients, frequently force staff to switch tasks or attempt challenging multitasking scenarios.

The situation in inpatient settings paints an equally concerning picture. Nurses typically spend 22% of their workday interacting with EHR systems, averaging about 144 minutes during each 12-hour shift. Physicians face an even heavier documentation burden, with EHR tasks consuming approximately 37% of their workday in both inpatient and outpatient settings. In primary care, the impact is particularly pronounced, with physicians spending an average of 36.2 minutes on EHR documentation for every 30-minute patient visit.

The COVID-19 pandemic has significantly exacerbated these challenges. Key impacts include:

These increasing time demands have created serious implications far beyond daily healthcare delivery. While EHRs can improve efficiency in certain areas such as information gathering and reviewing, the overwhelming time burden often leads to reduced patient interaction and contributes to provider burnout. Moreover, the fragmented nature of EHR documentation and inconsistent data entry practices directly impact medical research capabilities. When providers are rushed to complete documentation, vital patient information may be buried in unstructured notes or recorded inconsistently, making it extremely difficult to identify suitable candidates for clinical trials. This not only slows the pace of medical research but also creates missed opportunities for patients who could benefit from innovative treatments. The challenges of mining EHR data for research purposes are particularly acute in conditions requiring detailed longitudinal data or specific patient characteristics. The emotional exhaustion resulting from extensive EHR use, combined with the downstream effects on research and clinical trial recruitment, highlights the urgent need for more efficient solutions that better serve providers, patients, and the broader healthcare research community.


How Traditional AI/ML and GenAI Are Solving Healthcare's Most Complex Challenges

Traditional AI/ML along with the emerging field of GenAI have the potential to address many of the shortcomings of previous technological implementations while offering new capabilities that could truly transform patient care. To fully realize this potential, however, several key considerations must be addressed.

Optimizing Real Time Alerts

Firstly, real-time alerts must be carefully calibrated to inform relevant staff of risks without overwhelming them, thus avoiding alarm fatigue. This targeted approach ensures that critical information reaches the right personnel without causing unnecessary disruptions.

Data Utilization

Secondly, data utilization should serve a dual purpose: preventing immediate incidents and informing proactive interventions based on long-term insights. This comprehensive approach allows healthcare providers to not only react to current situations but also anticipate and mitigate future challenges.

Generate ROI

Lastly, for the program to be scalable, AI should not be viewed merely as an additional expense, but as a means to reduce operational costs and generate savings. These savings can then be reinvested to fund further technological advancements, creating a virtuous cycle of innovation and improvement in patient care.

Breaking Down Barriers 

Critical to this transformation is the elimination of data silos and institutional barriers that have historically hindered healthcare innovation. However, success hinges on establishing robust data standardization practices and maintaining pristine data quality across systems. Without clean, standardized data that can flow seamlessly between different healthcare entities and systems, even the most sophisticated AI solutions will fall short of their potential. This underscores the need for industry-wide collaboration on data standards and quality control protocols as foundational steps toward realizing the full promise of GenAI in healthcare.

By addressing these aspects, we can harness the full potential of Generative AI and other emerging technologies to create a more efficient, effective, and patient-centered healthcare system.

To illustrate this, let's consider some concrete examples of how Generative AI could revolutionize various aspects of healthcare:

1. Streamlining Administrative Tasks

GenAI can significantly reduce the administrative burden on healthcare providers:

  • Automated Documentation: GenAI can generate initial drafts of clinical notes, discharge summaries, and other documentation based on patient interactions, which clinicians can then review and finalize. This can drastically reduce the time spent on paperwork.
  • Intelligent Scheduling: AI algorithms can optimize appointment scheduling, taking into account factors like patient preferences, urgency of care, and provider availability.
  • Smart EHR Interfaces: GenAI can power more intuitive, context-aware EHR interfaces that adapt to individual clinician preferences and workflows.

AWS Case Study: Netsmart is transforming healthcare documentation through AWS' advanced AI solutions. Recognizing that providers spend approximately 40% of their workweek on documentation rather than patient care, Netsmart leveraged Amazon Bedrock and AWS HealthScribe to significantly reduce this administrative burden. Their solution combines Amazon Chime SDK's foundation with HealthScribe's ability to convert provider-patient conversations into clinical progress notes, particularly valuable for managing the extensive narrative information typical in behavioral health. By implementing these AWS services along with Claude 2 models through Bedrock, Netsmart has created a secure, HIPAA-compliant documentation system that allows providers to focus more on patient interaction while maintaining comprehensive clinical records. This integration of AWS's AI capabilities has proven especially effective in the behavioral health domain, where preserving the nuanced context of patient interactions is crucial for quality care delivery.

2. Enhancing Clinical Decision Support

GenAI can provide more sophisticated and personalized clinical decision support:

  • Personalized Treatment Recommendations: By analyzing vast amounts of patient data, research literature, and clinical guidelines, GenAI has the capability to generate personalized treatment recommendations tailored to individual patient characteristics.The accuracy and timeliness of input data is critical as inaccurate or outdated information can lead to inappropriate treatment recommendations, missed contraindications, or failure to account for crucial patient-specific factors that could result in adverse outcomes. Robust data validation processes and continuous monitoring of data quality are therefore essential to ensure patient safety and maintain trust in AI-assisted clinical decision support systems.
  • Predictive Analytics: GenAI models has the capability to identify patients at high risk for certain conditions or complications, allowing for more proactive and preventive care. One such real-world example is the Advanced Alert Monitor (AAM), an AI system developed by Kaiser Permanente to analyze hospital patient data (lab values, vital signs, clinical records) to predict potential health deterioration within 12 hours, resulting in prevention of over 500 deaths annually and a 10% reduction in high-risk readmissions by allowing early intervention aligned with patient care goals. Building on this foundation, advanced GenAI systems could further enhance predictive capabilities by incorporating natural language processing of clinical notes, real-time integration of wearable device data, genetic information, and social determinants of health, while also providing explainable recommendations for intervention strategies and automatically adjusting risk thresholds based on patient-specific factors and historical outcomes.
  • Literature Synthesis and Administrative Inbox Support: GenAI can continuously analyze and synthesize the latest medical literature, providing clinicians with up-to-date, evidence-based insights relevant to their patients, making evidence-based practice more accessible than ever before. Equally impactful on daily operations, GenAI can transform administrative workflow by intelligently filtering, prioritizing, and organizing clinical correspondence in providers' inboxes, potentially saving hours of administrative time daily and reducing burnout while ensuring important patient care communications are never missed.

AWS Case Study: The Fred Hutch Cancer Center demonstrates Amazon Comprehend Medical's impact on clinical trials to develop more personalized treatments. With estimated 60-80% of critical research information existing in narrative text form - from family histories to detailed care notes - processing this unstructured data traditionally took hours per record. Now, using this AWS service, their research teams can analyze these records in seconds, rapidly extracting vital clinical information from provider notes and accelerating patient matching for trials. This breakthrough gives researchers immediate access to crucial insights when they need them, accelerating the development of lifesaving therapies for patients.

3. Improving Patient Experience, Engagement, and Education

GenAI can enhance patient experience, engagement, and education:

  • Personalized Patient Education: GenAI can create hyper-personalized educational materials for patients based on their specific condition, treatment plan, and health literacy level.
  • Virtual Health Assistants: AI-powered chatbots could provide patients with 24/7 access to basic health information, medication reminders, and appointment scheduling assistance.
  • Remote Monitoring: While analytical AI can analyze data from wearable devices and home monitoring equipment, alerting healthcare providers to potential issues before they become serious; GenAI can generate natural, personalized communications to patients about their health metrics, translate complex medical data into easy-to-understand explanations, and when combined with multi-modal capabilities, can analyze patient-submitted photos or videos of symptoms, provide visual guides for proper use of medical devices, and generate customized wellness recommendations based on both numerical health data and visual inputs from the patient's environment.

AWS Case Study: Beth Israel Deaconess Medical Center (BIDMC) demonstrates how machine learning services can address a critical healthcare challenge - creating a frictionless experience for patients while allowing providers to focus more on care delivery and less on administrative tasks. Recognizing that inefficiencies in hospital operations can lead to medical errors and reduced quality of care, BIDMC implemented AWS solutions to transform their workflows. Through Amazon SageMaker and Amazon Comprehend Medical, they eliminated redundant paperwork for patients and automated time-consuming documentation tasks for providers, saving hours of manual work daily and reducing the frustration of repeatedly filling out similar forms. The system now automatically identifies missing consent forms and critical H&P documentation, preventing surgical delays and improving patient experience. Additionally, their ML models help predict patient appointment attendance and emergency department surges, enabling proactive patient outreach and resource allocation. This AWS-powered digital transformation has delivered tangible benefits: providers spend less time on paperwork, patients experience fewer delays and scheduling issues, and the hospital can deliver more timely and efficient care. 

4. Accelerating Medical Research

Medical research is being transformed by ML, traditional analytical AI, and newer GenAI approaches, each playing distinct but complementary role:

Drug Discovery Traditional analytical AI has long excelled at methodical screening processes, using predefined rules and established molecular patterns to systematically search existing databases of drug compounds. In contrast, GenAI brings a more creative and exploratory approach – it can generate entirely new molecular structures by learning patterns from successful drugs, effectively "imagining" novel drug candidates that have never existed before. While traditional AI asks "Does this match what we know?", GenAI asks "What else could be possible?" and can explain its reasoning for suggesting unexpected chemical combinations.

Clinical Trial Optimization Where traditional AI excels at the mathematical aspects of trial design – using statistical models to match patient demographics with trial criteria and predict dropout rates based on historical data – GenAI transforms how we interact with and understand trial data. It can generate natural language descriptions of ideal candidate profiles, rewrite complex trial protocols for clarity, and synthesize insights from thousands of unstructured trial notes. Perhaps most importantly, it can simulate potential trial scenarios to identify unforeseen challenges that statistical models might miss.

Genomic Analysis Traditional AI approaches genomics through pattern matching against known genetic markers and statistical analysis of genetic variations – a powerful but relatively rigid approach. GenAI takes this further by generating hypotheses about gene interactions, creating narrative explanations of complex genetic pathways, and proposing novel mechanisms of action by synthesizing insights from genomics, proteomics, and medical literature. It can effectively "connect the dots" across disparate biological datasets in ways that traditional pattern-matching algorithms cannot.

AWS Case Study: Moderna exemplifies how a digital-first biotechnology company can leverage AWS to revolutionize drug development and patient care. By building their entire infrastructure on AWS cloud services, Moderna has created an integrated ecosystem where data serves as a strategic asset, enabling rapid iteration and scalable innovation. Their "digitize everything" approach combines AWS's machine learning and AI capabilities to accelerate drug development insights, while AWS Data Exchange enhances their ability to analyze real-world data (RWD) at unprecedented speed and scale.

What sets Moderna apart is their seamless integration of multiple AWS services, including a sophisticated Natural Language Understanding (NLU) AI engine that enables intelligent, omnichannel communication across their entire stakeholder ecosystem - from healthcare providers to patients. Their future-ready platform extends from early-stage drug development through to patient experience, with conversational AI powering personalized digital interactions and automated workflows. This comprehensive AWS-powered digital transformation has enabled Moderna to achieve remarkable agility in drug development, drastically reduce time-to-insight for research data, and create more engaging patient experiences through technology.

By treating data as a strategic asset and leveraging AWS's advanced capabilities, Moderna has established a model for how biotechnology companies can use cloud technology to drive innovation, scale operations, and improve healthcare outcomes.

The Importance of Human Expertise and Ethical Guardrails

Artificial Intelligence's arrival in healthcare has sparked both hope and skepticism among frontline workers. While AI holds the potential to streamline workflows and boost productivity - much like the initial promises of Electronic Health Records (EHRs) - many healthcare professionals remain cautious. Their skepticism isn't unfounded; the implementation of EHRs, while digitizing patient records, increased administrative burden for many clinicians and nurses. Now, as major healthcare systems like HCA begin incorporating AI technologies, nurses are taking proactive steps to ensure new tech implementations genuinely support patient care rather than add to their workload. Recent contract negotiations between HCA and the National Nurses Organizing Committee reflect this dynamic, with nurses securing unprecedented protections regarding the deployment of AI in their facilities. These agreements signal a growing recognition that while AI may be inevitable in healthcare settings, its implementation must be guided by those directly involved in patient care.

This context underscores why human expert reasoning must remain central to any AI integration in healthcare. AI should augment and support clinical decision-making, not replace it. The complexity of healthcare, the uniqueness of each patient, and the ethical considerations involved in medical decisions necessitate the involvement of trained healthcare professionals who can interpret and validate AI-generated insights within the broader context of patient care.

Given the highly regulated nature of the healthcare industry and lessons learned from previous technological implementations, strong guardrails are essential. These should include:

  1. Rigorous Testing and Validation: All AI systems should undergo thorough testing and validation before clinical implementation, with direct input from frontline healthcare workers who will use these systems.
  2. Explainable AI: Priority should be given to AI models that provide clear explanations for their recommendations, enabling clinicians to understand and verify the reasoning behind AI-generated insights while maintaining their clinical judgment.
  3. Regular Audits: AI systems should undergo regular audits to ensure they're performing as intended, with particular attention to how they impact workflow efficiency and patient care quality.
  4. Data Privacy and Security: Robust measures must protect patient data, ensuring compliance with regulations like HIPAA while maintaining the trust of both healthcare workers and patients.
  5. Collaborative Implementation: Healthcare providers should not only receive thorough training on AI tools but should also be actively involved in decisions about their implementation and ongoing evaluation.

AWS Case Study: AWS has developed a suite of services designed to help healthcare organizations maintain compliance and security. For instance, AWS HIPAA-eligible services enable covered entities and their business associates subject to HIPAA to use the secure AWS environment to process, maintain, and store protected health information. Additionally, AWS offers tools like Amazon GuardDuty for threat detection and AWS Security Hub for security and compliance monitoring.

This approach to AI implementation, grounded in lessons learned from past technological transitions and guided by frontline healthcare workers' expertise, offers the best path forward for realizing AI's potential while avoiding the pitfalls of previous healthcare technology rollouts.

Conclusion: A Cautiously Optimistic Future

As a clinician who has experienced the challenges of poorly implemented healthcare technology, I approach the promise of GenAI with cautious optimism. The potential benefits are immense – from reducing administrative burdens to enhancing clinical decision-making and accelerating medical research. However, we must learn from past mistakes and ensure that these technologies are developed and implemented with a deep understanding of the healthcare environment and the needs of both patients and providers.

By combining the power of GenAI with human expertise and implementing strong ethical and regulatory guardrails, we have the opportunity to create a healthcare system that truly serves both patients and providers. It's an exciting time in healthcare, and I look forward to seeing how we can harness these technologies to improve patient outcomes and make the practice of medicine more efficient and rewarding.

The path to successful GenAI implementation in healthcare requires deep expertise in both technology and clinical workflows. At Caylent, we have guided healthcare organizations through successful AI transformations while maintaining focus on what matters most - improved patient outcomes and provider satisfaction. Get in touch with our healthcare technology specialists to learn how we can help you navigate this journey and realize the full potential of GenAI in your organization.

Generative AI & LLMOps
Kimberly Schaefer

Kimberly Schaefer

Kimberly Schaefer is a Principal Strategist at Caylent, focused on healthcare and life sciences. Offering a unique blend of hands-on nursing experience and extensive expertise in healthcare technology, Kimberly is skilled in turning complex healthcare challenges into actionable, technology-enabled solutions that drive measurable outcomes. With a deep understanding of both clinical workflows and business imperatives, she specializes in leveraging data and innovation to deliver competitive advantages for healthcare organizations. Kimberly is passionate about advancing the healthcare industry through strategic problem-solving, collaboration, and the application of cutting-edge technologies.

View Kimberly's articles

Learn more about the services mentioned

Caylent Catalysts™

Generative AI Strategy

Accelerate your generative AI initiatives with ideation sessions for use case prioritization, foundation model selection, and an assessment of your data landscape and organizational readiness.

Caylent Catalysts™

Generative AI Ideation Workshop

Educate your team on the generative AI technology landscape and common use cases, and collaborate with our experts to determine business cases that maximize value for your organization.

Accelerate your GenAI initiatives

Leveraging our accelerators and technical experience

Browse GenAI Offerings

Related Blog Posts

Experiences as a Tech Intern at Caylent

Read about the experiences our summer technology fellow had at Caylent, where she explored cloud computing, generative AI, web development, and more.

Culture
Generative AI & LLMOps

OpenAI vs Bedrock: Optimizing Generative AI on AWS

The AI industry is growing rapidly and a variety of models now exist to tackle different use cases. Amazon Bedrock provides access to diverse AI models, seamless AWS integration, and robust security, making it a top choice for businesses who want to pursue innovation without vendor lock-in.

Generative AI & LLMOps

Caylent Leverages AWS to Support the Development of BrainBox AI’s ARIA

ARIA is the World's First Generative AI-Powered Virtual Building Assistant Designed to Enhance Building Efficiency and Reduce Greenhouse Gas Emissions

Caylent Announcements
Generative AI & LLMOps