AI OVERVIEW
The field of AI is dynamic, constantly evolving and inherently complex. Several workflows are described below that will help case managers and other healthcare professionals better understand AI.
Machine Learning (ML): ML is probably the most common AI-related application that case managers encounter today. ML incorporates historical information that the computer utilizes for inferences in an if-then scenario to provide information regarding potential outcomes. Thus, these algorithms empower case management systems to learn from historical data and make predictions without being explicitly programmed. In case management, ML models can be utilized for:
- Predictive Analytics: Forecasting case outcomes based on past data patterns, assisting in resource allocation and decision-making. Examples of predictive analytics may apply for suggested patterns for recommended post-acute placement, or interpretation for discharge milestones based on clinical markers.
- Risk Assessment: Identifying high-risk cases or individuals that may require special attention or intervention. The most common of these would be the utilization of readmission risk scores or high-risk population health calculations to determine which patients should be evaluated for case management interventions.
- Automated Documentation: Generating case reports, summaries, or documentation based on case details and precedents. This technology uses rules to pull from key components in the case management text fields to provide a summary of services provided.
Natural Language Processing (NLP): NLP enables machines to understand, interpret and generate human language. NLP may be our current favorite toy as we learn more about ChatGPT and other, similar products that extract various amounts of information from the internet and provide answers to various questions. In case management, NLP finds applications in:
- Text Analysis: Parsing and extracting relevant information from unstructured case notes, emails or documents to aid in decision-making. Future case managers will be able to pull in multiple types of information from electronic health records (EHR) to obtain a synthesized picture of what is going on with a patient without having to manually sort through the patient records.
- Sentiment Analysis: Assessing the sentiment or emotional tone within case communications, which can provide insights into client satisfaction or potential issues. A common use case is for patient satisfaction surveys, which can evaluate patient responses and sort them into positive, negative and neutral categories. Sentiment analysis can pick up on key inferences from documentation and conversations to provide feedback based on the individual’s tone (Ramirez, et. al, 2019). In the future, this will help inform patient interventions and treatment strategies.
- Chatbots and Virtual Assistants: Offering 24/7 support to clients by answering queries, providing information and guiding them through case-related processes. Such chatbots and virtual assistants are not as common in healthcare settings. They are more dominant in commercial businesses where they may be used for tasks such as booking a hotel reservation. In the case management field, chatbots and virtual assistants will likely be used as a means to support patients who have questions about their discharge instructions or want to triage an issue with their care team. This offers flexibility for patients who may have questions outside of office hours.
Robotic Process Automation (RPA): RPA involves the use of software robots to automate repetitive tasks and streamline workflows (Itransition, 2024). The best use for RPA is for any task that is completed on the computer in a consistent and repetitive fashion. In case management, RPA can be applied for:
- Data Entry and Validation: Automating the input of case details into management systems and validating information for accuracy. This may occur while retrieving and documenting information from other clinicians working with the patient.
- Routine Administrative Tasks: Managing mundane tasks such as scheduling appointments, sending reminders or updating case statuses. This process is more common for case managers who are pulling reports from one system and “automatically” sending them to another. In some post-acute vendor software, this exists without case management manual intervention by pulling specific hospital notes to send to the post-acute facilities on a scheduled basis.
- Workflow Orchestration: Coordinating tasks across multiple systems or departments, ensuring smooth case progression and collaboration.
Decision Support Systems (DSS): DSS leverage AI algorithms to assist human decision-makers by providing relevant information, insights, and recommendations (Itransition, 2024). In case management, DSS has potentially the greatest opportunity for future benefits. DSS aids in:
- Case Prioritization: Recommending which cases should be addressed first based on factors like urgency, complexity or risk. DSS would allow for caseloads to adjust based on readmission risk scores and would be modified continuously based off the prioritizations for patients needing timely intervention. DSS already exists by some firms in the clinical documentation integrity (CDI) and utilization review specialty, which employ an active worklist with front-facing information for the specialists on the potential next steps of action.
- Resource Allocation: Suggesting optimal allocation of personnel, time or funding resources to address cases. In healthcare, DSS is primarily emerging in tools that look at acute care bed management. It can help evaluate key insights based off the patient’s clinical factors to identify their best location for bed assignment and when they should be moving toward discharge. By far, this is the greatest hope toward real-time progression of care management and has the potential to eliminate the overreliance on manual human intervention.
- Compliance Monitoring: Analyzing cases against regulations, organizational policies and other emerging standards to ensure adherence to those requirements and mitigate compliance risks.
CHALLENGES
Implementing new technologies in healthcare is often challenging. This is especially true when leveraging AI applications for case management programs.
The Consumer Technology Association (CTA) has identified several key challenges that healthcare professionals must address (CTA: ANSI/CTA Standards, 2020). These include:
- Optimizing Data Availability and Data Interoperability: In big data platforms that often drive robust and dynamic AI models, identifying the right data at the proper time is important. Bad data in leads to bad data out.
- Promoting Accurate Predictions or Output: ChatGPT and similar applications can generate false information such as citations and references. AI “hallucinations” are a major concern, so the accuracy of AI-generated information must be carefully peer-reviewed and checked through a quality-control process.
- Avoiding Data or Algorithm Shift: As noted by the CTA, “The performance of AI algorithms is impacted by the distribution shift—a change of statistical property of the data encountered when the algorithm is deployed compared to the data used to train the algorithm.” Therefore, it is important to implement QC protocols that track and adjust for these types of shifts to ensure the AI application is still accurate, and hopefully getting more accurate (not less).
- Fostering Clinical Trust: When case managers and other providers use automation to foster clinical decision-making for their patients, trust becomes an important factor for clinicians to accept and use those areas of delegation. AI and related IT applications are not always transparent in terms of the methodologies used when automating workflows and generating treatment recommendations.
- Ensuring Regulatory Oversight: Beyond the initial oversight offered by several federal agencies such as the U.S. Food and Drug Administration, other government agencies are developing regulatory schemes for AI applications in healthcare. Ensuring that there is meaningful oversight is going to be critical especially due to some press accounts of the dangers associated with AI. The White House has announced a series of initiatives on how to regulate AI, including several recent executive orders (Executive Orders, 2023 and 2024). The European Union also recently approved the first comprehensive law internationally to govern AI (EU AI Act, 2023).
A central value that case managers hold is offering a personal touch to coordinate care to promote better health. However, IT systems can potentially undermine case manager efforts when not configured or implemented properly. CMSA has learned over the years through its health information technology (HIT) studies that EHR systems, including the care management software system, sometimes distract more from care coordination activities than help (HIT Studies, 2012). Case managers have reported that EHR systems decreased their efficiency and took away time with patients because they were spending more time logging in data or battling other aspects of automation. When not implemented or used properly, AI will also likely undermine the high-touch care coordination that is a hallmark of case management programs.
When AI is implemented effectively, however, case managers can use their EHR systems to coordinate care much more effectively. One prime example is leveraging IT systems to develop, implement and update care treatment plans that factor in all of their patients’ co-morbidities. In a similar fashion, AI provides an opportunity to take case management to the next level. Often, IT solutions like this are a real game-changer.
THE POTENTIAL
The integration of various AI technologies into case management processes offers significant potential to enhance efficiency, accuracy and decision-making. From predictive analytics and natural language processing to robotic process automation and decision support systems, the diverse applications of AI are revolutionizing how patients are managed across different settings in the healthcare landscape. By embracing these technologies, organizations can optimize resource utilization, improve client outcomes and navigate complex cases with greater agility and effectiveness.
REFERENCES
Consumer Technology Association (CTA) “ANSI/CTA Standards: Definitions of Artificial Intelligence in Health Care,” ANSI/CTA-2089 and ANSI/CTA-2089.1 (February 2020).
Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, White House (October 30, 2023): See White House briefing room or see the following link retrieved March 31, 2024 at https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/. More recently the White House Office of Management and Budget issued a follow-up order announcing the first government-wide policy to oversee AI (March 28, 2024): See White House briefing room or see the following link retrieved March 31, 2024 at https://www.whitehouse.gov/wp-content/uploads/2024/03/M-24-10-Advancing-Governance-Innovation-and-Risk-Management-for-Agency-Use-of-Artificial-Intelligence.pdf.
EU AI Act: first regulation on artificial intelligence. Topic European Parliament (2023): https://www.europarl.europa.eu/topics/en/article/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence.
Habehh H, Gohel S., “Machine Learning in Healthcare,” Curr Genomics. 2021 Dec 16;22(4):291-300. doi: 10.2174/1389202922666210705124359. PMID: 35273459; PMCID: PMC8822225.
HIT Studies—CMSA co-sponsored with Schooner Strategies four studies in 2008, 2010, 2012 and 2022 examining how Health Information Technology (HIT) applications are impacting the practice of case management. For a summary of those studies, see Carneal G, Stricker R., “How Health Information Technology Is Impacting the Practice of Case Management,” CMSA Today (June 2022). To download the HIT studies, see https://schoonerstrategies.com/articles/publications/.
Itransition (February 24, 2024) “Machine Learning in Healthcare: Use Cases, Examples, & Algorithms,” Retrieved on March 24, 2024 from https://www.itransition.com/machine-learning/healthcare. Ramírez-Tinoco, F.J., Alor-Hernández, G., Sánchez-Cervantes, J.L., Salas-Zárate, M.d., Valencia-García, R. (2019). “Use of Sentiment Analysis Techniques in Healthcare Domain,” In: Alor-Hernández, G., Sánchez-Cervantes, J., Rodríguez-González, A., Valencia-García, R. (eds) “Current Trends in Semantic Web Technologies: Theory and Practice. Studies in Computational Intelligence”, vol 815. Springer, Cham. https://doi.org/10.1007/978-3-030-06149-4_8
Image credit: ISTOCK.COM/NUTTAPONG PUNNA