Expert Tips for Successful Gen AI Implementation in Healthcare
11 min readGenerative AI (Gen AI) is set to dramatically change healthcare by offering transformative solutions to enhance patient care and improve operations. To successfully implement Gen AI, it will be important to identify high-value, low-risk applications, while also ensuring security and compliance. Data availability and organizational readiness are critical for measurable success. We talked to Peter Burns, Director of Consulting and Healthcare Solutions at SoftServe, to get his insights on how payer and provider organizations can successfully implement Gen AI.
What are the biggest challenges healthcare organizations face when adopting Gen AI?
Peter Burns: The challenges fall into three categories: Data availability and integration, the transparency and application of models, and regulatory and ethical concerns.
1. Data availability and integration: One major challenge is combining different data types to get a complete view of healthcare entities and draw accurate conclusions based on all that data. Integrating data for Gen AI insights is tough and keeping it current and aligned with other systems is even harder. Right now, there's no good way to do this. Gen AI insights are only as good as the data they use.
2. Transparency and application of models: When you apply Gen AI models in healthcare, it's important to create and train them properly, draw explainable conclusions, and gain acceptance in their application. Gen AI models often lack transparency, meaning we don't always see how data links to conclusions. To encourage trust, people must be able to see how the conclusion was reached. And, if the solutions aren't integrated well and accepted into administrative processes like care or payment administration, the conclusions won’t be as helpful. For the best chance of successful adoption, we need to demystify Gen AI applications and help workers see them as just another tool that can help people achieve success in their daily goals.
3. Regulatory and ethical concerns: Healthcare has long struggled with data privacy, security, and regulatory compliance. You need to protect the data used to train models and ensure processes that involve protected health information (PHI) remain secure. You need to follow new artificial intelligence (AI) regulations, and existing healthcare regulations, too. On top of that, you need to make sure that AI conclusions are not biased by race, gender, socioeconomic status, or other classifications.
What's an under-explored application of Gen AI?
I am fascinated by the capability and application of digital twin technology. In a healthcare context, it could be applied to the human body and its environment, but it would require an incredible amount of data. In many ways, we still don't have a complete view of a person. Information like EHRs, medical records, admissions, discharges, and transfers (ADTs), claims, clinical data, specialty visits, behavioral health records, labs, prescriptions, financial details, nutritional data, personality preferences, social determinants of health, and more are often stored in different systems, if gathered at all. These systems often don't communicate, leading to a limited view of a person. Despite advancements in digital twins for other industries, the next level of data expansion for describing a patient's health is often seen as beyond our capability.
As the unified, comprehensive patient profile evolves, it will open many Gen AI opportunities. The ability to generate disease predictions, clinical trials management tools, prescription recommendations, and more will be much more specific and personalized. Learning models will improve at an astounding rate with broad, unified datasets that accurately represent a person and their environment. As preventive care and predictive analysis become standard, this will become more important.
In what ways does Gen AI improve patient care and support clinical decision-making?
Gen AI presents significant opportunities to improve patient care, particularly through predictive modeling of care conditions to generate preventive care plans, enhanced diagnostics management for better clinical decisions, and the creation of personalized treatment plans for better patient engagement and outcomes.
All three opportunities should improve population health and reduce stress on the healthcare system and its providers. For example, using AI to generate early warning signs for chronic diseases can shift treatment plans to preventive or lower acuity stages of a disease. Early detection means we can act sooner, leading to better prognosis, lower care costs, and most importantly, a better quality of life for patients.
Improving care also helps healthcare delivery staff. Enhanced diagnostics lead to more effective treatments, fewer adverse reactions, and more efficient use of resources. This reduces condition severity and administrative follow-up, allowing providers to focus more on proactive and preventive care. This should lead to shorter wait times and better patient experience.
How can Gen AI simplify administrative and clinical workflows?
Healthcare administrative processes are onerous, but Gen AI can simplify them. Current EHR technologies can dictate and interpret a patient’s visit and generate notes automatically. Future advancements will improve the ability of these engines to recognize clinically relevant details, add health specifics to the medical record, and translate health issues into diagnoses and billing codes on the claim. All these capabilities can reduce the administrative burden on healthcare staff and improve the results of the provider and patient interaction.
Gen AI can also help with regulatory compliance. The complexity and changing nature of healthcare regulations make it ideal for AI monitoring. AI can find changes, analyze new requirements, and interpret internal documents to find necessary changes. This can greatly reduce the manual effort in maintaining compliance.
Additionally, Gen AI can simplify care management workflows. Payers and providers analyze data to find useful tests or preventive actions. AI can make organizing and interpreting patient data easier and more effective. As conclusions are drawn from the data, Gen AI can interpret patient preferences and create personalized outreach to encourage patients to be proactive and engaged with their health.
What ethical issues should providers consider when they integrate Gen AI?
The big question today is how can we use Gen AI to drive care effectively and fairly while enhancing (but not replacing) human judgment? Since the early days of IBM Watson, it's been clear AI shouldn't make care decisions without oversight. Human judgment is still of utmost importance in clinical decision-making. For now.
Ethical considerations have more to do with “Do you serve all types of populations in a similar way?” This is a complex question because we don’t serve all populations the same way today, regardless of Gen AI. There are fine lines to walk. In the meantime, it's good that current discussions include a fairness doctrine to ensure AI helps all populations equally.
In many applications like voice recognition, physician accountability partly relies on Gen AI's interpretations, raising ethical questions about visits and treatments. When Gen AI makes interpretations or conclusions, it can change patient care dynamics. Clear accountability, transparency, and adherence to ethical standards are just the starting points for Gen AI in healthcare.
What’s the first step healthcare organizations should take to get started?
To make Gen AI work, you've got to align your data and ready the organization. Our study shows — and many companies we talk to agree — that data and operational alignment issues are two of the main reasons Gen AI projects fail.
Start by cleaning and organizing your data. Then, integrate data sources before feeding them into the Gen AI model. Ensure the organization is ready for the changes and the opportunities that the new technology offers. Once the data is solid and the organization is bought in, focus on specific, high-value, low-risk use cases to create advantages and prioritize based on your business goals.
5 STEPS TO GET STARTED
Clean and organize your data
Align the organization around the incorporation of the new technology
Integrate data sources before feeding them to the Gen AI model
Focus on specific, high-value, low-risk use cases
Select an experienced technology partner
Why is it important to select an experienced partner?
Selecting the right technology partner for Gen AI implementation is important. You need a partner with proven ability, proven methodologies, and a deep understanding of its potential and limitations. You need a partner who understands your business, helps with organizational readiness, and brings other specialties to the table as well.
SoftServe is uniquely qualified to implement Gen AI and related technologies. We offer technical expertise, an industry understanding, and a large team of data scientists to bring the technology to life in your organization. Our engineers are backed by R&D, which keeps us at the forefront of new advancements. And we also have strong vendor relationships for Gen AI, including strategic partnerships with Amazon, Google Cloud, Microsoft, and NVIDIA.
Learn how to successfully implement Gen AI in your organization.