Generative AI in Healthcare: Insights from Dr. Liz Kwo

For many patients, healthcare means a labyrinth consisting of administrative tasks, clinical decisions, trips back and forth between various medical specialists and a state of anxiety caused by questions like “is my medical insurance going to cover for all of this?” or “how much money will I have to pay out of pocket?”

The good news is that our healthcare - with a stated goal to provide the best medical care for each patient - has found reliable support in achieving this objective: generative Artificial Intelligence.  

I want to look at the most relevant uses of generative AI in healthcare and the requirements healthcare stakeholders should meet in order to make the most of it.

What is generative Artificial Intelligence?

Gen-AI is a technology that uses deep-learning algorithms to create code, text, audio and other types of new content.

Gen-AI can analyze vast amounts of unstructured data, meaning information not organized according to a predefined model, such as medical recordings, graphics and clinical notes and generates results that can be used combined with structured data like insurance claims.

What do healthcare stakeholders need to do before adopting generative AI?

1. Assess the organization’s landscape

Gen-AI requires a thorough assessment of a healthcare organization’s technological resources, operations and employee capabilities. This assessment is vital for establishing how gen-AI can serve the organization. To determine that, healthcare leaders must identify the most relevant processes in their organizations. 

The input of department heads reviewing with individual contributors in charge of data and technology can provide the insights as an effective upleveling of skill, and not as a fragmented solution.

2. Partner with the right companies

Adopting generative AI means that healthcare organizations will need to make several changes, the most relevant being:

• learn to use gen-AI driven platforms;

• train employees and help them acquire the skills needed to operate gen-AI applications and processes involving this technology;

• implement easy-to-use applications that will not increase the workload of the staff or keep them away from patients longer that needed;

• educate employees regarding the evolving nature of their jobs and reassure them that gen-AI is there to optimize a part of their work and not to replace them;

• maintain a human in supervising the functioning of the generative AI technology, who can intervene when errors occur;

• enter partnerships with technology companies that comply with regulatory requirements such as the Health Insurance Portability and Accountability Act (HIPAA);

• cooperate with other healthcare entities that posses large amounts of data sets in order to optimize this technology for everyone.

3. Address the risks

Given the importance of data protection, health organizations looking to integrate gen-AI in their operations must make sure the risks of data breach are mitigated properly. The sensitivity of this data requires a type of protection that gen-AI may not always provide, making it the healthcare leaders’ job to identify and implement the right level of data security.

Another risk is that the information delivered by generative AI needs to be reviewed for accuracy and never cause harm. People can oversee the process to decrease that risk. And since human errors factor in, healthcare leaders can do that by defining risk and legal terms to govern the implementation and use of gen-AI in the organization, including but not limited to accountability and regulatory compliance.

Although the steps needed to implement gen-AI may seem long, the effort is worth it: the automation of tiresome and susceptible-for-errors operational tasks for instance contributes to modernization of healthcare infrastructure and delivers a huge volume of clinical data to providers at their fingertips.

Plus, these organizations will have improved efficiency and a higher degree of experience in using state-of-the-art technologies for improving the health outcomes of patients. 

How is generative Artificial Intelligence already helping medical (tech) companies?

1. Bayer Pharmaceuticals and the clinical trials

Gen-AI has the potential to speed up the process of new drugs’ discovery, decreasing the time and costs associated with it. Bayer Pharmaceuticals has began to investigate how gen-AI solutions like Google Cloud’s Vertex AI and Med-PaLM 2 can turn this objective into reality. This helps researchers find and access large sets of data, mining them for establishing potential correlations and automating the drafting of clinical trial communications. This decreases the time spent on reading, comparing and making connections throughout years of research materials and takes scientists faster to the development of new revolutionary drugs and treatments.

2. Meditech and the Electronic Health Records

Medical records contain complex, high-volume data that help clinicians establish a diagnosis, ask for specialty consults, decide treatment, prescribe drugs and monitor the evolution of a patient's health.

However, many times the patient data is stored across several systems and that makes it difficult for clinicians to access this information completely and in time. Meditech Expanse is an initiative designed to boost the collection of patient data from different sources with the help of gen-AI in order to obtain a comprehensive view of a patient’s records. Automation of clinical documentation like tracking the changes in nursing shifts improves the efficiency of care and saves medical staff time they can dedicate to patients.

3. HCA Healthcare and the administrative tasks

Administrative tasks like patient handoffs between nurses is an essential, yet time consuming process where nurses who end their shift bring up to speed nurses who begin their shift by communicating them relevant patient information like vital signs, test lab results and evolution of their treatment. 

HCA Healthcare has started to work with Google Cloud to automate and standardize this handover with the help of gen-AI in order to save time and improve patients’ health.

Organizations like Huma Therapeutics, Ginkgo Bioworks and Infinitus Systems Inc. are also using generative Artificial Intelligence to tackle different struggles of healthcare and implement new solutions to improve the quality of care provided.

What are the uses of generative AI in healthcare?

Gen-AI can assist private payers and hospitals and physicians in achieving their objectives.

For private payers, gen-AI is an efficient tool for:

• creating online questionnaires with questions regarding members benefits and coverage;

• suggesting clinicians based on parameters like location and coverage of patients;

• centralize clinical notes and medical and information for healthcare managers;

• generating care plans for insured members;

• creating content for outbound nonclinical communications;

• comparing provider networks and product features;

• generating reports about provider performance;

• creating standard communication forms like claim denials or welcome letters; 

• generating denied claims issues and identify solutions;

• issuing summaries and outcomes for prior authorization requests;

• analyzing consumer distribution to create customized plans and products;

• improving sales support to help potential members understand coverage and choose the right plan;

• developing “first draft” product overviews for employers, broker, Affordable Care Act and Medicare Advantage members;

• generating corporate reports in standard formats;

• providing coverage updates for policyholders and centralizing updated legal and risk processes when regulations change;

• automating accounting by extracting relevant data;

• generating KPIs and reports across corporate functions.

Hospitals and physicians can use gen-AI to:

• leverage structured and unstructured data to generate summaries, videos or images for patient education;

• outline discharge information and follow-up instructions for post-acute care;

• improve documentation accuracy

• recommend customized risk measures for patients based on their medical history and medical literature;

• centralize for primary care clinicians the notes issued by medical specialists;

• evaluate market performance and make comparisons based on external data and resources;

• issue value-based care contracts depending on market characteristics;

• create customized training materials for clinicians;

• generate care coordination notes;

• issue workflows and schedules for different processes and locations;

• generate dictations and messages;

• use coding to automate repetitive tasks;

• automate reimbursement coding based on clinicians’ notes;

• issue summaries detecting coding errors in claims; 

• draft procurement contracts and vendor communications;

• generate purchase orders based on stock level;

• support HR efforts by creating offer letter and generating education materials for new employees;

• create legal, financial or compliance reports.

What are the most promising uses of generative Artificial Intelligence in healthcare?

These past years have seen the efforts for integrating gen-AI into healthcare in full bloom: tech giants like Oracle, Microsoft and Amazon have gone out of their way to develop services and products that can alleviate the burden of healthcare employees’ shoulders.

1. Note drafting 

Healthcare providers are allocating significant amounts of time to writing notes about the conversations they have with their patients, notes that must be documented accordingly. Currently, gen-AI can automatically draft these notes, leaving providers with more time to search for treatments for their patients.

2. Identify SDoH

A study released this year in January revealed that large language models of gen-AI can identify SDoH like employment status or housing from clinician notes, enabling the identification of patients who need help for improving their health. 

The information in these notes is often not organized in health records and there it is difficult to identify the patients who need additional help. This is where algorithms come in, noticing details that providers may miss and prove relevant for improving the health outcomes.

3. Check symptoms online

Online symptom checkers have been on the mind of groups like 98point6 and K-Health for a while, part of the effort to provide more specific health information, without asking a doctor. Currently, online symptom checkers are connected to Digital Health providers and communicate with licensed clinicians to prescribe medication, order tests and lab investigations or prescribe reliable instructions for self-care.

4. Educate patients

Many patients have trouble remembering everything their providers tell them, resulting in poor compliance and poor adherence to prescribed medication.

To have an impact, the communication about their condition, medication and treatment must be customized, consistent and compelling.

Gen-AI can support this objective by generating and delivering customized content, focused on the specific illness, medication and life conditions of each patient instead of using a one-to-fit-all patients education approach.

Gen-AI can provide detailed information that goes beyond standard inquiries like “what is meningitis?” It can provide answers to questions like “how is Pulmonary vasculitis diagnosed?” or explain the terms you find on imaging or lab results.

5. Explain patient coverage and benefits to healthcare providers

Healthcare providers are faced with a maze of conditions where services rendered must meet criteria in order to be compensated by the payer. The patient must have active coverage, the provider must be in-network and the healthcare service rendered must be covered by the patient’s payer plan. With variables such as deductibles or out-of-pocket payments in the patients’ health insurance plans, it can be very time consuming for healthcare providers to determine what they need to do to get paid for their work.

Gen-AI is starting to extract, collect and organize all this information, making it accessible and comprehensible by answering even difficult questions like those related to enrollment or renewal of health plans.

Some precautions are needed

Regardless of how revolutionary and helpful gen-AI is and can be for the healthcare sector, some precautions are needed when using it to make sure the results it provides are accurate, complete and relevant.

For instance, some AI platforms DO not have up-to-date information for users without premium (paid) subscriptions. A healthcare professional must be aware of the latest medical developments and since the medical world is evolving. 

Source checking is another factor we must consider when using gen-AI. The main advantage of a Google inquiry for instance is that the results displayed indicate their source to fact check a reputable site. With gen-AI, things are not always as clear, which requires refining and double-checking the information provided.

In other cases, users have reported that although the information provided by generative Artificial Intelligence platforms was correct, the cited sources don’t always feature answers to the questions they addressed. In other instances, the links indicated don’t exist or if they do, when you click on them, you get the “page not found” result which can seriously affect the reliability of the information offered by gen-AI.

Conclusion

Despite its potential issues, generative AI is leveraged by the 5P stakeholders - payers, providers, patients, pioneers, and policymakers - to help them reduce their expenses and radically transform the way we provide better health services, more informed medical decisions and improved health outcomes. 

The next few years will see a surge of gen-AI developments and I for one look forward to seeing how it will help improve the quality of healthcare around the world.

Sources provided by you:

Additional sources: