Establishing and Sustaining Health Care findings from "Consumers' Trust in Generative AI" underline that without trust, one cannot realize the transformative potential of generative artificial intelligence in health care.
There is strong customer confidence, but at a paralyzing standstill as far as deployment goes in strategies to foster trust in AI-driven healthcare decisions.
A recent survey, polling over 2,000 U.S. citizens this past March 2024, shows a number of key conclusions. General usage of gen AI by customers has actually decreased to 37 percent in 2024 from 40 percent in 2023, at a time when 66 percent of consumers using it for health-related purposes think it's going to save individual care costs and appointment wait times.
Consumers as much as in the past no longer say they lack faith in information provided by gen AI. Thirty percent of consumers, up from 23 percent, said this in 2024. This number is higher for millennials: 30% in 2024 versus 21% in 2023, and greatly higher for Baby Boomers: 32% in 2024 versus 24% in 2023.
It has been emerging lately that while consumers might all seem very optimistic about the possibility that AI will increase accessibility and lower costs, adoption rates are flat due to growing distrust. Customers majorly rely on physicians' advice when making gen AI-related health decisions.
It requires the management of very sensitive personal data. The output from artificial intelligence needs to be accurate and reliable. Complex laws related to the privacy of patients, security, and ethical concerns need to be addressed. There are consumer complaints about the privacy and security of their data that need to be dealt with. How gen AI is being used with health decisions needs to be communicated clearly.
There is a real need for strategies in AI-driven healthcare to be forthright about the use cases, advantages, and limitations of generative AI in treatment unique to patients. A count of 80% of the consumers, the Deloitte report indicates, would like knowledge of the generative AI their health provider uses in finding treatment choices and impacting care decisions.
Transparency is key, however, said Deloitte AI practice leader Bill Fera: "Hospitals should be looking to educate their patients on how and why generative AI is being used at their organization."
Educational programs featuring seminars, flyers, and one-on-one discussions could help quiet patient and provide anxieties and skepticism about what generative AI can and can't do.
All AI applications must, therefore, adhere to the existing data privacy laws, such as the GDPR within the EU and the Health Insurance Portability and Accountability Act within the US. Patients would always want to be fully sure that information pertaining to personal health is safe and would not be misused.
Another recent paper underscored the relevance of data privacy and security in AI applications. It added that "generative AI systems in healthcare have to be HIPAA compliant with regards to the disclosure of data and secure from breaches."
One way for you to engender strategies to foster trust in AI-driven healthcare decisions while protecting patient privacy is to provide patients autonomy over their data, from opting for or rejecting AI-driven procedures. This makes them feel their privacy is taken care of.
Strategies to foster trust in AI-driven healthcare decisions can be stronger if human therapists remain integral to the AI process. AI information is more likely to be trusted by patients if it has been reviewed and validated by their healthcare providers or practitioners.
It was Rep. Ted Lieu who resounded the need for human control in AI, saying, "Patient care settings can benefit a great deal from the use of generative AI in the delivery of health care." But still, a human must check twice the content generated by AI.
A paper demonstrating the utility of Gen AI techniques in EMRs was just published by researchers at Mass General Brigham. The research did point out some technological limits that would require human surveillance.
There should be good bondage between the patient and the health provider. Providers must explain to the patients in detail how AI is being used to enhance treatment so that the patients understand that AI is a tool to help, not replace, their healthcare team.
“Patient-provider connectedness has to stay front and center," says University Hospitals Chief Medical Officer Patrick Runnels. "On the back end, generative AI is helping us sort out your care, but you'll always have a connection with your nurse, social worker, or doctor."
Partnering with credible community-based organizations can increase reach and establishing strategies to foster trust in AI-driven healthcare decisions. These organizations can serve as trusted brokers to deliver dependable information to the patient.
There is a need to be certain that AI systems are fair and remain promotional of health equity. This must involve training models of AI on diversified data sets but with ongoing surveillance for bias.
It is what, according to a December 2023 article in Nature, proposes "GREAT PLEA" for ethical principles: Governability, Reliability, Equity, Accountability, Traceability, Privacy, Lawfulness, Empathy, and Autonomy.
Establishing trust can be encouraged through ethical tenets that ensure principles like safety, reliability, and fairness in the development and implementation of AI in healthcare. Two growing organizations providing guidelines and best practices for ethical and responsible use of AI are the Responsible AI Institute and the Coalition for Health AI.
The technology can be improved, and confidence built, by actively soliciting and acting upon patient feedback on all AI applications.
A fully functional feedback management mechanism that allows users to provide feedback on the performance of the AI, ensuring quality in AI delivery and user concerns, is embedded in Google's Gemini model.
Regular auditing of the AI systems would further increase confidence that the AIs are working not only securely but also accurately; this would involve checking for any possible errors or failures and acting promptly to rectify these.
Good quality data supplied in training an AI system is its bedrock. Diverse data sets contribute to the limitation of the precision of AI and help avoid bias; this increases users' confidence in the systems. Maintaining model integrity and having results that reflect real-world medical settings will require uncompromising commitment to data quality.
For this commitment to be effectively expressed, the organization will have to be open in communication. For example, at Caresyntax, it has constantly worked on the inclusion of a myriad of medical scenarios in datasets. This process helped implement AI algorithms with surgical team guidance and real-time intraoperative decision support.
This has increased patient safety and reduced surgical variability.
Once adequate trust is built in narrow AI by healthcare systems, this can quite significantly affect patient outcome. However, it is not only the patient who will gain from such tech. Our world equally stands to benefit from narrow AI so much as this eventually leads to less energy and resource wastes.
The impact of the healthcare industry on climate change is huge, as it accounted for 4.6% of global greenhouse gas emissions in the year 2017. As known, almost all hospitals are relatively energy-intensive facilities, and generating hospital clinical waste of 50-70 % originates from the ORs. This is believed to generate 9.7 million tons of CO2 annually in the US, UK, and Canada.
However, our own data actually shows that some of these very limited AI interventions in the OR—harnessed through a platform like Caresyntax—significantly reduce energy use and waste from being more efficient in surgeries. Narrow AI can make a significant contribution to a greener, more ecologically sensitive healthcare sector.
The potential that generative AI has to offer in healthcare is in changing the way patient care is conducted and decreasing costs while increasing efficiency. However, there is one critical barrier during its wide adoption—building trust. Consumers and health providers have raised their voices related to data privacy and the reliability of AI-generated insight and ethical considerations.
It will be required to exercise transparency, educate patients and healthcare providers about the potential and limitations of AI. Data security and privacy guarantees the continuing role of humans within AI processes. This includes patient relationships and adherence to ethical guidelines, active seeking and implementation of feedback for continuous improvement, and standards for unbiased use of data.
1. What are the principal concerns that hamper the full adoption of generative AI for healthcare?
This means that most of the trust issues surround this domain: data privacy, reliability of AI output, or ethical implications. Other concerns of these patients and providers are data security breaches and possible biases in AI decision-making.
2. How might healthcare organizations improve transparency with regard to the utilization of generative AI?
Healthcare providers should be transparent in describing their AI use in patient care, including hospitals' benefits and limitations. Educational efforts accompanied by transparent and understandable information will soothe fears and misconceptions.
3. Why human control in generative AI applications in healthcare
Human oversight offers substantiation that the suggestions generated by Al are safe for implementation. This aids in preserving the confidence of patients and assures the correctness of medical decisions.
4. What are the ethical concerns for the deployment of AI in healthcare?
Ethically conformable AI means fairness, accountability, transparency, and privacy constraints. Healthcare organizations have to improve fairness in AI algorithms and preserve patient rights to autonomy and protection of informed consent.
5. How might patient feedback improve AI applications for health
Soliciting and implementing patient feedback refines AI systems to better align with the requirements and expectations of the user. This will also help in building trust by demonstrating that their concerns and suggestions are being heard and acted upon.