Andriy Sambir, CEO of Linkup Studio.
As the CEO of a digital product development company, I often collaborate with healthcare organizations. Recently, I’ve noticed a surge in inquiries from hospitals, clinics, pharmaceutical businesses and health insurers wanting to incorporate AI into their workflows.
Based on the insights from my team and our client interactions, here are some pivotal considerations for bringing AI into the healthcare arena.
1. The Need For AI
Many companies jump on trends without fully understanding how to utilize the technology. Ask yourself: Why do you need AI, and what do you expect to improve?
2. Quality And Source Of Data
AI operates based on the information it’s given. Thus, it is essential to ensure that the data it uses is reliable and trustworthy for optimal results.
For instance, Google’s DeepMind developed an AI to detect eye diseases using data from Moorfields Eye Hospital. This vast, high-quality dataset enabled the AI to achieve results that were as accurate as human experts and, in some cases, better.
In my company’s practice, we developed a mobile app identifying skin diseases, and we used data from over 240,000 people to ensure better patient results. However, we still made sure to note that patients should turn to a specialist for diagnosis.
In case your healthcare institution uses previous medical care data, always ensure that it is of the highest quality possible and comes from a reliable source. This will help your AI produce accurate results.
3. Privacy And Ethical Implications
It’s important to align your AI tool with HIPAA requirements. Besides that, AI can sometimes raise ethical questions, especially when there’s potential for bias in diagnosis or treatment recommendations.
For example, researchers from UC Berkeley studied an AI system used by some hospitals to predict patient hospitalization risks from 2013 to 2015. They discovered that the system predicted white patients would need extra care more often than Black patients with similar health conditions. This bias stemmed from past human hospitalization decisions, potentially influenced by healthcare specialists’ racial biases.
Therefore, in my consultations, I always advise to consider privacy and ethical oversight and consistently evaluate AI’s fairness to reduce its biased decisions.
4. Clinical Validation And Testing
Before any large-scale implementation, AI solutions should be thoroughly validated in clinical settings—effectiveness should not be based only on someone’s assumptions.
A case study compared the consistency of IBM’s Watson for Oncology treatment schemes with healthcare professionals. It found that around 16% of treatment methods the AI recommended were not supported or recommended by medical practitioners.
The lesson is that clinical validation in real-world settings is paramount to ensure AI’s safety and efficacy.
5. Staff Training And Acceptance
I’ve found it is easier to integrate AI if employees understand its value. One of the main problems with implementing AI into healthcare is involving staff throughout the implementation process of AI systems.
However, a recent study found that artificial intelligence screens breast cancer, as well as two radiologists working together. These results will help specialists reduce their workload and help to get even better results. Despite any initial hesitation and distrust, proper training and familiarization with the system can help your medical staff improve their performance.
6. Integration With Current Systems
Seamless integration of AI with your existing electronic health record (EHR) systems and workflows is crucial as well. In my experience, medical institutions considering integrating AI into their EHR systems most commonly need to invest significantly in new infrastructure to support these requirements. This may include purchasing new hardware, such as servers and graphics processing units (GPUs), or expanding their existing infrastructure.
Clients also usually need to upgrade their software to support AI integration. Thus, AI systems can be expensive as they require quite a lot of processing power to train and make predictions.
7. Return On Investment (ROI)
I always recommend carefully weighing AI’s benefits against implementation costs and further maintenance.
Take, for instance, the healthcare network known as CommonSpirit Health. They introduced an AI-powered messaging system to enhance patient communication and care planning. After implementing this tool, new mothers’ hospital stays dropped by 10%, and early deliveries fell by 37% among the Medicaid population. In orthopedics, hospitalization time was reduced by 45%, with a notable 71% decrease in one-month readmissions.
However, in other situations, AI might be less efficient and can be costlier than its benefits.
8. Continual Learning And Updates
Often, company leaders think AI will be a one-time investment. It is not like that. Models need to be updated as medical knowledge evolves and new data becomes available.
Examining prices overall, I’ve found a custom AI system typically costs between $20,000 and $1 million. An MVP falls within the $8,000 to $15,000 range. Based on my discussions with clients and industry peers, the annual maintenance cost for an AI system averages between $5,000 and $100,000.
9. Proper Expectations
I’ve seen many companies treat AI as a magic bullet to solve a large portion of their business issues. However, it’s better to approach new technologies with caution.
One lesson from the well-known startup Theranos and the subsequent convictions of its founders is a reminder not to market our new solutions as game-changers unless they are fully proven and tested.
AI can have shortcomings, and in healthcare, the outcomes of mistakes can be grave.
10. Niche Professionals
I advise you to assign an SME (subject matter expert) to collaborate with the software development team, offering medical expertise and assisting in AI algorithm creation and validating results.
For the software development team, I suggest opting for specialized healthcare tech professionals over generalists. Choosing a team that’s already tackled your kind of challenges before is ideal. Working with generalists can be time-consuming, as you’ll need to explain your needs in detail and devote more time. Specialized teams often offer a smoother process.
AI unquestionably streamlines healthcare businesses. To excel in this endeavor, the best route is to make informed decisions throughout the implementation process based on the considerations mentioned above.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here