By Anna Bergeron ’24 Associate
AI has opened the doors for innovation, and is now used in countless different ways and in almost every industry. Companies like Google, Microsoft, and Amazon are racing to be at the front of this wave of innovation by funding early stage AI companies. As a result, AI-related startups have almost doubled the share of total startup funding they received from 11% in 2022 to 26% in 2023 to date. AI is increasingly being seen in the healthcare industry as it can be used in many areas of healthcare services, including administrative services, patient adherence and engagement, and diagnostics and treatment recommendations. The newest, and maybe most pivotal applications of AI in the healthcare industry is machine learning being used to diagnose and create treatment plans for patients. The Healthcare Technology market was valued at $174.3B in 2022, and has a 18.2% projected CAGR over the next 7 years. This large projected CAGR is fueled by new technological innovations like AI.
Machine learning is used to evaluate a patient’s condition, and consider other factors about the patient to guess the most likely diagnosis and the treatment plan with the highest probability of success. This application of machine learning has actually been a focus of its development since the 1970’s, however models developed up until recently were usually focused on detection of one disease or one type of cancer. Recently this has changed with models being developed that intend to diagnose and recommend treatment for a wider variety of conditions and illnesses. This technology is still in its very early stages, but early models have proven to be able to diagnose conditions with better accuracy than humans. So, will this completely change how people go to the doctor in the future?
The answer is complicated. The statistical approach that machine learning models use has proven to be very effective in making correct diagnoses; but there are many barriers to entry for models like this and many risks once implemented. This could mean that AI diagnosing isn’t the industry-altering innovation that it appears to be at first.
The primary barriers to entry concern the adoption of these models. First, many industry experts have questioned whether a cultural shift from patient-doctor relationships to a more automated and less personal system is actually plausible. Next, the fact that the daily use of these models will require some sort of approval by a regulatory body which does not exist yet. The World Health Organization has released general guidelines for the use of AI in healthcare, and has called for there to be more regulation on AI using healthcare data; but there is not yet a timeline for these regulations. Regulation of these models is vital, because in order for machine learning models to accurately diagnose patients they need to use quality data. If models are using bad or incorrect data, imperfections and biases will be amplified by the model in its diagnosis and treatment plans. Another aspect of data regulation is data privacy, as models will need to adhere to HIPAA, and there are not yet rules outlined for how specifically AI models can do this.
With AI coming to popularity, especially in early stage companies, it is difficult to say what the future holds for the AI-based companies entering the healthcare technology market. There are a lot of risks and changes that will need to take place in order for AI to become even close to the standard for doctors visits.
Anna Bergeron is a senior from Portsmouth, New Hampshire. She is pursuing a dual major in Economics and International Affairs, with minors in Spanish and Business. She is involved at UNH as a Paul Honor’s student, a member of the Women in Business club, and works as a lifeguard and swim lesson instructor for Campus Recreation. She is excited for her second semester with the fund to continue learning about private equity investment and entrepreneurship.