We’ve all used the internet as a self-diagnosis tool to find out the severity of our symptoms. Since Google is known for providing confusing and unsubstantiated information, symptom-checking apps are increasingly becoming a relatively safe alternative to internet self-diagnosis.
Tim Price, chief product officer at Infermedica, which develops the Symptomate app, outlines the amount of information the application can provide.
Price says:[The app] Provide the patient with a range of illnesses that may be causing their symptoms and provide triage recommendations regarding the level of care the patient can receive and its urgency. We also provide patient education materials to help you proactively manage it yourself. Determine whether they are using care or caring for themselves at home. ”
GlobalData predicts that the regulated healthcare app market will reach a value of approximately $12.2 billion by 2030. These apps can not only provide possible differentiation but also triage recommendations, which begs the question of how these apps are developed and where they get their information from. And how accurate are they?
Clinical and peer-reviewed data behind the app
Symptom check applications leverage a variety of technologies, including artificial intelligence (AI) and machine learning. The K-Health app was built using a clinical language model similar to ChatGPT, said Ran Scholl, the company’s co-founder and chief product officer.
Apps like Symptomate and Sensely build algorithms using Bayesian statistical modeling, which models probabilistic relationships between conditions, symptoms, risk factors, lab tests, and other medical concepts. They both behave quite differently and have different levels of monitoring and reproducibility.
Access the most comprehensive company profiles on the market with GlobalData. Save research time. Increase competitiveness.
Company Profile – Free Sample
thank you!
You will receive a download email shortly
We are confident in the unique quality of our company profile. However, we want you to make the decision that is most beneficial for your business, so we are offering free samples that you can download by submitting the form below.
by global data
The language model for the K-Health app was built using anonymized electronic medical records provided by the Mayo Clinic and the Israel Health Maintenance Authority. Scholl said the app not only uses specialized data sets, but also works “in that every time a patient sees something, every time our AI does something, it’s verified by a doctor. He’s not like ChatGPT.” [it]”.
Price said the Symtomate app’s Bayesian statistical model was created using published literature, including “peer-reviewed research, epidemiological data, and cases from some of our customers.” Additionally, the app’s interviews will be used to provide a feedback loop and inform future designs, “all of which have been expertly curated and validated” by a team of over 60 physicians. I added.
Adam Odesky, CEO of conversational AI platform Sensely, which also uses Bayesian modeling, said the model was created using input from the Mayo Clinic, the World Health Organization, the UK National Health Service and other Sensely partners. I did.
Complex and diverse regulations
Apps are regulated differently in some regions, even within the same region, due to differences in development, use cases, and scope. Symptomate and Sensely are both classified as medical devices in the US, but that doesn’t mean much. Price said Sympromate is a Class I medical device under the Food, Drug, and Cosmetic (FD&C) Act, “but the FDA currently exercises enforcement discretion over this type of software. Now, InferMedica’s software products must undergo regulatory filing, review, and approval before being brought to market.”
Furthermore, “Due to the specific purpose of use, [for Symptomate] It is classified as informational rather than diagnostic or therapeutic. Therefore, the FDA has said it will not mandate the process to become a medical device,” Price said. However, this is easier said than done, as the lack of a clearly defined process can make enforcement difficult.
The process is further complicated in the European Union (EU), where Symptomate and Sensely are classified as Class IIa (under MDR) and Class I (under MDD) medical devices, respectively. However, Odesky noted that Sensely will soon be regulated as a Class II medical device in the EU. This creates a confusing pathway for symptom checker apps to regulate.
However, the regulation of these applications is confusing and often has minimum standards. Mr. Price said the onus is on manufacturers to set internal standards as high as possible so that they are willing to recommend products to families who really need care.
Large-scale language models with supervised learning allow AI to derive probabilities for various diagnoses, but they have significant drawbacks due to little or no reproducibility and the inability to accurately document the analysis. Accompanying. Odesky is skeptical about the use of large-scale, unsupervised language models, saying that “generative AI is particularly prone to significant hallucinations,” and that without proper oversight and governance by experienced physicians, they cannot be used in medical settings. said it was potentially dangerous to use.
Price and Odesky said generative AI and large-scale language models, especially those without clinical supervision, need stricter regulation because they are difficult to “truly robustly clinically validate.” I agree. [these] Due to lack of predictability and reproducibility.
The future of the field
Symptom-checking apps increase people’s access to healthcare and have millions of users. Price said more than 1.5 million people have used Symptomate in the past year. In addition to a huge user base, these apps are more accurate than internet searches. In 2023, K-Health published data on the app’s accuracy, showing that 84.2% of cases were correctly diagnosed.
When we talk about the future, we envision these apps going beyond simple integration with primary care to improve efficiency. Odesky believes these apps could work with other medical devices, such as blood pressure cuffs and pulse oximeters, to provide primary-level care, such as managing chronic diseases.
“We can create virtual clinical teams that fundamentally address the problem of understaffing and underfunding for health care, so that it’s not just the lucky few who can’t afford or live on it,” Price said. We hope that everyone will have the opportunity to access basic primary health care.” In areas where it exists, care may or may not be available. ” The future of these apps is far-reaching and unpredictable.