Have you ever gone on a “diagnostic journey” – sorting through dozens of WedMD papers, consulting multiple doctors, asking friends and family – to determine the cause of your health-related symptoms? ?
Like 80% of Americans, you might be too.
That’s where Ada Health comes in. This is a mobile he app that allows people to quickly diagnose their symptoms by evaluating their disease against a database of 3,600 symptoms. And AI is being used to make that happen.
Since its launch in 2016, Ada’s app now covers 99.5% of all medically diagnosable conditions. The study examined more than 10,000 symptoms and risk factors across medical fields, including mental health, pregnancy, rare diseases, and pediatrics, and included more than 31,000 different ICD-10 codes (codes for diagnosis and claims processing procedures). system). This app has over 13 million users worldwide.
“Scouraging all the possibilities and information available online can lead to a great deal of confusion and anxiety,” says Dr. Claire Novorol, co-founder and chief medical officer of Ada Health. “Our AI-based system assessment and severity-based care navigation help eliminate inefficiencies in patient visits and accelerate time to diagnosis, treatment, and care.”
Novorol founded Ada Health with Daniel Nasras, who is now CEO, and Dr. Martin Hirsch, a former medical researcher who sought to apply probabilistic reasoning approaches to clinical diagnosis. Novorol said he felt there was a more personalized, detailed and reliable way to self-assess symptoms than online searches, which is why he created the Ada app.
This app utilizes a probabilistic inference system using Bayesian techniques, one of the core concepts of computer science and AI. Its probabilistic inference takes into account the existing evidence base. Novorol explained that the most frequently entered symptoms are fairly common, such as cough, headache, abdominal pain, fatigue, and nausea. But what sets Ada apart is the product’s ability to distinguish between common and uncommon causes and identify potentially rare conditions.
“This is a long tail of thousands of atypical symptoms and symptoms,” she says. “At the same time, we consistently deliver safe and appropriate outcomes. We don’t send every patient to emergency care, which is costly and overburdens the system.”
Novorol said Ada is a “white box system,” a form of explainable AI. Whatever the proposal, be transparent about how the technology arrived at its results. It identifies medical reasoning and logic. “This level of rigorous validation is essential in our field,” she said. “We take safety very seriously.”
For clinicians, Ada has become an essential tool in the pursuit of patient-first care. Novorol explained that Ada eases the burden on overworked clinicians by helping patients find the right treatment faster. For healthcare providers, electronic medical records are automatically imported, saving them the time-consuming task of manually documenting medical records. In the EU, Ada’s apps are technically considered Class II medical devices under the provisions of the Medical Devices Regulation. It is one of the few AI-based symptom assessment platforms included in such a designation.
Ada epitomizes the move by healthcare companies to deploy AI to achieve what the National Institutes of Health describes as the “four-fold purpose” of improving population health, improving the patient care experience, and enhancing the caregiving experience. is. According to his 2023 report from Accenture, a human-technology model is the answer, creating omnichannel connections with patients and reducing the burden on clinicians. It is estimated that 70% of healthcare worker tasks could be enhanced or automated by technology such as AI.
“In the medical field, AI goes far beyond research and data processing,” Novorol says. “We are seeing the transformative impact of AI in diagnostics, medical image interpretation, drug discovery, remote monitoring and robotics, and candidate identification. This is an important foundation for our health technology efforts.”
For Novorol, evolving the Ada product means testing new types of AI. They are currently experimenting with large-scale language models (LLM) and how it can complement Ada. She explained that the LLM is an exciting horizon in the health tech industry. Traditionally, LLM has had clear limitations that have made it difficult to adopt it in clinical practice. In other words, the LLM is a “black box system” and the decision-making process is opaque, Novorol said. But now there may be a way to leverage their approach in a responsible way.
“We work closely with clinicians in multiple health systems around the world, and we hear that physicians are very open to embracing the benefits of AI,” Novorol said. “However, transparency and explainability are especially important in clinical settings. They want to ensure the highest levels of safety and quality of care when implementing AI technology.”