Healthcare Trends - Data Sharing
Articles published by MIT Technology Review, Stony Brook University, Healthcare Weekly, Accu-Chek, and International Business Times indicate that several organizations are using people's health-related data in the prediction of illnesses, diseases, or medical conditions. These organizations, however, are mostly universities and established corporate entities, not corporate startup teams or non-traditional health startups. Two startups that are using people's health-related data to predict illness are Medopad and AIME.
Sources Backing Up The Trend
1. MIT Technology Review
- An article published by MIT Technology Review in May 2019 indicates that researchers at the Nokia Bell Labs in Cambridge, United Kingdom, found that digital records of grocery bills can be utilized in predicting or identifying geographic areas with high prevalence of high blood sugar, high cholesterol, and high blood pressure.
- The classifier that the researchers were able to develop using grocery bills from Tesco in London has a 91% accuracy rate and the potential to be utilized as "a cheap and scalable tool for health surveillance."
2. Stony Brook University
- An article published by Stony Brook University in June 2019 indicates that researchers from Stony Brook University and Penn Medicine found that Facebook posts, with owner or patient consent, can be used in predicting 21 conditions and diseases, including hypertension, diabetes, depression, and anxiety.
- According to this article, there is a link between diseases and language patterns, and Facebook posts are more reliable than demographic information in predicting some diseases.
3. Healthcare Weekly
- An article published by Healthcare Weekly in March 2019 indicates that the University of Michigan and Apple have teamed up to launch the Michigan Predictive Activity and Clinical Trajectories (MIPACT), a study that looks into whether Apple Watch data, along with other health information, can be used to predict whether a person will develop a disease.
- A thousand individuals have so far enrolled to participate in the study, but it is expected that several thousands more will join in the following year.
- An article published by Accu-Check in January 2019 indicates that Roche and IBM Watson Health have developed a new predictive model that makes use of real-world data and that offers a more accurate way of predicting chronic kidney disease, a long-term, diabetes-related complication.
- The article highlights the increasing value of predictive analytics and real-world data in the area of diabetes care.
5. International Business Times
- An article published by the International Business Times in May 2019 indicates that for big data and machine learning to accurately predict diseases, there should be sufficient retrospective data and non-traditional health data such as syndromic surveillance data and social media data.
- The geotagging capture technology of social media enables the mapping of outbreak hotspots.
Startups Demonstrating The Trend
- Medopad, a London-based startup that has earned the support of pharmaceutical company Bayer and is on track to become a unicorn, has developed an app that compiles and analyzes health data from patient wearables, mobile devices, and medical bodies to predict chronic diseases.
- Leveraging both big data and machine learning, the startup aspires to "understand, treat, and ultimately prevent ill health."
- Medopad recently acquired rival Sherbit, a startup that also utilizes personal data collected from sensors, apps, and devices in uncovering health insights.
- AIME, which stands for Artificial Intelligence in Medical Epidemiology, is a startup based in the United States that leverages machine learning and big data analytics to predict in real time the time and location of infectious disease outbreaks. The end goal is to detect outbreaks in advance and keep them in check.
- AIME, the system, has a bot named REDINT that scours over 40 databases of weather, geographical, and epidemiological data.
- According to Dr. Helmi Zakariah of AIME, AIME's deployment throughout the healthcare sector is crucial to its success. The system will be ineffective if it does not have "a continuous stream of new disease incidence data" that it can use to continue learning.
While we were able to find two relevant startups, we were unable to find sources that strongly support the trend that non-traditional health startups or corporate startup teams are using people's health-related data in predicting illness. Several organizations are using real-world data, such as grocery bills, Facebook posts, and data from wearables, in predicting diseases, but most of them are universities and established corporate entities, not startups. This is what we have gathered from examining sources published in 2019.
To find the desired sources, we initially looked for articles or reports discussing the prediction of illnesses, diseases, outbreaks, and medical conditions. There are several sources that cover the topic, but we were able to narrow the number down by considering only those that mention or imply usage of real-world data, and that match the specified geographies. As previously mentioned, we came across only two relevant startups, Medopad and AIME. The rest of the sources we found involved the Nokia Bell Labs, the Stony Brook University and Penn Medicine, the University of Michigan and Apple, and Roche and IBM Watson Health. We decided to present the findings in these sources above, as they still support the trend of using people's health-related data in disease prediction.
Statistics supporting the specified trend are, unfortunately, not publicly available. We looked for surveys of startups employing big data analytics in healthcare, but was unable to find any. We looked for qualitative information as well by looking for expert opinions, but this approach proved ineffective too. The only additional sources we were able to find that we believe are helpful are articles indicating that people are becoming more willing to share their personal health-related data. Both a study published in BMC Medical Informatics and Decision Making and an article published by Modern Healthcare indicate this.