StuffThatWorks, a startup leveraging AI, crowdsourcing, and machine learning to match patients with treatments, today closed a $9 million seed round. CEO Yael Elish says the proceeds will be used to accelerate go-to-market efforts as the company’s platform experiences pandemic-driven growth.
In response to the worsening global health crisis, patients and providers have sought out digital health and medical solutions to problems induced by COVID-19. In regions under lockdown, remote visits are now one of the key ways for patients to connect with specialists. Moreover, data and health platforms that pair providers with support data analytics have become critical for information-sharing, research, and analysis.
StuffThatWorks was cofounded by Elish, a founding member and former head of product at Google-owned Waze; CTO Ron Held, previously on the Israel Defense Force’s intelligence team; and chief data scientist Yossi Synett. Elish spent years helping a family member cope with a medical condition that also took a toll on her family’s life. After months of online research, he discovered a medical treatment that was more effective than what she’d been prescribed.
This experience inspired Elish to launch StuffThatWorks, a portal that taps AI to enable patients to share personal treatments and discover options that work best for them. The platform’s algorithms transform members’ data about chronic, rare, and “orphan” diseases, including diabetes and ADHD, into personalized treatment effectiveness insights. The idea is that the more people who contribute to StuffThatWorks, the greater “authority” it can accrue.
An estimated 7% of Google’s daily searches are health-related, and the Pew Research Center says over 70% of people in the U.S. look up their medical symptoms on Google. But the resources they might draw answers from run the gamut from unvetted to jargony, and the consequences of misinformation and misinterpretations can be disastrous. A 2017 study found that when cancer patients turn to therapies advocated by less-reputable online sources, like supplements in place of conventional therapies, they are 2.5 times more likely to die.
As might be expected, people with trouble accessing health care services are more likely to seek out — and follow — questionable online advice. A 2015 article in the Journal of Medical Internet Research suggests that improving the accuracy and reliability of publicly available internet resources could “substantially” bolster health outcomes.
That’s the idea behind StuffThatWorks, where 180,000 contributors have volunteered over 10 million data points and 150,000 “experience-based” pages across 110 different communities. Elish asserts that every condition has hundreds of possible treatments but that the effectiveness is impacted by factors like age, gender, comorbidities, and even climate and lifestyle. Because of this, he says, producing statistically viable information about which treatments work best requires collecting data from millions of people worldwide on an ongoing basis — a project that’s too costly for any one organization or institution.
StuffThatWorks encourages people in “condition communities” to share their anonymized perspectives about a condition via brief surveys. After the platform collects 100 answers about treatments, their effectiveness, and side effects, initial insights about a condition become visible to the community, including the age of onset, symptoms, and aggravating factors. At several hundred contributors, StuffThatWorks’ predictive algorithms kick in, ranking treatments by effectiveness. Once thousands of people contribute, the platform anticipates effective treatments for both subgroups and individuals.
StuffThatWorks employs natural language processing to analyze responses and generate insights about symptoms, comorbidities, treatments, and more. Its models look at the attributes of the treatment in question (e.g., the type of treatment, chemical compound, and pharmaceutical classes), as well as the attributes of the person who reported that treatment. This accounts for individual biases toward certain types of treatments, as well as biases that arise from the demographic distributions. As more data is collected, the models become more personalized, and they’re able to better account for such biases.
The platform incentivizes participation with a points-based system — members earn recognition for posting and responding to questions, joining discussions, rating insights, and contributing data. The company hopes people will eventually be able to use points to inform decisions about what medical research should be pursued, Elish said, although those plans remain in the preliminary stages.
Guaranteeing members’ contributions remain “scientifically rigorous” is the responsibility of StuffThatWorks’ six-person medical and advisory board, which includes graduates of Stanford Medical School and Cornell University. A team of data scientists performs regular checks for validity and accuracy, Elish says, and they work to ensure users retain control over all of the data they share.
StuffThatWorks says it’s fully compliant with “all applicable privacy laws,” including the European Data Protection Regulations (GDPR) and the California Consumer Privacy Act (CCPA). The company retains the services of a data protection officer responsible for keeping up with evolving privacy standards and practices. All processing and analysis is performed on a separate database made up of anonymized, normalized, and aggregated data that doesn’t include any personally identifying information. Access to personal health information is restricted to key StuffThatWorks personnel for the purpose of maintaining database integrity, with login access tracked to prevent abuse.
StuffThatWorks also claims it will never sell personal information to any third-party companies or institutions. (In compliance with the CCPA, there’s a toggle within the platform’s dashboard where users can request their data never be sold, which is set to “on” by default.) But the company says it’s collaborating with a “limited number” of researchers, medical organizations, and patient advocacy groups on research.
“People are the ones that hold in-depth knowledge about themselves, their condition, and the way treatments affect them,” Elish said. “Collecting this knowledge in an organized and structured way is the only way to compare effectiveness in scale. And when that’s done across all chronic conditions it creates a gold mine of data that can dramatically advance and facilitate research to the benefit of both the patient and the medical community.”
Bessemer Venture Partners, 83North, and Ofek Ventures contributed to the seed investment in StuffThatWorks. In the near future, the two-year-old Tel Aviv-based company plans to hire engineers to lay the groundwork for growth.