InsideTracker’s mission is to help you add healthy years to your life using data and actionable changes. So what exactly goes into the recommendations that makes them so meaningful? In short, it’s the rigorous science review and the precision with which they’re tailored to each and every one of you. But because you’re like us, and strive to understand the “how” and “why” behind science and technology, let’s dive deeper. In this article, Dr. Renee Deehan, InsideTracker’s VP of Science and Artificial Intelligence sheds some light on the scientific process behind InsideTracker’s recommendations.
Take a peek behind the hood and let us show you how it all happens.
What is SegterraX?
SegterraX is the system behind the InsideTracker platform. It contains the knowledge from tens of thousands of scientific research papers reviewed by InsideTracker scientists and delivers recommendations based on the unique combination of your current habits, blood biomarker levels, DNA, and fitness tracker data.
The AI-based engine works based on rules created by InsideTracker scientists that dictate how recommendations are provided to users. These rules are based on analysis of over 2,500 peer-reviewed scientific publications (curated from over 35,000 relevant papers) as well as demographic data from over hundreds of thousands of healthy individuals.
InsideTracker doesn’t just tell you where your health needs attention—it also tells you how to improve. Based on your data, SegterraX provides both quantitative measures (and how to interpret them) and expert recommendations that studies have shown to work to improve suboptimal areas of health.
And SegterraX is continuously refined, building upon existing research with the latest happenings in the field of health and longevity.
How artificial intelligence is used to provide recommendations
InsideTracker uses two types of artificial intelligence (AI): machine learning and expert reasoning. Machine learning is probably what most people think of when it comes to AI. It utilizes algorithms to mimic the way that humans learn.
Expert reasoning is different. According to Dr. Deehan, “Expert reasoning is the type of AI that truly teaches a computer how to think like a human and in particular, a human with very specific domain expertise.” Think a registered dietitian, exercise physiologist, or aging researcher for example. This allows the engine to actually interpret papers the way an expert would.
The combination of these types of AI allow for both the knowledge accumulation and creation, and the delivery of tailored recommendations based on your health data.
InsideTracker's scientific review process: 100+ person-years of cumulative work
While InsideTracker’s platform relies on AI for efficiency, the expertise of the scientists deserves a major highlight. “AI allows these recommendations to be delivered to you in seconds, but in reality it took over a century in person-years to program the engine to do that,” says Dr. Deehan. And that doesn’t even include all of the time researchers spent in clinical trials, writing papers, and publishing those studies—or the time experts have spent obtaining their education.
Let’s walk through the scientific research and review process to understand what goes into creating an InsideTracker recommendation.
First, scientists work for weeks sifting through papers to identify interventions that have worked on people. While science conducted in non-human subjects (like mice or cells) serves its purpose in research, the science used and reviewed for InsideTracker recommendations are based on studies conducted on humans.
And while the Science Team conducts literature reviews examining papers that use populations with specific disease (like type 2 diabetes or cardiovascular disease), there’s a concerted effort to understand where interventions work in generally healthy people. This makes InsideTracker’s recommendations relevant for those seeking optimal health.
Step 2: Interpretation and scientific weighting
Scientists then weigh the strength of the evidence and decide if there’s enough to support a recommendation. Some questions considered may include:
- What populations were the studies conducted with? Perhaps a narrow group of just men or just women, or maybe individuals were aged 40-65. Sex, age ranges, ethnicities, and race are all factors considered when justifying the strength of evidence. These age ranges and other population characteristics help scientists determine who the potential recommendation is suited for, based on who is most likely to improve a measure of health using this intervention. For example, if balance training was an intervention found in the literature to be effective specifically in postmenopausal women over 65 who exercise once a week, you would only see this recommendation if you share these demographics and have indicated on your health profile that you exercise at this frequency.
- What was the population size? The size of the group studied can show whether the intervention was effective for many people or only for a relatively small selection. Studies conducted in a small group of individuals may need to be validated with other studies to increase confidence in the results.
- Is there conflicting evidence? While a number of studies may support a given intervention for a particular group of people, the Science Team also considers any evidence that opposes the effectiveness and weighs this to determine whether there is substantial scientific basis for a recommendation.
Step 3: Present findings for expert consensus
Once the research and interpretation has been conducted, findings are presented to a larger group of scientists to determine whether the intervention reviewed has sufficient scientific backing to become a recommendation and which people the recommendation is supported for. Once consensus has been reached, the recommendation is added to the engine.
What isn't presented is just as important as what is
Hundreds of recommendations have met InsideTracker’s strict criteria, but much of the research isn’t substantial enough to warrant a place in the SegterraX engine. And this is actually a good thing. “Our job here is to help you cut through the clutter and really identify the best actions for you to take in order to increase your healthspan,” says Dr. Deehan.
But if the research does not currently support a recommendation, the work spent reviewing the topic doesn’t go to waste. Instead, the Science Team continues to monitor the research on an ongoing basis. Here are some examples:
- Resveratrol supplementation: Research has shown that consistent supplementation with resveratrol can reduce fasting blood glucose by up to 35 mg/dL—but only in those with high levels at baseline. So resveratrol supplements would only be recommended by InsideTracker for people with high glucose (above 125mg/dL) or high HbA1c (above 7.5mg/dL) at baseline.
- NMN supplementation: While there is some promising research to support supplementation in humans, the evidence is not yet substantial enough to justify a recommendation. InsideTracker scientists will continue to monitor new studies for NMN in humans, evaluating and adding research to the engine.
How are recommendations personalized?
At InsideTracker, the core of who we are is science. But also core to who we are—personalization. That’s because no two bodies are the same, and data can tell us how our own body is responding to an intervention.
There are a few notable ways that InsideTracker makes recommendations personalized to you:
- Your data are the basis for your recommendations. This includes blood biomarkers, genetics, data from wearable devices, and answers to your health profile questionnaire.
- Rules are created based on the scientific analysis that instruct SegterraX when to present these recommendations. Rules may include criteria based on sex, age, race, ethnicity, menopausal status, biomarker level, nutritional preferences, and more.
- Impact scores tell you what’s the biggest bang for your buck. For example, a recommendation to incorporate more olive oil may improve your high LDL and total cholesterol, but it also could improve your fasting blood glucose and hsCRP (a marker of general inflammation). Impact scores help set priority for the most impactful recommendations to add to your Action Plan for efficient and precise ways to optimize your health
The health and longevity space has grown substantially over the past decade, providing so many options that you may have difficulty choosing the right tool to help you live healthier. This product review discusses how InsideTracker can help, but here’s a summary of what makes SegterraX unique from other technology platforms:
- Data asset deck: With over 35,000 peer-review papers reviewed and a database with points on hundreds of thousands of healthy people, the data that InsideTracker is able to harness to provide recommendations is substantial.
- Knowledge curation: It takes decades to build the knowledge supporting InsideTracker recommendations and for the expert reasoning AI to mature.
- Experts shape the tech: While AI is a powerful tool that makes it possible to provide expert knowledge almost instantly, it’s the hours (and years) of research and review that make providing these targeted, scientifically-supported recommendations possible.