By Nate Worrell
In 2014, my wife had a serious health crash. She was stuck in bed battling gut pain, bloating, fatigue, anxiety, and depression. Her lab tests were “normal,” and the medications she tried had side effects worse than the struggles she faced.
The most effective remedy for her ailments ended up being a series of dietary adjustments. This sparked a health journey that went deep into the heart of nutrition sciences, and as we waded through the quagmire of information, I couldn’t help but think, “I’m glad I’m an actuary.”
In some ways, diet is easy. “Eat more vegetables” is a foolproof approach. Or to use an actuarial maxim, “Garbage in, garbage out.”
Yet the U.S. struggles with obesity. Heart disease and diabetes rank high among causes of non-accidental mortality and chronic illness. The most common tools to attack these conditions are drugs, surgeries, and diets. Drugs can get expensive and carry side effects. Surgeries are often costly and risky. As a contrast, several studies suggest that the right food choices can reverse certain diseases, with very little risk and almost no side effects. And the cost is effectively zero—we must eat, after all!
So if food is safe, effective, and inexpensive, why isn’t it our primary course of treatment? The answers to that question are contentious and convoluted, and not really the intent of this discussion. Instead, the question to ask is, “How can actuaries help?”
Actuaries and professionals who build careers around using insurance and financial instruments to manage and mitigate the financial hardships of disease and death have a lot to gain by considering the role of food in reducing risks of morbidity and mortality. Understanding what delays, reduces, or even eliminates chronic illness, or what can speed up recovery, has compounding effects. As experience improves and companies reduce loss ratios, products can become cheaper. Can actuaries help create a world where people feel better, work more, and get to keep more of what they earn?
To set the table for actuaries to inspire further discussion and innovation around these themes, this article will present a five-course meal, addressing various aspects of the dietary world in terms actuaries will be familiar with. Many of the ideas are meant to be a springboard for further thought, not necessarily a landing point.
- First Course “Data”—A spicy blend of nutrition studies complemented with a dash of dietary tracking tools.
- Second Course “First-Principles-Based Nutrition”—An intricate medley of biology, chemistry, genetics, and epigenetics.
- Third Course “Models and Measurement”—A rich and hearty exploration of outcomes and analysis.
- Fourth Course “External Factors”—An exotic mix of behavioral economics paired with lifestyle considerations.
- Fifth Course “Caveats and Cautions”—A tangy offering that underscores a respect for the complexity of the subject matter at hand.
Grab your fork, spoon, and spreadsheets—let’s dive in.
First Course: “Data”
Customer: “Waiter, there’s a fly in my soup!”
Waiter: “I’ll bring you a fork.”
Actuary: “I.T. Team, there’s a bug in my data!”
I.T. Team: “Only one?”
This starter plate has two dishes: nutrition science and food/consumption data.
Dietary studies run into the following challenges:
- Extrapolating from animal studies to humans
- Homogeneity of sample group
- Researcher funding conflicts of interest
- Credibility of sample sizes or study lengths
- Biases: survivorship bias, self-reporting bias, etc.
The diet battle arena tends to follow a certain pattern. A documentary or new book claims scientific proof behind its golden food approach. Then there will be a media headline citing scientific studies that discredit original finding. Then the internet explodes. The whole thing can seem confusing, contradictory, and even contentious. How does one avoid getting sucked into the food fight between Holistic Hippies and Big Food?
Actuarial Standard of Practice (ASOP) No. 23, Data Quality, offers key ideas:
- “Data selection” should consider appropriateness, whether it is sufficiently current, consistency, comprehensiveness, alternative sources, known limitations, and sampling techniques.
- “Data review” involves getting comfortable with the data elements, particularly any that have a significant impact on results.
- “Data use” acknowledges that data isn’t perfect and guides the actuary through the appropriate considerations and disclosures when dealing with imperfect information.
- Communication, reliance, and documentation standards complete the ASOP.
Actuaries love codifying things and creating algorithms. Imagine a process to evaluate data that comes from dietary experiments. Each study gets a score based on the following criteria:
To normalize, set the ideal study as a double-blind, randomized, controlled study of humans over the course of 10 years. The system weights the credibility of studies based on their proximity to this ideal experiment.
To illustrate further:
Assume there are 10 claims about the benefits of kale and 10 claims about the benefits of whole grains.
If the study outcome is positive for the food item under review, assign a value of 1 × study weight. If the study outcome is negative, then use –1 × study weight. Mixed results can get a score of 0. Ultimately, each item gets a bottom-line number.
Say kale gets a value of 0.75 and whole grains end up at 0.10. Both seem to be helpful, with scores greater than zero, but there is more confidence about kale.
With this sort of evaluation technique, we get some information from the studies that make it to the headlines, but we also know how strong that information is. It’s not a perfect system, but it’s better than assuming equal weight for every claim, and it can be agnostic about where the claim came from.
As a final note on food studies, most are done in a reductionist manner, trying to analyze some input X on an output Y. Different approaches—particularly observational studies—can help us understand what happens with food choices in the real world, where there’s a lot more going on than a collection of piecewise ingredients.
Part two of the appetizer is food and consumption data.
For insurers who want to take interest in what people eat, perhaps for underwriting or predictive modeling purposes, what sort of data exists, and how useful is it?
For years, dieticians and nutritionists used surveys and food journaling as a primary tools to understand what people ate. These tools are not perfect, because people aren’t always great at remembering things or might withhold information they don’t want to share. However, the process is getting easier and more comprehensive with technology. A photo log taken from your phone may help reduce the effects of poor memory, for example.
Secondly, your next meal will come with a side of Big Data.
Already restaurants are trying to predict what you will order, or will suggest to you what dish would align best with your patterns.
Mobile apps are beginning to use artificial intelligence to map nutritional information to photos of your food, and you can scan product barcodes to include ingredients and nutrition details. Machine learning algorithms may take these food images to the next level to use what you are eating to generate a risk score, or even to predict disease.
Additionally, there are companies developing ingestible devices that can evaluate your digestive juices and scan the intestinal wall as they pass through your system. Not sure about swallowing a mini-computer? Tufts University has developed a tooth-mounted sensor that offers the possibility of monitoring nutritional intake as you consume it.
Look at the back of your napkin. See any equations on it?
The actuarial world has its fair share of “rules of thumb,” approximations, and factors-based approaches. While helpful for quick checks or speedier calculations, they lack the required detail to really get after the underlying movement of a liability. Breaking calculations into their fundamental ingredients becomes necessary. Much like a chef constructing a recipe, actuaries need to understand how all the ingredients and their combinations affect the outcomes—a task that is much harder than it sounds.
For food and health, “first principles” comes down to following:
- Biology and chemistry of the foods themselves, and of the impact of these nutrients to the systems of human body
- Genetic and epigenetic influences, particularly to digestion
Chemistry and Biology
Sarah Ballantyne, PhD, is a scientific researcher with a special emphasis on how food can help heal autoimmune conditions. For her, the question is less about what you do or don’t you eat, but how food satisfies the body’s nutritional requirements.
“Food provides all the building blocks used to make every cell, tissue, organ and structure in our bodies,” Dr. Ballantyne says. “Food provides the raw materials for millions of chemical reactions inside our bodies at every moment. And food provides energy needed to sustain life.”
Studying food in this way leads to things like labeling Alzheimer’s disease as diabetes Type 3. Food consumption influences the insulin response. And an element of Alzheimer’s is poor insulin regulation. What was previously considered solely a neurological condition now has a strong dietary component.
Again and again, this proves to be the case. There are several heart-related conditions that are less about the heart itself and more directly related to foods we eat. It all comes back to the building blocks.
Categorizing food based on its nutritional content is nothing new. The U.S. government has been at it for years, and they have quite an extensive dataset. More recently, Joel Furnham, MD, author of Eat to Live, created a nutritional density index to identify foods that maximize nutrients per calorie. This simple ratio can be a great input into any dietary assessment framework.
While a nutrient-based orientation is helpful, it is far from complete. And as will be discussed later, just knowing what something is made of doesn’t mean you know exactly how the body will use it.
One of the hottest topics in food research involves the gut biome.
The intestine is not just a digestion tool; it is also a second nervous system and immunity engine. Inside our guts are trillions of bacteria. These microbes do many things.
One particularly fascinating role is that some of them consume brain chemicals like dopamine and serotonin, and others produce them.
These microbes also influence the body’s immune system, triggering inflammatory responses that manifest in things like rheumatoid arthritis, psychological disorders like schizophrenia, mood disorders, and various bowel issues.
Scientists are just beginning to understand and appreciate the role these little critters have in disease and aging. What is pretty widely accepted is that poor food choices can throw this system out of balance and can lead to an increased risk of certain diseases.
Companies like uBiome and the Human Longevity Institute are combining genetic sequencing of gut bacteria with machine learning tools to predict health outcomes. It may not be long before your bathroom features a smart toothbrush or even a smart toilet that runs diagnostic tests as you get ready for work.
Genetics and Epigenetics
Genetic analysis facilitates understanding how your unique body will respond to certain foods. Celiac disease and Type 1 diabetes are common examples of maladies with strong genetic components.
As a taste of what may be coming down the road, scientists are exploring the use of CRISPR, the DNA editing bacteria, to combat obesity. A recent study from the University of California used a variant that “turns up the volume” of genes instead of altering the genome. The experiment showed dramatic weight-loss results on mice when they amplified a particular gene.
Furthermore, epigenetics relates to how external factors affect the expression of genes. Things like stress levels, quality of sleep, and how much natural light we get end up affecting the expressions of genes. Food and nutrients have epigenetic effects on things like depression, inflammation, appetite, and more.
With countless chemical reactions, a universe of microbes, and a complicated DNA code, a complete understanding of the first principles of food and the human body may not be possible. However, knowing about these elements may still be useful.
How might an actuary use these building blocks? Think of a financial report that shows a loss of income due to adverse morbidity experience. Imagine a dietary report that attributes your fatigue to a lack of iron and other micronutrients. Detailed analysis reveals leafy greens quarter over quarter are down 25%. You take action to increase your kale consumption and give up donuts. In the next period, you just might see results improve!
Third Course: “Modeling and Measurement”
Now that we have examined the data and explored underlying mechanisms, it’s time to combine these ingredients to produce something especially delicious.
The third course in this discussion features concepts of dietary modeling that are actuarial comfort foods—measuring tools, efficient frontiers, target portfolios, and sensitivity analysis.
Multiple Measuring Tools
In a pricing exercise there are several questions to answer: What profit target are we aiming for—breakeven point, return on investment, or profit margin? Do we want to meet this objective in aggregate, allowing subsidization across the portfolio, or do we want to hit the target at each cell?
Similarly, in assessing the food and health model, what sort of outcome do we want to measure? Weight loss? Number of sick days? Occurrence of cancer? Longevity?
To be sure, eating nothing would yield fantastic weight loss results, but the strategy takes a terrible toll on mortality!
One way to look at trade-offs might be to take inspiration from the investment world and create some sort of efficient frontier. Instead of bonds and stocks, look at the concentrations of meat and veggies. A frontier can cover the dietary spectrum from vegan diet to carnivore diet.
Australian engineer Marty Kendall developed a Nutrition Optimizer, which operates under a similar concept. Above is a chart of various nutritional profiles, plotting nutrient density against insulin load.
Insulin is how the body manages energy storage. And foods with higher insulin loads put more strain on the body’s systems. Some nutrient-dense diets may also be energy-dense diets. The frontier shows where nutrient density peaks without straining insulin requirements.
Similar types of frontiers may be possible. For instance, plotting weight loss percentage against risk of cardiac disease. The latter would be an outcome of actuarial/nutritional models.
Ultimately, just as investors have different risk appetites, a diet plan can reference someone’s actual appetite.
In the investment world, risk appetites and portfolios change. The same could be true for diets.
Is there a need for rebalancing? What works best for a 20-something might not work as well at age 70. How many more prunes are in grandpa’s portfolio than in yours? As an illustration, here are some focus areas that an age-based nutrition portfolio might support.
After the 2008 market crash, funds started focusing on volatility instead of return. Weight-loss-oriented diets are tempting with their promises, yet many become “yo-yo” diets and the weight loss is temporary. Many dieters are afraid of “crashes.” More sustainable diets—the kinds you find in the Blue Zones or the “Whole Food Plant Based” approach—could be representative of a low-volatility diet portfolio.
As good as actuaries are, it is rare for profits to emerge exactly as priced. Maybe the business mix is different, perhaps the economic environment changed, maybe managerial actions shifted.
A diet model would be similarly evaluated.
How does a certain diet perform under different stresses? Increase fat 10%, decrease salt 5%? Different budgets? What happens if there’s a sugary binge due to an emotional event? Where are the strengths and weaknesses of the chosen approach? How do the dynamics vary by genetic makeups, age cohorts, and so on?
Fourth Course: “External Factors”
Suppose a chef prepares grilled sea bass over roasted veggies. Everything is cooked to perfection and seasoned exquisitely. But the server is rude, the food comes out late and cold, and you are eating while a construction crew is operating jackhammers and pouring asphalt. How much would you enjoy that meal?
The final part of the evaluation process requires a less quantitative set of actuarial tools. Some external risk factors and elements aren’t easily quantifiable. In these times, we reach for that strange ingredient we call actuarial judgment.
What are the things beyond food that make a food plan effective?
From Dan Buettner’s work with the Blue Zone research, we know that it is not diet alone that leads people to healthier living. Lifestyle factors play a role as well. Everything from quality of sleep to socialization to belief systems can amplify or restrict the effects of healthy eating.
At some point, these lifestyle items need to become contextual or even central items to the model.
From there, we have to dive in to psychology. Why do some people prefer meat over vegetables? Why aren’t we eating more bugs? Why do some diets succeed and others fail?
Game theory and behavioral economics reveal the challenges of delivering healthy dietary protocols effectively to the reluctant consumer. Insurance and retirement savings face challenges similar to nutrition:
- A “this won’t happen to me” mindset
- A preference for immediate consumption over delayed gratification
- Reluctance to give something up (loss aversion)
- Social norms that are more harmful than helpful
How could an actuary combat these things? Some ideas:
- Show some useful framing/anchoring stats on health care costs of preventable diseases. How much broccoli could you buy with that prescription money?
- Consider bundling or combining products that couple healthy foods with other goods.
- How about an auto-enrollment type of feature? What if your grocery bagger added extra greens to your order while you were checking out, courtesy of your insurance company?
- What about some nudges? What if annual statements came with coupons for fresh produce?
- To combat loss aversion, should we consider different terminology? The word “diet” itself has a bunch of connotations, implying someone is overweight and needs to become slender. What if instead, insurers presented options for “peak performance eating” or “leveling up your health” or perhaps “life extension plan.” The idea here is to focus on what is being added versus what is being removed.
It is important to remember that information is not enough. Whether selling insurance or advocating for a set of eating habits, we are dealing with people, and reason doesn’t always win. Intangibles like culture and emotional state can make a world of difference.
Fifth Course: Caveats and Cautions
Just as a nice beverage or sweet treat completes a meal, an actuarial article would not be complete without some caveats.
It’s possible that trying to model the food process is a fool’s errand.
Food and the human body are a lot more complicated than the sum of their parts. To quote T. Colin Campbell, PhD, and author of Whole and The China Study, “Nutrition is not a mathematical equation in which two plus two is four. The food we put in our mouths doesn’t control our nutrition—not entirely. What our bodies do with that food does.”
Foods themselves have inherent variability in their makeup. Furthermore, there is a different physiological experience between eating an orange and taking a vitamin C tablet. And more than that, each instance of eating an orange has a different metabolic response. The human body could be considered a bit of a “black box.” It is a system of systems. Patterns can emerge, but the causal links and path that led to a specific outcome ultimately may not be traceable.
Any dietary claim or effect of a specific nutrient or a specific gene needs to be weighed in the context of the whole system. Understanding each building block has value, but not at the expense of missing the big picture.
Considerations for randomness, complexity, and uncertainty are inherent components in any actuarial work. Corporate portfolios and assessments of solvency use scenarios and stochastic analysis. Flood maps or hurricane landfall predictions come with ranges of possibilities. Most predictive analytic methods quantify an element of “noise.” Similarly, any framework for dietary analysis has to have an appropriate acknowledgement of the complexity inherent in the process.
At the end of the day, the best model, if even possible, is still an approximation to a wonderfully intricate set of interactions.
Taking Home the Leftovers
U.S. News recently ranked 31 diet plans according to the opinions of a panel of doctors and nutritionists. In that effort there were mentions of difficulties with bias and ambiguity of terms. Furthermore, “best” ended up with several variations: easiest to follow, most effective for weight loss, etc. For example, the Mediterranean Diet ended up best overall but 17th in terms of weight loss.
What’s clear is that complexity and uncertainty are strong flavors in the diet soup.
It just so happens that actuaries have an acquired taste for uncertainty and noise.
Can we bring an objective and rigorous approach to the kitchen table to complement the work of dieticians and nutritionist? Is a framework for assessing diets, inspired by actuarial principles, feasible? Would it be useful?
How might such a framework work its way into insurance processes? Does the future include nutrition-based underwriting? Will food plans come about to help manage loss ratios and health care spending? Will retirement plans contain a nutrition protocol that aims to improve health to maximize retirement savings?
To conclude, let’s revisit a few key elements.
Bite the bytes: Food data can require more scrubbing than a bag of organic potatoes. Gathering data may be a challenge, but analyzing it in an unbiased way is the central value that actuaries can offer. For nutrition information sources, consider:
- Actuaries for Sustainable Healthcare;
- org from Dr. Greger (author of How Not to Die); and
- NHANES study done by the U.S. Census.
Chew on the models: A reliable tool will require health experts and food scientists and maybe a room full of computers. Granular analysis needs to be complemented with a big-picture perspective.
Check the lighting: A dietary model goes beyond just food—it’s also an emotional journey woven into the fabric of our identities and knitted with the consequences our lifestyle choices.
Perhaps as actuarial influence grows and becomes more reliable for assessing whether SlimFast beats Weight Watchers or if Jenny Craig is superior to Jared from Subway, we will see more Chief Actuaries become Chef Actuaries.
Until then, bon appétit!
Nate Worrell is an associate director at Moody’s Analytics.
Whole: Rethinking the Science of Nutrition; T. Colin Campbell; 2014.
Paleo Principles: The Science Behind the Paleo Template; Sarah Ballantyne; 2017.
How Not to Die: Discover the Foods Scientifically Proven to Prevent and Reverse Disease; Michael Greger; 2015.
Actuaries for Sustainable Health Care: www.actuariesforsustainablehealth
U.S. Nutrition Database: https://ndb.nal.usda.gov/ndb/
“U.S. News Best Diets: How We Rated 41 Eating Plans”; U.S. News and World Report; January 2, 2019.
“The Food Industry’s Influence In Nutrition Research”; NPR; September 17, 2016.
“Nutrition research is deeply biased by food companies. A new book explains why.”; Vox; November 11, 2018.
“How to optimise the insulin load of your diet for better diabetes management”; Optimising Nutrition.
Body Weight Planner: https://www.niddk.nih.gov/bwp
“Evaluation of Simulation Models that Estimate the Effect of Dietary Strategies on Nutritional Intake: A Systematic Review”; The Journal of Nutrition; May 2017.
It’s All in a Name: How to Boost the Sales of Plant-Based Menu Items; World Resource Institute’s Better Buying Lab; February 2019.
“CRISPR joins battle of the bulge, fights obesity without edits to genome”; Science; December 13, 2018.