The Better Assessment of Dietary Intake, Now

A recent article in Science, citing a recent study in Nature Food, contends that we need a new and better way to measure dietary intake- accurately, and scalably.? I agree, and there is one.? There is one that is apparently still just a bit too new for the accomplished reporter at Science to have heard of it.?

There is a way to assess dietary intake that is simple, sensible, scalable (infinitely*), and scientifically valid.? There is a way that is fast, easy, and - even - fun.? There is a way that can make diet quality (the single leading predictor of premature death from all causes and of chronic disease risk in the United States today, and in much of the modern world) a vital sign, accessible to us all the way our blood pressure numbers are.? There is a way that can make this crucial information a fixture in every medical record so that it serves as a cue to action, and an obligation, to every health care professional.? There is a way that can change the standard of medical practice- just as the blood pressure cuff did.

I know about the new method, while the reporter at Science did not (yet)- because I invented it (U.S. Patent 11,328,810 B2; U.S. Patent 12,073,935 B2; other patents pending) - with the help of some pretty high-powered friends.? So, disclosure: ?this is my baby we are talking about.? My hope is that making diet quality the vital sign it deserves to be, and now can be, will be my professional legacy.

There has long been awareness of the inaccuracies (and other, even more important limitations noted below) of the prevailing methods of dietary intake assessment based on recall, with expert opinion only differing on the severity and implications of such error.? Some have argued the methods are so inaccurate as to be useless, with which I disagree.? Others have argued they generate useful and interpretable information despite their limitations, and I am in this camp.

The new study, and the reporting on it, highlight one particular liability of recall-based methods (i.e., the widely used 24-hour recall surveys, and semi-quantitative food frequency questionnaires): the under-reporting of total energy intake.? This one error reverberates to many more, of course, because estimates of all food and nutrient intake levels are relative to the estimate of total calories consumed on average each day.?

The critique is based on comparing recall-based reporting of portions, which then add up to total daily calories, to an estimate based on a sophisticated equation derived from something called the “doubly labeled water” (DLW) technique.? The details of this method, which has some limitations of its own, are not important here.? We may simply note that by tracking the use of “labeled” oxygen in human metabolism, a fairly accurate estimate of energy (i.e., calorie) use can be derived.?

The doubly labeled water method itself, used in select research labs, is costly and cumbersome, and by no means the remedy to inaccuracies in recall surveys.? Rather, it just served to highlight those inaccuracies, and as a goad to innovation.? We need new and better methods, and as noted- we have one.

For starters, the new approach- diet quality photo navigation (DQPN; and no, taking photos is NOT required)- does not ask people to estimate their total food, and thus calorie, intake.? Rather, just like the recommendation in Science, it is based on a sophisticated equation estimating total energy needs.? The formula we use is derived from something called indirect calorimetry, rather than doubly labeled water, but the general idea is much the same.? In both cases, a robustly validated and quite reliable estimate of energy intake can be generated from simple, objective inputs.

The distinctions of the new method, brought to daylight by a company I founded, go well beyond how energy intake is determined.? There is general consensus that for 24-hour recalls (i.e., remember and report everything you ate yesterday, both what and how much) to reflect habitual dietary intake with acceptable accuracy, it takes at least three of them.? Each one takes roughly 20 minutes to complete, and the preference is that the three days not be consecutive.? In other words, just gathering these data takes an hour, spread out over multiple days, after which- those reports must first be analyzed by an expert (generally a registered dietitian nutritionist, or RDN) at the n-of-1 level to translate them into quantities of foods and nutrients, and ideally (but not always), an objective measure of overall diet quality.? Just imagine how routinely blood pressure would be measured if getting the numbers took over an hour of work by you, another hour by an expert conducting an analysis of your inputs, spread out over more than a week.?

Alternatively, the same information can be garnered from food frequency questionnaires.? These offer the benefit of capturing your input in just one sitting, but that sitting can take up to two hours (we used this method extensively in my research lab over the years).? This method requires you to remember just how much of exactly just what and exactly how often you ate over the past month, or months, or year.? The exercise is…well, let’s be kind and call it painstaking.? And here, too, the input must be subsequently analyzed to make sense of it, so again there is a delay of a week or even two between starting the effort and getting the answers.

The new approach, based on pattern recognition rather than recall, is routinely completed in as little as one minute (and very rarely takes more than three).? All answers- diet type, objectively measured diet quality (using the Healthy Eating Index 2020), approximate intake in servings of all food groups, and approximate intake of over 200 nutrients- are generated instantaneously.? These performance characteristics, adding up to “reliable answers in a minute,” are quite comparable to other vital signs- and nothing at all like other methods of dietary assessment.

How does this new method work?? In brief, we spent years building a “diet map,” populated by images that reflect a very wide range of real-world diets at different tiers of objective nutritional quality.? Your job is to look at two at a time (computer algorithms personalize this so you only look at images relevant to you) and say which looks more like things you typically eat.? This may remind you of the eye test where you have to say which of two images is more in focus, as well it should; we had that in mind.? We play this “game” until the best possible fit is achieved, and then- all the details of your diet can be reported back to you, and/or transmitted to your health care provider- immediately.

This new way is, of course, not perfect- and in fact, we have on our product roadmap some quite powerful ways of making it even better.? Why not deploy them now?? For all the same reasons the iPhone1 and the iPhone16 have noteworthy differences.?

Evolving a digital innovation into the optimal version of itself takes time, money, and lessons learned from real-world experience with each iteration.? We are delighted with how well DQPN works, and also eager to make it even better.? What’s most gratifying is that ways of evolving it are clear and accessible, because the approach is robust.? Humans have a native aptitude for pattern recognition, and DQPN capitalizes on that.? In contrast, humans have a native deficiency in the area of detailed recollection, so methods based on that weakness can never be made strong.

Still, the approach is not “perfect” now.? Neither is the measurement of blood pressure – and just look at how useful that is.? Blood pressure measurement at one time no more perfectly represents blood pressure all the time than a one-day recall reflects habitual diet.? Blood pressure can be affected by eating, drinking, activity, anxiety, stress, and time of day- to name only a few.? Blood pressure can be distorted by the size of an arm, the fit of a cuff, the methods of the clinician, and any stiffness in the brachial artery wall (a condition famously known as ‘Osler’s pseudo-hypertension’).? So, blood pressure readings are far from perfect- yet monumentally useful and important.? They reliably indicate dangerous extremes, trends, and relative values in a population.? They correlate strongly with health outcomes.

All of this is true of dietary intake assessment as well, even despite the inaccuracies of recall methods- which is why I am in the camp that thinks these data, though flawed, are still very useful.? Our pattern recognition approach correlates with recall methods just as we hoped it would: in a manner indicating it is correcting for known error.? The Diet ID assessment consistently estimates higher energy intake than recall methods, and tends to indicate a slightly lower overall diet quality.? It has performed well not only in comparison to the prevailing recall methods, but directly against biomarkers of cardiometabolic health and nutrient intake, too.

In addition to its objectively good performance, Diet ID offers tremendous advantages of efficiency, ease, and convenience.? Taking a minute to choose “this image, or that” is incomparably more user-friendly than logging everything you eat for a week; taking and uploading photos of everything you put in your mouth and then hoping software can sort out all the ingredients; or wracking your brain to remember just how many of Sandwich A on what bread and how much spread you had over the past six months- and then doing that for every other food item, too.??

Were a multidisciplinary group of experts- nutrition, medicine, biometrics, statistics, psychology, neuroscience, engineering- convened today and tasked with devising a reliable, efficient, scalable way to assess dietary intake and quality, we may be quite confident they would not land on methods requiring the recall of details.? Psychology research on the topic reveals that humans are notoriously deficient at capturing and remembering details with any consistency and accuracy.

The “legacy” methods of dietary assessment, however- those that have long been in use, and currently prevail- are predicated on just such fallibilities.? As if bounded by the limitations of evolutionary biology, the field has tinkered with these methods for decades, rather than replacing them.? Evolutionary biology- the most exceptional of all engineers- cannot back up and start over.? The result, along with such wonders as ourselves, and the manufacture of anteaters from ants, giraffes from leaves, Sequoias from soil and sunlight - is some stunning anatomical idiosyncrasies, such as the recurrent laryngeal nerve and our inside-out retinae (just go ahead and Google those; they are not the primary topic here).

Human engineers, though only capable of lesser wonders (thus far, at least)- have the advantage of Mulligans.? We can back up and start over when a better way presents itself. That does not mean, however, it is easy; the status quo tends to perpetuate itself.? There is, as well, the brute force of inertia; a great deal of effort and energy is required to radically redirect a large mass of object, work, or precedent.? In contrast, the perpetuation of forward motion by any large mass is easy- however ill-fated it may prove to be.

In such context, we may note in passing that when the blood pressure cuff was invented, the House of Medicine initially dismissed and rejected it, asserting that not having needed it yesterday, there could be no possible use for it today or tomorrow.? True innovation shakes up long-held convictions, however misguided, and must always climb such hills, and overcome such resistance.?

All of which is part of the reason this innovation in dietary assessment is not already more widely known.? That will change- but it takes time.

From my perspective, having spent years focused on the matter, the advance of dietary assessment based on pattern recognition can be, and should be, a matter of both trust and verification.? We may reasonably trust that basing the ascertainment of this vital information on something humans do natively, routinely, and well (i.e., pattern recognition) makes far more sense than it ever did to base it on something humans do uncomfortably, begrudgingly, and poorly (i.e., detailed recollection).? We may thank psychologists for the bountiful research on this topic, and Malcolm Gladwell for translating it into a very compelling read.

Stated bluntly, if tasked with building scalable dietary assessment de novo, we would be far more justified basing it on pattern recognition than recall.? We would never build the methods that have long prevailed.

Trust, supported by so much empiricism, is a good place to start- but not the place to stop.? A number of studies have been published by us, and others, to date.? The method is deployed into dozens of diverse, independent, university studies now.? We welcome collaborations that can put DQPN to the harsh test of the metabolic ward.

The DQPN method is not limited to assessing current dietary intake; the same approach can be used to identify an ideal, personalized “goal” diet. That combination- immediate access to the details of both where your diet is now, and where you would like it to be- confers all the same empowerment, convenience, and personalization as GPS.? Your baseline diet is your location now; that can be determined in as little as a minute.? Your goal diet is your destination, and that, too, can be determined in a minute, or less, based on the same selection among images of dietary patterns, and two key filters: (1) what health objectives would you like to address with diet?; and (2) of the dietary patterns responsive to those health objectives, which looks most appealing?? With both location and destination established in just a couple of minutes, the requisite changes to your diet to “get there from here” can be plotted, providing you a highly personalized journey, coaching (digital, and/or human), and in the food-as-medicine context, dietary prescription.? In this context, what used to be “behavior modification” evolves into “behavioral navigation,” empowering you to take your diet exactly where you want it to go.? And yes…our platform does all this, evolving the standard for food-as-medicine.? Of note and importance, our diet map is multicultural, meaning both on-boarding and goal-setting accommodate a wide and growing range of ethnic dietary patterns.?

The use of any map to locate yourself is only as “exactly right” as the map is richly detailed.? On a given map, you might locate your neighborhood.? On another, more detailed map- you might be able to find your room in your house on your street in your neighborhood.? Our diet map is already rich, but we are enriching it all the time.? That means ever more people with an ever broader range of dietary patterns will be able to arrive at a result that doesn’t just feel “about right,” but ever closer to “exactly right.”

Yes, I have a hammer and see nails.? But that’s because the nails were there all along, and someone needed to invent the hammer.? Diet quality is the single leading predictor of all-cause premature mortality and total chronic disease risk in the modern world.? If that does not warrant routine measurement and management, it would be hard to imagine what would.? If that nail doesn’t need to be driven…no nail does.? We needed a new hammer.

And, yes- I have skin in this game, but that’s because I put my skin at risk to do something I felt was pressing and important.? After nearly 30 years in academic medicine, I left to become an entrepreneur.? I founded a company, put in my life savings to launch it, and worked as its CEO without compensation for 7 years because I believe in the importance of making diet quality a vital sign.? It has transformational potential, no less than the blood pressure cuff.

I am not asking academic researchers to take my word for any of this, despite the robust body of publications we have already.? I am inviting them to study it for themselves.? Let it run the gauntlet of every relevant test.

And after a successful run there…let it elevate the standard of care, by making diet quality the vital sign it deserves to be.? We manage what we measure, and modern medicine cannot afford – in human or economic terms- to neglect the measurement and management of diet quality.? Time to embrace invention in the service of this necessity.

?

-fin

*In one deployment with our parent company, Tangelo, we completed over 10,000 comprehensive dietary intake assessments in a largely low-literacy, English-as-a-second-language population, in a span of under two weeks.? Completion rate approximated 100%.? To the best of our knowledge, this is the most efficient, population-level assessment of dietary intake in human history.

Dr. David L. Katz is a board-certified specialist in Preventive Medicine/Public Health, and past president of the American College of Lifestyle Medicine.??He is the founder of Diet ID, Inc, which developed and validated diet quality photo navigation.? He serves as Chief Medical Officer for Tangelo, a leading food-as-medicine company focused preferentially on the needs of underserved and food insecure populations.