The Whisk Health Score & Diabetes Friendly label
A simple score for any recipe based on USDA and NHS healthy eating definitions.
Whisk uses deep-learning based NLP to map extensive nutritional datasets from the USDA and EU sources onto our Food GenomeTM food ontology. This dataset allows us to understand the nutritional composition of millions of ingredients and recipes in detail and to use it to create smarter food experiences.
In partnership with the Academic Health and Science Network and in response to demand from NHS Digital, Ascensia Diabetes Care and Health Exchange to simplify the process of identifying healthy recipes we have built a Health Score that distills NHS and USDA definitions of healthy eating into a simple but sophisticated score out of 10.
Whisk also calculates the Glycemic Index and Glycemic Load values of recipes which we use in combination with the Health Score to tag healthy recipes as “Diabetes Friendly” to help people find recipes that are tailored to their needs.
Whisk Health Score
Whisk’s Health Score can be seen in action on the Whisk showcase site and is based on encouragement and limitation of 15 nutrients (macro and micro) per 100 kcal of recipe.
Our formula analyzes the overall amount of carbohydrates, fibres, sugars, fats, unsaturated and saturated fat, trans fat, proteins, sodium, potassium, calcium, iron, vitamin A, Vitamin C and cholesterol. It practically calculates nutrient density which identifies the proportion of nutrients in foods and scores range it from 2 to 10.
In practice, the Health Score is calculated based on nutritional values of recipes for macro and micronutrients per 100 kcal, where each of nutrient is multiplied with its own coefficient. Each nutrient has a coefficient which is a single unique number derived from nutrient ratios, investigation of papers, nutritionists advice and other sources. Coefficients are also limited by reference intake of nutrients (RDA).
Nutrients which are encouraged have positive values of coefficients, while nutrients which are limited have negative values of coefficients.
Whisk’s health score formula is considering that positive/negative impact from nutrients is not the same (e.g. having 5x Vitamin A in recipe doesn’t neutralise 5x Saturated Fat). The impact of nutrients is smoothed with logarithmic function.
Matching final score to the range 2-10 is accomplished with sigmoid function.
- Low health score (0.0-5.9)
- Medium health score (6.0-7.9)
- High health score (8.0-10.0)
Glycemic Index and Glycemic Load
Our Low GI and Low GL tags use a calculation to define the Glycemic Index (GI) and Glycemic Load (GL) of meal.
The GI is a relative ranking of carbohydrate in foods according to how they affect blood glucose levels. Carbohydrates with a low GI value (55 or less) are more slowly digested, absorbed and metabolised and cause a lower and slower rise in blood glucose and, therefore usually, insulin levels.
Classification of GI per individual food portion:
Low: 55 or less
Moderate: 56 – 69
Glycemic Load combines both the quantity and quality of carbohydrates. It is also the best way to compare blood glucose values of different types and amounts of foods. The formula for calculating the GL of a particular food or meal is:
GL = GI x Carbohydrate (g) content per portion ÷ 100.
For the whole day, a low GL diet has a GL less than 100 g/% for people consuming 8,700 kJ. Therefore, for people consuming 3 meals per day, a low GL meal would have a GL ≤ 33 g/%.
Whisk’s rules for GI and GL tags:
Low GI ≤ 55 per serving Low GL < 10 per serving
We tested our formula on various recipes with ‘diabetes friendly’ tags on different recipe sites such as:
We work with doctors, nutritionists and dieticians to validate the accuracy and credibility of our calculations.
"As a dietitian, I find the Health Score to be a meaningful way to designate healthy recipes. It is also a helpful way to highlight recipes that should be used with consideration, due to a less optimal nutrient profile. The Health Score is able to take into account nutrition variables and compile those into one easy-to-see score. Using this score can help consumers make more informed decisions about the foods they prepare."
Jennifer Flak, MS, RD, LD.