Today, Friday 4 March, is World Obesity Day. To mark the occasion, we have taken a look at how the advancing fields of artificial intelligence and machine learning are influencing obesity prediction and intervention.
Wouldn’t it be great to be able to predict – and by extension prevent – obesity?
Obesity is a complex and multifactorial disease, whose onset can be affected by any combination of genetic, biological, environmental or socioeconomic factors1 – many of which feed into one another. Machine learning technologies are now showing great promise2 in both predicting future obesity and identifying a wide range of causative factors.
A particularly wide-ranging study,3 published in 2022, was able to integrate genetic and lifestyle data into machine learning programs assessing hundreds of variables, allowing the authors to deliver projections of future obesity. The scope and scale of the study enabled the researchers to evaluate the impact on future obesity of both genetic and dietary factors, as well as combinations of the two. Significantly, the study found that specific components of a patient’s diet appeared to have a greater overall effect on the probability of obesity than their broader dietary patterns. In addition, the researchers determined that the epigenetic process of DNA methylation (where the activity of DNA is altered without any change to the DNA code) appeared to be a more important factor in obesity than defined gene mutations such as single nucleotide polymorphisms.
Machine learning projection models can incorporate publicly available data on environmental, lifestyle4 and genetic profiles5 to assess patients’ risk of obesity and highlight key risk factors, by establishing relationships between critical data points without the need for a priori assumptions or interpretations that may be required with traditional data processing models.6
Machine learning yields more accurate predictions of obesity than statistical modelling not only in adults, but also in children and adolescents.7 In addition to offering greater precision in data modelling, machine learning programs are able to identify a wider range of variables that may affect the risk of obesity. These may range from pocket money and smartphone use to the feeding style and attitudes to food established by a child’s caregiver, to their academic performance or quality of sleep.
External factors are not limited to the influence of a child’s parents or teachers: one study implemented a machine learning process, drawing on primary care data held in electronic health records, to evaluate the levels of attention paid by clinicians to child patients with obesity and their impact on patient BMI.8 The algorithm used in the study showed that nearly half of patients aged between six and 12 with overweight or obesity did not receive any BMI-specific attention from clinicians at well-child visits – including further screening, follow-up appointments or advice on weight loss. The researchers noted that wider application of the algorithm could aid in identifying gaps in healthcare delivery and provide decision support for busy primary care physicians.
Are there other uses of AI which could aid in the fight against obesity?
Machine learning seems to show the most promise in predicting obesity before it begins. Conversely, AI algorithms have been successfully deployed to treat existing cases, whether by enhancing users’ capacity to self-monitor, implementing predictive analysis to devise personalised plans and goals, or using real-time analytics to drive behavioural interventions. Adolescents have reported positive results when engaging with an AI-powered ‘behavioural coaching’ chatbot,9 while conversational AI programs can be integrated into a patient’s weight loss regime to deliver personalised behavioural ‘nudges’, encouragement and reminders.10
Elsewhere, researchers have developed an AI-powered recipe recommendation programme which incorporates users’ dietary and time constraints, health needs and any food preferences or dislikes to deliver personalised results to search queries.11 On a much grander level, a US-based team has developed an AI program capable of assessing the level of obesity in a given region through satellite imagery by identifying key patterns in the built environment. These findings could then potentially be used to inform public policy and locally targeted interventions.12
Why is obesity important?
Adult obesity directly increases the risk to patients of many noncommunicable diseases,13 including cardiovascular and musculoskeletal disease, diabetes and some cancers. Obesity in children is a significant risk factor for continued obesity and potential disability in adulthood, but also carries a broader range of risks, from breathing difficulties and heightened risk of incurring fractures to hypertension, insulin resistance and even premature death.
In 1975, 3.2% of men and 6.4% of women worldwide were considered obese; by 2014, figures had risen to 10.8% of men and 14.9% of women. If obesity rates continue to increase at the same pace, it is projected that 18% of men and 21% of women will be obese by 2025.14 Childhood obesity increased substantially during the COVID-19 pandemic, with England’s National Child Measurement Programme reporting considerable increases in the number of obese children in Year 6 (ages 10 to 11) and Reception (ages 4 to 5) between the 2019–20 and 2020–21 school years.15
Given the rapidly increasing prevalence of obesity, and the risks posed to the health of adult and child patients, it is imperative that the medical community continue to embrace new and innovative techniques to combat obesity, its causes and its effects.
- Public Health England. Excess weight and COVID-19: insights from new evidence. 2020. Available at: https://bit.ly/3CclEsw [Accessed March 2022].
- DeGregory KW, Kuiper P, DeSilvio T, et al. A review of machine learning in obesity. Obes Rev 2018;19:668–685. https://doi.org/10.1111/obr.12667
- Lee Y-C, Christensen JJ, Parnell LD, et al. Using machine learning to predict obesity based on genome-wide and epigenome-wide gene–gene and gene–giet Interactions. Front Genet 2022;12:783845. https://doi.org/10.3389/fgene.2021.783845
- Thamrin SA, Arsyad DS, Kuswanto H, et al. Predicting obesity in adults using machine learning techniques: an analysis of Indonesian basic health research 2018. Front Nutr 2021;8:669155. https://doi.org/10.3389/fnut.2021.669155
- Curbelo Montañez CA, Fergus P, Hussain A, et al. Machine learning approaches for the prediction of obesity using publicly available genetic profiles. Liverpool John Moores University 2017. Available at: https://bit.ly/3HHKy4D [Accessed March 2022].
- Fu Y, Gou W, Hu W, et al. Integration of an interpretable machine learning algorithm to identify early life risk factors of childhood obesity among preterm infants: a prospective birth cohort. BMC Med 2020;18:184. https://doi.org/10.1186/s12916-020-01642-6
- Colmenarejo G. Machine learning models to predict childhood and adolescent obesity: a review. Nutrients 2020;12:2466. https://doi.org10.3390/nu12082466
- Turer CB, Skinner CS, Barlow SE. Algorithm to detect pediatric provider attention to high BMI and associated medical risk. J Am Med Inform Assoc. 2019;26(1):55–60. https://doi.org/10.1093/jamia/ocy126
- Stephens TN, Joerin A, Rauws M, Werk LN. Feasibility of pediatric obesity and prediabetes treatment support through Tess, the AI behavioral coaching chatbot. Transl Behav Med 2019;9(3):440–447. https://doi.org/10.1093/tbm/ibz043
- Grasso SV. Solving the obesity epidemic: is artificial intelligence the answer? Open Access Government 2020. Available at: https://bit.ly/3Movn3w [Accessed March 2022].
- Fadelli I. Researchers develop a system that can recommend personalized and healthy recipes. TechXplore 2021. Available at: https://bit.ly/3Kapx3S [Accessed March 2022].
- Maharana A, Nsoesie EO. Use of deep learning to examine the association of the built environment with prevalence of neighborhood adult obesity. JAMA Netw Open 2018;1(4):e181535. https://doi.org/10.1001/jamanetworkopen.2018.1535
- World Health Organization. Obesity and overweight. 2021. Available at: https://bit.ly/3IJSnb0 [Accessed March 2022].
- NCD Risk Factor Collaboration. Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19.2 million participants. Lancet 2016;387(10026):1377–1396. https://doi.org/10.1016/S0140-6736(16)30054-X
- NHS Digital. Significant increase in obesity rates among primary-aged children, latest statistics show. 2021. Available at: https://bit.ly/3pDJoka [Accessed March 2022].