Advantages of using AI in nutrition and health
Artificial intelligence in healthcare
The role of the importance of nutrition in health has increased over the last decade. This surge in consumer and public interest is a result of increased access to nutrition and health information as well as government initiatives which focus on disease prevention in an effort to reduce healthcare spending. While public health recommendations focused on a one-size all approach and messages to reach the masses, advances in technologies and innovations have increased the potential of delivering individualized recommendations. In this opinion piece, I will discuss how recent advances in technologies are creating a paradigm shift in how we engage, collect, research and influence end-user behaviour with regards to nutritional information as well as cover their inherent pitfalls.
The current situation in Personalised healthcare
Whilst traditional practice has resulted in a treasure trove of healthcare data locked inside electronic health records, paper records and individual practice preferences, new tools have the ability to combine various data sources from sensors, apps, wearables, diagnostic, metabolic profiles and even research databases to be able to initiate early intervention, track progress and tailor treatment using a personalised nutrition approach. The technologies that have allowed for this significant progress to be made include artificial intelligence, machine learning, deep learning and neural networks powered by advances in cloud-computing. As information can be processed in a matter of seconds, it means that end-users have the ability to access information on their responses to foods or the suitability thereof very quickly, or in real-time.
Nutrition technologies that have advanced rapidly, include; food scanners (D'Auria et al., 2019), image recognition for barcodes and nutritional analysis of foods, voice recognition, omics tests such as microbiome, DNA and metabolites, telehealth platforms, menu planning platforms, digital assistants to name a few (Hamideh et al., 2019).
The advantages of Artificial Intelligence in nutrition
The use of AI in nutrition is still relatively new with the most cited research in this area is that of personalised nutrition and glucose monitoring (Zeevi et al., 2015), but slowly there has been an increase in research. The advantage of using artificial intelligence in the area of nutrition is especially obvious in clinical practice, where user information can be collected and processed quickly through using surveys, online forms and scoring systems which traditionally requires manual data-entry. This could potentially free up time for the practitioners to spend on therapy and building trusting relationships with service users.
Secondly, by combining collected data with other data sources such as sensors, wearables and dietary intake apps, healthcare professionals can have a better understanding of a patient/client’s actual lifestyle and activity, rather than relying on self-reported information that can swallow up a lot of consultation time. Artificial intelligence has the ability to reveal hidden trends and patterns in a population group or within the individual that may not be obvious to the practitioners without spending time conducting research in data records that are not organized in a structured format. This means that the opportunity to bring patient care closer to the patients is greatly advanced.
The potential of AI in nutrition also includes usefulness as apps can be used as a motivational visual tool for users to understand and keep track of how they are performing based on their health and nutrition goals.
The current challenges of new Personalised nutrition technologies
Despite the excitement surrounding the use of new technologies in practice, these are often not without problems. The accuracy of wearables and trackers have already been questioned, whilst logging of dietary intake has an atrocious adherence rate (Case et al., 2015)
There is existing bias in the data sources that are used to train machine learning algorithms, and these biases include; research used, culture, ethnicity, racial, social economic status, religious groups need to be kept into consideration when developing recommendations (Chen, Szolovits and Ghassemi, 2019)
Previous research has indicated the healthcare professionals are involved in only 5% of mobile health apps developed (Research2guidance, 2015), hence the use of health professional expertise should always be questioned .
Digital nutrition products tend to fall in the grey area of lifestyle and wellness in terms of regulatory control and therefore raises the issue of ethics, transparency and explain- ability of complex algorithms that may not serve those who could benefit.
Other regulatory issues such as privacy, storage, link-ability to other data sources, consent are also of equal importance.
The future of the Personalised nutrition market
As the GDPR regulation is still relatively fresh, and policy-makers are making an effort to increase the focus on prevention, many issues still need to be addressed. One such area is the uptake of new technologies by nutrition professionals which remains slow (Abrahams et al., 2018). Nutrition experts play a key role as critics and validators of algorithms; recommenders of suitable technologies for end-users and translators of research into actionable messages. As advances in technology are changing both how nutrition services are delivered and accessed, without upskilling the workforce (including other professionals delivering nutrition advice such as Dr’s), the benefits may not be reaped by the public who currently trust healthcare professionals.
This means that nutrition organizations, educators, policy-makers and regulators need to make a concerted effort to ensure that digital and technological literacy is embedded in any initiatives.
Conclusions
We are in the midst of a digital technology revolution. It is clear that advances in technology and the power of cloud computing show no sign of slowing down and is driving the development of innovative new products, platforms and apps that have the ability to increase access to nutrition information, reduce healthcare costs and increase engagement. It is important that data used to develop AI algorithms are of high quality, use accurate unbiased information and are transparent. Any advances in technologies need to benefit society as a whole, they need to address important health inequalities questions in order to promote health and reduce the disease burden for all. More research is needed on the long-term benefits of using AI -driven platforms as well as the engagement of healthcare professionals with nutrition innovations.
References
Abrahams, M., Frewer, L., J., Bryant, E. and Stewart-Knox, B. (2018) 'Perceptions and experiences of early-adopting registered dietitians in integrating nutrigenomics into practice', British Food Journal, 120(4), pp. 763-776.
Case, M. A., Burwick, H. A., Volpp, K. G. and Patel, M. S. (2015) 'Accuracy of smartphone applications and wearable devices for tracking physical activity data', Jama, 313(6), pp. 625-6.
Chen, I. Y., Szolovits, P. and Ghassemi, M. (2019) 'Can AI Help Reduce Disparities in General Medical and Mental Health Care?', AMA J Ethics, 21(2), pp. E167-179.
D'Auria, E., Abrahams, M., Zuccotti, G. V. and Venter, C. (2019) 'Personalized Nutrition Approach in Food Allergy: Is It Prime Time Yet?', Nutrients, 11(2).
Hamideh, D., Arellano, B., Topol, E. J. and Steinhubl, S. R. (2019) 'Your digital nutritionist', Lancet, 393(10166), pp. 19.
Research2guidance (2015). mHealth App developer economics. Available at https://research2guidance.com/product/mhealth-developer-economics-2015 (Accessed 28th February 2019)
Zeevi, D., Korem, T., Zmora, N., Israeli, D., Rothschild, D., Weinberger, A., Ben-Yacov, O., Lador, D., Avnit-Sagi, T., Lotan-Pompan, M., Suez, J., Mahdi, J. A., Matot, E., Malka, G., Kosower, N., Rein, M., Zilberman-Schapira, G., Dohnalová, L., Pevsner-Fischer, M., Bikovsky, R., Halpern, Z., Elinav, E. and Segal, E. (2015) 'Personalized Nutrition by Prediction of Glycemic Responses', Cell, 163(5), pp. 1079-94.