Improving Health Behaviours in the Digital Era

In this week’s essay, independent contributor Edric Ramirez-Valdez of the University of Cambridge’s School of Clinical Medicine delves into another aspect of healthcare in the digital era —specifically, how the data and analytics revolution can help tackle the global (and partially preventable) burden of non-communicable diseases.

As the discovery of cures for infectious and acute conditions increases and their impact on health worldwide wanes, a tidal wave of non-communicable diseases (NCDs) has emerged — leading to NCDs becoming the largest cause of premature death globally. To give an idea of scale, the WHO has reported that 41 million people die (70% of yearly deaths) as a result of heart attacks, stroke, chronic respiratory diseases, diabetes, cancers and mental disorders. Common to these conditions are a well-established array of ‘behavioural determinants’ — factors that can be reduced or eliminated through behavioural or lifestyle changes — which radically impact one’s risk of developing them. These behavioural determinants are what we might consider the “usual suspects” of developed lifestyles in 2021: quality of diet, exercise, alcohol, smoking, and sleep.

Beyond the obvious damaging effects of NCDs on lifespan, wellbeing, and quality of life for patients and their loved ones, there are also tremendous associated resources being funneled towards addressing them. In the United States, RAND and the CDC report that 90% of the $3.5 trillion budget in annual health expenditure goes towards those with chronic and mental health conditions, whilst in the UK the BMA estimates that 40% of the uptake of health services in England could be preventable through behavioural changes in society.

Given the unprecedented investment and innovation in digital health, what role does the data and analytics revolution have in helping tackle this enormous, likely preventable burden of NCDs? Evidence suggests that providing patients with information is a key component in changing behaviours and improving long-term health outcomes, although that alone is not a complete strategy for improving population-level health behaviours. Understanding the limitations of patient education and possible opportunities to improve behaviours requires exploring these questions from an interdisciplinary perspective, drawing on insights we now have from behavioural science, public health, and digital medicine.

Behaviour is greatly influenced by an impulsive and automatic system

While many people know the types of decisions they could make to improve their general health (like eating more fruits and vegetables, engaging in physical exercise every day, etc.), most of us can’t resist the unhealthy option when faced with a choice, despite the best of goals. This gap between our intentions and behaviour is a documented phenomenon with an entire dedicated corpus of research.

For example, one psychological framework proposed by Strack & Deutsch and popularised by Daniel Kahneman in his 2011 book, Thinking Fast & Slow, suggests that there are two systems — consisting of (A) reflective and (B) impulsive components — which, when combined, result in an individual’s behaviour. These two components operate through different mechanisms. The (B) impulsive component is fast, emotional and unconscious, relying on little cognitive resources to work. In contrast, the (A) reflective component is rational, conscious and considered. If Strack & Deustch’s framework is true, it has profound implications for the effectiveness of how and why we provide rational information as a way to change individual behaviours. Providing rational information would primarily appeal to an individual’s reflective system, yet this work suggests a major facet of decision-making lies in our subconscious.

Indeed, academic studies have repeatedly shown that decision making can be significantly influenced by appeals to the impulsive system. Two examples of this are particularly interesting in the context of the data revolution: the first focuses on the importance of environmental cues to our impulsive system. A clear demonstration of this is in the relationship between food proximity and snacking behaviour, which shows that in general population samples, how physically close food is to a person is correlated with the likelihood of that person consuming it — an effect that is unlikely to be moderated by self-control.

A second example highlights our natural tendency towards inertia, in that individuals tend to show a bias towards the status quo. A neat demonstration of this phenomena comes from Davidai, Gilovich and Ross in 2012, who found substantially higher rates of organ donation from patients when they faced an opt-out, rather than opt-in, system — frequently exceeding 90% in opt-out countries and remaining less than 15% in most opt-in countries. Environmental cues and tendencies towards inertia are just two examples out of many underscoring the importance of the automatic component of human behaviour in making decisions. With these findings in mind, it seems clear that for interventions to be effective, health policy, technology, and practice must be designed to interact with the impulsive component of our psyche.

Noise in our digital age dilutes the impact of providing information

As anyone who has experienced the waves of national lockdowns accompanying the Covid-19 pandemic knows, engagement with social and digital media has never been greater. With this exponential growth in time spent on social media, paralleled by growth in publications and digital content, competition for attention is at an all time high. This race has serious implications in the context of influencing health behaviours.

First is the reality of information overload and social media fatigue. One fitting illustration of this has been throughout the Covid-19 pandemic, during which there has been an unprecedented proliferation of information coupled with seemingly capricious public health guidance. A December 2020 Nature Analysis reported that as Covid-19 cases grew exponentially, so too did the body of coronavirus-related peer-reviewed publications and preprints — with one estimation indicating over 200,000 articles had been published in the first twelve months since the Sars-CoV-2 virus was defined. Daily changes in information coupled with increasing misinformation have had understandably confusing and detrimental effects, both on variability in behaviours and individuals’ mental stamina.

Beyond the pandemic-specific overload is also the realm of what media and digital content people choose to devote their attention to. A 2018 Pew Research study found that over 80% of parents with a child younger than 12 years old allow their child on YouTube, with 35% reporting regular YouTube engagement. A 2014 study in the American Journal of Preventative Medicine, meanwhile, highlighted FTC findings that food and beverage companies were spending $1.8 billion per year on advertising their products to young people in 2009. Since then, many of these companies have turned to social media advertising as a way to take advantage of influencers’ rapid growth in youth engagement.

Tying these trends together, an NYU study from November 2020 investigated the extent to which children are being exposed to unhealthy food advertisements on child influencer YouTube videos. The researchers identified the ‘top kid influencers’ from 2019 and analysed over 400 of their most popular videos, which collectively garnered over 1 billion views. They found that over a third of these videos included advertising for unhealthy or fast-food products. Targeted advertising to young children in this way has the potential to greatly impact individual and family consumer choices during a particularly malleable time in these children’s lives. Public health measures centered around providing important behavioural health information, as well as policy legislation, need to keep up. Legislators and public health officials must learn to adapt their messages to these types of digital channels, given their pervasiveness and impact in today’s society.

Physician engagement and social support are bottlenecks to successful behaviour changes

In 2009, the English National Health Service (NHS) introduced the NHS Health Checks programme as a preventative risk measure against a surge of cardiovascular disease predicted in England. A recent 2020 systematic review of the programme explored those factors found most critical for influencing patients’ behaviour. The most prevalent factor identified — and a factor also reported in other studies — was the importance of having a supportive physician who understands the patient’s lifestyle and finds the physical time and space to deliver a check-up.

Given epidemic levels of physician burnout (as recently discussed in the NEJM and Lancet), it is clear that over-demand on physicians not only contributes to worse patient experience, but also to worse health outcomes. It is critical we craft health policy alongside technologies that allow physicians to meaningfully work together with their patients towards improved health behaviours.

Importantly, the NHS Health Checks review also highlighted a range of social support and reinforcement activities (e.g. getting lunch with a friend) as critical determinants to successful behaviour changes. These findings are not surprising. Promoting health behaviour changes through social support has already been explored in Preventative Medicine, where researchers have demonstrated that intergenerational transmission of obesity (i.e., when obese parents pass along behaviours that contribute to obesity to their children) decreases with the presence of social support.

Do digital technologies have the potential to overcome these barriers?

In the face of the multifaceted challenge that is improving health behaviours, there is still plenty to be optimistic about — arguably in large part due to innovations coming from incumbents and hundreds of new healthcare AI startups developing targeted solutions to pain points along the patient journey (especially regarding human tendencies towards impulsive and automatic behaviours). Beyond these innovations is also a growing literature studying them, paving the way for iterative learning that informs health policy and demonstrates the potential efficacy of these new types of digital public health interventions. In the section below, I describe a few of these exciting new innovations:

Wearables tackling the intention-behaviour gap

Knowing that environmental cues are critical for encouraging behaviours and engaging with our automatic tendencies, health insurers and pharmacies in the USA, UK and South Africa have been increasingly leveraging wearable technologies such as Apple Watches and Fitbits to provide members with “nudges” to engage with their health. These ‘nudges’ range from gamifying elements of healthy lifestyles, such as personalising activity goals and offering rewards (e.g. Amazon vouchers), to providing reminders to refill prescriptions and for patients to see their primary care physician. Encouragingly, Rand Europe found that Vitality members on an Apple Watch programme were active for four additional days per month compared to non-users, while United Healthcare claimed to save $220 in annual medical costs per member on their wearables programme.

Whilst this move by insurance companies to provide their customers with wearables for enhancing health behaviours seems to be a win for all parties involved, it has been met with skepticism by some writers. With the huge expansion in demand from big tech for healthcare data, users are understandably cautious when being offered wearables by private companies that collect their personal information. Coupled with the reality that tens of millions of healthcare records are being compromised annually by data breaches in the US alone, security and privacy concerns are paramount when weighing up the pros and cons of wearables provided by insurers. Transparency, consistency and performance surrounding data storage and usage will therefore be essential for the continued growth and success of such schemes.

Beating advertisers at their own game

Evidence is building that digital health advertising interventions in the crowded internet marketing arena could lead to large scale population health improvements. A 2018 randomised controlled trial on almost 794,000 participants was the first to empirically demonstrate the possible impact of health-promoting advertisements on the internet. The researchers presented randomly selected individuals using the Bing search engine with professionally developed health-promoting campaign advertisements, and controls with unchanged “status quo” advertisements. The study found a 50% increase in individuals searching for health-promoting and weight loss keywords after being shown the campaign when compared to the control group. Critically, the authors also commented on the potential of precision public health — advertising based on sociodemographic data and lifestyle preferences to more effectively influence large-scale population health behaviours.

With clear evidence that public health interventions in the digital arena are capable of pronounced population-level benefits, combined with the reality that the private sector is able to outspend public health bodies when it comes to advertising (e.g. junk food advertising spend in the UK is 30 times the amount the government spends on promoting healthy eating), the question then becomes how public health organisations can best campaign and legislate to compete in the digital marketing space. Firstly, public health bodies should leverage the latest advances in content communication techniques — including search engine optimisation, algorithmic targeting and A/B testing, and content curation to maximise virality (e.g. creating emotive, interactive and positive short videos which are likely to be shared online). It would be critical that the deployment of such techniques is compliant with data protection legislation (e.g. GDPR and CCPA) and transparent in order to reduce the likelihood of public backlash.

Secondly, it is important to establish and maintain health partnerships with major tech companies that have a large digital footprint. One example of this is highlighted in the 2017–2020 Public Health England Marketing Strategy, in which the agency detailed co-creating an online health hub with Amazon that was promoted across Amazon’s channels. Finally, legislation must continuously be revised to target the development of NCDs, reflecting learnings from previous policies and the literature. In July 2020, the UK Department of Health & Social Care published a policy paper on “tackling obesity”, which included a plan to ban High Fat, Sugar or Salt (HFSS) products being advertised on TV and online before 9pm by 2022. Their methods are yet to be outlined for executing the online ban but these types of policies will have serious ramifications for the advertising of products that contribute to NCD incidence.

Improving the physician-patient interaction and generating social support

With the burden of updating Electronic Medical Records (EMR) constituting a major part of physician burnout, there is demand for technology designed to reduce this burden and free up time for physicians to instead make meaningful, supportive connections with their patients. As Natural Language Processing (NLP) techniques improve at unprecedented rates, one promising AI solution to address physician burnout rests in Ambient Clinical Intelligence (ACI). The idea behind ACI is to leverage NLP agents capable of “listening” to the audio of a physician-patient interaction during a consultation, and automatically generate an EMR for that consultation for the doctor to review. This AI assistance could radically reduce the time physicians spend on administrative tasks, allowing them instead to focus on their patients (which may also lead to decreased physician burnout as well as improved job satisfaction). There are several ongoing efforts to develop and pilot this type of solution, including the 2019 Nuance-Microsoft partnership seeking to accelerate ACI rollout to hundreds of thousands of physicians worldwide.

Beyond these examples, it is worth noting that the ever-increasing accessibility of online forums and virtual groups presents additional opportunities for individuals to engage in digital smart health communities. These virtual groups can foster encouragement and social support centered around proactive health and wellbeing management like more established organizations, such as WW International, have done for years. As some may remember from their meteoric rise during the 2020 Covid-19 pandemic, new community-based health and fitness tech like Peloton are also creating real world-digital hybrid communities — and if start-ups Mirror or Tempo are anything to go by, this is a trend we should expect to see growing in upcoming years.

Looking Ahead

As the examples here show, the technological landscape for improving health behaviour changes is clearly an exciting — and promising — one. Even today, a rapidly evolving technological ecosystem sensitive to both medical and human needs is creating paradigm-shifting applications already being trialled in communities around the world. What makes this shift particularly interesting is that it engages with a spectrum of pain points around non-communicable diseases — from gamifying health behaviours, through digital marketing, and even assisting physician productivity.

Amidst this excitement, however, we must remember to pay special attention to how inequalities in health outcomes by socioeconomic status may be exacerbated by striated access to technological solutions, as well as the security and privacy concerns surrounding the collection, usage and distribution of our personal data.

TL;DR

• Non-communicable diseases which have major behavioural components (e.g. tobacco use, physical inactivity, the harmful use of alcohol and unhealthy diets) are responsible for 41 million yearly deaths (70%) worldwide.
• Encouraging healthier behaviours is complicated by human behavioural psychology and an increasingly complex digital landscape that is exploited by unhealthy brands.
• Tech, data, and AI interventions from the public and private sector are gaining momentum in meeting these challenges — including wearables “nudging” healthy lifestyles, precision public health, and enhanced social support from physicians and digital communities.
• Data security and privacy concerns remain top-of-mind when evaluating the trajectory of these technological developments.

About the Author: Edric Ramirez-Valdez

Edric Ramirez-Valdez is a graduate student reading Medicine at the University of Cambridge’s School of Clinical Medicine. He is passionate about the adoption of technological developments for the betterment of human health in the clinic and general public. Before coming to Cambridge, Edric worked as a pharmaceutical consulting analyst at IQVIA and a consultant at QuantumBlack, the Machine Learning and AI centre of excellence in McKinsey & Company. He holds a first class BSc in Biological Sciences from Imperial College London and is an independent contributor for GDI.