medicine and pills with blue bg

Nudges Aren’t One-Size-Fits-All. Here’s How Personalizing Them Could Change Health Care

In the last decade, health care has been moving away from a traditional one-size-fits-all approach and towards personalized medicine. While the concept of precision medicine usually brings to mind things like genetic profiling for targeted therapeutic interventions (for example, tailored drug dosages), it can also be applied to facilitating more health-promoting behaviors among patients.1

In a recent article for Harvard Business Review, Shirley Chen and Mitesh Patel discuss how the rise in digital health tools such as mobile health apps, wearable technologies, and other remote patient-monitoring tools offers new channels for personalized care.2 Such tools provide the opportunity to collect a plethora of patient data, which can, in turn, provide insight into health behaviors, preferences, personality traits, motivations, barriers, area of residence, etc.

A lot of these data points are not exactly “clinical,” but can still complement the patient’s medical history. This allows for the creation of what Chen and Patel call “behavioral phenotypes,” which can then pave the way for personalized nudging in health care.

What are personalized nudges?

Personalized nudging is nudging that is optimized based on a patient’s behavioral phenotype, to maximize the effectiveness of the intervention. Personalized nudging can take two forms: choice personalization and delivery personalization.3 Choice personalization concerns the options presented to the patient, and delivery personalization refers to the method of nudging. Patient-centered nudging is powerful because nudges don’t work the same for everyone, and what works for a particular patient may not be effective for another. 

Nudging in health care can be applied towards improving medication adherence, physical activity, program engagement, and can even mental health. Artificial intelligence (AI) and machine learning algorithms allow for nudging interventions to be personalized and delivered at scale by automatically stratifying patients, incrementally learning from patient activity, and delivering personalized messages and recommendations. In essence, AI can learn and predict the most effective nudges for a particular patient and deliver them accordingly.

Here are some digital health startups that have already leveraged this technology to overcome challenges associated with patient behavior modification. 

Behavioral Science, Democratized

We make 35,000 decisions each day, often in environments that aren’t conducive to making sound choices. 

At TDL, we work with organizations in the public and private sectors—from new startups, to governments, to established players like the Gates Foundation—to debias decision-making and create better outcomes for everyone.

More about our services

Wellth

Wellth is tackling the problem of medication and care plan non-adherence using behavioral economics and AI. Wellth’s approach is grounded in principles of behavioral economics such as loss aversion and the endowment effect. Patients that are a part of their program are given monetary credit at the beginning of each month, and are asked to check in daily to track their health and medication intake. If a patient misses a check-in, they lose $2 of credit.

These extrinsic motivators are complemented with intrinsic motivators and personal touches (such as social and family support) to make the program more effective in building healthy habits.

For patients with chronic conditions, however, it is important that behavior change is sustained over a longer period of time. Wellth uses AI to learn how different patients respond to various nudging interventions and how their behavior changes over time. The platform also regularly polls patients to understand their socioeconomic circumstances. This real-time data collection and continuous process of iteration both help to address each patient’s unique obstacles in a personalized manner, that can also be rolled out at scale to reach thousands of patients at once.

AllazoHealth

AllazoHealth is also focused on the problem of medication non-adherence, but their AI takes a slightly different approach. Along with iteratively adapting patient interventions based on the data it has collected, AllazoHealth is able to predict the risk of non-adherence for a particular patient, as well as which interventions are likely to be most impactful for them. It does this by learning from large, diverse datasets, which include information from more than 23 million patients. This helps to personalize nudging right from the start.

AllazoHealth also places a large emphasis on the time at which an intervention is delivered, and their mode of delivery. Time of day can influence the effectiveness of a nudge. For example, it is more effective to deliver a reminder to take a medication when the patient is awake and at home, or perhaps immediately after a meal.

Lirio

Lirio partners with health systems and payers to deliver patient-centered communication, to drive patient behavior change in the right direction. Like the technologies described above, Lirio uses machine learning to process data about each patient, personalize messages, measure engagement, modify the message, and repeat. Though their messages are pre-encoded by behavioral scientists, they can be broken apart into their component elements and then re-assembled into thousands of unique combinations. This allows their machine learning technology to curate content appropriately based on a patient’s unique needs. 

Lirio’s platform can work with a variety of health issues such as cancer screenings, chronic care, and even wellness visits and is focused solely on hyper-personalized messaging to optimize engagement. Recently, Lirio leveraged their technology to increase acceptance of COVID-19 vaccination, demonstrating its potential for various other applications in health care. 

WoeBot

WoeBot is quite different from the companies described above, but leverages similar principles to deliver targeted interventions. WoeBot is an AI-powered mental health chatbot designed to help users improve their mental health. WoeBot is able to chat with users in a conversational manner, to provide 24/7 support and companionship.

This company’s technology is able to incrementally learn from the conversations it has with patients, using natural language processing (NLP) techniques to understand their emotional and cognitive states. It then uses this information to deliver targeted interventions that reflect the dynamic nature of mental health. These interventions span cognitive behavioral therapy (CBT), interpersonal psychotherapy, and dialectical behavior therapy.

LivNao

LivNao is another mental health startup that leverages personalized nudging, but again from a slightly different starting point. Through deep learning, LivNao aims to detect changes in mental health with zero input from users, using only passive data. They pride themselves in being able to identify and collect data that would traditionally only come from clinically validated questionnaires, simply by analyzing user engagement. Their predictive analytics then allow them to prompt specific members early with interventions that are included in their benefits. They target employers and insurance companies to improve productivity, reduce claims costs, and overall improve health outcomes. 

Work at TDL on MyHeartCounts

At TDL, we recently started working on a research project with the goal of improving the MyHeartCounts application. MyHeartCounts (MHC) is a smartphone-based mobile cardiovascular health research application originally developed at the Stanford University School of Medicine, in partnership with Google.

MHC is capable of providing unique and granular insights into how the balance of activity, sedentary behavior, and sleep contributes to cardiovascular health outcomes. Research on the app has thus far been limited to its capabilities as a data collection tool. At TDL, we are now working towards leveraging this data to deliver personalized interventions. One major challenge that the app has faced is user engagement and traction: the data collection is only effective if users are engaging with the app. TDL is working to better understand the specific barriers and drivers that affect engagement, in order to inform the design of an improved user interface and drive positive behavioral change across diverse populations within Canada.

Conclusion

Artificial intelligence technology allows for precisely designed messaging and tailored interventions to be delivered at scale to optimize engagement in health care. While the last decade has exploded with various new digital health technologies, the next must focus on optimizing behavior change for the diverse behavioral profiles of patients. Startups are beginning to recognize this, and personalized nudging is paving the way.

This comes with its own risks and drawbacks. For example, the collection of granular patient data can be used by insurance companies to hike premiums or deny coverage to high-risk patients. Patients may also express greater concerns about data privacy, posing a barrier to adoption and continued engagement.

The use of AI can also be susceptible to biases, and this is relevant even in the context of identifying behavioral patterns. It will continue to be important to train algorithms using large, diverse datasets during the development of similar products and interventions. As health care continues moving forward in this direction of personalized and digital care, many of the regulatory steps already being taken will directly support and de-risk efforts in personalized nudging.

References

1. Jethwani K, Kvedar J, Kvedar J. Behavioral phenotyping: a tool for personalized medicine. Per Med. 2010;7: 689–693.

2. Chen XS, Patel MS. Digital Health Tools Offer New Opportunities for Personalized Care. 18 Nov 2020 [cited 27 Dec 2021]. Available: https://hbr.org/2020/11/digital-health-tools-offer-new-opportunities-for-personalized-care

3. Mills S. Personalized nudging. Behavioural Public Policy. 2020. pp. 1–10. doi:10.1017/bpp.2020.7

About the Author

Sanketh Andhavarapu portrait

Sanketh Andhavarapu

Staff Writer

Sanketh is an undergraduate at the University of Maryland: College Park pursuing an individualized major in neuroeconomics with a minor in innovation and entrepreneurship. He is also the co-founder of an app startup centered around behavioral economics to improve clinician wellness. Sanketh hopes to apply concepts from behavioral science and neuroeconomics to improve patient and physician decision-making, and in turn, improve the overall quality of healthcare.

Read Next

Notes illustration

Eager to learn about how behavioral science can help your organization?