Digital Health and Addiction

Abstract

Digital technologies are rapidly changing how we understand and promote health. A robust and growing line of research has examined how digital health may enhance our understanding and treatment of addiction. This manuscript highlights innovations in the application of digital health approaches to addiction medicine, with a particular emphasis on advances in (1) real-time measurement of drug use events, (2) real-time measurement of the confluence of factors that surround drug use events, and (3) research examining how real-time measurement can inform responsive, in-the-moment interventions to prevent and treat substance use disorder. Although this manuscript focuses on addiction medicine as one exemplar of the striking impact of digital health, science-based digital health offers generalizable solutions to scaling-up unprecedented models of precision healthcare delivery across a broad spectrum of diseases across the globe.

Introduction

The Promise of Digital Health

Digital technology is transforming our world and promises to change how we understand and promote health. The explosion of digital devices and “big data” analytics allows for the unprecedented collection and interpretation of enormous amounts of granular data about everyday behaviors and the context in which they occur.

Digital health [1, 2]refers to the use of data captured via digital technology (sometimes referred to as “digital exhaust” data)[3] to both understand people’s health-related behavior and provide personalized health care resources. Smartphones and smartwatches enable passive, unobtrusive ecological sensing that provides continuous measurement of individuals’ behavior and physiology, such as sleep, social interactions, physical activity, electrodermal activity, and/or cardiac activity[4]. Individuals can provide consumer-generated data in response to queries on mobile devices (sometimes referred to as “ecological momentary assessment” or EMA) to reflect (timestamped) snapshots of their daily lives. Rich information can be gleaned via EMA about individual’s context, mood, behavior, social interactions, pain, sleep, stress levels among other factors (See Figure 1 for examples of EMA questions in a digital health app). And voluminous social media data allow for a rich understanding of individuals’ social networks, communications, and behavior. These data sources (separately or in combination) may provide new insights into digital (behavioral or biological) markers of health-related phenomena and their evolution over time[5]. Indeed, methods such as digital phenotyping[6] or reality mining allow for moment-by-moment quantification of individual-level information collected via personal digital devices as individuals live their daily lives.

Figure 1:

Sample EMA Questions- Types and Formats

In traditional medicine, clinical assessments are typically conducted by trained professionals using structured diagnostic instruments, medical devices, and clinically-validated assessments. Digital health data enhance these conventional sources of clinical data with ecologically-valid data captured in naturalistic contexts. And indeed, digitally-derived data are increasingly considered an essential part of an “evolving health data ecosystem[7, 8]” and promise to markedly impact both discovery science as well as translational science.

Digital Health and Addiction Medicine

One particularly vibrant area of research in digital health has been in the field of addiction medicine[9–13]. Research in addiction medicine is increasingly providing new insights into the neurobiology of addiction[14, 15], including the neurocircuitry underlying the reinforcing value of drugs as well as the molecular biology of the actions of drugs of abuse. This has informed a plethora of research focused on developing and evaluating various pharmacotherapeutic agents that target mechanistic processes underlying addiction as well behavioral treatments intended to disrupt processes that maintain addictive behavior and promote a new behavioral repertoire.

Along with these advances in the neurobiological understanding of addiction, a robust line of research has examined how digital health may enhance our understanding and treatment of addiction. Indeed, the very first “prescription digital therapeutic” given market authorization by the U.S. Food and Drug Administration (reSET® followed by reSET-O®; Pear Therapeutics) is for the treatment of substance use disorders and was approved based on strong empirical support from rigorous research[16, 17]. This development pioneers a new domain in which FDA-sanctioned software is deemed safe and effective in the treatment of disease. This enables the delivery of evidence-based substance use disorder treatment anytime and anywhere. This is particularly important given the insufficient addiction medicine clinical workforce to meet the population-level need – a phenomenon that is particularly apparent in the current U.S. opioid crisis.

And research using wearable and mobile data is enabling the development of dynamic models of substance use behavior in real-time and in response to changing environmental, social, physiological, and intrapersonal factors. This manuscript highlights innovations in the application of digital health approaches to addiction medicine, with a particular emphasis on research innovations in the last 2 years and discusses promising new directions in this field. Consistent with the theme of this Special Issue on incorporating wearable and mobile data into systems biology research, we specifically focus on: advances in (1) real-time measurement of drug use events, (2) real-time measurement of the confluence of factors that surround drug use events, and (3) research examining how real-time measurement can inform responsive, in-the-moment interventions to prevent and treat substance use disorder. A summary of these studies is provided in Table 1.

Table 1.

Summary of Studies on Digital Health Applied to Addiction Medicine

Real-Time Measurement of Drug Use via Mobile Sensing
Publication	Substance Measured	Data Source
Leffingwell, T.R., et al., 2013 (Reference 22)	alcohol detection	electrochemical fuel cells measuring alcohol vapor released from skin
Nomura, A., et al., 2019 (Reference 23)	cigarette smoking detection	exhaled carbon monoxide concentration levels
Holtyn et al., 2019 (Reference 24)	cocaine detection	photoplethysmography (PPG) sensor measuring heart rate interbeat interval; accelerometer measuring physical activity; gyroscope measuring motion
Simons et al., 2015 (Reference 25)	alcohol detection	electrochemical fuel cells measuring alcohol vapor released from skin
Bae et al., 2017 (Reference 26)	alcohol detection	56 sensor features related to time, movement patterns (e.g., accelerometry, rotation), communication (e.g., phone calls, texts), and psychomotor impairment (e.g., keystroke speed)
Krishnan, N., et al., 2019 (Reference 27)	cigarette smoking detection	exhaled carbon monoxide concentration levels
Real-time Measurement of the Confluence of Factors that Surround Drug Use Events
Publication	Factors Measured	Data Source
Preston, K.L., et al., 2018 (Reference 32)	craving, stress, drug use	ecological momentary assessment self-report
Epstein, D.H., et al., 2009 (Reference 33)	exposure to drug cues, mood changes, drug use	ecological momentary assessment self-report
Shiffman, S. et al., 2004 (Reference 34)	negative mood, smoking	ecological momentary assessment self-report
Marsch et al., 2020 (Reference 11)	sleep, stress, pain, craving, withdrawal, context, mood, drug use, momentary self-regulation, medication adherence	ecological momentary assessment self-report; smartwatch and smartphone sensing (e.g., location, activity, sleep heart rate, phone usage); social media
Linking Real-Time Measurement to Inform Responsive, In-the-Moment Substance Use Interventions
Publication	Intervention	Data Source
Hebert, E.T., et al., 2020 (Reference 39)	tailored messages responsive to imminent smoking risk	ecological momentary assessment self-report
McClure, J.B., et al., 2016 (Reference 40)	tailored advice for managing nicotine withdrawal symptoms and medication side-effects	ecological momentary assessment self-report
Hebert, E.T., et al., 2018 (Reference 41)	interventions tailored to stress, smoking urge, and cigarette availability	ecological momentary assessment self-report

Real-Time Measurement of Drug Use.

Substance use is traditionally measured via self-report (which is subject to recall bias and error) or urine toxicology testing (which is intrusive[18]). Additionally, neither of these traditional methods provides temporal precision about when use events occurred and for what duration. Wearable mobile sensor technologies allow for the continuous measurement of physiological data in the natural environment. These data may provide a richer understanding of substance use patterns as well as more granular outcomes measurement in clinical trials evaluating therapeutics for the treatment of substance use disorders[19–21].

To date, mobile sensing devices have been developed to detect alcohol use transdermally (via electrochemical fuel cells that measure alcohol vapor released from the skin) and can provide near real time data to alcohol researchers[22]. Smoking can be measured via carbon monoxide smoke detectors linked to a smartphone[23]. And promising data are emerging demonstrating that cocaine use can be measured via the application of dynamical systems models to data collected via smartwatches measuring heart rate (e.g., interbeat intervals) and physical activity data[24]. Overall, research focused on real-time measurement of drug use has highlighted the intra-day dynamics of substance use across and within individuals (e.g., drinking speed and quantity [25, 26]) and underscored the feasibility of providing near real-time biomarker feedback to individuals (e.g., feedback on adherence to smoking cessation[27, 28]).

Real-time measurement of the confluence of factors that surround drug use events.

Behavior, including substance use behavior, occurs in a context. This may include, for example, situational demands (including both internal and external triggers for engaging in substance use), changing life circumstances over time, changing resources (e.g., social support) and opportunities to engage in risk behavior[29]. Digital health allows us to capture rich physiological, behavioral, intrapersonal and social data from individuals on a longitudinal basis and in many contexts. And, increasingly sophisticated data analytics can turn these rich data into meaningful insights about individual health via advanced analytic approaches such as machine learning and deep learning approaches to prediction as well as dynamic models that examine time-varying relationships among intensive longitudinal data[30].

A dynamic understanding of the confluence of factors impacting substance use behavior may allow us to markedly refine and advance theoretical models of addiction and offers great promise for transforming the field of addiction medicine. Indeed, the Director of the U.S. National Institute on Drug Abuse has called for addiction treatment models that are dynamic and personalized including a call to “develop algorithms to identify indicators of relapse risk and incorporating them into wearable devices and smartphones”[31].

As one example, a recent digital health EMA study helped elucidate the relationship between craving, stress, and drug use[32]. Specifically, EMA data from persons in treatment for opioid use disorder who were asked to respond to survey prompts or initiate surveys on mobile devices throughout the day revealed that stress increased before reported stressful events but not before drug use events. In contrast, craving increased hours before drug use as well as before stressful events and remained at high levels after such events. These results underscored a stronger link between craving and drug use than between stress and drug use – a result not well understood by structured clinical assessments captured on an infrequent, episodic basis. Other EMA research has demonstrated that drug triggers (e.g., exposure to drug cues or mood changes) increased for hours before cocaine use events but not before reports of cocaine craving. In contrast, drug triggers increased for hours before reports of heroin craving but not before heroin use events[33]. And lapses to smoking among smokers trying to quit were associated with increases in negative mood for many days (not hours) before a smoking lapse event[34]. These results provide new insights into how self-reported EMA data collected in real time can help us understand one’s risk profile for substance use events and how that may differ across various sub-populations of substance users.

In addition to using EMA to measure factors surrounding drug use events in real time, additional digital assessment tools may be of value. Indeed, understanding the separate and combined utility of EMA, passive sensing and social media data in predicting clinically meaningful events for persons with substance use disorders is an important and understudied area of scientific inquiry. One ongoing study is examining how rich data collected from EMA (e.g., inquiring about sleep, stress, pain, craving, withdrawal, context, mood, substance use, momentary self-regulation, and medication adherence), smartwatch sensing (of location and distance traveled, physical activity, sleep, and heart rate), smartphone sensing (of App usage, audio/conversation, call/text, GPS, screen on/off, phone lock/unlock, phone notification information, Wi-Fi & Bluetooth logs, sleep, ambient light, and proximity) and social media data may (separately and combined) aid in the understanding of the interplay of factors that may precipitate events of medication/treatment non-adherence and relapse among individuals in treatment for opioid use disorder[11]. This study may reveal what subset of digitally-derived data may be particularly useful to include as part of broader outcomes measurement in clinical intervention studies.

Linking real-time measurement to inform responsive, in-the-moment substance use interventions.

Digital health approaches can also be used to prompt the delivery of interventions in real-time. Such interventions, sometimes referred to as ecological momentary interventions (EMIs[35]) or Just-in-Time Adaptive Interventions (JITAIs[36]) are designed to be directly responsive to the health needs of an individual, such as when they are craving a substance of abuse. Such interventions are intended to capitalize on states of opportunity when an individual may be most receptive to an intervention and to deliver the right type and amount of therapeutic support at the right time[37].

A digital JITAI targeting substance use can learn about factors associated with risk of relapse event and measure ongoing relapse risk via the digital assessment methods described earlier and then provide in-the-moment interventions designed to promote protective factors and prevent relapse events[38]. Thus, JITAIs are not static interventions that always work the same way across individuals or even within an individual, but rather they can enable dynamic computational models to predict and respond to people’s changing needs, goals, and clinical trajectories over time.

Research on adaptive, in-the-moment interventions for substance use disorders is still nascent but is gaining momentum and producing promising results, particularly in the realm of smoking cessation. A recent pilot study [39]examined a JITAI that used EMA data to assess risk of imminent smoking lapse and then provided tailored messages that were responsive to an individual’s risk profile (e.g., advised to chew a piece of nicotine gum). Results suggested that tailored messages may hold promise for better engaging individuals and help them stop smoking compared to usual care. Another pilot study [40]found that a JITAI that adaptively tailored advice for managing nicotine withdrawal symptoms and medication side-effects was used more often and shown to be more acceptable among smokers seeking to quit compared to a non-adaptive psychoeducational intervention. Additionally, a study[41] conducted with patients in a smoking cessation clinic showed that a JITAI that measured risk for smoking lapse and provided responsive interventions tailored to stress, smoking urge, and cigarette availability resulted in greater reductions in those triggers of lapse compared to non-tailored interventions (although effect sizes were moderate).

Conclusions and Future Directions

Digital health is rapidly changing how we understand and treat addiction. The ecological sampling and precision assessment that digital health approaches afford is increasingly providing new insights into the nature of addiction. And digital therapeutics offer anytime/anywhere treatment to individuals seeking to change drug-taking behavior. Delivering interventions on digital platforms is scalable, personalizable, and can greatly extend the reach and impact of addiction medicine services. Given the high prevalence of addiction and the limited addiction treatment workforce, digital health enables the delivery of science-based therapeutic addiction resources to any corner of the globe.

Although this manuscript focused on the promise of digital health in the realm of addiction medicine, this highlights only one exemplar of the striking impact of digital health. Indeed, a robust body of scientific research has illuminated trans-diagnostic, dynamic processes of health behavior via digital data capture and has underscored the robust, replicable effects that digital treatments can have in treating a broad array of health conditions. Additionally, although precision medicine has largely embraced individuals’ genetic or molecular profiles when offering personalized health care, including individual’s digital profiles, that capture everyday behavior and physiology, promises to transform the field of precision medicine.

Digital health thus provides us with tremendous opportunity to transcend our siloed models of understanding and treating disease to embrace the full spectrum of health and wellness. Indeed, harnessing science-based digital health offers considerable promise for offering generalizable solutions to scaling-up a science-based understanding of health across a broad spectrum of diseases/disorders and informing unprecedented models of precision healthcare delivery accessible around the world.

Highlights.

• The application of digital health to addiction medicine is a vibrant area of research.

• Digital therapeutics provide science-based substance use disorder treatment anytime and anywhere.

• Wearable and mobile data are enabling the development of dynamic models of substance use behavior in real-time and in response to changing environmental, social, physiological, and intrapersonal factors.

• Digital health promises to transform precision medicine approaches to treatment.