A Review of the Literature on Remote Monitoring Technology in Incentive-Based Interventions for Health-Related Behavior Change

Abstract

Use of technology (e.g., Internet, cell phones) to allow remote implementation of incentives interventions for health-related behavior change is growing. To our knowledge, there has yet to be a systematic review of this literature reported. The present report provides a systematic review of the controlled studies where technology was used to remotely implement financial incentive interventions targeting substance use and other health behaviors published between 2004 and 2015. For inclusion in the review, studies had to use technology to remotely accomplish one of the following two aims alone or in combination: (a) monitor the target behavior, or (b) deliver incentives for achieving the target goal. Studies also had to examine financial incentives (e.g., cash, vouchers) for health-related behavior change, be published in peer-reviewed journals, and include a research design that allowed evaluation of the efficacy of the incentive intervention relative to another condition (e.g., non-contingent incentives, treatment as usual). Of the 39 reports that met inclusion criteria, 18 targeted substance use, 10 targeted medication adherence or home-based health monitoring, and 11 targeted diet, exercise, or weight loss. All 39 (100%) studies used technology to facilitate remote monitoring of the target behavior, and 26 (66.7%) studies also incorporated technology in the remote delivery of incentives. Statistically significant intervention effects were reported in 71% of studies reviewed. Overall, the results offer substantial support for the efficacy of remotely implemented incentive interventions for health-related behavior change, which have the potential to increase the cost-effectiveness and reach of this treatment approach.

The leading causes of morbidity and mortality in developed countries (e.g., cardiovascular disease, diabetes, hypertension) are increasingly attributable to behavior or lifestyle (e.g., Higgins, 2014). Consequently, new and more effective interventions to promote health-related behavior change are sorely needed. Behavioral economic interventions in which financial incentives (e.g., cash, vouchers exchangeable for goods or services) are provided contingent on objective evidence of behavior change (e.g., biochemically verified smoking abstinence) are gaining considerable attention in efforts to promote health-related behavior change (e.g., Higgins, Silverman, Sigmon, & Naito, 2012). These interventions leverage the same preference for short-term gains that often underpin substance abuse, unhealthy food choices, and other unhealthy behavior patterns by providing monetary rewards for shorter-term healthy choices (Higgins et al., 2012; Lowenstein, Brennan, & Volpp, 2007). Such interventions are commonly referred to under the heading of Contingency Management (CM) in the substance abuse area (e.g., Higgins, Silverman, & Heil, 2008). These incentive-based interventions have been shown to be highly effective at producing healthy behavior change, including, for example, abstinence from substance abuse (Dallery, Glenn, & Raiff, 2007; Higgins et al., 1994), increases in physical activity (Kurti & Dallery, 2013; Pope & Harvey-Berino, 2013), greater medication adherence (Rigsby et al., 2000; Sorensen et al., 2007) and improvements in attendance at therapy sessions (Ledgerwood, Alessi, Hanson, Godley, & Petry, 2008).

Frequent and objective monitoring of attainment of behavior change goals and differential delivery of the incentives depending on goal attainment is a critically important element of an effective incentive program. When targeting changes in substance use, for example, drug metabolites or other byproducts of drug use are typically monitored in breath or body fluids and incentives provided when testing indicates no recent use. Such frequent monitoring can be burdensome on clinic staff and clients/patients alike if frequent clinic or home visits are involved. Moreover, those factors can also limit the geographical range over which clinical services can be delivered thereby denying treatment for individuals residing in remote or otherwise difficult to reach settings. An important strategy that is emerging to reduce or eliminate these burdens and barriers is the use of the Internet, Smartphones, or related technologies for remote monitoring of behavior change and delivery of incentives. These strategies are being used in incentive-based interventions for substance abuse (e.g., Alessi & Petry, 2013; Meredith, Grabinski, & Dallery, 2011), weight loss (e.g., Unick, Leahey, Kent, & Wing, 2015), as well as other health-related behavior problems. To our knowledge, there has not yet been a systematic review of this emerging literature, which is the purpose of the present report.

We focused on studies using technology to objectively monitor the target of the intervention or as a delivery system to provide reinforcement when the targeted goal is attained (Dallery & Raiff, 2011). Another way that technology has been integrated into incentives interventions is to supplement the incentives with feedback to participants on progress or in combining the incentives intervention with another formal treatment (e.g., computer-based cognitive behavior therapy). Although some studies included in the present review used technology for these purposes in addition to monitoring behavior and/or delivering incentives, all included reports were required to use technology to accomplish one or both of the main elements of incentives interventions remotely and we focus specifically on how technology was used for these purposes. In addition to reviewing the efficacy of these interventions, we also characterized the particular health-related behaviors and populations targeted, reinforcement schedules employed, study research designs and sample sizes, and discernable trends in the literature over the years covered by the review.

Methods

Search Strategy

The review covered an approximately 11-year time period (April 2004–May 2015) that corresponds to the time-period covered in other reviews by our group on use of financial incentives for treating substance use disorders where we noted the growth in use of remotely-implemented CM interventions (Davis et al., 2015; Higgins, Sigmon, & Heil, 2011). All studies identified for potential inclusion in this review were retrieved using a detailed search strategy. First, we used PubMed, the search engine of the U.S. National Library of Medicine, as well as PsycINFO, an electronic database offered through the American Psychological Association that focuses primarily on behavioral and social sciences. The search strategy was tailored appropriately for each database (see Appendices), but in both cases involved searching for subject headings that matched the four concepts of (a) remote behavior monitoring technology (e.g., “accelerometers”, “Internet”, “email”, “Smartphone”, “remote monitoring application”); (b) contingent financial incentives (e.g., “financial incentives”, “contingency management”, “vouchers”, “deposit contract”, “gift cards”); (c) a disorder or condition related to the target behavior (e.g., “substance abuse”, “obesity”, “hypertension”); and (d) the target behavior of the treatment (e.g., “cigarette”, “tobacco use cessation”, “body weight”, “walking”, “antihypertensive medication”). The years searched included April 2004–May 2015.

Reports generated by the above search strategy were reviewed by at least two of the authors. For inclusion, reports had to meet the following criteria: (a) used technology to accomplish one or both of the main elements of incentives interventions (i.e., monitoring behavior, delivering incentives) remotely; (b) used financial incentives; (c) was published in a peer-reviewed journal; (d) included an experimental comparison condition; and (e) was an original, prospective experimental study that reported previously unpublished data. Retrospective data analyses and secondary analyses of previously published data were excluded. With respect to the requirement that the study include an experimental comparison condition, this condition could be a non-contingent incentives condition, a different treatment intervention (e.g., cognitive-behavioral therapy, treatment as usual), or varying intensities of the incentives intervention. For those studies that employed a within-subjects design, the comparison could be a no-intervention baseline phase that preceded and followed the intervention, or a multiple-baseline design wherein the timing of the incentives intervention was staggered in time across different targets or different participants. Because the incentives interventions we reviewed often included components (e.g., feedback about progress) whose independent influence on treatment efficacy has never been assessed, we did not require that studies have isolated the effects of incentives apart from those common elements for inclusion.

Second, we reviewed bibliographies of those articles that met the above criteria and were therefore included in the present review. Disagreements about whether a report met the above criteria were resolved through discussion until consensus was reached. Given heterogeneity across studies in the experimental designs used, participant populations targeted, and intervention durations and outcomes, a meta-analysis was not conducted. Instead, we characterized the literature in terms of efficacy, the health-related behaviors and populations targeted, purposes for using technology and specific technologies used, reinforcement schedules employed, as well as trends in the literature over the time period covered by this review. We also characterized the research designs and sample sizes used in included studies, the proportion of studies reporting significant treatment effects, and, if evaluated, whether treatment effects were maintained at a follow-up assessment occurring after incentives were discontinued.

All summaries below regarding the proportion of studies that produced a statistically significant treatment effect are limited to studies that used inferential statistics to detect differences between one or more treatment groups relative to a control group. Studies in which treatment effects were determined using other methods (e.g., visual analysis of the data) were excluded from these summaries as were studies in which two levels of the incentives intervention were compared to one another in the absence of a control condition. Maintenance of treatment effects refers to: (a) statistically significant differences favoring the incentives intervention over the comparison condition during treatment that remained significant at follow-up (between-subjects designs); (b) significant improvements in the target behavior produced by the intervention that remained significant at follow-up (i.e., pre-intervention versus post-intervention [within-subjects designs with a return to baseline]); and (c) significant improvements in the target behavior remained significant at follow-up, with no differences between the end of treatment versus follow-up (within-subjects multiple baseline designs).

Results

The literature search identified 534 articles, from which 39 were selected for inclusion in the review. The main reasons for exclusion were: (a) incorporating technology in financial incentives interventions that were not delivered remotely (e.g., delivering computer-based therapy plus incentives in an outpatient research clinic; Budney et al., 2015; (b) using technology to monitor behavior remotely but not in the context of incentive-based interventions (e.g., Internet-based weight loss programs without financial incentives; e.g., Unick, Leahey, Kent, & Wing, 2015); or (c) delivering a remote incentives-based intervention without a comparison condition in between-subjects designs, or in the absence of a return to baseline or staggered introduction of the intervention in within-subjects designs [e.g., Leahey & Rosen, 2014]).

The target behavior was substance use in 18 of the 39 studies included (46.2%). Ten (25.6%) studies targeted medication adherence or adherence to home-based health monitoring (e.g., blood pressure, blood glucose), and 11 (28.2%) targeted diet, exercise, or weight loss. All 39 studies incorporated mobile technology into incentive-based health interventions to monitor the target behavior remotely. In addition, twenty-six of these studies (66.7%) also incorporated technology in the delivery of incentives. Of the 31 studies included in our calculation of the proportion of studies indicating treatment effects, 22 (71%) reported statistically significant treatment effects.

Substance Abuse

Study targets and populations

A total of 18 studies focused on substance use (Table 1), with 14 (77.8%) targeting cigarette smoking and four (22.2%) targeting alcohol use. The populations targeted by these interventions were primarily adult substance abusers. However, populations with co-morbid conditions or vulnerabilities were also targeted, including smokers with PTSD (Hertzberg et al., 2013), pregnant smokers (Harris & Reynolds, 2015; Ondersma et al., 2012), homeless Veterans who were smokers (Carpenter et al., 2015), and rural smokers (Stoops et al., 2009).

Table 1.

Interventions Targeting Substance Use Disorders

Study	Design/ Sample Size	Schedule	Duration of Treatment (weeks)	Max Earnings	Use of Technology	Target Behavior	Frequency of Monitoring/ Reward Delivery	Statistically Significant Treatment and Follow- Up Effect	Population
Dallery & Glenn (2005)	WS n = 4	Escalating with reset and bonus	3	$171.50	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/2x daily updates on earnings/vouchers immediately redeemable	Treatment: N/A Follow-Up: N/A	Smokers
Glenn & Dallery (2007)	WS n = 14	Escalating with reset and bonus	1	$56.25	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/2x daily updates on earnings/vouchers immediately redeemable	Treatment: Yes Follow-up: N/A	Smokers
Dallery et al. (2007)	WS n = 20	Escalating with reset and bonus	4	$171.50	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/2x daily updates on earnings/vouchers immediately redeemable	Treatment: Yes Follow-up: N/A	Smokers
Reynolds et al. (2008)	WS n = 4	Escalating with reset and bonus	4	$314.75	Monitor behavior, deliver incentives	Cigarette smoking abstinence	3x daily breath CO samples/≥ 2x daily emails about earnings/participants paid weekly	Treatment: N/A Follow-up: N/A	Adolescent smokers
Dallery et al. (2008)	WS n = 8	Escalating with reset and bonus	2	$78.80	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/2x daily updates on earnings/vouchers immediately redeemable	Treatment: N/A Follow-up: N/A	Smokers
Stoops et al. (2009)	BS n = 68	Escalating with reset and bonus	7	$830.75a	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/immediate feedback and vouchers	Treatment: Yes Follow-up: N/A	Rural smokers
Meredith et al. (2011)	WS n = 15	Escalating with reset and team bonus	2	$161.50	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/immediate feedback and vouchers	Treatment: Yes Follow-up: N/A	Smokers
Barnett et al. (2011)	WS n = 13	Escalating with reset	2	$154.00	Monitor behavior, deliver incentives	Reduced alcohol intake	Continuous (30-min intervals) transdermal alcohol levels/daily emails about earnings/vouchers available immediately	Treatment: Yes Follow-up: N/A	Heavy drinkers
Ondersma et al. (2012)	BS n = 110	Fixed amount and self-initiated on max of 5 prenatal care visits	10	$50.00	Monitor behavior, deliver incentives	Cigarette smoking abstinence	Breath CO and urinary cotinine > 1 week between measures/Vouchers given immediatelyb	Treatment: No Follow-up: No	Pregnant smokers
Alessi & Petry (2013)	BS n = 30	Escalating with reset	4	$340.00	Monitor behavior, deliver incentives	Alcohol abstinence	1–3 daily breath alcohol concentration tests/earnings updated at least daily and vouchers available immediately	Treatment: Yes Follow-up: N/A	Non-dependent drinkers
Dallery et al. (2013)	BS n = 77	Escalating with reset and bonus	7	$530.00	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/immediate feedback and vouchers	Treatment: Yes Follow-up: No	Smokers
Hertzberg et al. (2013)	WS n = 22	Escalating with reset and bonus	4	$530.00	Monitor behavior	Cigarette smoking abstinence	2x daily breath CO samples/vouchers delivered at end of treatment	Treatment: No Follow-up: No	Smokers with PTSD
Meredith & Dallery (2013)	Mixedc n = 32	Escalating with reset and bonus individually and in teams	2	$112.50	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/immediate feedback/vouchers delivered 1 week after treatment	Treatment: Yesd Follow-up: N/A	Smokers
Dougherty al. (2014)	Mixede n = 26	Fixed	8	$300.00	Monitor behavior	Reduced alcohol intake	Continuous (30-min intervals) transdermal alcohol levels/vouchers available weekly	Treatment: Yes Follow-up: Yes	Heavy drinkers
Carpenter et al. (2015)	WS n = 20	Escalating with reset and bonus	4	$466.00	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/vouchers delivered after incentive condition was terminated	Treatment: N/A Follow-up: N/A	Homeless veteran smokers
Dallery et al. (2015)	Mixedc n = 43	Escalating with reset and bonus individually and in teams	7	$530.00	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/immediate feedback and vouchers	Treatment: N/A Follow-up: No	Smokers
Dougherty et al. (2015)	WS n = 80	Fixed	12	$600.00	Monitor behavior	Reduced alcohol intake	Continuous (30-min intervals) transdermal alcohol levels/vouchers available weekly	Treatment: Yes Follow-up: Yes	Heavy drinkers
Harris & Reynolds (2015)	BS n = 17	Escalating with reset and bonus	4	CNBD	Monitor behavior, deliver incentives	Cigarette smoking abstinence	2x daily breath CO samples/immediate feedback and vouchers	Treatment: No Follow-up: No	Pregnant smokers

Note: Design refers to between-subjects (BS) or within-subjects (WS); n refers to sample size (across all groups); Schedule refers to the reinforcement schedule; Duration of Treatment refers to the number of weeks over which incentives contingent on the target behavior could be earned; Maximum earnings refers to the maximum amount that could be earned contingent on engaging in the target behavior (CNBD= maximum earnings could not be determined). Average or expected amounts are reported for studies using probabilistic schedules (e.g., prize-based); Use of technology refers to the purpose for which technology was used (“Other” may refer to additional components of the incentives intervention, or additional treatments in combined interventions); Statistically significant treatment effects and effects at follow-up= defined in all studies as differences in ≥ one primary outcome (p < .05) for comparisons between a control condition versus ≥ one incentives condition.

^aDid not specify amount earnable during an initial shaping phase, so total represents maximum earnings calculated without the shaping phase.

^bNot explicitly stated in the article.

^cIncentives delivered within subjects, access to online forum differed between groups.

^dControl differed from the two incentives conditions (independent vs interdependent group contingencies), which did not differ from one another.

^eIncentives delivered within subjects, sequence order differed between groups.

Technologies used in monitoring substance use

The most common method of remote monitoring involved wireless submission of data indicating participant’s substance use status. For example, in a majority of the cigarette smoking interventions, participants submitted videos of themselves taking a breath carbon monoxide (CO) test with results discernible via a web camera (e.g., Dallery & Glenn, 2005; Dallery, Meredith, and Glenn, 2008; Meredith et al., 2011; Stoops et al., 2009). These videos were submitted via email or uploaded to a secure website for later review by the researchers. Those studies targeting alcohol use monitored intake via the Secure Remote Alcohol Monitoring (SCRAM) bracelet, which detects metabolites of alcohol that are excreted through sweat (e.g., Barnett et al., 2011; Dougherty et al., 2014, 2015). These data were collected continuously while participants wore the bracelet, and later downloaded and reviewed by the researchers at periodic in-person assessments.

Technologies used in delivering reinforcement

In 15 (83.3%) studies, technology was also used for remote delivery of financial incentives contingent on biochemically verified abstinence, and the means of doing so was similar across all 15 studies. More specifically, participants were provided with statements of their recent and/or accumulated earnings via either email (e.g., Dallery & Glenn, 2005), text message (e.g., Alessi & Petry, 2013), or by accessing a study website (e.g., Barnett et al., 2011; Dallery et al., 2007, 2008; Dallery, Raiff, & Grabinski, 2013).

Reinforcement schedules

Consistent with in-person CM interventions, the technology-based substance use interventions in this review relied almost exclusively on differential reinforcement of other behavior (DRO) schedules to reinforce drug abstinence, in which the amount of reinforcement escalated contingent on biochemically confirmed, consecutive abstinent samples (Higgins et al., 1991; Roll, Higgins, & Badger, 1996; Roll & Higgins, 2000). All of the studies that used escalating schedules also employed a reset contingency, where submitting a positive sample reset the participant’s voucher earnings to their starting value, however submitting a pre-determined number of negative samples following a slip restored participant’s voucher values to the magnitude they were at before the slip. A majority of the studies using escalating schedules with resets also provided a bonus after participants submitted a certain number of consecutive negative samples (e.g., a $5.00 bonus following every three consecutive negative samples; Dallery & Glenn, 2005). One exception to the above was Ondersma et al.’s (2012) study targeting pregnant cigarette smokers, in which “incentivization attempts” were self-initiated by participants on a maximum of five prenatal care visits, with a fixed $50.00 provided for abstinent samples. Another exception was Dougherty et al. (2014; 2015) where participants received fixed, non-escalating payments on a weekly basis contingent on abstinence from alcohol use.

The frequency of monitoring the target behavior ranged from a minimum of five times during the course of a 10-week intervention (Ondersma et al., 2012), to continuous monitoring of alcohol use in real time using the SCRAM device (e.g., Barnett et al., 2011), with a majority of studies requiring participants to submit discrete samples at least twice daily (e.g., Dallery & Glenn, 2005; Dallery et al., 2008; Glenn & Dallery, 2007 [Table 1]). In addition to frequent monitoring of the target behavior, most studies also minimized the delay between verification of participant’s goal attainment and reinforcement delivery. More specifically, participants in a majority of studies received feedback about their earnings directly after providing evidence of the target behavior, and were able to redeem their vouchers immediately upon learning of their recent earnings. Exceptions to this, however, included six studies in which the delay to receiving vouchers ranged from one week (e.g., Dougherty et al., 2014; Reynolds, Dallery, Shroff, Patak, & Leraas, 2008) up to the end of a four-week treatment (Hertzberg et al., 2013; Carpenter et al., 2015).

Research designs and sample sizes

Five of the 18 (27.8%) reports targeting substance abuse employed between-subjects designs with random assignment to an intervention versus either a control or other experimental treatment condition, 10 reports used within-subjects designs, and the remaining three reports used mixed designs in which incentives were evaluated in a within subjects design but another aspect of the treatment differed between groups (e.g., access to a social forum, Dallery et al., 2015, Meredith & Dallery, 2013; sequence in which study conditions were experienced, Dougherty et al., 2014).

Sample sizes varied considerably across reports, from four smokers in Dallery and Glenn’s (2005) initial assessment of the feasibility of delivering an Internet-based financial incentives intervention to reduce cigarette smoking, to 110 pregnant cigarette smokers in Ondersma et al. (2012).

During- and post-treatment efficacy

Of the 13 studies that used inferential statistics to evaluate intervention effects, 10 of 13 (76.9%) reported significant effects of the intervention on substance use outcomes (e.g., proportion of drug-negative biological specimens, longest duration of abstinence). In the three with negative results, incentives did not produce greater abstinence than a control treatment (Ondersma et al., 2012) or a baseline condition in a within-subjects design (Carpenter et al., 2015; Hertzberg et al., 2013). These studies were smoking-cessation interventions conducted among smokers with co-morbid conditions (i.e., pregnant women, homeless Veterans, and individuals with PTSD, respectively), which may suggest that more tailored or intensive treatments are needed to promote abstinence among these relatively difficult-to-treat populations. For example, there is ample evidence supporting the efficacy of incentives for promoting smoking cessation among pregnant women in trials involving more intensive and larger magnitude incentives than were used in the Ondersma et al. report (see Higgins et al., 2012a).

Seven of the 13 (53.8%) studies above conducted follow-up assessments. Treatment effects remained significant at follow-up in only two (28.6%) reports (Dougherty et al., 2014, 2015), both of which targeted alcohol intake among heavy drinkers with no other comorbidities present. In the initial study, follow up drinking status was objectively verified while in the latter study it was based on participant self-report. In the other six studies, statistically significant treatment effects were not observed at follow-up assessments. The lack of support for the maintenance of treatment gains in these studies is presumably due to the fact that, across all studies, no components of the treatment were available and no alternative treatments were provided during the period of time between the termination of the intervention and the follow-up assessment.

Trends in research targeting substance use

One trend observed across the 18 reports targeting substance use was an increase in the number of interventions targeting hard to reach or difficult to treat populations in the later years encompassed by the review (e.g., rural populations, pregnant women). All of the reports targeting vulnerable populations were published between 2008–2015. As availability of Internet and particularly Smartphone services increases (Smith, 2013), technology-based incentives interventions may become an increasingly feasible and efficacious method for surmounting barriers to reducing substance use in these sometimes hard-to-reach vulnerable populations.

A second trend observed in this literature was the emergence of interventions targeting alcohol. No studies published between 2004–2009 targeted drinking, primarily because of two deficiencies in technology: (a) a reliable method for detecting intermittent alcohol use other than breath alcohol sensors, and (b) a device that would allow such monitoring to be done remotely. The emergence of the SCRAM device surmounted both obstacles.

A third trend observed across the present reports involved capitalizing on newer technologies (e.g., study websites, systems for transmitting data) to support incentives interventions in ways that permitted researchers to improve upon some of the important determinants of intervention effectiveness. For example, in Dallery and Glenn’s (2005) initial Internet-based incentives study, participants submitted videos of themselves blowing into a breath CO monitor by email, which had to be reviewed by the researchers before participants received feedback about their progress and study earnings. By contrast, participants in later studies by this same group submitted breath CO samples over a website (Motiv8) which provided immediate, automatic feedback about their progress in the form of a cumulative progress graph and an account activity box showing their recent and overall earnings (Dallery et al., 2008). This development decreased the delay between monitoring of the target behavior and the delivery of financial incentives, which has been shown to be a significant moderator of treatment effect size (Lussier et al., 2006). Further developments of this website included the addition of a communication forum in which participants in group-based incentives interventions could write messages to one another and view each other’s cumulative progress graphs (Dallery et al., 2015). Future studies conducted by this research group may incorporate a Smartphone-enabled CO monitor, which demonstrated good reliability and validity in pilot tests (Meredith et al., 2014) and may reduce the response effort among participants in incentives interventions by permitting them to submit evidence of behavior change using a device that is becoming ubiquitous and is carried by many individuals during all waking hours.

Medication Adherence and Home-Based Health Monitoring

Study targets and populations

Of the 10 studies included in this topic area (Table 2) four targeted adherence to antiretroviral medications among HIV-positive individuals (Barnett, Sorensen, Wong, Haug, & Wong, 2009; Moore et al., 2015; Rosen et al., 2007; Sorensen et al., 2007), two targeted blood glucose monitoring among adolescents with Type I diabetes (Raiff & Dallery, 2010; Stanger et al., 2013), two targeted Warfarin adherence among individuals with poor anticoagulation control (Kimmel et al., 2012; Volpp et al., 2008), one targeted adherence to antihypertensive medications (Petry, Alessi, Byrne, & White, 2015), and one targeted self-monitoring of blood glucose, blood pressure, and weight (Sen et al., 2014). Other than the two studies targeting adolescents with diabetes, all eight of the other studies targeted adults, with four of eight (50%) targeting especially vulnerable populations. More specifically, two of the four samples of HIV-positive individuals were also receiving methadone maintenance therapy, and the other two samples of HIV-positive individuals were current illicit drug users.

Table 2.

Interventions Targeting Medication Adherence or Home-Based Health Monitoring

Study	Design/Sample Size	Schedule	Duration of Treatment (weeks)	Max Earnings	Use of Technology	Target Behavior	Frequency of Monitoring/Reward Delivery	Statistically Significant Treatment and Follow-up Effect	Population
Rosen et al. (2007)	BS n = 56	Prize-based/escalating with reset and bonus	16	$800.00	Monitor behavior	Antiretroviral adherence	Continuous monitoring of pill bottle use/Prize draws available every 4 weeks	Treatment: Yes Follow-up: No	HIV+ substance abusers
Sorensen et al. (2007)	BS n = 66	Escalating with reset	12	$1172.40	Monitor behavior	Antiretroviral adherence	Continuous monitoring of pill bottle use/vouchers delivered every 2 weeks	Treatment: Yes Follow-up: No	HIV+ methadone maintained
Volpp et al. (2008)	WS n = 20	Fixed/lottery-based	12	$252.00–$420.00	Monitor behavior, deliver incentives	Warfarin adherence	Continuous monitoring of pill bottle use/daily feedback/voucher delivery time-frame CNBD	Treatment: N/A Follow-up: N/A	Adults prescribed warfarin
Barnett et al. (2009)	BS N = 86	Escalating	12	$1172.00	Monitor behavior	Antiretroviral Adherence	Continuous monitoring of pill bottle use/vouchers delivered every 2 weeks	Treatment: Yes Follow-up: No	HIV+ Opioid Maintained
Raiff & Dallery (2010)	WS n = 4	Fixed with bonus	1	$55.00	Monitor behavior, deliver incentives	Blood glucose monitoring	Minimum of 4 daily glucose tests/immediate feedback/vouchers delivered at end of study	Treatment: N/A Follow-up: N/A	Adolescents with type 1 diabetes
Kimmell et al. (2012)	BS n = 100	Fixed/lottery-based	24	$401.00–$504.00	Monitor behavior, deliver incentives	Warfarin adherence	Continuous monitoring of pill bottle use/daily feedback/voucher delivery time-frame CNBD	Treatment: No Follow-up: No	Adults prescribed warfarin
Stanger et al. (2013)	WS n = 17	Escalating with reset and bonus	12	$590.00	Monitor behavior	Blood glucose monitoring	Continuous monitoring of blood glucose testing/weekly feedback and voucher delivery	Treatment: Yes Follow-up: Yes	Adolescents with type 1 diabetes
Sen et al. (2014)	BS n = 75	Fixed lottery-based	12	$117.60–$235.20	Monitor behavior, deliver incentives	BPa, weight, & blood glucose monitoring	Continuous health monitoring/daily feedback/voucher delivery next-day	Treatment: Yesb Follow-up: Yes	Adults with diabetes
Moore et al. (2015)	WS n = 10	Escalating with reset	12	$544.00	Monitor behavior, deliver incentives	Antiretroviral adherence	Continuous monitoring of pill dispenser/immediate feedback of earnings/vouchers paid twice weekly	Treatment: No Follow-up: N/A	HIV+ substance abusers
Petry et al. (2015)	BS n = 29	Escalating	12	$444.00–$468.00	Monitor behavior	Anti-hypertensive Adherence	Daily monitoring of pill-taking/daily feedback of earnings/voucher delivery CNBD	Treatment: Yes Follow-up: Yes	Patients with Hypertension

Note: Design refers to between-subjects (BS) or within-subjects (WS); n refers to sample size (across all groups); Schedule refers to the reinforcement schedule; Duration of Treatment refers to the number of weeks over which incentives contingent on the target behavior could be earned; Maximum earnings refers to the maximum amount that could be earned contingent on engaging in the target behavior (CNBD= maximum earnings could not be determined). Average or expected amounts are reported for studies using probabilistic schedules (e.g., prize-based); Use of technology refers to the purpose for which technology was used (“Other” may refer to additional components of the incentives intervention, or additional treatments in combined interventions); Statistically significant treatment effects and effects at follow-up= defined in all studies as differences in ≥ one primary outcome (p < .05) for comparisons between a control condition versus ≥ one incentives condition.

^aBlood pressure.

^bControl differed from the two incentives conditions (low- versus high-magnitude vouchers), which did not differ from one another.

Technologies used in monitoring medication adherence or health monitoring

All 10 studies (100%) used technology to monitor the target behavior remotely. Among the medication adherence reports, four of seven studies monitored pill taking by measuring openings of electronic or medication event monitoring system (MEMS) caps (Barnett et al., 2009; Moore et al., 2015; Rosen et al., 2007; Sorensen et al., 2007). MEMS are pill bottles or containers fitted with microcircuitry that provide time stamps when the container is opened or closed. In these four studies, this information was stored on the device for later review by the researchers, who tracked and delivered incentives for medication adherence. Similar technology was also employed in the Stanger et al. (2013) study with diabetic adolescents, in which a blood glucose monitor detected and stored data about glucose monitoring events, which were reviewed at periodic assessments at which teens received incentives contingent on that stored information.

Technology like MEMS but slightly more advanced was employed in the Volpp et al. (2008) and Kimmel et al. (2012) studies (i.e., the Informedix Med-eMonitor System^™) on Warfarin adherence. Rather than storing the data on the device until it could be reviewed in person by the researcher and participant, the Med-eMonitor^™ in these studies communicated with a central database to denote each instance of pillbox opening, and participants received an automatically generated message requesting that they confirm whether they had ingested their medication. Similarly, the three wireless biometric devices with which participants in Sen et al. (2014) monitored glucose, blood pressure, and weight automatically transferred data to a website that was accessible by both participants and researchers.

A different approach than the above was employed in the last two reports, in which researchers remotely observed participants while they engaged in the target behavior. In Raiff and Dallery (2010), diabetic adolescents submitted videos of themselves monitoring their blood glucose via a web camera four times per day. Similarly, Petry et al. (2015) measured medication adherence by requiring participants to submit videos of themselves ingesting the antihypertensive medications using the video capabilities of a cell phone.

Technologies used in delivering reinforcement

In addition to using technology to monitor the target behavior, five of the 10 (50%) studies in this area also incorporated technology in the remote delivery of reinforcement (Kimmel et al., 2012; Moore et al., 2015; Raiff & Dallery, 2010; Sen et al., 2014; Volpp et al., 2008). Participants in all five studies received messages about their earnings, which were automatically generated shortly after participants engaged in the target behavior in all but one case (Sen et al., 2014). Worth noting is that participants in the Sen et al. (2014) study received personalized texts or emails from the researchers about their earnings. In addition to receiving statements about their earnings, participants in Moore et al. (2015) also received electronically loaded increments on debit cards contingent on antiretroviral adherence.

Reinforcement schedules

Six of 10 (60%) studies used an escalating schedule of reinforcement contingent on pillbox opening or health monitoring. Of the four studies that used other schedules, three of the four studies (Kimmel et al., 2012; Sen et al., 2014; Volpp et al., 2008) employed lottery systems in which participants had a larger probability of winning a small sum of money for engaging in the target behavior, or a smaller probability of winning a large sum of money (e.g., a one in five chance of winning $10.00 and a one in 100 chance of winning $100.00 for taking Warfarin, Volpp et al., 2008). In the fourth study that used a non-escalating schedule (Raiff & Dallery, 2010), participants received incentives worth a fixed amount (i.e., $1.00) for every glucose testing video submitted, along with a $3.00 bonus upon submitting their fourth video.

In all but two of the medication adherence and home-based health monitoring studies, engagement in the target behavior was monitored continuously and detected passively. In the two studies that required active input by the user, one required glucose testing videos to be submitted four times daily (Raiff & Dallery, 2010), and the other required participants to submit videos of themselves taking antihypertensive medications once per day (Petry et al., 2015). In contrast to the incentives interventions targeting substance use disorders, there was greater variability across studies in this topic area in the frequency with which feedback about participant’s earnings was provided, and in the delays to receiving incentives. In six studies, for example, participants received feedback about their earnings either immediately (Kimmel et al., 2012; Raiff & Dallery, 2010; Moore et al., 2015; Sen et al., 2014; Volpp et al., 2008) or within the same day (Petry et al., 2015), but the delay before participants could redeem their vouchers ranged from the next day (Sen et al., 2014) to the end of the study (Raiff & Dallery, 2010). In the other four studies, participants met with the researchers periodically throughout the treatment to transfer the data stored on their devices onto a computer where the researcher could review them. Vouchers in these studies were typically delivered at these meetings, and thus were redeemable at delays that varied from twice weekly (Moore et al., 2015) to every four weeks (Rosen et al., 2007).

Research designs and sample sizes

Six reports (60%) randomized participants to a financial incentives versus either a control or other treatment comparison condition (e.g., supportive counseling) using between-subjects designs (Barnett et al., 2009; Kimmel et al., 2012; Petry et al., 2015; Rosen et al., 2007; Sen et al., 2014; Sorensen et al., 2007). The other four studies employed within-subjects designs. Sample sizes ranged from four teens diagnosed with Type I diabetes in Raiff and Dallery (2010) to 101 adults with poor anticoagulation control in Kimmel et al. (2012).

During- and post-treatment efficacy

Eight of 10 (80%) studies evaluated the effectiveness of the intervention via statistical analyses of appropriate comparison conditions, with six of eight (75%) reporting significant increases in adherence or self-monitoring. Those studies that reported null effects included Moore et al.’s (2015) intervention to promote antiretroviral adherence among HIV-positive, illicit drug abusers, and Kimmel et al.’s (2012) intervention to increase warfarin adherence. The former attributed the lack of effect to low power given a small sample size (N = 10) while the latter attributed it to a ceiling effect (i.e., higher than anticipated pre-intervention adherence levels reduced the opportunity to observe significant increases in adherence).

Six of the eight studies above included follow-up assessments, of which three (50%) supported the maintenance of treatment gains (Petry et al., 2015; Sen et al. 2014; Stanger et al., 2013). Significant increases in glucose monitoring, along with associated decreases in blood glucose levels, were maintained three months after the incentives intervention was discontinued in the Stanger et al. study. Similarly, adherence to antihypertensive medications in the Petry et al. study differed significantly between the end of treatment versus at the three and six month follow-ups. In Sen et al., which included both low and high-magnitude incentive conditions, participants in both incentives conditions maintained treatment-produced increases in glucose-, blood pressure- and weight-monitoring at an assessment conducted three months following termination of incentives. Interestingly, treatment-produced increases in self-monitoring at the four and five-month follow-ups persisted in the low incentives condition, but not among participants in the high incentives condition where they decreased to control levels at the later follow-ups.

Treatment effects were not maintained at follow-up in the remaining three studies (Barnett et al., 2009; Rosen et al., 2007; Sorensen et al., 2007). Although voucher or cash incentives were no longer provided for medication adherence during the time between treatment completion and follow-up in any of these studies, participants in Sorensen et al. (2007) received lottery-based incentives for completing medication coaching sessions during the month leading up to the follow-up assessment.

Trends in research targeting medication adherence and home-based health monitoring

In contrast to the literature on technology-based incentives interventions targeting substance use, which grew substantially between the earlier and later half of the period surveyed, the number of medication adherence or home-based health monitoring studies grew minimally between 2004–2009 (four studies) and 2010–2015 (six studies). While this literature did not change substantially in quantity, several notable changes in qualitative aspects of these studies were discernible and are outlined below.

One trend in the present literature involved capitalizing on newer technologies to improve upon some of the limitations in earlier incentive-based interventions targeting medication adherence or other health monitoring behavior. Three of four studies published between 2004–2009 used the MEMS system to monitor medication adherence, and the Med-eMonitor^™ system used in the fourth study differed only slightly from the MEMS system. Because MEMS only detects pillbox opening, one cannot ascertain whether the medication was actually taken, or whether the intended individual opened the container. Given the emphasis on objectively verifying the target behavior in incentives interventions, methods of directly observing participants take their medications would circumvent these limitations of MEMS. Studies that used technology to facilitate direct observation of the target behavior in medication adherence or health monitoring studies (e.g., video-based methods) emerged in the later half of the period covered by this review (2010–2015) (Raiff & Dallery, 2010; Petry et al., 2015). As Smartphones grow increasingly commonplace, monitoring medication taking via cell phone, as was done in Petry et al., may become the predominant method for verifying this study target in future research.

In addition to capitalizing on newly available technologies to circumvent limitations in monitoring medication adherence remotely, Moore et al.’s (2015) method of remotely loading participant debit cards represents an especially effective method for minimizing the delay between engaging in the target behavior and earning reinforcement.

Diet, Exercise, and Weight Loss

Study targets and populations

The specific behaviors targeted across the 11 diet, exercise, and weight loss interventions (see Table 3) differed substantially, with one study targeting both diet and exercise (Spring et al., 2012), three targeting the duration of engagement in physical activity (Finkelstein, Brown, Brown, & Buchner, 2008; Hunter, Tully, Davis, Stevenson, & Kee, 2013; Weinstock, Capizzi, Weber, Pescatello, & Petry, 2014), four (36%) targeting steps per day (Donlin-Washington et al., 2014; Kullgren et al., 2014; Kurti & Dallery, 2013; Petry, Andrade, Barry, & Byrne, 2013), and three (27%) targeting weight loss (Almeida et al., 2015; Kullgren et al., 2013; Leahey et al., 2015). Seven (64%) studies targeted sedentary adults with no additional vulnerabilities or medical conditions, three (27%) studies specifically targeted overweight and obese adults (Almeida et al., 2015; Kullgren et al., 2013; Leahey et al., 2015), and one (9%) study targeted sedentary, hazardous drinking college students (Weinstock et al., 2014).

Table 3.

Interventions Targeting Diet, Exercise or Weight Loss

Study	Design/Sample Size	Schedule	Duration of Treatment (weeks)	Max Earnings	Use of Technology	Target Behavior	Frequency of Monitoring/Reward Delivery	Statistically Significant Treatment and Follow-up Effect	Population
Finkelstein et al. (2008)	BS n = 51	Variable based on activity levels achieved	4	$100.00	Monitor behavior	Aerobic activity	Continuous monitoring of activity/Vouchers delivered at end of treatment	Treatment: Yes Follow-up: No	Sedentary older adults
Spring et al. (2012)	BS n = 204	Fixed	3	$175.00	Monitor behavior	Exercise/diet	Daily goal monitoring/vouchers delivered after a 20-week follow-up	Treatment: No Follow-up: N/A	Adults with poor diet/low activity
Hunter et al. (2013)	BS n = 406	Fixed	12	£84.00	Monitor behavior	Minutes of activity	Continuous monitoring of physical activity/vouchers delivered at weeks 6 and 12	Treatment: Yes Follow-up: Yes	Sedentary Adults
Kullgren et al. (2013)	BS n = 105	Fixed/split among team	24	$600.00–$3,000a	Monitor behavior, deliver incentives	Weight loss	Monthly weight-ins and voucher deliveries	Treatment: Nob Follow-up: No	Obese adults
Kurti & Dallery (2013)	WS n = 6	Escalating	4–10	$102.50	Monitor behavior, deliver incentives	Steps per day	Continuous monitoring of steps/daily feedback/vouchers delivered at end of treatment	Treatment: N/A Follow-up: N/A	Sedentary adults
Petry et al. (2013)	BS n = 45	Prize-based/escalating with reset and bonus	12	CNBD	Monitor behavior	Steps per day	Continuous monitoring of steps/weekly feedback and prize draw	Treatment: Yes Follow-up: Yes	Sedentary older adults
Donlin-Washington et al. (2014)	WS n = 15	Prize-based/fixed	1	CNBD	Monitor behavior	Steps per day	Continuous monitoring of steps/3x weekly prize draws	Treatment: Yes Follow-up: N/A	College students
Kullgren et al. (2014)	BS n = 92	Lottery-based/fixed	16	$336.00	Monitor behavior, deliver incentives	Steps per day	Continuous monitoring of steps/weekly feedback and voucher delivery	Treatment: No Follow-up: No	Older adults
Weinstock et al. (2014)	BS n = 31	Prize-based escalating with reset and bonus	8	$230.00	Monitor behavior	Exercise activities	Variable monitoring of activitiesc/weekly feedback and prize draws	Treatment: Yes Follow-up: No	Sedentary, hazardous-drinking college students
Almeida et al. (2015)	BS n = 1,790	Variable based on magnitude of weight loss	24	CNBD	Monitor behavior	Weight loss	Quarterly weight-ins and voucher delivery	Treatment: No Follow-up: N/A	Overweight and obese adults
Leahey et al. (2015)	BS n = 268	Lottery-based	36	$50.00–$100.00	Monitor behavior, deliver incentives	Weight loss	Weekly monitoring of weight and activity/vouchers delivered at end of treatment	Treatment: Yes Follow-up: Yes	Overweight and obese adults

Note: Design refers to between-subjects (BS) or within-subjects (WS); n refers to sample size (across all groups); Schedule refers to the reinforcement schedule; Duration of Treatment refers to the number of weeks over which incentives contingent on the target behavior could be earned; Maximum earnings refers to the maximum amount that could be earned contingent on engaging in the target behavior (CNBD= maximum earnings could not be determined). Average or expected amounts are reported for studies using probabilistic schedules (e.g., prize-based); Use of technology refers to the purpose for which technology was used (“Other” may refer to additional components of the incentives intervention, or additional treatments in combined interventions); Statistically significant treatment effects and effects at follow-up= defined in all studies as differences in ≥ one primary outcome (p < .05) for comparisons between a control condition versus ≥ one incentives condition.

^aLower number= max possible for individual contingency group; higher number= max possible for team contingency group.

^bNo significant differences between control vs the two incentives arms (individual, group), which also did not differ from one another.

^cMonitoring conducted continuously for pedometer and periodically for aerobic/resistance training.

Technologies used in monitoring the target behavior

All eleven (100%) studies used technology to monitor the target behavior remotely. Coinciding with the diversity in target behaviors was a diverse array of technologies used to monitor these targets. One (9%) study required participants to self-report their daily diet and exercise practices on a website (Spring et al., 2012), four (36%) used accelerometers or other wireless sensor technology to capture bodily movements which were converted using algorithms into minutes of physical activity (Finkelstein et al., 2008; Hunter et al., 2013; Petry et al., 2013; Weinstock et al., 2014), three (27%) used the Fitbit® advanced pedometer to measure participant’s steps per day (Kullgren et al., 2014; Kurti & Dallery, 2013; Donlin-Washington et al., 2014), and three (27%) used wireless scales to measure participant’s weight (Almeida et al., 2015; Kullgren et al., 2013; Leahey et al., 2015).

Technologies used in delivering reinforcement

Four (36%) of the 11 studies incorporated technology in the remote delivery of reinforcement (Kullgren et al., 2013; 2014; Kurti & Dallery, 2013; Leahey et al., 2015). In all four cases, participants received statements about their earnings either by email contingent on engaging in the target behavior (e.g., meeting step goals on the requisite five of seven days per week [Kullgren et al. 2014; Kurti & Dallery, 2013]), or via a website that generated the messages automatically contingent on meeting a monthly weight-loss goal (Kullgren et al., 2013; Leahey et al., 2015).

Reinforcement schedules

There was substantial variability in the reinforcement schedules used in the reports on diet, exercise, or weight loss. Two (19%) studies provided fixed, non-escalating payments for meeting diet, exercise, or weight loss goals (Kullgren et al., 2013; Spring et al., 2012); three (27%) used escalating schedules with reset contingencies for consecutive time periods over which exercise goals were met (Kurti & Dallery, 2013; Petry et al., 2013; Weinstock et al., 2014); one (9%) used percentile schedules to gradually increase steps per day (Donlin-Washington et al., 2014); two (18%) involved variable payments depending on participant’s average number of aerobic minutes per week in one case (Finkelstein et al., 2008) and percent weight loss in the other (Almeida et al., 2015); one (9%) equated minutes of physical activity with points which could be redeemed at a later date (Hunter et al., 2013); and the final two (18%) studies employed lottery systems in which participants could earn different amounts of money contingent on meeting exercise or weight loss goals (Kullgren et al., 2014, Leahey et al., 2015, respectively).

Consistent with the variability in reinforcement schedules employed across the 11 diet, exercise, and weight loss studies, these studies also varied in terms of the frequency of behavior monitoring and delays to reinforcement. For example, behavior was monitored continuously in some of the studies targeting duration of engagement in physical activity (e.g., Finkelstein et al., 2008) and steps per day (e.g., Kurti & Dallery, 2013), but monthly or quarterly in the studies targeting weight loss (Kullgren et al., 2013, Almeida et al., 2015, respectively). Of course, it should be noted that this variability may be related to differences in the nature of these two study targets, where weight would be expected to change more gradually over time relative to the other targets. With respect to the delay between engaging in the target and receiving reinforcement, participants in four (37%) studies could not redeem their incentives until the end of treatment (Finkelstein et al., 2008; Kurti & Dallery, 2013; Leahey et al., 2015; Spring et al., 2012). Similarly, participants in Hunter et al. (2013) had only two opportunities during the 12-week intervention to exchange the points that they accumulated for engaging in physical activity for rewards. Those studies that involved relatively short delays between researcher verification of participant’s engagement in the target and delivery of incentives included studies where participants earned incentives at delays ranging from once weekly (e.g., Petry et al., 2013; Weinstock et al., 2014) up to three times per week (Donlin-Washington et al., 2014).

Research designs and sample sizes

All reports used between-subjects designs with the exceptions of Kurti and Dallery (2013) and Donlin-Washington et al. (2014). Of the nine studies that used between-subjects designs, five randomized participants to an incentives condition versus either a control condition (Finkelstein et al., 2008; Hunter et al., 2013; Leahey et al., 2015; Petry et al., 2013; Spring et al., 2012) or another treatment comparison condition (Almeida et al., 2015; Weinstock et al., 2014). Participants in the last two between-subjects studies (Kullgren et al., 2013; 2014) were assigned to either a non-contingent incentives condition versus one of two levels of the incentives condition. In both studies, the incentives arm included an individual- and a group-based condition, where the group condition involved access an online forum in one study (Kullgren et al., 2013), and exposure to a contingency in which a $500.00 incentive was split among members who achieved their weight loss goals in the other study (Kullgren et al., 2014). The sample sizes in these studies ranged from six sedentary older adults in Kurti and Dallery (2013) to 1,790 overweight/obese adults across 28 worksites in Almeida et al. (2015).

During- and post-treatment efficacy

Ten of 11 studies (90.9%) conducted statistical analyses to evaluate treatment effects, of which six (60%) reported significant effects (Finkelstein et al., 2008; Hunter et al., 2013; Donlin-Washington et al., 2014; Leahey et al, 2015; Petry et al., 2013; Weinstock et al., 2014). In two instances these differences were based on self-reported attainment of the target behavior while the four others were based on objective measures (Hunter et al., 2013; Weinstock et al. 2014). The remaining four studies reported that incentives had no influence on participant’s percent weight loss (Almeida et al., 2015; Kullgren et al., 2013), step counts (Kullgren et al., 2014), or diet and exercise behaviors (Spring et al., 2012).

Seven of the 10 studies above (70%) also conducted follow-up assessments, of which three (42.9%) reported that statistically significant treatment effects were present at follow-up (Hunter et al., 2013; Leahey et al., 2015; Petry et al., 2013). In contrast, intervention-produced increases in activity were not maintained in Finkelstein et al. (2008) or Weinstock et al. (2014), and the non-significant effect of treatment in the two Kullgren et al. (2013, 2014) studies remained non-significant at follow-up. In all of these studies, no treatment components remained in place following discontinuation of incentives, although participants in Kullgren et al. (2014) were permitted to keep their pedometers for the two-month period between treatment completion and follow-up.

Trends in research targeting diet, exercise, and weight loss

With the exception of Finkelstein et al. (2008), the other ten studies in this section were published in a relatively short three-year span between 2012–2014. Therefore, the development of technology-based financial incentives interventions targeting these behaviors appears to be a very recent trend in the literature.

Related to the recent emergence of these reports, a second trend observed was variability in the definitions of exercise used in these studies, and consequently, differences in the technologies used to monitor exercise remotely. Achieving greater consistency with respect to how exercise is operationally defined is necessary for comparing interventions targeting this behavior. Such consistency might be achieved by taking into account the current national guidelines that specify the level of physical activity required to attain health benefits (e.g., decreasing risk for cardiovascular disease). For example, the Centers for Disease Control (CDC, 2008) recommends that adults achieve at least 150 minutes of moderate intensity activity per week (i.e., 30 minutes per day on at least five days), which has been shown to be approximately equal to walking 10,000 steps per day (Le-Masurier, Sidman, & Corbin, 2003). However, only three studies delivered incentives for targets that were broadly consistent with this goal (Finkelstein et al., 2008; Kullgren et al., 2014; Kurti & Dallery, 2013), and all three of those studies were primarily designed to promote the acquisition of increased activity levels as opposed to maintaining them.

Given the shorter period over which the interventions in this topic area have evolved relative to those targeting substance use or medication adherence/home-based health monitoring, there has been little opportunity to observe whether technologies used in monitoring the target behavior have undergone substantial change over time. However, the fact that one of the first two studies targeting diet and exercise (Spring et al., 2012) relied on self-reported behavior whereas later studies capitalized on technological advancements such as Fitbits^™ (Donlin-Washington et al., 2014; Kullgren et al., 2014; Kurti & Dallery, 2013) may suggest that exercise-based incentives interventions that use newly developing and conveniently available technology are on the horizon. In addition, two of the seven reports targeting exercise or weight loss involved a group-based intervention arm (Kullgren et al., 2013, 2014). Perhaps such group approaches will comprise a reasonable portion of interventions targeting diet, exercise, and weight loss in future research. Doing so may appeal to a preference reported in the literature for home-based exercise interventions that incorporate social components (Brawley, Rejeski, & King, 2003; Wilcox, Castro, King, Housemann, & Brownson, 2000).

Discussion

The use of technology to facilitate the remote implementation of incentive-based interventions to improve health is a burgeoning area of research, especially interventions targeting substance use, home-based health monitoring, and diet, exercise, and weight loss. Considered collectively, the number of technology-based financial incentives interventions grew from 11 studies published between 2004–2009, to 28 studies published between 2010–2015. Of the 31 reports statistically assessing treatment effects, 22 (71%) reported significant changes in the target behavior, providing strong support for the promise of remotely implemented financial incentives interventions.

A majority of the reports included in the present review targeted substance abuse, of which 83.3% produced statistically significant increases in abstinence rates. The three studies that reported no difference between an incentives condition versus either a control or other treatment comparison condition targeted special populations (Carpenter et al., 2015; Hertzberg et al., 2013; Ondersma et al., 2012). Whether these interventions failed to reduce substance use because of the methodology employed, small sample sizes, or because the populations targeted were particularly difficult to treat, was unclear. Nonetheless, the growth over time in technology-based incentives studies targeting vulnerable populations suggests that using technology to surmount geographical and socioeconomic barriers to treatment delivery may be a growing trend in the literature. This approach may be particularly promising among vulnerable populations in which the prevalence of substance use is disproportionately high (e.g., cigarette smoking among disadvantaged individuals), but who nonetheless have increasing access to technology (e.g., Internet, Smartphones [Smith, 2013]).

Of all study targets examined in this review, technology-based incentives studies targeting medication adherence have arguably demonstrated the least change over time in terms of the diversity of technologies used to facilitate remote monitoring of the target behavior. Specifically, the predominant method of monitoring medication adherence across studies was the MEMS system, which indicates whether a pill container was opened but not whether the medication was actually taken. Studies such as Petry et al. (2015), in which participants submitted videos of themselves ingesting their medications, help circumvent the limitations of MEMS but place a greater burden on the participant, as the target behavior is not verified automatically using this method. Interestingly, these were the only two methods of remote behavior monitoring represented across the studies targeting medication adherence. Developments in technology to detect medication taking passively, such as digital pills which produce a voltage during digestion and communicate this information to external sensors (e.g., Bosworth, 2012), should permit more accurate, remote monitoring in future incentive-based interventions targeting medication adherence.

Those reports targeting exercise demonstrated the greatest variability in how the target behavior was defined, and in the technologies used to monitor the target remotely. These differences may be attributable to the fact that all but one of the interventions targeting this class of behavior were published between 2012–2014, thereby reducing the opportunity to observe changes over time. As mentioned previously, future incentive-based interventions targeting exercise could achieve greater uniformity in how exercise is operationalized by referring to current exercise guidelines (e.g., CDC, 2008). Given the growing obesity crisis in the United States, it is critical to develop interventions that produce clinically significant improvements in this well-established predictor of negative health outcomes.

With respect to technology-based incentives interventions targeting substance use, the relatively poor maintenance of treatment effects at follow-up underscores the need for greater research attention to this topic. In particular, it will be important for future research to develop methods of programming for the maintenance of treatment gains. Examples of how this might be accomplished are evident in studies that integrated the community reinforcement approach (CRA) with contingent incentives for cocaine abstinence (e.g., Higgins, Wong, Badger, Odgen, & Dantona, 2000; Higgins et al., 2007; Garcia-Fernandez et al., 2011; Secades-Villa et al., 2011). In these studies, participants received in-person counseling that focused on developing naturalistic sources of reinforcement that would sustain behavior change once the contrived financial incentives were withdrawn, with the results indicating improvements is both sustained abstinence and psychological functioning.

Another model to sustain treatment effects among those with substance use disorders is one wherein incentives remain in place longer-term as developed by Silverman and colleagues (Silverman et al., 2005). Serious consideration should be given to the possibility that keeping incentive-based interventions in place for extended durations may be cost-effective. Just as chronic medication regimens for conditions such as diabetes and hypertension cost substantially less than the medical crises that may result from discontinuing these therapies, it is certainly conceivable that chronic incentives interventions to sustain drug abstinence, medication adherence, or other health-related behavior change in high-risk populations may be a cost-effective but to date unexplored treatment option. Importantly, careful consideration must be given in designing programs that are intended to remain in place long term. For example, incentive-based wellness programs offered by employers (e.g., programs that provide rewards for using the gym) are not cost-effective when employee engagement rates are low, as a majority of the money spent on the program ends up being earned by people who would have engaged in the rewarded behavior even without the incentive, rather than by those who are non-adherent or not engaged (Loewenstein, Asch, & Volpp, 2013). However, such programs can be made more cost-effective by modifying particular design features of the intervention (e.g., separating incentive payments from paychecks to make the incentives more salient to participating employees).

Given the growing contribution of modifiable lifestyle factors to risk for chronic disease and premature death, the need for improved health-related behavior change interventions is clear. Remotely managed financial incentive programs hold substantial promise for reducing practical barriers and increasing the reach of efficacious health-promoting treatments. Of course, additional research will be needed to refine how we can better capitalize on technological advancements to accurately monitor and reinforce targeted changes in health behaviors. Moreover, cost-effectiveness studies will be critically important in evaluating the general strategy of using financial incentives to promote health-related behavior change as well as the relative merits of remote versus in-person incentives interventions. Accomplishing these research tasks has the potential to contribute substantially to reducing the advent impact of unhealthy lifestyles on individual and population health.

Appendices

Appendix 1. PubMed Search Strategy

1	(“Employee Incentive Plans”[Mesh]) OR “Reward”[Mesh]
2	compensation[tiab] OR “contingency management”[tiab] OR “deposit contract”[tiab] OR “deposit contracts”[tiab] OR “financial incentive”[tiab] OR “financial incentives”[tiab] OR “financial reinforcement”[tiab] OR “gift card”[tiab] OR “gift cards”[tiab] OR lottery[tiab] OR “loyalty card”[tiab] OR “loyalty cards”[tiab] OR monetary[tiab] OR money[tiab] OR payment[tiab] OR payments[tiab] OR prize[tiab] OR prizes[tiab] OR voucher[tiab] OR vouchers[tiab]
3	1 OR 2
4	(((((“Behavior Therapy”[Mesh]) OR “Health Promotion”[Mesh]) OR “Patient Compliance”[Mesh]) OR “Therapy, Computer-Assisted”[Mesh]) OR “Tobacco Use Cessation”[Mesh]) OR “Treatment Outcome”[Mesh]
5	Abstinen*[tiab] OR “addiction treatment”[tiab] OR “addiction treatments”[tiab] OR adheren*[tiab] OR “alcohol use reduction”[tiab] OR “incentive based”[tiab] OR intervention[tiab] OR “patient compliance”[tiab] OR “self report”[tiab] OR “smoking cessation”[tiab] OR “smoking reduction”[tiab] OR “tobacco use cessation”[tiab] OR “tobacco use reduction”[tiab]
6	4 OR 5
7	(((((((((((“Accelerometry”[Mesh]) OR “Breath Tests”[Mesh]) OR “Computer-Aided Design”[Mesh]) OR “Computer-Assisted Instruction”[Mesh]) OR “Computer Peripherals”[Mesh]) OR “Computers, Handheld”[Mesh]) OR “Internet”[Mesh]) OR “Mobile Applications”[Mesh]) OR “Monitoring, Physiologic”[Mesh]) OR “Telemedicine”[Mesh]) OR “Telephone”[Mesh]) OR “Video Recording”[Mesh]
8	Accelerometer[tiab] OR breathalyzer[tiab] OR “cell phone”[tiab] OR “cell phones”[tiab] OR cellphone[tiab] OR cellphones[tiab] OR computer[tiab] OR computers[tiab] OR “bottle caps”[tiab] OR “electronic measurement”[tiab] OR “electronically monitored”[tiab] OR email[tiab] OR “e-mail”[tiab] OR fitbit[tiab] OR internet[tiab] OR “medication event monitoring system”[tiab] OR mems[tiab] OR “message board”[tiab] OR “message boards”[tiab] OR “mobile application”[tiab] OR “mobile applications”[tiab] OR “mobile phone”[tiab] OR “mobile phones”[tiab] OR “mobile technology”[tiab] OR “online discussion forum”[tiab] OR “online discussion forums”[tiab] OR “online forum”[tiab] OR “online forums”[tiab] OR pedometer[tiab] OR pedometers[tiab] OR phone[tiab] OR phones[tiab] OR “remote monitoring”[tiab] OR “smart phone”[tiab] OR “smart phones”[tiab] OR telephone[tiab] OR telephones[tiab] OR texting[tiab] OR transdermal[tiab] OR video[tiab] OR web[tiab] OR wireless[tiab]
9	7 OR 8
10	((((((((((((((((((“Alcohol Abstinence”[Mesh]) OR “Alcohol Drinking”[Mesh]) OR “Alcohol-Related Disorders”[Mesh]) OR “Anticoagulants”[Mesh]) OR “Antihypertensive Agents”[Mesh]) OR “Blood Glucose”[Mesh]) OR “Blood Glucose Self-Monitoring”[Mesh]) OR “Body Weight”[Mesh]) OR “Health Promotion”[Mesh]) OR “Hypertension”[Mesh]) OR “Methadone”[Mesh]) OR “Narcotics”[Mesh]) OR “Nicotine”[Mesh]) OR “Opioid-Related Disorders”[Mesh]) OR “Substance-Related Disorders”[Mesh]) OR “Tobacco Use”[Mesh]) OR “Tobacco Use Cessation”[Mesh]) OR “Walking”[Mesh]) OR “Weight Loss”[Mesh]
11	alcohol*[tiab] OR “antihypertensive medication”[tiab] OR “blood glucose”[tiab] OR cannabis[tiab] OR cigarette*[tiab] OR “drug use”[tiab] OR “illicit drug”[tiab] OR “illicit drugs”[tiab] OR marijuana[tiab] OR methadone[tiab] OR nicotine[tiab] OR obes*[tiab] OR opioid[tiab] OR opioids[tiab] OR “physical activity”[tiab] OR smok*[tiab] OR “substance abuse”[tiab] OR “substance use disorder”[tiab] OR “substance use disorders”[tiab] OR tobacco[tiab] OR walk*[tiab] OR warfarin[tiab] OR “weight loss”[tiab]
12	10 OR 11
13	3 AND 6 AND 9 AND 12

Appendix 2. PsycINFO Search Strategy

SU.EXACT(“Incentives”) OR SU.EXACT(“Contingency Management”) OR
SU.EXACT(“Monetary Incentives”) OR TI,AB,IF(compensation OR “contingency management”
OR “deposit contract*” OR “financial incentive*” OR “financial reinforcement” OR “gift card*” OR
lottery OR “loyalty card*” OR monetary OR money OR payment* OR prize* OR voucher*)
AND
SU.EXACT.EXPLODE(“Mobile Devices”) OR SU.EXACT.EXPLODE(“Websites”) OR
SU.EXACT.EXPLODE(“Computer Software”) OR SU.EXACT.EXPLODE(“Computer
Applications”) OR SU.EXACT.EXPLODE(“Technology”) OR SU.EXACT.EXPLODE(“Electronic
Communication”) OR SU.EXACT.EXPLODE(“Internet Usage”) OR
SU.EXACT.EXPLODE(“Online Therapy”) OR SU.EXACT.EXPLODE(“Internet”) OR
TI,AB,IF(accelerometer OR breathalyzer OR “cell phone*” OR computer* OR “bottle cap*” OR
“electronic measurement” OR “electronically monitored” OR email OR “e-mail” OR fitbit OR
internet OR “medication event monitoring system*” OR MEMS OR “message board*” OR
“mobile application*” OR “mobile phone*” OR “mobile technology” OR “online discussion
forum*” OR “online forum*” OR pedometer* OR phone* OR “remote monitoring” OR “smart
phone*” OR telephone* OR texting OR transdermal OR video OR web OR wireless)
AND
SU.EXACT.EXPLODE(“Narcotic Drugs”) OR SU.EXACT.EXPLODE(“Drug Usage”) OR
SU.EXACT.EXPLODE(“Physical Activity”) OR SU.EXACT.EXPLODE(“Body Weight”) OR
SU.EXACT.EXPLODE(“Walking”) OR SU.EXACT.EXPLODE(“Nicotine”) OR
SU.EXACT.EXPLODE(“Cannabis”) OR SU.EXACT.EXPLODE(“Glucose”) OR
SU.EXACT.EXPLODE(“Drug Abuse”) OR SU.EXACT.EXPLODE(“Addiction”) OR
TI,AB,IF(alcohol* OR “antihypertensive medication” “blood glucose” OR cannabis OR cigarette*
OR “drug use” OR “illicit drug*” OR marijuana OR methadone OR nicotine OR obese OR
obesity OR opioid* OR “physical activity” OR smoke OR smoking OR “substance abuse” OR
“substance use disorder*” OR tobacco OR walk* OR warfarin OR “weight loss”)
AND
SU.EXACT.EXPLODE(“Treatment”) OR SU.EXACT.EXPLODE(“Methadone Maintenance”) OR
SU.EXACT.EXPLODE(“Intervention”) OR SU.EXACT.EXPLODE(“Behavior Change”) OR
SU.EXACT.EXPLODE(“Drug Abstinence”) OR SU.EXACT.EXPLODE(“Reinforcement”) OR
SU.EXACT.EXPLODE(“Smoking Cessation”) OR SU.EXACT.EXPLODE(“Treatment
Compliance”) OR SU.EXACT.EXPLODE(“Drug Rehabilitation”) OR
SU.EXACT.EXPLODE(“Treatment Effectiveness Evaluation”) OR
SU.EXACT.EXPLODE(“Behavior Therapy”) OR TI,AB,IF(abstinence OR abstinent OR
“addiction treatment” OR adherence OR adhere OR “alcohol use reduction” OR “incentive
based” OR intervention OR “patient compliance” OR “self report” OR “smoking cessation” OR
“smoking reduction” OR “tobacco use cessation” OR “tobacco use reduction”)
SU.EXACT.EXPLODE(“Treatment”) OR SU.EXACT.EXPLODE(“Methadone Maintenance”) OR
SU.EXACT.EXPLODE(“Intervention”) OR SU.EXACT.EXPLODE(“Behavior Change”) OR
SU.EXACT.EXPLODE(“Drug Abstinence”) OR SU.EXACT.EXPLODE(“Reinforcement”) OR
SU.EXACT.EXPLODE(“Smoking Cessation”) OR SU.EXACT.EXPLODE(“Treatment
Compliance”) OR SU.EXACT.EXPLODE(“Drug Rehabilitation”) OR
SU.EXACT.EXPLODE(“Treatment Effectiveness Evaluation”) OR
SU.EXACT.EXPLODE(“Behavior Therapy”) OR TI,AB,IF(abstinence OR abstinent OR
“addiction treatment” OR adherence OR adhere OR “alcohol use reduction” OR “incentive
based” OR intervention OR “patient compliance” OR “self report” OR “smoking cessation” OR
“smoking reduction” OR “tobacco use cessation” OR “tobacco use reduction”)