Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Dec 1.
Published in final edited form as: Health Psychol. 2021 Feb 11;40(12):897–908. doi: 10.1037/hea0001057

Digitally Characterizing the Dynamics of Multiple Health Behavior Change

Bonnie Spring 1, Tammy K Stump 1, Samuel L Battalio 1, H Gene McFadden 1, Angela Fidler Pfammatter 1, Nabil Alshurafa 1, Donald Hedeker 2
PMCID: PMC8355237  NIHMSID: NIHMS1727589  PMID: 33570978

Abstract

Objective:

We applied the ORBIT model to digitally define dynamic treatment pathways whereby intervention improves multiple risk behaviors. We hypothesized that effective intervention improves the frequency and consistency of targeted health behaviors, and that both correlate with automaticity (habit) and self-efficacy (self-regulation).

Methods-Study 1:

Via location scale mixed modeling we compared effects when hybrid mobile intervention did versus did not target each behavior in the Make Better Choices 1 (MBC1) trial (n=204). Participants had all of four risk behaviors: low moderate-vigorous physical activity (MVPA) and fruit and vegetable consumption (FV); high saturated fat (FAT) and sedentary leisure screen time (SED). Models estimated the mean (location), between-subject variance, and within-subject variance (scale).

Results:

Treatment by time interactions showed that location increased for MVPA and FV (Bs=1.68, .61 ps<.001) and decreased for SED and FAT (Bs=−2.01,−.07, ps<.05) more when treatments targeted the behavior. Within-subject variance modeling revealed group by time interactions for scale (taus=−.19, −.75, −.17, −.11 ps<.001): all behaviors grew more consistent when targeted.

Methods-Study 2:

In the MBC2 trial (n=212) we examined correlations between location, scale, self-efficacy, and automaticity for the three targeted behaviors.

Results:

For SED, higher scale (less consistency), but not location correlated with lower self-efficacy (r=−.22, p=.014) and automaticity (r=−.23, p=.013). For FV and MVPA, higher location, but not scale, correlated with higher self-efficacy (rs=.38, .34, ps<.001) and greater automaticity (rs=.46, .42, ps<.001).

Conclusions:

Location scale mixed modeling suggests that both habit and self-regulation changes probably accompany acquisition of complex diet and activity behaviors.

Introduction

Poor-quality diet and physical inactivity are the most prevalent risk factors for chronic diseases, including diabetes, cardiovascular disease, and cancers (Adams et al., 2017; Arena et al., 2015; Bauer et al., 2014; Baruth et al., 2011; Lloyd-Jones et al., 2010; Mendis, et al., 2015; Mozaffarian et al., 2015; Myint, et al., 2009; Schuit, et al., 2002; Spring et al., 2014; Spring et al., 2013). In turn, chronic diseases are the main causes of premature death and disability and the leading drivers of the U.S.’s $3.3 trillion annual health care costs (Centers for Disease Control and Prevention, 2019). Just 1 in 10 U.S. adults consumes the recommended intake of fruits and vegetables (Lee-Kwan et al., 2017), and fewer than one-third meet dietary guidelines to consume less than 10% of calories from saturated fats (American Heart Association, 2015; Sacks et al., 2017; U.S. Department of Health and Human Services and U.S. Department of Agriculture, 2015). Only about half meet public health recommendations for moderate-vigorous physical activity (MVPA; Centers for Disease Control and Prevention, 2018) and more than 50% exceed two sedentary hours per day watching television (Fedewa et al., 2015). Moreover, risk behaviors co-occur: the average adult reports at least two; 25% report three or more (Baruth et al., 2011; Chou, 2008; Meader et al., 2017; Schuit et al., 2002).

Public health guidelines for diet and physical activity advise consumers to accumulate a total amount of a food commodity (e.g., servings/cups of fruits/vegetables) or type of activity (total minutes MVPA) on a regular basis. They also specify how eating and activity behaviors should optimally be distributed over time: e.g., whether target levels need to be met daily or weekly in order to achieve a health benefit (USDHS, 2010, 2015, 2018; Dunton, 2018). What neither current guidelines nor behavioral theories specify well is how to design interventions so that they produce the temporal pattern of changes that leads to a sustainable modification of behavior. As others have noted, existing psychological theory and analytic methods are more adept at characterizing differences between people than fluctuations within a person over time (Riley et al 2011; Dutton & Atienza, 2009; Dunton, 2018).

This is an era when digital sensing affords continuous, real-time detection of changes in a person’s behavioral state and surrounding context. New technologies and analytic methods allow us to measure and conceptualize the dynamics of behavior change in ways that can inform our theories and interventions (Riley et al., 2011). Whereas traditional examination of intervention effects has focused on overall changes to the mean level or frequency of a behavior, we can now leverage technological advances to identify more complex behavioral patterns. The problem being addressed is how to identify behavioral features that are evident before the end of treatment and that can successfully predict whether a healthy behavioral change will be maintained after treatment is withdrawn. Later, we elaborate how we plan to draw predictors of maintenance from a broad array of behavioral features (e.g., changes in day-to-day variability, rate of change in variability, responses to “lapses” in behavior). The present study takes a first step toward that goal by characterizing patterns of change in daily variability (scale) and level (location) of targeted behaviors within a successful intervention.

This paper’s premise is that breakthrough advances for complex health behavior change interventions will be facilitated by introducing more dynamic concepts and tools into two domains of the science of behavior change. The first needed development is an expansion of the manner in which traditional statistical methods characterize within-person change. The second is integration into psychological theory of dynamic constructs that explain how an intervention induces a pattern of transitions between psychological and behavioral states that leads to the acquisition of durable behavioral change. When stated dynamically in algorithmic form, adaptive treatment decision rules (Murphy, 2005) can specify how change in the patient’s response pattern and change in the intervention should come to reciprocally determine each other over time (Bandura,1985).

Intervention Development Model – Orbit Phase 1

Step 1 in the ORBIT treatment development model (Czajkowski et al., 2015) is to identify a clinically meaningful question. Our question is how to maximize healthy change in multiple diet and physical activity risk behaviors for the least possible resource expenditure. Effective interventions for diet and PA risk behaviors exist but are intensive: involving multiple treatment sessions, each lasting from 10–90 minutes (Curry et al., 2014; Curry et al., 2018). Patients (Becker et al., 2017; Jensen, et al., 2012) and payers (Arterburn et al., 2008; Jones et al., 2015) consistently name high cost and burden (long treatment duration, high time commitment) as top barriers to uptake of intensive behavioral treatments. Yet, even after intensive lifestyle intervention, behavioral improvements often are not maintained (Arterburn et al., 2008; Perri & Practice, 1998), prompting recourse to the most common maintenance strategy: continuing to offer behavioral treatment (Perri & Practice, 1998), which further augments cost and participant burden.

Accordingly, a critical goal for the science of health behavior change is to detect when, during treatment, an intervention has produced durable behavioral improvement that might indicate when intervention delivery can begin to be tapered and then discontinued. Likewise, an ability to detect during the post-treatment maintenance phase when a previously stable healthy behavior pattern is beginning to “wobble,” could indicate, just in time, when intervention should be reinstated. Such knowledge would allow treatment dosing decisions to be made by assessing dynamic processes that underlie the acquisition and maintenance of durable healthy eating and PA patterns.

Our specific intervention development goal is to optimize the existing Make Better Choices (MBC) intervention for multiple diet and physical activity change (Spring et al. 2012, 2018) to achieve and maintain the maximal healthy behavior improvement that is attainable for the least resource utilization (cf., Spring et al, 2020). All of the work shown here represents activity undertaken in either Orbit Phase 1a (Define: during which the scientific foundation of the behavioral treatment is defined), or ORBIT Phase 1b (Refine: when candidate targets for treatment components are specified and the hypothesized pathway by which treatment produces benefit is formulated) (Czajkowski et al., 2015). Our Phase 1 treatment development process involves performing secondary analysis of two prior MBC clinical trials (Spring et al, 2012, 2018). In ORBIT Phase 1a, we use these data to define what it means to achieve healthful behavior change. We apply location-scale modeling to identify two behavioral features (level and temporal consistency) that change as different interventions produce acquisition of guidelines-concordant diet and activity behaviors. These secondary analyses were prompted by the premise that the acquisition of healthy behavior change is more likely to be durable when intervention causes the behavior to both reach its targeted level and be enacted consistently. In Phase 1b, we sought to refine our understanding of this behavior change patterning by learning how what is measured as changes in behavioral level and consistency fits into the nomological network of psychological constructs (Cronbach & Meehl, 1955) that are thought to explain the successful acquisition of sustained healthy behavior change. Two constructs in this network pertain to the acquisition of self-regulation skills and the acquisition of habits. For behavioral consistency to be useful as a new construct, it should converge with these related constructs but not overlap entirely.

Orbit Phase 1A. Acquisition of Behavioral Consistency

In this phase, we perform secondary analysis of prior clinical trial data to define what it means to achieve multiple health behavior change. We proposed in the introductory section of this manuscript that boosting the growth of dynamic behavior change interventions requires expanding the manner in which traditional statistical methods measure within-person change. Most treatment evaluations have inferred healthy habit acquisition from changes in the frequency or rate of a behavior: increases for healthy behaviors; decreases for unhealthy ones. Many statistical analysis techniques for intensive longitudinal data are well-suited to model this type of improvement in behavioral rate. In contrast, another aspect of intra-individual change (improvement in behavioral consistency) has not been measured routinely. Yet, for many medical and public health guidelines, maintaining a stable homeostatic range of a biomarker or behavior is at least as important as attaining a mean target level (Riley et al, 2011; Dunton, 2018; American Diabetes Association, 2020). Consider, for example, two individuals with diabetes and acute coronary syndrome who have the same average blood glucose levels across the month, one whose glucose is within a healthy range on most days and the other whose glucose fluctuates between healthy, hyperglycemic, and hypoglycemic. Fewer adverse cardiac events are expected for the patient who maintains a stable glucose level than the one whose glucose shows great within-person variability (Gerbaud, Darier, Montaudon, Beauvieux, et al, 2019). Thus, we sought an analytic technique that could measure within-person variability in order to study how it is impacted by behavioral intervention. Such an analytic technique could, for instance, let us differentiate interventions and individuals that not only improve fruit and vegetable consumption to a targeted level on average but also improve consumption to a similar, high level of fruits and vegetables on most days.

Mixed-effects location scale modeling (a statistical analysis technique for intensive longitudinal data; Hedeker et al., 2008; Hedeker et al., 2012) appeared to offer the needed capabilities. Location scale modeling quantifies change in within-person variability of behavior (scale), unlike most analytic techniques which only assess changes in absolute level (location). The availability of this method let us evaluate the hypothesis that effective interventions produce both improvement in location (increases for healthy behaviors; decreases for unhealthy ones) and decrease in scale (i.e., increased behavioral consistency) of targeted behaviors. We studied this by applying location scale modeling to data from a previously published trial of behavioral interventions to improve multiple diet and PA behaviors (Spring et al., 2010, 2012).

The Make Better Choices 1 (MBC1) Trial

The Make Better Choices (MBC1) study enrolled 204 adults with unhealthy levels of sedentary leisure screen time (SED), fruit/vegetable intake (FV), saturated fat intake (FAT) and moderate vigorous physical activity (MVPA). All participants had all four risk behaviors and were randomized to one of four hybrid interventions involving an app plus remote coaching: 1) increase FV and MVPA, 2) decrease FAT and SED, 3) decrease FAT and increase MVPA, 4) increase FV and decrease SED (Spring et al., 2010, 2012). The aim was to determine which combination of one diet and one PA intervention target produced the maximum improvement in all four risk behaviors. Based on behavioral choice theory (Bickel & Vuchinich, 2000), we hypothesized that the increase FV and decrease SED intervention condition would yield maximum improvement in a composite measure combining all four behaviors because of substitute and complementary relationships between FV, SED, and the other behaviors. Specifically, we predicted and found that in addition to directly increasing FV, this intervention indirectly decreased FAT: FV partially substituted for (crowded out) FAT, probably as a result of increased satiety due to heightened fiber intake. We also observed a complementary relationship between SED and FAT, such that decreasing leisure screen time was accompanied by decreased FAT, at least partially because reducing television viewing also decreased the hand-to-mouth snacking with which participants paired it. By the end of treatment, the group that was asked to increase FV and decrease SED improved on a standardized composite score reflecting all behaviors more than the groups in other treatment conditions that were asked to change different pairs of diet and PA behaviors (p<.001), and the difference was maintained through the 6-month follow-up period (Spring et al., 2012).

Methods

All participants were asked to wear an accelerometer and use a custom-designed app to self-monitor dietary intake, MVPA, and SED daily during a 2-week Baseline phase, early phase treatment [Rx1: 1 week] when goals were set to 50% of final level; and later phase treatment [Rx2: 2 weeks] when goals were set to 100%. The intervention for all participants included three weeks of telephone coaching as well as a mobile application designed based upon Control Systems Theory (Carver & Scheier 1982) to help participants set diet and PA goals, self-monitor their behavior and receive feedback about progress, and earn financial incentives for self-monitoring and goal attainment. In the 6-month follow-up phase, participants were only incentivized to self-monitor on the following schedule: daily for one week at week 4, 3 consecutive days for weeks 5 and 6, biweekly for 6 weeks, and monthly until 6-month follow-up. Goal attainment was no longer incentivized. Study procedures were approved by the Institutional Review Boards at the University of Illinois at Chicago and Northwestern University.

Analyses were run separately for each of the four behaviors, with participants divided into one of two groups (Targeted or Not Targeted) with regard to whether their assigned intervention targeted that behavior or targeted other behaviors. For instance, for analysis of FV, the Targeted group included those in the increase FV and increase MVPA intervention condition as well as those in the increase FV and decrease SED condition. For FAT, the Targeted group included those in the decrease FAT and SED intervention condition, and those in the decrease FAT, increase MVPA condition. For MVPA, the Targeted group included those in the increase FV and MVPA intervention condition, and those in the decrease FAT, increase MVPA condition. For SED, the Targeted group included those in the decrease FAT and SED intervention condition, and those in the increase FV and decrease SED condition. Here we tested the hypothesis that each behavior would show an interaction between treatment group and time, such that improvement in the behavior’s rate and consistency would be greater for the group whose intervention targeted that behavior than for those for whom the behavior was not targeted, demonstrating treatment differentiation for the enactment of specific diet and PA behavioral improvements.

Each of the four behavioral outcomes was analyzed using a mixed-effects location scale model, as implemented in the MIXREGLS software program (Hedeker & Nordgren, 2013). To better satisfy the normality assumption of the model, a square root transformation was used for the count outcomes: fruit and vegetable consumption (FV; servings), MVPA (minutes), sedentary leisure behavior (SED; minutes), and arc sin transformation was used for the percentage outcome (FAT; % daily calories attributable to saturated fat). The mean, BS variance, and WS variance models included terms for group (0/1), time (0=baseline, 1=Rx1, 2=Rx2) and group by time interaction. The mean model corresponds to a regression model, while the variance models (BS and WS) are log-linear regression models so that the resulting variances are always positive (Hedeker et al., 2008). Group was dummy-coded separately for the four outcomes, such that group=1 for participants in the conditions that targeted that outcome. For example, in the analysis of MVPA, group=1 for those in the conditions that were targeted to increase MVPA, and 0 otherwise. For SED, group=1 for the conditions that were targeted to decrease sedentary behavior, and 0 otherwise. For FV, group=1 for the conditions that were targeted to increase FV, and 0 otherwise. For FAT, group=1 for the conditions that were targeted to decrease saturated fat consumption, and zero otherwise. With this coding of the outcomes, the main effect of group represents the group difference at baseline, the time effect represents the time trend for the group=0 (i.e., not targeted) conditions, and the group by time interaction represents the difference in the time trend for the group=1 (i.e., targeted) conditions, relative to group=0 condition. Thus, the group by time interaction is of greatest interest here, as it indicates the effect of targeting behaviors on the outcomes across the time intervals.

Results

For MVPA (Figure 1), there were significant effects of time (β = 0.25, p = 0.008) and the group by time interaction (β = 1.68, p = 0.001) in the mean model. This indicates that all subjects increased MVPA levels across the three time intervals (baseline, Rx1, Rx2), but that the conditions that were designed to target increasing MVPA did so to a much greater degree (cf. Figure 1). In terms of the BS variance modeling, there was a significant effect of time (alpha = 0.22, p = 0.001) indicating that subjects became more heterogeneous across the three time intervals. For the WS variance modeling, there was only a significant group by time interaction (tau = −0.19, p = 0.001). This indicates that the consistency of MVPA within subjects did not change across days for the conditions in which MVPA was not targeted, but MVPA became more consistent for the conditions in which it was targeted (i.e., the WS variance was reduced across the time intervals for the targeted MVPA conditions). Exponentiating this estimate yields a variance ratio effect of exp (−0.19) = 0.83. This indicates that the WS variance was reduced by 17% (100%−83%=17%), such that the level of PA became more consistent with each successive time interval after baseline for the targeted MVPA group, relative to the group for which MVPA was not targeted.

Figure 1.

Figure 1.

Open in a new tab

Changes in mean and WS variance for square root transformed minutes of MVPA over time between groups whose intervention targeted vs. did not target increasing MVPA.

Similar findings were also observed for the other three behaviors, such that the mean level of targeted healthy behaviors (FV) increased over time and the mean level of targeted unhealthy behaviors (SED and FAT) decreased over time. In the conditions in which they were targeted, behaviors also became more consistent over time. Table 1 presents results for the location scale mixed models run separately for each behavioral outcome, in terms of the effects on the location (mean) and scale (WS variance). As predicted, these results revealed increases in mean levels of targeted healthy behaviors (PA and FV), decreases in mean levels of targeted unhealthy behaviors (SED and FAT) (i.e., improved location), and decreases in WS variance (i.e., increased consistency, improved scale) for all targeted behaviors.

Table 1.

Results of location scale models run separately for each behavioral outcome.

MVPA FV SED FAT
Mean (Location) Model
Intercept (β) 5.28*** .80*** 13.37*** 3.29***
Time (β) .25*** .11*** −.68*** −.18***
Group (β) .30 .12* −.31 −.08*
Group*Time (β) 1.68*** .61*** −2.01*** −.07*
WS Variance (Scale) Model
Intercept (tau) 3.06*** −.53*** 3.30*** −.74***
Time (tau) −.03 .01 −.13*** −.002
Group (tau) .10 .18* −.03 .03
Group*Time (tau) −.19*** −.75*** −.17*** −.11***
Open in a new tab
***

p<.001,

**

p<.01,

*

p<.05

Note. MVPA: minutes of moderate-vigorous physical activity. FV: fruit and vegetable servings. SED: minutes of sedentary leisure screen time. Fat: percent of calories consumed from saturated fat

Discussion

These findings provide preliminary support for the use of location scale mixed modeling to evaluate improved behavioral frequency and increased consistency during the acquisition of healthy diet and activity changes induced by an effective intervention. It is noteworthy that both location and scale improved for each of the quite different health behaviors that the interventions targeted. Location improvements characterized both increasing behavioral levels for low-rate healthy behaviors (FV, MVPA) and decreasing behavioral levels for high-rate unhealthy ones (FAT, SED). Regardless of the directionality of the targeted behavior change, interventions that effectively improved the behavior’s location also lowered its scale, diminishing within-person variability across time. Stated differently, effective interventions that achieved guidelines-recommended levels of dietary intake or PA also achieved guidelines-recommended consistency of the behavior across occasions (i.e., in this case, days). It might be asked why we consider consistent non-action or minimal action for an unhealthy behavior to be a good thing, once effective intervention has reduced behavioral rate (location) to low or zero. It is because, just as single or rare slips to an unhealthy behavior increase the odds of a full relapse, continued, consistent lapse-free intervals build self-efficacy that is protective against relapse (Kirchner et al., 2012; Larimer et al., 1999). In sum, the introduction of location-scale modeling illustrates the first scientific development that we found needed to support more dynamic behavioral interventions: improved modeling of within-person change.

ORBIT Phase 1B: How is behavioral consistency related to psychological constructs thought to underlie healthful behavior change?

The second scientific development that we found needed is the integration into psychological theory of dynamic constructs that explain how a pattern of transitions between psychological and behavioral states could lead to the acquisition of behavioral changes (particularly those that might be maintained). In this instance, we reanalyze prior clinical trial data to learn whether changes to the absolute level and consistency of various diet and PA behaviors co-occur with changes to psychological variables previously identified as important accompaniments of behavior change. When behavioral consistency emerges, does self-regulation improve? Do habits form and strengthen? Both? In other words, we refine understanding of the MBC intervention’s treatment pathway by identifying which psychological variables are associated with improved behavioral level and consistency. In doing so, we sought to provide evidence for behavioral consistency’s recognition as a distinct construct and to explore its placement within the nomological network of related constructs (Cronbach & Meehl, 1955).

Two main classes of psychological theory are especially relevant to achieving changes to healthy behavior that become consistently enacted at the daily level: self-regulation theory (Bandura, 1991; Baumeister et al., 2007; Kanfer & Gaelick-Buys, 1991) and habit theory (Aarts & Dijksterhuis, 2000; Gardner, 2015). Self-regulation is the process of volitionally exerting control over the self, by inhibiting competing responses, in order to change or sustain a pattern of thought, feeling, or behavior (Baumeister et al.,1994). An intervention to improve self-regulation might teach skills to execute generalizable behavior change strategies or techniques that have utility across many different risk behaviors and contexts. Such behavior change techniques (BCTs) have been characterized in comprehensive taxonomies put forward by Michie and colleagues (Abraham & Michie, 2008; Michie et al., 2011; Michie et al., 2013). Our recent meta-review of meta-analyses examining interventions to foster diet and physical activity changes for health or weight loss, found that goal-setting and self-monitoring were by far the most commonly evaluated BCTs (Spring et al., 2019). None of the specific 14 BCTs analyzed was consistently related to diet and activity improvements, however (Spring et al., 2019). On the other hand, in multiple studies, increased self-efficacy has been a consistently observed consequence of effective interventions that train self-regulatory skills. Self-efficacy often mediates the beneficial effects of self-regulatory interventions on diet and PA behavioral outcomes (Schneider et al., 2016; Darker et al., 2010). Hence, we consider greater self-efficacy to be an indicator of better self-regulation.

Self-regulation is a goal-driven and effortful undertaking, particularly when it involves pursuit of multiple different behavior change goals that require complex new action patterns. Given the considerable burden of executing several System 2 deliberative, conscious, goal-directed cognitive processes at once (Hagger, 2016; Kahneman, 2011), it would be helpful if intervention could be designed to reduce the burden of multiple behavior change. One possible avenue is to repeat targeted behaviors sufficiently often in the presence of a consistent context that the actions become habits. Habits are cue-driven, triggered automatically by contextual stimuli, undergirded by System 1 automatic processes, no longer requiring conscious effort to initiate (Kahneman, 2011; Lally et al., 2008; Lally & Gardner, 2013). When life demands are high, being able to rely on habitual responding frees up cognitive resources to address other challenges that require deliberation and reflection (Neale et al., 2013).

Consistency has slightly different meanings and is operationalized differently when framed from the vantage point of habit theory versus self-regulation theory. As noted above, the appeal of self-regulatory strategies or BCTs is that they are generalizable: consistently executable and effective across contexts that are variable or changing. In that respect, temporal self-regulatory consistency maps well onto public health guidance that prescribes performing a behavior daily or weekly regardless of whether the person is in usual surroundings or in a radically different time zone or continent with different cues. Self-regulatory consistency that causes a behavior to be repeated regularly over time regardless of context is thought to be achieved deliberatively by the imposition of goal striving

The premise that consistency (scale) is also a marker of habit acquisition emerged from animal models of stimulus-response (S-R) learning, whereby stimuli (e.g., light, tone) were thought to “hijack” control of an associated conditioned response, rendering actions predictable (Skinner, 1953; Watson, 1925). For humans, the stimuli thought to hijack responding and induce consistency are cues such as an environmental trigger, time of day, or immediate prior action. Also generalized from animals to humans is frequency (location), the other longstanding habit acquisition marker. As stimulus-response pairings are learned, behaviors are thought to become produced, changing in rate (frequency) as the number of cue exposures increase or decrease. In humans, the initial phase of habit learning is thought to rely on goals, motives, and rewards. As responding becomes habitual, those influences are posited to diminish. The behavior becomes automatically triggered by contextual cues (Kwasnicka et al. 2016; Lally & Gardner, 2013; Wood, 2017), although some influence of rewards and goals may persist (Aarts & Dijksterhuis, 2000; Phillips, 2019). Hence, consistency, from the perspective of habit theory means that specific physical or psychological contextual cues reliably trigger specific behavioral responses. For example, location-scale modeling has been applied to test whether negative mood consistently cues cigarette smoking, on the one hand, and, on the other hand, whether smoking consistently triggers improved mood (Hedeker et al., 2008).

One set of signs that behavioral control has changed from deliberative (effortful and difficult to sustain) to habitual (automatic and persistent) is that responding speeds up and mental effort reportedly decreases (Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977). Another is that different brain neural circuitry becomes activated. Whereas initial learning activates a network that links motor regions (basal ganglia, caudate) with the midbrain and prefrontal cortex (associated with self-control and planning), continued performance shifts activation to a neural network connecting the putamen of the basal ganglia with the sensorimotor cortices and parts of the midbrain (Knowlton & Patterson, 2016; Tricomi et al., 2009; Yin & Knowlton, 2006). Hence, responding appears to transition from being regulated by higher cortical structures to being underpinned by sensorimotor circuits.

Habit theory performs well for understanding the acquisition of uncomplicated behaviors that become conditioned to discrete cues. Debate is ongoing about whether habit formation also can explain the acquisition of more complex health-promoting diet and PA behaviors (Mullan & Novoradovskaya, 2018; Trafimow, 2018). To the extent that habitual responses are defined as those that have become completely cue-driven and free from cognitive influence, it is unlikely that habit alone will suffice to explain the acquisition of complex behaviors. Eating and PA behaviors appear to remain under partial cognitive control, never acquiring full automaticity (Gardner & Lally, 2018; Rhodes & Rebar, 2018), and continuing to need some degree of volitional self-regulation. We note, though, that there are interesting hybrid forms of regulation. For example, when using implementation intentions, a self-regulatory strategy, an individual deliberatively self-administers a cue that she consistently pairs with a specific response. By doing so repeatedly, the person functionally administers a stimulus-response conditoning protocol whereby the cue may eventually come to trigger the response automatically.

Habit research has a long history of objectively measuring habit formation. Goal independence, behavioral continuation after withdrawal of reward, altered frequency and consistency of behaviors are all objective markers of habit formation that trace from animal to human research. Recently, however, habit acquisition has become equated with the attainment of subjective automaticity (Kwasnicka et al., 2016; Lally & Gardner, 2013; Wood, 2017), shifting research attention away from overt behavioral measurement and toward assessment of subjective experience (Verplanken & Orbell, 2003). The concern that automatic mental processes are, by definition, inaccessible to conscious awareness (Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977) is mitigated somewhat by the finding that the trajectory of change in automaticity is directly preceded and predicted by the individual’s trajectory of behavioral frequency and consistency (Gardner & Lally, 2018; Lally et al., 2010).

To refine our understanding of the MBC intervention’s treatment pathway, we analyzed whether the improved level and consistency produced in each diet and PA behavior was associated with greater automaticity (operationalizing habit strength) or greater self-efficacy (operationalizing self-regulation).

The Make Better Choices 2 (MBC2) Trial

The MBC1 trial demonstrated that a multiple behavior change intervention targeting improvements in fruits/vegetables and sedentary leisure screen time maximized overall healthy change in the two targeted behaviors as well as in saturated fat intake. MVPA did not improve as a tag-along, incidental change, but it did improve to guideline-recommended levels in other intervention conditions that targeted it directly. These results suggested a need for intervention to target MVPA directly in addition to FV and SED in order to bring about improvement in all 4 risk behaviors. On the other hand, there was reason for concern that addressing 3 behaviors at once (FV, SED, PA) could prove overwhelming (Schulz et al., 2014; Spring et al., 2004). Hence, the MBC2 trial compared two different versions of the MBC intervention to a contact control intervention that addressed stress and sleep. A Simultaneous MBC treatment condition intervened on all of FV, SED, and PA throughout a 12-week intervention period. A Sequential MBC treatment condition targeted FV and SED for the first 6 weeks and added PA for the last 6 weeks. We hypothesized that both Simultaneous and Sequential MBC intervention would improve diet and PA as compared to control, and that Sequential would produce greater benefit because its response demands were more manageable. In fact, both Simultaneous and Sequential MBC treatment produced comparably large diet and PA improvements to guideline levels, which were sustained to final follow-up at 9 months (Spring et al, 2018). Of particular relevance here, the MBC2 study assessed both automaticity and self-efficacy for each behavior at the end of the treatment phase. This enabled us to evaluate whether the level or the consistency of each diet and PA behavior was associated with greater habit formation or better self-regulation.

Methods

The MBC2 trial enrolled 212 adults, all of whom demonstrated unhealthy levels of SED, FV, FAT, and moderate vigorous physical activity (PA) during a 1 week baseline period of self-monitoring performed using a custom smartphone app and worn accelerometer (Spring et al., 2012). Study procedures were approved by the Institutional Review Board at Northwestern University. During a 12-week intervention, participants used the app to set diet and PA goals, self-monitor their eating and PA, receive feedback about progress from the app, and receive remote telephone coaching. Goals for the Sequential and Simultaneous intervention conditions were tapered until targeted behaviors reached guideline levels. For the Simultaneous condition, goals for all three criterion health behaviors (FV, SED and MVPA) were tapered to guideline levels from weeks 1 through 6 of the intervention period, after which participants were instructed to maintain goal level performance for the subsequent 6 weeks of the intervention period. For the Sequential condition, FV and SED goals were tapered to guideline levels during weeks 1 through 6, and maintained thereafter, whereas MVPA goals were introduced and tapered up to guideline level during weeks 7 through 12. Hence, goals for all three targeted behaviors for both Simultaneous and Sequential conditions had been fully tapered to guideline levels during the last two weeks of treatment (weeks 11 and 12).

The week following the intervention period (week 13) was the first study follow-up assessment period. Participants were instructed to continue self-monitoring their behaviors during this period but they received no additional coaching. They also completed self-report questionnaires about perceived automaticity and self-efficacy for each targeted diet and PA behavior (FV, SED, MVPA). Automaticity was assessed using the automaticity sub-scale of the Self-Report Habit Index (SRHI; Verplanken & Orbell, 2003), which shows good internal consistency (α=.92; Gardner et al., 2012). Self-efficacy was assessed using a measure developed to evaluate self-perceived ability to enact health behavior changes (Marcus et al., 1992). The measure also showed good internal consistency for behaviors targeted by MBC intervention (α=.94–.96; Schneider et al., 2016). Minor adjustments were made to adapt wording to the specific targeted behaviors (FV, MVPA, and SED).

Our goal was to evaluate associations between location and scale for each targeted behavior with self-efficacy and automaticity for that behavior. Hence, we computed location and scale for each behavior during the last 2 weeks of the intervention phase (weeks 11 and 12), when goals for all three criterion behaviors in both diet and PA intervention conditions had been fully tapered to guideline levels. Given their lack of difference, data from the Simultaneous and Sequential MBC intervention conditions were combined for this analyses. Data from the control Stress and Sleep condition were not included in these analyses since that group self-monitored behaviors other than diet and PA during weeks 11 and 12.

Location and scale estimates were computed for each of the three targeted behaviors (FV, SED, MVPA) during the last two weeks of the intervention (weeks 11 and 12) using mixed-effects location scale modeling implemented in the MIXREGLS software program (Hedeker & Nordgren, 2013). Models were run separately for each of the three targeted behaviors (FV, SED, MVPA). As for the MBC1 data, square root transformations were applied to each count outcome (FV servings, SED minutes, MVPA minutes) to better approximate normality. Sex was included as a covariate, dummy coded such that male was the reference group. Bivariate correlations measured the associations between estimates of location and scale and measures of self-efficacy and automaticity for each behavior. To account for possible interactions between location and scale in their associations with either self-efficacy or automaticity, we regressed self-efficacy and automaticity for a given behavior onto (1) location, (2) scale, and (3) the interaction between location and scale for that behavior. If the interaction term was significant, we planned to probe it with post-estimation analyses.

Results

Bivariate correlations between self-reported self-efficacy and automaticity for SED (r = .57, p < .001), FV (r = .59, p < .001), and MVPA (r = .48, p < .001) revealed significant and positive associations with comparably large effect sizes, indicating considerable shared variance between self-efficacy and automaticity for each behavior.

For SED, higher scale (more variability, less consistency) was significantly associated with both lower self-efficacy (r = −.22, p = .014) and lower automaticity (r = −.23, p = .013). However, location (amount or level) of SED was not associated with either self-efficacy (r = −.08, p = .409) or automaticity (r = −.15, p = .096). Further, there was no interaction between location and scale in predicting either self-efficacy (Estimate = .17, t = 1.28, p = .20) or automaticity (Estimate = .02, t = 0.08, p = .93). This lack of an interaction indicates that, for SED, the association between greater temporal variability (i.e., scale) of SED with lower self-efficacy and lower automaticity did not vary based on the total quantity (i.e., location) of sedentary behavior.

For FV and MVPA, scale was not associated with either self-efficacy (FV, r = .06, p = .547; MVPA, r = .02, p = .777) or automaticity (FV, r = .12, p = .200; MVPA, r = −.10, p = .286). On the other hand, for both FV and MVPA, greater location (level) was significantly associated with higher self-efficacy (FV, r = .38, p < .001; MVPA, r = .34, p < .001) and greater automaticity (FV, r = .46, p < .001; MVPA, r = .42, p < .001). Again, there was no evidence of an interaction between location and scale in predicting self-efficacy (FV, p = .737; MVPA, p = .820) or automaticity (FV, p = .402; MVPA, p = .878). This lack of interactions indicates that the positive relationship between the amount (i.e., location) of each behavior and increased self-efficacy and automaticity for the same behavior, did not vary based on how consistently the behaviors were enacted.

Discussion

These results from the MBC2 trial indicate that self-regulatory strength and habit strength are strongly positively correlated for the complex diet and physical activity changes we studied. The findings suggest that the acquisition of complex diet and physical activity behaviors probably rests on both good self-regulatory skills and strong habits. Notably, specific diet and physical activity risk behaviors differed in whether it was their improved level or their improved consistency as a result of behavioral intervention that correlated with greater self-efficacy and automaticity. For both healthy behaviors that occurred at a low level initially (FV, MVPA), it was progressing to perform a greater amount of the behavior that correlated with greater feelings of self-efficacy and automaticity. In contrast, for the unhealthy behavior, sedentary leisure screen time, that occurred at too high a level initially, it was improving the consistency of time spent on the behavior across days that correlated with increased self-efficacy and automaticity. We note that because this measure of consistency was taken at the end of the intervention period, when behavioral improvements had largely been accomplished, low scale meant being consistent across days in practicing a lower level of the behavior (which was mostly television viewing) than had been practiced at baseline.

Conclusions

The overarching goal of this research is to optimize a behavioral intervention to improve multiple diet and physical activity risk behaviors in a sustainable, resource-efficient way. This manuscript illustrates the application of Phase 1 of the ORBIT model with the goal of understanding and refining the important intervention targets and outcomes in a multiple diet and physical activity intervention.

In Orbit Phase 1a, we sought statistical analysis techniques that could operationalize and quantify improvement in behavioral frequency and consistency using dense digital data obtained from mobile tools. Having identified mixed location scale modeling as a method of intensive longitudinal analysis that meets our needs, we conducted secondary analyses of data from a previously published study. In the MBC1 study, participants utilized an app, telephone coaching, and incentives to achieve targeted goals for pairs of four unhealthy diet and physical activity risk behaviors, all of which characterized them at the study’s outset. As hypothesized, results showed that targeting each specific behavior via behavioral intervention improved its frequency and consistency over time to a greater extent than behaviors not targeted by the intervention. The findings suggest that the induction of improved behavioral location as well as decreased scale (increased consistency) hold promise as a dynamic behavioral phenotype for diet and physical activity improvement. Particularly if future research shows that improved location and decreased scale reflect a pattern of behavioral acquisition that has potential to be sustainable, then the consistency of a behavior as well as its level may both become intervention targets.

In Orbit Phase 1b, we consulted prior research literature to identify candidate treatment pathways through which behavioral intervention could produce potentially sustainable improvements in diet and physical activity behaviors. We identified relevant constructs from two bodies of psychological science: one characterizing self-regulation and one characterizing habit acquisition. Habitual responses triggered by a stable contextual cue have the advantage of proceeding automatically without effort, freeing up mental capacity to devote to more effortful deliberative processing. Self-regulation requires more burdensome, effortful processing, but offers the advantage of enabling complex diet and activity patterns behaviors to be executed appropriately across variable contexts. Both self-regulatory skills and habit strength tended to covary for each improved behavior, suggesting that both may undergird successful performance. Interestingly it was an increase in the level of behaviors that began at a low rate that correlated with greater feelings of self-efficacy and automaticity about the behavior, as if the challenge of initiating change in these behaviors primarily involved building up a large stimulus-response repertoire by which to increase the behavior’s frequency. On the other hand, it was the consistency of suppressing an unhealthy behavior that covaried with feelings of self-efficacy and automaticity about it, as if relentless inhibitory control was the pathway to success in improving such behaviors.

A novel contribution of this work has been to introduce the concept of behaviorally consistent improvement as a potential dynamic operational definition of the acquisition of a complex behavioral pattern. An understanding of how treatment influences behavioral dynamics would have implications for how interventions could be designed and delivered in ways that minimize treatment burden and cost while still maximizing sustained healthy behavior change. For example, the operational definition of behaviorally consistent improvement could serve as a tailoring variable to guide the design of adaptive treatment algorithms (e.g. that remove intervention components for those who exhibit sustained improvement or add them for those who fail to attain consistent behavioral improvement; Almirall, Nahum-Shani, Sherwood, & Murphy, 2014)

We envision a day when algorithms using a rich array of behavioral features, many inferred from digital sensing, will be able to track behavior change and adaptively guide intervention. These developments will be made possible by applying new statistical analysis techniques and machine learning to measure behavioral dynamics in new ways. For example, in addition to measuring an individual’s level of a behavior when an intervention starts and its slope of change over time, we can also statistically analyze how consistently the person displays the behavior at the start and end of the intervention and their rate of change in this variability over time. We can cluster these features in a way that accurately characterizes people who will go on to maintain a healthy behavioral change post-intervention, versus those who will not. This allows us to apply machine learning methods to derive additional features such as specific ranges of the behavior’s level and consistency during the intervention that characterize future maintainers. We can even use the ranges to extract additional features that might predict post-treatment behavioral maintenance, such as the time that it took to first reach the target ranges of mean behavioral level and consistency and the duration of time during the intervention that their behavior remained within the target range. An algorithm using this rich array of features might signal, during or at the end of an intervention, which people have improved a targeted behavior in a manner that is durable and unlikely to regress to prior levels following removal of treatment. A similar dynamic treatment algorithm might support time-varying dosing decisions based on the real-time detection of behavioral patterning that predicts progression toward durable healthy change. Such advances may allow interventions to become more resource-efficient and scalable because we can tell when an individual has achieved a sustainable pattern of health behavior improvement that allows intervention to be tapered, and then stopped, with low risk of relapse.

The current work has strengths in introducing a novel application of location scale mixed modeling to quantify the acquisition of consistent behavioral improvements that have the potential to be sustainable. Use of this method was made feasible by the re-purposing and de-aggregation of dense digital data collected from mobile health tools, which we had initially averaged over week-long intervals to analyze intervention outcomes. In addition to being resource-efficient, this strategy has enabled us to extract new information about behavioral dynamics that was invisible before de-aggregating and analyzing the data via location scale methods.

The current work also has limitations, chief among which is the absence of evidence about whether behaviorally consistent improvement in diet and activity behaviors bears an association with maintenance of behavioral improvements after intervention ends. No data on behavioral maintenance were presented here because our Phase 1 activities concern identifying and understanding the dynamic phenotype(s) of behavioral acquisition. Our next step will be to examine whether the features that define the successful acquisition of multiple diet and activity change also define and predict its maintenance. A second limitation is that our assessments of diet and physical activity behaviors were based primarily upon self-report. It can be noted that previous studies support the validity of self-report diet and activity assessments administered via a mobile device (Beasley et al., 2005; Clark et al., 2009; Sternfeld et al., 2012). Also, to further encourage honest reporting, in MBC1 we also implemented a bogus pipeline protocol whereby participants submitted grocery receipts, accelerometer data, and urine samples that they believed would be used to evaluate their self-reports (Roese & Jamieson, 1993). Still, replicating this work using objective sensor data to detect and monitor all behaviors will be an important step forward.

In sum, this work illustrates the use of the ORBIT model as a supportive framework to guide early phase basic research needed to prepare to optimize an existing behavioral intervention. The goal of the intervention will be to initiate and, ultimately maintain, the improvement of multiple prevalent chronic disease diet and physical activity risk behaviors in a resource-efficient manner. The promise of this work lies partly in the potential that it may lead toward greater basic science understanding of dynamic mechanisms that underlie the acquisition of needed complex healthy lifestyle patterns that hold up against the challenge of changing contexts. It also lies in the practical prospect of being able to design adaptive data-based treatment algorithms to guide sound decisions about the magnitude and timing of intervention dosing. To fulfill both objectives, the next phase of this research program will incorporate optimization designs (cf.,Collins et al, 2007; Pfammatter, Nahum-Shani, DeZelar, e al, 2019; Spring et al, 2020) that evaluate how to balance intervention effectiveness with resource efficiency.

Funding:

This study was funded by National Heart, Lung, and Blood grant HL075451 and by National Institute of Diabetes and Digestive and Kidney Diseases grant DK108678 to Dr. Spring. The work was also supported, in part, by National Cancer Institute grant T32 CA193193 (PI: Dr. Spring, providing salary support for Dr. Stump).

References

  1. Aarts H, & Dijksterhuis A (2000). Habits as knowledge structures: automaticity in goal-directed behavior’. Journal of personality and social psychology, 78(1), 53–63. [DOI] [PubMed] [Google Scholar]
  2. Abraham C, & Michie S (2008). A taxonomy of behavior change techniques used in interventions. Health psychology, 27(3), 379–387. [DOI] [PubMed] [Google Scholar]
  3. Adams ML, Grandpre J, Katz DL, & Shenson D (2017). Linear association between number of modifiable risk factors and multiple chronic conditions: Results from the Behavioral Risk Factor Surveillance System. Preventive medicine, 105, 169–175. [DOI] [PubMed] [Google Scholar]
  4. Almirall D, Nahum-Shani I, Sherwood NE, & Murphy SA (2014). Introduction to SMART designs for the development of adaptive interventions: with application to weight loss research. Translational behavioral medicine, 4(3), 260–274. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. American Diabetes Association. (2020). 15. Diabetes Care in the Hospital: Standards of Medical Care in Diabetes—2020. Diabetes Care, 43(Supplement 1), S193–S202. [DOI] [PubMed] [Google Scholar]
  6. American Heart Association. (2015). Saturated Fat. Retrieved from https://www.heart.org/en/healthy-living/healthy-eating/eat-smart/fats/saturated-fats.
  7. Arena R, Guazzi M, Lianov L, Whitsel L, Berra K, Lavie CJ, … Shurney D (2015). Healthy Lifestyle Interventions to Combat Noncommunicable Disease-A Novel Nonhierarchical Connectivity Model for Key Stakeholders: A Policy Statement From the American Heart Association, European Society of Cardiology, European Association for Cardiovascular Prevention and Rehabilitation, and American College of Preventive Medicine. Mayo Clin Proc, 90(8), 1082–1103. doi: 10.1016/j.mayocp.2015.05.001 [DOI] [PubMed] [Google Scholar]
  8. Arterburn D, Westbrook EO, Wiese CJ, Ludman EJ, Grossman DC, Fishman PA, … Drewnowski A (2008). Insurance coverage and incentives for weight loss among adults with metabolic syndrome. Obesity, 16(1), 70–76. [DOI] [PubMed] [Google Scholar]
  9. Bandura A (1985). Model of causality in social learning theory. In Cognition and psychotherapy (pp. 81–99). Springer, Boston, MA. [Google Scholar]
  10. Bandura A (1991). Social cognitive theory of self-regulation. Organizational behavior and human decision processes, 50(2), 248–287. [Google Scholar]
  11. Baruth M, Wilcox S, Egan BM, Dowda M, Laken M, & Warren TY (2011). Cardiovascular disease risk factor clustering among African American adults. Ethnicity & disease, 21(2), 129–134. [PMC free article] [PubMed] [Google Scholar]
  12. Bauer UE, Briss PA, Goodman RA, & Bowman BA (2014). Prevention of chronic disease in the 21st century: elimination of the leading preventable causes of premature death and disability in the USA. Lancet, 384(9937), 45–52. doi: 10.1016/S0140-6736(14)60648-6 [DOI] [PubMed] [Google Scholar]
  13. Baumeister RF, Heatherton TF, & Tice DM (1994). Losing control: How and why people fail at self-regulation. San Diego, CA: Academic Press. [Google Scholar]
  14. Baumeister RF, Vohs KD, & Tice DM (2007). The strength model of self-control. Current directions in psychological science, 16(6), 351–355. [Google Scholar]
  15. Beasley J, Riley WT, & Jean-Mary J (2005). Accuracy of a PDA-based dietary assessment program. Nutrition, 21(6), 672–677. [DOI] [PubMed] [Google Scholar]
  16. Becker WC, Dorflinger L, Edmond SN, Islam L, Heapy AA, & Fraenkel L (2017). Barriers and facilitators to use of non-pharmacological treatments in chronic pain. BMC family practice, 18(1), 41–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Bellg AJ, Borrelli B, Resnick B, Hecht J, Minicucci DS, Ory M, … Czajkowski S (2004). Enhancing treatment fidelity in health behavior change studies: best practices and recommendations from the NIH Behavior Change Consortium. Health Psychology, 23(5), 443–451. [DOI] [PubMed] [Google Scholar]
  18. Bickel WK, & Vuchinich RE (2000). Reframing Health Behavior Change with Behavioral Economics: New York: Psychology Press. [Google Scholar]
  19. Borrelli B, Sepinwall D, Ernst D, Bellg AJ, Czajkowski S, Breger R, … Ogedegbe G (2005). A new tool to assess treatment fidelity and evaluation of treatment fidelity across 10 years of health behavior research. Journal of consulting and clinical psychology, 73(5), 852–860. [DOI] [PubMed] [Google Scholar]
  20. Carver C, & Scheier M (1982). Control theory: A useful conceptual framework for personality-social, clinical, and health psychology. Psychological Bulletin, 92(1), 111–135. [PubMed] [Google Scholar]
  21. Centers for Disease Control and Prevention. (2018). Early Release of Selected Estimates Based on Data From the 2017 National Health Interview Survey.
  22. Centers for Disease Control and Prevention. (2019). Chronic Diseases in America. Retrieved from https://www.cdc.gov/chronicdisease/resources/infographic/chronic-diseases.htm
  23. Chou KL (2008). The prevalence and clustering of four major lifestyle risk factors in Hong Kong Chinese older adults. Journal of aging and health, 20(7), 788–803. [DOI] [PubMed] [Google Scholar]
  24. Clark BK, Sugiyama T, Healy GN, Salmon J, Dunstan DW, & Owen N (2009). Validity and reliability of measures of television viewing time and other non-occupational sedentary behaviour of adults: a review. Obesity reviews, 10(1), 7–16. [DOI] [PubMed] [Google Scholar]
  25. Collins LM, Murphy SA, & Strecher V (2007). The multiphase optimization strategy (MOST) and the sequential multiple assignment randomized trial (SMART): new methods for more potent eHealth interventions. American journal of preventive medicine, 32(5), S112–S118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Cronbach L and Meehl P (1955). Construct validity in psychological tests, Psychological Bulletin, 52, 4, 281–302 [DOI] [PubMed] [Google Scholar]
  27. Curry SJ, Grossman DC, Whitlock EP, & Cantu A (2014). Behavioral counseling research and evidence-based practice recommendations: U.S. Preventive Services Task Force perspectives. Ann Intern Med, 160(6), 407–413. doi: 10.7326/M13-2128 [DOI] [PubMed] [Google Scholar]
  28. Curry SJ, Krist AH, Owens DK, Barry MJ, Caughey AB, Davidson KW, … & Kubik M (2018). Behavioral weight loss interventions to prevent obesity-related morbidity and mortality in adults: US preventive services task force recommendation statement. Jama, 320(11), 1163–1171. [DOI] [PubMed] [Google Scholar]
  29. Czajkowski SM, Powell LH, Adler N, Naar-King S, Reynolds KD, … Epel E (2015). From ideas to efficacy: The ORBIT model for developing behavioral treatments for chronic diseases. Health Psychology, 34, 971–981. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Darker CD, French DP, Eves FF, & Sniehotta FF (2010). An intervention to promote walking amongst the general population based on an ‘extended’ theory of planned behaviour: a waiting list randomised controlled trial. Psychology and Health, 25(1), 71–88 [DOI] [PubMed] [Google Scholar]
  31. Dunton GF, & Atienza AA (2009). The need for time-intensive information in healthful eating and physical activity research: a timely topic. Journal of the Academy of Nutrition and Dietetics, 109(1), 30–35. [DOI] [PubMed] [Google Scholar]
  32. Dunton GF (2018). Sustaining health-protective behaviors such as physical activity and healthy eating. Jama, 320(7), 639–640. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Fedewa SA, Sauer AG, Siegel RL, & Jemal A (2015). Prevalence of major risk factors and use of screening tests for cancer in the United States. Cancer Epidemiology and Prevention Biomarkers, 24(4), 637–652. [DOI] [PubMed] [Google Scholar]
  34. Gardner B, Abraham C, Lally P, & de Bruijn GJ (2012). Towards parsimony in habit measurement: Testing the convergent and predictive validity of an automaticity subscale of the Self-Report Habit Index. International Journal of Behavioral Nutrition and Physical Activity, 9(1), 102–114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Gardner B (2015). A review and analysis of the use of ‘habit’ in understanding, predicting and influencing health-related behaviour. Health psychology review, 9(3), 277–295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Gardner B, & Lally P (2018). Modelling habit formation and its determinants. In The Psychology of Habit (pp. 207–229): Cham, Switzerland: Springer. [Google Scholar]
  37. Gerbaud E, Darier R, Montaudon M, Beauvieux M-C, Coffin-Boutreux C, Coste P, Douard H, Ouattara A, Catargi B (2019). Glycemic Variability Is a Powerful Independent Predictive Factor of Midterm Major Adverse Cardiac Events in Patients With Diabetes With Acute Coronary Syndrome. Diabetes Care. 42(4), 674–681. [DOI] [PubMed] [Google Scholar]
  38. Hagger MS, & Chatzisarantis NL (2016). The trans-contextual model of autonomous motivation in education: Conceptual and empirical issues and meta-analysis. Review of educational research, 86(2), 360–407. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Hedeker D, Mermelstein RJ, & Demirtas H (2008). An application of a mixed-effects location scale model for analysis of Ecological Momentary Assessment (EMA) data. Biometrics, 64, 627–634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Hedeker D, Mermelstein RJ, & Demirtas H (2012). Modeling between- and within-subject variance in Ecological Momentary Assessment (EMA) data using mixed-effects location scale models. Statistics in Medicine, 31, 3328–3336. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Hedeker D, & Nordgren R (2013). MIXREGLS: A program for mixed-effects location scale analysis. Journal of statistical software, 52(12), 1–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Jensen CD, Aylward BS, & Steele RG (2012). Predictors of attendance in a practical clinical trial of two pediatric weight management interventions. Obesity, 20(11), 2250–2256. [DOI] [PubMed] [Google Scholar]
  43. Jones E, Zur J, Rosenbaum S, & Ku L (2015). Opting out of Medicaid expansion: Impact on encounters with behavioral health specialty staff in community health centers. Psychiatric Services, 66(12), 1277–1282. [DOI] [PubMed] [Google Scholar]
  44. Kahneman D (2011). Thinking, fast and slow. New York, NY, US, Farrar, Straus and Giroux. [Google Scholar]
  45. Kanfer FH and Goldstein AP (1991). Helping people change: A textbook of methods, 4th ed. Elmsford, NY, US, Pergamon Press. [Google Scholar]
  46. Kirchner TR, Shiffman S, & Wileyto EP (2012). Relapse dynamics during smoking cessation: recurrent abstinence violation effects and lapse-relapse progression. Journal of Abnormal Psychology, 121(1), 187–197. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Knowlton BJ, Patterson TK (2016) Habit Formation and the Striatum. In Behavioral Neuroscience of Learning and Memory (pp. 275–295): Cham, Switzerland: Springer. [DOI] [PubMed] [Google Scholar]
  48. Kwasnicka D, Dombrowski SU, White M, & Sniehotta F (2016). Theoretical explanations for maintenance of behaviour change: a systematic review of behaviour theories. Health psychology review, 10(3), 277–296. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Lally P, Chipperfield A, & Wardle J (2008). Healthy habits: efficacy of simple advice on weight control based on a habit-formation model. International journal of obesity, 32(4), 700–707. [DOI] [PubMed] [Google Scholar]
  50. Lally P, Van Jaarsveld CH, Potts HW, & Wardle J (2010). How are habits formed: Modelling habit formation in the real world. European journal of social psychology, 40(6), 998–1009. [Google Scholar]
  51. Lally P, & Gardner B (2013). Promoting habit formation. Health Psychology Review, 7(sup1), S137–S158. [Google Scholar]
  52. Larimer ME, Palmer RS, & Marlatt GA (1999). Relapse prevention: An overview of Marlatt’s cognitive-behavioral model. Alcohol Research & Health, 23(2), 151–160. [PMC free article] [PubMed] [Google Scholar]
  53. Lee-Kwan SH, Moore LV, Blanck HM, Harris DM, & Galuska D (2017). Disparities in State-Specific Adult Fruit and Vegetable Consumption - United States, 2015. MMWR Morb Mortal Wkly Rep, 66(45), 1241–1247. doi: 10.15585/mmwr.mm6645a1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Lichstein KL, Riedel BW, & Grieve R (1994). Fair tests of clinical trials: A treatment implementation model. Advances in Behaviour Research and Therapy, 16(1), 1–29. [Google Scholar]
  55. Lloyd-Jones DM, Hong Y, Labarthe D, Mozaffarian D, Appel LJ, Van Horn L, … & Arnett DK (2010). Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association’s strategic Impact Goal through 2020 and beyond. Circulation, 121(4), 586–613. [DOI] [PubMed] [Google Scholar]
  56. Marcus BH, Selby VC, Niaura RS, & Rossi JS (1992). Self-efficacy and the stages of exercise behavior change. Research quarterly for exercise and sport, 63(1), 60–66. [DOI] [PubMed] [Google Scholar]
  57. Meader N, King K, Wright K, Graham HM, Petticrew M, Power C, … Sowden AJ (2017). Multiple risk behavior interventions: meta-analyses of RCTs. American journal of preventive medicine, 53(1), e19–e30. [DOI] [PubMed] [Google Scholar]
  58. Mendis S, Davis S, & Norrving B (2015). Organizational update: the world health organization global status report on noncommunicable diseases 2014; one more landmark step in the combat against stroke and vascular disease. Stroke, 46(5), e121–122. doi: 10.1161/STROKEAHA.115.008097 [DOI] [PubMed] [Google Scholar]
  59. Michie S, Abraham C, Whittington C, McAteer J, & Gupta S (2009). Effective techniques in healthy eating and physical activity interventions: a meta-regression. Health Psychology, 28(6), 690–701. [DOI] [PubMed] [Google Scholar]
  60. Michie S, Ashford S, Sniehotta FF, Dombrowski SU, Bishop A, & French DP (2011). A refined taxonomy of behaviour change techniques to help people change their physical activity and healthy eating behaviours: the CALO-RE taxonomy. Psychology & health, 26(11), 1479–1498. [DOI] [PubMed] [Google Scholar]
  61. Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W, … & Wood CE (2013). The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Annals of behavioral medicine, 46(1), 81–95. [DOI] [PubMed] [Google Scholar]
  62. Moncher FJ, & Prinz RJ (1991). Treatment fidelity in outcome studies. Clinical psychology review, 11(3), 247–266. [Google Scholar]
  63. Mozaffarian D, Benjamin EJ, Go AS, Arnett DK, Blaha MJ, Cushman M, … & Huffman MD (2015). Executive summary: heart disease and stroke statistics—2015 update: a report from the American Heart Association. Circulation, 131(4), 434–441. [DOI] [PubMed] [Google Scholar]
  64. Mullan B, Novoradovskaya E (2018). Habit Mechanisms and Behavioural Complexity. In The Psychology of Habit (pp. 71–90): Cham, Switzerland: Springer. [Google Scholar]
  65. Murphy SA (2005). An experimental design for the development of adaptive treatment strategies. Statistics in medicine, 24(10), 1455–1481. [DOI] [PubMed] [Google Scholar]
  66. Myint PK, Luben RN, Wareham NJ, Bingham SA, & Khaw KT (2009). Combined effect of health behaviours and risk of first ever stroke in 20,040 men and women over 11 years’ follow-up in Norfolk cohort of European Prospective Investigation of Cancer (EPIC Norfolk): prospective population study. BMJ, 338, b349. doi: 10.1136/bmj.b349 [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Neal DT, Wood W, & Drolet A (2013). How do people adhere to goals when willpower is low? The profits (and pitfalls) of strong habits. Journal of Personality and Social Psychology, 104(6), 959–975. [DOI] [PubMed] [Google Scholar]
  68. Perri MG (1998). The maintenance of treatment effects in the long-term management of obesity. Clinical Psychology: Science and Practice, 5(4), 526–543. [Google Scholar]
  69. Pfammatter AF, Nahum-Shani I, DeZelar M, Scanlan L,H. McFadden HG, Siddique J, Hedeker D, Spring B (2019). SMART: Study Protocol for a Sequential Multiple Assignment Randomized Controlled Trial to Optimize Weight Loss Management. Contemporary Clinical Trials, (82):36–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Phillips LA (2019). “Defining Habit: Advantages and Disadvantages of Using Time Stability to Assess Habit Strength.” Annals of Behavioral Medicine, 53(S1). [Google Scholar]
  71. Piercy KL, Troiano RP, Ballard RM, Carlson SA, Fulton JE, Galuska DA, … & Olson RD (2018). The physical activity guidelines for Americans. Journal of the American Medical Association, 320(19), 2020–2028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Riley WT, Rivera DE, Atienza AA, Nilsen W, Allison SM, & Mermelstein R (2011). Health behavior models in the age of mobile interventions: are our theories up to the task?. Translational behavioral medicine, 1(1), 53–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Rhodes RE, & Rebar AL (2018). Physical activity habit: Complexities and controversies. In The psychology of habit (pp. 91–109): Cham, Switzerland: Springer. [Google Scholar]
  74. Roese NJ, & Jamieson DW (1993). Twenty years of bogus pipeline research: A critical review and meta-analysis. Psychological bulletin, 114(2), 363–375. [Google Scholar]
  75. Sacks FM, Lichtenstein AH, Wu JH, Appel LJ, Creager MA, Kris-Etherton PM, … & Stone NJ (2017). Dietary fats and cardiovascular disease: a presidential advisory from the American Heart Association. Circulation, 136(3), e1–e23. [DOI] [PubMed] [Google Scholar]
  76. Schneider KL, Coons MJ, McFadden H, Pellegrini CA, DeMott A, Siddique J, … & Spring B (2016). Mechanisms of change in diet and activity in the Make Better Choices 1 trial. Health Psychology, 35(7), 723–732 Epub 2016/04/08. [DOI] [PMC free article] [PubMed] [Google Scholar]
  77. Schneider W, & Shiffrin RM (1977). Controlled and automatic human information processing: I. Detection, search, and attention. Psychological review, 84(1), 1–66. [Google Scholar]
  78. Schuit AJ, van Loon AJM, Tijhuis M, & Ocké MC (2002). Clustering of lifestyle risk factors in a general adult population. Preventive medicine, 35(3), 219–224. [DOI] [PubMed] [Google Scholar]
  79. Schulz DN, Kremers SP, Vandelanotte C, Van Adrichem MJ, Schneider F, Candel MJ, & de Vries H (2014). Effects of a web-based tailored multiple-lifestyle intervention for adults: a two-year randomized controlled trial comparing sequential and simultaneous delivery modes. Journal of medical Internet research, 16(1), e26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  80. Shadish WR, Cook TD, & Campbell DT (2002). Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton Mifflin. [Google Scholar]
  81. Shiffrin RM, & Schneider W (1977). Controlled and automatic human information processing: II. Perceptual learning, automatic attending and a general theory. Psychological review, 84(2), 127–190. [Google Scholar]
  82. Skinner BF (1953). Science and human behavior. Oxford: Macmillan. [Google Scholar]
  83. Spring B, Doran N, Pagoto S, Schneider K, Pingitore R, & Hedeker D (2004). Randomized controlled trial for behavioral smoking and weight control treatment: effect of concurrent versus sequential intervention. Journal of consulting and clinical psychology, 72(5), 785–796. [DOI] [PMC free article] [PubMed] [Google Scholar]
  84. Spring B, Schneider K, McFadden HG, Vaughn J, Kozak AT, Smith M, … & Hedeker D (2010). Make Better Choices (MBC): study design of a randomized controlled trial testing optimal technology-supported change in multiple diet and physical activity risk behaviors. BMC public health, 10(1), 586–597. [DOI] [PMC free article] [PubMed] [Google Scholar]
  85. Spring B, Schneider K, McFadden HG, Vaughn J, Kozak AT, Smith M, … Lloyd-Jones DM (2012). Multiple behavior changes in diet and activity: a randomized controlled trial using mobile technology. Arch Intern Med, 172(10), 789–796. doi: 10.1001/archinternmed.2012.1044 [DOI] [PMC free article] [PubMed] [Google Scholar]
  86. Spring B, Ockene JK, Gidding SS, Mozaffarian D, Moore S, Rosal MC, … Stroke N (2013). Better population health through behavior change in adults: a call to action. Circulation, 128(19), 2169–2176. doi: 10.1161/01.cir.0000435173.25936.e1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  87. Spring B, Moller AC, Colangelo LA, Siddique J, Roehrig M, Daviglus ML, … & Liu K (2014). Healthy lifestyle change and subclinical atherosclerosis in young adults: Coronary Artery Risk Development in Young Adults (CARDIA) study. Circulation, 130(1), 10–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  88. Spring B, Pellegrini C, McFadden HG, Pfammatter AF, Stump TK, Siddique J, … & Hedeker D (2018). Multicomponent mHealth intervention for large, sustained change in multiple diet and activity risk behaviors: the make better choices 2 randomized controlled trial. Journal of medical Internet research, 20(6), e10528, 1–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  89. Spring B (2019). Sound health care economics: Provide the treatment needed (not less, not more). Health Psychology, 38(8), 701–704. [DOI] [PubMed] [Google Scholar]
  90. Spring B, Pfammatter A, Hoffman SA, McFadden G,H, Hedeker D, Stump TK, & Collins LM (2020). A factorial experiment to optimize remotely delivered behavioral treatment for obesity: Results of the Opt-IN study. Obesity,. [DOI] [PMC free article] [PubMed] [Google Scholar]
  91. Sternfeld B, Jiang S-F, Picchi T, Chasan-Taber L, Ainsworth B, & Quesenberry JC (2012). Evaluation of a cell phone-based physical activity diary. Medicine and science in sports and exercise, 44(3), 487–495. [DOI] [PubMed] [Google Scholar]
  92. Trafimow D (2018). The automaticity of habitual behaviours: Inconvenient questions. In The psychology of habit (pp. 379–395): Cham, Switzerland: Springer. [Google Scholar]
  93. Tricomi E, Balleine BW, & O’Doherty JP (2009). A specific role for posterior dorsolateral striatum in human habit learning. European Journal of Neuroscience, 29(11), 2225–2232. [DOI] [PMC free article] [PubMed] [Google Scholar]
  94. U.S. Department of Health and Human Services and U.S. Department of Agriculture. (2010). Dietary Guidelines for Americans. 8th ed. Washington, DC: US Dept of Health and Human Services; US Dept of Agriculture; 2010. [Google Scholar]
  95. U.S. Department of Health and Human Services and U.S. Department of Agriculture. (2015). 2015 – 2020 Dietary Guidelines for Americans- 8th Edition. Retrieved from https://health.gov/dietaryguidelines/2015/guidelines/ [Google Scholar]
  96. U.S. Department of Heath and Human Services. (2018). Physical Activity Guidelines for Americans .2nd ed. Washington, DC: US Dept of Health and Human Services. [Google Scholar]
  97. Verplanken B, & Orbell S (2003). Reflections on past behavior: a self-report index of habit strength. Journal of applied social psychology, 33(6), 1313–1330. [Google Scholar]
  98. Watson JB (1925). Behaviorism. New York: The People’s Institute. [Google Scholar]
  99. Wood W (2017). Habit in personality and social psychology. Personality and Social Psychology Review, 21(4), 389–403. [DOI] [PubMed] [Google Scholar]
  100. Yin HH, & Knowlton BJ (2006). The role of the basal ganglia in habit formation. Nature Reviews Neuroscience, 7(6), 464–476. [DOI] [PubMed] [Google Scholar]

RESOURCES