Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Sep 1.
Published in final edited form as: Psychopharmacology (Berl). 2017 Jun 8;234(17):2631–2642. doi: 10.1007/s00213-017-4663-0

Context and Craving During Stressful Events in the Daily Lives of Drug-dependent Patients

Kenzie L Preston 1,*, William J Kowalczyk 1, Karran A Phillips 1, Michelle L Jobes 1, Massoud Vahabzadeh 2, Jia-Ling Lin 2, Mustapha Mezghanni 2, David H Epstein 1
PMCID: PMC5709189  NIHMSID: NIHMS883197  PMID: 28593441

Abstract

Rationale

Knowing how stress manifests in the lives of people with substance-use disorders could help inform mobile “just in time” treatment.

Objectives

To examine discrete episodes of stress, as distinct from the fluctuations in background stress assessed in most EMA studies.

Methods

For up to 16 weeks, outpatients on opioid-agonist treatment carried smartphones on which they initiated an entry whenever they experienced a stressful event (SE) and when randomly prompted (RP) three times daily. Participants reported the severity of stress and craving and the context of the report (location, activities, companions). Decomposition of covariance was used to separate within-person from between-person effects; reffect sizes below are within-person.

Results

Participants (158 of 182; 87%) made 1,787 stress-event entries. Craving for opioids increased with stress severity (reffect = 0.50). Stress events tended to occur in social company (with acquaintances, 0.63, friends, 0.17, or on the phone, 0.41) rather than with family (spouse, −0.14; child, −0.18), and in places with more overall activity (bars, 0.32; outside, 0.28; walking, 0.28) and more likelihood of unexpected experiences (with strangers, 0.17). Being on the internet was slightly protective (−0.22). Our prior finding that being at the workplace protects against background stress in our participants was partly supported in these stressful-event data.

Conclusions

The contexts of specific stressful events differ from those we have seen in prior studies of ongoing background stress. However, both are associated with drug craving.

Keywords: cocaine, heroin, ecological momentary assessment, stress, craving

The commonly held assumption that stress plays a causal role in addiction and relapse has been given biological plausibility by consistent findings of overlap in the neural circuitry affected by stress and drug use (Cleck and Blendy 2008) and by lab-animal studies showing that acute stressors such as footshock increase the likelihood of drug-seeking behavior (Bossert et al. 2013; Cleck and Blendy 2008). In humans with addiction, stress responses in a laboratory predict susceptibility to relapse (Sinha et al. 2006).

The idea of stress-induced relapse has led to treatments intended to mitigate the effects of stress. For example, laboratory studies showing that alpha-2 adrenergic agonists decrease stress-induced reinstatement of drug seeking in rats (Erb et al. 2000; Highfield et al. 2001) and stress-induced drug craving in humans (Fox et al. 2012; Jobes et al. 2011) led our research clinic to conduct a randomized clinical trial that found that clonidine increased time to lapse and longest duration of abstinence during treatment for heroin addiction (Kowalczyk et al. 2015). Most evidence-based psychosocial treatments for addiction, such as cognitive-behavioral therapy (Carroll 1998) and Acceptance and Commitment Therapy (ACT) (Hayes et al. 2013)(Brewer et al. 2009; Chiesa and Serretti 2014; Hoppes 2006; Luoma et al. 2012), incorporate the teaching of techniques for responding adaptively to stress.

These kinds of treatments, which have until now been administered only in a clinician’s office during scheduled visits, or sometimes by desktop computer (Carroll et al. 2014) are becoming mobile— adapted for delivery when and where they are needed (Lindhiem et al. 2015; Nahum-Shani et al. 2016). Our knowledge base has not caught up to this change (Hekler et al. 2016; Riley et al. 2011). Much of what we know about addiction and stress is based on retrospective studies in which weeks or months intervene between the event and its recollection, leaving ample time for recall bias or the formation of post hoc explanations for behavior. These problems are now being surmounted by Ecological Momentary Assessment (EMA) (Shiffman et al. 2008), in which people report their moods and behaviors in real time on mobile devices. The EMA data we collected during our clonidine trial (Kowalczyk et al. 2015) provided mechanistic insight we could not otherwise have had, showing that clonidine seemed to act by decoupling stress from craving.

In our clonidine trial, we assessed only changes in “background” levels of stress—meaning levels of stress as a tonic state, rated in randomly prompted reports throughout the day. This is how stress was assessed in all published EMA studies we know of, until our group began to incorporate assessment of participant-defined episodes of stress. We ask participants to initiate an EMA report of every stressful event they undergo, as soon as it happens, while we continue to assess levels of background stress in randomly prompted reports (Furnari et al. 2015). This helps ensure that we do not miss the instances of peak stress. In this paper, we use those data to examine patterns of stressful events in the lives of people with opioid-use disorders during up to 16 weeks of agonist maintenance. Their data enabled us to examine drug craving in the context of discrete episodes of stress as well as fluctuations in randomly assessed background stress.

Patients and Methods

Participants

The participants were methadone- or buprenorphine-treated opioid-dependent patients at a treatment-research clinic in Baltimore, MD. All were participating in a 46-week natural-history study of stress, geographical location, and drug use. Our prior publications of data from a subset of participants in this study focused primarily on ambulatory physiological monitoring (Hossain et al. 2014; Kennedy et al. 2015; Rahman et al. 2014) and an examination of the relationship between stress and drug use (Furnari et al. 2015).

During screening, participants completed the Addiction Severity Index (ASI; McLellan et al. 1985) and the Diagnostic Interview Schedule (DIS IV; Robins et al. 1995) and were given physical examinations and psychological testing. The main inclusion criteria were: age 18 to 75 years, physical dependence on opioids, and (due to the behavioral-geography focus of the parent study) residence in Baltimore City or one of the surrounding counties. The main exclusion criteria were: history of any DSM-IV psychotic disorder, bipolar disorder, or current Major Depressive Disorder; current dependence on alcohol or sedative-hypnotics; cognitive impairment precluding informed consent or valid self-report; conditions that preclude urine collection; or medical illness or medications that would compromise research participation. This study was reviewed and approved by the National Institutes of Health Addictions Institutional Review Board. Participants gave prior written informed consent and were paid for completing the research components of this study.

Procedure

Participants attended clinic five to seven days a week; methadone or buprenorphine was administered daily, and individual counseling was given weekly. Urine was collected under observation and screened for illicit drugs thrice weekly. Urine was tested for the presence of heroin and other opioids (through testing for morphine), oxycodone (through a more specific assay), buprenorphine, cocaine metabolite, amphetamines, and cannabinoids.

After two weeks of treatment, each participant was trained to use a smartphone as an electronic diary and then carried it for up to 16 weeks, during weeks 3–18 of the study. The smartphones were programmed with our electronic-diary software (Vahabzadeh et al. 2004; Vahabzadeh et al. 2012) to prompt participants three times per day to answer a series of questions. The timing of the prompts was random, but constrained by the individual participant’s self-reported typical waking hours for each day of the week. Participants also initiated entries whenever they experienced a stressful event (SE).

For the stress-event entries, we asked participants to initiate an entry any time they felt “more stressed, overwhelmed, or anxious than usual” and to check whichever applied when they initiated an entry (multiple responses were possible). To categorize severity of each stress event, participants were asked to designate as a “hassle,” “something that could spoil the day,” or “something that could do more than spoil the day,” and to rate “how bad was that feeling” on a 0–10 scale. They indicated the cause of the feeling by selecting from a list: conflict with someone; had too much to do; problems with money; surroundings were unsafe or threatening; was inconvenienced; injury or health problem; legal problem; just started thinking about stressful things; or “other” (a “fill in the blank” option). In each of these SE entries, participants rated how much they craved opioids and cocaine, and reported whom they were with, what they were doing, and where they were, from dropdown lists. The list of questions from the stress event entries is shown in Table 1.

Table 1.

EMA Stress-Event Questions

Question Response Options
How do you feel? (check all that apply) stressed; overwhelmed; anxious
How bad is the feeling?* 0–10 (0 = not bad at all, 10 = the worst you’ve ever felt)
Was the feeling mostly because… (choose one) you had a conflict with someone; you had too much to do; you had problems with money; your surroundings were unsafe or threatening; you were inconvenienced; you had an injury or health problem; you had a legal problem; you just started thinking about stressful things; other [fill in the blank]
How big a problem was this? (choose one) a passing annoyance/hassle; something that could spoil the day; something that could do more than spoil the day
How well can you deal with the feeling right now? 1–5 (1 = not at all, 5 = extremely well)
Right now, do you crave heroin/other opiates (Percocet, oxycodone, etc.)?** 1–5 (1 = not at all, 5 = extremely)
Right now, do you crave cocaine?** 1–5 (1 = not at all, 5 = extremely)
Where were you when the use/stress started? (choose one)*** home; work; vehicle (car, bus, train); waiting for ride, bus, etc.; another’s home; abandominium/vacant home; outside hanging out; walking from one place to another; restaurant; store; church; shelter; bar/club; clinic/doctor; other [fill in the blank]
Who were you with when the stress started? (check all that apply)*** no one; spouse / partner; child(ren); other family; coworkers friends; acquaintances; clinic staff / patients; strangers; out-the-gate partner
Were you with at least one person who you have ever used with?**** Yes; No
Were you with at least one person who still uses?**** Yes, No
What were you doing when the stress started? [check all that apply]*** resting/sleeping; working; walking/riding/travelling; eating or preparing food; watching TV/videos/ DVD; listening to music; talking on phone; talking/ socializing; arguing; household chores or personal hygiene; shopping/errands; child care / elder care; sports / games / other recreation; reading; thinking/ planning; waiting; copping; other illegal activities; legal hustling; on internet; other [fill in the blank]
Open in a new tab
*

In the RP entries participants rated “How much stress are you feeling right now?” on a 1–5 (1 = not at all, 5 = extremely) scale.

**

Wording of these questions in RP entries was “Within 5 minutes of the ED beeping / Since you got to your present location, did you crave…” and was rated on the same 1–5 scale.

***

Wording of these questions in RP entries was “Who are you with?” [check all that apply], “What were you doing when the beep occurred” [check all that apply], and “Where are you now?” (choose one). Response options for stress event and RP entries were identical.

****

Questions in RP entries were the same.

Participants also initiated an EMA entry any time they used a drug for nonmedical purposes, and completed an end-of-day questionnaire before bed. Data from these entries are not reported here.

In random-prompt (RP) entries, participants rated their stress level and how much they craved opioids and cocaine and indicated whom they were with, what they were doing, and their location from dropdown lists. (Participants were also asked to rate their mood using a list of 28 adjectives and to report whether they had recently seen, been offered, or seen others using drugs, alcohol, or tobacco; data from these measures are not reported here.) The participants were paid $10 each week for completing at least 82% of their random prompts, or were given a warning if they did not meet the criterion. Participants who did not meet the 82% completion criterion for 2 weeks after being warned were removed from the study and assisted with transfer into community-based addiction treatment.

Data Analysis

We tested associations among variables using either general or generalized linear mixed models.

Specifically, we used general linear mixed models (SAS Proc Mixed) when the dependent variables were numerical ratings of craving and stress. The predictors in these models were: categories of stress, such as “anxious” versus “overwhelmed”; categories of stress severity, such as “hassle”; and numerical ratings of stress severity (as a predictor of numerical ratings of craving).

We used generalized linear mixed models (SAS Proc Glimmix) when the dependent variables were dichotomous measures. This was the case for most of our analyses, in which we were testing whether stress-event entries were disproportionately associated with particular contexts. Although causation would almost certainly be bidirectional, the models required that we designate predictors and dependent variables. The predictor in each model was a possible context (e.g., being with friends vs. not being with friends), and the dichotomous dependent variable was the type of EMA entry (stress-event entry versus randomly prompted entry). Thus, these models tested whether stress events were over- or underrepresented in a given context relative to random-prompt entries, a proxy for the amount of monitored time spent in that context.

In all the mixed models, we used a spatial-power error structure, which accounts for within-person correlations as a function of the interval between successive entries. Denominator degrees of freedom were conservatively set to reflect the number of participants rather than the number of measurement occasions, as recommended by Bolger and Laurenceau (2013). The predictors were treated as fixed, but each model used a random intercept. Alpha was set at .05, two-tailed, for all analyses. To adjust for the familywise error rate in multiple tests of significance when we compared contexts of stress-event and random-prompt entries, we entered all obtained p values into the SAS procedure Multtest to obtain false-discovery-rate (FDR) p values, which are the ones we report.

Where appropriate, we used F values to calculate effect-size r values (reffect values) as specified by Rosnow et al. (2000), with associated 95% confidence intervals. The Rosnow et al. (2000) procedure for calculating reffect values has not been formally tested in the context of intensive longitudinal data and is dependent on the method used to specify denominator degrees of freedom. Therefore, our reported reffect values should be interpreted with caution in terms of cross-study comparisons (e.g., eventual inclusion in meta-analyses), but are useful for comparisons of effects within our results.

We also checked whether the relationships we found were artifacts of between-person differences rather than within-person associations. Mixed models can account for this problem when participants differ widely on the outcome variable, but do not automatically account for it when people differ widely on the predictor (van de Pol and Wright 2009). Where possible, we checked for this with a model-based method (van de Pol and Wright 2009) in which momentary values are centered on person-level means; for those models, we treated dichotomous predictors (0,1) as numeric variables so person-level “means” could be calculated. When we had no model-based method appropriate for separation of within-person from between-person effects, we checked by descriptively examining how many participants showed patterns of results consistent with the overall pattern (Branch 2014).

Results

Participants and EMA reporting and compliance

Data were collected between July 14, 2009 and June 4, 2015. Of 226 individuals who signed consent, 182 provided EMA data for at least two weeks, 12 provided EMA data for less than two weeks, and 32 left the study before providing any EMA data. Of the 182 participants who provided EMA data, 158 participants (87%) reported at least one stress-event entry, for a total 1,787 stress-event entries. The demographic and drug-use histories of all 182 participants who provided EMA data and the 158 who made stress-event entries were similar (Table 2).

Table 2.

Clinical and demographic characteristics

All
Participants		 Participants
with Stress Events
N 182 158
Opioid agonist treatment
  Methadone 107 (58.8%) 91 (57.6%)
  Buprenorphine/naloxone 75 (41.2%) 67 (42.4%)
Sex
  males 135 (74.2%) 113 (71.5%)
Age mean (SD) years 41.9 (9.6) 42.4 (9.7)
Race
  African American 64.8% 64.6%
  White 34.0% 32.9%
Education mean (SD) years 12.1 (1.5) 12.1 (1.6)
Marital Status
  Married 13.3% 14.6%
  Never married 60.8% 59.9%
  Separated/divorced/widowed 26.0% 25.5%
Employment Status
  Full Time 46.4% 44.0%
  Part Time 22.6% 22.9%
  Unemployed 25.4% 26.8%
  Retired/disability/controlled 5.5% 6.4%
Drug Use History
  Days used in last 30
    Heroin mean (SD) 19.3 (11.9) 19.1 (11.9)
    Other Opioid mean (SD) 8.1 (10.3) 8.0 (10.3)
    Cocaine mean (SD) 4.5 (8.5) 4.6 (8.5)
  Years Using
    Heroin mean (SD) 14.5 (10.2) 14.5 (10.5)
    Other Opioid mean (SD) 1.4 (2.6) 1.5 (2.8)
    Cocaine mean (SD) 6.0 (8.0) 6.2 (8.1)
Route of Administration
  Heroin (n=176) (n=146)
    Intranasal 60.6% 61.6%
    Intravenous 38.8% 37.7%
  Other Opioids (n=143) (n=115)
    Intranasal 7.4% 7.8%
    Oral 91.2% 90.4%
    Smoking 1.5% 1.7%
  Cocaine (n=136) (n=115)
    Intranasal 24.0% 24.4%
    Smoking 54.1% 54.8%
    Intravenous 21.8% 20.9%
Urine drug screen results
  Opioids mean % negative (SD) 61.8% (38.2) 61.0% (36.7)
  Cocaine mean % negative (SD) 66.2% (39.6) 65.2% (39.8)
  Cannabis mean % negative (SD) 85.8% (29.3) 85.9% (29.0)
Open in a new tab

Overall, the participants provided EMA data for a mean (± SD) of 101.0 (± 26.6) days (range 14–156) over 18,379 study days. On average, they completed 2.4 (± 0.4) random-prompt entries per day, for a total of 44,724 entries, with an overall compliance rate of 81%. The 158 participants who made at least one stress-event entry provided EMA data for a mean (± SD) of 102.2 (± 25.5) days (range 14–156) over 16,152 study days and completed 2.4 (± 0.4) random-prompt entries per day for total of 38,768 entries, with an overall compliance rate of 80%.

Stress events

Adjectives (stressed, overwhelmed, anxious) were missing for 47 of the 1,787 stress entries due to a technical error. Of the remaining 1,740 entries, 46.0% were stressed only, 20.0% anxious only, 19.9% overwhelmed only, and 14.0% were assigned more than one adjective (8.7% stressed and overwhelmed, 1.7% stressed and anxious, 0.2% overwhelmed and anxious, and 3.4% all three). In a mixed regression, mean stress ratings differred significantly across the four categories (in ascending order: anxious, 3.89 ± .21; stressed, 4.46 ± .18; overwhelmed, 4.74 ± .21; multiple adjectives, 5.80 ± .25), F(3,208) = 22.95, p < .0001). In Tukey pairwise comparisons, each of those four means was different from each of the others, except “stressed” versus “overwhelmed.” Additional models for decomposition of covariance showed that these adjective/severity relationships were present both within and between participants (reffect sizes for the various pairwise comparisons were 0.28 to 0.58 within participants, 0.06 to 0.36 between participants, reffects > 0.16 were statistically significant).

Across all stress-event entries, the participants’ mean (SEM) rating of “how bad” the stress event felt was 5.04 (0.07) on a 0–10 scale. Categorical descriptions of severity were: 45.4% “hassle,” 30.3% “something that could spoil the day,” and 24.4% “something that could do more than spoil the day.” Across those three categories, there was a linear increase in mean ratings of “how bad” the event felt (0–10 scale), F(1,193) = 99.77, p < .0001, reffect = .58 (CL95 .47, .67) (Figure 1a). An additional model for decomposition of covariance showed that this relationship was present both within and between participants (reffect size 0.80 within, 0.51 between). Even so, when we examined the raw data person by person (Branch 2014), we found that the relationship was not perfect: of the 118 pts who used more than one categorical label for severity across occasions, only 49% gave numerical ratings that increased as would be expected with severity category, for 39%, numerical ratings were flat or fluctuated across categories, and for 12%, numerical ratings actually decreased across categories (see Discussion).

Fig. 1.

Fig. 1

Open in a new tab

Severity and causes of stressful events. a. Bars show percentages of stressful events categorized by participants as “a hassle,” “something that could spoil the day,” or “something that could do more than spoil the day.” Filled circles show mean numerical rating of stress severity for each severity category. b. Bars show percentages of stressful events attributed to each of nine possible causes. Filled circles show mean numerical ratings of stress severity for each cause. Brackets indicate the standard error of the mean.

Participants were asked to choose the source of the stressful event from a dropdown list with an “other” option. The sources, in decreasing order of frequency, were: interpersonal conflict, just thinking about stressful things, money problems, having too much to do, being inconvenienced, “other,” injury or health problems, legal problems, and unsafe surroundings (Figure 1b). Across sources of stress, ratings of “how bad” varied without a clear relationship to frequency (Figure 1b). The “other” events made up 11% of the entries and had the numerically highest mean severity rating. We classified the “other” events into seven categories: issues related to family or significant others (3.0%), general problems (2.9%), study-related issues (1.5%), anticipation of positive events (0.6%), death of or violence toward a family member or friend (0.6%), concerns about drug use or maintaining abstinence (0.5%), and indeterminate (1.6%).

Craving in stress-event entries

Across levels of acute stress (rated 0–10) in stressful events (Figure 2), there were linear increases in both opioid craving (b = 0.13 ± 0.02, CL95 0.10, 0.16, F[1,156] = 63.78, p < .0001) and cocaine craving (b = 0.03 ± 0.01, CL95 0.005, 0.05, F[1,156] = 5.62, p = .019). Additional models for decomposition of covariance showed that the relationship between stress severity and heroin craving was present both within and between participants (reffect = 0.50 within, 0.26 between). For cocaine craving, however, the relationship with stress was accounted for mostly by between-participant differences (reffect = 0.14 within, 0.22 between; reffects < 0.16 were not statistically significant).

Fig. 2.

Fig. 2

Open in a new tab

Drug-craving severity as a function of stress-event severity Craving was rated 1–5 scale (1 indicating no craving) for opioids and cocaine. Means and standard errors are from a mixed regression run solely to generate those estimates for display, treating the 0–10 stress rating as an 11-level categorical predictor rather than a continuous predictor.

Context of stress-event entries vs. background (RP) entries

Figure 3 shows raw data (i.e., data across all EMA entries, ignoring person-level differences) for the contexts (companions, location, activities) of the stress-event entries and RP entries. The figure does not show categories reported in fewer than 3% of entries and not differing between stress and RP entries; thus, not included are three locations (abandominium, church, shelter) and seven activities (child/elder care, sports or recreation, buying drugs, illegal activities, hustling for money, reading, twelve-step activities). The figure is intended to illustrate the overall prevalence of each context. In our generalized linear mixed models, we analyzed the data the other way around, with each context as the predictor and the entry type (stress event versus RP) as the outcome—a choice reflecting our view that causation, though likely bidirectional, may flow more from context to mental state than vice versa. The analyses could be operationalized either way with similar results. (Our use of the terms risk factor and protective factor is purely statistical.) Results of the analyses are in Table 3.

Fig. 3.

Fig. 3

Open in a new tab

Contexts of stress-event and random prompt entries. The bars show raw percentages of the entries in which each context occurred in the 158 participants who made any stress entries. Asterisks mark contexts that were differentially associated with stress in generalized linear mixed models (SAS Proc Glimmix), but note that in those models (Table 2), the predictor was the presence/absence of each context and the outcome was the proportion of each entry type (stress versus RP) in that context. Here we show the data in the opposite way (proportion of entries in each context, as a function of entry type) to give a clear view of the prevalence of each context. The data are shown at the raw momentary level (i.e., as 38,768 RP entries and 1,740 stress-event entries) without regard to the person-level differences that are taken into account by the models in Table 2.

Table 3.

Companions, location, and activities during self-initiated reports of stress events (SEs) versus those present during randomly prompted entries (RPs)

F(1,156) FDR p OR (95%CI) Yes: % SEs No: % SEs reffect within reffect between
Companion type
  Risk factors for stress
    Acquaintances 109.21 .0110 2.83 (2.32, 3.44)* 6.8 2.5 0.63* 0.24*
    Coworkers 15.36 .0006 1.61 (1.26, 2.04)* 4.1 2.6 0.28* 0.13
    Gate partner 9.06 .0006 2.65 (1.61, 4.34)* 6.7 2.6 0.31* 0.07
    Strangers 5.70 .0368 1.31 (1.05, 1.64)* 3.4 2.6 0.17* 0.15
    Friends 5.60 .0370 1.24 (1.04, 1.48)* 3.2 2.6 0.17* 0.13
  Protective from stress
    No one 10.71 .0033 0.83 (0.74, 0.93)* 2.4 2.9 −0.26* −0.01
    Child 6.72 .0033 0.74 (0.58, 0.93)* 2.0 2.7 −0.18* −0.17*
  Not associated with stress
    Spouse/partner 4.28 .07 0.85 (0.73, 0.99) 2.4 2.8 −0.14 −0.12
    Other family 0.44 .61 0.85 (0.81, 1.11) 2.5 2.7 −0.07 −0.12
    Clinic staff or patients 0.01 .97 1.01 (0.77, 1.32) 2.7 2.7 0.01 0.09
Location type
  Risk factors for stress
    Bar or club 18.87 .0004 3.83 (2.05, 7.18)* 10.1 2.8 0.32* 0.24*
    Outside hanging out 13.72 .0004 1.69 (1.28, 2.24)* 4.5 2.8 0.28* 0.03
    Walking 13.14 .0004 1.54 (1.22, 1.96)* 4.3 2.8 0.28* 0.02
    Waiting for ride 9.30 .0004 1.32 (1.10, 1.57)* 3.7 2.8 0.22* 0.25*
  Protective from stress
    Home 86.96 .0004 0.61 (0.55, 0.68)* 2.2 3.6 −0.59* −0.18*
    Another’s home 37.68 .0004 0.46 (0.36, 0.59)* 1.4 3.0 −0.44* 0.02
    Other (not on dropdown list) 17.67 .0004 0.65 (0.53, 0.80)* 1.9 3.0 −0.32* 0.03
  Not associated with stress
    Work 2.09 .24 0.86 (0.70, 1.06) 2.5 2.9 −0.11 −0.01
    Abandominium 2.08 .26 1.69 (0.77, 3.73) 4.7 2.9 0.11 0.09
    Vehicle 1.67 .29 1.10 (0.95, 1.28) 3.1 2.8 0.10 0.09
    Shelter 1.29 .39 1.82 (0.62, 5.35) 5.1 2.9 0.09 −0.01
    Store 1.02 .42 1.16 (0.87, 1.56) 3.3 2.0 0.09 −0.05
    Restaurant 0.51 .60 0.82 (0.47, 1.42) 2.4 2.9 −0.06 0.05
    Church 0.42 .61 1.32 (0.55, 3.16) 3.7 2.9 0.05 0.09
    Clinic 0.03 .90 0.98 (0.79, 1.21) 2.8 2.9 −0.02 0.11
Activity type
  Risk factors for stress
    Arguing 288.72 .0004 8.23 (6.43, 10.53)* 17.9 2.6 0.80* 0.19*
    Other (not on dropdown list) 51.42 .0004 2.49 (1.94, 3.21)* 6.3 2.6 0.49* 0.06
    Waiting 46.30 .0004 1.88 (1.56, 2.25)* 4.8 2.6 0.48* 0.04
    Walking/riding 36.95 .0004 1.42 (1.27, 1.59)* 3.5 2.5 0.42* 0.28*
    On phone 33.14 .0004 1.90 (1.52, 2.37)* 4.9 2.7 0.41* 0.18*
    Talking/socializing 26.45 .0004 1.44 (1.25, 1.66)* 3.7 2.6 0.38* 0.07
  Protective from stress
    Resting/sleeping 95.34 .0004 0.48 (0.41, 0.55)* 1.5 3.1 −0.62* 0.01
    Watching TV 57.29 .0004 0.57 (0.49, 0.66)* 1.7 3.0 −0.51* −0.23*
    Eating 27.42 .0004 0.55 (0.44, 0.69)* 1.6 2.8 −0.40* −0.16*
    On Internet 8.20 .0110 0.56 (0.38, 0.84)* 1.6 2.8 −0.22* −0.02
  Not associated with stress
    Shopping/errands 3.88 .09 1.33 (0.99, 1.78) 3.6 2.7 0.16 0.01
    Thinking 3.53 .11 1.17 (0.99, 1.37) 3.1 2.7 0.13 0.09
    Copping (buying drugs) 2.59 .18 1.78 (0.87, 3.65) 4.7 2.7 0.12 0.11
    Hustling for money 1.03 .42 1.44 (0.70, 2.93) 3.9 2.7 0.08 0.04
    Listening to music 0.98 .42 1.14 (0.88, 1.48) 3.1 2.7 0.08 0.02
    Chores/hygiene 0.96 .42 1.12 (0.90, 1.39) 3.0 2.7 0.08 −0.04
    Illegal activities 0.33 .66 1.23 (0.59, 2.56) 3.3 2.7 0.04 0.07
    Sports/games/recreation 0.28 .67 0.87 (0.51, 1.48) 2.4 2.7 −0.04 −0.01
    Working 0.18 .73 0.96 (0.79, 1.16) 2.6 2.7 −0.04 0.01
    Child or elder care 0.09 .82 1.07 (0.68, 1.67) 2.9 2.7 0.02 0.01
    Reading 0.02 .90 0.98 (0.70, 1.35) 2.7 2.7 −0.02 0.05
Open in a new tab

Note. Results in the first four columns are from Glimmix models in which the predictor was the presence versus absence of the companion and the dependent variable was the type of entry (stress event versus random prompt). In the raw data from the 158 participants who made stress entries, 4.3% of all EMA entries were stress entries (1,740 / (1,740+38,768)). The columns “Yes: % SEs” and “No: % SEs” show model-adjusted rates of stress entries, in the presence versus absence of the each type of companion. Model-adjusted rates tended to be lower than the rate in the raw data; we report them to make the statistical output more concrete.

The last two columns, “reffect within” and “reffect between,” are from additional Glimmix models in which each overall association was decomposed into relationships that occurred over time within participants (reffect within) versus differences between participants (reffect between). F values from those models were used to calculate reffect sizes, which can take the same range of values as Pearson correlation coefficients.

*

FDR-adjusted p value < .05.

As shown in Table 3, the companion types most strongly associated with stress events were acquaintances (this was one of the strongest associations in the dataset) and coworkers, along with “out the gate” partners (companions who help acquire a morning dose of drug), strangers, and friends. Being alone or being with a child were each associated with a lower probability of stress. All these associations were stronger at the within-person level than the between-person level. Being with a spouse/partner or other family member was, if anything, slightly protective against stress, but not to a statistically significant degree. Being with clinic staff or patients was not associated with stress.

The locations most strongly associated with stress events were bars/clubs and outdoor locations. Home, and others’ homes, tended to be protective against stress (Table 3). The workplace was, if anything, a slightly protective location (though this was not statistically significant), even though being in the company of coworkers was associated with stress. Almost all these associations were stronger at the within-person level than the between-person level.

The activity most clearly associated with stress was arguing—this was the strongest contextual association in the dataset (Table 3). Outdoor and social activities were also associated with stress, as were “other” activities, which, based on fill-in responses, included family conflicts, family and personal health crises, and courtroom situations. Stress events were less likely when participants were resting, watching TV, eating, or on the internet. All these associations were stronger at the within-person level than the between-person level. Stress events were not differentially associated with any of the other activities we assessed, though shopping/errands, thinking, and copping (buying drugs) appeared marginally likely to be risk factors. Working, as an activity, showed no sign of an association with stress.

Discussion

The primary purpose of this paper is to present a detailed first description of the overall patterns of stressful events reported by individuals with opioid use disorders during treatment. Because of the way we asked participants to report every stressful event as it happened, this is the first data set of its kind.

We asked participants to assign a numerical severity rating to each stressor on a 0–10 scale, but, anticipating that the results might be hard to interpret, we also asked them to rank the severity of each stressor in everyday language—as “a hassle,” “something that could spoil the day,” or “something that could do more than spoil the day.” The usefulness of this classification was borne out by comparison of the two sets of responses. The relationship was strong overall, but in more than half of the participants, it was not entirely orderly and was sometimes even contradictory. We take this to mean that a 0–10 scale without verbal anchor points, used repeatedly without access to one’s own prior responses on it, can lead to inconsistencies across measurement occasions. Accompanying such a scale with an everyday-language item (or incorporating everyday-language anchors into the scale) seems likely to produce results that more straightforwardly capture a respondent’s experiences across time.

The severity of stressful events in our participants was associated with greater craving for drugs. This finding is consistent with our earlier work on background stress in a similar cohort of patients in opioid agonist treatment (Kowalczyk et al. 2015; Preston and Epstein 2011). The detection of craving at the time of stressful events is not surprising, but is important because craving is a key feature of addiction: it predicts treatment outcome (Higley et al. 2011; Tsui et al. 2014) and has been added as a criterion for diagnosis of substance-use disorders in DSM-5 (Murphy et al. 2014). This study confirmed prior studies showing that craving increases with stress when recorded at random times and suggests that the craving may be even higher at the time of a specific stressful event compared to background stress. The relationship was stronger for heroin craving than for cocaine craving; this difference had been present to a lesser degree in our earlier work with background stress (Preston and Epstein 2011).

One difference between these and our earlier EMA findings concerns stress in the workplace. In our earlier study we found that background stress was lower at the workplace than at any other location, and that mood was better in the presence of coworkers than with other companions (Epstein and Preston 2012)—a finding we took to mean that the workplace, for our population, can be a respite from chaotic home environments (Damaske et al. 2016). In the current analyses of stressful events, not so. Being in the workplace appeared somewhat protective (though this did not reach statistical significance); being with coworkers was clearly associated with an increased likelihood of stressful events; the activity of working had no association with the likelihood of stressful events. Being at home was strongly protective. This discrepancy probably reflects the difference between fluctuations in mental state versus occurrence of specific external stressors. Arguably, neither should go unassessed (Deutsch 1986).

The “external event”-driven nature of our data is underscored by the other situational correlates of stress-event entries: they tended to occur in social company (with acquaintances or friends, or on the phone) rather than with family (spouse/partner, child), and they tended to occur in places where there is more overall activity (bars/clubs, outside hanging out, walking) and more likelihood of unexpected experiences (in the presence of strangers). These are the kinds of environmental exposures we are trying to quantify with our ongoing studies of activity space using GPS (Epstein et al. 2014). A more nuanced picture of what happens at home, with spouses and other partners or family members, could emerge from dyadic EMA, in which both members of a relationship or household respond independently to the same mobile assessments (Dunton et al. 2016; Laurenceau and Bolger 2012). The robust association of stress events with the presence of “out the gate” partners (drug-acquistion companions) may partly reflect the fact that, as shown in laboratory studies, drug cues may be stressors in people who are trying to abstain (Sinha et al. 2006; Sinha et al. 2003).

The absence of any overall protective effect of listening to music might be an artifact of the relative crudeness of the item we used. An EMA study that focused on music listening (in a general-population sample) showed that music reduces stress only in social situations or when listened to alone for the specific purpose of stress reduction (Linnemann et al. 2016).

The slightly stress-protective effect of internet use contrasts with studies that have shown deleterious long-term effects of overuse of social media (Sriwilai and Charoensukmongkol 2016; Zheng and Lee 2016). However, the association in our data was at the momentary level: it can be summarized as “during internet use, there were fewer stressors” rather than “for people with more overall internet use, there were fewer stressors.” This is consistent with findings that social-media use can buffer acute stress (Rus and Tiemensma 2016).

When we asked participants to give reasons for stress events, the most frequent responses included busyness and interpersonal conflicts, both in keeping with the event-driven nature of the data. Participants also frequently cited money problems; this accords with the unsurprising prior finding, in healthy volunteers, that having at least sustenance-level cash available is an independent predictor of life satisfaction (Ruberton et al. 2016). It also accords with prior findings that financial stress can be a major barrier to recovery from addiction (Siahpush et al. 2009). Distinct from the other types of stress events was the high frequency of reports brought on by having “just started thinking about stressful things.” These can be taken as instances of ruminative stress, which may be especially deleterious to emotional health (Snyder and Hankin 2016) and is specifically associated with craving (Caselli et al. 2010; Caselli et al. 2013). In future studies we plan to analyze stress responses as a function of trait differences such as proneness to anxiety. One of the more rare causes of stress events was the feeling that one’s surroundings were unsafe—but when that did occur, it led to some of the highest ratings of stress severity. This finding supports the importance of studying and trying to address environmental precipitants of stress and drug seeking.

A limitation of this study is that we cannot know how many stress events went unreported. When we tested the EMA program ourselves to check that it was working properly, we found that at the peak of a stressful event, we did not always have time to make an entry. This is likely to have been the case for our study participants as well. Nevertheless, most participants (87%) made at least one stress entry, and there were no discernible differences between those who did and those who did not, an area we intend to explore in future analyses. The methods appear to have worked insofar as the data are internally consistent. Stress ratings increased with other measures of severity, and associated responses, such as being with a coworker at work while working, had effects in the same direction, even if not always meeting statistical significance.

In hindsight, one thing we would have done differently to reduce burden and increase data quality would be to let participants add a few of their own categories to our dropdown list of stress causes. Addition of biosensors to detect physiological signs of stress, a technique we have started to explore (Kennedy et al. 2015; Sarker et al. 2016), could both validate the stress events and indicate the circumstances under which participants fail to report stressful events. Finally, programming the electronic diary to deliver timed follow-up questions could provide information on the duration of each stress episode, the extent of consequent craving, and the occurrence of associated drug use. We analyzed some of those stress/drug associations in more detail in a prior publication (Furnari et al. 2015); this paper complements that one by providing an overview of the stress events themselves.

The kind of description presented here is a necessary step toward developing an intervention (Sidman 1960). There is already evidence, in healthy adults, that a mobile, automated “just in time” intervention (JITI) for a stress event can buffer the effects of the next day’s stressors (Donald et al. 2016). Next steps from our current findings to a stress-targeted JITI for addiction should include situation- and person-specific analyses (Russell and Odgers 2016) so that the intervention can be matched to the event.

Acknowledgments

Funding and Disclosure

This study was supported by the Intramural Research Program of the National Institute on Drug Abuse, National Institutes of Health.

Footnotes

The authors do not have any competing financial interests related to the research presented.

References

  1. Bolger N, Laurenceau J-P. Intensive Longitudinal Methods. New York: Guilford; 2013. [Google Scholar]
  2. Bossert JM, Marchant NJ, Calu DJ, Shaham Y. The reinstatement model of drug relapse: recent neurobiological findings, emerging research topics, and translational research. Psychopharmacology. 2013;229:453–76. doi: 10.1007/s00213-013-3120-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Branch M. Malignant side effects of null-hypothesis significance testing. Theory & Psychology. 2014;24:256–277. [Google Scholar]
  4. Brewer JA, Sinha R, Chen JA, Michalsen RN, Babuscio TA, Nich C, Grier A, Bergquist KL, Reis DL, Potenza MN, Carroll KM, Rounsaville BJ. Mindfulness training and stress reactivity in substance abuse: results from a randomized, controlled stage I pilot study. Subst Abus. 2009;30:306–17. doi: 10.1080/08897070903250241. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Carroll KM. A Cognitive-Behavioral Approach: Treating Cocaine Addiction. Washington, DC: National Institutes of Health; 1998. [Google Scholar]
  6. Carroll KM, Kiluk BD, Nich C, Gordon MA, Portnoy GA, Marino DR, Ball SA. Computer-assisted delivery of cognitive-behavioral therapy: efficacy and durability of CBT4CBT among cocaine-dependent individuals maintained on methadone. Am J Psychiatry. 2014;171:436–44. doi: 10.1176/appi.ajp.2013.13070987. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Caselli G, Ferretti C, Leoni M, Rebecchi D, Rovetto F, Spada MM. Rumination as a predictor of drinking behaviour in alcohol abusers: a prospective study. Addiction. 2010;105:1041–8. doi: 10.1111/j.1360-0443.2010.02912.x. [DOI] [PubMed] [Google Scholar]
  8. Caselli G, Gemelli A, Querci S, Lugli AM, Canfora F, Annovi C, Rebecchi D, Ruggiero GM, Sassaroli S, Spada MM, Watkins ER. The effect of rumination on craving across the continuum of drinking behaviour. Addict Behav. 2013;38:2879–83. doi: 10.1016/j.addbeh.2013.08.023. [DOI] [PubMed] [Google Scholar]
  9. Chiesa A, Serretti A. Are mindfulness-based interventions effective for substance use disorders? A systematic review of the evidence. Subst Use Misuse. 2014;49:492–512. doi: 10.3109/10826084.2013.770027. [DOI] [PubMed] [Google Scholar]
  10. Cleck JN, Blendy JA. Making a bad thing worse: adverse effects of stress on drug addiction. J Clin Invest. 2008;118:454–61. doi: 10.1172/JCI33946. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Damaske S, Zawadzki MJ, Smyth JM. Stress at work: Differential experiences of high versus low SES workers. Soc Sci Med. 2016;156:125–33. doi: 10.1016/j.socscimed.2016.03.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Deutsch F. Calling a freeze on “stress wars”: There is hope for adaptational outcomes. American Psychologist. 1986;41:713–714. [Google Scholar]
  13. Donald JN, Atkins PWB, Parker PD, Christie AM, Ryan RM. Daily stress and the benefits of mindfulness: examining the daily and longitudinal relations between present-moment awareness and stress responses. Journal of Research in Personality. 2016;65:30–37. [Google Scholar]
  14. Dunton GF, Dzubur E, Huh J, Belcher BR, Maher JP, O’Connor S, Margolin G. Daily associations of stress and eating in mother-child dyads. Health Educ Behav. 2016 doi: 10.1177/1090198116663132. [DOI] [PubMed] [Google Scholar]
  15. Epstein DH, Preston KL. TGI Monday?: drug-dependent outpatients report lower stress and more happiness at work than elsewhere. Am J Addict. 2012;21:189–98. doi: 10.1111/j.1521-0391.2012.00230.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Epstein DH, Tyburski M, Craig IM, Phillips KA, Jobes ML, Vahabzadeh M, Mezghanni M, Lin JL, Furr-Holden CD, Preston KL. Real-time tracking of neighborhood surroundings and mood in urban drug misusers: application of a new method to study behavior in its geographical context. Drug Alcohol Depend. 2014;134:22–9. doi: 10.1016/j.drugalcdep.2013.09.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Erb S, Hitchcott PK, Rajabi H, Mueller D, Shaham Y, Stewart J. Alpha-2 adrenergic receptor agonists block stress-induced reinstatement of cocaine seeking. Neuropsychopharmacology. 2000;23:138–50. doi: 10.1016/S0893-133X(99)00158-X. [DOI] [PubMed] [Google Scholar]
  18. Fox HC, Seo D, Tuit K, Hansen J, Kimmerling A, Morgan PT, Sinha R. Guanfacine effects on stress, drug craving and prefrontal activation in cocaine dependent individuals: preliminary findings. J Psychopharmacol. 2012;26:958–72. doi: 10.1177/0269881111430746. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Furnari M, Epstein DH, Phillips KA, Jobes ML, Kowalczyk WJ, Vahabzadeh M, Lin JL, Preston KL. Some of the people, some of the time: field evidence for associations and dissociations between stress and drug use. Psychopharmacology. 2015;232:3529–37. doi: 10.1007/s00213-015-3998-7. [DOI] [PubMed] [Google Scholar]
  20. Hayes SC, Levin ME, Plumb-Vilardaga J, Villatte JL, Pistorello J. Acceptance and commitment therapy and contextual behavioral science: examining the progress of a distinctive model of behavioral and cognitive therapy. Behav Ther. 2013;44:180–98. doi: 10.1016/j.beth.2009.08.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hekler EB, Michie S, Pavel M, Rivera DE, Collins LM, Jimison HB, Garnett C, Parral S, Spruijt-Metz D. Advancing models and theories for digital behavior change interventions. Am J Prev Med. 2016;51:825–832. doi: 10.1016/j.amepre.2016.06.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Highfield D, Yap J, Grimm JW, Shalev U, Shaham Y. Repeated lofexidine treatment attenuates stress-induced, but not drug cues-induced reinstatement of a heroin-cocaine mixture (speedball) seeking in rats. Neuropsychopharmacology. 2001;25:320–31. doi: 10.1016/S0893-133X(01)00227-5. [DOI] [PubMed] [Google Scholar]
  23. Higley AE, Crane NA, Spadoni AD, Quello SB, Goodell V, Mason BJ. Craving in response to stress induction in a human laboratory paradigm predicts treatment outcome in alcohol-dependent individuals. Psychopharmacology. 2011;218:121–9. doi: 10.1007/s00213-011-2355-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Hoppes K. The application of mindfulness-based cognitive interventions in the treatment of co-occurring addictive and mood disorders. CNS Spectr. 2006;11:829–51. doi: 10.1017/s1092852900014991. [DOI] [PubMed] [Google Scholar]
  25. Hossain SM, Ali AA, Rahman M, Ertin E, Epstein D, Kennedy A, Preston K, Umbricht A, Chen Y, Kumar S. Identifying drug (cocaine) intake events from acute physiological response in the presence of free-living physical activity. IPSN. 2014;2014:71–82. [PMC free article] [PubMed] [Google Scholar]
  26. Jobes ML, Ghitza UE, Epstein DH, Phillips KA, Heishman SJ, Preston KL. Clonidine blocks stress-induced craving in cocaine users. Psychopharmacology. 2011;218:83–8. doi: 10.1007/s00213-011-2230-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kennedy AP, Epstein DH, Jobes ML, Agage D, Tyburski M, Phillips KA, Ali AA, Bari R, Hossain SM, Hovsepian K, Rahman MM, Ertin E, Kumar S, Preston KL. Continuous in-the-field measurement of heart rate: Correlates of drug use, craving, stress, and mood in polydrug users. Drug Alcohol Depend. 2015;151:159–66. doi: 10.1016/j.drugalcdep.2015.03.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kowalczyk WJ, Phillips KA, Jobes ML, Kennedy AP, Ghitza UE, Agage DA, Schmittner JP, Epstein DH, Preston KL. Clonidine maintenance prolongs opioid abstinence and decouples stress from craving in daily life: A randomized controlled trial with Ecological Momentary Assessment. Am J Psychiatry. 2015;172:760–7. doi: 10.1176/appi.ajp.2014.14081014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Laurenceau J-P, Bolger N. Analyzing diary and intensive longitudinal data from dyads. In: Mehl M, Conner TS, editors. Handbook of Research Methods for Studying Daily Life. Guilford; New York: 2012. pp. 407–422. [Google Scholar]
  30. Lindhiem O, Bennett CB, Rosen D, Silk J. Mobile technology boosts the effectiveness of psychotherapy and behavioral interventions: a meta-analysis. Behav Modif. 2015;39:785–804. doi: 10.1177/0145445515595198. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Linnemann A, Strahler J, Nater UM. The stress-reducing effect of music listening varies depending on the social context. Psychoneuroendocrinology. 2016;72:97–105. doi: 10.1016/j.psyneuen.2016.06.003. [DOI] [PubMed] [Google Scholar]
  32. Luoma JB, Kohlenberg BS, Hayes SC, Fletcher L. Slow and steady wins the race: a randomized clinical trial of acceptance and commitment therapy targeting shame in substance use disorders. J Consult Clin Psychol. 2012;80:43–53. doi: 10.1037/a0026070. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. McLellan AT, Luborsky L, Cacciola J, Griffith J, Evans F, Barr HL, O’Brien CP. New data from the Addiction Severity Index. Reliability and validity in three centers. J Nerv Ment Dis. 1985;173:412–23. doi: 10.1097/00005053-198507000-00005. [DOI] [PubMed] [Google Scholar]
  34. Murphy CM, Stojek MK, Few LR, Rothbaum AO, Mackillop J. Craving as an alcohol use disorder symptom in DSM-5: an empirical examination in a treatment-seeking sample. Exp Clin Psychopharmacol. 2014;22:43–9. doi: 10.1037/a0034535. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Nahum-Shani I, Smith SN, Spring BJ, Collins LM, Witkiewitz K, Tewari A, Murphy SA. Just-in-Time Adaptive Interventions (JITAIs) in mobile health: Key components and design principles for ongoing health behavior support. Ann Behav Med. 2016 doi: 10.1007/s12160-016-9830-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Preston KL, Epstein DH. Stress in the daily lives of cocaine and heroin users: relationship to mood, craving, relapse triggers, and cocaine use. Psychopharmacology. 2011;218:29–37. doi: 10.1007/s00213-011-2183-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Rahman M, Bari R, Ali AA, Sharmin M, Raij A, Hovsepian K, Hossain SM, Ertin E, Kennedy A, Epstein DH, Preston KL, Jobes M, Beck JG, Kedia S, Ward KD, al’Absi M, Kumar S. Are wer there yet? Feasibility of continuous stress assessment via wireless physiological sensors. ACM BCB. 2014;2014:479–488. doi: 10.1145/2649387.2649433. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Riley WT, Rivera DE, Atienza AA, Nilsen W, Allison SM, Mermelstein R. Health behavior models in the age of mobile interventions: are our theories up to the task? Transl Behav Med. 2011;1:53–71. doi: 10.1007/s13142-011-0021-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Robins LN, Cottler LB, Bucholz KK, Compton WM., III . The Diagnostic Interview Schedule, Version IV. St. Louis, MO: Washington University; 1995a. [Google Scholar]
  40. Robins LN, Cottler LB, Bucholz KK, Compton WM. The Diagnostic Interview Schedule, Version IV. St. Louis, MO: Washington University; 1995b. [Google Scholar]
  41. Rosnow RL, Rosenthal R, Rubin DB. Contrasts and correlations in effect-size estimation. Psychol Sci. 2000;11:446–53. doi: 10.1111/1467-9280.00287. [DOI] [PubMed] [Google Scholar]
  42. Ruberton PM, Gladstone J, Lyubomirsky S. How your bank balance buys happiness: The importance of “cash on hand” to life satisfaction. Emotion. 2016;16:575–80. doi: 10.1037/emo0000184. [DOI] [PubMed] [Google Scholar]
  43. Rus HM, Tiemensma J. Social media as a shield: Facebook may buffer acute stress. Psychoneuroendocrinology. 2016;71:65–66. [Google Scholar]
  44. Russell MA, Odgers CL. Antisocial behavior among children in poverty: Understanding environmental effects in daily life. In: Pietz CA, Mattson CA, editors. Violent Offenders: Understanding and Assessment. Oxford University Press; Oxford, UK: 2016. [Google Scholar]
  45. Sarker H, Tyburski M, Rahman M, Hovsepian K, Sharmin M, Epstein DH, Preston KL, Furr-Holden CD, Milam A, Nahum-Shani I, al’Absi M. Finding significant stress episodes in a discontinuous time series of rapidly varying mobile sensor data; Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems; 2016. pp. 4489–4501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Shiffman S, Stone AA, Hufford MR. Ecological momentary assessment. Annu Rev Clin Psychol. 2008;4:1–32. doi: 10.1146/annurev.clinpsy.3.022806.091415. [DOI] [PubMed] [Google Scholar]
  47. Siahpush M, Yong HH, Borland R, Reid JL, Hammond D. Smokers with financial stress are more likely to want to quit but less likely to try or succeed: findings from the International Tobacco Control (ITC) Four Country Survey. Addiction. 2009;104:1382–90. doi: 10.1111/j.1360-0443.2009.02599.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Sidman M. Tactics of Scientific Research: Evaluating Experimental Data in Psychology. New York: Basic Books; 1960. [Google Scholar]
  49. Sinha R, Garcia M, Paliwal P, Kreek MJ, Rounsaville BJ. Stress-induced cocaine craving and hypothalamic-pituitary-adrenal responses are predictive of cocaine relapse outcomes. Arch Gen Psychiatry. 2006;63:324–31. doi: 10.1001/archpsyc.63.3.324. [DOI] [PubMed] [Google Scholar]
  50. Sinha R, Talih M, Malison R, Cooney N, Anderson GM, Kreek MJ. Hypothalamic-pituitary-adrenal axis and sympatho-adreno-medullary responses during stress-induced and drug cue-induced cocaine craving states. Psychopharmacology. 2003;170:62–72. doi: 10.1007/s00213-003-1525-8. [DOI] [PubMed] [Google Scholar]
  51. Snyder HR, Hankin BL. Spiraling out of control: Stress generation and subsequent rumination mediate the link between poorer cognitive control and internalizing psychopathology. Clin Psychol Sci. 2016;4:1047–1064. doi: 10.1177/2167702616633157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Sriwilai K, Charoensukmongkol P. Face it, don’t Facebook it: Impacts of social media addiction on mindfulness, coping strategies and the consequence on emotional exhaustion. Stress Health. 2016;32:427–434. doi: 10.1002/smi.2637. [DOI] [PubMed] [Google Scholar]
  53. Tsui JI, Anderson BJ, Strong DR, Stein MD. Craving predicts opioid use in opioid-dependent patients initiating buprenorphine treatment: a longitudinal study. Am J Drug Alcohol Abuse. 2014;40:163–9. doi: 10.3109/00952990.2013.848875. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Vahabzadeh M, Epstein DH, Mezghanni M, Lin J-L, Preston KL. An electronic diary software for ecological momentary assessment (EMA) in clinical trials; Proceedings of the 17th IEEE Symposium on Computer-Based Medical Systems (CBMS); 2004. pp. 167–172. [Google Scholar]
  55. Vahabzadeh M, Mezghanni M, Lin J-L. Context aware mobile device software for substance abuse interventions and behavioral modification. Federal Register. 2012;77:48997–48998. [Google Scholar]
  56. van de Pol M, Wright J. A simple method for distinguishing within- versus between-subject effects using mixed models. Animal Behaviour. 2009;77:753–758. [Google Scholar]
  57. Zheng X, Lee MKO. Excessive use of mobile social networking sites: Negative consequences on individuals. Computers in Human Behavior. 2016;65:65–76. [Google Scholar]

RESOURCES