Background Maladaptive behaviors may be more fully understood and efficiently prevented by ambulatory tools that assess people’s ongoing experience in the context of their environment. Methods To demonstrate new field-deployable methods for assessing mood and behavior as a function of neighborhood surroundings (Geographical Momentary Assessment; GMA), we collected time-stamped GPS data and Ecological Momentary Assessment (EMA) ratings of mood, stress, and drug craving over 16 weeks at randomly prompted times during the waking hours of opioid-dependent polydrug users receiving methadone maintenance. Locations of EMA entries and participants’ travel tracks were calculated for the 12 hours before each EMA entry were mapped. Associations between subjective ratings and objective environmental ratings were evaluated at the whole neighborhood and 12-hour track levels. Results Participants (N=27) were compliant with GMA data collection; 3,711 randomly prompted EMA entries were matched to specific locations. At the neighborhood level, physical disorder was negatively correlated with negative mood, stress, and heroin and cocaine craving (ps <.0001 to .0335); drug activity was negatively correlated with stress, heroin and cocaine craving (ps .0009 to .0134). Similar relationships were found for the environments around respondents’ tracks in the 12 hours preceding EMA entries. Conclusions The results support the feasibility of GMA. The relationships between neighborhood characteristics and participants’ reports were counterintuitive and counter-hypothesized, and challenge some assumptions about how ostensibly stressful environments are associated with lived experience and how such environments ultimately impair health. GMA methodology may have applications for development of individual- or neighborhood-level interventions.

Background Ambulatory physiological monitoring could clarify antecedents and consequences of drug use and could contribute to a sensor-triggered mobile intervention that automatically detects behaviorally risky situations. Our goal was to show that such monitoring is feasible and can produce meaningful data. Methods We assessed heart rate (HR) with AutoSense, a suite of biosensors that wirelessly transmits data to a smartphone, for up to four weeks in 40 polydrug users in opioid-agonist maintenance as they went about their daily lives. Participants also self-reported drug use, mood, and activities on electronic diaries. We compared HR with self-report using multilevel modeling (SAS Proc Mixed). Results Compliance with AutoSense was good; the data yield from the wireless electrocardiographs was 85.7%. HR was higher when participants reported cocaine use than when they reported heroin use (F(2,9) = 250.3, p<.0001) and was also higher as a function of the dose of cocaine reported (F(1,8) = 207.7, p<.0001). HR was higher when participants reported craving heroin (F(1,16)=230.9, p<.0001) or cocaine (F(1,14)=157.2, p<.0001) than when they reported of not craving. HR was lower (p<.05) in randomly prompted entries in which participants reported feeling relaxed, feeling happy, or watching TV, and was higher when they reported feeling stressed, being hassled, or walking. Conclusions High-yield, high-quality heart-rate data can be obtained from drug users in their natural environment as they go about their daily lives, and the resultant data robustly reflect episodes of cocaine and heroin use and other mental and behavioral events of interest.

No abstract

Management of daily stress can be greatly improved by delivering sensor-triggered just-in-time interventions (JITIs) on mobile devices. The success of such JITIs critically depends on being able to mine the time series of noisy sensor data to find the most opportune moments. In this paper, we propose a time series pattern mining method to detect significant stress episodes in a time series of discontinuous and rapidly varying stress data. We apply our model to 4 weeks of physiological, GPS, and activity data collected from 38 users in their natural environment to discover patterns of stress in real-life. We find that the duration of a prior stress episode predicts the duration of the next stress episode and stress in mornings and evenings is lower than during the day. We then analyze the relationship between stress and objectively rated disorder in the surrounding neighborhood and develop a model to predict stressful episodes.

Just-in-time adaptive interventions (JITAIs), typically smartphone apps, learn to deliver therapeutic content when users need it. The challenge is to "push" content at algorithmically chosen moments without making users trigger it with effortful input. We trained a randomForest algorithm to predict heroin craving, cocaine craving, or stress (reported via smartphone app 3x/day) 90 min into the future, using 16 weeks of field data from 189 outpatients being treated for opioid-use disorder. We used only one form of continuous input (along with person-level demographic data), collected passively: an indicator of environmental exposures along the past 5 h of movement, as assessed by GPS. Our models achieved excellent overall accuracy-as high as 0.93 by the end of 16 weeks of tailoring-but this was driven mostly by correct predictions of absence. For predictions of presence, "believability" (positive predictive value, PPV) usually peaked in the high 0.70s toward the end of the 16 weeks. When the prediction target was more rare, PPV was lower. Our findings complement those of other investigators who use machine learning with more broadly based "digital phenotyping" inputs to predict or detect mental and behavioral events. When target events are comparatively subtle, like stress or drug craving, accurate detection or prediction probably needs effortful input from users, not passive monitoring alone. We discuss ways in which accuracy is difficult to achieve or even assess, and warn that high overall accuracy (including high specificity) can mask the abundance of false alarms that low PPV reveals.npj Digital Medicine (2020) 3:26 ; https://doi.