Search

Background: Just-in-time adaptive interventions (JITAI) aim to prevent smoking lapse using tailored support delivered via mobile technology in the moments when it is most needed. Effective smoking cessation JITAI rely on the development of accurate decision rules that determine when someone is most likely to lapse. The primary goal of the present study was to identify the strongest predictors of first lapse among smokers undergoing a quit attempt., Methods: Smokers attending a clinic-based smoking cessation program (n = 74) were asked to complete ecological momentary assessments five times daily on study-provided smartphones for 4 weeks post-quit. A three-stage modeling process utilized Cox proportional hazards regression to examine time to lapse a function of 31 predictors. First, univariate models evaluated the relationship between each predictor and time to lapse. Second, the elastic net machine learning algorithm was used to select the best predictors. Third, backwards elimination further reduced the set of predictors to optimize parsimony., Results: Univariate models identified seven predictors significantly related to time to lapse. The elastic net algorithm retained five: perceived odds of smoking today, confidence in ability to avoid smoking, motivation to avoid smoking, urge to smoke, and cigarette availability. The reduced model demonstrated inadequate approximation to the non-penalized baseline model., Conclusions: Accurate estimation of moments of high risk for smoking lapse remains an important goal in the development of JITAI. These results demonstrate the utility of exploratory data-driven approaches to variable selection. The results of this study can inform future JITAI by highlighting targets for intervention., (Copyright © 2020 Elsevier B.V. All rights reserved.)

Introduction: Machine learning algorithms such as elastic net regression and backward selection provide a unique and powerful approach to model building given a set of psychosocial predictors of smoking lapse measured repeatedly via ecological momentary assessment (EMA). Understanding these predictors may aid in developing interventions for smoking lapse prevention., Methods: In a randomized-controlled smoking cessation trial, smartphone-based EMAs were collected from 92 participants following a scheduled quit date. This secondary analysis utilized elastic net-penalized cox proportional hazards regression and model approximation via backward elimination to (1) optimize a predictive model of time to first lapse and (2) simplify that model to its core constituent predictors to maximize parsimony and generalizability., Results: Elastic net proportional hazards regression selected 17 of 26 possible predictors from 2065 EMAs to model time to first lapse. The predictors with the highest magnitude regression coefficients were having consumed alcohol in the past hour, being around and interacting with a smoker, and having cigarettes easily available. This model was reduced using backward elimination, retaining five predictors and approximating to 93.9% of model fit. The retained predictors included those mentioned above as well as feeling irritable and being in areas where smoking is either discouraged or allowed (as opposed to not permitted)., Conclusions: The strongest predictors of smoking lapse were environmental in nature (e.g., being in smoking-permitted areas) as opposed to internal factors such as psychological affect. Interventions may be improved by a renewed focus of interventions on these predictors., Implications: The present study demonstrated the utility of machine learning algorithms to optimize the prediction of time to smoking lapse using EMA data. The two models generated by the present analysis found that environmental factors were most strongly related to smoking lapse. The results support the use of machine learning algorithms to investigate intensive longitudinal data, and provide a foundation for the development of highly tailored, just-in-time interventions that can target on multiple antecedents of smoking lapse.

Despite a persistent link between homelessness, mental illness, and increased psychiatric hospitalizations, there is a dearth of research examining risk factors for higher utilization of mental health treatment among this vulnerable population. This study utilizes machine learning in electronic health record data to examine admissions from 2010 to 2019 at a regional psychiatric safety-net hospital among individuals experiencing homelessness compared to non-homeless individuals. Of the 64 social, demographic, clinical and treatment predictors examined, 12 were identified as the strongest predictors of homelessness status at psychiatric hospital admission. Admissions for patients identified as homeless were driven by social determinants of health. The strongest predictor was having no income (+267% increase in the odds of being homeless). The study results demonstrate how machine learning can be used to explore electronic health record data collected during psychiatric hospital admission, including social determinants of health, to elucidate the clinical and social needs of homeless individuals seeking psychiatric treatment. These findings have implications for the development of targeted prevention/treatment interventions to address homeless patients' needs during psychiatric hospitalization to improve social and health outcomes. [ABSTRACT FROM AUTHOR]

Senior financial exploitation (FE) is prevalent and harmful. Its often insidious nature and co-occurrence with other forms of mistreatment make detection and substantiation challenging. A secondary data analysis of N = 8,800 Adult Protective Services substantiated senior mistreatment cases, using machine learning algorithms, was conducted to determine when pure FE versus hybrid FE was occurring. FE represented N = 2514 (29%) of the cases with 78% being pure FE. Victim suicidal ideation and threatening behaviors, injuries, drug paraphernalia, contentious relationships, caregiver stress, and burnout and victims needing assistance were most important for differentiating FE vs non-FE-related mistreatment. The inability to afford housing, medications, food, and medical care as well as victims suffering from intellectual disability disorder(s) predicted hybrid FE. This study distinguishes socioecological factors strongly associated with the presence of FE during protective service investigations. These findings support existing and new indicators of FE and could inform protective service investigation practices. [ABSTRACT FROM AUTHOR]

Hepatitis C virus (HCV) in the U.S. has tripled in the prior five years, and injecting drug use is the primary risk for HCV, with up to 90% of older and former people who inject drugs (PWIDs) testing positive. Laboratory testing of HCV for any PWIDs is the gold standard, however many PWIDs lack access to health treatment or services. Identifying risks of HCV via a data science approach would aid community health workers (CHW) to rapidly link those most at risk of infection with treatment. This study employed a data-science approach to determine the strongest risk factors of HCV in a sample of Mexican-Americans WIDs n = 221 (96 negative/125 positive). Data included 238 demographic and psychosocial predictors. A Random Forest machine learning algorithm demonstrated significant prediction improvement over baseline no information rate comparison. Strongest risks for positive HCV included sharing drug-use equipment and younger age at first heroin use; receiving drug-education during incarceration was protective. A ROC curve fit to the prediction yielded an area under the curve of 0.77. Predictive variables of HCV in the present analysis can be obtained via screening by CHW. Identification of patients most at risk of HCV within community settings can maximize treatment utilization. [ABSTRACT FROM AUTHOR]

Background and Aims: Cognitive impairments are primary hallmarks symptoms of bipolar disorder (BD). Whether these deficits are markers of vulnerability or symptoms of the disease is still unclear. This study used a component-wise gradient (CGB) machine learning algorithm to identify cognitive measures that could accurately differentiate pediatric BD, unaffected offspring of BD parents, and healthy controls. Methods: 59 healthy controls (HC; 11.19 ± 3.15 yo; 30 girls), 119 children and adolescents with BD (13.31 ± 3.02 yo, 52 girls) and 49 unaffected offspring of BD parents (UO; 9.36 ± 3.18 yo; 22 girls) completed the CANTAB cognitive battery. Results: CGB achieved accuracy of 73.2% and an AUROC of 0.785 in classifying individuals as either BD or non-BD on a dataset held out for validation for testing. The strongest cognitive predictors of BD were measures of processing speed and affective processing. Measures of cognition did not differentiate between UO and HC. Conclusions: Alterations in processing speed and affective processing are markers of BD in pediatric populations. Longitudinal studies should determine whether UO with a cognitive profile similar to that of HC are at less or equal risk for mood disorders. Future studies should include relevant measures for BD such as verbal memory and genetic risk scores. [ABSTRACT FROM AUTHOR]

Background: Depressive symptoms may contribute to cocaine use. However, tests of the relationship between depression and severity of cocaine use have produced mixed results, possibly due to heterogeneity in individual symptoms of depression. Our goal was to establish which symptoms of depression are most strongly related to frequency of cocaine use (one aspect of severity) in a large sample of current cocaine users. We utilized generalized additive modeling to provide data-driven exploration of the relationships between depressive symptoms and cocaine use, including examination of non-linearity. We hypothesized that symptoms related to anhedonia would demonstrate the strongest relationship to cocaine use.Method: 772 individuals screened for cocaine use disorder treatment studies. To measure depressive symptoms, we used the items of the Beck Depression Inventory, 2nd Edition. Cocaine use frequency was measured as proportion of self-reported days of cocaine use over the last 30 days using the Addiction Severity Index.Results: Models identified 18 significant predictors of past-30-day cocaine use. The strongest predictors were Crying, Pessimism, Changes in Appetite, Indecisiveness, and Loss of Interest. Noteworthy effect sizes were found for specific response options on Suicidal Thoughts, Worthlessness, Agitation, Concentration Difficulty, Tiredness, and Self Dislike items.Conclusions: The strongest predictors did not conform to previously hypothesized "subtypes" of depression. Non-linear relationships between items and use were typical, suggesting BDI-II items may not be monotonically increasing ordinal measures with respect to predicting cocaine use. Qualitative analysis of strongly predictive response options suggested emotional volatility and disregard for the future as important predictors of use. [ABSTRACT FROM AUTHOR]

Highlights • Machine learning was used to optimize modeling of patient aggressive events in a large set of electronic health records in a safety net psychiatric facility. • The best-performing algorithm (penalized generalized linear modeling) achieved an area under the curve = 0.7801. • The strongest predictors of patient aggressive events were homelessness, having witnessed abuse, and prior assault conviction. • A cost-optimized probability threshold of an aggressive event was generated to assist with allocation of hospital resources. Abstract Recent advances in data science were used capitalize on the extensive quantity of data available in electronic health records to predict patient aggressive events. This retrospective study utilized electronic health records (N = 29,841) collected between January 2010 and December 2015 at Harris County Psychiatric Center, a 274-bed safety net community psychiatric facility. The primary outcome of interest was the presence (1.4%) versus absence (98.6%) of an aggressive event toward staff or patients. The best-performing algorithm, penalized generalized linear modeling, achieved an area under the curve = 0.7801. The strongest predictors of patient aggressive events included homelessness (b = 0.52), having been convicted of assault (b = 0.31), and having witnessed abuse (b = −0.28). The algorithm was also used to generate a cost-optimized probability threshold (6%) for an aggressive event, theoretically affording individualized hospital-staff coverage on the 2.8% of inpatients at highest risk for aggression, based on available hospital operating costs. The present research demonstrated the utility of a data science approach to better understand a high-priority event in psychiatric inpatient settings. [ABSTRACT FROM AUTHOR]

Background: Immune system abnormalities have been repeatedly observed in several psychiatric disorders, including severe depression and anxiety. However, whether specific immune mediators play an early role in the etiopathogenesis of these disorders remains unknown.Methods: In a longitudinal design, component-wise gradient boosting was used to build models of depression, assessed by the Mood-Feelings Questionnaire-Child (MFQC), and anxiety, assessed by the Screen for Child Anxiety Related Emotional Disorders (SCARED) in 254 adolescents from a large set of candidate predictors, including sex, race, 39 inflammatory proteins, and the interactions between those proteins and time. Each model was reduced via backward elimination to maximize parsimony and generalizability.Results: Component-wise gradient boosting and model reduction found that female sex, growth- regulated oncogene (GRO), and transforming growth factor alpha (TGF-alpha) predicted depression, while female sex predicted anxiety.Limitations: Differential onset of puberty as well as a lack of control for menstrual cycle may also have been responsible for differences between males and females in the present study. In addition, investigation of all possible nonlinear relationships between the predictors and the outcomes was beyond the computational capacity and scope of the present research.Conclusions: This study highlights the need for novel statistical modeling to identify reliable biological predictors of aberrant psychological behavior. [ABSTRACT FROM AUTHOR]

Given datasets with a large or diverse set of predictors of aggression, machine learning (ML) provides efficient tools for identifying the most salient variables and building a parsimonious statistical model. ML techniques permit efficient exploration of data, have not been widely used in aggression research, and may have utility for those seeking prediction of aggressive behavior. Objectives: The present study examined predictors of aggression and constructed an optimized model using ML techniques. Predictors were derived from a dataset that included demographic, psychometric and genetic predictors, specifically FK506 binding protein 5 (FKBP5) polymorphisms, which have been shown to alter response to threatening stimuli, but have not been tested as predictors of aggressive behavior in adults. Methods: The data analysis approach utilized component-wise gradient boosting and model reduction via backward elimination to: (a) select variables from an initial set of 20 to build a model of trait aggression; and then (b) reduce that model to maximize parsimony and generalizability. Results: From a dataset of N = 47 participants, component-wise gradient boosting selected 8 of 20 possible predictors to model Buss-Perry Aggression Questionnaire (BPAQ) total score, with R2 = 0.66. This model was simplified using backward elimination, retaining six predictors: smoking status, psychopathy (interpersonal manipulation and callous affect), childhood trauma (physical abuse and neglect), and the FKBP5_13 gene (rs1360780). The six-factor model approximated the initial eight-factor model at 99.4% of R2. Conclusions: Using an inductive data science approach, the gradient boosting model identified predictors consistent with previous experimental work in aggression; specifically psychopathy and trauma exposure. Additionally, allelic variants in FKBP5 were identified for the first time, but the relatively small sample size limits generality of results and calls for replication. This approach provides utility for the prediction of aggression behavior, particularly in the context of large multivariate datasets. [ABSTRACT FROM AUTHOR]

Research indicates that high utilizers of the health care system are more likely to have mental illness, to be from socially disadvantaged groups, and to have limited access to community-based services. In this retrospective study, three definitions of high utilization were examined: (1) across time: non-high utilization versus high-utilization, (2) single year versus multi-year, and (3) year-to-year. Univariate logistic regression models were fit to a set of 20 theory-selected predictors of high utilization. An optimal multiple predictor model was then derived via penalized multiple logistic regression (via elastic net, a machine learning algorithm). Three factors were identified in the optimized model as increasing the likelihood of high utilization: having a diagnosis of schizophrenia, having a co-occurring personality disorder diagnosis, and having less than a high school education. Given the complex needs of psychiatric high utilizers, innovative approaches should be considered to improve patient outcomes and reduce costly psychiatric hospitalizations. [ABSTRACT FROM AUTHOR]

Adults experiencing homelessness are more likely to have an alcohol use disorder compared to adults in the general population. Although shelter-based treatments are common, completion rates tend to be poor, suggesting a need for more effective approaches that are tailored to this understudied and underserved population. One barrier to developing more effective treatments is the limited knowledge of the triggers of alcohol use among homeless adults. This paper describes the use of ecological momentary assessment (EMA) to identify predictors of "imminent drinking" (i.e., drinking within the next 4 h), among a sample of adults experiencing homelessness and receiving health services at a homeless shelter. A total of 78 mostly male (84.6%) adults experiencing homelessness (mean age = 46.6) who reported hazardous drinking completed up to five EMAs per day over 4 weeks (a total of 4557 completed EMAs). The study used machine learning techniques to create a drinking risk algorithm that predicted 82% of imminent drinking episodes within 4 h of the first drink of the day, and correctly identified 76% of nondrinking episodes. The algorithm included the following 7 predictors of imminent drinking: urge to drink, having alcohol easily available, feeling confident that alcohol would improve mood, feeling depressed, lower commitment to being alcohol free, not interacting with someone drinking alcohol, and being indoors. The research team used the results to develop intervention content (e.g., brief tailored messages) that will be delivered when imminent drinking is detected in an upcoming intervention phase. Specifically, we created three theoretically grounded message tracks focused on urge/craving, social/availability, and negative affect/mood, which are further tailored to a participant's current drinking goal (i.e., stay sober, drink less, no goal) to support positive change. To our knowledge, this is the first study to develop tailored intervention messages based on likelihood of imminent drinking, current drinking triggers, and drinking goals among adults experiencing homelessness. [ABSTRACT FROM AUTHOR]