Examining Passively Collected Smartphone-Based Data in the Days Prior to Psychiatric Hospitalization for a Suicidal Crisis: Comparative Case Analysis

BACKGROUND

Digital phenotyping has seen a broad increase in application across clinical research; however, little research has implemented passive assessment approaches for suicide risk detection. There is a significant potential for a novel form of digital phenotyping, termed screenomics, which captures smartphone activity via screenshots.

OBJECTIVE

This paper focuses on a comprehensive case review of 2 participants who reported past 1-month active suicidal ideation, detailing their passive (ie, obtained via screenomics screenshot capture) and active (ie, obtained via ecological momentary assessment [EMA]) risk profiles that culminated in suicidal crises and subsequent psychiatric hospitalizations. Through this analysis, we shed light on the timescale of risk processes as they unfold before hospitalization, as well as introduce the novel application of screenomics within the field of suicide research.

METHODS

To underscore the potential benefits of screenomics in comprehending suicide risk, the analysis concentrates on a specific type of data gleaned from screenshots-text-captured prior to hospitalization, alongside self-reported EMA responses. Following a comprehensive baseline assessment, participants completed an intensive time sampling period. During this period, screenshots were collected every 5 seconds while one's phone was in use for 35 days, and EMA data were collected 6 times a day for 28 days. In our analysis, we focus on the following: suicide-related content (obtained via screenshots and EMA), risk factors theoretically and empirically relevant to suicide risk (obtained via screenshots and EMA), and social content (obtained via screenshots).

RESULTS

Our analysis revealed several key findings. First, there was a notable decrease in EMA compliance during suicidal crises, with both participants completing fewer EMAs in the days prior to hospitalization. This contrasted with an overall increase in phone usage leading up to hospitalization, which was particularly marked by heightened social use. Screenomics also captured prominent precipitating factors in each instance of suicidal crisis that were not well detected via self-report, specifically physical pain and loneliness.

CONCLUSIONS

Our preliminary findings underscore the potential of passively collected data in understanding and predicting suicidal crises. The vast number of screenshots from each participant offers a granular look into their daily digital interactions, shedding light on novel risks not captured via self-report alone. When combined with EMA assessments, screenomics provides a more comprehensive view of an individual's psychological processes in the time leading up to a suicidal crisis.

On the Selection of Item Scores or Composite Scores for Clinical Prediction

Recent shifts to prioritize prediction, rather than explanation, in psychological science have increased applications of predictive modeling methods. However, composite predictors, such as sum scores, are still commonly used in practice. The motivations behind composite test scores are largely intertwined with reducing the influence of measurement error in answering explanatory questions. But this may be detrimental for predictive aims. The present paper examines the impact of utilizing composite or item-level predictors in linear regression. A mathematical examination of the bias-variance decomposition of prediction error in the presence of measurement error is provided. It is shown that prediction bias, which may be exacerbated by composite scoring, drives prediction error for linear regression. This may be particularly salient when composite scores are comprised of heterogeneous items such as in clinical scales where items correspond to symptoms. With sufficiently large training samples, the increased prediction variance associated with item scores becomes negligible even when composite scores are sufficient. Practical implications of predictor scoring are examined in an empirical example predicting suicidal ideation from various depression scales. Results show that item scores can markedly improve prediction particularly for symptom-based scales. Cross-validation methods can be used to empirically justify predictor scoring decisions.

Associations among Drug Acquisition and Use Behaviors, Psychosocial Attributes, and Opioid-Involved Overdoses: A SEM Analysis

Aims

This study sought to develop and assess an exploratory model of how demographic and psychosocial attributes, and drug use or acquisition behaviors interact to affect opioid-involved overdoses.

Methods

We conducted exploratory and confirmatory factor analysis (EFA/CFA) to identify a factor structure for ten drug acquisition and use behaviors. We then evaluated alternative structural equation models incorporating the identified factors, adding demographic and psychosocial attributes as predictors of past-year opioid overdose. We used interview data collected for two studies recruiting opioid-misusing participants receiving services from a community-based syringe service program. The first investigated current attitudes toward drug-checking (N = 150). The second was an RCT assessing a telehealth versus in-person medical appointment for opioid use disorder treatment referral (N = 270). Demographics included gender, age, race/ethnicity, education, and socioeconomic status. Psychosocial measures were homelessness, psychological distress, and trauma. Self-reported drug-related risk behaviors included using alone, having a new supplier, using opioids with benzodiazepines/alcohol, and preferring fentanyl. Past-year opioid-involved overdoses were dichotomized into experiencing none or any.

Results

The EFA/CFA revealed a two-factor structure with one factor reflecting drug acquisition and the second drug use behaviors. The selected model (CFI = .984, TLI = .981, RMSEA = .024) accounted for 13.1% of overdose probability variance. A latent variable representing psychosocial attributes was indirectly associated with an increase in past-year overdose probability (β=.234, p = .001), as mediated by the EFA/CFA identified latent variables: drug acquisition (β=.683, p < .001) and drug use (β=.567, p = .001). Drug use behaviors (β=.287, p = .04) but not drug acquisition (β=.105, p = .461) also had a significant, positive direct effect on past-year overdose. No demographic attributes were significant direct or indirect overdose predictors.

Conclusions

Psychosocial attributes, particularly homelessness, increase the probability of an overdose through associations with risky drug acquisition and drug-using behaviors. To increase effectiveness, prevention efforts might address the interacting overdose risks that span multiple functional domains.

Examining the dynamic relationship between nonsuicidal self-injury and alcohol use experiences

INTRODUCTION

Nonsuicidal self-injury (NSSI) is a prevalent and concerning behavior and its risk pathways require a greater understanding, particularly in predicting short-term risk. Although the literature has supported a between-person link among NSSI and alcohol use, limited research has directly examined the nuances of this relationship at the within-person level using intensive longitudinal data.

METHOD

Utilizing two independent samples (total n = 85), the current study examined bidirectional, concurrent and prospective risk relationships between NSSI and alcohol, considering both urges and behavior engagement, via ecological momentary assessment.

RESULTS

Findings demonstrate concurrent, within-person relationships between NSSI urges and alcohol urges, as well as alcohol use. Alternatively, prospective between-person findings demonstrated negative relationships between NSSI urges and alcohol use, as well as alcohol urges and NSSI acts; however, this may represent suppression effects as associations were positive with the removal of autoregressive effects.

CONCLUSIONS

Together, findings support proximal risk relationships between NSSI and alcohol experiences that, for urges in particular, is bidirectional.

Measurement invariance and response consistency of single-item assessments for suicidal thoughts and behaviors

The present study aimed to expand the literature on single-item assessments for suicidal thoughts and behaviors (STBs) by examining measurement invariance of commonly used single-item assessments of suicidal ideation (SI), planning (SP), and attempts (SA) with respect to race and ethnicity. Predictive invariance with respect to depression, and multi-item measures of passive and active SI were also explored. Measurement invariance was examined across (a) Black and White respondents and (b) Hispanic/Latinx and non-Hispanic/-Latinx respondents. Participants (N = 1,624; 51.66% male) were recruited from Mechanical Turk and Prime Panels. Participants were administered four distinct single-item measures each for SI, SP, and SA across three timeframes (past month, past year, lifetime). Items were drawn from well-known large-scale studies (e.g., National Comorbidity Survey) and common suicide risk assessments. Multiple group confirmatory factor analysis was used to examine measurement invariance; regression with group by measure interactions were used to evaluate predictive invariance. Measurement invariance was observed for both Black (N = 534) and White (N = 1,089) respondents as well as Hispanic/Latinx (N = 335) and non-Hispanic/-Latinx (N = 1,288) respondents across single-item outcomes. Thus, SI, SP, and SA rates can be defensibly compared between Black and White and Hispanic/Latinx and non-Hispanic/-Latinx respondents within studies; however, comparison of SI and SP rates across studies with differing assessment prompts should be met with caution. Multiple single-item STB measures demonstrated predictive bias across race and ethnicity suggesting potential differential screening capabilities. Elevated SI, SP, and SA rates for Hispanic/Latinx individuals were also observed. Findings reiterate the importance of minor language differences in single-item STB assessments. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Supervised latent Dirichlet allocation with covariates: A Bayesian structural and measurement model of text and covariates

Text is a burgeoning data source for psychological researchers, but little methodological research has focused on adapting popular modeling approaches for text to the context of psychological research. One popular measurement model for text, topic modeling, uses a latent mixture model to represent topics underlying a body of documents. Recently, psychologists have studied relationships between these topics and other psychological measures by using estimates of the topics as regression predictors along with other manifest variables. While similar two-stage approaches involving estimated latent variables are known to yield biased estimates and incorrect standard errors, two-stage topic modeling approaches have received limited statistical study and, as we show, are subject to the same problems. To address these problems, we proposed a novel statistical model-supervised latent Dirichlet allocation with covariates (SLDAX)-that jointly incorporates a latent variable measurement model of text and a structural regression model to allow the latent topics and other manifest variables to serve as predictors of an outcome. Using a simulation study with data characteristics consistent with psychological text data, we found that SLDAX estimates were generally more accurate and more efficient. To illustrate the application of SLDAX and a two-stage approach, we provide an empirical clinical application to compare the application of both the two-stage and SLDAX approaches. Finally, we implemented the SLDAX model in an open-source R package to facilitate its use and further study. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

The impact of social connection on near-term suicidal ideation

While predominant suicide theories emphasize the role of social connectedness in suicidal thinking, there is a need to better understand (a) how specific aspects of social connection relate to suicidal ideation and (b) the timeframe over which these relationships persist. The current study examined ecological momentary assessment data over a 30-day period from 35 participants with past-year suicidal thoughts or behaviors (mean age = 25.88; 62.9% women; 68.6% White) to address these questions. Results demonstrated that absence of social contact was associated with next timepoint suicidal ideation, even after considering the suicidal ideation autoregressive effect (i.e., concurrent), with effects strongest in the short-term. Findings provide preliminary evidence of the need to assess for the presence of social contact, and for assessments to occur in close proximity (i.e., a few hours), to capture the true dynamics of risk for suicidal ideation. Although needing replication, results suggest importance of just-in-time interventions targeting suicidal ideation.

A critique of using the labels confirmatory and exploratory in modern psychological research

Psychological science is experiencing a rise in the application of complex statistical models and, simultaneously, a renewed focus on applying research in a confirmatory manner. This presents a fundamental conflict for psychological researchers as more complex forms of modeling necessarily eschew as stringent of theoretical constraints. In this paper, I argue that this is less of a conflict, and more a result of a continued adherence to applying the overly simplistic labels of exploratory and confirmatory. These terms mask a distinction between exploratory/confirmatory research practices and modeling. Further, while many researchers recognize that this dichotomous distinction is better represented as a continuum, this only creates additional problems. Finally, I argue that while a focus on preregistration helps clarify the distinction, psychological research would be better off replacing the terms exploratory and confirmatory with additional levels of detail regarding the goals of the study, modeling details, and scientific method.

Comparing the role of perceived burdensomeness and thwarted belongingness in prospectively predicting active suicidal ideation

OBJECTIVE

The Interpersonal Theory of Suicide has been foundational in guiding current suicide literature. Despite recent research underscoring fluctuations of suicidal ideation within hours, there have been few studies examining the key constructs of perceived burdensomeness and thwarted belongingness within an intensive framework. Thus, the current study aimed to add cumulative knowledge regarding the within-person relationship between perceived burdensomeness, thwarted belongingness, and active suicidal ideation as assessed within an ecological momentary assessment design.

METHOD

A final sample of 35 individuals with a past-year history of suicidal thoughts or behaviors completed brief surveys four times per day for 30 days.

RESULTS

Findings highlighted that the addition of covariates may offer small improvements in modeling subsequent suicidal ideation, while controlling for SI at the prior time. Further, both thwarted belongingness and perceived burdensomeness were associated with next timepoint suicidal ideation, and their interaction added little incremental value.

CONCLUSION

Findings demonstrate the potential importance of thwarted belongingness in predicting suicidal ideation. Further, results highlight that the main effects of thwarted belongingness and perceived burdensomeness, rather than their interaction, may be more important to consider in relation to active suicidal ideation.

Explorations of Individual Change Processes and Their Determinants: A Novel Approach and Remaining Challenges

Over the past 40 years there have been great advances in the analysis of individual change and the analyses of between-person differences in change. While conditional growth models are the dominant approach, exploratory models, such as growth mixture models and structural equation modeling trees, allow for greater flexibility in the modeling of between-person differences in change. We continue to push for greater flexibility in the modeling of individual change and its determinants by combining growth mixture modeling with structural equation modeling trees to evaluate how measured covariates predict class membership using a recursive partitioning algorithm. This approach, referred to as growth mixture modeling with membership trees, is illustrated with longitudinal reading data from the Early Childhood Longitudinal Study with the MplusTrees package in R.

Evaluating the item-level factor structure of anhedonia

BACKGROUND

Anhedonia has long been theorized to be a multidimensional construct, focusing on domains of reward stimuli and temporal relationship to reward. However, little empirical work has directly examined whether there is support for this assertion.

METHODS

The study used data from young adults from four independent samples (n = 2098). Participants completed multiple measures of anhedonia.

RESULTS

We used rigorous conducted exploratory and confirmatory factor analyses on items from six commonly used anhedonia measures to examine dimensions underlying anhedonia. Results suggested a four-factor solution with factors reflecting social reward, social disinterest, status/achievement, and physical/natural reward. The identified factors reflected broad content domains of pleasure, but not specific reward processes. The four factors were modestly associated with one another, suggesting a weak common underlying anhedonia trait that manifests across multiple dimensions. Factor scores were associated with personality measures, reward-related indices, and depression symptoms, supporting the validity of the factors.

LIMITATIONS

Participants were all young adults and we assessed anhedonia only at the level of self-report.

CONCLUSION

Anhedonia is a multidimensional construct. However, the dimensions of anhedonia only distinguish domains of, but not temporal processes of anhedonia. Future work should continue to refine the structures underlying the construct of anhedonia through iterative theory- and data-driven research and examine associations between anhedonia and clinical outcomes.

Examining momentary associations between behavioral approach system indices and nonsuicidal self-injury urges

BACKGROUND

The current study aimed to examine the concurrent and prospective relationships between the three hypothesized components of behavioral approach system (BAS) sensitivity: drive, reflecting the motivation to pursue one's desired goals; reward responsiveness, reflecting sensitivity to reward or reinforcement; and fun-seeking, reflecting the motivation for pursuing novel rewards in a spontaneous manner, and NSSI urge severity.

METHODS

A sample of 64 undergraduates with a history of repetitive NSSI completed an ecological momentary assessment protocol. During this period of time, participants reported on the BAS-constructs of drive, reward responsiveness, and fun-seeking, as well as NSSI urge severity on a momentary basis at three random intervals each day for a period of ten-days.

RESULTS

Drive and reward responsiveness, but not fun-seeking, were concurrently positively associated with NSSI urge severity. However, no associations between BAS facets and prospective NSSI urges were found.

LIMITATIONS

This study was limited by its use of single items to assess the BAS-constructs of drive, reward responsiveness, and fun-seeking.

CONCLUSIONS

Our findings indicate that feeling strongly impacted by rewards and having a strong sense of drive toward goal attainment may represent cognitive risk states that are associated with increased within-person NSSI risk. However, their lack of prospective prediction may suggest that these cognitive states are associated only on a momentary basis with NSSI urges and may not confer risk.

How we ask matters: The impact of question wording in single-item measurement of suicidal thoughts and behaviors

The present study aimed to extend prior literature on single-item assessment by examining response consistency (1) between several commonly used single-item assessments of suicidal ideation, planning, and attempts, and (2) across three timeframes (past month, past year, and lifetime) commonly employed in the literature. Participants (N = 613) were recruited from an online community, Amazon Mechanical Turk (mTurk). Participants were administered three sets of four distinct single-items assessing suicidal ideation, suicidal planning, and suicide attempt history, respecitvely. Items were drawn from well-known large-scale studies (e.g., National Comorbidity Survey; World Health Organization Mental Health Survey Initiative, Youth Risk Behavior Survey) and commonly used suicide risk assessments (i.e., Self-Injurious Thoughts and Behaviors Interview). Through examinations of intraclass correlations and confirmatory factor analyses, findings suggested mixed response agreement across most outcomes and timeframes. Response inconsistency among items assessing suicidal ideation and among items assessing suicidal planning were partly attributed to minor, yet important, language differences. Given findings that even minor language changes in suicidal ideation and planning items may inflate or restrict prevalence estimates in a meaningful way, it will be important for researchers and clinicians alike to pay close attention to the wording of single items in designing research studies, interpreting findings, and assessing patient risk.

A prospective examination of COVID-19-related social distancing practices on suicidal ideation

BACKGROUND

The COVID-19 pandemic has spurred the implementation of several public safety measures to contain virus spread, most notably socially distancing policies. Prior research has linked similar public safety measures (i.e., quarantine) with suicide risk, in addition to supporting the role of social connection in suicidal thoughts and behaviors; consequently, there is a need to better understand the relationship between widespread social distancing policies and suicide risk. The current study aimed to examine the prospective association between COVID-19-related social distancing practices and suicidal ideation.

METHODS

Participants (N = 472) completed measures of suicidal ideation and impacts of social distancing practices at baseline and two weeks later.

RESULTS

After controlling for general psychosocial distress (i.e., depression, social connectedness), cross-lagged regression models identified prospective, bidirectional relationships between perceived impacts of social distancing on one's mental health and both passive and active suicidal ideation. The impact of social distancing on work/social routine was not associated with suicidal ideation.

CONCLUSIONS

Overall, findings suggest the importance of an individual's perception regarding the effect of social distancing on their mental health, rather than the disruption to work or social routine, in suicide risk. Findings highlight potential targets for suicide risk prevention and intervention.

Reliable Trees: Reliability Informed Recursive Partitioning for Psychological Data

Recursive partitioning, also known as decision trees and classification and regression trees (CART), is a machine learning procedure that has gained traction in the behavioral sciences because of its ability to search for nonlinear and interactive effects, and produce interpretable predictive models. The recursive partitioning algorithm is greedy-searching for the variable and the splitting value that maximizes outcome homogeneity. Thus, the algorithm can be overly sensitive to chance associations in the data, particularly in small samples. In an effort to limit chance associations, we propose and evaluate a reliability-based cost function for recursive partitioning. The reliability-based cost function increases the likelihood of selecting variables that are more reliable, which should have more consistent associations with the outcome of interest. Two reliability-based cost functions are proposed, evaluated through simulation, and compared to the CART algorithm. Results indicate that reliability-based cost functions can be beneficial, particularly with smaller samples and when more reliable variables are important to the prediction, but can overlook important associations between the outcome and lower reliability predictors. The use of these cost functions was illustrated using data on depression and suicidal ideation from the National Longitudinal Survey of Youth.

Future Time Perspective in Mid-to-Later Life: The Role of Personality

OBJECTIVES

Future time perspective (FTP), or the way individuals orient to and consider their futures, is fundamental to motivation and well-being across the life span. There is a relative paucity of studies, however, that explore its contributing factors in mid-to-later life, specifically. Therefore, uncovering which variables contribute to individual differences in FTP, as well as the ways these variables interact, is paramount to developing a strong understanding of this construct during this life-span stage.

METHOD

This study used three data mining techniques (ie, elastic net, decision tree, and tree ensemble analyses) to simultaneously test several potential contributors identified in the literature, including the five-factor personality domains, several health indices, and age.

RESULTS

Personality, especially neuroticism, extraversion, conscientiousness, and agreeableness, had the most influence on FTP. Age and health were not among the most salient FTP contributors in mid-to-later life. Furthermore, decision tree analyses uncovered interactive effects of personality; several profiles of neuroticism, extraversion, and/or conscientiousness were linked with differing FTP levels.

DISCUSSION

Although the extant literature has indicated that FTP, age, and health are inextricably tied, these results indicate that there is more variability to be explained in FTP, perhaps especially when looking within specific age groups.

Preliminary investigation of the association between COVID-19 and suicidal thoughts and behaviors in the U.S

Evidence suggests that the negative consequences of COVID-19 may extend far beyond its considerable death toll, having a significant impact on psychological well-being. Despite work highlighting the link between previous epidemics and elevated suicide rates, there is limited research on the relationship between the COVID-19 pandemic and suicidal thoughts and behaviors. Utilizing an online survey, the current study aimed to better understand the presence, and extent, of the association between COVID-19-related experiences and past-month suicidal thoughts and behaviors among adults in the United States recruited via Amazon Mechanical Turk (n = 907). Results support an association between several COVID-19-related experiences (i.e., general distress, fear of physical harm, effects of social distancing policies) and past-month suicidal ideation and suicide attempts. Further, a significant proportion of those with recent suicidal ideation explicitly link their suicidal thoughts to COVID-19. Exploratory analyses highlight a potential additional link between COVID-19 and suicidal behavior, suggesting that a portion of individuals may be intentionally exposing themselves to the virus with intent to kill themselves. These findings underscore the need for suicide risk screening and access to mental health services during the current pandemic. Particular attention should be paid to employing public health campaigns to disseminate information on such services to reduce the enormity of distress and emotional impairment associated with COVID-19 in the United States.

The use of text-based responses to improve our understanding and prediction of suicide risk

OBJECTIVE

Text-based responses may provide significant contributions to suicide risk prediction, yet research including text data is limited. This may be due to a lack of exposure and familiarity with statistical analyses for this data structure.

METHOD

The current study provides an overview of data processing and statistical algorithms for text data, guided by an empirical example of 947 online participants who completed both open-ended items and traditional self-report measures. We give an introduction to a number of text-based statistical approaches, including dictionary-based methods, topic modeling, word embeddings, and deep learning.

RESULTS

We analyze responses from the open-ended question "How do you feel today?", detailing characteristics of the responses, as well as predicting past-year suicidal ideation.

CONCLUSIONS

We see the analysis of text from social media, open-ended questions, and other text sources (i.e., medical records) as an important form of complementary assessment to traditional scales, shedding insight on what we are missing in our current set of questionnaires, which may ultimately serve to improve both our understanding and prediction of suicide.

Using ordinal regression for advancing the understanding of distinct suicide outcomes

OBJECTIVE

As recent advances in suicide research have underscored the importance of studying distinct suicide outcomes (i.e., suicidal thinking vs. behavior), there is a need to consider the theoretical meaningfulness of our statistical approach(es). As an alternative to more popular statistical methods, we introduce ordinal regression, detailing specific forms that are well-aligned to examine outcomes specific to suicide research.

METHOD

Ordinal regression models allow for assessment of the influences of covariates on the experience of lower (i.e., suicidal ideation) to higher (i.e., suicidal planning) suicide risk outcomes.

RESULTS

As an empirical application, we fit a sequential ordinal regression model with 17 theoretically selected covariates and modeled category specific effects for each covariate.

CONCLUSIONS

Results detailed from depression and presence of nonsuicidal self-injury demonstrate the utility of ordinal regression in consideration of transitions across suicide outcomes. Ordinal regression models may be particularly informative in identifying risk factors unique to each suicide outcome, which has the potential to meaningfully inform theoretical models of suicide and suicide risk prediction.

Regularized structural equation modeling with stability selection

Regularization methods such as the least absolute shrinkage and selection operator (LASSO) are commonly used in high dimensional data to achieve sparser solutions. Recently, methods such as regularized structural equation modeling (SEM) and penalized likelihood SEM have been proposed, trying to transfer the benefits of regularization to models commonly used in social and behavioral research. These methods allow researchers to estimate large models even in the presence of small sample sizes. However, some drawbacks of the LASSO, such as high false positive rates (FPRs) and inconsistency in selection results, persist at the same time. We propose the application of stability selection, a method based on repeated resampling of the data to select stable coefficients, to regularized SEM as a mechanism to overcome these limitations. Across 2 simulation studies, we find that stability selection greatly improves upon the LASSO in selecting the correct paths, specifically through reducing the number of false positives. We close the article by demonstrating the application of stability selection in 2 empirical examples and presenting several future research directions. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Examining the factor structure of the Big Five Inventory-2 personality domains with an adolescent sample

We examined the within-domain structure of the five domains of personality measured by the Big Five Inventory-2 with data collected from an adolescent sample (N = 838). Three possible factor models were tested: a single factor, correlated facets, and a single factor with correlated residuals. We examined each model controlling for acquiescence, a response bias in which respondents tend to agree/disagree regardless of item content, using two approaches: acquiescence factor and within-person centering of item-level responses. Across each domain, results indicated both the correlated facets and correlated residuals models demonstrated acceptable fit. Accounting for acquiescent responding was generally associated with improved model fit. However, consistent with past struggles in measuring open-mindedness in adolescents, the correlated residuals model with acquiescence as a factor for open-mindedness failed to converge. Regularized structural equation modeling was conducted on this model for open-mindedness and suggested certain residual covariances that contributed to estimation difficulties should be constrained to zero. Advantages of models are discussed with implications for studying the Big Five personality domains in adolescents. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Machine Learning and Psychological Research: The Unexplored Effect of Measurement

Machine learning (i.e., data mining, artificial intelligence, big data) has been increasingly applied in psychological science. Although some areas of research have benefited tremendously from a new set of statistical tools, most often in the use of biological or genetic variables, the hype has not been substantiated in more traditional areas of research. We argue that this phenomenon results from measurement errors that prevent machine-learning algorithms from accurately modeling nonlinear relationships, if indeed they exist. This shortcoming is showcased across a set of simulated examples, demonstrating that model selection between a machine-learning algorithm and regression depends on the measurement quality, regardless of sample size. We conclude with a set of recommendations and a discussion of ways to better integrate machine learning with statistics as traditionally practiced in psychological science.

Exploratory Mediation Analysis of Dichotomous Outcomes via Regularization

Exploratory mediation analysis via regularization, or XMed, is a recently developed technique that allows one to identify potential mediators of a process of interest. However, as currently implemented, it can only be applied to continuous outcomes. We extend this method to allow application to dichotomous outcomes, including both mediators and dependent variables. Simulation results show that XMed can achieve the same sensitivity as more conventional methods for mediation analysis such as the Sobel test, percentile bootstrap, and bias-corrected bootstrap, but in general requires only half the sample size to do so. We demonstrate the implementation of this approach using an illustrative example examining the relationship between youth behavioral/emotional problems and alcohol use.

P6417Increasing the accuracy of a machine learning algorithm for STEMI diagnosis by incorporating demographic variables

Background

Our previous work demonstrated the diagnostic value of Artificial Intelligence (AI) -driven algorithms for ST-Elevation Myocardial Infarction (STEMI). In the present research, we explore the importance of demographic data inclusion, in order to achieve a more accurate diagnosis.

Purpose

To demonstrate that incorporation of demographic variables into the sample records will augment the accuracy of AI-based protocols for STEMI diagnosis.

Methods

An observational, retrospective, case-control study. Demographic data (age and gender) male/female ratio 1.3, ages 98–18 years was added to the sample records. Sample: 8,511 EKG records, previously diagnosed as normal, abnormal (over 200 conditions) or STEMI. Records excluded other patient and medical information. The sample was derived from the private International Telemedical Systems (ITMS) database. LUMENGT-AI Algorithm was employed. Preprocessing: detection of QRS complexes by wavelet system, segmentation of each EKG into individual heartbeats (90,592 total beats) with fixed window of 0.4s to the left and 0.9s to the right of main QRS; Classification: A 1-D convolutional neural network was implemented, “STEMI” and “Not-STEMI” classes were considered for each heartbeat, individual probabilities were aggregated to generate the final label for each record. Training & Testing: 90% and 10% of the sample was used, respectively. Experiments: Intel PC i7 8750H processor at 2.21GHz, 16GB RAM, Windows 10 OS with Nvidia GTX 1070GPU, 8GB RAM.

Results

The model yielded an accuracy of 97.1%, a sensitivity of 96.8%, and a specificity of 97.5%.

Conclusions

The ability of AI-guided algorithms to diagnose STEMI is increased by expanding the morphological variables with demographic data. This approach may be applied to improve the EKG diagnosis of other cardiovascular entities and improve clinical management.

P2426Validating the diagnostic value of a machine learning algorithm for STEMI detection

Background

We have previously reported the use of Artificial Intelligence (AI) guided EKG analysis for detection of ST-Elevation Myocardial Infarction (STEMI). To demonstrate the diagnostic value of our algorithm, we compared AI predictions with reports that were confirmed as STEMI.

Purpose

To demonstrate the absolute proficiency of AI for detecting STEMI in a standard12-lead EKG.

Methods

An observational, retrospective, case-control study. Sample: 5,087 EKG records, including 2,543 confirmed STEMI cases obtained via feedback from health centers following appropriate patient management (thrombolysis, primary Percutaneous Coronary Intervention (PCI), pharmacoinvasive therapy or coronary artery bypass surgery). Records excluded patient and medical information. The sample was derived from the International Telemedical Systems (ITMS) database. LUMENGT-AI Algorithm was employed. Preprocessing: detection of QRS complexes by wavelet system, segmentation of each EKG into individual heartbeats (53,667 total beats) with fixed window of 0.4s to the left and 0.9s to the right of main QRS; Classification: A 1-D convolutional neural network was implemented, “STEMI” and “Not-STEMI” classes were considered for each heartbeat, individual probabilities were aggregated to generate the final label for each record. Training & Testing: 90% and 10% of the sample were used, respectively. Experiments: Intel PC i7 8750H processor at 2.21GHz, 16GB RAM, Windows 10 OS with NVIDIA GTX 1070 GPU, 8GB RAM.

Results

The model yielded an accuracy of 97.2%, a sensitivity of 95.8%, and a specificity of 98.5%.

Conclusion(s)

Our AI-based algorithm can reliably diagnose STEMI and will preclude the role of a cardiologist for screening and diagnosis, especially in the pre-hospital setting.

An examination of individual forms of nonsuicidal self-injury

Nonsuicidal self-injury (NSSI) is a growing public health concern, and there is an increasing need to better characterize and identify severe NSSI behavior. One readily accessible, yet understudied, avenue for improving the assessment of NSSI severity is through the examination of individual forms, or methods, of the behavior. The present study aimed to address this gap in the literature by investigating the relationship between 12 different NSSI methods with three NSSI severity indicators and three distinct suicidal thoughts and behaviors among 1,436 undergraduate students with a history of NSSI (70.90% female, M age = 20.69, SD = 3.32). Results across six decision tree analyses highlighted the use self-hitting / punching, in addition to cutting oneself, as the most informative NSSI methods for differentiating outcome severity. Gender differences were only found for the outcome of suicidal ideation. The present study provides preliminary evidence that the examination of individual NSSI methods may be useful in identifying individuals at risk for negative correlates of NSSI, including NSSI-related hospital visits, unintended serious injury, and suicidal behavior. Upon replication in longitudinal work, findings have important clinical utility by providing a potential marker of prognosis and the need for higher levels of care.

A Practical Guide to Variable Selection in Structural Equation Models with Regularized MIMIC Models

Methodological innovations have allowed researchers to consider increasingly sophisticated statistical models that are better in line with the complexities of real world behavioral data. However, despite these powerful new analytic approaches, sample sizes may not always be sufficiently large to deal with the increase in model complexity. This poses a difficult modeling scenario that entails large models with a comparably limited number of observations given the number of parameters. We here describe a particular strategy to overcoming this challenge, called regularization. Regularization, a method to penalize model complexity during estimation, has proven a viable option for estimating parameters in this small n, large p setting, but has so far mostly been used in linear regression models. Here we show how to integrate regularization within structural equation models, a popular analytic approach in psychology. We first describe the rationale behind regularization in regression contexts, and how it can be extended to regularized structural equation modeling (Jacobucci, Grimm, & McArdle, 2016). Our approach is evaluated through the use of a simulation study, showing that regularized SEM outperforms traditional SEM estimation methods in situations with a large number of predictors and small sample size. We illustrate the power of this approach in two empirical examples: modeling the neural determinants of visual short term memory, as well as identifying demographic correlates of stress, anxiety and depression. We illustrate the performance of the method and discuss practical aspects of modeling empirical data, and provide a step-by-step online tutorial.

The use of machine learning in the study of suicidal and non-suicidal self-injurious thoughts and behaviors: A systematic review

BACKGROUND

Machine learning techniques offer promise to improve suicide risk prediction. In the current systematic review, we aimed to review the existing literature on the application of machine learning techniques to predict self-injurious thoughts and behaviors (SITBs).

METHOD

We systematically searched PsycINFO, PsycARTICLES, ERIC, CINAHL, and MEDLINE for articles published through February 2018.

RESULTS

Thirty-five articles met criteria to be included in the review. Included articles were reviewed by outcome: suicide death, suicide attempt, suicide plan, suicidal ideation, suicide risk, and non-suicidal self-injury. We observed three general aims in the use of SITB-focused machine learning analyses: (1) improving prediction accuracy, (2) identifying important model indicators (i.e., variable selection) and indicator interactions, and (3) modeling underlying subgroups. For studies with the aim of boosting predictive accuracy, we observed greater prediction accuracy of SITBs than in previous studies using traditional statistical methods. Studies using machine learning for variable selection purposes have both replicated findings of well-known SITB risk factors and identified novel variables that may augment model performance. Finally, some of these studies have allowed for subgroup identification, which in turn has helped to inform clinical cutoffs.

LIMITATIONS

Limitations of the current review include relatively low paper sample size, inconsistent reporting procedures resulting in an inability to compare model accuracy across studies, and lack of model validation on external samples.

CONCLUSIONS

We concluded that leveraging machine learning techniques to further predictive accuracy and identify novel indicators will aid in the prediction and prevention of suicide.

Reconsidering important outcomes of the nonsuicidal self-injury disorder diagnostic criterion A

OBJECTIVE

The nonsuicidal self-injury (NSSI) disorder diagnostic criteria have been the focus of empirical study. However, Criterion A (i.e., required frequency and timeframe) has received relatively limited attention. The current study aimed to examine the relationship between past 12-month NSSI frequency and eight NSSI behavior features among individuals with past 12-month and 1-month NSSI.

METHOD

Participants were 723 undergraduate students reporting at least 1 past 12-month NSSI act and completed online questionnaires. Decision trees and structural equation model trees were utilized to examine the relationship between NSSI frequency and behavior features.

RESULTS

Results highlight several potential subgroups: high (i.e., greater than 49 acts), moderate-to-high (i.e., 19-48 acts), low-to-moderate (i.e., 7-18 acts), and low (i.e., fewer than 6 acts) frequency subgroups.

CONCLUSIONS

Findings suggest that increasing the NSSI disorder criterion A frequency cutoff or requiring at least one past month NSSI act may better demarcate individuals with more severe NSSI behavior.

Using Exploratory Data Mining to Identify Important Correlates of Nonsuicidal Self-Injury Frequency

OBJECTIVE

Non-suicidal self-injury (NSSI) has been linked to many adverse outcomes, with more frequent NSSI increasing the likelihood of impairment, severity, and more serious self-harming behavior (e.g., suicidality; Andover & Gibb, 2010; Darke et al., 2010). Despite the determined importance of NSSI frequency in understanding the severity of one's behavior, there is still a need to identify which constructs may be influential in predicting frequency. The current study aimed to fill this gap by identifying which correlates are most important in relation to NSSI frequency through two exploratory data mining methods.

METHOD

Seven hundred twelve undergraduate students with a history of NSSI completed self-report measures of NSSI behavior, suicidality, cognitive-affective deficits, and psychopathology symptomology.

RESULTS

Both exploratory data mining methods, lasso regression and random forests, demonstrated number of NSSI methods to be the factor with the most importance in relation to lifetime NSSI frequency. Once this variable was removed, suicide plan and depressive symptomology were significant correlates across methods.

CONCLUSIONS

The current findings support the literature between NSSI frequency and NSSI methods, but also implicate suicide plans, an often-overlooked factor, and depression in NSSI severity.

Recursive Partitioning with Nonlinear Models of Change

In this article, we introduce nonlinear longitudinal recursive partitioning (nLRP) and the R package longRpart2 to carry out the analysis. This method implements recursive partitioning (also known as decision trees) in order to split data based on individual- (i.e., cluster) level covariates with the goal of predicting differences in nonlinear longitudinal trajectories. At each node, a user-specified linear or nonlinear mixed-effects model is estimated. This method is an extension of Abdolell et al.'s (2002) longitudinal recursive partitioning while permitting a nonlinear mixed-effects model in addition to a linear mixed-effects model in each node. We give an overview of recursive partitioning, nonlinear mixed-effects models for longitudinal data, describe nLRP, and illustrate its use with empirical data from the Early Childhood Longitudinal Study-Kindergarten Cohort.

Identifying the relative importance of non-suicidal self-injury features in classifying suicidal ideation, plans, and behavior using exploratory data mining

Individuals with a history of non-suicidal self-injury (NSSI) are at alarmingly high risk for suicidal ideation (SI), planning (SP), and attempts (SA). Given these findings, research has begun to evaluate the features of this multi-faceted behavior that may be most important to assess when quantifying risk for SI, SP, and SA. However, no studies have examined the wide range of NSSI characteristics simultaneously when determining which NSSI features are most salient to suicide risk. The current study utilized three exploratory data mining techniques (elastic net regression, decision trees, random forests) to address these gaps in the literature. Undergraduates with a history of NSSI (N = 359) were administered measures assessing demographic variables, depression, and 58 NSSI characteristics (e.g., methods, frequency, functions, locations, scarring) as well as current SI, current SP, and SA history. Results suggested that depressive symptoms and the anti-suicide function of NSSI were the most important features for predicting SI and SP. The most important features in predicting SA were the anti-suicide function of NSSI, NSSI-related medical treatment, and NSSI scarring. Overall, results suggest that NSSI functions, scarring, and medical lethality may be more important to assess than commonly regarded NSSI severity indices when ascertaining suicide risk.

The Relationship Between Nonsuicidal Self-Injury Age of Onset and Severity of Self-Harm

This study examined how age of nonsuicidal self-injury (NSSI) onset relates to NSSI severity and suicidality using decision tree analyses (nonparametric regression models that recursively partition predictor variables to create groupings). Those with an earlier age of NSSI onset reported greater NSSI frequency, NSSI methods, and NSSI-related hospital visits. No significant splits were found for suicide ideation or attempts, although those with an earlier onset were more likely to have a suicide plan. Overall, findings suggest that onset of NSSI before age 12 is associated with more severe NSSI and may be a crucial age for prevention efforts.

Comparison of frequentist and Bayesian regularization in structural equation modeling

Research in regularization, as applied to structural equation modeling (SEM), remains in its infancy. Specifically, very little work has compared regularization approaches across both frequentist and Bayesian estimation. The purpose of this study was to address just that, demonstrating both similarity and distinction across estimation frameworks, while specifically highlighting more recent developments in Bayesian regularization. This is accomplished through the use of two empirical examples that demonstrate both ridge and lasso approaches across both frequentist and Bayesian estimation, along with detail regarding software implementation. We conclude with a discussion of future research, advocating for increased evaluation and synthesis across both Bayesian and frequentist frameworks.

Exploratory Mediation Analysis via Regularization

Exploratory mediation analysis refers to a class of methods used to identify a set of potential mediators of a process of interest. Despite its exploratory nature, conventional approaches are rooted in confirmatory traditions, and as such have limitations in exploratory contexts. We propose a two-stage approach called exploratory mediation analysis via regularization (XMed) to better address these concerns. We demonstrate that this approach is able to correctly identify mediators more often than conventional approaches and that its estimates are unbiased. Finally, this approach is illustrated through an empirical example examining the relationship between college acceptance and enrollment.

A Comparison of Methods for Uncovering Sample Heterogeneity: Structural Equation Model Trees and Finite Mixture Models

Although finite mixture models have received considerable attention, particularly in the social and behavioral sciences, an alternative method for creating homogeneous groups, structural equation model trees (Brandmaier, von Oertzen, McArdle, & Lindenberger, 2013), is a recent development that has received much less application and consideration. It is our aim to compare and contrast these methods for uncovering sample heterogeneity. We illustrate the use of these methods with longitudinal reading achievement data collected as part of the Early Childhood Longitudinal Study-Kindergarten Cohort. We present the use of structural equation model trees as an alternative framework that does not assume the classes are latent and uses observed covariates to derive their structure. We consider these methods as complementary and discuss their respective strengths and limitations for creating homogeneous groups.

Development and validation of empirically derived frequency criteria for NSSI disorder using exploratory data mining

Research suggesting nonsuicidal self-injury (NSSI) may belong in a distinct diagnostic category has led to the inclusion of NSSI disorder in the DSM-5 section for future study. There has been limited research, however, examining the validity of Criterion A (the frequency criterion). The current study aimed to examine the validity of the frequency criterion of NSSI disorder through the use of an exploratory data mining method, structural equation modeling trees, as a way to determine a NSSI frequency that optimally discriminates pathological NSSI from normative behavior among undergraduate students (n = 3,559), 428 who engaged in NSSI in the previous year. The model included psychopathology symptomology found to be comorbid with NSSI and cognitive-affective deficits commonly associated with NSSI. Results demonstrated a first split between individuals with 0 and 1 act of NSSI in the past year, as was expected. Among individuals with 1 or more previous acts, the optimal split was between those with 5 and 6 NSSI acts in the past year. Results from the current study suggest that individuals with 6 acts of NSSI in past year, compared with those with 5 acts or less, may represent a more severe group of self-injurers. These individuals reported higher levels of related psychopathology symptomology and cognitive-affective deficits, in addition to decreased quality of life. Findings have potential implications for the proposed frequency criteria of NSSI disorder and how pathological NSSI is characterized. (PsycINFO Database Record

Regularized Structural Equation Modeling

A new method is proposed that extends the use of regularization in both lasso and ridge regression to structural equation models. The method is termed regularized structural equation modeling (RegSEM). RegSEM penalizes specific parameters in structural equation models, with the goal of creating easier to understand and simpler models. Although regularization has gained wide adoption in regression, very little has transferred to models with latent variables. By adding penalties to specific parameters in a structural equation model, researchers have a high level of flexibility in reducing model complexity, overcoming poor fitting models, and the creation of models that are more likely to generalize to new samples. The proposed method was evaluated through a simulation study, two illustrative examples involving a measurement model, and one empirical example involving the structural part of the model to demonstrate RegSEM's utility.

Using Classification and Regression Trees (CART) and random forests to analyze attrition: Results from two simulations

In this article, we describe a recent development in the analysis of attrition: using classification and regression trees (CART) and random forest methods to generate inverse sampling weights. These flexible machine learning techniques have the potential to capture complex nonlinear, interactive selection models, yet to our knowledge, their performance in the missing data analysis context has never been evaluated. To assess the potential benefits of these methods, we compare their performance with commonly employed multiple imputation and complete case techniques in 2 simulations. These initial results suggest that weights computed from pruned CART analyses performed well in terms of both bias and efficiency when compared with other methods. We discuss the implications of these findings for applied researchers.