Episode forecasting in bipolar disorder: Is energy better than mood?

Machine learning applied to the prediction of relapse, hospitalization, and suicide in bipolar disorder using neuroimaging and clinical data: A systematic review

BACKGROUND

Bipolar disorder (BD) is associated with increased morbidity/mortality. Adverse outcome prediction might help with the management of patients with BD.

METHODS

We systematically reviewed the performance of machine learning (ML) studies in predicting adverse outcomes (relapse or recurrence, hospital admission, and suicide-related events) in patients with BD. Demographic, clinical, and neuroimaging-related poor outcome predictors were also reviewed. Three databases (PubMed, Scopus, and Web of Science) were explored from inception to July 2023.

RESULTS

Eighteen studies, accounting for >30,000 patients, were included. Support vector machine, decision trees, random forest, and logistic regression were the most frequently used ML algorithms. ML models' area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, and specificity ranged from 0.71 to 0.98, 72.7-92.8 %, and 59.0-95.2 % for relapse/recurrence prediction (4 studies (3 on relapses and 1 on recurrences). The corresponding values were 0.78-0.88, 21.4-100 %, and 77.0-99.7 % for hospital admissions (3 studies, 21,266 patients), and 0.71-0.99, 44.4-97.9 %, and 38.9-95.0 % for suicide-related events (10 studies, 5558 patients). Also, one study addressed a combination of the interest outcomes. Adverse outcome predictors included early onset BD, BD type I, comorbid psychiatric or substance use disorder, circadian rhythm disruption, hospitalization characteristics, and neuroimaging parameters, including increased dynamic amplitude of low-frequency fluctuation, decreased frontolimbic functional connectivity and aberrant dynamic functional connectivity in corticostriatal circuitry.

CONCLUSIONS

ML models can predict adverse outcomes of BD with relatively acceptable performance measures. Future studies with larger samples and nested cross-validation validation should be conducted to reach more reliable results.

Prediction of patient admission and readmission in adults from a Colombian cohort with bipolar disorder using artificial intelligence

Introduction

Bipolar disorder (BD) is a chronically progressive mental condition, associated with a reduced quality of life and greater disability. Patient admissions are preventable events with a considerable impact on global functioning and social adjustment. While machine learning (ML) approaches have proven prediction ability in other diseases, little is known about their utility to predict patient admissions in this pathology.

Aim

To develop prediction models for hospital admission/readmission within 5 years of diagnosis in patients with BD using ML techniques.

Methods

The study utilized data from patients diagnosed with BD in a major healthcare organization in Colombia. Candidate predictors were selected from Electronic Health Records (EHRs) and included sociodemographic and clinical variables. ML algorithms, including Decision Trees, Random Forests, Logistic Regressions, and Support Vector Machines, were used to predict patient admission or readmission. Survival models, including a penalized Cox Model and Random Survival Forest, were used to predict time to admission and first readmission. Model performance was evaluated using accuracy, precision, recall, F1 score, area under the receiver operating characteristic curve (AUC) and concordance index.

Results

The admission dataset included 2,726 BD patients, with 354 admissions, while the readmission dataset included 352 patients, with almost half being readmitted. The best-performing model for predicting admission was the Random Forest, with an accuracy score of 0.951 and an AUC of 0.98. The variables with the greatest predictive power in the Recursive Feature Elimination (RFE) importance analysis were the number of psychiatric emergency visits, the number of outpatient follow-up appointments and age. Survival models showed similar results, with the Random Survival Forest performing best, achieving an AUC of 0.95. However, the prediction models for patient readmission had poorer performance, with the Random Forest model being again the best performer but with an AUC below 0.70.

Conclusion

ML models, particularly the Random Forest model, outperformed traditional statistical techniques for admission prediction. However, readmission prediction models had poorer performance. This study demonstrates the potential of ML techniques in improving prediction accuracy for BD patient admissions.

Identifying patient-specific behaviors to understand illness trajectories and predict relapses in bipolar disorder using passive sensing and deep anomaly detection: protocol for a contactless cohort study

BACKGROUND

Predictive models for mental disorders or behaviors (e.g., suicide) have been successfully developed at the level of populations, yet current demographic and clinical variables are neither sensitive nor specific enough for making individual clinical predictions. Forecasting episodes of illness is particularly relevant in bipolar disorder (BD), a mood disorder with high recurrence, disability, and suicide rates. Thus, to understand the dynamic changes involved in episode generation in BD, we propose to extract and interpret individual illness trajectories and patterns suggestive of relapse using passive sensing, nonlinear techniques, and deep anomaly detection. Here we describe the study we have designed to test this hypothesis and the rationale for its design.

METHOD

This is a protocol for a contactless cohort study in 200 adult BD patients. Participants will be followed for up to 2 years during which they will be monitored continuously using passive sensing, a wearable that collects multimodal physiological (heart rate variability) and objective (sleep, activity) data. Participants will complete (i) a comprehensive baseline assessment; (ii) weekly assessments; (iii) daily assessments using electronic rating scales. Data will be analyzed using nonlinear techniques and deep anomaly detection to forecast episodes of illness.

DISCUSSION

This proposed contactless, large cohort study aims to obtain and combine high-dimensional, multimodal physiological, objective, and subjective data. Our work, by conceptualizing mood as a dynamic property of biological systems, will demonstrate the feasibility of incorporating individual variability in a model informing clinical trajectories and predicting relapse in BD.

Motor activity patterns can distinguish between interepisode bipolar disorder patients and healthy controls

BACKGROUND

Bipolar disorder (BD) is linked to circadian rhythm disruptions resulting in aberrant motor activity patterns. We aimed to explore whether motor activity alone, as assessed by longitudinal actigraphy, can be used to classify accurately BD patients and healthy controls (HCs) into their respective groups.

METHODS

Ninety-day actigraphy records from 25 interepisode BD patients (ie, Montgomery-Asberg Depression Rating Scale (MADRS) and Young Mania Rating Scale (YMRS) < 15) and 25 sex- and age-matched HCs were used in order to identify latent actigraphic biomarkers capable of discriminating between BD patients and HCs. Mean values and time variations of a set of standard actigraphy features were analyzed and further validated using the random forest classifier.

RESULTS

Using all actigraphy features, this method correctly assigned 88% (sensitivity = 85%, specificity = 91%) of BD patients and HCs to their respective group. The classification success may be confounded by differences in employment between BD patients and HCs. When motor activity features resistant to the employment status were used (the strongest feature being time variation of intradaily variability, Cohen's d = 1.33), 79% of the subjects (sensitivity = 76%, specificity = 81%) were correctly classified.

CONCLUSION

A machine-learning actigraphy-based model was capable of distinguishing between interepisode BD patients and HCs solely on the basis of motor activity. The classification remained valid even when features influenced by employment status were omitted. The findings suggest that temporal variability of actigraphic parameters may provide discriminative power for differentiating between BD patients and HCs while being less affected by employment status.

The futility of long-term predictions in bipolar disorder: mood fluctuations are the result of deterministic chaotic processes

BACKGROUND

Understanding the underlying architecture of mood regulation in bipolar disorder (BD) is important, as we are starting to conceptualize BD as a more complex disorder than one of recurring manic or depressive episodes. Nonlinear techniques are employed to understand and model the behavior of complex systems. Our aim was to assess the underlying nonlinear properties that account for mood and energy fluctuations in patients with BD; and to compare whether these processes were different in healthy controls (HC) and unaffected first-degree relatives (FDR). We used three different nonlinear techniques: Lyapunov exponent, detrended fluctuation analysis and fractal dimension to assess the underlying behavior of mood and energy fluctuations in all groups; and subsequently to assess whether these arise from different processes in each of these groups.

RESULTS

There was a positive, short-term autocorrelation for both mood and energy series in all three groups. In the mood series, the largest Lyapunov exponent was found in HC (1.84), compared to BD (1.63) and FDR (1.71) groups [F (2, 87) = 8.42, p < 0.005]. A post-hoc Tukey test showed that Lyapunov exponent in HC was significantly higher than both the BD (p = 0.003) and FDR groups (p = 0.03). Similarly, in the energy series, the largest Lyapunov exponent was found in HC (1.85), compared to BD (1.76) and FDR (1.67) [F (2, 87) = 11.02; p < 0.005]. There were no significant differences between groups for the detrended fluctuation analysis or fractal dimension.

CONCLUSIONS

The underlying nature of mood variability is in keeping with that of a chaotic system, which means that fluctuations are generated by deterministic nonlinear process(es) in HC, BD, and FDR. The value of this complex modeling lies in analyzing the nature of the processes involved in mood regulation. It also suggests that the window for episode prediction in BD will be inevitably short.

DSM-5 and ICD-11 criteria for bipolar disorder: Implications for the prevalence of bipolar disorder and validity of the diagnosis - A narrative review from the ECNP bipolar disorders network

This narrative review summarizes and discusses the implications of the Diagnostic and Statistical Manual of Mental Disorders (DSM)-5 and the upcoming International Classification of Diseases (ICD)-11 classification systems on the prevalence of bipolar disorder and on the validity of the DSM-5 diagnosis of bipolar disorder according to the Robin and Guze criteria of diagnostic validity. Here we review and discuss current data on the prevalence of bipolar disorder diagnosed according to DSM-5 versus DSM-IV, and data on characteristics of bipolar disorder in the two diagnostic systems in relation to extended Robin and Guze criteria: 1) clinical presentation, 2) associations with para-clinical data such as brain imaging and blood-based biomarkers, 3) delimitation from other disorders, 4) associations with family history / genetics, 5) prognosis and long-term follow-up, and 6) treatment effects. The review highlights that few studies have investigated consequences for the prevalence of the diagnosis of bipolar disorder and for the validity of the diagnosis. Findings from these studies suggest a substantial decrease in the point prevalence of a diagnosis of bipolar with DSM-5 compared with DSM-IV, ranging from 30-50%, but a smaller decrease in the prevalence during lifetime, corresponding to a 6% reduction. It is concluded that it is likely that the use of DSM-5 and ICD-11 will result in diagnostic delay and delayed early intervention in bipolar disorder. Finally, we recommend areas for future research.

How Does Adding the DSM-5 Criterion Increased Energy/Activity for Mania Change the Bipolar Landscape?

According to DSM-IV, the criterion (A) for diagnosing hypomanic/manic episodes is mood change (i.e., elevated, expansive or irritable mood). Criterion (A) was redefined in DSM-5 in 2013, adding increased energy/activity in addition to mood change. This paper examines a potential change of prevalence data for bipolar I or II when adding increased energy/activity to the criterion (A) for the diagnosis of hypomania/mania. Own research suggests that the prevalence of manic/hypomanic episodes drops by at least one third when using DSM-5 criteria. Whether this has positive or negative impact on clinical practice and research still needs further evaluation.

Big data in severe mental illness: the role of electronic monitoring tools and metabolomics

There is an increasing interest in the development of effective early detection and intervention strategies in severe mental illness (SMI). Ideally, these efforts should lead to the delineation of accurate staging models of SMI enabling personalized interventions. It is plausible that big data approaches will be instrumental in describing the developmental trajectories of SMI by facilitating the incorporation of data from multiple sources, including those pertaining to the biological make-up of affected subjects. In this review, we first aimed to offer a perspective on how big data are helping the delineation of personalized approaches in SMI, and, second, to offer a quantitative synthesis of big data approaches in metabolomics of SMI. We finally described future directions of this research area.

Correlates, Course, and Outcomes of Increased Energy in Youth with Bipolar Disorder

OBJECTIVES

Compare longitudinal trajectories of youth with Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV Bipolar Disorder (BD), grouped at baseline by presence/absence of increased energy during their worst lifetime mood episode (required for DSM-5).

METHODS

Participants from the parent Course and Outcome of Bipolar Youth study (N = 446) were assessed utilizing The Schedule for Affective Disorders and Schizophrenia for School-Age Children (KSADS), KSADS Mania Rating Scale (KMRS), and KSADS Depression Rating Scale (KDRS). Youth were grouped at baseline into those with increased energy (meeting DSM-5 Criteria A for mania) vs. without increased energy (meeting DSM-IV, but not DSM-5, Criteria A for mania), for those who had worst lifetime mood episode recorded (n = 430). Youth with available longitudinal data had the presence/absence of increased energy measured, as well as psychiatric symptomatology/clinical outcomes (evaluated via the Adolescent Longitudinal Interval Follow-Up Evaluation), at each follow-up for 12.5 years (n = 398).

RESULTS

At baseline, the increased energy group (based on endorsed increased energy during worst lifetime mood episode; 86% of participants) vs. the without increased energy group, were more likely to meet criteria for BD-I and BD Not Otherwise Specified, had higher KMRS/KDRS total scores, and displayed poorer family/global psychosocial functioning. However, frequency of increased energy between groups was comparable after 5 years, and no significant group differences were found on clinical/psychosocial functioning outcomes after 12.5 years.

LIMITATIONS

Secondary data limited study design; groupings were based on one time point.

CONCLUSIONS

Results indicate no clinically relevant longitudinal group differences.

Modelling mood disorders: An ACE solution?

OBJECTIVES

The treatment of mood disorders remains sub-optimal. A major reason for this is our lack of understanding of the underlying pathophysiology of depression and bipolar disorder. A core problem is the lack of specificity of our current diagnoses. This paper discusses the history of this problem and posits a solution in the form of a more sophisticated model.

METHOD

The authors review the notable historical works that laid the foundations of mood disorder nosology; discuss the more recent influences that shaped modern diagnoses; and examine the evidence that mood disorders are characterised by multidimensional and longitudinal symptom profiles.

RESULTS

The ACE model considers mood disorders as a combination of symptoms across three domains: Activity, Cognition, and Emotion; that vary over time. This multidimensional and longitudinal perspective is consistent with the prevalence of complex clinical presentations, such as mixed states, and highlights the importance of recurrence in mood disorders.

CONCLUSIONS

The ACE model encourages researchers to characterise patients from a number of equally important perspectives and, by doing so, add specificity to the treatment of mood disorders.