Evidence for mood instability in patients with bipolar disorder: Applying multilevel hidden Markov modelling to intensive longitudinal ecological momentary assessment data

Abstract

Background. Bipolar disorder (BD) is a chronic psychiatric condition characterized by large shifts in mood, energy, and cognitive functioning. Recently, the conceptualization of BD has shifted from alternating discrete episodes to a chronic cyclical mood instability model. Recog-nizing and quantifying this mood instability may improve care and calls for high-frequency measures coupled with advanced statistical models. Methods. To uncover empirically derived mood states, a multilevel hidden Markov model (HMM) was applied to 4-month ecological momentary assessment (EMA) data in twenty pa-tients with BD, yielding ~9.820 assessments in total. EMA data comprised self-reported ques-tionnaires (5 per day) measuring manic and depressive constructs using 12 items. Manic and depressive symptoms were further assessed by weekly self-reported questionnaires (i.e., Alt-man Self-Rating Mania Scale and Quick Inventory for Depressive Symptomatology Self-Report). Alignment between uncovered mood states and weekly questionnaires was assessed with a multilevel linear model.Results. HMM uncovered four mood states (euthymic, manic, mixed, and depressive) that aligned with weekly symptom scores. States switched more frequently than weekly data sug-gested. The average state duration was 17.1h (SD=78.0) for euthymic, 16.3h (SD=70.8) for depressive, 9.1h (SD=11.6) for manic, and 8.8h (SD=10.9) for mixed. In almost half of the patients, significant mood instability was observed. Large individual differences were observed in state duration and switching.Conclusions. The results indicate that mood instability is a key feature of BD, which should be considered in theoretical and clinical conceptualizations of the disorder.

Evidence for mood instability in patients with bipolar disorder: Applying multilevel hidden Markov modeling to intensive longitudinal ecological momentary assessment data.

“…Importantly, the total sample size (Number of Participants × Number of Assessments) is comparable with previous studies that explore switches in latent mood states of patients with BD (e.g., Cochran et al, 2016;Prisciandaro et al, 2019;Yee et al, 2021). Indeed, previous research in MHMMs has demonstrated that parameters are accurately estimated with smaller samples when the number of assessments is sufficiently large (McClintock, 2021;Mildiner Moraga & Aarts, 2024). However, clinically, the sample was small and quite heterogeneous (e.g., both BD Type I and II, males and females, with different comorbidities and medication types), and participants were selected based on a clinical history of frequent manic or depressive episodes.…”

Towards a Generative Model for Emotion Dynamics

“…However, many other models or descriptive statistics can be used to capture relevant features of the time series. In this context, an interesting class of models that has not been widely used in the emotion dynamics literature are Hidden Markov Models (HMMs), which model time series as generated by unobserved latent time-varying states (for a recent application in the context of emotion time series, see Mildiner Moraga et al, 2023). The number and nature of latent states found by fitting HMMs to empirical data could potentially be used to inform the situation part of our generative model, or to fine-tune the modeling of emotion variables as consisting of distinct states.…”