Table 2.
Description of models and corresponding data for verification.
| Study | DD or TD | What it Models | Model type or method used | Data used to test model |
|---|---|---|---|---|
| Huber et al. [21] | TD | Episodes (binary occurrence vs. absence) | Deterministic and stochastic dynamical systems | None |
| Mohan [35] | TD | Mood oscillations/variations (untreated vs. treated) | Deterministic dynamical system | None |
| Conte et al. [11] | TD | Mood oscillations/variations (“latent” and “acclaimed phases”) Deterministic/stochastic contributions to mood variations in BD vs. healthy control. | Deterministic and stochastic dynamical systems | Qualitative description of the results of Gottschalk et al. [47] |
| Daugherty et al. [37] | TD | Mood oscillations/variations (treated vs. untreated, interactions between two patients with BD) | Deterministic dynamical system | None |
| Nana [34] | TD | Mood oscillations/variations (treated vs. untreated) | Deterministic dynamical system | None |
| Goldbeter [38] | TD | Mania and depression as independent, interacting systems. Mood oscillations/transitions (effect of antidepressants simulated) | Deterministic dynamical system | None |
| Bonsall et al. [28] | DD | Time-series of mood variability in stable and unstable BD (by fitting linear and nonlinear AR models to data) | Fitting linear and nonlinear AR models to data | QIDS-SR time-series (one measure per week over a 220-week period from 23 individuals with BD, divided into “stable mood” (n = 11) and “unstable mood” (n = 12) |
| Frank [41] | TD | Oscillations in second messenger systems | Deterministic dynamical system | None |
| Goldbeter [42] | TD | Mania and depression as independent, interacting systems. Mood oscillations/transitions (effect of antidepressants simulated) | Deterministic and stochastic dynamical systems | None |
| Hadaeghi, et al. [36] | TD | Mood oscillations/variations (treated vs. untreated) | Deterministic dynamical system | None |
| Steinacher & Wright [10] | TD | Time-course of behavioral activation/approach in BD, using both deterministic and stochastic systems | Deterministic and stochastic dynamical systems | Qualitative description of results from Wright et al. [50] |
| Koutsoukos & Angelopolous [51] | TD | Energy (mood) oscillations/variations generated from a theoretical mood “pendulum” (effect of mood-stabilizers considered) | Deterministic dynamical system | None |
| Bonsall et al. [39] | DD + TD | Time-series of mood variability (by fitting linear and threshold AR models to time-series data). Mood fluctuations using both deterministic and stochastic dynamical systems (relaxation oscillators fit to time-series data) | Fitting linear and threshold AR models to data Deterministic and stochastic dynamical systems | QIDS-SR time-series from 61 individuals with BD (one measure per week for 79–233 weeks). n = 42 used for AR models, n = 19 for relaxation oscillator models. |
| Ortiz et al. [26] | DD | Time-series of mood, anxiety and energy in BD vs. healthy control (by fitting AR models to time-series data). | Fitting AR models to time-series data | Time-series data of self-reported mood, anxiety and energy levels using a visual analog scale from 30 individuals with BD, and 30 healthy controls, (two measures per day, for 90 days) |
| Cochran et al. [44] | DD | Clinical course of BD by fitting discrete-time Markov chain model with discretized mood states to longitudinal data. | Discrete-time Markov chain model | Data from the Prechter Longitudinal Study of Bipolar Disorder at the University of Michigan [58] (n = 209 individuals with bipolar I disorder) |
| Hadaeghi et al. [52] | TD | Circadian activity variation in BD | Deterministic dynamical system | Actigraphic data from n=15 subjects, but model not fit to group level data, and comparisons between model output and data are shown for single subject only. |
| Bayani et al. [31] | TD | Circadian activity pulse trains in BD | Deterministic dynamical system | None |
| Cochran et al. [40] | DD + TD | Mood variations | Patient-level statistics to test a set of hypotheses, followed by a proposed stochastic dynamical system | Self-report ASRM and Patient Health Questionnaire for Depression (PHQ-9), collected every 2 months from 178 individuals with BD, for at least 4 years |
| Chang & Chou [53] | TD | Relationship between mood sensitivity and realized/expected value. Simulated QIDS-SR16 scores. | Deterministic dynamical system | None |
| Cochran et al. [78] | TD | Time-course of mood variations in BD using stochastic models | Stochastic dynamical systems | None |
| Ortiz et al. [27] | DD | Time-series of mood, anxiety and energy in BD, unaffected first-degree relatives, and healthy controls (by fitting AR models to time-series data). | Fitting AR models to time-series data | Time-series data of self-reported mood, anxiety and energy levels using a visual analog scale (two measures per day, for 90 days) in 30 individuals with BD, 30 unaffected first-degree relatives and 30 healthy controls |
| Prisciandaro et al. [43] | DD | Empirically-derived mood states and transition probabilities in BD (using hidden Markov modeling) | Hidden Markov modeling | Longitudinal data from STEP-BD study [79] (n = 3918 for transition probability analyses, n=3229 for analyses involving baseline covariates) |
| Doho et al. [33] | TD | Neural activity related to circadian function in BD and the effect of chronotherapy on neuronal activity | Deterministic dynamical system | None |
| Nobukawa et al. [32] | TD | Frontal neural activity and circadian activity in BD and healthy control, and effect of chronotherapy | Deterministic and stochastic dynamical systems | None |
| Moore et al. [30] | DD | Forecasting time-series of QIDS-SR scores in BD | Fitting statistical models to time-series data. Forecasting using AR, exponential smoothing, Gaussian process regression | QIDS-SR and ASRM time-series from 100 individuals with BD (one measure per week for 3.5 years). Only QIDS-SR scores were used for forecasting. |
| Moore et al. [29] | DD | Forecasting time-series of QIDS-SR scores in BD | Fitting statistical models to time-series data. Linear and nonlinear forecasting using: persistence, exponential smoothing, AR, gaussian process regression, locally constant prediction, local linear prediction | QIDS-SR time-series from eight individuals with BD (one measure per week for 5 years) |
AR autoregressive, ASRM Altman self-rating mania scale, BD bipolar disorder, DD data-driven, PHQ-9 patient health questionnaire, QIDS-SR quick inventory of depressive symptoms, STEP-BD systematic treatment enhancement program for bipolar disorder, TD theory-driven.