Abstract

BACKGROUND: The path to depressive symptom improvement during therapy is often complex, as many individuals experience periods of instability and discontinuous symptom change. If the process of remission follows complex dynamic systems principles, early warning signals (EWS) may precede such depressive symptom transitions.

AIMS: We aimed to test whether EWS, in the form of rises in lag-1 autocorrelation and variance, occur in momentary affect time series preceding transitions towards lower levels of depressive symptoms during therapy. We also investigated the presence of EWS in patients without symptom transitions.

METHODS: In a sample of 41 depressed individuals who were starting psychological treatment, positive affect and negative affect (high and low arousal) were measured five times a day using ecological momentary assessments (EMA) for four months (521 observations per individual on average; yielding 25,197 observations in total), and depressive symptoms were assessed weekly over six months. We used a moving window method and time-varying autoregressive generalized additive modeling (TV-AR GAM) to determine whether EWS occurred in these momentary affect measures, within-persons.

RESULTS: For the moving-window autocorrelation, 89% of individuals with transitions showed at least one EWS in one of the variables (versus 62.5% in the no-transition group), and the proportion of EWS in the separate variables was consistently higher (~44% across affect measures) than for individuals without transitions (~27%). Rising variance was found for few individuals, both preceding transitions (~11%) and for individuals without a transition (~12%).

CONCLUSIONS: The process of symptom remission showed critical slowing down in at least part of our sample. Our findings indicate that EWS are not generic across all affect measures and may have limited value as a personalized prediction method.

Keywords: critical slowing down; depression; idiographic change; replicated single-subject design; depressive symptom improvement; early warning signals; critical transitions

Abstract

Empirical evidence is mounting that monitoring momentary experiences for the presence of early warning signals (EWS) may allow for personalized predictions of meaningful symptom shifts in psychopathology. Studies aiming to detect EWS require intensive longitudinal measurement designs that center on individuals undergoing change. We recommend that researchers: (a) define criteria for relevant symptom shifts a priori to allow specific hypothesis testing; (b) balance the observation period length and high-frequency measurements with participant burden by testing ambitious designs with pilot studies; (c) choose variables that are meaningful to their patient group and facilitate replication by others. Thoroughly considered designs are necessary to assess the promise of EWS as a clinical tool to detect, prevent or encourage impending symptom changes in psychopathology.

Keywords: Psychopathology, ecological momentary assessment, symptom change, early warning signals, critical transitions

Abstract

OBJECTIVE: We examined individual overall trajectories of change and the occurrence of sudden gains in daily self-rated problem severity and the relation of these patterns to treatment response.

METHOD: Mood disorder patients (N = 329, mean age = 44, 55% women) completed daily self-ratings about the severity of their complaints as a standard part of treatment, using the Therapy Process Questionnaire (TPQ). Per individual, the best-fitting defined (linear, log-linear, 1-step) trajectory was tested for significance: for change over time, and for specificity of the best-fitting trajectory. Two-hundred and three cases had ICD-10 Symptom Rating (ISR) depression scores posttreatment: a score ≤1 identified 114 treatment responders. Relation to response was examined for sudden gains and type of change trajectory.

RESULTS: 138 cases (42%) had a significant decrease in problem severity, of which 54 cases (16%) had a defined trajectory: 50 cases with one-step improvement, and 4 with a linear improvement in daily problem severity. Sudden gains occurred in 28% of the total sample, and within 58% of improvement patterns. Specifically, sudden gains occurred in 68% of significant 1-step trajectories and 25% of the linear cases. Sudden gains and nonspecific change trajectories were significantly more frequent for treatment responders.

CONCLUSIONS: At the day-level, patterns of improvement are nonlinear for most patients. Sudden gains occur within various forms of overall change and are associated with treatment response. Clinically relevant improvements in depression occur both gradually and abruptly, and this finding allows for the possibility that the remission process functions according to dynamical systems principles.

Keywords: depression; treatment response; idiographic change pattern; sudden gains; daily assessment

Abstract

Whereas sudden gains and losses (large shifts in symptom severity) in patients receiving psychotherapy appear abrupt and hence may seem unexpected, hypotheses from complex-systems theory suggest that sudden gains and losses are actually preceded by certain early-warning signals (EWSs). We tested whether EWSs in patients’ daily self-ratings of the psychotherapeutic process predicted future sudden gains and losses. Data were collected from 328 patients receiving psychotherapy for mood disorders who completed daily self-ratings about their therapeutic process using the Therapy Process Questionnaire (TPQ). Sudden gains and losses were classified from the Problem Intensity scale of the TPQ. The other items of the TPQ were used to compute the EWSs. EWSs predicted an increased probability for sudden gains and losses in a 4-day predictive window. These results show that EWSs can be used for real-time prediction of sudden gains and losses in clinical practice.

Keywords: early-warning signals; sudden gains; mood disorders; complex systems; psychotherapy; open materials; preregistered

Abstract

More instability (MSSD) and variability (SD) of negative affect (NA) have been related to current and future depressive symptoms. We investigated whether NA instability and variability were predictive of the rate of symptom improvement during treatment and of reaching remission status. Forty-six individuals with major depressive disorder completed six days of ecological momentary assessments (10 beeps/day) before starting a combination of pharmacotherapy and supportive therapy. During and after treatment, the Hamilton Depression Rating Scale (HDRS) diagnostic interview was performed monthly for 18 months. The rate of change in HDRS scores over five months (during treatment) and remission status (HDRS ≤7) over eighteen months were predicted by the SD, MSSD and Mean of NA, and relevant baseline predictors. Higher Mean NA and baseline symptom severity, but not NA variability or instability, predicted stronger depressive symptom reduction during treatment. Higher odds of remitting were only associated with lower Mean NA. Higher mean NA may indicate more room for improvement and thus for a steeper rate of symptom change, while lower mean NA levels may imply that remission status is more readily reached. To resolve the inconclusive findings for instability and variability as predictors of symptom improvement replication in larger samples is required.

Keywords: emotion dynamics; depression; ecological momentary assessment; symptom improvement; treatment response; remission; destabilization; instability; variability; negative affect

Abstract

Identifying relevant symptom shifts as they enfold can be challenging, as the time period over which they take place is not uniform for all people. This paper proposes an adaptation of the well-established Reliable Change Index (RCI) that allows researchers and clinicians to explore the presence of symptom changes of varying durations in individual patients’ time series: the Duration-adjusted RCI (DaRCI). The DaRCI takes the RCI cut-off score for change between two points as a starting point, and proportionally extends this measure over multiple observations, while maintaining reliability at a given confidence level. Researchers must choose the relevant time period between two observations, and additional increments are added accordingly.

To illustrate the ability of this method to detect changes of various durations, simulated depressive symptom time series with varying degrees of discontinuity and overall mean change in scores were used. The results show that the DaRCI thresholds over two, three and four observations were effective at identifying the simulated change periods over multiple time points, starting from relatively gradual change slopes (picking up reliable changes in 20-60% of simulated time series if the overall change was large enough), to highly discontinuous changes (up to 100% accuracy).

The DaRCI may be particularly useful for identifying shifts in symptoms that appear relatively abrupt, which can help indicate when a patient is showing significant improvement or deterioration. Its ease of use makes it suitable for application in the clinical context, and is a promising method to explore different change durations in clinical populations.

Keywords: duration of change; reliable change index (RCI); intensive longitudinal data; routine outcome measurements; symptom change; repeated assessments; within-person; threshold method; simulation

Abstract

Based on dynamical systems theory, the current study aimed to investigate if recurrence of depression is systematically preceded by within-person early warning signals (EWS) in positive and negative affect. Ecological momentary assessments were collected 5 times a day for a period of 4 months (averaging 524 assessments per individual) in 37 formerly depressed individuals discontinuing antidepressant medication. EWS (increases in window autocorrelation and variance) preceded recurrence of depression in 32.9% of the participants across robustness checks. Compared to participants that remained in remission, participants with a recurrence showed (1) significantly more positive trends in the variance, but not in autocorrelation, and (2) the average number of significant EWS was over three times larger across tested affect variables. Although the results provide the first systematic evidence that EWS occur more often before the recurrence of depression, the low sensitivity of EWS poses a substantial challenge for clinical applications.

Keywords: early warning signals, critical slowing down, dynamical systems theory, experience sampling methodology, depression, critical transitions

Abstract

Identifying relevant symptom shifts as they enfold can be challenging, as the time period over which they take place is not uniform for all people. This paper proposes an adaptation of the well-established Reliable Change Index (RCI) that allows researchers and clinicians to explore the presence of symptom changes of varying durations in individual patients’ time series: the Duration-adjusted RCI (DaRCI). The DaRCI takes the RCI cut-off score for change between two points as a starting point, and proportionally extends this measure over multiple observations, while maintaining reliability at a given confidence level. Researchers must choose the relevant time period between two observations, and additional increments are added accordingly.

To illustrate the ability of this method to detect changes of various durations, simulated depressive symptom time series with varying degrees of discontinuity and overall mean change in scores were used. The results show that the DaRCI thresholds over two, three and four observations were effective at identifying the simulated change periods over multiple time points, starting from relatively gradual change slopes (picking up reliable changes in 20-60% of simulated time series if the overall change was large enough), to highly discontinuous changes (up to 100% accuracy).

The DaRCI may be particularly useful for identifying shifts in symptoms that appear relatively abrupt, which can help indicate when a patient is showing significant improvement or deterioration. Its ease of use makes it suitable for application in the clinical context, and is a promising method to explore different change durations in clinical populations.

Keywords: duration of change; reliable change index (RCI); intensive longitudinal data; routine outcome measurements; symptom change; repeated assessments; within-person; threshold method; simulation

Preregistration

Brief background:

Early Warning Signals (EWS) are theoretical indicators of critical slowing down and low resilience in a complex dynamical system, where a rise in EWS indicates that the likelihood of an upcoming critical transition is higher. EWS have shown promise in anticipating large shifts in depressive symptoms in a few studies already. However, no study so far has tested whether rises in autocorrelation at lag-1 (AR(1)) and variance (SD) can be found within-persons prior to depressive symptom improvement.

Examining this is clinically and theoretically important, as it could provide patient and therapist valuable insight into whether the system is sensitive to positive changes (is destabilizing), and may be close to recovery, even if it may not be overtly noticeable in the symptoms yet. Moreover, it takes the next step from viewing mental disorder from a complex systems angle, to actually testing whether the principles of dynamical systems apply to transitions in psychological systems.

Research question:

Can we find indicators of critical slowing down (CSD) in the form of early warning signals (EWS) like increases in AR(1) and variance before transitions towards improvement in depressive symptoms?

Hypotheses:

• Does the autoregressive coefficient at lag-1 (AR(1)), in moment-to-moment mood assessments, increase before a sudden shift in depressive symptoms on the SCL-90?

• Does variability (SD), in momentary mood measurements, increase before a sudden shift in depressive symptoms on the SCL-90?

Description

AIM: The Transitions in Depression (TRANS-ID) Recovery study has gathered intensive longitudinal data in a group of individuals with depression during psychological therapy. The study was designed to gather high-resolution time series to monitor individual change processes in great detail, to allow for personalized predictions of shifts in depressive symptoms.

METHOD: The data collection combined experience sampling methods to assess momentary affect and behavior (five 27-item questionnaires a day, for four months), ambulatory assessment of physical activity and heart rate (continuous, for four months), and depression symptom assessments (weekly for six months, and monthly for the six months thereafter, twelve months total). In addition to a baseline diagnostic interview, baseline questionnaires covered a range of constructs, including overall psychopathology symptoms, medication use, psychological treatment history, alexithymia, life events, quality of life, and chronotype. After the four-month ambulatory assessment period, a personal report of the experience sampling data was given to each participant, and a semi-structured qualitative interview was conducted to evaluate participants’ own retrospective experience of symptom changes during the research period.

CONCLUSION: The TRANS-ID Recovery study procedures and materials are described in detail in this document. The study protocol was approved by the Medical Ethical Committee of the University Medical Center Groningen (reg. number: NL58848.04.16).

Preregistration

Brief background:

When a dynamical system approaches a tipping point, the stabilizing forces tend to decrease and the rate at which the system can recover from perturbations slows down (a phenomenon known as Critical Slowing Down). Early Warning Signals (EWS) are theoretical indicators of Critical Slowing Down and low resilience in a complex dynamical system, where a rise in EWS indicates that the likelihood of an upcoming critical transition is higher.

Two earlier studies by Wichers and colleagues have demonstrated the potential of EWS for anticipating large shifts from a low level of depressive symptoms towards a higher level of depressive symptoms in individual patients. We call such shifts ‘transitions in depression’. In these studies, mood was measured prospectively using the Experience Sampling Method (ESM), multiple times a day over the course of several months, starting when depression was still in remission, and continuing until after a large increase in depressive symptoms occurred. This data was uniquely able to test for the presence of EWS in psychology, as it was the first to capture both the transition in depression itself and the period that led up to it with intensive longitudinal data. As hypothesized, a rise in EWS was found at the within-person level for both patients that experienced a transition in depression; such rises were absent in the majority of the five patients that did not experience a transition.

As only a total of seven participants divided over two studies have been tested so far, there is a pressing need to investigate if EWS can indeed anticipate transitions in depression in a much larger sample. In the current study, similar intensive longitudinal data has been obtained for 56 participants. The aim is to test if lag-1 autocorrelation and variance in ESM time series increase before these transitions occur, and if these same EWS are absent in participants who did not experience a transition in depression. Transitions in depression were defined for this sample in a previous study.

Research question:

Can we find Early Warning Signals (EWS) that indicate Critical Slowing Down in the form of lag-1 autocorrelation and variance in ESM data obtained in the period leading up to transitions towards higher levels of depressive symptoms?

Hypotheses:

• Lag-1 autocorrelation and variance of ESM mood scores obtained before transitions towards a higher level of depressive symptoms will show an increase that is both statistically significant and larger than a minimum effect size of interest (tau > .1).

• In patients who did not experience a transition in depressive symptoms, rises in lag-1 autocorrelation and variance that are both relevant and significant will be found less frequently than in patients who experienced a transition.

• The average magnitude of the trend in lag-1 autocorrelation and variance will be higher in the group of participants that experienced a transition in depression than in the group of participants that did not experience a transition in depression.

Description

AIM: The Transitions in Depression (TRANS-ID) Tapering study has gathered intensive longitudinal data in a group of individuals who were previously prescribed antidepressant medication for depressive symptoms, which they tapered during the study period. The study was designed to gather high-resolution time series to monitor individual change processes in great detail, to allow for personalized predictions of shifts in depressive symptoms.

METHOD: The data collection combined experience sampling methods to assess momentary affect and behavior (five 28-item questionnaires a day, for four months), ambulatory assessment of physical activity and heart rate (continuous, for four months), and depression symptom assessments (weekly for six months). In addition to a baseline diagnostic interview, baseline questionnaires covered a range of constructs, including overall psychopathology symptoms, medication use, psychological treatment history, alexithymia, life events, quality of life, and chronotype. After the four-month ambulatory assessment period, a personal report of the experience sampling data was given to each participant, and a semi-structured qualitative interview was conducted to evaluate participants’ own retrospective experience of symptom changes during the research period.

CONCLUSION: The TRANS-ID Tapering study procedures and materials are described in detail in this document. The study protocol was approved by the Medical Ethical Committee of the University Medical Center Groningen (ABR no. NL58469.042.16).