Background
Mental disorders are directly and indirectly associated with a large part of overall morbidity and mortality worldwide [
1]. Once developed, many mental disorders tend to become chronic or recur [
2]. Hence, prevention of these disorders is crucial.
Still, our current understanding of the development of psychopathology is limited, due to a substantial amount of different factors involved in this process (such as variations in individual differences and environmental factors) and complex, non-linear interactions between these factors. Such complexity behind psychopathological processes hampers accurate identification of people at risk. Embracing this complexity may be the way forward in understanding psychopathology and its development. A promising approach to obtain accurate risk estimations comes from the theory of complex systems. Examples of such complex systems are ecosystems, which are known to make shifts from a forest state to a swamp state, or the financial market, which can experience a sudden collapse [
3,
4]. Although such changes are results of numerous mechanistic interactions, complex systems theory states that the stability of a system, i.e., how hard it is for a large change to occur, can be quantified in one characteristic: an index of resilience. This overall estimate of resilience of the system can be assessed by its capacity to recover from minor perturbations that occur. This phenomenon is called “critical slowing down” (CSD) and refers to the process whereby the system becomes increasingly slower in its capacity to recover [
3,
4]. Indicators of CSD have been shown to predict (non)-critical transitions as well as gradual change in various sorts of complex systems, whether they are financial markets, oceans, climate, or brain activity [
3,
5,
6]. If these principles work for psychopathology as well, we can assume that higher instability in the system (in this case, mental health), and thus lower resilience, means that it is more difficult to remain in a current healthy state and that this is related to, on average, higher levels of symptoms in the near future in this group of people.
Therefore, we expect that when speed of recovery from small perturbations is slower over time, this signals a lower stability of people’s current mental health state and, therefore, a higher likelihood of a change in the level of symptoms. Application of this approach to mental health, therefore, may help to improve personalized assessment of risk and resilience to psychopathology before new symptoms have arisen [
7‐
10].
Supporting this line of reasoning, some previous studies examined indirect indicators of the process of critical slowing down, such as rising temporal autocorrelation and variance [
3], in the micro-dynamics of affect states. These studies indeed showed that temporal autocorrelations and/or variances are increased in people with higher levels of psychopathological symptoms compared to people with lower levels of symptoms or healthy controls [
11‐
15]. Moreover, a study by Wichers and Groot has shown on the individual level how a change in these indicators directly preceded a transition to a state with more psychopathological symptoms [
16] Thus, there is initial empirical support suggesting that mental health may behave according to the laws of complex dynamic systems based on indirect measures of critical slowing down [
10,
15,
17].
However, hardly any studies in psychopathology have examined the phenomenon of critical slowing down using direct measures of this process, i.e., direct measurements of the speed of recovery from minor perturbations in the system. For that, a design is needed that allows for the prospective and detailed assessment of the impact of minor perturbations in the flow of daily life on mental states. To our knowledge, only one recent study, by Vaessen and colleagues [
18], examined in this way the speed of the affect recovery from daily stressors in groups with various levels of psychopathology. They found that speed of affect recovery was slower in people at early stages of psychosis compared to healthy volunteers and people with already developed psychosis. Although this study was not written explicitly from a complex systems perspective, results may support the predictions from that theory. This is because both healthy controls and people with established psychosis can be assumed to be in more stable states than those at early stages of psychosis. Therefore, as a next step, it is important to test the hypothesis that speed of recovery, as an indicator of the process of critical slowing down, indeed predicts the future development of psychopathology. The current study will therefore, for the first time, use “speed of recovery from minor perturbations to the system” as a direct dynamic indicator of the process of critical slowing down to examine whether this measure predicts future change in levels of psychopathology.
In order to examine this question, we used a sample of adolescents from the general population with relatively low levels of happy childhood experiences, representing an increased risk for psychopathology [
19]. These adolescents come from the TWINSSCAN data set which includes baseline time-series data on affect states and daily unpleasant events, combined with baseline and follow-up assessments of (subclinical) psychopathology in a large sample of adolescents. Using a similar approach to measure the concept of “speed of recovery” as Vaessen and colleagues [
18], we examined how quickly people recovered in terms of their experienced affect states from small negative events, reflecting minor perturbations, that happened throughout the day (e.g., spilled coffee, traffic jams).
In sum, the aim of this study is to examine whether the speed of recovery from small perturbations in daily life differs between adolescents with different future trajectories of psychopathology. We expect the speed of affect recovery from daily life unpleasant events to be slower in adolescents who will develop more psychopathological symptoms over 1 year than in adolescents who remain on similar levels of symptoms over 1 year.
Discussion
This study examined whether the speed of affect recovery from small perturbations in daily life differs between adolescents with different mental health trajectories in the following year. Results show that, in individuals who will develop more symptoms in the following year, their negative affect recovered slower after unpleasant events than in people with no increase in symptoms in the following year. For positive affect, group differences were not significant. Stratified by group, the effect of unpleasant events on both negative and positive affect was detectable 90 min longer in the Increase group than in the Stable group. The analysis of the individualized estimations of speed of recovery similarly showed an association between speed of negative affect recovery and individual future symptoms change, but did not reach significance.
Complex systems principles and psychopathology
This study supports the idea that the complex systems approach can be applied to psychopathology. This approach assumes that the system can shift between alternative states, such as between having different levels of symptoms. These results add to the growing body of research suggesting that dynamic indicators of stability of complex systems, which estimate the process of “critical slowing down,” may be also applicable to mental health. The current results have shown that a direct measure hereof—namely speed of affect recovery from small perturbations—predicted mental health outcomes. Moreover, although at baseline the two groups were similar in levels of symptomatology, they already differed in this dynamic measure of resilience. Thus, the dynamic examination of speed of recovery may capture some additional information compared to simple mean levels of stressors, affect states, and levels of symptoms. Therefore, in the future, a complex systems approach to mental health may contribute to a more accurate and reliable prediction of risk and resilience in psychopathology.
The dynamic concept of resilience
Psychological resilience is a popular topic in contemporary mental health research, as many scholars believe that focusing on protective mechanisms may yield insights for prevention and treatment [
35‐
37]. However, most studies attempt to examine resilience using static measurements, such as retrospective questionnaires estimating personal competences, acceptance of change, social abilities and support, coping strategies, levels of optimism, and meaning in life [
38‐
40]. However, the concept of resilience, in most of its definitions, is about people’s ability to withstand adverse circumstances, making the concept a dynamic one [
11,
41,
42]. Although static measures certainly may tap into important aspects of resilience, they are unlikely fit to fully capture a dynamic concept. Defining resilience from a complex systems perspective has the advantage that it can be assessed in a direct, dynamic way, by prospectively measuring the impact of minor perturbations on the system. Although replication is warranted, the dynamic assessment of resilience may become a valuable tool to assess and monitor change in psychological resilience both for research and clinical practice.
Methodological issues
The current study has several methodological issues. First, as the data came from a twin sample, it is possible that twins may have different dynamics of affect than non-twins, and therefore the findings may not be fully generalizable. Moreover, the phenotype of slower (or faster) affect recovery may have a shared hereditary component. However, despite being a twin cohort, we could only use those participants who also had follow-up measurements. Thereby, although interesting, this sample is strongly underpowered for any hereditary investigations. Second, the approach that we took for creating individualized affect recovery indicators has both benefits and limitations. The additional benefits of this approach were (i) the creation of one indicator that reflected recovery over several time points, (ii) a possibility to test the predictive value of this indicator on the individual level, and (iii) a possibility to obtain potentially clinically relevant estimations of effect sizes (i.e., how differences in the speed of recovery were associated with change in SCL-90 scores). The limitation of this approach, however, was a reduction of power due to the loss of the multilevel structure of the data, as this approach was performed with one score representing the speed of recovery per individual (although the time-series data allowed us to retain more power due to the lower standard deviations of the variables which were constructed based on multiple observations, compared to a hypothetical cross-sectional study with only one variable per person). Therefore, the borderline significance of the association between this AUCb score and future level of symptoms may be also due to the lack of power. Finally, symptom trajectories were measured with only two assessments, 1 year apart, which adds much noise to the data. Therefore, the results of this study should be considered preliminary until reproduced with more data observations and higher temporal precision.
Clinical translation and future directions
The above method of assessing people’s current resilience state may have clinical value, not only as a way to monitor individual resilience but also as a new potential target for intervention and prevention strategies. There are, however, some important steps in the process of translating this study outcome to clinical practice. First, findings need to be translated from the group level to the individual level. The differences between individuals concerning affect dynamics may be substantial [
43] and it is very important to investigate which changes are of clinical relevance and for whom. The results of this study represent the average effect over many, and therefore the overall effect is an average of individual differences in affect dynamics. Moreover, individuals may also differ in the moment when they precisely developed symptoms, and this moment was not assessed in the current study as only a single follow-up measure was used. Thus, new personalized designs, in which people are continuously and intensively monitored with regard to daily stress, affect and symptoms over extended periods of time, are required to establish whether CSD indicators indeed consistently anticipate relevant symptom changes. Although our study represents a first step towards testing this hypothesis, an important next step is to reproduce these findings at the individual level.
Second, we can assume that speed of recovery, as an indicator of system stability, is not a constant but will change over time. If we thus want to monitor changes in people’s resilience, we should measure how the speed of recovery from daily unpleasant events changes over time within individuals. This would require a design in which individuals are monitored with ESM over a longer period of time (e.g., several months). Feasibility of such designs in patients has recently been established (unpublished communication).
Finally, for this study, we assume that CSD, because it signals instability of the system, is relevant in predicting vulnerability to psychopathology. With the current design, it was not possible to assess directly whether a sudden transition occurred and, if so, at what moment in time. Therefore, for future studies, it is important to attempt to follow participants through transitions between states and to directly assess the timing and shape of this transition and the changes in the speed of recovery with respect to them.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.