The pooled sample comprised participants from 8 previously conducted studies that used a similar protocol and are part of the ESM merge file. These studies included individuals with a mental disorder, that is, psychotic disorder [17,33-38] or depressive disorder with residual symptoms [39]; at-risk individuals, that is, with familial [17,34,36,40] or psychometric risk for psychosis [19,38]; and controls without a personal or family history of mental disorder [17,19,34,36,38,40]. The samples and procedures to obtain diagnoses and risk status of the participants have been described elsewhere (see Multimedia Appendix 2).

All 8 studies received approval by their respective medical ethics committees in the Netherlands and Belgium as stated in the original references and all procedures were performed in accordance with the ethical standards of the responsible medical ethics committee. This study was registered on OSF (Open Science Framework) before data access [41].

Stata version 16.0 (StataCorp LLC) was used for statistical analysis [43]. Experience sampling data have a 3-level structure with individual assessments (level 1) nested within days (level 2), which are, in turn, nested within individuals (level 3). Group differences on level 3 variables (ie, age and gender) were examined using 1-way ANOVAs and chi-square tests as appropriate, whereas group differences on levels 1 and 2 were examined using Stata’s “mixed” command for multilevel models.

To test the hypotheses, the procedure described by Vaessen et al [20] was followed. Trajectories of positive affect in response to the first stressor of a day were examined to rule out the potential cumulative impact of consecutive stressors throughout a day on positive affect. A new predictor variable “time_since” was created for each stress measure to mark the time points when positive affect was measured in relation to the first stressor of the day. The time point when the stressor occurred (ie, stress reactivity) was set to t₀, the time point prior to this (ie, t_–1) served as the baseline, and all time points following the stressor were set to t_1–n. First, to test H1, in a separate model for each group using time_since to predict positive affect, we compared all time points t_0–n with baseline. Second, to test H2, group was added as a predictor in the model and group comparisons of positive affect were calculated at baseline and t₀. Third, to test H3, an interaction between time_since and group was specified in the model to compare affective reactivity at t₀ between groups. Last, affective recovery was compared between groups (H4) using the average decrease of positive affect from baseline across the recovery period. Specifically, a recovery period of 2 prompts was specified as the average deviation of positive affect at these time points from baseline positive affect.

For each momentary stressor (event-related, activity-related, and composite stress), separate models were fitted. For each model, observations were excluded (i) for participants who never reported the specific type of stress, (ii) for days on which the specific type of stress was not reported, and (iii) for days on which the specific type of stress was reported on the first prompt of the day so that no baseline measure was available.

All models were adjusted for age (centered using the grand mean) and gender (for unadjusted models, see Multimedia Appendix 4). As a sensitivity analysis, the analysis was repeated controlling for subsequent stressors. To this end, dichotomous control variables were created for event- and activity-related, or composite stress indicating the presence (=1) or absence (=0) of the respective stressor at all time points t_n>0. We used Simes correction [44] to account for multiple tests of significance regarding our 3 stress measures, as all models testing our specific hypotheses were repeated for each stress measure. Therefore, according to the Simes procedure, the most significant P value within each model was compared with α=.05/3=.02 and the second most significant P value was compared with α=.05/2=.03. Results that remain significant after Simes correction are marked with footnotes in tables. A significance level of P<.05 was set for all remaining P values.

The sample comprised 921 participants. This includes 422 individuals with a mental disorder (ie, 293 with psychotic disorder and 129 with remitted depressive disorder with residual symptomatology), 246 at-risk individuals (ie, 178 with familial and 68 with psychometric risk), and 253 controls. Participants completed a total of 42,778 prompts. Average compliance was 75% (45/60 prompts) for patients, 78% (47/60 prompts) for at-risk individuals, and 82% (49/60 prompts) for controls (F_2,918=13.02, P<.001). Across groups, 2304 prompts were not completed within 15 minutes after the signal or all positive affect and stress items were missing (χ²₂=21.2, P<.001). In addition, 34 participants completed less than 20 prompts over 6 days (χ²₂=2.8, P=.24) and 75 participants never reported any type of stress (χ²₂=0.01, P=.95) and were therefore excluded from the analysis.

Hence, the analytic sample consisted of 811 participants (patients/at-risk/controls: n=367/217/227) with a total of 39,903 valid prompts (patients/at-risk/controls: n=16,122/9997/10,784). Sample characteristics of the analytic sample are depicted in Table 1.

Figure 1 shows the trajectories of positive affect in response to composite stress in all groups. A main effect of group was observed for all stress measures (event-related stress: χ²₂=1575.64, P<.001; activity-related stress: χ²₂=48.69, P<.001; composite stress: χ²₂=200.9, P<.001). There were differences in baseline levels of positive affect across all groups, consistent with H2 (Table 4). However, patients and at-risk individuals did not differ as hypothesized. Patients had significantly lower baseline levels of positive affect than at-risk individuals (P values for all stress types <.001). There was no evidence for a 2-way interaction (time_since × group) at t₀ for any stress measure. This indicated that the associations of event-related stress, activity-related stress, or composite stress with positive affect, that is, the initial positive affective reactivity, did not differ across individuals at different stages of mental disorder, leaving H3 unsupported. As all groups had returned to baseline levels of positive affect by t₂ following activity-related and composite stress, marking the end point of the continuous recovery period, t₁–t₂ were included in the between-group analysis. When examining differences in the average deviation of positive affect from baseline levels during the recovery period t₁–t₂ in response to activity-related stress and composite stress, we did not find evidence for between-group differences (Table 4). This indicated that positive affective recovery, operationalized as an average deviation from baseline, was similar across the groups at different stages of mental disorder, leaving H4 unsupported.

This study aimed to investigate trajectories of positive affect in response to daily life stress across different transdiagnostic clinical stages in a pooled sample of patients with a mental disorder, individuals at psychometric or familial risk, and controls. All groups showed a similar trajectory of positive affect in response to momentary stress, as indicated by a decrease in positive affect to event-related, activity-related, or composite stress, and a continuously lower level of positive affect before recovering to baseline level in response to activity-related or composite stress (H1). We observed a continuous recovery period of 180 minutes on average in patients and at-risk individuals, whereas controls required 90 minutes on average to recover. Comparisons across groups revealed that patients with a mental disorder and at-risk individuals had lower baseline levels of positive affect in daily life compared with controls (H2). Contrary to our prediction, patients had lower levels of positive affect compared with at-risk individuals. Differences in positive affective reactivity to daily stress between groups (H3) and in positive affective recovery fell short of statistical significance (H4).

Several methodological considerations should be taken into account when interpreting the reported findings. First, because this study used existing data, participants with different clinical characteristics were pooled to form transdiagnostic groups as an approximation to representing subclinical and clinical stages of mental disorder based on the literature of clinical staging. To further support the staging approach and ensure that participants with different clinical characteristics form a group regarding severity of symptoms or functional impairment as suggested by clinical staging, latent class analysis may be used in future analysis to identify groups with similar behavioral patterns. Furthermore, participants may be recruited according to recently developed criteria for clinical staging as there is first evidence for their validity as a way of identifying individuals in early stages with predictive power for transition between stages [3].

Second, the dichotomous operationalization of stress as the presence or absence of a stressor does not account for the degree of unpleasantness of a reported activity or event, which reduces variance. An activity or event rated as –3 may impact positive affect longer than an activity or event rated as –1. Similarly, Vaessen et al [45] showed that emotional reactivity to mild, but not intermediate or strong stressors was related to symptom levels in adolescents 1 year later, suggesting that the degree of unpleasantness of a stressor may need to be accounted for in future studies on affective recovery.

Third, stress reactivity at t₀ was modeled in a cross-sectional manner, that is, ratings of stress and positive affect measured at the same time point were used to define stress reactivity. Therefore, temporal order between the first stressor of a day and an associated decrease in positive affect remains unclear as a stressor may lead to a decrease in positive affect, or vice versa. Yet, the cross-sectional modeling does not restrict interpretations regarding the recovery period, which was of main interest in this study, operationalized using time points chronologically before and after the occurrence of stress.

Relatedly, the exploratory finding that positive affective recovery within groups may be accounted for by cumulative stress at the following time points should be interpreted with caution. A recent review showed that experiencing positive affect can impact the neural signaling of stress, which may lead to less self-reported stress [46]. As the temporal order between cumulative stress and positive affect measured at the same time point remains unclear, it may, in turn, be possible that being in the recovery period, that is, in a state of decreased positive affect, may lead participants to report more stress.

Last, the composite stress measure combining event- and activity-related stress may hold restrictions. Both stress types may be related to affective recovery in different ways. Specifically, event-related stress is a retrospective judgment of the most important event that happened since the last prompt. As the time points were approximately 90 minutes apart, the unpleasant event might have happened up to 90 minutes before the rating, meaning that an immediate drop in positive affect after the event and the beginning of the recovery period might not have been recorded by the random sampling procedure. Activity-related stress, by contrast, measures the unpleasantness of the current activity. The sampling procedure does not reveal when an unpleasant activity started or for how long it was continued after the measurement, which may also influence positive affect ratings at baseline or during the recovery period. We found no recovery period after event-related stress and effect sizes were lower at t₀ for event-related stress than at the same time point for activity-related stress (Table 2), indicating that the recovery period for event-related stress may have already begun before reporting the event. Taken together, the sampling procedure in this study may have been limited in detecting differences in positive affective recovery between groups. For future research, a design with more frequent measurements or a hybrid event- and time-contingent sampling procedure may provide more fine-grained modeling of affective recovery.

In line with previous research [15,47], our study showed that levels of positive affect differed between individuals with a mental disorder, individuals at-risk for developing a mental disorder, and controls across the continuum of mental health, thus broadening findings to a transdiagnostic staging approach for the first time. While all groups reported stress reactivity and a period of affective recovery in response to activity-related and overall stress that was descriptively longer within the patient group and at-risk individuals, these differences fell short of statistical significance in between-group analysis comparing average deviations of positive affect from baseline levels. Yet, levels of positive affect in patients and at-risk individuals were generally lower across the entire recovery period (Figure 1). This may suggest that reactivity of similar magnitude and a recovery period of similar length may be associated more strongly with risk and disorder when operating on lower overall levels of positive affect.

Furthermore, the magnitude of differences in positive affective reactivity and recovery between groups might have been too small to be detected with the number of observations per day in our models. In addition, criteria other than clinical status might be relevant to index risk and identify group differences in trajectories of positive affect in response to minor stressors, such as childhood adversities. In line with the stress sensitization hypothesis [12], stress reactivity as a behavioral marker for stress sensitization has been shown to be amplified in individuals exposed to severe adversity across the life course [22,48-51]. For instance, stress reactivity in early and later stages of psychopathology has been reported to be greater in individuals exposed to high levels of childhood adversity than in controls exposed to high levels of adversity, suggesting they were more resilient [22]. Future research should investigate whether this holds true for differences in positive affective recovery as a transdiagnostic marker for momentary resilience, that is, the ability to recover from minor stressors in the moment. Differences in affective recovery across stages may only become evident when viewed in the context of exposure to adversities across the life course.

To our knowledge, this is the first study to transdiagnostically investigate the trajectories of positive affect after minor stressors in daily life. It has been shown that positive and negative affect can be conceptualized as 2 distinct factors [52] that are related to positive and negative events in daily life in different ways. For example, negative events were found to be less strongly related to positive than to negative affect [53]. Similarly, Wichers et al [54] found that physical activity, which may be regarded as a positive activity, was related to momentary positive affect, but unrelated to momentary negative affect. Adding to previous findings [20], we found a shorter period of positive affective recovery after a negative event than was found for negative affective recovery after a negative event. Furthermore, group differences in negative affective recovery between at-risk individuals and individuals with a mental disorder were not reflected in our findings regarding positive affective recovery. Taken together, this may suggest that the trajectories of positive and negative affect in response to minor daily stressors constitute separate psychological mechanisms. This underlines the differential role of positive affect for psychological well-being [24,27] and highlights the need to investigate, in more detail, how positive and negative affective recovery compare in stages of mental disorder.

In this study, we found first evidence for different trajectories of positive affect following minor daily stressors in a transdiagnostic sample covering the continuum of mental health. Whether positive affective recovery on the scale of minor stressors in daily life may be a putative indicator for momentary resilience should be investigated further in the context of childhood adversity, specifically focusing on healthy, that is, resilient, individuals exposed to adversities. When disentangling this putative protective mechanism further, trajectories of affective recovery may potentially serve as a target for ecological momentary interventions, a mobile health approach using mobile devices to deliver interventions in daily life [55,56]. Targeting affective recovery, intervention components may potentially be presented in moments when participants experience stress helping them to recover, and ultimately foster resilience in early and later stages of psychopathology. Targeting this putative momentary mechanism in ecological momentary interventions allowing the use of experimental designs in daily life [57] may allow us to understand more fully the role of affective recovery in pathways to severe mental disorder. This will provide evidence for the effectiveness and feasibility of scalable interventions for transdiagnostic populations.