The impact of daily social support on the associations between daily negative events and daily suspiciousness across different clinical stages of psychosis

independently moderate the associations between the unpleasantness of daily negative events and daily suspiciousness in the lagged models


Introduction
The clinical staging framework proposes that psychosis develops gradually through subsequent stages (McGorry et al., 2006).The early clinical stages are characterized by less frequent and/or intense and less psychotic-specific manifestations, followed by stages characterized by attenuated, subthreshold psychotic-like experiences.In more advanced clinical stages, diagnostic criteria of psychotic disorder are met (Shah et al., 2020).
One of the most prevalent subclinical psychotic experiences is paranoia, describing the attribution of malevolent intentions to others (Freeman, 2007).Paranoia manifests with different degrees of severity, from mild suspiciousness to persecutory delusions (i.e.beliefs that others would intentionally cause harm (Freeman and Garety, 2000)).Paranoia manifestations are present in the general population and across the clinical stages for psychosis and they precede the development of severe forms of psychotic symptoms (Bebbington et al., 2013;Elahi et al., 2017;Ronald et al., 2014;Wong and Raine, 2018).Several studies have suggested that paranoia fluctuates at the micro-level (i.e. at shorter periods of time, such as within and between days) and macro-level (i.e.months and years) (Ben-Zeev et al., 2011;Thewissen et al., 2008).To investigate paranoia at the micro-level, studies use ambulatory assessments (AA).AA require participants to rate their experiences once or several times a day and yield a large number of assessments per individual, allowing the investigation of within-person processes.The associations between macro and micro-level paranoia have been rarely investigated, but previous research (Schreuder et al., 2020;van der Tuin et al., 2023) have suggested medium associations between macro-and micro-level psychotic symptoms, suggesting that these measures overlap to some extent, but also capture different processes (i.e.momentary experiences vs remembering experiences (Christensen et al., 2003).Paranoia, measured both at macro-and micro-level, has been associated with several risk factors, for example high anxiety, low self-esteem, traumatic life events, aberrant social processing, sleep problems and anomalous internal experiences, just to name a few (Collip et al., 2011;Fusar-Poli et al., 2017;Lüdtke et al., 2023;Meisel et al., 2018;Rosen, 2016;So et al., 2018;Thewissen et al., 2011;Wong et al., 2014).Although previous studies focused extensively on the risk factors associated with paranoia, the investigation of protective factors in daily life remains scarce.
Social support is considered one protective factor associated with lower mental health problems (Lakey and Orehek, 2011) and psychotic manifestations (Robustelli et al., 2017;Wang et al., 2023).At the microlevel, research has suggested that increased daily social support is associated with lower daily paranoia (Monsonet et al., 2022;Steenhuis et al., 2023).However, the mechanisms through which daily social support impacts daily paranoia remain unknown.Social support could directly decrease daily mental health problems (Lakey and Orehek, 2011) or indirectly, by buffering the effect of daily negative events (Ilies et al., 2011;Kaurin et al., 2021).Daily negative events exacerbate daily psychotic experiences (Klippel et al., 2018;Monsonet et al., 2022;Palmier-Claus et al., 2012;Vaessen et al., 2019), thus, daily social support might impact daily paranoia by buffering their effect.To date, this hypothesis has not been empirically tested.The clinical staging framework assumes that the risk and protective factors sustaining or buffering the psychotic manifestations are stage-specific (McGorry et al., 2018), but it remains unclear if the effect of daily social support differs across the clinical stages.Focusing on the micro-level manifestations of paranoia and the factors associated with them, can provide insights on how psychotic experiences manifest and develop in daily life.Furthermore, better understanding of the effect of daily social support on the association between daily negative events and daily paranoia could contribute to the theoretical understanding of psychosis and how it may develop along the continuum of severity.
In the present study, we focus on the daily manifestations of paranoia as suspiciousness, thus throughout the paper, the latter term will be used.We aim to address two research questions: Is the unpleasantness of daily negative events associated with daily suspiciousness (RQ1)?Is there an effect of daily social support on the association between daily negative events and daily suspiciousness and does it differ across the clinical stages (RQ2)?Both contemporaneous (same day t) and lagged effects (i.e. the effects of the unpleasantness of negative events and social support during the previous day (t-1) on current day (t) suspiciousness) will be investigated, using four subgroups representing clinical stages (i.e. from a psychometrically defined risk group to an UHR subgroup).
For RQ1, we expect that: i) participants will report higher than average suspiciousness on the days with more unpleasant negative events than on average (Hypothesis 1a-contemporaneous effect); ii) participants will report higher than average suspiciousness when on the previous day they experience more unpleasant negative events than on average (Hypothesis 1b-lagged effect).
For RQ 2, we expect that: i) the effect of social support on the association between the unpleasantness of negative events and suspiciousness s during the current day (t) will differ among the subgroups (Hypothesis 2a-contemporaneous effect); ii) the effect of previous day (t-1) social support on the associations between the unpleasantness of previous day (t-1) negative events cand current day (t) suspiciousness will differ among the subgroups (Hypothesis 2b-lagged effect).Because of the lack of studies investigating these relations in early clinical stages, the direction of any possible differences among the subgroups is not specified.
If the hypotheses of RQ2 are not confirmed, a follow-up research question will be tested (RQ3): Do the daily social support and clinical stage moderate the associations between the unpleasantness of daily negative events and daily suspiciousness?We expect that: i) the effect of the unpleasantness of negative events and suspiciousness during the current day (t) will differ among the subgroups and will be weaker when participants report more social support than on average during the same day (t) (Hypothesis 3a-contemporaneous effect); ii) the effect of the unpleasantness of negative events on the previous day (t-1) and suspiciousness during the current day (t) will differ across the subgroups and will be weaker when participants report more social support than on average during the previous day (t-1) (Hypothesis 3b-lagged effect).The relation between daily negative events and daily suspiciousness is not linear across the clinical stages (e.g., Palmier-Claus et al., 2012), thus the direction of any possible differences among the subgroups is not specified.The research questions and hypotheses are presented in Fig. 1.

Participants
Data from the baseline daily diary period and the baseline (i.e.before the daily diary) assessment of the Mapping Individual Routes of Risk and Resilience (MIRORR) study was used.MIRORR was designed to study a broad range of transdiagnostic symptom manifestations, functioning, coping mechanisms and resilience factors on both the micro and macrolevel, combining two daily diary periods and three annual follow-up measures.MIRORR is anchored in the clinical staging model and aims to disentangle the characteristics that are common and different across early at-risk stages (i.e.stage 0 to 1b from the clinical staging model (McGorry et al., 2006)), to better understand disorder progression across these stages (Booij et al., 2018).
For the daily diary assessments, participants reported each evening, for 90 consecutive days, the unpleasantness of negative events, social support and suspiciousness using a digital questionnaire they could access using a link received on their smartphones.Participants had a time limit of 90 min to complete the surveys.On average, participants completed 84 daily surveys (ranging from 69 to 90), resulting in a total of 7977 complete surveys across the entire sample.Before the daily assessment period, clinical interviews were conducted by trained researchers.Inclusion criteria in the study were: being aged between 18 and 35 years old, having the ability to read and speak Dutch fluently, and providing informed consent.Exclusion criteria were: previous or current psychotic episode according to DSM-IV criteria, significant hearing or visual problem impairments, and pregnancy.
Participants were assigned to four subgroups that represent clinical stages for psychosis.Subgroup 1 (n = 25) represents a psychometrically at-risk subgroup, including participants from the general population that scored high on the Community Assessment of Psychotic Experiences Questionnaire (CAPE; Konings et al., 2006).Subgroups 2, 3 and 4 included help-seeking individuals recruited from mental health facilities, assessed with the Prodromal Questionnaire 16-item version (PQ-16; Ising et al., 2012) and the Comprehensive Assessment of At-Risk Mental State interview (CAARMS; Yung et al., 2005).Participants in subgroup 2 (n = 27) had a score < 6 in the PQ-16 and were not assessed with the CAARMS.Participants in subgroup 3 and 4 scored ≥6 in the PQ-16 and were subsequently assessed with CAARMS.Participants in subgroup 3 (n = 24) scored ≥6 in the PQ-16, but were not classified as meeting the UHR prognostic criteria on the CAARMS.Participants in subgroup 4 (n = 20) scored ≥6 in the PQ-16 and were classified as meeting the UHR prognostic criteria.Fig. S1 from the Supplementary Materials presents the inclusion procedure in the subgroups.Details Fig. 1.Illustration of the two main research questions (RQ1 in solid black lines and RQ2 in dotted black lines) and the follow-up research question (RQ 3 in dotter grey lines).On the left, the contemporaneous effects (same day associations planned to be tested in the Hypothesis 1a, 2a and 3a) and on the right, the lagged effects are presented (the effect of unpleasantness of previous day negative events and social support on the current day suspiciousness planned to be tested in the Hypotheses 1b, 2b, 3b).
about the MIRORR study design and procedures can be found elsewhere (Booij et al., 2018;Wigman et al., 2022).

Daily diary items
Three daily diary items were selected to measure the unpleasantness of negative events -"How unpleasant was the most important negative event of the day?", suspiciousness -"I felt suspicious today" and perceived social support-"Did you feel supported today?".All items were scored on a 100-point visual analogue scale (VAS; 0-"not at all", 100-"very much).The item assessing the unpleasantness of the daily negative event was rated on a scale from 0-very unpleasant to 100neutral and it was reversed, so that higher scores indicate higher unpleasantness of the event.

General psychopathology and psychotic experiences
The Symptom Checklist Revised (SCL-90; Derogatis and Unger, 2010) was used to assess the psychological symptoms in the past week.Subclinical psychotic experiences were assessed using the Community Assessment of Psychotic Experiences (CAPE; Konings et al., 2006).The CAPE consists of 42 questions assessing the frequency and the distress of psychotic symptoms.For the present study, the total score and persecutory ideation frequency score were used to describe the subgroups' characteristics at baseline (before the daily diary assessment).The persecutory ideation subscale is composed by 7 items measuring suspiciousness, ideas of reference, and persecutory delusions.The items are presented in the Supplementary Materials in Table S1.For a detailed description of the four subgroups regarding the subscales of these questionnaires, see Wigman et al. (2022).

Statistical analysis
Because the normality assumption of the scores of general psychopathology and psychotic experiences was not met (Shapiro-Wilk and Anderson-Darling tests yield p values >0.05) the subgroups were compared using Kruskal-Wallis tests.Multilevel analyses were used to compare the subgroups on their scores of daily items.
The analyses were preregistered before the data analysis (https://osf.io/9uqjp).Multilevel models were fitted to test the contemporaneoussame day associations, and temporal or lagged effects, from one day to another.Multilevel models account for the hierarchical structure in the data, where the daily responses are clustered within individuals (Bolger and Laurenceau, 2013).Contemporaneous and lagged effects were tested in separate models.For the lagged effects, a lag of one day was used.Prior to analysis, all time-varying predictors (i.e.variables that were measured every day: social support and unpleasantness of daily negative events) were person-mean centered (PMC) by subtracting each person mean (PM) from the raw scores.The subgroup variable was included in the analysis dummy-coded (i.e.four variables taking the values 0 and 1, depending on the group participants were assigned to).In all models, time, centered at the first day of the study, was included as a covariate to control for linear trends.In the models testing the lagged effects, previous day suspiciousness was included to control for autoregressive effects.Random slopes for the predictors and covariates were added in all models.Random slopes were also added for the interaction terms at first, but due to convergence problems, these were excluded from the final analyses.Different error-covariance structures were tested and, based on the Akaike and Bayesian Information Criteria (with lower values indicate a better fit), the unstructured covariance matrix was chosen for all models.
To test the hypotheses of RQ1, two multilevel models were run.In Model 1a, the scores of current day suspiciousness1 were regressed on the current day PMC scores of the unpleasantness of negative event).In Model 1b, the scores of current day suspiciousness were regressed on the current day PMC scores of the unpleasantness of the negative event.
To test the hypotheses of the RQ2, two multilevel models were run.In Model 2a, scores of current day suspiciousness were regressed on a three-way interaction between the current day PMC scores of the unpleasantness of negative event, current day PMC score of social support and the subgroup (i.e.dummy-coded subgroup variables).In Model 2b, scores of suspiciousness at t were regressed on a three-way interaction between the PMC unpleasantness of negative events at t-1, PMC social support at t-1 and subgroup (i.e.dummy-coded variables).Because the Model 2a showed a statistically significant triple interaction, we conducted follow-up analyses to interpret these results.First, thesubgroups were compared among themselves, by changing the reference subgroup in the multilevel models.Second, the double interaction between the current day unpleasantness of negative events and current day social support were tested in each subgroup separately.The Johnson-Neyman technique was used to probe the statistically significant interaction by calculating the upper and lower limits of the moderator (i.e., current day social support) where the effect was significantly different from zero (Hayes, 2018).Because Model 2b did not show a statistically significant triple interaction, the hypotheses of RQ3 were tested, by running two multilevel models.In Model 3b, the scores of current day suspiciousness were regressed on two double interactions: i) one interaction between the previous day PMC unpleasantness of the negative event and the previous day PMC social support and ii) one interaction between the previous day PMC unpleasantness of the negative events and subgroup (i.e.dummy-coded subgroup variable).In the sensitivity analysis, the person means (PM) of daily social support and of unpleasantness of the daily negative events and gender were added in all the models showing statistically significant results.We used p < 0.05 as the inference criteria.The analyses were run using the nlme package (Pinheiro et al., 2021) in R.
Our focus on the within-person associations, thus the power of our analyses is mostly based on the number of observations for each individual rather than on the number of participants.Using different studies recommend at least 60 observations per individual for multilevel and vector autoregressive (VAR) models (e.g.Rosmalen et al., 2012).Thus, with 90 time points for each individual, we expect sufficient power to detect the within-person effects.Regarding the cross-level interactions, a recent simulation analysis (Wardenaar, 2021) suggests that comparing subgroups with 25 participants and 60 time points per individual yields enough power (>0.8) to detect medium effect sizes (i.e.0.30).Thus, with our sample size, we expect sufficient power to detect the cross-level interaction effects.In addition, previous studies (e.g.van  all analyses, we will use the scores of suspiciousness after being person-mean centered (PMC).However, as we included a random intercept in the model, person-mean centering of the outcome variable became unnecessary.We run two sets of analyses using PMC and the raw scores of suspiciousness and the results of the main analyses of the study remain unchanged.

RQ 1: is the unpleasantness of daily negative events associated with daily suspiciousness?
There was a significant contemporaneous effect of the unpleasantness of negative events on suspiciousness (B = 0.10, p < 0.001): During the days when participants rated the negative events as more unpleasant than on average, they also experienced more suspiciousness than on average.The unpleasantness of the negative event on the previous day had no significant effect on the current day suspiciousness (B = 0.009, p = 0.12).Table 2 presents these results.Ns = not significant.
§ Low = primary or lower secondary education, medium = upper secondary education, high = university/collage education.

Table 2
The results of the multilevel models testing the RQ1 (Model 1a and Model 1b), estimating the contemporaneous and lagged effects of the unpleasantness of daily negative events on daily suspiciousness in the whole sample.In italics and bold are marked the p < 0.001.PMC = person-mean centered.
Time was centered at the first day of the daily study.

RQ 2: is there a moderating effect of daily social support on the association between the unpleasantness of daily negative events and daily suspiciousness and does it differ across the at-risk subgroups for psychosis?
There was a significant triple interaction among the subgroup, unpleasantness of the negative event and social support during the current day (F (3,7868) = 3.28, p = 0.02).The follow-up multilevel models comparing the four subgroups indicated that subgroup 3 differed from subgroups 2 and 4 in the moderation effect (Table 3).After the FDR correction, only the difference between subgroup 3 and subgroup 2 remained statistically significant (B = -0.002,p = 0.003).
The multilevel models run in each subgroup to test the double interaction between the current day unpleasantness of the negative events and current day social support showed a significant two-way interaction in subgroup 3 (B = − 0.001, p = 0.006) (Table 4).Johnson-Neyman test indicated that the association between the unpleasantness of negative events and suspiciousness during the current day was statistically different from 0 (p < 0.05) when social support was below 34.06.This suggests that in subgroup 3, on the days with more social support than on average, there was a weaker association between the unpleasantness of negative events and suspiciousness compared to the days with lower social support than on average.In the other subgroups, the interactions were not significant (p > 0.05).But a main effect of both social support and unpleasantness of negative on suspiciousness was shown.Fig. 2 presents the contemporaneous associations between the unpleasantness of negative events and suspiciousness in each subgroup, plotted at one standard deviation above and below the means.
No significant triple interaction was shown (F (3,7868) = 0.33, p = 0.803) among the subgroup, previous day unpleasantness of negative events and previous day social support, thus RQ3 hypotheses were tested.

RQ 3: do daily social support and at-risk subgroups for psychosis independently moderate the associations between the unpleasantness of daily negative events and daily suspiciousness in the lagged models?
The results showed no significant double interaction between the previous day unpleasantness of negative events and previous day social support (F(1, 7211) = 0.86, p = 0.353) or between previous day unpleasantness of negative events and subgroup F(3,7211) = 1.189, p = 0.312).

Sensitivity analysis: including gender and person means as covariates
Model 1a, Model 2a and the follow-up analyses of Model 2a were rerun including gender and the person mean (PM) of current day unpleasantness of negative events and current day social support as covariates.The effects remain similar in strength and significance (Tables S2 and S3 in the Supplementary Materials).

Findings
The present study aimed to investigate the buffering effect of daily social support on the association between the unpleasantness of daily negative events and daily suspiciousness and whether this effect differed among clinical stages (i.e. from a psychometrically at-risk subgroup to individuals meeting the UHR criteria).On days when participants reported more unpleasant negative events than on average, they also reported more suspiciousness than on average.Furthermore, we found that the moderation effect of the social support on the association between the unpleasantness of negative event and suspiciousness on the same day differed across the clinical stages.
Regarding the first research question, our findings are in accordance with previous research suggesting that daily negative events are Table 3 The results of the multilevel models testing the RQ 2 (Model 2a and Model 2b) and the follow-up multilevel models (after the statistical significant triple interaction showed in Model 2a), comparing the at-risk subgroups on the effect of current day social support on the association between the unpleasantness of current day negative events and current day suspiciousness.For the Model 2b, no follow-up comparisons were conducted given that the triple interaction was not statistically significant.
Model 2a: Differences among the subgroups in the effect of social support at t on the association between unpleasantness of negative events at t and suspiciousness at t Overall triple interaction unpleasantness negative events (t) x social support (t)  associated with increased daily paranoia (Hermans et al., 2020).The mechanisms through which daily stressors increase daily suspiciousness are not very well understood.Previous research suggests that at-risk individuals present a reactive autonomic nervous system, characterized by greater anticipatory stress (Baumeister et al., 2021) and slower recovery (Vaessen et al., 2019) of the daily psychotic manifestations in response to daily stressors.Daily stressors might activate negative beliefs about others' intentions, which take the form of paranoia (Lincoln et al., 2009(Lincoln et al., , 2017)).In the MIRORR study, although the unpleasantness of daily negative events did not statistically differ among the subgroup, the number of negative events and the level of stress reported daily were higher in the clinical subgroups than in the general population subgroup, which might indicate increased stress sensitivity (Wigman et al., 2022).Furthermore, our results suggest that the unpleasantness of daily negative events is associated with increased daily suspiciousness in all subgroups, but this association becomes slightly stronger across the atrisk continuum (seen in the main effects in Table 4 and by steeper slopes in Fig. 2).Future studies need to investigate the mechanisms through which daily negative events impact daily suspiciousness.One avenue of study could focus, for example, on the processes that mediate this within-person process, such as anxiety, physiological activation, attribution style biases that can impact the interpretation of negative events (Lüdtke et al., 2023).Regarding the second research question, the moderating effect of daily social support differed across the subgroups: the participants from subgroup 3, who present psychotic experiences without meeting the UHR criteria, differed from the participants from subgroup 2 (participants with few psychotic experiences) and participants from subgroup 4 (participants meeting UHR criteria).Interestingly, when looking within each subgroup, the moderation effect of daily social support was statistically significant only in subgroup 3.For the participants in subgroup 3, on days with more social support than on average, the association between the unpleasantness of negative events and suspiciousness was weaker than on days with less social support than on average.These findings echo previous studies suggesting that social support can buffer the effect of daily negative events on daily emotional experiences (Collip et al., 2011;Ilies et al., 2011;Kaurin et al., 2021).While daily social support moderated the effect of daily negative events on suspiciousness in subgroup 3, in the other subgroups, daily social support was associated with lower daily suspiciousness, independently of the level of unpleasantness of the negative events.Together, these results suggest that the mechanisms through which daily social support impacts daily suspiciousness differ across the clinical stages.For instance, in subgroups 1, 2 and 4, daily social support might reduce daily suspiciousness by directly reappraising negative beliefs about others' intent (Freeman, 2007).On the other hand, in subgroup 3, daily social support decreases daily suspiciousness by buffering the effect of daily negative events.This indirect effect might be explained by several characteristics of subgroup 3. First, subgroup 3 presents similar levels of persecutory ideation as subgroup 4 (see Fig. S2 in the Supplementary Material), but without meeting the UHR criteria.Thus, they might lack the specialized support and care offered to the more clinical populations.In addition, these participants might also lack the support they need from their social network, as previous studies conducted on the MIRORR dataset found that participants from subgroup 3 report needing more support in daily life (Wigman et al., 2022).Furthermore, participants in subgroup 3 might present increased psychotic reactivity to stressors, being in line with previous studies reporting stronger psychotic response to daily stressors in early at-risk stages of psychosis compared to more clinical populations (Vaessen et al., 2019;van der Steen et al., 2017).Therefore, in subgroup 3 daily suspiciousness might be triggered to a greater extent by the unpleasantness of daily negative event, while in the other subgroups daily suspiciousness might be underpinned by other processes, such as anomalous experiences, anxiety etc.Thus, in individuals from subgroup 3 social support might increase the adaptive coping strategies, such as reappraising any stressors (Davis and Brekke, 2014;Uchino, 2009) and provides the material and  emotional resources necessary to deal with daily negative events (Tempier et al., 2013).Contrary to our expectations and previous results, our study did not reveal any statistically significant lagged effects, while testing both RQ2 and RQ3.Previous studies using AA indicate also lagged effects of previous day negative events on current day paranoia (Klippel et al., 2018) and a buffering effect of previous day social support on current day stress (Kaurin et al., 2021).Our null findings might be partially explained by the fact that our results are based on daily diary data, in which participants reported once a day their experiences, whereas previous studies used more frequent sampling.When using daily diaries, the time interval between two reports is longer, resulting in weaker lagged effects and stronger effects within days.It is also possible that the effects targeted in the present study unfold over shorter periods of time (i.e.hourly), reflected in the statistically significant contemporaneous results.The current results are in line with other work on daily diary data showing most effects at the contemporaneous level (Groen et al., 2022;van der Tuin et al., 2021).Furthermore, the lagged effect of the unpleasantness of negative events and suspiciousness might be stronger in more clinical subgroups, such as used in previous studies (Klippel et al., 2018).Given that we focused on earlier at-risk subgroups for psychosis, the effects might be less strong, thus they did not reach the statistically significant threshold.At the moment, our understanding of how these constructs fluctuate over different time spans is suboptimal (Myin-Germeys and Kuppens, 2022).It seems unlikely to assume that all momentary effects change at the same pace.For this purpose, future studies could combine different sampling rates (e.g.daily and momentary assessments, but also weekly or monthly) and different methodologies (e.g.experimental, where the impact of social support can be studied using social stress or threat tasks, self-and other reports of social support).Additionally, richer theoretical frameworks, embedded in clinical practice, are needed to guide future research on social support, suspiciousness and unpleasantness of negative events.

Strengths and limitations
A first strength of the study lies in the extended period of the daily assessment (90 days) and the nature of the sample, including both general population and clinical subgroups, representing different clinical stages for psychosis.This design allowed the investigation of a threeway interaction using both within-person (i.e., social support) and between-person moderators (i.e., subgroup).One limitation of the study is that the number of participants in each subgroup was relatively small.However, the research questions focus on the within-person associations thus, with 90 observations per individual, we had enough power to detect a medium effect size (Wardenaar, 2021).Nevertheless, the results of the present study should be replicated with larger samples to investigate their robustness.Furthermore, because of the daily sampling frequency, relations at shorter time spans could not be studied.This design, however, allows longer study periods potentially leading to a better reflection of real life and the possibility of greater variability in reported events and experiences.Furthermore, some individuals presented high or low scores for the daily diary items across the entire diary period.Although this is a commonly observed phenomenon, especially when assessing negative affect items, this might affect the reliability of our model estimation (von Klipstein et al., 2023).Although these floor and ceiling effects could be observed for some participants, all participants presented some variability in their scores.Furthermore, although the rates of missing daily diary data are low (on average 9 % of missing data per participant), the reasons why the participants did not complete these surveys remain unknown.Some studies suggest that the missingness in AA studies is not predicted by the measured construct (Sun et al., 2020), however, the non-response patterns should be addressed by future research.
Another limitation lies in the fact that we investigated the effect of the unpleasantness of the most negative event of the day on daily suspiciousness, using an item focusing on a broad range of events.Previous studies have suggested that different daily stressors (e.g.social or event related stressors) have specific impact on daily psychotic experiences (Klippel et al., 2017;van der Steen et al., 2017).To further disentangle Fig. 2. Associations between the unpleasantness of current day negative events and current day suspiciousness at one standard deviation above and below the mean of current day social support in each subgroup.
the buffering effect of daily social support, future studies need to investigate its effect on the association between different types of daily negative events and daily suspiciousness.Furthermore, we focused on the perceived social support during the day (I felt supported), without investigating the available social support (i.e.support network).Previous studies have suggested that daily perceived and available social support have different effects on the association between daily stressors and daily emotions (Kaurin et al., 2021;Uchino, 2009).Additionally, we tested the moderation effect of social support measured in the same day as the unpleasantness of daily negative events, for both contemporaneous and lagged effects.It is possible that social support measured at t-1 moderates the association between unpleasantness of negative events at t-2 and suspiciousness at t. Future research needs to address if the types of social support also impact differently daily psychotic manifestations and at what time scale.

Conclusions
The present study showed that the effect of protective factors (i.e.daily social support) on daily psychotic manifestations (i.e.daily suspiciousness) differs across the clinical stages for psychosis, thus contributing to a better understanding of the clinical staging model.Our results highlight the effect of daily negative events on daily suspiciousness and the buffering effect of social support, especially in individuals presenting with psychotic symptoms, without meeting the full UHR prognostic criteria.Awareness that different mechanisms may underlie psychotic experiences at different clinical stages might improve the theoretical understanding of psychosis continuum.
der Stouwe et al., 2014) used similar sample sizes to test double and triple interactions among categorical and continuous variables, with two subgroups, each containing 27 participants and 96 time points per participant).

Table 1
Demographic and clinical characteristics of the whole sample and the four subgroups.

Table 3
(continued )Model 2a: Differences among the subgroups in the effect of social support at t on the association between unpleasantness of negative events at t and suspiciousness at t

Differences among the subgroups in the effect of social support at t-1 on the association between unpleasantness of negative events at t-1 and suspiciousness at t
For clarity raisons, the results for the comparison of the same pair of subgroups were left out from the table.In italics are marked the p < 0.05, in bold are marked the values p < 0.01 and italics and bold are marked the p < 0.001.PMC = person-mean centered.Time was centered at the first day of the daily study.p < 0.05 after the FDR correction. *

Table 4
Results of the follow-up multilevel models estimating the double interaction between same day unpleasantness of negative events and same day social support in each subgroup separately.
In italics are marked the p < 0.05, in bold are marked the values p < 0.01 and italics and bold are marked the p < 0.001.PMC = person-mean centered.Time was centered at the first day of the study.