A three timepoint network analysis of COVID-19's impact on schizotypal traits, paranoia and mental health through loneliness

This is a very timely investigation on the relationship between schizotypy, paranoia, loneliness and mood disturbances during the COVID-19 pandemic. The three-wave design and large multi-site sample are unique and precious. I hope the manuscript would do justice to the amount of work that has already been put in by clarifying the following issues:

1. The conceptuliastion of some words, such as 'mental health', 'wellbeing', and 'symptoms' seems to be blurred throughout the manuscript. For example, it is debatable whether loneliness is considered a symptom (see Abstract) just like anxiety and depression. While anxiety and depression have established cutoffs and are typically considered as sympotms within numerous clearly defined psychiatric disorders, the construct of loneliness may or may not be clinical/ symptomatic. In this paper, loneliness has been phrased as 'symptom' in Abstract, but 'problem' on p. 3, and 'feelings' on p. 4. Another example is p. 4 (last paragraph): 'four studies have investigated paranoia and schizotypal personality traits in relation to mental health during the pandemic' - it is not clear what 'mental health' is referred to here. From the abstract, my guess is that 'mental health' means anxiety, depression, and loneliness, but it wasn't made clear. Moreover, it wasn't clear why then paranoia wouldn't also be part of mental health?

2. On a related note, while the authors set out to consider 'psychotic-like experiences' as indexed by schizotypal personal disorder and paranoia (see p. 3), the studies cited focused on mistrust and suspicion only, i.e. concepts of paranoia rather the PLE (which is broader). Freeman et al (2020) was cited (p. 4, paragraph 2) as follows: 'Psychotic-like experiences as highlighted in a large representative sample of UK adults in April 2020...'. However, while Freeman et al (2020) used a paranoia measure (R-GPTS) and a trust barometer, they did not include a PLE measure. It would be easier for readers to follow if the constructs of concern are discussed with more clarity.

3. On p. 4 (paragraph 2), a range of variables have been suggested to be consequences of lockdown restrictions (e.g. loneliness, anxiety and PLE), but it wasn't clear how the authors think that these variables may contribute to each other. Even though network analysis is a data-driven approach, a bit more theoretical discussion of the expected directions of associations would still be helpful for interpretation of results.

4. On p. 6 (paragraph 1), the authors specified the lockdown periods in the UK. However, this was a multi-site sample and it wasn't clear whether the same periods would be relevant to lockdown measures in other sites. If not, then it needs to be specified in the Introduction and Discussion sections so as to facilitate interpretation of results.

5. Hypothese: what are 'social networks' in the context of hypothesis 2? [check also the expression of 'psychological networks' on p. 9 last paragraph]. Hypotheses are supposed to be tested for or against, but the way hypothesis 2 is phrased isn't testifiable. In addition, why is the 3-wave design not mentioned in the hypotheses?

6. Since the 3-wave design is a major design element, which certainly reflects the amount of work involved in this study, it would seem fitting for more discussion on the 3-wave design and use of network analysis to be included in the Introduction section. In particular, as network analyses can be done in multiple ways, it would be helpful if the authors link the specific type of network analysis with the research question in the Introduction section. e.g. Why were the 3 time points needed? Should readers expect to see 3 separate networks? Were the strengths of edges (within each network) of interest, or the changes in edges across networks?

7. This manuscript will benefit from thorough proof-reading and grammar check.

This study investigated the relation between schizotypal traits and various mental health variables during the COVID19 pandemic. The method of network analysis highlighted the important role of loneliness in linking schizotypal traits and poor mental health outcomes. The longitudinal design further demonstrated stable network structures over time despite reductions in overall symptom levels.

Overall, this was a well-written manuscript, and the method and results sections are clear and easy to follow. However, given the rich longitudinal data collected and the network approach used, there are several additional questions that could be addressed to make this paper stand out from among the others in this area.

Below are more specific comments and suggestions for revision:

Significance of this study : One of my major concern is how this study could contribute uniquely to the impact of COVID19 on paranoia and schizotypal traits. The focus of the analysis was the relation between schizotypal traits/paranoia and other mental health outcomes over time, but little is known concerning COVID19-related variables. For example, whether and how would paranoia be associated with social distancing and length of lockdown? If the data were collected without the impact of COVID19 (in pre-pandemic periods), would we have similar findings?

Framing of primary measures: Throughout the manuscript, the authors regarded “schizotypal traits” and “paranoia” as two different concepts/variables to index psychotic-like experiences. I am not sure about this, because paranoia/suspiciousness is a sub-dimension of schizotypal traits (as measured by SPQ). Even in the short version of the SPQ-B, there are 4 items specifically assessing paranoia, which were included in both the cognitive-perceptual and the interpersonal factor of the SPQ-B.

Additional analyses should be considered: Given the longitudinal data, the authors could consider using cross-lagged panel network modeling to explore longitudinal associations between different variables. This method can provide further insight into which node was most strongly predicted by other variables, and also which node shows the strongest power to predict other symptoms. Such analysis could help us to better understand the causal relationship between schizotypal traits and mental health outcomes.

See relevant studies using cross-lagged panel network modeling:

Bringmann, L. F., Lemmens, L. H. J. M., Huibers, M. J. H., Borsboom, D., & Tuerlinckx, F. (2015). Revealing the dynamic network structure of the Beck Depression Inventory-II. Psychological Medicine, 45(4), 747–757. https://doi.org/10.1017/S0033291714001809

Savelieva, K., Komulainen, K., Elovainio, M., & Jokela, M. (2021). Longitudinal associations between specific symptoms of depression: Network analysis in a prospective cohort study. Journal of Affective Disorders, 278, 99–106. https://doi.org/10.1016/j.jad.2020.09.024

Groen, R. N., Snippe, E., Bringmann, L. F., Simons, C. J. P., Hartmann, J. A., Bos, E. H., & Wichers, M. (2019). Capturing the risk of persisting depressive symptoms: A dynamic network investigation of patients’ daily symptom experiences. Psychiatry Research, 271, 640–648. https://doi.org/10.1016/j.psychres.2018.12.054

Further, the authors did not introduce the method of network comparison across three waves in the Method section. Please add this part. Note that we should account for the dependence of measurements within the same individual when comparing networks at different time points, so this analysis is a bit different from network comparison across groups.

Results:

1. The authors should give more details about the samples of the three waves. As a considerable proportion of participants dropped out at Wave 2 and Wave 3, it is better to clarify whether there are any differences in demographic characteristics and mental health outcomes between those dropping out and those who completed 3 waves of surveys.
2. The authors only showed the results of centrality indices (strength) for Wave 1 data. How about the other 2 waves? Does the node of “loneliness” stably show high centrality in the network and serve as a bridge connecting schizotypal traits and mental health?
3. As shown in Figure 2, the network seems to become less densely connected over time (reduced global strength from Wave 1 to Wave 3). Therefore, it is a bit strange that the global strength of Wave1 network (3.99) is smaller than that of the Wave 2 network (4.02). Please check if the result was correctly presented. Also, have the authors compared the network structures between Wave 1 and Wave 3? Is it possible that the network differences become significant over a longer time period (i.e., 1 year)?

Discussion:

Based on the main findings summarized in the first paragraph of the Discussion, I feel confused and not very convinced how the authors could come to the conclusions that “intervening on self-perceived loneliness - an influential variable across all participant groups which may have improved during the easing of lockdown - may break the negative associations between paranoia/schizotypy and negative mental health symptoms, but externalizing symptoms may still remain.”

In the second paragraph, the authors tried to explain why schizotypal traits were correlated with loneliness. The results that “both paranoia and the interpersonal dimension of schizotypy were strongly associated with loneliness in the network” could support the two interpretations proposed by the authors.

The authors used two entire paragraphs to explain the changes in self-reported loneliness during the pandemic. I agree this is an interesting finding but I cannot see why and how this result contributes to the main purposes of the current study. Maybe more emphasis should be put on the bridge function of loneliness linking schizotypal traits and mental health outcomes. (Moreover, I feel it hard to understand why individual differences could explain the evolution of self-perceived loneliness.)

Although loneliness may serve as a bridge symptom in the network, loneliness was not the node with the highest strength. Both depression and anxiety had high centrality in the network. Targeting the node with the strongest influence can lead to significant changes of the network structure and the levels of other symptoms. Therefore, should we also consider the interventions targeting depressive and anxiety symptoms?

The authors did not discuss the results of the more densely connected network for individuals with high schizotypal traits compared with individuals with low levels of schizotypy/social mistrust. This is an interesting finding worth more detailed discussion.

Minor issues:

Introduction:

Page 4 line 1, “both of which……” It is unclear what constructs the authors were referring to. Are they “paranoia and schizotypal traits”?

Page 4 line 9 (“It is conceivable……”) The second paragraph in the Introduction provided evidence to show the COVID19 pandemic caused heightened levels of social distrust. However, it seems a bit far-fetched to come to the hypothesis that “lockdown will have a bigger effect for individuals with higher levels of schizotypal traits and paranoia compared to their peers”.

Page 4 the last two lines: “a similar group reported increases in schizotypal……” This sentence is ambiguous and I am not sure what “a similar group” meant. Does the author mean “a similar proportion of individuals reported increases in schizotypal traits (compared with the proportion of individuals reporting the experience of schizotypal traits for the first time)”? In addition, the work from Knoelle and colleagues (2021) is not in the References list. Please add this reference.

Page 5 2 ^nd paragraph (the one to introduce the network analysis) line 4: What does “comparison across interactions ” mean?

Page 6 the 2 ^nd hypothesis “the social network”?

Methods:

Page 7 line 2 “two waves of data collection” should be “three waves”

Page 8 2.2.1 Why the total score of the SPQ-B ranges from 0-44 (not 0-22)? Also for the loneliness questionnaire, why the score range is 20-77 but not 20-80? Please check.

When calculating the internal consistency of all the scales in this study, which wave of the data did the authors use?

Page 10 2.3 Data analysis The authors showed bivariate relationships for Wave 2 data in Table3. It should be the results of Wave1, right?

Page 11 NCT paragraph – “The significance threshold was set at p or adjusted p< 0.05”?

Results:

For table 4 and 5, please add in the notes what the BOLD characters indicate.

Page 17-18: “network structure invariance test: M; global strength invariance: S” Should clearly state what the M and S referred to.

Discussion:

Page 21 “Whether this is purely due to the COVID easing of restrictions taking place during time 3….” It is unclear what “this” refers to in this sentence.

Page 22 Ill-formed sentence: “This may suggest that there are individual differences in the length of lockdown on self-perceived levels of loneliness ……”

Page 22 the last line: “This was not measures….” Should be “measured”