An online self-guided cognitive intervention for unwanted intrusive thoughts about harming infants in new parents: initial randomised controlled trial with mediation analysis

ABSTRACT Approximately one-fifth of new parents struggle with unwanted intrusive thoughts (UITs) about intentionally harming their child. This study evaluated the initial efficacy, feasibility and acceptability of a novel online self-guided cognitive intervention for new parents with distressing UITs. Self-recruited parents (N = 43, 93% female, age 23–43 years) of children 0–3 years reporting daily distressing and impairing UITs were randomized to the 8-week self-guided online cognitive intervention or to waiting-list. The primary outcome was change on the Parental Thoughts and Behaviour Checklist (PTBC) from baseline to week 8 (post-intervention). The PTBC and negative appraisals (mediator) were assessed at baseline, weekly, post-intervention and at the 1-month follow-up. Results showed that the intervention led to statistically significant reductions in distress and impairment associated with UITs at post-intervention (controlled between-group d = 0.99, 95% CI 0.56 to 1.43), which were maintained at the 1-month follow-up (controlled between-group d = 0.90, 95% CI 0.41 to 1.39). The intervention was deemed to be feasible and acceptable by the participants. Change in negative appraisals mediated reductions in UITs but the model was sensitive to mediator-outcome confounders. We conclude that this novel online self-guided cognitive intervention can potentially reduce the distress and impairment associated with UITs in new parents. Large-scale trials are warranted. Abbreviations: UITs: Unwanted Intrusive Thoughts PTBC: Parental Thoughts and Behaviour Checklist.


Introduction
Becoming a parent is generally regarded as a meaningful and happy period in life but it may also set the stage for increased levels of worry.About half of new parents have occasionally experienced distressing, unwanted intrusive thoughts (UITs) about intentionally harming their child (e.g.thoughts about stabbing or burning the child; Brok et al., 2017;Fairbrother et al., 2022).These thoughts are regarded as a normal and transient psychological phenomenon (Brok et al., 2017;Fairbrother et al., 2022).However, for about one-fifth of parents, these cognitions can become disproportionately distressful in intensity or duration, and interfere with parental function and personal relationships (Abramowitz et al., 2003(Abramowitz et al., , 2010;;Fairbrother & Woody, 2008).In their most extreme form, individuals experiencing such "taboo obsessions" fulfill criteria for a diagnosis of obsessive-compulsive disorder (OCD).
UITs about intentionally harming one's child are associated with more parental stress than other kinds of intrusive thoughts, such as thoughts about accidentally harming the infant (Fairbrother & Woody, 2008;Fairbrother et al., 2018).UITs about intentionally harming one's child are also associated with symptoms of depression, impaired quality of life and parental dysfunction (Challacombe et al., 2016;Gezginç et al., 2008;Miller et al., 2015).Importantly, there are reports indicating that parents who experience these distressing thoughts tend to not disclose their thoughts to healthcare personnel due to shame or fear that their children will be removed from their care (Bayrampour et al., 2017;Challacombe & Wroe, 2013;Lawrence et al., 2017).This is unfortunate, as there is evidence that cognitive-behavioral therapies (CBT) can be efficacious in reducing distress and impairment associated with these cognitions in new parents.A trial by Challacombe et al. (2017) showed that time-intensive CBT was more effective than treatment as usual in reducing symptoms (N = 34; between group effect size d = 1.32) in a sample of postpartum mothers with OCD.A randomized controlled trial by Timpano et al. (2011) showed that a brief psychological intervention, consisting of 6 weekly group meetings of 1.5 hours each, reduced OCD symptoms in post-partum women (n = 38) compared to an active control (n = 33, between-group effect size d = 0.41-0.73).Another study showed that a brief online self-help psychoeducation for new mothers specifically targeting UITs related to intentional harm decreased levels of self-concealment and shame, i.e. the participants became more open about their symptoms (Melles & Keller Dupree, 2022).
Our research group has recently developed an online, therapist-guided intervention specifically targeting taboo obsessions in individuals with a formal diagnosis of OCD.The main focus of this intervention is to change negative appraisals of intrusive thoughts (e.g."Having these thoughts is a sign that I'm an immoral person") to more functional and realistic beliefs (e.g."It's just a random thought and it doesn't say anything about me as a person").This cognitive change is achieved through standard psychological techniques such as psychoeducation, behavioural experiments, and cognitive restructuring exercises.A pilot study including 19 OCD patients with taboo obsessions showed that the intervention was highly feasible.The majority of the participants were able to understand and apply the cognitive framework to their own intrusive thoughts.Sixty-three percent were classified as treatment responders at post-treatment.Reductions in intrusive thoughts were driven by the hypothesized mechanism of change (change in negative appraisals; Olofsdotter Lauri, Aspvall, Bagøien Hustad, et al., 2022).Because the majority of parents struggling with UITs may not necessarily meet diagnostic criteria for OCD or seek help for their difficulties, one logical next step would be to adapt this intervention to a selfguided format targeting a broader population of parents who are not in contact with health services.The present study was designed to address this gap in the literature.
A recent survey of the population of interest, that is, parents of infants and toddlers (N = 594) found that about one-fifth of the sample reported difficulties controlling their obsessions and/or that the obsessions had had a negative impact on the relationship/ attachment to their child (Olofsdotter Lauri, Aspvall, Serlachius, et al., 2022).Only 5% of these participants had sought and received help for their intrusive thoughts and 51% of the full sample responded that they would absolutely or probably like to try online interventions for these intrusive thoughts.The results of this survey suggest that there is a demand for self-guided online interventions in new parents who struggle with intrusive thoughts related to intentional harm.
This randomized controlled trial evaluated the initial efficacy, feasibility and acceptability of a novel, online self-guided cognitive intervention targeting new parents who report UITs about harming their children (aged 0-3 years).A secondary aim was to evaluate if the hypothesized cognitive mechanism of change (change in negative appraisals) mediated improvements in UITs.

Trial design
This randomized controlled trial compared an 8-week self-guided online intervention to a waiting-list control.Controlled follow-up assessments were conducted 12 weeks post randomization (1-month follow-up).Participants in the waitlist group were subsequently offered the intervention.Recruitment was national, with participants eligible for participation across Sweden.The study was approved by the Swedish Ethical Review Authority (dnr 2021-01540) and was registered at Open Science Framework (https://osf.io/jtzma/)prior to the start of the trial.All participants provided written informed consent.

Participants
The study was available to adult parents of children aged 0-3 years, reporting daily UITs about intentionally harming their child, which were experienced as being distressing and interfering with daily function (e.g.difficulties concentrating or finding joy in everyday situations, avoidance).We used a cut-off of ≥10 points on the Parental Thoughts and Behaviour Checklist (PTBC) and at least one yes response to one of four items investigating (a) the impact of UITs on perceived control over the thoughts, (b) self-image, (c) social relationships and (d) attachment to one's child. 1 Exclusion criteria were: non-Swedish speaking; severe depression (defined as >32 points on the Montgomery Åsberg Depression Rating Scale-Self report [MADRS-S]); and elevated suicidal risk (defined as >4 points on item 9 on the MADRS-S).
Of 110 screened individuals, 43 (female = 40, 93%) were included in the study (Figure 1), and randomized to the intervention (n = 22) or waiting-list (n = 21).Most participants (n = 26, 60%) reported that they had been diagnosed with one or two psychiatric disorders prior to study entry.Depression was the most commonly reported disorder (n = 12, 28%).Only three participants (7%) had received a previous diagnosis of OCD (Table 1).

Recruitment and inclusion
The study was advertised on the social media platform Facebook (9 February 2022 to 21 March 2022) with a link to the study webpage.After providing informed consent, participants were asked to complete a screening battery of self-rated scales to assess eligibility, consisting of the PTBC (Thiséus et al., 2019), MADRS-S (Svanborg & Åsberg, 1994), four items investigating the impact of UITs on perceived control over the thoughts, self-image, social relationships and attachment to one's child, and general background information.Each screening was carefully reviewed by the study coordinator and principal investigator, who immediately called participants scoring >32 points, or >4  points on item 9 on the MADRS-S total scale, reporting previous serious mental disorder that could pose a risk to study participation, or ticked in a box in the online platform that the participant wanted an additional telephone assessment by study personnel.Individuals who were not eligible for participation were informed via text message and received information on the reason for exclusion and contact information to the study personnel if they needed more guidance on how to seek help for their problems.
Eligible participants were asked to complete the baseline assessment battery and were subsequently randomly assigned 1:1 to either intervention or waiting-list control group.Participants were excluded if they had not completed the baseline assessment.
Participants who reported severe depression or elevated suicidality in the baseline assessment were immediately contacted by study personnel.All participants received information about what to do in case of worsening of symptoms and could call study personnel during office hours.The allocation was done according to a prespecified randomization list (created using random.org).The randomization list was not available to the study coordinator who included the participants.

Intervention
The intervention was previously tested using a therapist-supported format (Olofsdotter Lauri, Aspvall, Bagøien Hustad, et al., 2022) and is based on the cognitive model of obsessions (Rachman, 1997(Rachman, , 1998)).Briefly, negative appraisals (interpreting the intrusive thought as personally significant) are seen as core mechanism that fuels the obsession (Rachman, 2003).The main aim of the intervention is therefore to adjust these negative appraisals to more realistic beliefs.The intervention was an 8-week program delivered via a secure web platform and consisted of 6 text-based modules (described in detail below), also including audio files and interactive features such as worksheets.Each module also includes homework tasks with digital worksheets for the participants to work with for a few days up to one week.The participants worked with the intervention on their own without any therapist contact, but they could contact study personnel if needed.The participants received an automated text message every week with a reminder of the weekly measures to be completed within the online platform.In case the participants had not logged in for one to two weeks, the study coordinator sent a personalized text message to the participant.All modules were open for the participants from the start, and they were instructed to work with the modules chronologically and in their own pace.
The first module consists of general psychoeducation about the nature of UITs and a general overview of the cognitive model.This framework emphasizes how negative appraisals of intrusive thoughts play a key causal role in the development and maintenance of intrusive thoughts.The participant is encouraged to register daily UITs in an online diary, as well as how they interpret the meaning of the UITs.The second module provides an in-depth walkthrough of the cognitive model of intrusive thoughts.The participant is encouraged to apply this general model to their own UITs using an online work sheet.The third module provides the participant with a cognitive restructuring exercise called "Theory A vs Theory B".In this exercise, the participant is instructed to write down their current interpretation of the UITs (e.g."Theory A: I'm a dangerous person because I have these thoughts").The online worksheet also includes a column where the participant can elaborate on the evidence for this interpretation (e.g."I have read an online article about people who have obsessive thoughts and actually acted on them").The next step in this exercise is to elaborate on alternative interpretations (e.g."Theory B: My thoughts are only thoughts and thus not anything I need to pay any attention to") and support for this theory (e.g."The information in the module said that about 87-95% of the population have thoughts similar to mine.If it was true that people who experience these thoughts act on them, it would be much more common that parents murder their children.And it is not.").An important aspect of module three was that the participant is instructed to challenge the interpretation of, rather than directly challenge, the content of the intrusive thought.Module four includes a behavioral experiment exercise where the participant is encouraged to test out what happens if they do not respond to the intrusive thoughts and record if this new strategy can reduce distress.Module five continues with more examples of behavioral experiments.The participant is asked to plan, execute and evaluate different experiments to test the validity of his/her appraisals of the UITs.After completing a behavioral experiment, the participant writes down what he/she learned from the experiment in the online worksheet from module three (e.g."I have tested if I would lose control if I really focused on the thoughts for 30 minutes.I felt a lot of anxiety, but I did not lose my mind and I did not act on the thoughts.

Conclusion: Even if it feels like I am going crazy because of the anxiety, I can keep my sanity. Maybe I just misinterpret my anxiety sometimes?"
).The fifth module also provides the participant with case examples of common difficulties when working with behavioral experiments.Module six includes psychoeducation about setbacks.The participants are asked to summarize their own progress and make an individual plan for maintenance of gains.

Primary outcome measure
The primary outcome measure was a revised version of the self-rated PTBC (Thiséus et al., 2019), which is a measure of UITs and related behaviors in new parents.It includes two parts.The first part contains a checklist of obsessions and related behaviors.The current study only included items and examples of intentional harm thoughts related to the infant and toddler (e.g.thoughts about intentionally drowning the child or stabbing the child to death) as parents with these kinds of thoughts was the target population.The second part includes 10 items assessing severity of the obsessions and behaviors, on a 5-point scale (0-4 points) with a total score of 0-40 points, with higher scores indicating more symptoms.The PTBC has shown good psychometric properties (Chronbach's α = 0.91) and a similar factor structure as the PTBC clinical interview (Thiséus et al., 2019).PTBC was administered pre-intervention (week 0), weekly during the intervention (weeks 1-7), at week 8 (primary end point) and at 1-month follow-up.Chronbach's alpha in the current sample was α = 0.91.

Secondary outcome measures
The Personal Significance Scale (PSS; Rachman, 2003) was used to measure the hypothesized mechanism of change in cognitive therapy (reductions in negative appraisals).The PSS includes 23 items rated on a 9-point scale (scores ranging from 0 = "Not at all" to 8 = "Definitely") assessing the level of experienced personal significance of the UIT's.The current study included a shortened five-item version, ranging from 0 to 40 points which has shown a strong correlation (r = .86)with the full scale and a good internal consistency with the full scale (α = 0.88) in a Swedish sample of patients with OCD (Olofsdotter Lauri, Aspvall, Bagøien Hustad, et al., 2022).Chronbach's alpha in the current sample was α = 0.85.Included items were: 1. "Do these thoughts reveal something important about you?", 2. "Do these thoughts mean that you are a dangerous person?",3."Do these thoughts mean that you might lose control and do something awful?", 4. "Do these thoughts mean that you might go crazy one day?", 5. "Would other people condemn or criticize you if they knew about your thoughts?".
Symptoms of depression were assessed by using the MADRS-S (Svanborg & Åsberg, 1994).The self-rated scale is a 7-point, 9-item questionnaire ranging from 0 to 40 points.Higher score indicates a higher level of depression.The self-rated scale has a good internal consistency of α = 0.84 (Fantino & Moore, 2009).Chronbach's alpha in the current sample was α = 0.85.
Symptoms of sleep difficulties were assessed by the Insomnia Severity Index (ISI; Bastien et al., 2001).The ISI consists of a 5-point, 5-item questionnaire ranging from 0 to 28 points, higher score indicating a higher level of insomnia.The scale has an adequate internal consistency of α = 0.74 (Bastien et al., 2001).Chronbach's alpha in the current sample was α = 0.88.
The Work and Social Adjustment Scale (WSAS; Mundt et al., 2002) was used to assess functional impairment due to the participants UITs within five different areas: work/ school, home management, social leisure activities, private leisure activities and close relationships.Each area is measured with one item, on a 9-point scale (0 = "Not at all" to 8 = "Very severely").WSAS has shown a good internal consistency from 0.70 to 0.94 as well as a test test-retest value of 0.73 (Mundt et al., 2002).Chronbach's alpha in the current sample was α = 0.85.
Parental function was assessed by the Tool to measure Parenting Self-Efficacy (TOPSE; Kendall & Bloomfield, 2005).The self-rated scale assesses perceived parental self-efficacy in four different areas: affection/emotion, play/enjoyment, empathy/understanding and how the parent perceives his/her parenting skills.Each area is assessed by a 6-item subscale, every subscale ranging from 0 to 60 points.Each subscale have shown an good internal consistency of α = 0.81 and above (Kendall & Bloomfield, 2005).Chronbach's alpha in the current sample was α = 0.96.
All secondary outcome measures were administered at baseline week 0, postintervention week 8 (primary end point), and at 1-month follow-up.The PSS was also administered weekly during the intervention (weeks 1-7).Participants who scored >32 points or >4 points on item 9 on the MADRS-S total scale at any assessment point were contacted by study personnel for a psychiatric assessment.

Feasibility, acceptability
Feasibility was assessed by number of completed modules, number of dropouts, and by assessing the participants understanding of and ability to apply the cognitive framework to their own situation.Acceptability were assessed at week 8 using an adapted version of the Client Satisfaction Questionnaire (CSQ; Attkisson & Zwick, 1982).Included items were for example: "How would you rate the quality of intervention you received?"(1 = "Poor" to 4 = "Excellent"), "Overall‚ how satisfied are you with the intervention you have received?"(1 = "Quite dissatisfied" to 4 = "Very satisfied"), "If a friend were in need of similar help‚ would you recommend our program to him or her?" (1 = "No‚ definitely not" to 4 = "Yes‚ definitely"), as well as an open-ended question for comments and feedback.

Adverse events
Adverse events were assessed using a self-reported adverse events questionnaire at week 8 that was used in a drug trial (Andersson et al., 2015).The questionnaire included questions about any unexpected or unwanted events and/or effects the participants had experienced during study participation and follow-up questions about when, how often and for how long the event/effects occurred.Immediate and long-term consequences were assessed on a 4-point scale (0 = "Did not affect me at all", 3 = "Did affect me very much").Each adverse event was continuously monitored by study staff during the study period.If a participant reported a potential serious adverse event, the study coordinator and/or the principal investigator immediately called the participant for an in-depth assessment of the event.

Data analysis
Our previous pilot study in a sample of individuals with OCD had a within-group effect size of d = 1.67 (Olofsdotter Lauri, Aspvall, Bagøien Hustad, et al., 2022).This trial used a randomized trial design with control group and a mainly subclinical sample.We therefore expected the effect sizes of this trial to be lower than in the pilot study.Twenty participants in each group were considered sufficient to estimate the initial effects of the intervention (d ≈ 0.8, 80% power) and to provide reliable feasibility estimates.
In the primary efficacy analysis, we used all weekly data from weeks 0-8 to estimate the change over time (set as a continuous variable) on the PTBC.This was done using a mixed effects regression framework (intention-to-treat) with random intercepts (Stata 16.1).Next, we estimated the between-group effect size at week 8 using the m_effectsize script in Stata ("net install m_effectsize, from (http://www.imm.ki.se/biostatistics/stata) replace").This script estimates effect sizes by dividing the estimated change score in the mixed-effects regression analysis by the pooled standard deviation at baseline.One thousand bootstrap replications are used to construct a 95% confidence interval around this effect size estimate.To estimate the longer term-effects on the PTBC, we repeated the analyses and added the 1-month follow-up in the mixed effects model.The secondary outcomes were analysed in identical fashion as the primary outcome but with the difference that we did not use weekly measurements on the WSAS, MADRS-S, ISI and TOPSE.Descriptive analyses were used to assess feasibility and acceptability.
Regarding the secondary aim, we investigated if changes on the PSS during week 0 to 8 mediated reductions on the PTBC.This was done by obtaining the individual slope change on both the outcome and the mediator in a mixed effects regression framework.The mixed effects model used maximum likelihood estimations with all available data and included fixed effects of time as well as random intercepts and random slopes.Next, the value from the assessment at week 8 was subtracted from the baseline value in each individual linear slope.Indirect (mediation) effects were subsequently estimated with robust standard errors using the medeff command in Stata 16.1 (Hicks & Tingley, 2011;Imai, Keele, & Tingley, 2010).Given the relatively small sample size and lack of pilot data to estimate power for time-lagged mediation, we modelled a simultaneous parallel mediation association.Sensitivity analyses (testing to what extent the results may be affected by mediator-outcome confounding) were conducted by fixing the residual correlation between the mediator and the outcome from r = −.9 to .9 using the Medsens command in Stata 16.1 (Hicks & Tingley, 2011;Imai, Keele, & Tingley, 2010;Imai, Keele, & Yamamoto, 2010).

Efficacy
Participants randomized to the intervention improved significantly more on the primary outcome measure (PTBC) compared to waitlist participants, with a large between-group effect size at week 8 (primary end point, p < .001,d = 0.99, 95% CI 0.56 to 1.43).The differences on all secondary outcome measures also favoured the intervention group.The effects of the intervention were maintained or increased at the 1-month follow-up (between-group effect size d = 0.90, 95% CI 0.41 to 1.39, Table 2).For a detailed description of weekly measures see Table 3.

Feasibility, acceptability (intervention group only)
One participant (4.5%) from the intervention group dropped out from the study and 18 (82%) completed the self-rated assessments at post-intervention.Eleven participants (50%) completed 4 to 6 modules (M = 3.0, SD = 1.93).Four participants (18%) did not complete any modules at all.Detailed information is shown in Table 4. Six participants (27%) contacted study personnel with questions through the online platform during the intervention period.Three of the questions (50%) were about practical issues (e.g.how to access the next module), and three questions (50%) were about difficulties to comprehend or work with the intervention material.Online worksheet data were available for 12 (54%) participants.All participants but two (91%) were assessed by two raters to have successfully understood and applied the cognitive framework to their own situation.All 18 participants (100%) who completed the self-rated assessments at postintervention rated the quality of the intervention as good or excellent, and a large majority (16/18; 89%) reported being mostly or very satisfied with the intervention and would recommend the intervention to a friend.

Adverse events
There was one reported adverse event in the intervention group during the controlled study period of 12 weeks.The corresponding number in the control group were five events.One adverse event was assessed by the study staff to be not directly related to study participation (e.g. car crash) and four events were assessed to be mild and possibly related to the study participation (e.g.increased anxiety at start of intervention, increased frequency of UITs, feeling angry, mental images with taboo themes).One participant from the waiting-list control reported the occurrence of an anger outburst towards the child.The event was not considered to be due to participation in the study.

Mediation
Meditation analysis showed that the indirect effect was significant (estimate 0.68, 95% CI 0.07 to 1.57; proportion mediated = 18%, 95% CI 13% to 27%).This association was highly sensitive to mediator-outcome confounders with a residual correlation as low as r ≈ 0 for the confidence intervals to cross zero (see Figure 2), suggesting that the mediated effect was not robust to unmeasured confounders.

Discussion
The current trial was the first to investigate if an online self-guided cognitive intervention can help new parents to better cope with distressing and impairing UITs about harming their child.Individuals who were allocated to the intervention had significantly less distressing and impairing UITs, depressive symptoms, and sleep impairments, as well as improved general-and parental functioning at the primary end point.These effects were maintained up to 12 weeks after randomization.As hypothesized by the cognitive model, change in negative appraisals during the intervention mediated reductions in UITs.However, the mediation model was highly sensitive to assumptions about mediator-outcome confounding and should therefore be regarded as preliminary.The typical participant in this study was a woman in her thirties, with one child around one year of age and currently on parental leave.The most common previous psychiatric diagnosis in the sample was depression.Very few had received an OCD diagnosis despite elevated levels of UITs and related behaviors at baseline.This suggests that our easily accessible intervention, which does not require extensive contact with services, has the potential to reach a wider population of parents who struggle with intrusive thoughts but who, for various reasons, have not sought or received help for their problems.This might be especially important for individuals with intrusive thoughts of infant-related harm who are hesitant to seek help due to shame or fear of having their children removed (Bayrampour et al., 2017;Challacombe & Wroe, 2013;Lawrence et al., 2017).
The between-group effect sizes found in this trial were in the large range at week 8 (d = 0.99) and sustained at the 1-month follow-up (d = 0.90).These effects are smaller than those of a previous trial investigating the efficacy of face-to-face CBT for new mothers with full-blown OCD (Challacombe et al., 2017), but slightly larger than those reported by Timpano et al. (2011) who included an sample of 71 pregnant women considered to be at risk for OCD receiving a prevention program for obsessive compulsive symptoms.The intervention used in this study has previously been tested in a therapist-guided format for individuals with OCD and taboo obsessions (withingroup d = 1.67, 95% CI 0.62 to 2.72; Olofsdotter Lauri, Aspvall, Bagøien Hustad, et al., 2022).The treatment content in this and our previously published pilot study is identical, but adaptations of the length and the included examples were made to fit the unguided format and the specific study population of new parents.In the pilot study, 63% of participants completed all modules, whereas the corresponding number in this trial was only about 40%.One possible explanation for the somewhat lower module adherence found in this current trial was the amount of therapist support provided (Karyotaki et al., 2021;Lundström et al., 2022).In our previous pilot study, the patients had regularly contact with a dedicated therapist who responded within 24 h to messages, whereas in the current study there was no therapist support, unless the participants explicitly requested it.
Despite the fact that the majority of the participants did not complete all modules, the intervention group in this trial had large and statistically significant reductions of UITs.This indicates that it might be possible to condense the intervention even further and, for example, only include the key aspects such as psychoeducation about UITs of infant-harm and an overview of the cognitive model with case examples.One idea for future research could be to sequentially increase the dosage of treatment, i.e. start with brief psychoeducation, and individuals who need more support could gradually receive more extensive interventions and, in some cases, additional guidance from a clinician.
The mediation analysis in this study showed that changes in negative appraisals only had a modest relationship with reductions in intrusive thoughts (proportion explained = 18%).This was unexpected and we speculate that it may be due to floor effects (ours was mainly subclinical sample with low symptom severity at baseline).One next step to investigate this issue further could therefore be to analyse moderated mediation, i.e. does the mediation association become stronger for those individuals who score high on negative appraisals and/or intrusive thoughts at baseline?Investigating this issue in more detail could provide clinicians with knowledge for whom this intervention works optimally.Another interesting venue for future research could be to analyse other kinds of mediators, e.g.does reduction in intrusive thoughts mediate improved parental-self efficacy?Also, does improved parental self-efficacy in turn lead to protective long-term effects against developing other mental health problems as suggested by some longitudinal data (Albanese et al., 2019)?These are all interesting questions that should be addressed in future research.
Eligibility for participation consisted on reviewing participants' self-report on validated scales.Clinical interviews were only conducted with individuals who were unsure if they experienced UITs or who wished to talk to study personnel for other reasons (e.g.felt need for more information about study participation).One advantage of this approach is that it can lower the threshold for treatment seeking and thus improve scalability.However, this strategy could also potentially increase the risk of including individuals who are in need for more acute or intensive psychiatric care.Future research should evaluate to what extent self-report is sufficient to assess the presence of UITs, for example by performing full clinical interviews in a random subgroup of participants.
Limitations of this trial were the modest sample size and the use of a waiting-list control.This control group had some ecological validity as most parents with impairing UITs do not receive any kind of support.Nevertheless, a more stringent test of our hypothesis would require an active comparator.Another limitation was the predominantly female sample; fathers suffer from UITs related to their child to the same extent as mothers (Abramowitz et al., 2001(Abramowitz et al., , 2003;;Olofsdotter Lauri, Aspvall, Serlachius, et al., 2022) but seem to be more reluctant to seek help (Mojtabai, 2007).Future research should focus on developing recruitment strategies for fathers with distressing UITs.A final limitation was the use of a parallel mediation analysis, which opens up the possibility for reverse causation (that reductions in UITs mediated changes in negative appraisals).The sensitivity analysis did indeed indicate that the mediation model was highly sensitive to unmeasured confounders.Future trials should therefore use a large sample size and investigate time-lagged trajectories of change in the hypothesized mediator of change, and also include multiple competing mediators in the model, as recommended by Kazdin (2007).

Conclusion
With these limitations in mind, the outcomes from this trial provide initial evidence that a brief, online self-guided cognitive intervention that targets negative appraisals could significantly reduce distress and impairment associated with unwanted intrusive thoughts about intentionally harming one's child among new parents.The next step is to conduct a large-scale trial with active control condition and long-term follow-ups to investigate the potential long-term benefits of this intervention.

Figure 1 .
Figure 1.Participant flow throughout the trail.

Table 1 .
Demographics and clinical characteristics.

Psychiatric diagnosis specified (self-reported)n (%)
a Interquartile range for the full sample was 10-26 points on the PTBC.

Table 2 .
Impact of the intervention on the primary and secondary outcome measures.
PTBC, Parental Thoughts and Behaviours Checklist; PSS, Personal Significance Scale; WSAS, the Work and Social Adjustment Scale; MADRS-S, Montgomery Åsberg Depression Rating Scale-Self report; ISI, Insomnia Severity Index; TOPSE, Tool to measure parenting self-efficacy.

Table 3 .
Impact of the intervention on the PTBC and PSS, weekly measures (weeks 0-12).

Table 4 .
Self-reported engagement with the modules from participants.
Each module ended with the question «Did you find this module helpful?».Table4displays each item response rate for each module by the participants (n = 18).