Study design

This study is part of the Integrated Health Monitor COVID-19, a five-year, multimethod research project on the health effects of the COVID-19 pandemic in the Netherlands (for details see van Duinkerken et al. [22]). The present study is based on the Dutch Public Health Monitor Adults and Elderly monitor (DPHM) conducted in September to December 2022. The DPHM consists of a comprehensive survey encompassing various questions about health, well-being, and lifestyle, and is normally conducted every four years. During the Integral Health Monitor COVID-19, the DPHM is conducted every two years and focused on topics related to the COVID-19 pandemic.

Procedures

The data collection for the DPHM2022 took place through a random sample drawn from the Basic Register for Population (BRP). The BRP is based on the digitized Municipal Personal Records Database and includes personal data (names, addresses, marital status) for every person that is enlisted in a municipality. The random sample that is drawn from the BRP forms a representative sample of the Dutch population. Individuals of 18 years and older were invited by letter in the fall of 2022 to complete the questionnaire digitally or on paper. Those invited to partake in the study received a letter outlining study’s purpose, emphasizing the voluntary nature of participation without the need for providing a reason for refusal. The participants provided written consent. They completed the survey from September to December 2022. The net response rate of the survey was 34% (N = 364,557).

Measures

Exposure to potentially traumatic and stressful events.

Respondents were presented with a list of eleven potentially traumatic and other life events (from now on ‘events’) they could have experienced during the pandemic (from March 2020 up until the moment of responding (between September-December 2022). This list includes events specific to the COVID-19 pandemic and events that are not related to the pandemic. Table 1 presents an overview of these events and if these events fulfill the DSM-5 exposure criterion, fulfill the ICD-11 exposure criterion, or fulfills neither exposure criterion (marked by “✓”).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Examined events and categorization of events.

https://doi.org/10.1371/journal.pone.0347315.t001

Subsequently, when respondents indicated experiencing one or more of these events, they were asked whether they were still affected by these events. If the respondent answered ‘yes,’ and there were multiple events causing distress, they were asked to identify the most distressing event. Respondents were also asked to specify the time since the occurrence of this most distressing event. Some of these events may meet the ICD-11 criteria for a traumatic event. In the main analysis, only the actual death of a loved one was included, while severe illness and hospitalizations (whether related to COVID-19 or not) were excluded. Sensitivity analyses were conducted to assess whether including these events in the ICD-11 category altered the results.

Ethics statement

The study was subject to ethical approval by the Medical Research Ethics Committee (MREC) of Amsterdam UMC. It was deemed exempt from further review as it was declared not subject to the WMO (Medical Research Involving Human Subjects Act). All participants have given written informed consent before they responded to the survey.

Demographic characteristics

In total, 72,851 respondents (approximately 20% of the total sample) were exposed to and still felt affected by one of the events that they had experienced during the pandemic. This sample answered the PCL-5 and thus are included in this study. Their demographic information is included in S2 Table. A large proportion of respondents is older than 65 years old (39.9%), the sample includes relatively many women (59.6%) and about 17% of respondents had a migration background.

PCL-5 sum score across different event categories

The mean total sum score for the PCL-5 was 12.36. When comparing the categories, those answering the PCL-5 for an event that fulfilled the DSM-5 exposure criterion had the highest score (15.82). The events fulfilling the DSM-5 exposure criterion had a higher sum score (15.82) than the events that did not fulfill the DSM-5 exposure criterion. The events fulfilling the ICD-11 exposure criterion had a higher average sum score (12.95) than those that did not fulfill the ICD-11 exposure criterion. These differences between categories were mainly explained by differences in the intrusion domain (3.97 versus 3.12, and 3.60 versus 2.85 respectively) and the avoidance domain (1.66 versus 1.28, and 1.42 versus 1.22 respectively). The exception to this was the events that fulfilled the DSM-5 exposure criterion, as they scored considerably higher on average for each domain than the events that did not fulfill the DSM-5 exposure criterion. These sum scores are presented in Table 2.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Sum scores of the PCL-5.

https://doi.org/10.1371/journal.pone.0347315.t002

Sensitivity analyses were executed to account for possible confounders with regards to time since the event and demographics. The association between fulfilling the DSM-5 exposure criterion and the total PTSD mean score is confounded by age and financial difficulties, and that pattern is seen for all sub domains. However, the association between fulfilling the DSM-5 exposure criterion and the sum score on the PCL-5 is already small (3.81), and the change when adding age (from 3.81 to 3.17) or financial difficulties (from 3.81 to 3.16) to the model is limited. The association between fulfilling the ICD-11 exposure criterion and the sum score on the PCL-5 is confounded by age. However, the association between fulfilling the ICD-11 exposure criterion and the sum score on the PCL-5 is even smaller (1.11), and the change when adding age (from 1.11 to 1.23) is limited. More details are included in the S3 Table.

DSM-5 prevalence across different event categories

The overall prevalence of PTSD symptoms based on DSM-5 criteria was 9.74%. This prevalence was higher among respondents who experienced an event that fulfilled the DSM-5 exposure criterion (16.54%) compared to those whose most distressing event did not meet this criterion (9.05%). A similar, though less pronounced, pattern was observed when applying the ICD-11 exposure criterion (10.37% vs. 9.19%). Events involving direct life threat or violence, such as sexual violence (61.32%), physical violence or threats (35.40%), and life-threatening accidents (17.25%), were associated with the highest prevalence. The prevalence by event is presented in Table 3. These results indicate that the average prevalence of PTSD symptoms is higher for events that meet the exposure criteria. However, even for events that do not fulfil this criterion, the symptom patterns observed still correspond to the DSM-5 symptom clusters, suggesting that such experiences can also be associated with clinically relevant posttraumatic stress reactions.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Prevalence of probable PTSD based on the PCL-5.

https://doi.org/10.1371/journal.pone.0347315.t003

Factor loadings of the model

Standardized factor loadings ranged from 0.54 to 0.90, indicating that each item contributed significantly to its corresponding latent factor. The factor loadings were highest in the intrusion and avoidance domains, across all groups. Detailed information on the factor loadings is included in S4 Table.

Model fit of the different models

In Table 4, the goodness-of-fit indices are presented. The χ² was significant in all events. There were no large variations between events with regards to model fit. In multigroup analysis, placing constraints on the factor loadings and error variances between the groups did significantly change the χ². However, it did not considerately worsen other model fit indices. Changes in CFI and RMSEA were below its thresholds [27,28]. This implies there were no noteworthy differences between any of the group-comparisons with regard to factor structure, thus the models were invariant.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Goodness-of-fit indices of the four-factor model over different events.

https://doi.org/10.1371/journal.pone.0347315.t004

Several sensitivity analyses were performed. Firstly, the different inclusions in ICD-11 criteria did not make a difference: both models showed measurement invariance. Secondly, there were also no variations between COVID-19 related events and other events. Thirdly, running the models separately for each age group showed that measurement invariance is present across all age groups. These results can be found in S5 Table.

Findings

There is an ongoing debate about the exposure criterion of PTSD in the DSM-5 and ICD-11. Studies examining the prevalence of PTSD following the COVID-19 pandemic are criticized because they assessed stressful pandemic-related events which do not fulfill the exposure criterion of the DSM-5 and ICD-11. In this study, we argued that if stressful pandemic-related events are indeed distinct events compared to events that fulfill the exposure criterion, there should be measurement noninvariance of PTSD symptom clusters. The findings of this large, population-based study do not support this hypothesis. All four PTSD symptom clusters – intrusion, avoidance, negative alterations in cognition and mood, and hyperarousal – were observed across all groups, indicating measurement invariance of these clusters across the different event categories. Even though the χ² is significant, this could be caused by the large sample size leading to incorrectly rejecting the model [29].

The analysis of PCL-5 sum scores show that the scores in the two symptom clusters most directly related to the event (intrusion and avoidance) and average prevalence is are higher for the events fulfilling the exposure criteria. It is important to note that even events that meet the current PTSD exposure criteria in the DSM-5 and ICD-11 can vary in their associated risk of developing the disorder. For instance, some events, such as rape, are linked to a significantly higher risk of PTSD compared to others, like natural disasters [30].

The combination of our results suggests two things. First, all symptom clusters can be identified across the different types of events, which is consistent with the notion that PTSD symptoms reflect the same underlying latent structure irrespective of the precipitating event. Secondly, events that meet the exposure criteria, particularly those defined by the DSM-5, tend to result in higher scores of PTSD symptoms. This difference is mainly in the intrusion and avoidance domain which are ‘typical’ PTSD symptom clusters. This difference in scores between events that meet the ICD-11 exposure criterion and those that do not is even smaller. One key difference between the ICD-11 and DSM-5 criteria is that the loss of a loved one can fulfill the ICD-11 exposure criterion but not the DSM-5 criterion.

Earlier studies showed that individuals who experienced stressful events that do not fit the PTSD exposure criteria in the subsequent DSM and ICD versions, suffer from event-related intrusions, avoidance reactions, negative alterations in cognition and mood, and hyperarousal symptoms similar to individuals who experienced stressful events that do fit the exposure criteria. Ignoring the exposure criterion, these events can lead to PTSD and in some cases, they even lead to a higher prevalence of PTSD than events that did fit the exposure criterion [11–16]. The ongoing revisions of the exposure criterion in the DSM and ICD across different versions highlight its variability and subjectivity. These ongoing revisions lead to the striking situation where individuals who experience potentially traumatic events may be diagnosed with PTSD in one version of these manuals but not in another [13,31]. This illustrates that the exposure criterion is not fixed and is open to debate and potential improvement.

From a public health perspective, it is also important to consider the broader societal impact. Even though the average prevalence is lower for events that do not fulfil the exposure criterion, these events are more common in the general population. As shown in our data, experiences such as worrying about loved ones contracting COVID-19 or being unable to say farewell due to containment measures were reported by large groups, whereas direct violence or life-threatening accidents were relatively rare. Consequently, even if only a small proportion of those affected by these widespread events develop PTSD, the absolute number of individuals suffering from clinically relevant posttraumatic distress may still be considerable.

This study suggests that reliance on a narrow exposure criterion may be overly restrictive and that the exposure criterion may warrant reconsideration. Reassessing this criterion in light of a broader spectrum of potentially traumatic experiences may facilitate a more accurate understanding and identification of PTSD across diverse populations. The observed measurement invariance indicates that trauma-related reactions reflect similar underlying symptom clusters across event types, supporting the usefulness of a dimensional approach to trauma. At the same time, further research is needed to establish whether these findings generalize to clinically diagnosed samples.

In addition, our findings support the view that COVID-19-related experiences can be meaningfully examined from a trauma perspective and that PTSD represents a relevant outcome in assessments of the pandemic’s mental health impact when symptom levels exceed established cut-off values. Importantly, the current findings highlight that a thorough assessment of respondents’ exposure to events remains essential. Studies must always clarify to which events the reported intrusions, avoidance reactions, negative alterations in cognition and mood, and hyperarousal symptoms relate to. It must be noted that research on PTSD resulting from the pandemic too often neglect this important detail [32], making it difficult to determine if the examined symptoms are genuinely caused by a pandemic-related event or by another event. Without identifying the triggering event, the accuracy and relevance of these studies are compromised, ultimately hindering our understanding and ability to address pandemic-induced PTSD effectively.

Strengths and limitations

One of the key strengths of our study is the large sample size, which increases statistical power and precision of our results, making our conclusions more generalizable. Additionally, our study benefits from including various types of events, which allows for robust comparative analysis. Importantly, our sample is drawn from the general population rather than a biased group, such as individuals already seeking psychological support. This approach ensures that our findings are more reflective of the broader community and not limited to a specific subpopulation, thereby providing a more accurate representation of the experiences and behaviors of the general population.

The DPMH 2022 survey has limitations that are important to consider when interpreting the findings. One notable limitation is recall bias, where participants may not accurately remember past experiences and perceptions, potentially leading to biased responses. The time since exposure could range up to two and a half years, since the beginning of the pandemic. However, the study primarily focuses on PTSD symptoms experienced in the past four weeks, limiting the impact of potential recall bias. Another limitation is selection bias, since participation was voluntary, possibly resulting in a sample that does not fully represent the entire adult population. Despite efforts to diversify recruitment, reaching certain groups, particularly the most vulnerable, proved challenging. The sample consists of an overrepresentation of elderly, almost 40% of the sample is older than 65 years while 20.2% of the Dutch adult population was older than 65 years in 2022, and women, which form almost 60% of the sample while 50.3% of the adult Dutch population are women [33]. However, the focus of this study is on the factor structure of PTSD and not its prevalence, thus the lack of generalizability is less problematic. When looking at the age groups separately, measurement invariance is observed in all groups.

A common critique of studies that show that PTSD symptoms do arise after experiencing events that do not fulfill the exposure criterion is that they rely on self-report surveys rather than clinical interviews, which can lead to bias [3]. The same applies to to this study, all results are based on self-report measures and not on clinical diagnoses. Consequently, this study cannot draw conclusions about clinically confirmed PTSD diagnoses. At the same time, it is important to note that PTSD cannot be measured objectively [34], both self-report and clinical interviews rely on subjective symptom reports. The PCL-5, while not a diagnostic tool, assesses the same underlying construct of PTSD symptoms and has demonstrated good correspondence with clinician-based diagnoses, including in Dutch validation studies [35]. While the current study has attempted to investigate whether PTSD symptoms arise following events that do not meet the formal exposure criterion using measurement invariance within the PCL-5, replication with clinician-administered diagnostic interviews is needed to determine whether the exposure criterion should be reconsidered in the context of clinically diagnosed PTSD.

A limitation of this study is that the data were collected during the COVID-19 pandemic, an unusual period that may have influenced participants’ experiences. To enhance the generalizability of the findings, further research should examine similar traumatic events outside of the pandemic context, such as the sudden hospitalization or death of a loved one.

Implications

The debate surrounding the PTSD exposure criterion is ongoing, and based on past revisions of the DSM and ICD, further modifications can be expected. If an exposure criterion remains a requirement for diagnosis, the critical question is how to determine which events should be included or excluded. We argue that assessing measurement invariance across different types of stressful events is a key empirical method for guiding such decisions. Events that exhibit measurement invariance with those already included in the exposure criterion could be considered for inclusion, whereas those that do not may warrant exclusion. Future research should prioritize evaluating the measurement invariance of stressful events that remain under debate, as this would provide crucial data for refining the definition of potentially traumatic events.

Importantly, our proposal to include events that demonstrate measurement invariance of PTSD symptoms does not imply that, similar to events that do meet the current stressor criteria, all events are equally severe. Only events associated with symptom levels reaching clinically meaningful thresholds, such as those captured by the PCL-5, are relevant in this context. In other words, the focus remains on trauma-related reactions that meet a clinically significant threshold, ensuring that normal stress responses in the community are not inappropriately labeled as PTSD. As events currently recognized in the exposure criterion, such as war, sexual violence, and natural disasters, differ in their psychological impact and prevalence [30], this cannot be used as an argument against including new events within the exposure criterion. This is further supported by regression analyses to show differences in prevalence rates between event types, using the same list of events and database as the current study [36]. Expanding the scope of eligible events based on measurement properties does not imply that all qualifying events have the same psychological consequences. One notable example is the DSM-5 criterion that considers learning about the actual or threatened death of a close family member or friend traumatic only if the event was violent or accidental. This framework raises questions about whether deaths from severe life-threatening infections, cancer, or suicide should also qualify. Applying measurement invariance testing to such cases could provide an objective and empirically grounded approach to determining which events should be included in the PTSD exposure criterion. This approach offers a way to move beyond subjective determinations, contributing to a more scientifically rigorous and data-driven foundation for future revisions of PTSD diagnostic criteria. In future research, this approach should be applied to diverse events, populations, and diagnostic methods (e.g., clinical interviews rather than only self-report instruments).