This is a prospective longitudinal epidemiological study (MD-Care) with 3 measurement time points (Figure 1), examining the psychosocial impact of the attack on the Magdeburg Christmas market in Germany. The primary aim of the study is to examine the levels, longitudinal trajectories, and associated factors and predictors across the following three domains: (1) psychosocial, mental health, and somatic burdens; (2) individual resources and protective factors; and (3) mental health service use and support needs.

The overarching research aims and the 3 main domains of investigation were conceptualized prior to the start of data collection. However, more specific research questions and their methodological and analytical approach developed progressively over the course of the study. The findings are intended to highlight unmet care needs and guide the continued improvement of psychological emergency services.

To be eligible to take part in the study, participants had to be aged 18 years or older, have sufficient proficiency in the German language, and provide written informed consent. Affected individuals (eg, individuals directly affected by the attack, witnesses, first responders, relatives and acquaintances, and health care professionals) as well as unaffected individuals could take part in the study. Affected participants were asked to specify if they are directly affected as a survivor or on-site witness or indirectly affected as a relative or acquaintance of a directly affected person, a first responder (eg, paramedic, police officer, or firefighter), a member of the medical or psychosocial staff (eg, doctor or therapist), or other (asked to specify). Data were collected at 3 time points: 3 months after the attack, 6 months after the attack, and 15 months after the attack. From March to April 2025, individuals completed the baseline survey (t1). The second measurement point (t2) took place between June and July of 2025. The t3 survey was conducted between March and April 2026, 12 months after the baseline survey. Data were collected pseudonymously via the online platform LimeSurvey [25], with personal identifiers being managed by a data trustee. Recruitment channels for the baseline survey included the distribution of informational materials and study flyers (Multimedia Appendix 1) at the University Medical Center Magdeburg, Germany, as well as promotion of the study through social media and a TV feature to raise awareness of the study. No incentives were offered to participate in the study. For t2 and t3, participants were sent an email at an address which they had provided in t1, with an invitation and the link to participate in the next survey. The invitation emails were sent out a total of 3 times during the survey period as reminders to complete the survey. More details on the online surveys can be found in the CHERRIES (Checklist for Reporting Results of Internet E-Surveys; Checklist 2) [26]. Additionally, qualitative interviews will be conducted with a subset of participants after the completion of the t3 survey. At the end of the t2 survey, the participants were asked whether they would be willing to take part in an in-depth interview on the psychosocial effect of the attack. A total of 127 indicated their willingness to participate, 78 of whom were affected individuals. Recruitment for the qualitative interviews will start in the months following the t3 survey. Quantitative data collection was completed by May 2026.

As survey data were collected through a trusted third-party data management arrangement, individual responses could not be directly linked to participants by the study team for the purpose of actively monitoring distress during the surveys. However, at the beginning and end of each survey, all participants were informed that they could contact the study team at any time with questions or concerns arising from their participation. The interviews will be conducted by trained and professionally qualified interviewers who will be instructed to remain attentive to signs of distress during the interview. If an interviewer notices any signs of distress or concerns regarding a participant’s well-being, these will be addressed directly with the participant. Additionally, participants will be informed that they may withdraw from the interviews at any time, without giving a reason and without any consequences.

The MD-Care surveys include items designed to capture the study’s primary aim of examining the levels, longitudinal trajectories, and associated factors and predictors across (1) psychosocial, mental health, and somatic burdens; (2) individual resources and protective factors; and (3) mental health service use and support needs. The questionnaires in the surveys addressing these themes were presented in a randomized order to minimize response bias. The surveys comprise self-developed (ad hoc) items as well as standardized questionnaires. Self-developed items include questions about the involvement in the attack, individuals’ subjective burden caused by the incident (eg, distress, psychological and physical well-being, media exposure, and experience of pain), as well as questions on the use, satisfaction, and evaluation of psychosocial care after the attack. The standardized questionnaires include instruments assessing depressive, somatic, and anxiety symptoms, along with individuals’ resilience, self-efficacy, social support, and sense of coherence. All instruments and items used are listed and described in Table 1. The table also specifies whether each instrument was applied within a cross-sectional or longitudinal design across the 3 measurement points.

All analyses will be conducted using SPSS Statistics (IBM Corp) [42], the statistical software R (R Foundation for Statistical Computing) [43], and Microsoft Excel [44].

Initially, the t1 dataset was processed and cleaned. This entailed the following steps. First, all participants who did not complete the entire survey from the first to the last page were deleted from the dataset. Second, 1 participant had to be excluded due to an unrealistically short survey completion time. Third, the response patterns were checked to examine whether participants selected the same response option repeatedly within the questionnaires. This was assessed by calculating LongString scores, that is, the maximum number of consecutive identical responses [45]. Cases with excessively long strings were flagged for further analysis. Upon closer examination and detailed plausibility checks, no participants were excluded in this step. These steps were completed with the t2 and t3 datasets.

Descriptive statistics will be used to show participant characteristics and variables such as sociodemographic information and reported complaints and needs. This will include measures of central tendency (mean and median) and dispersion (SD and variance) to provide a detailed overview of the participants’ psychosocial status and well-being. For visualization purposes, data will be presented using boxplots and bar charts.

Inferential and multivariate statistical analyses will follow a primarily hypothesis-driven approach, testing already given theoretical mechanisms. Measures that are not based on any given theory and are new in this study will be analyzed in an exploratory manner to deduce new theories on psychological mechanisms when facing uncontrollable adversity, which should undergo confirmational analysis in future studies. Analyses will be organized into cross-sectional and longitudinal components to study structural associations and dynamical associations respectively.

Cross-sectional analyses will examine correlations and generalized linear models, considering the distributional properties of the outcome. Longitudinal analyses will use generalized linear mixed-effects models [46] and latent growth curve models [47] to assess contemporaneous effects and lagged-effects models [48] to assess effects across measurement time points; both analyses will take into consideration the multilevel data structure of this study. The latter allows for the exploration of the temporal ordering of effects across constructs, deriving possible causal mechanisms.

The analysis of missing data will depend on the model class [49-51]: full information maximum likelihood (FIML) will be used where applicable, whereas multiple imputation via chained equations will be used where FIML is not directly applicable. Longitudinal measurement invariance will be analyzed where applicable. Attrition patterns will be evaluated using the Little missing completely at random test [52] to assess whether data are missing completely at random, and by modeling dropout as a function of baseline demographics and affectedness, to evaluate whether the missing at random assumption underlying both approaches is plausible [53,54]. Table 2 shows more specifics of the statistical analysis plan.

Within each family of tests addressing a shared research question, P values (with a significance level of α=.05) will be adjusted for multiple comparisons. Sensitivity analyses will include a comparison of imputed versus complete-case estimates.

While longitudinal within-person analyses will constitute the primary approach to examining changes and temporal associations, between-group comparisons of affected and nonaffected individuals will be conducted for specific research questions where such comparisons are theoretically meaningful. For example, these analyses may examine whether the association between media exposure related to the event and psychological outcomes differs between affected and nonaffected individuals. Such comparisons will be considered complementary and interpreted cautiously, given the inherent differences between affected and nonaffected populations.

This study obtained approval from the Research Ethics Committee of the Medical Faculty at Otto von Guericke University Magdeburg (approval 02/25). All participants received an information notice and had the right to withdraw their consent at any time.

Several strengths should be considered when evaluating the findings of the present study. First, the study included a relatively large participant number with satisfactory retention for t2 (256/336, 76.2%) and t3 (193/336, 57.4%). Second, the inclusion of 3 measurement time points allows for longitudinal within-person analyses. Third, to our knowledge, this is one of very few studies investigating the psychosocial impact of a vehicle-ramming attack in Germany, thereby addressing an important gap in the literature. Finally, the inclusion of ad hoc items with open-response formats provided participants with the opportunity to report experiences that may not be captured by standardized questionnaires, potentially allowing for the identification of previously overlooked aspects of the psychosocial impacts of an attack.

A key limitation of the MD-Care study is the nonprobability, convenience sampling strategy [55] based on open recruitment through flyers, social media, and television. This approach does not provide a defined sampling frame or response denominator, making it impossible to assess participation rates. Furthermore, participation was voluntary and based on self-administered surveys, which introduces motivation bias and may contribute to an overestimation of the prevalence of health problems. As a result, the findings should be interpreted with caution and cannot be generalized to the broader population of affected and unaffected individuals. In the immediate aftermath of a disaster, probability sampling is often not feasible. In Germany, strict data protection regulations generally prevent researchers from obtaining timely access to population registers or contact information for affected individuals.

As the attack on the Magdeburg Christmas market was an unexpected and unplanned event, the design of this study did not start until after the incident on December 20, 2024. The t1 baseline survey started in March 2025, allowing just over 2 months for the planning and implementation of the study. Consequently, certain aspects of the t1 survey could have benefited from more careful consideration. These limitations will be discussed, where relevant, in future publications on this study.

The findings of this study will be disseminated beyond the scientific community and made available to relevant political and public health stakeholders to support the further development of psychosocial emergency care and crisis response structures. In particular, the results will be communicated to political representatives and relevant regional and national stakeholders, including the Ministry of Science (Wissenschaftsministerium) of the German federal state of Saxony-Anhalt. Lessons learned from the response have already contributed to discussion regarding the optimization of mass casualty incident concepts and psychosocial emergency care in the region and may inform preparedness and response planning in other regions [56].

For future work, it would be helpful to have precise and uniform definitions of who is considered affected vs not affected and directly affected vs indirectly affected, as, to our knowledge, no clear, established, and widely accepted definitions currently exist. A study examining the vehicle-ramming attack in Münster, Germany, on April 7, 2018 [57] considered individuals directly affected if they were survivors of the attack, witnessed it firsthand, or had a significant other who was a survivor or direct witness. Individuals were considered indirectly affected if they were not directly exposed but experienced distress after learning about the attack, such as concerns about their own or their family’s safety. The present study understands directly affected individuals as survivors or on-site witnesses of the attack and indirectly affected individuals as relatives or acquaintances of directly affected individuals, as well as first responders, medical or psychosocial staff, or others involved in the aftermath.

To improve psychosocial aftercare, it is vital to implement structured systems for identifying and monitoring affected individuals, as indicated in international guidelines [58-61]. Stene et al [58] concluded that standardized systems for registration, long-term follow-ups, and ongoing evaluation of the psychosocial care services can help ensure more consistent and effective responses to mass violence incidents. While this study does not focus on evaluating or implementing such standardized systems, the results may indirectly contribute ideas for future implementation. Malota et al [62] have already described a digital tool that helps distribute patients among hospitals more effectively. This reduced the risk of one hospital becoming too crowded. If emergency services (including psychosocial services) regularly use and practice with these types of systems, they may be better prepared for future large-scale emergencies.

The MD-Care study may contribute to the evidence landscape of the psychosocial impact of mass violence incidents and specifically vehicle-ramming attacks in Germany. By systematically disseminating the resulting evidence to relevant stakeholders, the study aims to contribute to the evidence base for future public health policy, service planning, and emergency preparedness.