In this retrospective study, data were extracted for all consecutive individuals exposed to violence examined by physicians in the Department of Forensic Medicine at Jean-Verdier Hospital in Bondy (Seine-Saint-Denis, France), in the Paris metropolitan area, between January 1, 2015, and December 31, 2015. These data are not publicly available due to privacy restrictions, as they contain information that could compromise the confidentiality of research participants. The data used in this study resulted from the combination of 2 types of sources. On the one hand, it included systematically collected, routine, and standardized characteristics such as age, sex, and circumstances of violence (eg, locations). On the other hand, it involved retrieving corresponding medical certificates from electronic health records and extracting additional characteristics not included in the standard collection through simple textual analysis techniques (eg, searching for terms, variants, and lexical fields while accounting for typing errors or spelling variations). These analyses are facilitated by the fact that the certificates are in digital form and standardized. The features extracted in this manner primarily concern psychological symptoms.

This research study was conducted retrospectively from data obtained for clinical purposes. An ethics approval of the project (reference number CERHUPO 2015-07-03) was given by the institutional review board (IRB 00001072) CPP Ile-de-France 2 on July 9, 2015. An information note for patients was displayed in each consultation room. This information was also available in the welcome booklet provided to each hospitalized patient. Patients were informed about the potential statistical use of their personal data, which would be anonymized and used solely for research purposes. They could opt out of this use by contacting the department manager.

Patients reporting deliberate assault and battery were assessed for eligibility. We excluded patients younger than 10 years, those reporting unintended violence or neglect, those examined more than 30 days after the reported incident, those examined by physicians with fewer than 300 patients per year, and those assessed for a second evaluation of their functional impairment. Only patients with complete data were included in the analysis.

Three self-evaluated Likert scales were used to measure pain, fear, and the perception of a life threat at 2 distinct end points: at the time of the assault and during the medical interview. Each Likert item had 7 levels (from 0=no pain to 6=maximum pain). The physician asked the questions in a general manner and could repeat them up to 2 times while providing a visual graduated scale for the patient to use in self-rating. Regarding the pain scale, the physician asked (translated from French) “On a scale of 0-6, where 0 means no pain and 6 represents the maximum amount of pain you can imagine, how would you rate the pain you experienced during the assault?” followed by, “Similarly, on a scale of 0-6, where 0 means no pain and 6 represents the maximum amount of pain you can imagine, how would you rate the pain you are experiencing now?” Questions about fear and the perception of a life threat were asked in a similar manner (see the “Questions Asked During Consultations” section in Multimedia Appendix 1).

We introduced 2 physician-evaluated Likert scales, each with 7 levels from 0 to 6. One scale was used to measure the intensity of global functional impairment at the time of the consultation, with 0 indicating no functional impairment and 6 representing maximum functional impairment. The other scale was used to assess the quality of interaction between the physician and the patient. Physicians were asked to rate the quality of interaction from 0 to 6, with 0 indicating the worst quality and 6 representing the most satisfying interaction. Finally, we introduced the psychosomatic index, a 7-level scale from 0 to 6. Physicians rated this index from 0 if physical trauma was the only source of functional impairment to 6 if psychological trauma was the only source of functional impairment [13].

The optimal data set partition identified 5 profiles (see the “Identification of Homogeneous Profiles” section in Multimedia Appendix 1; also see Figures S1 and S2 in Multimedia Appendix 1). Table 1 presents the characteristics of each profile. Profile A included mostly individuals exposed to single assailants and low-impact assaults, resulting in lower levels of functional impairment (TIW median 3, IQR 1-4) and rarely causing physical injury (538/779, 69.1%, patients without any somatic traumatic lesions), with a majority being women (474/779, 60.9%). Profile B mainly included males exposed to violence (564/749, 75.3%) involving multiple (585/749, 78.1%) and often unknown (464/749, 61.9%) assailants. Profile C consisted primarily of situations involving police violence (567/719, 78.9%), involving young (median age 22, IQR 18-28 years) and predominantly male (687/719, 95.5%) patients held in police custody (678/719, 94.3%). Profile D comprised mainly individuals exposed to single assailants and high-impact assaults, resulting in higher levels of functional impairment (TIW median 4 days, IQR 3-6 days) and systematic physical injury (1092/1092, 100%, patients with somatic lesions), with a majority being men (852/1092, 78.0%). Profile E included mostly women (703/841, 83.6%) who were exposed to repeated assaults (679/841, 80.7%) perpetrated by intimate partners (678/841, 80.6%).

Scales for each profile and patient-reported symptoms related to psychological trauma are described in Table 2. Self-rated scales had higher scores at the time of the assault compared with the consultation time (Table 2). Women who experienced repeated IPV (profile E) showed high scores for fear during both the assault (median 6, IQR 4-6) and consultation (median 5, IQR 2-6), as well as for the perception of a life threat during the assault (median 4, IQR 0-6) and consultation (median 2, IQR 0-6). Individuals exposed to violence involving multiple assailants (profile B) had high scores for the perception of a life threat during the assault (median 3, IQR 0-6). Young males exposed to police violence (profile C) reported low scores for pain during the assault (median 3, IQR 2-5) and low scores for fear both during the assault (median 3, IQR 0-5) and during the consultation (median 0, IQR 0-3). Women experiencing low-impact assaults (profile A) reported low scores for pain during the consultation (median 2, IQR 0-4).

We found significant differences (P<.001) in median ratings between physicians for each scale (Table S6 in Multimedia Appendix 1). Results for pain during the assault, comparing pairs of physicians in the overall population and within profiles of situations involving violence, are shown in Multimedia Appendix 2 (also see Table S3 in Multimedia Appendix 1). All physicians had examined more than 30 patients in each profile of situations involving violence (Table S2 in Multimedia Appendix 1). For pain during the assault, we observed significant differences between 16 out of 45 pairs of physicians in the overall population, while fewer than 4 pairs of physicians showed significant differences when assessed within profiles of situations involving violence (Multimedia Appendix 2). Results were similar for other self-rated scales, but most pairwise comparisons remained significant for physician-rated scales (Table S3 in Multimedia Appendix 1).

All pain, fear, and life threat scales were significantly associated with functional impairment as measured by TIW (Table 3).

In this study, we investigated tools for quantifying subjective elements related to functional impairment in individuals exposed to violence, based on a typology of situations involving violence derived from massively multivariate data. A clustering algorithm identified 5 remarkably stable profiles of violent situations, each with comparable sizes and characteristics consistent with clinical practice. This typology allowed us to evaluate the subjective scales with the assumption that within each type, the characteristics of violence and exposed individuals were homogeneous. Consequently, any remaining variability in scale results could reasonably be attributed to inter- and intrarater performance. All pain, fear, and life threat scales were significantly associated with functional impairment, indicating that they effectively measure elements contributing to the overall assessment of functional impairment. These scales highlighted different patterns of psychological trauma associated with the various profiles of situations involving violence. There were a few significant differences in ratings between physicians for self-rated scales within the same violent situation profiles. Finally, we found mild to good inter- and intrarater reproducibility for all scales within the same violent situation profiles.

Patients were clustered consistently within intuitive clinical categories (Table 1). Differences in scale scores between violent situation profiles should be examined in light of these clinical aspects (Table 2). First, individuals exposed to violence involving multiple assailants (profile B) more frequently reported a perception of a life threat during the assault, which aligns with the high intensity of these assaults, often involving blows that caused falls (541/749, 72.2%) or being knocked to the ground (135/749, 18%). Interestingly, while the perception of a life threat was higher during the assault, it did not seem to persist over time and was lower at the time of consultation. Women exposed to IPV (profile E) displayed a different pattern of fear and perception of a life threat, with higher scores recorded both during the assault and at the consultation. This persistence of symptoms is consistent with these patients experiencing mostly repeated assaults (679/841, 80.7%) by known, intimate assailants, which increases the credibility and fear of future attacks. The scales were consistent with patients’ reported symptoms such as fear, sleeping disorders, and loss of appetite, while providing richer and more comprehensible information. Second, young men held in police custody and reporting police violence (profile C) exhibited lower scores for fear both during the assault and at the consultation, as well as a lower score for pain during the assault (Table 2). These lower scores may reflect underlying psychosocial behaviors of pain or fear denial in this population. These findings are consistent with their lower functional impairment measurements (median 2, IQR 1-3 days of TIW). However, it has been previously shown [17] that their functional impairment was systematically evaluated at lower values and that they were examined in shorter intervals after the assault (median 5, IQR 3-9 hours) and under specific circumstances (in police custody).

There were differences in ratings between physicians in the overall population (Table S6 in Multimedia Appendix 1) for all scales and for most pairs of physicians (Table S3 in Multimedia Appendix 1). However, because all physicians were instructed to ask the questions in the same manner (see the “Questions Asked During Consultations” section in Multimedia Appendix 1), these observed differences were likely attributable to variations in the situations involving violence. For instance, some physicians more frequently consulted patients from cluster A than cluster D, possibly due to planning reasons or personal preferences. Supporting this interpretation, we found that within the same profiles of situations of violence, only a few significant differences for self-rated scales remained (Table S3 in Multimedia Appendix 1 and Multimedia Appendix 2).

Self-rated scales were associated with functional impairment as measured by TIW; specifically, TIW increased with higher scale scores. This association underscores the clinical relevance of these scales, as subjective elements reported by patients themselves were directly linked to TIW, which is otherwise solely determined by physician assessment. These scales, now shown to be associated with functional impairment, could provide standardized access to a patient’s perspective and enhance clinical examination.

The reproducibility of the scales within the same profiles of situations of violence was found to be mild to good, ranging from 0.43 to 0.66 for interrater and 0.46 to 0.66 for intrarater reproducibility across the 5 situations of violence profiles. Higher reproducibility (>70%) can be achieved with similar scales [11] in a classic design, such as when patients are evaluated at a 2-week interval using the same scales. These differences could arise from evaluating patients in real-life conditions rather than in the controlled environment of a research study. Additionally, our study design made it challenging to determine whether strong specificities of individual situations persisted among matched patients. Such remaining specificities could lead to systematic differences in their scale scores, thereby reducing reproducibility. Increasing the number of situations of violence profiles helped test this hypothesis. As the number of profiles grew, the remaining specificities among patients within the same profiles decreased. A sensitivity analysis revealed slightly higher reproducibility values when differentiating 800 profiles (Tables S5 and S7 in Multimedia Appendix 1). This suggests that the reproducibility results were not underestimated and that the situations of violence profiles were sufficiently detailed.

Reproducibility was assessed across similar situations involving violence rather than individual patients seen at different time intervals to stay close to real-life conditions, as a classic design was not feasible in this context. First, a classic design could not assess the reproducibility of subjective elements at consultations because a first consultation for violence cannot be replicated. Second, if 2 consultations occurred close together, patients might remember their responses for subjective elements during the assault, whereas if consultations were distant, memories might fade, affecting the accuracy of their responses.

This study has several shortcomings. First, as a monocentric study, it may not fully represent forensic practice across the country. However, the population served by the study is among the most diverse in France in terms of geographical origins, socioeconomic status, and culture, according to the National Institute for Statistics and Economic Studies (Insee). Second, many patients (1000/5180, 19.31%) were excluded due to missing data. However, missing data were comparable for baseline characteristics. Third, combining physical and psychological impairments poses a challenge, and we may have lacked the power to detect small effect sizes. Finally, our definition of situations involving violence could have been enhanced by including information related to psychological trauma. However, psychological trauma symptoms were not systematically reported by physicians and could neither be reliably used nor provide informative data for determining the typology.

To our knowledge, this study is the first to investigate the quantification of subjective elements related to functional impairment in individuals exposed to violence within the context of daily medical examinations in a judicial setting. Although the study is based on data collected 8 years ago, it remains relevant and current, as the activity of the forensic medicine department has not significantly changed since 2015, neither in volume nor in the reasons for the examination. The missions remain the same, and the practices have not significantly evolved in either direction. No modifications to the French penal code that could affect the relevance of this study have been noted either. Quantification of subjective elements provided direct access to patients’ opinions in a synthetic and reliable manner and revealed significant differences between situations involving violence, such as variations in fear and perception of a life threat among those exposed to IPV. Further studies are planned to explore whether these tools can lead to straightforward interpretations by magistrates and judicial police officers and to assess their impact on judicial decisions.

The pain, fear, and life threat scales were correlated with higher functional impairment, aligning with expert knowledge, and demonstrated fair reproducibility in real-life conditions for similar situations of violence. Subjective elements related to functional impairment in individuals exposed to violence can be quantified using Likert scales during medical interviews. High scores in fear and perception of a life threat, both during the assault and at the medical consultation, suggest greater impairment related to psychological trauma, particularly following IPV.