Prevalence of Weapons in the Health Care Setting: A Systematic Review and Meta-Analysis

Abstract

This study aimed to systematicically evaluate and quantify the prevalence of weapons in the health care setting. A systematic search of MEDLINE, Embase, Scopus, Web of Science, CINAHL, and EBSCO MegaFILE was performed from inception to January 12, 2024. The primary outcome was the prevalence of weapons in the health care setting on patients and/or visitors. Prevalence was pooled across studies and estimated using a random effects model. Subgroup analyses were done based on types of weapons, characteristics of weapon carriers, weapons screening/detection technology, and screened population characteristics. A total of 14 observational studies were included. All studies were from the United States and were published between 1984 and 2023. Weapons prevalence ranged from 0.4% to 26.3% among populations screened in the included studies. The overall pooled weapons prevalence was 4.0% (95% CI, 2.0%-7.8%). Most weapons were bladed (3.8%; 95% CI, 1.5%-8.9%), followed by other weapons (0.6%; 95% CI, 0.3%-1.3%), and firearms (0.1%; 95% CI, 0.02%-0.5%; P<.01). Weapons prevalence was 2.0% (95% CI, 0.7%-5.8%) among individuals entering the hospital setting, compared with 1.6% (95% CI, 0.7%-3.4%) of individuals entering the emergency department and highest (24.3%; 95% CI, 21.6%-27.2%) when major trauma patients were hand-searched. Prevalence was higher in males than that in females (11.1% vs 3.1%; P=.01). Weapons should be expected on individuals presenting to hospitals in the United States; however, prevalence varied widely based on the setting, type of patients, and detection method.

Article Highlights.

• •Weapons should be expected on individuals presenting to hospitals in the United States.
• •Weapons were detected among 4% of patients/visitors screened.

Workplace violence (WPV) against staff is a marked concern in health care^1,2 as workers in the health care and social service industries experience the highest rates of injuries caused by WPV and are 5 times as likely to experience an injury from WPV than workers overall.³ Although rare, acts of violence occurring in the health care setting that involve weapons result in significant morbidity and mortality to staff and bystanders.^4,5 These high-threat events such as health care–based shootings have been increasing in frequency since the early 2000s.^6,7

Given an increasingly volatile health care environment with significant concern for staff safety, institutions are looking toward novel strategies to protect staff. Weapons detection technology is rapidly advancing and becoming more prominent at entrances of health care institutions, screening patients and visitors for a multitude of weapons. This advancing technology offers a potential protective measure against high-threat violent events within the workplace, protecting staff, patients, and visitors. Previous literature has found weapons are commonly encountered by health care providers in the hospital environment;7, 8, 9 however, to date, there has been no study that has systematically synthesized data from all available studies to estimate the prevalence and types of weapons encountered within health care. The goal of this systematic review and meta-analysis was to determine the prevalence of weapons within the health care setting.

Methods

We reported this systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (2020) guideline checklist¹⁰ following a priori established protocol.

Eligibility Criteria

We included studies that met the following criteria: (1) observational studies; (2) reported prevalence of weapons within the health care (hospital) setting; and (3) included a denominator of number of patients/visitors screened or patient visits (volume) during the study period to calculate a rate of weapons detected. We included articles reported in any language and those conducted up until January 12, 2024. We excluded the following studies: (1) studies on the consequences of weapons in health care; (2) studies on hospital security guards carrying weapons; (3) qualitative studies; (4) commentaries lacking new unpublished data; (5) studies focused on knowledge and attitude without reporting data on the outcomes of interest; (6) studies in the nonhospital setting (eg, nursing homes and home health); and (7) studies in which a denominator of individuals screened or facility volume could not be determined.

Search Strategy

The following databases and/or platforms were searched on January 12, 2024: EBSCO CINAHL with full text (1963+), EBSCO MegaFILE (1800s), Ovid Embase (1974+), Ovid MEDLINE (1946+ including epub ahead of print, in-process, and other nonindexed citations), Scopus (1823+), and Web of Science Core Collection (Science Citation Index Expanded 1975+ and Emerging Sources Citation Index 2015+). Additional articles were identified by screening the bibliographies and citing references of the final studies that met the inclusion criteria. A gray literature search was performed in Google with review of the first 25 search results. Primary search terms were (weapons, firearms, knives, guns) AND (health facilities, hospitals, emergency department, outpatient clinics) AND (security screening, detection). No restrictions were set for language or date. The details of the full database search strategies are included in Supplement Tables 1 and 2 (available online at http://www.mcpiqojournal.org). The search strategies were developed by a medical librarian (D.J.G.) and clinician researcher with expertise in WPV (S.S.M.), covering the concepts of weapons, detection/screening, and health care settings and used a combination of keywords based on a previous literature review of the topic and standardized index terms.

Selection Process

Records identified during the search phase were exported to reference management software (EndNote) to enable the identification and removal of duplicates. Deduplication was performed in EndNote following the method by Bramer et al.¹¹ Records were then screened using Covidence, a cloud-based tool for systematic reviews. Covidence automatically removed additional duplicates. Two reviewers (S.S.M. and C.M.C.) independently performed the screening process at the title/abstract level as well as full text assessment of included records. Any disagreements during the screening process were resolved through discussion and consensus between the reviewers. The CitationChaser tool¹² identified 434 additional backward and forward citations for screening.

Data Extraction and Quality Assessment

Two reviewers (S.S.M. and C.M.C.) independently extracted data onto a standardized form in a Microsoft Excel spreadsheet. The extracted citation information; study location (geographic location); health care setting type (eg, psychiatric emergency department [ED], trauma center); study population (eg, all patients within a unit or only certain types of patients); study period; study design; weapon screening process (eg, metal detector, passive weapons screening technology [PWST], hand searches); study exclusion criteria (eg, prehospital patients not screened); definition of weapons used; overall number of weapons detected; overall number of individuals screened and/or unit patient volume during study period; number of specific types of weapons detected (eg, firearms, bladed, and other); and weapon-carrier characteristics (eg, gender). Disagreements were resolved by consensus.

The methodologic quality of each study was assessed by these 2 reviewers by consensus using previously published criteria for assessing risk of bias in prevalence studies that grades studies according to 10 assessment questions within the 2 domains of external and internal validity.¹³ An overall judgment on risk of bias (low, moderate, or high risk) was then made (Supplemental Table 2, available online at http://www.mcpiqojournal.org). For studies evaluating change in weapons prevalence with institutional change in method/process of weapons screening (eg, transition from targeted searches of certain patient populations to census searching of ED patient population with metal detectors),^8,14 we assigned risk of bias based on the postimplementation study methodology and used postimplementation data in our analysis.

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework was used to grade certainty in the pooled estimates.¹⁵ In GRADE, certainty in the estimates can be rated down due to methodologic limitations of the studies, imprecision (wide CIs), inconsistency, indirectness, or suspected publication bias.

Outcomes

The primary outcome was the overall prevalence of weapons in the (hospital) health care setting. Secondary outcomes of interest were prevalence of types of weapons, characteristics of weapon-carriers (eg, gender), weapons screening/detection technology, and screened population characteristics (eg, staff-directed searches, major trauma patients, all individuals entering facility). For subgroup analysis of different types of weapons, categories included firearms, bladed (knives, razors, etc) and other weapons (eg, Mace, Tasers, brass knuckles, and improvised weapons).

Statistical Analyses

Considering the heterogeneity in populations and settings across different studies, we used a random effects model with a maximum likelihood estimator for between study heterogeneity¹⁶ and a generalized linear mixed model with logit transformation for pooling across studies.¹⁷ Meta-analysis was conducted using the “meta”¹⁸ package in R software.¹⁹ We assessed heterogeneity between study-specific estimates using inconsistency index (I² statistic), which estimates the proportion of total variances across studies that is due to heterogeneity rather than chance. Heterogeneity was interpreted with I² values of 0% to 25%, 50% to 75%, and >75% consistent with low, moderate, and high heterogeneity, respectively. We investigated between study sources of heterogeneity using subgroup analyses by stratifying original estimates according to study characteristics as described earlier. In the event annual patient volume was provided but the study period was <365 days, a proportion of approximate patient volume during the study period was extrapolated based on number of days in the study period compared with provided patient volume over 365 days. In the event a study reported weapons prevalence per both patient volume and individual screens performed, we included the larger denominator (individual screens) in our analyses.

Results

Study Selection

From 1470 unique references identified using the prespecified systematic search strategy, 14 observational studies were included in the quantitative synthesis.^8,14,20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 Gray literature search did not identify additional prevalence studies. Figure 1 shows the schematic diagram of study selection. One study provided weapons data with an average annual ED patient volume but was excluded due to metal detector technology only being using for limited hours per day.³²

Figure 1.

Flowchart summarizing study identification and selection.

Study Characteristics

The Table^8,14,20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 describes the characteristics of the included studies. All 14 of these studies were performed in the United States and all were descriptive observational studies with the exception of one that was a letter to the editor detailing previously unpublished retrospective observational study data²² and another that was a descriptive analysis of weapons detection data within a survey study.²³ The median study period of the 14 studies was 339 (range, 35-5110) days. Study dates ranged from 1984 to 2023; 8 studies^14,20, 21, 22, 23^,27,28,31 were published before the year 2000, and 6 studies^8,24, 25, 26^,29,30 were published between 2000 and 2023. Four studies^20,21,23,27 from the 1980s described a weapons detection process consisting of hand searches of patients and/or their belongings; 3 of these were specific to psychiatric emergency patient populations. Two additional studies^28,31 described hand searches of patients and their belongings, but were specific to major trauma patients whose clothing and belongings had been fully removed in the trauma bay; in one of these studies, metal detector technology was incorporated toward the end of the study period.²⁸ The remaining studies involved patients and/or visitors passing through metal detectors^{8,14,22,29,30} or PWST.24, 25, 26^,33 Six studies^8,14,20,28, 29, 30 provided the definition of “weapon” they utilized for their study purpose. Five of these referenced including other items that could be used to cause bodily injury: 1 with further categorization of these items (eg screwdrivers),³⁰ 3 without particular item examples,^8,14,20 and 1 with limited examples (eg, “tools” and “other”).²⁹ Six studies^{20,21,23,27,28,31} provided the number of individuals screened overall, 5 provided patient volume during the study period^14,22 or annual patient volume from which study period patient volume was extrapolated,^24,26,29 and 3 studies^8,25,30 provided both number of individuals screened and patient volume.

Table.

Characteristics of Included Studies on the Prevalence of Weapons in the Health Care Setting

Study	Location/population	Study period	Design	Weapon screening process	Exclusions	Overall weapons per screens/volume	Weapon types	Weapon-carrier characteristics
Anderson et al20	Newark, NJ Emergency Psychiatric Unit	December 19, 1986, to July 22, 1987	Observational study	Hand search of patients admitted to hospital’s inpatient psych service	Patients refusing search	24 weapons/287 patient screens	Not reported	Male gender: 16/287 Female gender: 8/287
Goetz et al21	Portland, OR Academic ED	January 1, 1987, to August 31, 1988	Observational study	Hand search of patients and/or visitors identified as possibly dangerous by either security or medical personnel	None	89 weapons/500 screens	Firearms: 1 Bladed: 77 Other: 11 (included razor blades—number not quantified)	Male gender: 74/363 Female gender: 15/137
Irvin and Habas22	Detroit, MI Urban ED	August to October 1994 and 1995	Letter to the editor on unpublished observational study	Metal detector and package search	None	2800 weapons/33,665 patient visits	Firearms: 36 Bladed: 1779 Mace: 609 Other: 376	Not reported
McCulloch et al23	San Francisco, CA Psychiatric ED	1984 (4 mo)	Descriptive analysis of weapons detection within a survey study	Staff search patients’ belongings	None	14 weapons/175 patient screens	Not reported	Not reported
McGuire et al24: implementing	Rochester, MN Level 1 trauma center academic ED	January 19, 2022, to November 29, 2022	Observational study	PWST at main entrance	Prehospital arrivals	1518 weapons/67,101 patient visitsa	Firearms: 59 Bladed: 1233 TASER: 7 Mace: 167 Other: 52	Not reported
McGuire et al25: one	Rochester, MN Level 1 trauma center academic ED	April 1, 2022 to March 31, 2023	Observational study	PWST at main entrance	Prehospital arrivals	1741 weapons/247,926 screens (80,968 patient visits)	Firearms: 117 Bladed: 1217 Other: 407	Not reported
McGuire et al26: passive	Rochester, MN Level 1 trauma center academic ED	November 1, 2022 to December 6, 2022	Observational study	PWST at main entrance	Prehospital arrivals	166 weapons/7693 patient visitsa	Firearms: 6 Bladed: 130 TASER: 3 Mace: 27	Not reported
NcNiel and Binder27	San Francisco, CA University hospital-based psychiatric ED	14 mo (exact study period not described, published in 1987)	Observational study	Security search patients’ jackets and/or purses, conducting a more thorough search if clinically indicated	None	37 weapons/1012 patient screens	Not reported	Male gender: 26/1012 Female gender: 11/1012
Ordog et al28	Los Angeles, CA Urban university/county hospital ED (major trauma patients only)	1979-1993	Observational study	All major trauma patients had clothing removed and searched. In 1992, metal detectors implemented	Nontrauma patients	6881 weapons/26,134 patient screens	Firearms: 1123 Bladed: 5758	Male gender: 6018/26,134 Female gender: 733/26,134
Rankins and Hendey14	Fresno, CA Urban ED	1992-1996	Observational (pre/post) study	Walkthrough metal detector; ambulance arrivals wanded	Excluded household items such as screwdrivers and nail files	435 weapons/108,994 patient visits postintervention	Not reported	Not reported
Simon et al29	Atlanta, GA Health System (urban general hospital and pediatric hospital)	January 1, 2000, to August 31, 2000	Observational Study	Walkthrough metal detector	Identifiable hospital staff	3706 weapons/85,479 patient visitsa	Firearms: 4 Bladed: 2918 Mace: 255 Other: 529	Not reported
Smalley et al30	Cleveland, OH Health system (academic hospital, behavioral health hospitals, trauma/pediatric hospitals)	January 1, 2016, to March 17, 2017	Observational study	Walkthrough metal detector and handheld screening device (wand). All handbags underwent a manual search	Ambulance arrivals, patients who arrived with ischemic chest pain, stroke-like symptoms, or respiratory distress were excluded but were handwanded when feasible on room arrival	10,691 weapons/1,179,530 screens (346,323 patient visits)	Firearms: 34 Bladed: 7642 TASER: 250 Mace: 2231 Other: 534	Not reported
Vilke et al8	San Diego, CA Level 1 trauma center	March 15, 2021, to May 9, 2021	Observational (pre/post) study	Walkthrough metal detector and handheld screening device (wand). All handbags were visually inspected	For patients arriving by ambulance, security performed a bedside screening and baggage check	761 weapons/56,470 screens (31,901 patient visits) postintervention	Not reported	Not reported
Wasserberger et al31	Los Angeles, CA Trauma-receiving hospital (major trauma patients only)	1979-1987	Observational study	All major trauma patients had clothing removed and searched for weapons	Nontrauma patients	4796 weapons/21,456 major trauma patients screened	Firearms: 651 Bladed: 3564	Male gender: 4125/21,456 Female gender: 671/21,456

ED, emergency department; PWST, passive weapons screening technology.

^aPatient volume denominator was extrapolated for study period based on annual patient volume data provided.

Methodologic Quality of Included Studies

The risk of bias of included studies is shown in Supplemental Table 2. Risk of bias was high in 4 studies,^20,21,23,27 mostly due to the target population and/or sampling frame not being a close representation of the national population (eg targeted searches of psychiatric patients or patients/visitors identified as “possibly dangerous”), lack of random sample selection, and lack of reliability/validity of the weapon screening method (eg staff hand-searched patients’ jackets and/or purses). Risk of bias was moderate in 4 studies^22,26,28,31 and low in 6 studies.^{8,14,24,25,29,30}

Subgroup Analysis of Types of Individuals Screened

The overall weapons prevalence ranged from 0.4% to 26.3% among populations screened in the included studies. On pooled analysis, the overall weapons prevalence was 4.0% (95% CI, 2.0%-7.8%) (Figure 2). There was high heterogeneity among the studies (I²=100%). Prevalence of weapons varied by types of individuals screened. Prevalence was highest (24.3%; 95% CI, 21.6%-27.2%) among 2 studies that were specific to major trauma patients, followed by 1 study that had its staff hand search patients identified as “possibly dangerous” (17.8%; 95% CI, 14.6%-21.4%). Among 3 studies that were specific to handsearches of emergency psychiatric patients, prevalence of weapons was 6.0% (95% CI, 3.8%-9.4%). Among individuals entering the hospital setting, prevalence was 2.0% (95% CI, 0.7%-5.8%), compared with 1.6% (95% CI, 0.7%-3.4%) of individuals entering the ED setting.

Figure 2.

Subgroup analysis of prevalence of weapons per individual patients/visitors screened or department volume during study period, grouped by type (description) of individuals screened.

Subgroup Analysis by Patient Volume vs Individual Screens

Subgroup analysis based on whether overall weapons prevalence was reported per department volume (patients), or total individual screens (patients and visitors) revealed overall weapons prevalence of 2.4% (95% CI, 1.3%-4.4%) per department volume, compared with 5.3% (95% CI, 2.1%-13.2%) per individual screens (Figure 3).

Figure 3.

Subgroup analysis of overall weapons prevalence and whether it was reported per department volume (patients) or total individual screens (patients and visitors) during study period.

Subgroup Analysis of Types of Weapons

Nine studies characterized types of weapons detected. The overwhelming majority were bladed weapons (3.8%; 95% CI, 1.5%-8.9%), followed by other weapons (0.6%; 95% CI, 0.3%-1.3%), and firearms (0.1%; 95% CI, 0.02%-0.5%; P<.01).

Subgroup Analysis of Weapons Screening Methods

Prevalence of weapons also varied by method of weapons screening (Figure 4). Prevalence was again highest (24.3%; 95% CI, 21.6%-27.2%) among the 2 studies that were specific to major trauma patients whose clothing and belongings had been fully removed in the trauma bay, followed by 4 studies that had its staff hand search patients (8.2%; 95% CI, 4.5%-14.4%). Among individuals passing through metal detectors, prevalence was 1.8% (95% CI, 0.7%-4.7%) and among individuals screened by PWST, prevalence was 1.5% (95% CI, 0.8%-2.8%).

Figure 4.

Subgroup analysis of prevalence of weapons per individual patients/visitors screened or department volume during study period, grouped by method of weapons screening.

Weapon-Carrier Gender

Gender was the only weapon-carrier characteristic reported consistently by ≥2 (reported by 5) studies. Overall, male patients and/or visitors had a higher prevalence of carrying weapons (11.1%; 95% CI, 5.1%-22.6%), compared with females (3.1%; 95% CI, 1.6%-5.7%; P=.01).

Certainty in the Estimates

Using the GRADE framework, the certainty in the pooled prevalence estimates was moderate, primarily owing to the methodologic limitations of the included studies and inconsistency of the prevalence estimates that was not fully explained by subgroup analyses.

Discussion

This is the first systematic review and meta-analysis to evaluate the prevalence of weapons in the hospital setting. The prevalence of weapons was 4.0% from 14 studies, including 2.4% per department patient volume and 5.3% per individual screens. Not unexpected, prevalence varied significantly based on types of weapons detected, types of individuals screened, and screening methods/technology. Given differing methodology and screening procedures between studies (eg, early studies targeting psychiatric patients or patients identified by staff as possibly dangerous), we could not reliably analyze for temporal trends to determine whether the prevalence of weapons in health care is increasing over time in parallel with increasing active shooter events occurring throughout American society.^34,35

Several older studies in our review included methodology of targeting specific individual populations deemed to be at risk for weapons carrying, an action now recognized (and discouraged) as a form of profiling. Other studies included methodology not replicable among larger study populations, such as searching for weapons on major trauma patients who had had clothing and belongings fully removed in the trauma bay as part of their emergent trauma evaluation. Given these weapons screening methods are not replicable across current large-scale hospital-based visitor or patient populations, more relevant and reliable studies to focus on may include the 8 studies that reported on screening all individuals entering into the ED^8,14,22,24, 25, 26 or hospital environment^29,30 through metal detectors^{8,14,22,29,30} or PWST.24, 25, 26 Among these, weapons prevalence ranged from 2.0% on individuals entering the hospital setting to 1.6% among individuals entering the ED setting. It is important to note that several of these studies excluded patients arriving by emergency medical services owing to limitations of the technology used.

The results of this meta-analysis reflect a previous survey study by Kansagra et al³⁶ in which 20% of ED staff (clinicians and nurses) surveyed reported that guns or knives were brought into the ED on a daily or weekly basis.³⁶ Our findings are pertinent to the current health care environment, given the Occupational Safety and Health Act of 1970 requires employers such as hospitals assure safe and healthful working conditions for employees, including the contingency of an active shooter incident,³⁷ which has become an increasing threat to health care facilities and employees.⁴ Additionally, previous literature has shown that weapons screening technology at hospital entrances, such as through metal detectors and PWST are well-received by patients and visitors and even improve their sense of safety within the health care environment.^26,38,39 The results of this review suggest that weapons should be expected on individuals presenting to hospitals within the United States. Health care staff should recognize this potential threat and institutions should consider weapons detection processes to prioritize deterrence of weapons from the health care setting.

In addition to the limitations of the individual studies, our meta-analysis has several limitations. Despite conducting subgroup analyses to explore heterogeneity, the included studies had different time periods, study populations, screening procedures and technology, and definitions of weapons. The 2 health system studies^29,30 we categorized in our subanalysis as hospital populations likely included other patient population subcategories (eg, ED, psychiatric, and major traumatic patients) that we were unable to differentiate with the available data. We did not account for geopolitical variability among study sites. As such, it is possible that prevalence of weapons in health care is influenced by local ordinances, crime rates, state firearmcarry laws, and preferences and political views of communities in which individual studies were performed. It is also possible that the same patients have been repeat carriers of weapons in the same study or across studies, which can violate the statistical independence assumption; however, we do not have access to such data. The nature of our meta-analysis question regarding prevalence of weapons in health care precludes study designs such as randomized controlled trials and was only amenable to observational studies. Overall, we encountered high heterogeneity across studies as we expected in a prevalence meta-analysis. In order to capture as comprehensive dataset as possible, we included studies that did not include individual screens or exact patient volume during a specific study period <365 days, as long as the study included annual patient volume (and then extrapolated a proportion of approximate patient volume during the study period with this annual volume). We recognize that ED and hospital visits vary based on time of day as well as time of year, and our extrapolated denominator may not reflect a true individual denominator during the particular study period.

Conclusion

Weapons are commonly brought into medical facilities, although the prevalence varies widely based on types of weapons detected, types of individuals screened, and screening methods/technology. Future high-quality, unbiased research is needed to better determine risk factors and temporal trends for weapons in health care.

Potential Competing Interests

The authors report no competing interests.