The role of religiosity and spirituality in interpersonal violence: a systematic review and meta-analysis

Abstract

Objectives:

Religiosity and spirituality (R/S) have been negatively associated with several mental health problems, including delinquency. The study aimed to investigate the relationship between R/S and interpersonal violence using a systematic review.

Methods:

We conducted a descriptive systematic review followed by meta-analyses using seven different databases. We included observational studies that assessed the relationship between R/S and different types of interpersonal violence (physical and sexual aggression and domestic violence).

Results:

A total of 16,599 articles were screened in the databases and, after applying the eligibility criteria, 67 were included in the systematic review and 43 were included in the meta-analysis. The results showed that higher levels of R/S were significantly associated with decreased physical and sexual aggression, but not domestic violence. All selected studies evidenced sufficient methodological quality, with 26.8% being cohort studies. In the subanalyses, the role of R/S was more prevalent among adolescents.

Conclusion:

There is an inverse relationship between R/S and physical and sexual aggression, suggesting a protective role. However, these results were not observed for domestic violence. Healthcare professionals and managers should be aware of their patients’ beliefs when investigating interpersonal violence to create tailored interventions for reducing violent behavior.

Introduction

According to the World Health Organization, violence is the fourth leading cause of death worldwide among people aged 15-44 years, with approximately 1.3 million deaths registered annually.1 Non-fatal violence, such as assaults or physical, sexual, and/or psychological abuse is also very common, and its effects on survivors include mental health problems, such as higher levels of depression,2 post-traumatic stress disorder, increased anxiety and self-harming.3 It also causes physical health complications, including poor maternal and fetal outcomes for women,4 high-risk sexual behavior, and substance abuse.5 The consequences are more serious when traumatic experiences occur during childhood, showing a later association with illicit substance use,6 personality disorders and mental problems,7 and risky sexual behavior and criminal behavior.8,9

Thus, the adverse effects of violence should be considered a global mental health crisis with long-term social and economic consequences10,11 for which it is increasingly necessary to formulate control strategies.12 According to the DSM-5, multidimensional treatments incorporating cultural aspects should be considered when addressing the consequences of violence,11 i.e., understanding how people react to and interpret violence within their cultural context is a crucial factor in managing the consequences of violent acts.

Religiosity is the belief and practice of the doctrinal foundations of religion,13 while spirituality refers to a personal quest for the understanding of existential issues, which may not necessarily be linked to a particular religion.13 Spirituality can also be defined as the way people find meaning and purpose in life, and experience a connection with others and whatever they may define as sacred.14

Studies have shown that religiosity/spirituality (R/S) is correlated with enhanced psychological well-being, satisfaction, happiness, and lower depression, anxiety, and post-traumatic stress symptomatology.13,15,16 Consistent with these recommendations, spiritual and religious beliefs have been widely used as complementary treatments for mental health rehabilitation regarding depression, anxiety, substance abuse and suicide, yielding promising results.13,16

Moreover, R/S plays a protective role against violence and delinquency, deterring crime regardless of the type.17,18 For instance, nationally representative studies of adolescents and youth in the USA found fewer fights, gang fights, shootings, and stabbings among religious participants.19-21 Similarly, it has been reported that people with higher levels of R/S perpetrate fewer violent acts toward intimate partners,22 are less involved in risky sexual behavior,23 and more strongly condemn victimless crimes.24

The role of religion in deterring criminal behavior can be explained by belief in supernatural punishment/rewards (e.g., “I will not go to heaven if I harm others”),25 socialization,26,27 social support,28 and the encouragement of healthy behaviors and attitudes.29 The theory of social control proposes that for families, religious institutions act as educators and help construct normative beliefs that promote greater assistance, commitment, and involvement with society.28 Moreover, the rational choice theory suggests that religious individuals create self-impositions that increase the probability of feeling guilty about harmful attitudes and behavior, which reduces their expression toward others.30 Additionally, religious individuals usually associate with others who have similar beliefs, which positively reinforces and enhances morality.26,29

Nevertheless, the influence of R/S can move in different and even opposite directions within the same disease or condition.31 For instance, negative religious coping (e.g., “God is punishing me”) and religious fundamentalism may encourage violence. Saroglou32 published a meta-analytical review on the relationship between R/S and personality. The findings showed that intrinsic religiosity was positively associated with religious maturity and openness, while religious fundamentalism was negatively associated with openness.

To our knowledge, four systematic reviews have demonstrated a consistent, robust relationship between higher R/S and decreased delinquency and/or crime.17,18,31,33 However, most scales and validated instruments designed to assess delinquency entail illegal conduct, such as vandalism, propriety destruction, the sale and/or possession of drugs and weapons, and police detention, and violence may not necessarily be associated with delinquent acts. These constructs should be addressed separately. Therefore, there remains a paucity of reviews assessing R/S and interpersonal violence.

Thus, we aimed to fill this gap by investigating the relationship between R/S and interpersonal violence, including domestic violence, and physical and sexual aggression. By evaluating the real impact of R/S on interpersonal violence, our findings may help the design and implementation of preventive strategies to improve public health.

Methods

Study design and protocol registration

This systematic review and meta-analysis followed PRISMA guidelines.34 The protocol was registered in PROSPERO35 and is fully available on the National Institute for Health Research – Health Technology Assessment website (https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42018080979).

Eligibility criteria

Inclusion criteria

The main outcome in this review was any physically violent and/or aggressive act perpetrated against another person, i.e., interpersonal violence. According to the World Health Organization, interpersonal violence involves “violence between individuals, subdivided into family and intimate partner violence and community violence. The former category includes child maltreatment; intimate partner violence; and elder abuse, while the latter is broken down into acquaintance and stranger violence and includes youth violence; assault by strangers; violence related to property crimes; and violence in workplaces and other institutions.”36

Exclusion criteria

Articles assessing violence against property, risk behavior for violence, moral aspects of crime, or crime recidivism were excluded. We also excluded delinquency scales that assessed items of violence along with other criminal behaviors, such as the sale and/or possession of drugs, robbery, vandalism, and property crimes.

Concerning methodology, only studies that were published in peer-reviewed international indexed databases were included, since this type of article has more appropriate and robust scientific evidence. Additionally, manuscripts in languages other than English, Portuguese, or Spanish were excluded.

The PI(E)CO strategy for observational studies

The PICO components for our study were: Patients – general population who committed acts of interpersonal violence, regardless of sex, age, socioeconomic status or nationality; Exposure – individuals with high levels of R/S; Comparison – individuals with low levels of R/S. Outcomes – interpersonal violence outcomes (i.e., domestic violence and physical and sexual aggression).

Type of studies

Since our review investigated whether a relationship exists between R/S and interpersonal violence, only observational studies were assessed. These included: cohort, cross-sectional, and case-control studies.

Type of participants

We included studies investigating individuals who committed any type of violence against other individuals, with no restrictions regarding age, sex, previous history of criminal activity, or setting (e.g., individuals in prisons or reformatories).

Information sources

Seven different databases were used to search for and select publications regarding violent behavior and R/S from inception to November 11, 2020: Sociological Abstracts, Applied Social Sciences abstracts (ASSIA), National Criminal Justice Reference Service (NCJRS), PsycINFO, Scopus, PubMed, and Web of Science. Only publications in English, Spanish, or Portuguese were included. EndNote X4 software was used to search for and select the articles.

Search strategy

A Boolean expression was used to optimize the search for relevant studies according to the main objectives of the review. Pilot experiments were conducted within the databases to ensure the accuracy of the expression. The final version was: (spirit* OR religi* OR faith OR god) AND (violence OR violent behavior OR aggressive behavior OR deviant behavior OR delinquency OR delinquent behavior). The expressions developed for each database are listed in Supplementary Material S1 ^{(586.9KB, pdf)} , available online only.

Study selection phases

Article exclusion was performed by two independent reviewers in three phases.

Phase 1

Articles were assessed by title and abstract. Studies were excluded if they used a methodology not reported in the inclusion criteria. Studies were also excluded if they were considered irrelevant to the main theme (i.e., studies on terrorism, political violence, substance abuse, survivors of violence, suicide, genocide, and historical perspectives).

Phase 2

Full texts were obtained through online databases or via email request to the corresponding author and were subsequently read in full by the researchers. Articles that investigated types of interpersonal violence associated with any delinquency outcomes, or assessed attitudes toward violence and the tolerance of violence and/or crime were excluded. Furthermore, articles that assessed R/S combined with other independent variables, such as social support and happiness, were also excluded.

Phase 3

Some articles were excluded due to insufficient statistical data. We contacted the author via email if an article provided insufficient information to allow for inclusion in the meta-analysis. If we received no response after 10 emails, or if they still provided inadequate information, their studies were excluded from further analyses. Additionally, studies assessing the same outcomes and samples in different publications were excluded, including those on homicide and violent acts perpetrated in counties, cities, and/or countries where the researchers used population stratification.

Data collection process

The data were extracted by one researcher (JG), and included articles from Phase 1 were cross-coded by a second independent researcher (PL). Those included in Phase 3 were cross-coded by a different researcher (EM). Discrepancies were resolved by consensus.

Data items

We extracted the following data from the selected articles: authors, year of publication, study design, representativeness of the population, sample size, type of population, sex, age group of participants, and country in which the study was conducted.

Violence was classified into similar types of violent acts: physical aggression (fighting, attacking, assaulting), domestic violence (harming family members, such as children and spouse/partner), and sexual aggression (rape, forced sex). We then described the assessed outcome. R/S type was divided into organizational (i.e., religious affiliation, worship service attendance), non-organizational (i.e., private activities and behaviors such as prayer and reading, listening to, or watching religious content), intrinsic (i.e., commitment, any variable that included importance of religion, regardless of the other items assessed), and spirituality (i.e., spiritual well-being, spiritual intelligence). We then described the assessed outcome for each R/S type. Finally, we defined the results of each outcome as a protective or risk factor when the articles showed a significant or non-significant association with interpersonal violence, respectively.

Risk of bias in individual studies

Since there is no gold standard for quality assessment of observational studies,37 we used a critical appraisal tool38 to assess the risk of bias (Supplementary Material S2 ^{(586.9KB, pdf)} , available online only). The tool consists of 14 key components of epidemiological or observational studies used by the National Institutes of Health for cohort studies. However, because four items (6, 7, 10, and 13) did not apply to cross-sectional studies, a total of 10 items were used to assess the quality of this specific type of methodological design.

The instrument allows five possible responses for each item: yes, no, cannot determine, not applicable, and not reported. To rate the quality score, we attributed one point for each yes response. We then summed the points of each study and calculated an average. This value served as a cut-off point. Cross-sectional and cohort designs were calculated separately.

Studies scoring above the cut-off were considered to have sufficient methodological quality. The cut-off was determined using the mean of all studies included in this systematic review. To analyze the type of R/S measures used for interpersonal violence outcomes, we classified the eighth item of the scale more conservatively: “For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g., categories of exposure, or exposure measured as a continuous variable)?” We only attributed a yes response if the authors used a previously published valid instrument, rather than single items.

Summary measures

The effect size was determined using the unadjusted Pearson correlation coefficient (r) with a 95%CI.

In articles that provided unstandardized beta coefficients, we used them to indicate the effect size. When an article presented the results as an odds ratio (OR), we used a logarithmic formula (ln (OR)/1.81) to convert it to effect size, as validated in a previous study.39 We requested unstandardized coefficients from authors who presented their results in standardized coefficients. Those who did not respond to our email, did not provide sufficient information, or could not be contacted were excluded from the final meta-analysis. Similarly, articles that only described the association between violence and religiosity using descriptive analyses were excluded.

ProMeta 3.0 (Internovi, Cesena FC, Italy) was used to convert the OR and Cohen’s d into r.

Meta-analysis: synthesis of results and risk of bias across studies

OpenMeta software was used to perform the meta-analysis.40 Due to the high heterogeneity (I ²), the random effect statistic was selected, and sensitivity analysis consisted of stratifying the studies in different subgroup analyses.41 We aimed to determine whether the magnitude of the results was influenced by: 1) the interpersonal violence outcome (single item/combined items), 2) religiosity (organizational/non-organizational/intrinsic), 3) age (< 19/> 19 years), 4) the methodology (cross-sectional/longitudinal), 5) the representativeness of the sample (yes/no), and 6) study quality (lower/higher score).

Additionally, a random-effects meta-regression was performed to explore potential differences in the subgroup analyses (Q statistics). By nominating a reference subgroup, the p-value can indicate whether there is a statistically significant difference among the groups.41 Meta-regression coefficients and 95%CI were reported, and p-values < 0.05 were considered significant.

Results

Study selection

Figure 1 is a flow diagram of the article selection process. The initial search yielded 16,599 articles. In Phase 1, we excluded 16,392 articles, of which 3,984 were duplicates, 11,825 did not meet the inclusion criteria, and 583 had heterogeneous study designs. The 207 articles included in Phase 2 were then read in detail, after which 140 were excluded for not assessing interpersonal violence as a separate outcome from other delinquency and crime variables (122), assessed R/S combined with other independent variables such as social support (10), or assessed the occurrence of violence in countries and cities, rather than among individuals (8).

Figure 1. Flow diagram of the article selection process. ASSIA = Sociological Abstract, Applied Social Sciences abstracts; NCJRS = National Criminal Justice Reference Service.

Of the 67 articles included in Phase 3, the data of 18 were insufficient for inclusion in the meta-analysis. When we attempted to contact these authors, eight could not be reached, six no longer had access to the data, and four did not respond with the information requested. Another six studies were excluded due to reporting only descriptive statistics, stratifying the results by groups (i.e., high vs low religiosity groups), or for sharing the same sample and outcome. Ultimately, 43 studies were included in the final meta-analysis.

Study characteristics and results of individual studies

Table 1 presents the characteristics of the 67 included articles. The publication dates varied between 1985 and 2020, and 56.7% were published in the last decade (2011 to 2020). The studies were from the following regions: North America (76.1%), Asia (10.4%), South and Central America (5.9%), Europe (5.9%), Oceania (2.9%), and the Middle East (1.4%).

Table 1. Characteristics of studies evaluating the association between violence outcomes and religiosity/spirituality.

Author	Type of study	Representative	Sample	Population	Sex	Group age (years)	Country	Violence group	Type of violence	Violence outcome	Type of religiosity/spirituality	Religious/ spiritual outcome	Results†
Abdalla42	Cross-sectional	Yes	4,607	General	Both	15-64	Brazil	Physical aggression	Fighting	Items combined	Organizational	Single item	Protective
Adamczyk43	Longitudinal	Yes	2,199	Children and adolescents	Both	7-12th grades	United States	Physical aggression	Fighting, threats	Items combined	Non-organizational	Items combined	Protective
Altschul44	Longitudinal	No	845	Mothers	Female	23-26	United States	Domestic violence	Assaulting children	CTSPC	Organizational	Single item	Protective (W1)Protective (W2)
Banyard45	Cross-sectional	Yes	2,225	Children and adolescents	Both	7-10th grade	United States	Sexual aggression	Sexual coercion, unwanted sex, rape	Items combined	Spirituality	Single item	Protective
Benda46	Cross-sectional	No	1,093	Adolescents	Both	13-20	United States	Physical aggression	Attacking, fighting, threats	Items combined	Non-organizational	Items combined	NS
Benda47	Cross-sectional	No	600	Convicted in boot camp	Male	15-24	United States	Physical aggression	Assaults, sexual assaults	Items combined	Intrinsic	Items combined	NS
Benda21	Cross-sectional	Yes	3,335	Adolescents	Both	13-18	United States	Physical aggression	Attacking, fighting, rape	Items combined	Spirituality	SWBS	Protective
Bernat19	Longitudinal	Yes	2,263	Adolescents	Both	14-20	United States	Physical aggression	Attacking, fighting	Items combined	Organizational	Single item	NS (W1) NS (W2)
Brinkerhoff22	Cross-sectional	Yes	1,834	General	Both	> 18	Canada	Domestic violence	Attacking, fighting, threats	CTS	Organizational	Single item	NS NS
Clubb48	Cross-sectional	No	6,400	Children and adolescents	Both	9-19	United States	Physical aggression	Attacking, fighting, threats	Items combined	Intrinsic	Single item	NS (attendance)Protective (salience)
Corwyn49	Cross-sectional	No	600	Adolescents	Both	13-18	United States	Physical aggression	Attacking, fighting, threats, sexual coercion	Items combined	Non-organizational	Items combined	Protective
Cretacci50	Cross-sectional	Yes	6,500	Children and adolescents	Both	10-19	United States	Physical aggression	Fighting, threats	Items combined	Intrinsic	Items combined	Protective (commitment) NS (beliefs)
Cunradi51	Cross-sectional	Yes	1,440	General	Both	> 18	United States	Domestic violence	Attacking, fighting, threats	CTS	Non-organizational intrinsic	Single item	NS (non-organizational for male and female) NS (salience for males)Risk (salience for females)
Desmond52	Longitudinal	Yes	1,725	Children and adolescents	Both	11-17	United States	Physical aggression	Hitting	Single item	Intrinsic	Items combined	NS
Desmond53	Longitudinal	Yes	1,725	Children and adolescents	Both	11-17	United States	Physical aggression	Hitting	Single item	Intrinsic	Items combined	NS
Dick54	Cross-sectional	No	20,353	Adolescents	Both	10th	United States	Physical aggression	Attacking	NR	NR	NR	Protective
Dyslin55	Cross-sectional	No	436	General	Both	17-47	United States	Domestic violence	Assaulting parents	Child physical abuse risk	Organizational intrinsic	RLI	Risk (organizational)NS (intrinsic)
Ellison56	Cross-sectional	Yes	13,017	General	Both	> 18	United States	Physical aggression	Hitting	Single item	Organizational	Single item	Protective (once per week or more) NS (times per month or year)
Eshuys57	Cross-sectional	No	111	Inmates	Male	38.5 (14.2)	United States	Physical aggression sexual aggression	Number of victims of sexual and non-sexual offenses	Single item	Non-organizational intrinsic	Items combined	Protective (“stayers” vs atheists) NS (other religious groups)
Gonçalves58	Cross-sectional	Yes	4,607	General	Both	14-99	Brazil	Physical aggression domestic violence	Attacking, fighting, threats	Items combined	Intrinsic	Single item	Protective (affiliation)Protective (salience)
Hagen59	Longitudinal	No	795	Adolescents and young adults	Male	NR	United States	Sexual aggression	Sexual coercion, unwanted sex, rape, threats	SES	Non-organizational	Items combined	Protective (W4)
Hardy60	Cross-sectional	No	502	Children and adolescents	Both	10-18	United States	Physical aggression	Attacking, fighting, threats	RPQ	Non-organizational	Items combined	Protective (behavior)Protective (activities)
Hemphill61	Longitudinal	No	927	Children and adolescents	Both	10-11	United States and Australia	Physical aggression	Attacking, fighting	Items combined	Organizational	Single item	NS
Holmes62	Cross-sectional	Yes	110	Children and adolescents	Male	7-12	United States	Physical aggression	Fighting, threats	Items combined	Organizationalnon-organizational intrinsic	DUREL	NS (organizational and non-organizational)Protective (intrinsic)
Itani63	Cross-sectional	Yes	448	Adolescents and young adults	Both	NR	Lebanon	Physical aggression	Fighting, threats	Items combined	Intrinsic	CRS	Protective
Johnson64	Longitudinal	Yes	4,834	Children and adolescents	Both	11-21	United States	Physical aggression	Fighting, threats, hitting	Items combined	Intrinsic	Items combined	Protective (W1)Protective (W2)
Karimi65	Cross-sectional	No	50	Adolescents	Male	14-17	Iran	Physical aggression	Fighting, threats, hitting	BPAQ	Spirituality	SIS	Protective
Karriker-Jafe66	Cross-sectional	Yes	5,118	Children and adolescents	Both	11-16	United States	Physical aggression	Attacking, fighting, threats	Items combined	Intrinsic	Items combined	Protective
Katerndahl67	Cross-sectional	No	105	Couples	Both	45.7 (14.7) male 42.9 (14.0) female	United States	Domestic violence	Attacking, fighting, threats	The Conflict Tactics Scale	Non-organizational intrinsic	BMMRS	Risk (religious beliefs incompatibility)
Kingre68	Longitudinal	No	544	Adolescents and young adults	Both	18-20	United States	Sexual aggression	Sexual coercion, unwanted sex, rape	SES	Organizational	Single item	Protective
Leach69	Cross-sectional	No	62	Adolescents and young adults	Both	20.43 (3.11)	United States	Physical aggression	Fighting, threats, hitting	BPAQ	Organizational, intrinsic spirituality	ROS -Revised STS	NS (organizational)Protective (intrinsic and spiritual)
Linville70	Cross-sectional	No	235	Adolescents	Both	8-12th grades	United States	Physical aggression	Fighting	Single item	Organizational	Single item	NS
Lynch71	Cross-sectional	No	255	General	Male	>18	United States	Domestic violence	Attacking, fighting, threats	SVAWS	Intrinsic	Religious SRQ	Protective (identified religious self-regulation) Risk (introjected religious self-regulation)
Massarwi72	Cross-sectional	No	2,811	Children and adolescents	Both	13-18	Israel	Physical aggression	Attacking, fighting	Items combined	Intrinsic	Items combined	Protective
Michaelson73	Cross-sectional	Yes	24,307	Children and adolescents	Both	6-10th grades	Canada	Physical aggression	Fighting, threats	Items combined	Non-organizational	Single item	NS (W1) NS (W2)
Murshid74	Cross-sectional	Yes	3,186	General	Male	16-54	Bangladesh	Domestic violence	Attacking, fighting, threats	M-CTS	Organizational	Single item	NS
Padilla-Walker75	Cross-sectional	No	1,629	Adolescents	Both	16.08 (1.09)	United States	Physical aggression	Fighting, threats	Items combined	Non-organizational	Items combined	Protective
Park76	Longitudinal	Yes	2,895	Adolescents	Both	12-13	United States	Physical aggression	Attacking	Single item	Organizational	Single item	Protective
Peek77	Longitudinal	Yes	1,545	Adolescents	Male	Sophomore, junior, senior years	United States	Physical aggression	Assaulting parents	Single item	Intrinsic	Items combined	Protective
Peek78	Longitudinal	Yes	817	Adolescents	Male	Sophomore, junior, senior years	United States	Physical aggression	Attacking	Items combined	Organizational	Items combined	Protective (W1)Protective (W2)
Pickering79	Cross-sectional	No	865	Adolescents	Both	16.4 (1.2)	United States	Physical aggression	Attacking	Items combined	Non-organizational intrinsic	Items combined	Protective (Bible reading) NS (attendance, salience, relational practice)
Pitel80	Cross-sectional	Yes	1,784	Children and adolescents	Both	15.48 (0.45)	Slovakia	Physical aggression	Fighting	Single item	Intrinsic	Single item	NS NS
Pournaghash81	Cross-sectional	No	180	Couples	Both	25-45	Iran	Domestic violence	Attacking, fighting, threats	ADV	Non-organizational, intrinsic	Islamic Religious Tendency Scale, items combined	Protective (beliefs, practices, activities)Risk (religious disorganization)
Powell82	Cross-sectional	No	521	Children and adolescents	Both	5-11th grades	United States	Physical aggression	Fighting	Single item	Intrinsic	Items combined	Protective
Purwono83	Cross-sectional	No	238	Adolescents	Both	15-18	Indonesia	Physical aggression	Attacking, fighting	Items combined	Intrinsic	Items combined	Protective
René84	Cross-sectional	No	1,271	General	Both	> 18	United States	Physical aggression	Attacking, fighting	Items combined	Organizational non-organizational	Single item	NS (self-perception and intentionally)Protective (attendance)
Resnick85	Longitudinal	Yes	13,110	Children and adolescents	Both	7-12th grades	United States	Physical aggression	Attacking, fighting, threats	Items combined	Non-organizational	NR	Protective (female)NR (male)
Sadeghifard86	Cross-sectional	No	1,345	Adolescents and young adults	Both	18-28	Iran	Physical aggression	Attacking, fighting, threats	BPAQ	Spirituality	SAI	NS
Salas-Wright87	Cross-sectional	No	290	Children, adolescents and young adults	Both	11-25	El Salvador	Physical aggression	Attacking, fighting, homicide	NR	Intrinsic spirituality	CRC, ISS	Protective (religiosity)Protective (spirituality)
Salas-Wright20	Cross-sectional	Yes	90,047	Children and adolescents	Both	12-17	United States	Physical aggression	Attacking, fighting	Single item	Intrinsic	Items combined	Protective (attendance)Protective (beliefs)Protective (participation in religious groups)
Salas-Wright88	Cross-sectional	No	138	Adolescents and young adults	Female	13-24	United States	Physical aggression	Fighting	Single item	Non-organizational intrinsic	SCSORF	Protective (severe attacks) NS (fights)
Salas-Wright89	Cross-sectional	No	236	Children, adolescents and young adults	Both	12-25	United States	Physical aggression	Attacking, fighting	NR	Non-organizational intrinsic	SCSORF	NS
Schuster90	Longitudinal	Yes	1,593	Adolescents and young adults	Both	18-30	Chile and Turkey	Sexual aggression	Sexual coercion, unwanted sex, rape	SAV-S	Intrinsic	Single item	Protective (Chileans W1) NS (Chilenas W2 and Turkish W1/W2)
Sealock91	Longitudinal	No	298	Adolescents	Both	13-17	United States	Physical aggression	Attacking, fighting, threats	Items combined	Spirituality	Items combined	NS (W1) Protective (W2)
Shepperd92	Longitudinal	No	1,162	Adolescents	Both	15-19	United States	Physical aggression	Attacking, fighting, threats	RCRQ	Intrinsic	RCI	Protective (W2)Protective (W3)
Sinha93	Cross-sectional	Yes	2,004	Children and adolescents	Both	11-18	United States	Physical aggression	Attacking, fighting	Single item	Intrinsic	Single item	NS (attendance and participation)protective (salience)
Sloane94	Cross-sectional	Yes	1,121	Adolescents	Both	13-18	United States	Physical aggression	Fighting	Single item	Non-organizational intrinsic	Single item	NS (non-organizational)protective (intrinsic)
Solinas-Saunders95	Cross-sectional	Yes	14,499	Inmates	Both	16-84	United States	Physical aggression	Attacking, fighting	Single item	Non-organizational	Single item	NS
Stevens96	Cross-sectional	No	310	Adolescents	Both	13-19	Polynesia	Physical aggression	Attacking, fighting, threats	Proactive-Reactive Aggression Questionnaire	Non-organizational	RCI	Protective
Todhunter97	Cross-sectional	Yes	1,507	Adolescents and young adults	Male	18-26	United States	Physical aggression	Attacking, fighting, threats, sexual coercion	NR	Non-organizational	Items combined	NS
Tomaszewska98	Longitudinal	No	318	Adolescents and young adults	Both	19.7 (1.03)	Poland	Sexual aggression	Sexual coercion, unwanted sex, rape	SAV-S	Intrinsic	Items combined	NS (W1) NS (W2)
Tyler99	Cross-sectional	No	172	Adolescents and young adults	Both	19-26	United States	Physical aggression	Attacking, fighting, threats	Items combined	Intrinsic	Items combined	NS
Tzamalouka100	Cross-sectional	No	1,122	General	Both	18-65	Greece	Physical aggressionsexual aggression	Attacking, fighting, threats, rape, forced sex	Items combined	Non-organizational	Items combined	Protective (physical aggression) Protective (sexual aggression)
Velazquez101	Cross-sectional	No	345	Children and adolescents	Both	1-6th grades	Mexico	Physical aggression	Attacking, fighting, threats	CAS	Organizational	Escala del Ambiente Social Familiar	Protective
Weber102	Cross-sectional	No	457	Adolescents and young adults	Both	18-23	United States	Physical aggression	Attacking, fighting, threats	RAS	Intrinsic	FMS	NS
Wolf103	Cross-sectional	Yes	3,023	Children	Both	< 12	United States	Physical aggression	Attacking, hitting, threats	PCTSPC	Organizational	Single item	Risk
Yun104	Cross-sectional	No	4,864	General	Both	15.15 (1.61)	South Korea	Physical aggression	Attacking, fighting, threats	Items combined	Intrinsic	Items combined	NS

ADV = Domestic Violence Questionnaire; BMMRS = Brief Multidimensional Measure of Religiousness/Spirituality; BPAQ = Buss-Perry Aggression Questionnaire; BPAQ = Buss-Perry Aggression Questionnaire; CAS = Children’s Aggression Scale; CRC = Religious Coping Scale; CTS = Conflict Tactics Scale; CTSPC = Parent-Child Conflict Tactics Scale; CTSPC = Parent-Child Conflict Tactics Scales; DUREL = Duke University Religion Index; FMS = Faith Maturity Scale; ISS = Intrinsic Spirituality Scale; M-CTS = Modified Conflict Tactics Scale; NR = not reported; NS = non-significant; RAS = Relational Aggression Scale; RCI = Religious Commitment Inventory; RCRQ = Richardson Conflict Response Questionnaire; RLI = Religious Life Inventory; ROS = Religious Orientation Scale; RPQ = Reactive-Proactive Aggression Questionnaire; SAI = Spiritual Assessment Inventory; SAV-S = Sexual Aggression and Victimization Scale; SCSORF = Santa Clara Strength of Religious Faith Questionnaire; SES = Sexual Experiences Survey; SIS = Spiritual Intelligence Scale; SRQ = Self-Regulation Questionnaire; STS = Spiritual Transcendence Scale; SVAWS = Severity of Violence Against Women Scale; SWBS = Spiritual Well-Being Scale; W1, W2, W3 = Wave 1, 2, and 3.

^†Protective and risk results were statistically significant outcomes reported in the articles.

Regarding study design, 50 (73.2%) studies were cross-sectional and 17 (26.8%) were longitudinal. A total of 44.8% of the studies evaluated a probability representative sample. The total sample consisted of 269,910 individuals. Regarding outcomes, physical aggression was the most frequently assessed type (83.6% of the articles), followed by domestic violence and sexual aggression (10.4% each). The most frequently investigated R/S type was intrinsic (43.75%), followed by non-organizational (26.25%), organizational (21.25%), and spirituality (8.75%).

A total of 101 outcomes were assessed in the studies: R/S had a significant protective role in 55.4% and the results were non-significant in 38.6%. Six studies found that religious individuals had a significant risk of perpetrating violent acts (5.9% of the sample), of which two analyzed the negative outcomes of religiosity (introjected religious self-regulation and disorganized religiosity). Five of these studies assessed domestic violence, and one examined physical aggression.

Risk of study bias

The risk of study bias is presented in Table 2. The mean quality assessment score for cross-sectional studies was 7.42 (SD, 1.29), with 88% exceeding the cutoff point. The mean score for cohort studies was 11 (SD = 1.28), with only 65% exceeding the cutoff. There was at least one unreported response in 80% of the items in cross-sectional studies, while this occurred in only 28.6% of the cohort studies.

Table 2. Study quality assessment of all included articles using the NIH quality assessment tool for observational cohort and cross-sectional studies.

Author	1†	2	3	4	5	6	7	8	9	10	11	12	13	14	Score
Cross-sectional design
Abadlla42	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Banyard45	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	Yes	NA	Yes	9
Benda46	Yes	Yes	Yes	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	NR	NA	NR	8
Benda47	Yes	Yes	NR	Yes	NR	NA	NA	Yes	Yes	NA	Yes	No	NA	Yes	7
Benda21	Yes	Yes	Yes	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	10
Brinkerhoff22	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	Yes	NA	Yes	8
Clubb48	Yes	Yes	NR	No	NR	NA	NA	No	Yes	NA	Yes	No	NA	Yes	5
Corwyn49	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	No	NA	Yes	7
Cretacci50	Yes	Yes	Yes	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	10
Cunradi51	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Dick54	Yes	Yes	NR	Yes	NR	NA	NA	No	No	NA	No	NR	NA	No	3
Dyslin55	Yes	Yes	Yes	Yes	NR	NA	NA	Yes	Yes	NA	Yes	NR	NA	No	7
Ellison56	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	NR	NA	Yes	8
Eshuys57	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	NR	NA	No	6
Gonçalves58	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Hardy60	Yes	Yes	NR	Yes	NR	NA	NA	No	Yes	NA	Yes	Yes	NA	No	6
Holmes62	Yes	Yes	NR	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	NR	NA	Yes	8
Itani63	Yes	Yes	Yes	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	10
Karimi65	Yes	Yes	NR	Yes	NR	NA	NA	Yes	Yes	NA	Yes	NR	NA	No	6
Karriker-Jafe66	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	NR	NA	Yes	8
Katerndahl67	Yes	Yes	NR	Yes	NR	NA	NA	Yes	Yes	NA	Yes	No	NA	Yes	7
Leach69	Yes	Yes	NR	NR	Yes	NA	NA	Yes	Yes	NA	Yes	No	NA	No	6
Linville70	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	Yes	NA	Yes	8
Lynch71	Yes	Yes	NR	Yes	NR	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	8
Massarwi72	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Michaelson73	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	NR	NA	Yes	8
Murshid74	Yes	Yes	Yes	Yes	Yes	NA	NA	No	No	NA	Yes	NR	NA	Yes	7
Padilla-Walker75	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	NR	NA	Yes	7
Pickering79	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	NR	NA	Yes	7
Pitel80	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	Yes	NA	Yes	9
Pournaghash81	Yes	Yes	NR	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	9
Powell82	Yes	Yes	No	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	7
Purwono83	Yes	Yes	Yes	NR	NR	NA	NA	Yes	Yes	NA	Yes	NR	NA	Yes	7
René84	Yes	Yes	NR	Yes	NR	NA	NA	No	Yes	NA	Yes	No	NA	Yes	6
Sadeghifard86	Yes	Yes	NR	NR	Yes	NA	NA	Yes	Yes	NA	Yes	NR	NA	Yes	7
Salas-Wright87	Yes	Yes	NR	Yes	Yes	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	9
Salas-Wright20	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	NR	NA	Yes	8
Salas-Wright88	Yes	Yes	NR	NR	NR	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	7
Salas-Wright89	Yes	Yes	NR	NR	NR	NA	NA	Yes	Yes	NA	Yes	NR	NA	Yes	6
Sinha93	Yes	Yes	Yes	Yes	NR	NA	NA	No	Yes	NA	Yes	No	NA	Yes	7
Sloane94	Yes	Yes	NR	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	NR	6
Solinas-Saunders95	Yes	Yes	No	Yes	Yes	NA	NA	No	Yes	NA	Yes	Yes	NA	Yes	8
Stevens96	Yes	Yes	NR	Yes	NR	NA	NA	Yes	Yes	NA	Yes	Yes	NA	Yes	8
Todhunter97	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Tyler99	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Tzamalouka100	Yes	Yes	Yes	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	8
Velazquez101	Yes	Yes	NR	Yes	NR	NA	NA	No	Yes	NA	Yes	No	NA	Yes	6
Weber102	Yes	Yes	NR	NR	NR	NA	NA	Yes	Yes	NA	Yes	NR	NA	Yes	6
Wolf103	Yes	Yes	No	Yes	Yes	NA	NA	No	Yes	NA	Yes	No	NA	Yes	7
Yun104	Yes	Yes	Yes	NR	Yes	NA	NA	No	Yes	NA	Yes	Yes	NA	Yes	8
Longitudinal design
Adamczyk43	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	11
Altschul44	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	11
Bernat19	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	Yes	Yes	Yes	Yes	12
Desmond52	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	NR	Yes	Yes	12
Desmond53	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	NR	Yes	Yes	12
Hagen59	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	NR	Yes	No	11
Hemphill61	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	Yes	Yes	Yes	Yes	Yes	12
Johnson64	Yes	Yes	NR	Yes	NR	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	9
Kingre68	Yes	Yes	Yes	Yes	NR	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	12
Park76	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	11
Peek77	Yes	Yes	NR	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	NR	No	Yes	10
Peek78	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	NR	No	Yes	11
Resnick85	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	Yes	No	Yes	Yes	10
Schuster90	Yes	Yes	Yes	Yes	NR	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	10
Sealock91	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	Yes	No	No	Yes	9
Shepperd92	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	14
Tomaszewska98	Yes	Yes	Yes	Yes	NR	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	10

NA = not applicable; NR = not reported.

^†Items assessed according to National Institutes of Health (NIH) scale (available in Supplementary Material S2 ^{(586.9KB, pdf)} ).

Population recruitment and the inclusion and exclusion criteria (item 4) were similar between design types, with 14% non-reported in the cross-sectional studies and 0% in the cohort studies. Regarding the assessment of exposure levels (item 8), 20 studies (30%) used validated religious/spiritual scales. Validated instruments were used in 38% of the cross-sectional studies but in only 6% of the cohort studies. There was a high score for item 9, which assessed the clarity of the definitions and the reliability of the exposure variables: 48 (96%) for cross-sectional studies and 14 (82.4%) for cohort studies. Outcome assessor blinding was reported in 30% of the cross-sectional studies and in 23.5% of the cohort studies.

Two specific questions for cohort methodology determined whether the exposure of interest was assessed before the outcome (item 6) and whether there was a sufficient timeframe between waves (item 7). Both items were reported by all authors. In 47.1% of the articles, R/S variables were assessed several times during the study period, and only 17.6% of the studies reported dropout rates > 20%.

Synthesis of the results and risk of bias across studies

Of the 67 included studies, 24 were excluded from the meta-analysis. We contacted the authors of 18 of these studies for additional database information. Eight of these authors could not be reached, six no longer had access to the data, and four responded without providing the necessary information. We excluded three articles that analyzed separate age or religious groups and did not present the results for the total sample, in addition to two others that only provided descriptive analyses. The same religious and interpersonal violence outcomes were assessed using the same sample in two different publications, so we excluded one.

The remaining 43 studies were divided into three groups according to violence outcomes: physical aggression, domestic violence, and sexual aggression. Since some articles assessed more than one violence outcome, including more than one type of religious/spiritual variable, the data were overlapped in the analysis, which resulted in more comparison groups than studies for each outcome.

Sex was not included in the subgroup analyses because the results of most articles were presented as mixed groups of men and women, making it impossible to stratify the samples. Moreover, the subgroup analyses could not be performed by country, since 71.6% of the studies were conducted in the United States.

Physical aggression

The physical aggression analyses included 33 studies and 80 comparisons, totaling 1,221,897 individuals (Figure S1 ^{(586.9KB, pdf)} , available as online-only supplementary material). Higher levels of R/S were significantly associated with lower physical aggression (r = -0.12, 95%CI = -0.137 to -0.095). Due to the high heterogeneity (I ² = 99.16%, p < 0.001), subgroup analyses were performed (Table 3).

Table 3. Subgroup analyses and meta-regression data for the outcomes: physical aggression, domestic violence, and sexual aggression.

	Studies (n)	Comparisons (n)	Sample size for the comparisons (n)	Correlation r	95%CI	p-value	I 2 (%)	Q	Tau-square	Coefficient	95%CI	p-value
Physical aggression
All studies	33	80	1,221,897	-0.116	-0.137 to -0.095	0.001	99.16	9442.72	0.008	-
Violence outcome
Single item	8	21	1,102,332	-0.097	-0.119 to -0.075	0.001	99.11	2242.31	0.002	0.029	-0.041 to 0.099	0.413
Combined items	25	59	119,565	-0.122	-0.181 to -0.063	0.001	99.16	6870.61	0.052	Ref
Religiosity†
Organizational	11	17	289,987	-0.150	-0.205 to -0.094	0.001	99.42	2782.31	0.012	-0.017	-0.150 to 0.008	0.079
Non-organizational	14	35	604,719	-0.070	-0.087 to -0.053	0.001	96.51	973.33	0.002	Ref
Intrinsic	16	27	326,652	-0.145	-0.193 to -0.098	0.001	99.33	3743.22	0.014	-0.068	-0.136 to 0.000	0.052
Age, years
< 19	19	52	1,161,342	-0.097	-0.113 to -0.081	0.001	98.34	3075.34	0.003	Ref
Included > 19	14	28	60,555	-0.138	-0.239 to -0.038	0.007	99.44	4872.02	0.072	-0.042	-0.107 to 0.022	0.198
Methodology
Cross-sectional	25	65	1,195,767	-0.119	-0.143 to -0.096	0.001	99.31	9320.84	0.008	-0.019	-0.099 to 0.061	0.645
Longitudinal	8	14	26,13	-0.099	-0.135 to -0.063	0.001	87.90	107.47	0.004	Ref
Representative
Yes	15	37	6,172,91	-0.116	-0.145 to -0.086	0.001	99.59	8821.27	0.008	Ref
No	18	43	52,277	-0.117	-0.150 to -0.083	0.001	92.94	595.23	0.011	-0.001	-0.063 to 0.061	0.976
Study quality
Lower score	5	10	15,228	-0.079	-0.112 to -0.045	0.001	68.69	28.74	0.002	0.046	-0.048 to 0.139	0.337
Higher score	28	70	1,206,669	-0.122	-0.144 to -0.099	0.001	99.27	9405.60	0.008	Ref
Year of publication
≤ 2009	9	16	31,592	-0.109	-0.161 to -0.057	< 0.001	94.93	296.06	0.010	Ref
≥ 2010	24	63	1,190,305	-0.118	-0.142 to -0.095	< 0.001	99.31	9134.74	0.008	-0.009	-0.086 to 0.069	0.827
Domestic violence/intimate partner violence
All studies	8	23	23,137	-0.050	-0.200 to 0.099	0.511	99.70	7362.11	0.131	-
Violence outcome
Single item	0	-	-	-	-	-	-
Combined items	8	23	23,137	-0.050	-0.200 to 0.099	0.511	99.70	7362.11	0.131	-
Religiosity†
Organizational	6	10	13,613	-0.038	-0.313 to 0.236	0.784	99.67	2752.09	0.195	Ref
Non-organizational	3	7	2,861	-0.057	-0.317 to 0.203	0.669	99.65	1726.29	0.120	-0.016	-0.247 to 0.216	0.893
Intrinsic	4	6	6,663	-0.040	-0.230 to 0.150	0.681	99.44	1265.60	0.074	-0.026	-0.267 to 0.215	0.833
Age
Children	0	-	-	-	-	-	-			-
Adolescents	2	3	3,319	-0.113	-0.189 to -0.038	0.003	78.99	9.52	0.004	Ref
Adults	6	20	19,818	-0.041	-0.207 to 0.125	0.627	99.74	7352.41	0.141	0.071	-0.215 to 0.357	0.627
Methodology
Cross-sectional	7	21	21,447	-0.048	-0.207 to 0.111	0.556	99.72	7358.37	0.136	-
Longitudinal	1	-	-	-	-	-	-
Representativeness
Yes	4	9	17,098	-0.032	-0.298 to 0.235	0.817	99.74	3158.43	0.166	0.027	-0.172 to 0.226	0.789
No	4	14	6,039	-0.062	-0.230 to 0.107	0.475	99.55	2907.74	0.101	Ref
Study quality
Lower score	0	-	-	-	-	-	-
Higher score	8	23	23,137	-0.050	-0.200 to 0.099	0.511	99.70	7362.11	0.131	-
Year of publication
≤ 2009	3	11	6,878	0.060	-0.062 to 0.182	0.334	96.47	283.59	0.041	Ref
≥ 2010	5	12	16,259	-0.152	-0.368 to 0.064	0.168	99.83	6789.46	-0.143	-0.210	-0.387 to -0.033	0.020
Sexual aggression
All studies	4	8	6,025	-0.049	-0.077 to -0.021	0.001	13.55	8.09	0.000	-
Violence outcome
Single item	0	-	-	-	-	-	-
Combined items	4	8	6,025	-0.049	-0.077 to -0.021	0.001	13.55	8.09	0.000	-
Religiosity
Organizational	0	-	-	-	-	-	-
Non-organizational	1	-	-	-	-	-	-
Intrinsic/Spirituality	3	7	5,516	-0.044	-0.075 to -0.014	0.004	16.65	7.19	0.000	-
Age†
Children	1	-	-	-	-	-	-
Adolescents	3	6	3,482	-0.044	-0.075 to -0.014	0.004	16.65	7.19	0.000	-
Young adults	1	-	-	-	-	-	-
Methodology
Cross-sectional	1	-	-	-	-	-	-
Longitudinal	3	6	3,482	-0.044	-0.075 to -0.014	0.004	16.65	7.19	0.000	-
Representative
Yes	2	5	3,164	-0.037	-0.088 to 0.013	0.150	48.24	7.72	0.002	0.008	-0.042 to 0.059	0.743
No	2	3	1,145	-0.079	-0.137 to -0.021	0.007	0.00	0.17	0.000	Ref
Study quality
Lower score	2	6	3,291	-0.039	-0.081 to 0.004	0.075	30.04	7.14	0.001	0.014	-0.037 to 0.065	0.594
Higher score	2	2	2,734	-0.057	-0.095 to -0.020	0.003	0.00	0.66	0.000	Ref

Bold type denotes significant statistical difference.

Ref = reference category.

^†These analyses include overlap from studies that assessed more than one religiosity criterion.

All investigated subgroups showed statistically significant results with small effect sizes. However, the heterogeneity did not decrease in any of these analyses. Organizational and intrinsic religiosity had similar effect sizes (r = -0.15, 95%CI = -0.20 to -0.09; r = -0.14, 95%CI = -0.19 to -0.10, respectively), and non-organizational religiosity showed a lower effect size than the other two types (r = -0.07, 95%CI = -0.09 to -0.0.5). However, religiosity outcomes for the meta-regression analyses were not significant.

Domestic violence

The domestic violence subanalysis included eight studies and 23 comparisons, resulting in an overall sample of 23,137 individuals. Although less intimate partner violence was not associated with higher levels of R/S (r = -0.05, 95%CI = -0.200 to 0.099) (Figure S2 ^{(586.9KB, pdf)} , available as online-only supplementary material), there was significant heterogeneity among the studies (I ² = 99.70%, p < 0.001). Subgroup analyses, however, revealed a significant association among adolescents (r = -0.11, 95%CI = -0.189 to -0.038), with a heterogeneity of 78.99% (p < 0.005). Although no significant results were found for articles published until 2009 (r = 0.060, 95%CI = -0.062 to 0.182, p = 0.334) or after 2009 (r = – 0.152, 95%CI = -0.368 to 0.064, p = 0.168), there was a significant difference between older and newer articles of the meta-regression (p = 0.020).

Sexual aggression

Regarding sexual aggression, we analyzed four studies and carried out eight comparisons, totaling 6,025 individuals. There was a significant negative association between sexual aggression and higher R/S, although the effect size was smaller than that of physical aggression (r = -0.05, 95%CI = -0.077 to -0.021) (Figure S3 ^{(586.9KB, pdf)} , available as online-only supplementary material). Heterogeneity in this outcome was low and non-significant (I ² = 13.55%, p = 0.324). All authors used combined items as their interpersonal violence outcome. Most studies assessed intrinsic/spiritual variables (seven of eight comparisons) and investigated adolescents (six of eight comparisons). No significant difference was found between the studies in the subgroup analysis.

Discussion

The results of this systematic review and meta-analysis support the proposition that R/S plays a significant protective role against physical and sexual aggression. Nevertheless, R/S was only associated with less domestic violence among adolescents.

Previous meta-analyses investigating the involvement of religion in delinquency have found a consistently inverse relationship,17,18 which corroborates our findings. However, these meta-analyses focused on delinquent acts and criminal behavior, rather than exclusively violent acts against others. To our knowledge, this is the first systematic review and meta-analysis to explore the impact of R/S on different aspects of interpersonal violence.

Interestingly, our findings had different effect sizes for different aspects of interpersonal violence. Specifically, it was higher for physical than for sexual aggression and was non-significant for domestic violence outcomes. Previous studies have found that R/S has a larger effect size for victimless crimes (such as tax evasion,105-107 the selling and consumption of illegal substances,18 and robbery and vandalism43,90) than for crimes involving victims.18 According to our findings, it seems that the impact of R/S differs depending on the type of interpersonal violence, which could be explained by the complexity involved in domestic and sexual aggression, including barriers to reporting such crimes.108

Regarding physical aggression, all subgroup analyses (age, study design, representativeness) were significant, consistently showing that R/S plays a protective role against physical aggression. These findings have strong implications for health care professionals and managers. While no differences were found in religious subgroup analysis in the meta-regression, the effect sizes varied for organizational and intrinsic religiosity vs. non-organizational religiosity. The effect of organizational religiosity can be explained by the social control theory, which contends that the notion of divine punishment/reward combined with the social support of a formal religion can prevent believers from committing crimes.25,26 The concept of intrinsic religiosity involves the notion of self-control and the rational choice of healthy behaviors and attitudes27,29 as a result of internal reasoning and self-awareness. However, private non-organizational religiosity seems to have little preventive effect against acts of physical violence. This could be explained by the fact that, even though listening to religious music, reading sacred texts, and praying reduce undesirable symptoms,15 they may be insufficient in some contexts, and thus may not help prevent violence. This is also consistent with sociopsychological and evolutionary theories linking religiosity to prosociality (including variables such as social bonding, social support, and social monitoring).109,110

Although sexual aggression had a lower effect size than physical aggression, the subgroup analyses also indicated that R/S played a consistently protective role. Notably, this violence outcome showed the lowest heterogeneity, suggesting that these findings are related to intrinsic religiosity among adolescents. Since adolescents are at greater risk of sexual aggression,1 more studies have been published involving this specific population.111 Regarding intrinsic religiosity, this finding reinforces the aforementioned theories about self-control and rational choice.27,29

In contrast, the domestic violence meta-analysis showed no association with R/S variables, except among adolescents. This could be attributed to the fact that interpersonal violence is a complex multidimensional concept involving a number of causes.112 Thus, R/S may not prevent domestic violence due to overlapping influence from the cultural background.113 There are some explanations for such findings in the literature. First, some cultures and religions can be permissive or tolerant towards domestic violence114,115 in an effort to minimize the disruption of family units. Previous studies have supported this hypothesis, showing that fear of separation or ostracization may cause women to remain in unhealthy relationships.116,117 Second, in some cases, clergy may advise victims to resign themselves to the situation, rather than report it to the police, thus perpetuating the cycle of violence.118 Third, studies in Eastern cultures have found that both men and women agree that men can beat their partner if she refuses sex or retaliates during a fight.119 Similarly, in Western cultures, approval of corporal punishment for disciplining children is also common among religious conservatives.118

Despite these explanations, it should be pointed out that greater awareness about domestic violence has been achieved in recent decades,119 including the harmful effects of violence on mental health, which may interfere with the relationship between religiosity and domestic violence. This was observed in the meta-regression, since newer articles on this topic showed a trend toward significance for R/S as a protective factor (r = -0.152), unlike older articles (r = 0.060).

In five studies, domestic violence was the only outcome in which religiosity was a risk factor for violence. Three of them found religiosity to be a risk factor when assessing negative variables, such as religious incompatibility,67 disorganized religiosity,81 and introjected religious self-regulation.71 Previous research indicates that negative religious coping is associated with higher levels of depression, anxiety, and alcohol and drug consumption.120,121 Since the studies that investigated physical and sexual aggression did not assess negative religiosity, we cannot conclude that the risk is associated with domestic violence alone. Future studies should conduct a more detailed investigation of the role of negative religious coping and violence.

Notably, in the subanalyses, the results were only significant for all types of violence among adolescents. This is consistent with the current literature, which indicates that R/S plays a protective role against delinquency in this age group.17,18,31,33 A meta-analysis by Baier et al.18 showed that religiosity had a deterrent effect on delinquency among adolescents, which was moderated by the year of data collection, sample size, and the proportion of Whites in the sample. Similarly, Kelly et al.17 found a small-to-moderate average effect size between religiosity and delinquency, with similar results for church attendance. However, even though they explored heterogeneity through moderators, they found no significant difference among funded studies, sample type, and sample location.

Despite this promising evidence, the heterogeneity was significant in our analysis of both physical aggression and domestic violence, even after stratifying by subgroup. A previous meta-analysis17 also found high heterogeneity regarding religiosity and interpersonal violence outcomes, which was attributed to possible interference by different mediators.41 Nevertheless, there was low heterogeneity regarding sexual aggression outcomes in our sample because of three important subgroup similarities: it involved the fewest studies and comparisons, the outcomes were assessed through combined items/validated scales, and most studies investigated intrinsic religiosity and adolescent participants. These facts may have yielded more appropriate results.

The assessment of R/S variables showed a similar trend. Although several valid instruments have been developed to measure various constructs of both violent behavior122,123 and R/S outcomes,13,124 we found that they were infrequently used in the included studies, especially those with longitudinal designs. Hence, the consistent use of reliable and valid instruments is needed to elucidate this relationship, especially considering its clinical implications for public health.31

Concerning the studies’ methodological quality, the mean scores were good for both the cross-sectional and cohort designs. However, grouping separate constructs of R/S within the same variable, such as worship service attendance, salience, and beliefs, can produce invalid results, especially in cross-sectional studies.125 Furthermore, reported outcome assessor blinding was less than 30% in both designs. The authors rarely declared whether the individual performing the assessment was aware of the exposure status of the participants. This methodological parameter must be prepared in advance when designing a study and is easily manageable due to its simplicity.

Clinical implications

The evidence that R/S plays a protective role against interpersonal violence has clinical implications, both for health care professionals and health managers. Several studies have examined whether, why, and how physicians approach religion and spiritual topics with their patients in clinical practice.126-128 While this may significantly influence physical and mental health, physicians seldom address R/S and the beliefs of their patients, except among terminally ill patients.126 The most cited barriers are that this topic falls outside their scope of practice, they lack appropriate training, and that there are time constraints.126-127

Nevertheless, the impact of R/S is present throughout life.13 R/S can impact human health both positively and negatively.129 Therefore, strategies and adequate instruments for approaching R/S in clinical practice safely and reliably have been developed in recent years.128,130,131 Considering the patient’s history of R/S and its impact can provide physicians with helpful and tailored preventive strategies. This can reinforce positive religious coping or transform negative religious perspectives into a more constructive condition. Health managers should thus be aware of these findings and train their staff to address these issues in clinical practice.

For example, a previous qualitative study on women incarcerated for murdering their domestic abusers114 included individuals either raised in a home without religion or in an extremely religious home with rigid and aggressive moral conduct based on a punitive concept of God. It seems that traumatic episodes linked to religious issues are difficult to recover from. Health professionals should address these issues in a patient-centered, individualized, and nonjudgmental approach. The authors proposed an intervention based on spirituality (moral values, faith, and transcendence) to alleviate the convicts’ mental suffering. Despite negative prior religious experiences, participants transitioned from negative to positive religious strategies.

Understanding the patients’ religious/spiritual background can provide insight into how it relates to their present. Evidence shows that parental religiosity impacts the mental and physical health and behavior of adolescents, both positively and negatively.132-134 The family religious environment may thus hinder or encourage child development.

Clinical trials designed to prevent interpersonal violence through R/S have ethical limitations. However, some authors are exploring R/S interventions to reduce violence and misconduct in male prisons.135-137 The results have shown improvement in personal conduct, less fighting, and improved mental health outcomes for those who converted to a religious affiliation. More research is necessary to elucidate the actual long-term impacts on mental health and behavior. Nevertheless, such programs have already been implemented in institutions that can benefit from simple and low-cost interventions.

Future research

Most studies included in this review did not assess R/S as a central explanatory variable. Johnson et al.31 conducted the first systematic review regarding religiosity and delinquency, finding that although most studies examined religiosity as a central variable, they also investigated only one or two other dimensions of religiosity, mainly worship service attendance and the reported importance of religion.

The cohort studies assessing R/S and violence were designed to investigate nationally representative samples, including several other measures and outcomes during an interview assessment. Therefore, R/S was not previously predicted as an outcome that could impact violence: it was simply addressed as another variable. Future studies on R/S and violence should be designed to clarify this relationship using appropriate instruments for both dependent and independent variables.138 Even if the researchers choose to assess single questions, they must avoid summing all points in the same score when analyzing the data.

To explore the mechanisms of action of the preventive function of religiosity, future cohort studies should be specifically designed to address the impact of R/S on violence and clarify possible moderators during follow-up research. Furthermore, clinical trials for individuals who exhibit violent traits can help provide insight into whether R/S interventions can help improve rehabilitation by diminishing violent impulses.139,140 Finally, qualitative studies should also be considered as an avenue for understanding the role of R/S in human nature and how it can help improve behavior.

Limitations

Although 16,599 articles were screened in seven different health science and sociology databases, other relevant studies may have been overlooked. Moreover, we found no studies in languages other than Portuguese, English, and Spanish, but, again, articles in other languages may have been missed.

In addition, although we were able to carry out subgroup analyses, these were limited to age group, sex, measurements of violence, and R/S outcomes. The heterogeneity among studies was relevant, especially regarding the dependent and independent variables.

In conclusion, this meta-analysis found a significant negative association between R/S and physical and sexual aggression. Although R/S showed no effect on domestic violence, the subgroup analysis showed a significant negative association among adolescents. These findings have significant implications for health care professionals worldwide.

Disclosure

The authors report no conflicts of interest.