Ebonyi State is in the south-east region of Nigeria and is primarily an agricultural region, with an estimated population of 3,140,944 [19]. According to the Ebonyi State Ministry of Education, there are 299 registered secondary schools with a total school population of 367,001 and also higher number of public schools.

In-school adolescents aged 10–19 years were recruited using a multi-stage sampling method. Out of 13 local government areas (LGAs) in Ebonyi state, four were selected using simple random sampling by balloting; with two being rural and two being urban. In each of the selected LGA, three secondary schools were selected by simple random sampling from the list of secondary schools. One private and two public schools were selected from each L.G.A using a proportion of 1:2. Equal number of students (108) were allocated to each school from the sample size. From the earlier selected schools, a class from each level of study was selected randomly. However, for schools with only one class per level, the class was automatically selected. Subsequently, participants were proportionately allocated to every level of study in each school. All eligible students in the class were consecutively recruited into the study until the allocated sample size was met. For each level of study, all recruited participants were seated well-spaced from one another in a large classroom and the questionnaire self-administered with no interference from the investigator. These sessions lasted approximately 30 min each.

Sample size was calculated using Cochran formula for sample size for single proportion, considering 95% significance level, 3% margin of error(d) and an estimated proportion of 35% of PPV drawn from a previous study [20], while accounting for an expected 20% for non-response. This gave an estimated sample size of 1,217. Eventually, a total of 1,296 adolescents aged 10–19 years participated in the study; with a mean age of 15.0 ± 2.0 years and the proportion of males recruited 42%.

Data was collected between November 2020 and January 2021, using a pretested self-administered WHO/CDC Global school-based Health Survey (GSHS) questionnaire as adapted. The adaptation occurred by modifying a few questions from the original GSHS questionnaire to fit the cultural context of the study area. For example, in the question on drug/substance use, the term “marijuana” in the GSHS questionnaire was modified to a different term (“Igbo”) which is more popularly used within the study area. Furthermore, the language for some of the questions were rephrased for better understanding following a pre-test conducted.

Ethical approval was obtained from the Research and Ethics Committee of the Federal Teaching Hospital, Abakaliki, Ebonyi State (FETHA/REC/VOL2/2019/165). Permission was obtained from the school authorities. Written informed consent and assent were obtained from the parents and participants respectively.

The dependent variable – peer physical violence, was assessed using the question, “During the past 12 months, how many times were you in a physical fight?” Participants who had not been in a physical fight were coded as “0 = No” while those involved in 1 or more fights were coded as “1 = Yes.”

Independent variables including family socioeconomic status (as derived based on methods proposed by Olusanya and colleagues) [21], drug use, smoking, gambling, religiosity, bullying, suicide attempt, sexual exposure, weapon carrying, serious injury and alcohol use as described in Supplementary Appendix S1, were assessed.

All quantitative analyses were done using STATA software (Version 18; College Station, TX:Statacorp LLC). Descriptive analysis was done using mean and the standard deviation for continuous variables, while proportions and frequency were used to describe the categorical variables as when appropriate. Tables were used to display some of the data. Univariate and multivariate binary logistic regression models were fitted to determine factors associated with peer physical violence (PPV). The relationship between each of the independent variables and PPV was analysed in the unadjusted regression analysis at 5% significance level. The independent variables that were not significant in the unadjusted analysis were excluded for analysis in the adjusted model while the variables that were significantly associated (confidence intervals excluding the null value of one) with PPV in the unadjusted analysis were included in the final adjusted regression model at 5% significance level.

In the univariate binary logistic analysis, age, religiosity and all the measured health risk behaviours of the adolescents were predictive of engaging in PPV (Table 3).

In the multivariate binary logistic regression, adolescents aged 14–16 years (AOR: 0.63; 95% CI:0.46–0.85; p = 0.002) and 17–19 years (AOR:0.37; 95% CI: 0.25–0.53; p < 0.001) were significantly less likely to be involved in peer physical violence compared to the younger adolescents (10–13 years).

Adolescents involved in gambling (AOR: 1.56; 95% CI:1.13–2.16; p = 0.007), who smoke cigarettes (AOR: 1.85; 95% CI:1.01–3.40; p = 0.047) and those who have had serious injury in the past 1 year (AOR: 2.29; 95% CI:1.78–2.95; p < 0.001) were significantly more likely to be involved in peer physical violence. Adolescents that were bullied had a 1.6 times higher odds of being involved in PPV than their peers (AOR: 1.70; 95% CI:1.28–2.25; p < 0.001) while those who were bully perpetrators had about twice higher odds of being involved in PPV than their counterparts (AOR: 1.89; 95% CI: 1.37–2.61; p < 0.001). Conversely, adolescents that were religious had a 0.7 times less odds of being involved in PPV (AOR: 0.70; 95% CI: 0.53–0.92; p = 0.011) (Table 4).

The findings of this work would need to be interpreted knowing that this was a cross-sectional study in which it will not be possible to demonstrate the temporality of the relationship between the factors and physical fighting. However, the large sample size and the use of an internationally validated questionnaire are strengths of the study. In the future, it may be important to carry out a qualitative study as a way of providing further insights into potential social nuances around peer physical violence.

In conclusion, the prevalence of peer physical violence also known as physical fighting is high in Ebonyi State, Nigeria. The risk factors were young adolescent age, bullying, gambling, cigarette smoking and having had a serious injury. Being religious was significantly associated with protection. The findings of this study have provided evidence that can be used in designing and implementing interventions and policies aimed at curbing this pattern of violence.