For the present work, a secondary analytical cross-sectional analysis of the publicly accessible 2015–2017 Demographic and Family Health Survey (Encuesta Demográfica y de Salud Familiar, ENDES) was performed [23].
Records for women between 15 and 49 years of age who were randomly selected to respond to the domestic violence module of the ENDES individual survey for the period 2015–2017 and interviewed face-to-face by INEI personnel were included.
The ENDES used a two-stage probabilistic sample that was balanced, stratified and independent at the department level and analyzed by rural and urban areas. This type of sampling allows total estimates that are approximately equal to the characteristics of the reference population to be obtained [24].
The selection criteria were as follows. The inclusion criteria was women between 15 and 49 years who were selected to complete the violence module that included questions about physical and sexual violence and about where they lived for the longest until they were 12 years old, their current place of residence and the amount of time they lived there. The exclusion criterion was having lived abroad for most of their life before the age of 12 years [25].
Measurement of variables
The outcome variable was reported physical and/or sexual violence inflicted by the partner (IPV) at some time during the last 12 months. This variable was developed based on ten questions regarding having ever being pushed, slapped, hit with a closed hand, kicked, suffering from an attempted strangling or burns and being threatened and/or attacked with a knife or firearm. Additionally, the participants were asked about being forced to have sexual relations and/or perform other sexual acts without consent. All questions had to be answered with “yes” or “no”, and if the answer was yes to any of these questions, the women were categorized as having reported IPV. The questions in the ENDES are a modified version of the Conflict Tactics Scale (CTS-2), which has high reliability and comparability among different cultures [26, 27].
The main exposure variable was internal migration. This was operationalized using information from three questions: current place of residence, principal place of residence before 12 years of age and number of years of residence in the current place. Migrants were classified as those who reported having lived for 5 years or more in the current location, this fixed interval of time lived is one of the internal migration recommended best measures [28,29,30] and were categorized as follows according to changes in their primary residence and current residence as rural-rural migrants, urban-urban migrants, urban–rural migrants and rural–urban migrants, recent migrants were those who reported having lived less than 5 years in the current location without migration pattern specification. Nonmigrants were those who reported having resided in the same place all their lives.
In addition, other variables were included, including age in years, educational level, marital status, native language, paid occupation of the interviewee, socioeconomic level, age at first marriage/cohabitation, age when sexual intercourse started, number of children, alcohol consumption by their stable partner in the last 12 months, number of sexual partners in addition to their stable partner in the last 12 months and history of physical aggression by their father toward their mother [15].
Statistical power was calculated using Open Epi, version 3.01, considering 7571 women who were exposed to IPV and 56,288 who were not exposed, as well as the prevalence of IPV in the last 12 months among migrant women from China (19.04%) [15] and the prevalence of IPV during the last 12 months in women in the general population of Peru (10.8%) [31]. With a confidence interval of 95%, the power was greater than 80%.
The present study was reviewed and approved by a Ethics Committee of the Universidad Peruana de Ciencias Aplicadas.
Data analysis
For the statistical analysis, the STATA 16 MP program was used, considering a 95% confidence level.
A descriptive analysis of the sociodemographic characteristics of the population and the prevalence of physical violence and internal migration was performed, and internal migration subgroups were identified. Simple frequencies and weighted percentages are reported for the categorical variables, and means and standard deviations are reported for the numerical variables.
Then, a bivariate analysis using the Pearson chi-square test was performed to identify the associations of the categorical variables. Numerical variables were compared by regression and the Wald test.
To identify the association between internal migration and physical violence, a generalized linear model (GLM) of the family and the log Poisson link option was used, and the results are presented as prevalence ratios (PRs). A crude model and a model adjusted for confounding variables were performed; confounding variables were entered into the model according to epidemiological criteria. To evaluate the collinearity between the independent variables that were included in the final adjusted model, the variance inflation factor (VIF) test was used, and a value of ten was used to exclude a variable from the model. Additionally, a correlation analysis of the variables was performed; a correlation greater than 0.5 was found between education level and socioeconomic level (0.052), so the latter was excluded from the final adjusted model. For the statistical analysis, the stratified design of the primary study was taken into account, and adjustments for sample weights were made using the "svy" commands included in STATA.