Study setting and procedures
The study was conducted in a health and demographic surveillance system (HDSS) covering 22% (n = 54,647 people) of the urban and rural population of León municipality, Nicaragua [
20]. The HDSS is a dynamic cohort that provides yearly information on births, deaths, immigration, and emigration, allowing longitudinal follow-up of study participants.
The starting point of this research is a population-based survey assessing the magnitude and characteristics of IPV during pregnancy [
17]. In this study, all pregnant women living in the areas covered by the HDSS (n = 483) were identified by a screening question posted in the HDSS baseline data collection and then invited to participate. One woman refused to join the survey and two were found not to be pregnant. During data collection we identified that two households had a pair of pregnant women living together, due to the sensibility of the study only one woman in each household was randomly selected to be interviewed. Finally, 478 women were included. This paper includes a median follow-up time of 43 months (IQR 40–45 months) of all children born alive. Women were enrolled between 2002 and 2003.
Data collection
A modified version of the WHO Multi-country Study on Women’s Health and Domestic Violence against Women questionnaire [
1] was used to measure exposure to IPV. The WHO instrument was chosen because it had been extensively used to measure IPV magnitude in developing countries and the items within each IPV sub-scale have shown high internal consistency [
21]. The WHO questionnaire was modified by including questions about emotional, physical or sexual IPV during pregnancy
.
The questionnaire contains four sub-scales measuring different IPV types (emotional, physical, sexual and controlling behavior). Each sub-scale has specific questions depicting violent actions that a man can do to his female partner. Yelling, humiliation, intimidation, and threats were considered emotional IPV. Exposure to any violent partner acts such as slaps, pushes, punches, kicks, strangulation, and use of weapons were labeled as physical IPV. The partner use of force/threats or intimidation to have sexual relations with the woman was typified as sexual IPV. A woman was classified as experiencing IPV if she answered “yes” to any of the questions explored within each sub-scale.
Exposure to controlling behavior by the father of the child was assessed on a scale containing seven items describing the following actions: if the partner restricted the woman’s contacts with friends; if the partner restricted the woman’s contacts with family; if he insisted on knowing her whereabouts at all times; if he ignored and treated her indifferently; if he became angry if she spoke to another man; if he was constantly suspicious that she was unfaithful; and if he expected her to ask for his permission to seek healthcare for herself. These were later summarized into two categories: no controlling behaviors or exposed to 1–7 controlling behaviors.
Emotional, physical and sexual IPV were assessed if they occurred in a lifetime or during pregnancy. In order to enhance disclosure, data on emotional, physical and sexual IPV were collected twice during pregnancy. The average gestational age at the first visit was 19 weeks and 36 weeks at the second. Only lifetime exposure to controlling behavior by father of the child was collected.
In our study all IPV sub-scales showed high internal consistency with Cronbach alphas above 0.80 for emotional (0.83), physical (0.89), sexual (0.82) and controlling behavior by the father of the child (0.81) sub-scales. We also performed exploratory factor analysis taking into account the questions related to IPV (controlling behavior, emotional, physical and sexual violence). The factor analysis showed four factors with eigen values greater than one. A Varimax rotation was applied to these factors because they were considered to be independent. The first factor extracted represented questions exploring physical IPV, the second factor represented questions related to controlling behavior, the third factor represented questions related to sexual IPV and the fourth factor questions related to emotional IPV (data not shown). The four factors had a cumulative variance of 0.6271 . The analysis showed that the WHO instrument can distinguish between different forms of IPV in the Nicaraguan setting.
Data on birthweight were excerpted from hospital and clinic records. At a median child age of 43 months, standing height was measured using a vertical metric rule attached to the scale and a horizontal headboard. The metric rule was graduated at 0.1 cm intervals. During the visit, two standing height measurements were collected and their mean calculated.
Demographic data (women’s age, marital status, residency, education, parity, and socioeconomic status) were collected during pregnancy. Socioeconomic status was measured by means of the Unsatisfied Basic Needs Assessment index [
22]. This variable was constructed from four indicators: inadequate housing (if the family home had a dirt floor or walls constructed from materials other than cement), low school enrolment (if there was a child of school age in the household not in education), high dependency economy (if the ratio between non-working and working persons in the household was higher than one) and inadequate sanitary conditions (if no piped water was available indoors or there was no flush toilet). Households with two or more such indicators were considered poor.
Emotional distress during pregnancy was measured using the Self-Report Questionnaire (SRQ), a 20-item questionnaire exploring depressive and neurotic symptoms that has been validated in Nicaragua [
23] . Distress was defined as a score ≥7 [
23].
Data analysis
Z-scores are the preferred way to express anthropometric indices because they are used to calculate summary statistics [
24]. The program ANTHRO 2005 (WHO, 2006) was used to convert mean height follow-up measurements into height-for-age Z-scores. Height-for-age index was used to evaluate cumulative linear growth; low values were associated with impaired long-term growth.
Student’s t-test was used to compare differences in the mean height-for-age Z-scores values for the mother’s residency, education, parity, socioeconomic status, maternal emotional distress and child gender. Birthweight, child age, mother’s age at pregnancy were not normally distributed, thus Spearman correlation was used to assess the relationship between these variables and height-for-age Z-scores. Mann–Whitney U test was used to compare median mothers’ age across IPV groups (data not shown).
Analysis of covariance (ANCOVA) was used to assess the association between different types of IPV and height-for-age Z-scores adjusted for confounding factors. Height-for-age was normally distributed (Kolmogorov-Smirnov test p value > 0.05) and the variance was homogeneous among all groups (Levene’s test p value > 0.05). Mother’s educational level, age, residency, parity, socioeconomic status, emotional distress during, child age and birthweight were considered potential confounders. A variable was included in the model if it had a p value <0.20 in its association with IPV exposure and height-for-age Z-scores and if there was a difference of 5% or more between the crude and adjusted estimates. Birthweight could or could not be considerate an intermediate factor between exposure to IPV during pregnancy and subsequent growth during early childhood. To test this hypothesis, in preliminary analysis we included birthweight as a confounding factor, however the difference between crude and adjusted estimates was less than 1% (data not shown). Thus, birthweight was excluded of the final analysis. The confounding factors included in each final model are described in the footnotes of Table
1.
Table 1
Adjusted mean height-for-age Z-scores and 95% CI by exposure to different forms of Intimate Partner Violence (IPV) during pregnancy
Any IPV during pregnancy † ‡ | Yes | −1.03 (−1.20 to −0.87) | −1.14 (−1.32 to −0.96) | −1.09 (−1.21 to −0.98) |
| No | −0.96 (−1.18 to −0.73) | −0.76 (−0.98 to −0.54)* | −0.85 (−1.01 to −0.69)* |
Emotional IPV during pregnancy ‡ | Yes | −0.92 (−1.15 to −0.68) | −1.05 (−1.31 to −0.79) | −0.99(−1.16 to −0.81) |
| No | −1.05 (−1.21 to −0.89) | −0.97 (−1.14 to −0.80) | −1.01 (−1.12 to −0.89) |
Physical IPV during pregnancy § | Yes | −1.07 (−1.43 to −0.71) | −1.20 (−1.58 to −0.81) | −1.13 (−1.40 to −0.87) |
| No | −1.00 (−1.14 to −0.85) | −0.96 (−1.11 to −0.81) | −0.98 (−1.08 to −0.88) |
Sexual IPV during pregnancy ∥ | Yes | −1.02 (−1.53 to −0.51) | −1.04 (−1.63 to −0.44) | −1.03 (−1.42 to −0.64) |
| No | −1.01 (−1.15 to −0.86) | −0.99 (−1.14 to −0.84) | −1.00 (−1.10 to −0.89) |
Controlling behavior by partner ‡ | Yes | −1.03 (−1.20 to −0.85) | −1.23 (−1.42 to −1.03) | −1.13 (−1.26 to −1.00) |
| No | −0.98 (−1.19 to −0.77) | −0.74 (−0.94 to −0.54)* | −0.84 (−0.99 to −0.70)* |
In addition, multivariate analyses were stratified by the child gender. The Statistical Package for Social Sciences (Version 15; SPSS Inc., Chicago, IL) was used to analyze the data. All bivariate and multivariate analysis were considered significant if p < 0.05.