Dating violence and physical health: A longitudinal lens on the significance of relationship dynamics and anti-social lifestyle characteristics

Abstract

Background

Scholars have documented the significant physical health consequences of intimate partner violence. Yet, because existing research draws primarily on clinical samples of adult women, it is unclear whether exposure to dating violence is related to health detriments among young men and women. Furthermore, data limitations largely have precluded consideration of the mechanisms underlying these previously observed associations.

Aims

We sought to examine the direct association between dating violence and self-rated physical health during adolescence and across the transition to adulthood. We also directed attention to potential mediating and confounding factors, including negative relationship dynamics, anti-social lifestyle characteristics and physical health correlates.

Methods

Drawing on five waves of data from the Toledo Adolescent Relationships Study (n = 3746 person-periods), we used growth curve analyses to examine these associations among a sample of young men and women in dating relationships.

Results

Longitudinal analyses revealed that dating violence was associated with declines in self-rated physical health across the period from adolescence to young adulthood. This effect, however, was attenuated with the inclusion of negative relationship dynamics.

Conclusions

Findings suggest the need to further examine the physical health consequences of dating violence, with a particular focus on the relationship context and other potential confounding forces.

Introduction

The increased recognition of intimate partner violence (IPV) as both social and public health problems has led to a proliferation of research on its consequences. Much of this work stems from the psychological tradition and focuses on mental health consequences (i.e. depression, post-traumatic stress disorder and sleep disorders; e.g. Ansara and Hindin, 2011; Walker et al., 2011; Eshelman and Levendosky, 2012). A smaller body of research, primarily conducted by public health experts, examines the influence of IPV on physical health – often with the goal of developing screening approaches to identify IPV victims in healthcare settings (Campbell et al., 2008; Black, 2011; Vives-Cases et al., 2011).

Although the link between IPV and mental health has been more clearly delineated, researchers are still in the process of disentangling the pathways leading IPV victims to experience short- and long-term detriments to physical health. Few studies, however, have included (1) attention to other negative features of the relationship that may co-vary with the experience of IPV or (2) adequate controls for emotional well-being, lifestyle factors and other potential confounds (e.g. child abuse). Finally, prior research draws almost exclusively on samples of adult women, and thus we know little about the influence of IPV on the physical health of young men and women in dating relationships.

Background

Physical IPV and health

The majority of studies on IPV and health have focused on mental health outcomes; however, analyses of the use of medical care among women with a history of IPV also suggest that this population suffers severe physical health consequences as reflected in their rate of injury, consumption of medical care and number of emergency room visits (National Center for Injury Prevention and Control, 2003). Thus, researchers have begun to devote increasing attention to physical health outcomes, including those indicative of recent abuse (i.e. physical injuries) and more chronic physical health conditions. In addition to direct effects, acute and chronic stress and post-traumatic stress disorder have been identified as potential mechanisms in the IPV-physical health association (Campbell and Lewandowski, 1997; Campbell et al., 2008; Black, 2011). Beyond stress, attention to other aspects of individuals’ personal (i.e. emotional well-being and broader lifestyle) and romantic (i.e. relationship characteristics) lives may help us further develop our understanding of the association between IPV and health.

Emotional well-being and lifestyle factors

As they transition to adulthood, individuals become increasingly independent as their access to financial and social supports cease and parental and other adult supervision decreases (National Center for Health Statistics, 2009). Accordingly, young adults experience considerable freedom to make decisions with immediate and long-lasting health consequences. Although young adults are a relatively healthy population, lifestyle factors (e.g. alcohol/substance use, eating habits and exercise) may compromise their health status. Because substance use and other lifestyle factors have been linked to IPV itself, it is important to take these factors into account in examining the physical health consequences of exposure to violent relationships. It is also important to consider the influence of preexisting mental health issues, which may be implicated in observed associations between IPV and physical health.

Features of the romantic context

Research on intimate relationships suggests that cohabitation and marriage are generally advantageous for health (Hughes and Waite, 2009; Musick and Bumpass, 2012). Although relationships present opportunities for social support, integration and improved health behaviours, they may also negatively influence physical and emotional well-being by exposing individuals to violence and other negative dynamics that are commonly observed in violent relationships (i.e. controlling behaviours, verbal conflict and infidelity) (Giordano et al., 2010; Nemeth et al., 2012; Giordano et al., 2015). Individuals experiencing these negative dynamics may have elevated levels of stress and likely do not share in the health conferring benefits of being in an intimate relationship. Prior investigations of the IPV health link have seldom considered these other features of intimate relationships.

Teen and young adult dating violence

Rates of dating violence (DV) peak during the young adult period (Johnson et al., 2015a, 2015b); yet, research on IPV and health primarily has focused on samples of adult women in married and cohabiting relationships. Consequently, we know little about the effects of DV on younger individuals, including whether violence experienced in early dating relationships produces lasting effects on the physical well-being of teens and young adults. Additionally, although the adult literature emphasises the more severe consequences experienced by female victims (Carbone-Lopez et al., 2006; Caldwell et al., 2012), research on younger populations suggests a high level of mutuality in violent relationships and considerable fluidity in the experience of violence both within and across relationships (Capaldi et al., 2003; Johnson et al., 2015a, 2015b), underscoring the need for research to consider the implications of exposure to DV among young men and women.

Current study

Using data from the Toledo Adolescent Relationships Study, the current investigation seeks to extend the existing literature by examining the physical health consequences of DV among a large representative sample of young men and women. Focusing on the implications of physical DV, we draw on five waves of structured interview data to direct attention to specific pathways leading to health detriments. This includes consideration of the role of negative relationship dynamics that, in addition to the violence itself, may negatively influence well-being. We account for potential confounding variables including exposure to childhood abuse, correlates of physical and emotional well-being and lifestyle characteristics. We capitalise on the longitudinal data by examining the temporal patterning of DV effects on health.

Data and methods

Sample

The Toledo Adolescent Relationships Study initially was based on a stratified random sample of 1321 adolescents. These data were collected in the years 2001 (wave 1), 2002 (wave 2), 2004 (wave 3), 2006 (wave 4) and 2011 (wave 5). Wave 1 also included a questionnaire administered to a parent/guardian (typically mothers). The current analyses examined data from waves 1 through 5 with a few exclusions including respondents reporting their race as ‘other’, and the youngest (12 years) and oldest (29 years) observations as small cell sizes precluded meaningful analyses of these groups. Analyses are limited to dating relationships. The final analytic sample (3746 person-periods) represented an 11-year accelerated cohort design with three overlapping cohorts (ages 13 to 28 years).

Measures

Self-rated health, assessed across all five interviews, is a single item asking the respondents, ‘In general, how is your health?’ Responses ranged from (1) poor to (5) excellent.

Dating violence, also assessed across all five interviews, used four items assessing physical violence from the Conflict Tactics Scale (Straus et al., 1996). We created a dichotomous variable to indicate any reports of violence (victimisation, perpetration and mutual).

Anti-social lifestyle factors

Alcohol use, assessed across all five waves, is based on the question: ‘In the past 2 years (or 24 months), how often have you drunk alcohol?’ Responses ranged from (1) never to (9) more than once a day. Drug use referred to using ‘drugs to get high’ and is based on a similar question and response categories. Criminal justice system contact was a dichotomous variable based on a single item from the wave 5 interview, indicating whether the respondent had ever been arrested.

Negative relationship dynamics

Infidelity, assessed at all five waves, is based on the question: ‘How often have you gotten physically involved with other girls [guys]?’ We dichotomised responses to indicate any reports of infidelity. Controlling behaviours, assessed at all five interview waves, is the mean of two items: (1) ‘X sometimes wants to control what I do’ and (2) ‘X always tries to change me’. We measured verbal conflict across the five waves as the mean of two items asking how often respondents (1) had disagreements or arguments and (2) yelled or shouted at their current/most recent partner.

Physical health correlates

Depressive symptoms, measured across the five waves, used the six-item version of the Center for Epidemiological Studies’ depressive symptoms scale. We measured overweight by asking the respondents how they compared with other similarly aged peers. Responses ranged from (1) thinner than most to (4) very overweight. Respondents identifying as overweight were coded 1 and 0 otherwise.

We included several sociodemographic characteristics: gender, age – measured in years by using a continuous variable – and three dichotomous variables for race/ethnicity including non-Hispanic White (contrast category), non-Hispanic Black and Hispanic. Family structure indicated whether the respondents lived with two biological parents at the first interview. To control for socioeconomic status, we used the highest level of education reported by the parent. Primarily answered by women, we referred to this dichotomous measure as ‘mother’s education’, which indicated college or more. Coercive parenting, measured at wave 1, asked teens, ‘When you and your parents disagree about things, how often do they push, slap, or hit you?’ Dichotomised responses indicated any coercive parenting.

Analytic strategy

We employed linear growth curve analyses to model self-rated health as a function of lifestyle factors, relationship dynamics and other potential confounds. We estimated linear mixed-effects models by using restricted maximum likelihood estimation in SAS 9.3. We presented descriptive statistics for all variables (Table 1). Next, we estimated an unconditional means model to determine the amount of variation in physical health that occurs between- versus within-individuals. In our sample, the intraclass correlation was 0.42, indicating that about two fifths of the variation in self-rated health is attributable to differences between individuals, whereas the remaining three fifths is attributable to within-individual differences. Model 1 added the linear effect of age and controls (gender, race, family structure, mother’s education and coercive parenting), and model 2 examined the effect of DV net of these factors. Models 3–6 introduced the remaining variables in blocks, including anti-social lifestyle factors, negative relationship dynamics and physical health correlates. Supplemental analyses examined variation in effects over time and tested whether the observed associations were similar for men and women.

Table 1:

Descriptive statistics for study variables (n = 3746)

	Full sample			Men (n = 1852)		Women (n = 1894)
	Mean/percentage	SD	Range
Dependent variable
Self-rated health	3.76	0.94	1–5	3.88	***	3.65
Independent variables
Age	18.34	3.37	13–28	18.37		18.31
Dating violence	28.14%			32.02%	***	24.34%
Anti-social lifestyle factors
Alcohol use	2.90	2.19	1–9	3.03	***	2.77
Drug use	1.90	1.99	1–9	1.99	**	1.81
Criminal justice system contact	22.85%			29.97%	***	15.89%
Negative relationship dynamics
Infidelity	26.86%			27.05%		26.66%
Controlling behaviours	1.99	0.92	1–5	2.13	***	1.85
Verbal conflict	2.21	0.95	1–5	2.17	*	2.25
Physical health correlates
Depressive symptoms	2.43	1.28	1–8	2.31	***	2.55
Overweight	29.23%			24.83%	***	33.53%
Sociodemographic characteristics
Female	50.56%			—		—
Race (white)
Black	24.45%			24.89%		24.02%
Hispanic	10.38%			10.53%		10.24%
Family structure
Two bio parents	52.08%			55.94%	***	48.31%
Mother’s education
College or more	22.93%			22.14%		23.71%
Coercive parenting	22.72%			21.44%	†	23.97%

^†p < 0.10.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Results

Descriptive statistics for all variables (Table 1), examined separately by gender, are based on the 3746 person-periods. On a scale of 1–5, the average level of self-rated health across both individuals and age was 3.76. Thus, even within this relatively young sample, only 24% rated their health as excellent. Roughly 28% reported DV. Average levels of alcohol and drug use were relatively low, that is, about ‘once every two-three months’ and ‘once or twice a year’ respectively. Nearly a quarter of respondents reported having ever been arrested. More than a quarter reported partner infidelity, and levels of controlling behaviours and verbal conflict were moderately low, at 1.99 and 2.21 respectively. The average level of depressive symptoms during the study period was 2.43, indicating moderately low levels of depression. Approximately one third indicated that they were overweight or very overweight.

The multivariate analyses examined the association between DV and self-rated health, accounting for anti-social lifestyle factors, negative relationship dynamics, physical health correlates and sociodemographic characteristics. Table 2 presented growth curve models predicting self-rated health. Model 1 included the linear effect of time and the full roster of sociodemographic and family background information (all measured at the between-subject level). Model 2 added DV. Consistent with prior literature, DV is associated with declines in self-rated health across the period from adolescence to young adulthood. Models 3–5 entered the covariates for anti-social lifestyle, negative relationship dynamics and physical health correlates separately. In model 3, DV continued to exert a negative influence on self-rated health controlling for alcohol and drug use and criminal justice system contact. Additionally, of the lifestyle factors, drug use negatively influenced the self-rated health net of DV and sociodemographic characteristics. Model 4 examined the effect of DV net of other negative features of the relationship, including infidelity, controlling behaviours and verbal conflict. In this model, the negative relationship dynamics were significantly associated with self-rated health net of reports of DV. Moreover, controlling for these negative dynamics, the effect of DV on self-rated health was attenuated. This attenuation largely was driven by the addition of controlling behaviours to the model. The next model (model 5) examined the effect of DV accounting for physical health correlates, including depressive symptoms and overweight. Net of DV, depressive symptoms and overweight negatively influenced self-rated health. Controlling for these factors, DV was partially attenuated; however, the effect remains significant at the p < 0.10 level. A final model assessed the association between DV and self-rated health, accounting for anti-social lifestyle factors, negative relationship dynamics, physical health correlates and sociodemographic characteristics. In this model, DV was not significant. Furthermore, drug use, depressive symptoms, overweight and controlling behaviours all exerted independent effects on self-rated health.

Table 2:

Coefficients for the multilevel model of dating violence on self-rated health (n = 3746)^a

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Fixed effects, composite model
Within-subjects
Initial status	4.072***	4.095***	4145***	4.280**	4.355***	4.515***
Age	−0.017***	−0.016***	−0.016**	−0.014**	−0.014**	−0.014**
Dating violence		−0.081*	−0.070*	0.001	−0.055†	0.010
Anti-social lifestyle factors
Alcohol use			0.005			0.009
Drug use			−0.039***			−0.030***
Criminal justice system contact			0.022			0.022
Negative relationship dynamics
Infidelity				−0.057†		−0.028
Controlling behaviours				−0.082***		−0.064***
Verbal conflict				−0.029†		−0.021
Physical health correlates
Depressive symptoms					−0.102***	−0.089***
Overweight					−0.298***	−0.292***
Variance components
Level 1: within-person	0.475***	0.476***	0.477***	0.473***	0.483***	0.481***
Level 2: in intercept	1.661***	1.622***	1.586***	1.578***	1371***	1.334***
Level 2: in rate of change	0.003***	0.003***	0.003***	0.003***	0.003***	0.003***
AIC	9481.8	9480.4	9475.2	9457.8	9344.5	9343.2
BIC	9502.3	9500.9	9495.7	9478.3	9365.0	9363.7

^aModels include the following sociodemographic characteristics: gender, race, family structure, mother’s education and family violence.

^†p<0.10.

^*p<0.05.

^**p<0.01.

^***p<0.001.

In supplemental analyses, we considered whether the effect of DV on self-rated health varied over time. The interaction of age and DV was marginally significant and negative, suggesting that DV has a more negative effect on self-rated health at older ages. However, controlling for other factors, the main effect of DV was not significant at age 18.34 (the sample average), and although approaching significance at the maximum age of the sample (28), its effect remained non-significant (p = 0.11). This highlights the importance of including attention to the broader lifestyle and relationship contexts within which DV is likely to unfold. We also examined whether DV at any given point in time was related to more immediate health detriments or whether the experience of DV had lasting effects on self-rated health. Controlling for the most recent report of DV, the long-term health consequence of exposure to partner violence was not significant. Finally, we tested whether gender conditioned the effect of DV on health and found that exposure to DV had similar effects on the self-rated health of men and women.

Discussion

Consistent with prior research on the association between IPV and physical health, the results of the current analyses suggested that reports of DV were associated with declines in self-rated health across the period from adolescence to young adulthood for both men and women. However, these findings also highlighted that DV does not occur in isolation from other ongoing lifestyle and relationship dynamics. Using data from a large diverse sample, growth curve analyses showed that inclusion of a range of negative relationship dynamics, including infidelity, controlling behaviours and verbal conflict, explained the association between DV and health. Anti-social lifestyle factors and physical health correlates were also examined, and several of these factors helped explain the IPV health link and were predictors of health in their own right. These findings further suggested the need to examine the physical health consequences of DV, with a particular focus on the relationship context and other potential confounding forces.

Scholars increasingly have recognised the necessarily dyadic nature of IPV, directing attention to the relationship context within which violence occurs. Yet, because much of the prior research on partner violence and health has been conducted by scholars of public health, limited research has examined the potential impact of this broader relationship context. This is an important omission as, for example, the presence of intrusive controlling behaviours within a dating relationship indicates a pattern that is potentially recurring and in a sense all encompassing (and in turn may be experienced as more pernicious) relative to specific acts of violence. We do not conclude from these results that DV is limited as a public health problem, but instead findings underscore the importance of considering that these behaviours unfold within the context of a broader package or constellation of relationship dynamics that are physically costly or burdensome for health and well-being. We added to prior work that has investigated mental health consequences, and it is noteworthy that models included a time-varying control for depressive symptoms.

Our analyses bridged the period from adolescence to young adulthood to examine the effects of violence in dating relationships on the self-rated health of young men and women. In addition to the associations described in the preceding texts, we examined temporal effects and found some evidence to suggest that the effect of DV on physical health increased over time; however, after accounting for other key variables, its effect remained non-significant across the age range examined in this investigation. We also considered whether the experience of DV had lasting effects on the health of teens and young adults and found that the effect of violence on physical health was largely confined to the period in which it was reported. Finally, we tested whether the observed associations were similar for men and women and found that DV exerted a similar effect on the self-rated health of men and women across the study period.

This examination, to our knowledge, is the first to assess the role of relationship dynamics with respect to DV and physical health and one of a small number of studies to consider this association longitudinally within the context of a diverse sample of respondents interviewed first in adolescence and across the adult transition period. Yet, additional research is needed that explores specific linking mechanisms, including stress processes, and where other factors treated as confounds here (drug use and depression) fit in terms of sequencing and impact. Given the considerable fluidity in intimate relationships and the experience of DV during this transitional phase of the life course, it is also important to explore whether leaving a violent relationship or desisting from violence within a focal relationship is associated with improvements over time in physical and emotional health. As in the current study, a comprehensive approach to these more focused investigations will benefit from incorporating the broader relationship context and lifestyle factors into such assessments.