Critique of Factors Impacting Domestic Violence - Research Paper Example

? Critique of Factors Impacting Domestic Violence Number 2nd April Introduction Domestic violence is a crime that occurs across cultures and transcends socioeconomic status. Domestic violence can affect women or men of any age, although it is generally accepted as a phenomenon that occurs primarily against women. The importance of domestic violence lies in the fact that it is so serious, so common and can manifest itself in so many ways. The effects of domestic violence upon the victim can be terrible and long-lasting, with both psychological and physiological marks being left. The purpose of this paper is to explore the findings of scientific research about the results of domestic violence using four quantitative studies. The first of these, by Ellsberg, Jansen, Heise, Watts & Garcia-Moreno (2008) explores the links between intimate partner violence and the physical and mental health using results from the WHO multi-national study on the topic. Bonomi et al (2009) completed a similar study looking at the medical and psychosocial ailments of women who had been victim of domestic violence. Ackerson & Subramanian (2008) studied the more specific and perhaps more subtle effect of malnutrition in India and the links between this state and domestic violence. Similarly, the Hashemi & Beshkar (2011) study looked very specifically at the maxillofacial fractures of domestic violence victims in Tehran, Iran. These all come together to suggest that there are a wide variety of marks that domestic violence can leave on the victim’s body, not all of which would manifest themselves in the way that would be expected. Methods The wide variety of statistical tests used in the aforementioned studies represents the diverse nature of their content and hypotheses. The purpose of this section is to give a brief overview of the methods and statistical testing in each of these studies and then to give a critical analysis of their use and purpose. Ellsberg et al (2008) used the results of a WHO multi-national study on domestic violence, which employed trained female interviewers to talk to women in 15 sites in 10 countries. The women interviewed were asked whether they had been victim of domestic violence, and if the answer was yes then which injuries they sustained. Women were also asked to give a general estimate about their health overall, and a questionnaire designed to approximate their mental health. These results were then analysed by a multitude of statistical tests; bivariate descriptive analysis and multivariate logistic regression modelling for associations between violence and self-reported health problems, unadjusted and adjusted odds ratios (OR) with 95% CIs to calculate the likelihood of health problems in victims of domestic violence, and negative binomial-regression techniques to determine the significance of the SRQ-20 questionnaire for mental health outcomes. The use of so many different statistical analyses represents the different types of results which are inevitable in conducting such a study examining a wide-range of health outcomes. The one obvious problematic area is the use of the SRQ-20 questionnaire and the negative binomial-regression techniques. The SRQ-20 questionnaire is usually analysed based on culture and country specific cut-off points that have been determined after a number of years of analysis on people in that country. In this study, many of the countries had not received a SRQ-20 questionnaire rating and the negative binomial-regressions were calculated to ascertain the cut-off point for the mental health of the participants. Evidently, it would be preferable to use the more traditional way of analysing these results. Another problem with the statistical analyses used here is that they are all so different and therefore it may be difficult to get comparable results in this way, but it represents the difficulty of comparing things like overall health, mental health and physical ailments. Bonomi et al (2009) did a similar study investigating a number of different health outcomes for domestic violence victims. The thing that sets this apart from the Ellsberg et al (2009) study is that Bonomi et al (2009) found their own participants using their own sampling methods, which is useful because it means that the results were all found in the same way and therefore can be judged to be comparable. The study used 3568 women who fit a number of stringent selection criteria. ?2 tests and tests for trend were used to compare the demographic characteristics of the women identified; generalized linear models with a log link and binomial errors were used to estimate RRs between the diagnoses of the victims compared with the non-victims. The use of the X2 test is interesting as it is one of the more common inferential statistical tests. In this case, it seems extremely appropriate as it is used to draw links between demographics in the population. It tests the null hypothesis of the work that the frequency distribution of the demographics is a true representation of the theoretical distribution. This is needed in this case to give a basis for the linear models the study employs to find the relative risk of certain medical problems for women who were victims of domestic violence. The first of the two more specific studies, Hashemi & Beshkar (2011) analysed patients with maxillofacial injuries in one hospital in Tehran between 2004 and 2006. The cause of the fracture was recorded, with additional data being taken for those who reported that it was a result of domestic violence. Again, an X2 test was used to test the null hypothesis that the expected data would not be different from the recorded data. In this case, we would expect the proportion of maxillofacial fractures to represent the population. Again, the test is very appropriate in this situation as it analyses how well the data in the test fits with the data expected, and therefore can be used to make inferences about the type of people that are victim of maxillofacial fractures and the incidence of these occurring as a result of domestic violence. Ackerson & Subramanian (2008) studied the link between malnutrition and anaemia in domestic violence victims in India. The data used was from a previously conducted study (the Indian National Family Health Survey), taking into account the haemoglobin and weight counts of the women in the study and comparing it to reports of the incidence of domestic violence. The researchers felt that there may be clustering at all levels, and therefore used a logistic multilevel modelling procedure to adjust for this, a fairly standard procedure which is well-used and relevant in this case. The rest of the statistical analysis used a three-level model for the binary responses and these were linked as part of a formula to a set of predictors, which was composed of a fixed component and two random intercepts that related to the neighbourhood and the state of the respondent. All models were created by using penalized quasi-likelihood approximation with second-order Taylor linearization. The statistical analysis used here is well-described, but there may be some problems in the use of an analysis that uses such a mixed approach using variables that are defined by the researchers themselves, including the fact that the analysis may have been altered to give a more favourable or desirable response. The test could be repeated, but there are a number of other ways to analyse this, including using the more traditional X2 test to see if the expected values of anaemia and malnutrition in the population were comparable to the experimental values. Results Evidently, the scope of this paper means that the results of the four studies cannot be examined in any great detail, but an overview is given here. Ellsberg et al (2009) used a unadjusted and adjusted odds ratios (OR) with 95% CIs. The odds ratio was 1.6 and with the confidence interval became a range of 1.5 to 1.8. The results found that there were a number of health outcomes found in the participants falling into this domain, including difficulty walking (1·6 [1·5–1·8]), difficulty with daily activities (1·6 [1·5–1·8]), pain (1·6 [1·5–1·7]), memory loss (1·8 [1·6–2·0]), dizziness (1·7 [1·6–1·8]), and vaginal discharge (1·8 [1·7–2·0]). Mental health was also found to be affected, with suicidal thoughts (2·9 [2·7–3·2]), and suicidal attempts (3·8 [3·3–4·5]) being extremely high compared to the non-abused populations. The researchers felt it was appropriate to say that this suggested that domestic abuse victims were highly likely to have health complaints. Taking into account the statistical analysis and the data, this seems like a highly appropriate conclusion to draw when taking into account the high rates of mental health problems particularly. Bonomi et al (2009) presented the results of the medical and psychosocial diagnoses after the statistical analysis. They found that the results of the age-adjusted and non-adjusted models were very similar and chose to focus on the age-adjusted model as a result. This seems like an appropriate decision as adjustment is an important part of research allowing different characteristics which are variables in the population to be eliminated from research, preventing bias. It is useful that they acknowledge the fact that they have done this. The following are the results of the study (with the number in brackets being the relative risk of the illness described): psychosocial/mental disorders (substance use, 5.89; family and social problems, 4.96; depression, 3.26; anxiety/neuroses, 2.73; tobacco use, 2.31); musculoskeletal disorders (degenerative joint disease, 1.71; low back pain, 1.61; trauma-related joint disorders, 1.59; cervical pain, 1.54; acute sprains and strains, 1.35); and female reproductive conditions (menstrual disorders, 1.84; vaginitis/vulvitis/cervicitis, 1.56). The researchers decided to draw the conclusion that these relative risks were high enough to conclude that women who were victims were more likely to suffer from a whole variety of medical and psychosocial conditions than their non-abused counterparts. This is a sensible conclusion to draw given the results, although it is clear that the psychosocial/mental disorders are far more likely to occur and the relative risks of the physical disorders are small enough to be coincidental in some cases. This is not acknowledged in the results of this paper, but perhaps should be. Hashemi & Beshkar (2011) found that maxillofacial fracture was the result of domestic violence in 3.5% cases of admission for the problem, with all of these being women. Taking this into account, the researchers felt it was appropriate to ignore the male hospitalizations when talking about domestic violence fractures, an appropriate decision given the nature of the responses. After adjusting for gender, 13% of all maxillofacial fractures were a result of domestic violence, which is extremely high. The researchers felt that it was appropriate to conclude that maxillofacial fractures were often the result of domestic violence in women. Although this seems like a relevant conclusion, there is no indication of the incidence of domestic violence in the population as a whole to offer a comparative value, so it would be wise to include this data. Ackerson & Subramanian (2008) make strong distinctions between the adjusted and non-adjusted portions of their results. They found that women who reported more than one instance of domestic violence in the previous year had a substantially increased likelihood of having anemia (odds ratio (OR) = 1.22, 95 percent confidence interval (CI): 1.14, 1.30; p for trend < 0.0001) and also had an increased likelihood of being underweight (OR = 1.51, 95 percent CI: 1.41, 1.61; p for trend < 0.0001) in the unadjusted models. Using an adjusted model, however, the strength of the associations were not as strong (although still significant in many cases). The researchers draw the conclusion that women who are victims of domestic violence in the last year are more likely to be underweight or anemic. This is quite a strong conclusion to draw based on the results, with the links between the two being significantly weaker when using an adjusted model. As the results came from such a wide area of India with different demographic conditions, it is necessary to use an adjusted model and therefore the unadjusted results can be quite confusing and make the link appear stronger than it is in reality. Discussion The four studies above can be grouped into two groups of two; the general overview of both the medical and psychological health of domestic violence victims and the specific incidence of one type of injury in a certain area. The general overviews by Bonomi et al (2009) and Ellsberg et al (2009) were very similar in using a large amount of participants and measuring their mental and physical health. The benefit of the Bonomi et al (2009) study is that they found their own participants and therefore were more in tune with the sampling design, whereas that in Ellsberg et al (2009) was quite vague. Both used reliable and commonly used statistical tests, which were appropriate in both cases, particularly as the sample sizes were so large. The use of such large samples in itself was both a strength and a weakness. Using a large sample size is good because it reduces the likelihood of bias and strong conclusions can be drawn. However, the studies must be careful to adjust their results, particularly in the Ellsberg et al (2009) study where the results represented a number of different countries with different attitudes towards domestic violence. The Ackerson & Subramanian (2008) and Hashemi & Beshkar (2011) were far more specific and as a result did not have such large sample sizes. The benefit of this allows for more in depth analysis of the specific ailment (anemia and malnutrition or maxillofacial fractures respectively) and how it relates to domestic violence. The problem with the Ackerson & Subramanian (2008) study mostly centres around the use of a quite confusing statistical analysis which had very little reference to existing statistical models, which made the results quite hard to compare to others. The study does explain the mathematics behind the results, but does not go into much detail about the validity of this type of method and where it is used elsewhere in the literature. The Hashemi & Beshkar (2011) was much stronger in this aspect. However, the study was much more focused on ascertaining the percentage of maxillofacial structures that were resulting from domestic violence, whereas the other three studies focused on inferring illness from the domestic violence. This makes it stand out, and the results harder to compare to the others. It also is the only study to include men in the sample, which again makes it difficult to compare the inferences from the results. Future research could examine each of the different medical problems separately in a cross-sectional study of a number of different nations. This would help to remove the confusion from the larger studies, particularly Ellsberg et al (2009), about the different types of sampling, statistical measuring and variables used. Employing a simple X2 test in all of these would make the results comparable and the research would be able to draw conclusions about the likelihood of these ailments individuals when compared to the general population. These could be brought together in a meta-analysis to explore the links between physical and psychological injuries to domestic violence victims. Conclusion Evidently, each study used a different type of statistical analysis that was deemed to be appropriate for each data set. Generally, the reasoning behind the choice was not covered in the research paper, although in most cases (excepting Ackerson & Subramanian [2008]) the choice of statistical analysis used was a common one and had been covered for appropriateness and utility in a number of ways beforehand. Although we should be careful not to assume that a statistical analysis works simply because it has been used a number of times before, the use of the X2 test is appropriate in many cases of testing a null hypothesis, as is done in some of the studies above. The results of the four studies above all give results that suggest that individuals that are victims of domestic violence are likely to have a number of psychological and physical health effects that surpass the level of these found in the general population, as suggested by the thesis. References Ackerson, L. K., & Subramanian, S. V. (2008). Domestic Violence and Chronic Malnutrition Among Women and Children in India. American Journal of Epidemiology, 167(10), 1188–1196. doi:10.1093/aje/kwn049 Bonomi, A. E., Anderson, M. L., Reid, R. J., Rivara, F. P., Carrell, D., & Thompson, R. S. (2009). Medical and Psychosocial Diagnoses in Women With a History of Intimate Partner Violence. Arch Intern Med, 169(18), 1692–1697. doi:10.1001/archinternmed.2009.292 Ellsberg, M., Jansen, H. A., Heise, L., Watts, C. H., & Garcia-Moreno, C. (2008). Intimate partner violence and women’s physical and mental health in the WHO multi-country study on women’s health and domestic violence: an observational study. The Lancet, 371(9619), 1165–1172. doi:10.1016/S0140-6736(08)60522-X Hashemi, H. M., & Beshkar, M. (2011). The prevalence of maxillofacial fractures due to domestic violence – a retrospective study in a hospital in Tehran, Iran. Dental Traumatology, 27(5), 385–388. doi:10.1111/j.1600-9657.2011.01016.x Read More