Factors associated with elder abuse: a systematic review of the literature

Elder abuse is a complex public health problem. It is of fundamental importance to ascertain which factors are associated with each specific type of abuse, as a way of enabling the creation of evidence-based public policies. The aim of the present study was to systematically review the literature regarding analytical epidemiological studies of factors associated with elder abuse. Four databases were used for the bibliographic search: Pubmed, Scopus, Web of Science and Lilacs, with no limitations regarding year of publication. Articles were selected by pairs of researchers in two stages: reading of abstracts (3.121) and reading of complete articles (64). The total number of articles selected was 27. The risk of bias was evaluated. The factors associated with general violence were age, sex, marital status, educational level, income, family arrangement, social support, solitude, mental disorder, depression, suicide attempt, dependence on others in daily activities, cognitive function, chronic diseases, alcohol or drug abuse, among others. Elder abuse was found to be a multifactorial and complex phenomenon that should not therefore be viewed in a partial one-dimensional manner, but in such a way as to take into account all dimensions and the interdependence of these.


introduction
The demographic profile of human beings is undergoing a worldwide transformation and the number of elderly people now stands at 962 million individuals aged 60 and over and is expected to double by 2050 and triple by 2100 1 . Parallel to the population growth of the elderly, there is also an increase in abuse in this population, as these individuals become more vulnerable and dependent on others, either for performance of basic daily activities, or in terms of psychological or economic dependence, especially in the case of those with impaired cognitive faculties or with the natural limitations of aging itself, which diminish the ability to defend oneself and leave the individual prone to acts of aggression 2 .
According to the World Health Organization (WHO), violence is defined as the which either results in or has a high likelihood of resulting in injury, death, psychological harm, maldevelopment, or deprivation use of physical force or power, threatened or actual, against other individuals, groups or communities, in such a way as is likely to result in death, physical, psychological or sexual harm, maldevelopment, financial deprivation, negligence, abandonment or self-neglect 3 .
Physical violence is characterized by the use of physical force to make someone to do something against their will, to harm, or to cause pain, incapacity or death. Psychological violence is a verbal or gestural offense, which many involve terrorizing individuals, humiliating them, limiting their freedom or alienating them from social life. Sexual violence aims to obtain arousal, sex, or erotic behavior by way of grooming, physical violence or threats of a homosexual or heterosexual nature. Financial or economic abuse is the improper and non-consensual ownership of the physical or financial assets of the elderly 4 .
Abandonment is the absence of provision, by the government, institutions or family, of relief to an elderly person in need of care. Negligence is the omission of care for the elderly, by relatives or institutions and self-neglect is the elderly's own lack of care for themselves, which may jeopardize their health and safety 3,4 . In relation to mistreatment of the elderly, the WHO defines this as a one-off or repeated event, or even an absence of an appropriate act, that occurs in the context of a relationship of trust and causes injury, suffering or distress to the elderly 5 .
In addition to physical injuries, the effects of violence on health include disability, depression, physical health problems, smoking, high-risk sexual behavior, alcohol and drug abuse, a host of other chronic and infectious diseases, and death 2 . Abuse has a major impact on health systems and criminal justice, as well as social services. All types of abuse are strongly associated with social determinants, cultural and gender norms, unemployment, income inequality, limited education, greater access to firearms and other types of weapons, and excessive consumption of alcohol, among other factors 6 .
As elder abuse is a complex public health problem, it is of fundamental importance to ascertain which factors are associated with it, and, in particular, the specific factors associated with each type of abuse, as a way of ensuring the creation of well-substantiated evidence-based public policies.
Only 17% of the 133 countries studied in the World Report on the Prevention of Violence in 2014 carried out representative national population-based surveys of elder abuse. Most countries, therefore, adopt strategies to contain violence without addressing this specific problem 6 . The identification of associated factors contributes to the prevention of violence by reducing risk or generating protection for the elderly population.
The present study thus involved a systematic review of the literature regarding analytical epidemiological studies of factors associated with violence against the elderly.

Method
A systematic review of the literature was carried out, according to the guidelines contained in Main Items for Reporting Systematic Reviews and Meta-analyses (PRISMA) 7 and Meta-analysis of Observational Studies in Epidemiology (MOOSE) 8 . The review was guided by the following question: "What factors associated with elder abuse appear in the literature in analytical epidemiological studies?" Articles included were observational epidemiological studies of a cohort, case-control or cross-sectional type, whose outcome (dependent variable) was elder abuse. The variables associated with the outcome found in each study were considered. The following types of article were excluded: studies of elderly people with specific diseases; studies without multivariate analysis; and studies of specific populations (institutionalized elderly or home care).
The bibliographic search was guided by the descriptors "violence" and "aged, 80 and over", located in the list of Descriptors in Health Sciences, the Virtual Health Library (http://decs.bvs. br) and the Medical Subject Headings -Mesh, available from the US National Library of Medicine (http: // www. nlm.nih.gov/mesh/). Four databases were used for the bibliographic search: Pubmed, Scopus, Web of Science and Lilacs. In the Pubmed database, the following search key was used: (age, 80 and over [MeSH Terms]) AND violence [MeSH Terms] In Scopus the search key was: (KEY (violence) AND KEY (aged, 80 AND over)). In the Web of Science the search was guided by the following key: Topic: (violence) AND Topic: (aged, 80 and over). In Lilacs the search key was: "VIOLENCIA" [Subject descriptor] and "elderly of 80 years or more" [Subject descriptor]. All searches were performed on July 12, 2017, and there were no time or language limits, in order to identify as many articles as possible on the subject under review.
Using the descriptors, the search of the selected databases led to the identification of 3121 articles for potential inclusion in the systematic review. The selection of these articles was carried out in two stages: reading abstracts and reading the full article. Initially, a pilot study was conducted, which involved reading the first 100 abstracts found so as to establish the inclusion and exclusion criteria, and then the other abstracts were read. The abstracts were read by two researchers, authors of this study (MABS and PFF), independently, based on the inclusion and exclusion criteria pre-defined in the research protocol.
After reading the abstracts, the Kappa Index was applied to analyze the agreement between the two researchers and to validate protocol selection criteria. For the pilot of the first 100 abstracts a Kappa of 0.81 was found, and for all 3121 abstracts the Kappa was 0.57, representing near perfect and moderate agreement, respectively. Of the 3121 abstracts read, there was agreement on 33 for inclusion in the full reading phase and 3040 for exclusion. There were divergences regarding 48 abstracts and these were then read by a third researcher, also an author of the present study (VLS), and a consensus meeting was held with the three readers to refine the pre-defined criteria. After the meeting, there was consensus for inclusion of 31 abstracts and exclusion of 17. Finally, 64 abstracts were included in the second stage of full reading of the articles.
The full reading of the articles was carried out independently by the same two readers from the previous stage. There was agreement on 19 articles for inclusion in the review and 32 for exclusion. There was a divergence regarding 13 articles, which were subsequently read by a third researcher (VLS). A consensus meeting of the three readers agreed to include 8 articles on which there were divergent opinions. Four duplicate articles and four articles covering a special population were identified and these were also excluded, leaving at total of 27 articles (Figure 1). The Kappa index for this stage was 0.59, indicating moderate agreement.
Twenty-seven articles were thus selected for the systematic review, all from the Pubmed and Scopus databases. Data was extracted from the articles independently by two readers (MABS and PFF), using a protocol established by the researchers. The data extracted were: author(s), year, title, publication language, country, study objective, study population, studied age, study design, study period, sample size (elderly), type of violence, violence measured, type of aggressor, measure of violence, statistical analysis, associated factors and authors' conclusion. Statistical data were expressed in terms of relative risk (RR), odds ratio (OR), adjusted prevalence ratio (APR), confidence interval (95% CI) and p < 0.05.
Having extracted the data, the risk of bias in the articles was analyzed using the Newcastle-Ottawa Scale (NOS) 9 . The NOS scale measures the methodological quality of a study by the number of stars received in the selection of study groups, comparability of groups and verification of exposure/outcome. The original scale was devised for cohort studies. For cross-sectional studies, an adapted version of the case-control study scale was used. The risk of bias was assessed for each scale question as follows: if the answer was "Yes for low risk of bias," one star was allocated (*), if "No for high risk of bias," no star was not allocated. All items are worth one star (*), except for comparability, which can receive up to two stars. Cross-sectional studies can receive a maximum total of eight stars and cohort studies a maximum of nine (Chart 1).

results
Twenty-seven studies were included in this systematic review, of which 23 were of cross-sectional design, two case-control studies and two cohort studies. Most of the articles selected were produced in the United States, followed by China, Korea and Spain, with only one study each for other countries (Chart 2). The countries covered span three continents: the Americas, Asia and Europe. No studies were from Africa or Oceania. Only one article was written in the Portuguese language and one in Korean. All the others were written in English. The sample size of the studies ranged from 164 to 24,343 elderly individuals and the year of publication ranged from 1997 to 2016 (Chart 2).
In relation to the population studied, most studies included covered the elderly population as a whole. Among these, a cut-off point of 60 years or more was more common than a cut-off point of 65 years or more. Only one study covered elderly individuals aged 75 years or older and no article used a cut-off point of 80 or older (Chart 2).
The types of abuse studied varied. Most studies generalized and used terms such as general violence, mistreatment or abuse; others were more specific and cited the types of abuse studied, with emotional or psychological abuse and financial abuse being the most prevalent, followed by physical abuse, neglect, self-neglect, sexual abuse and verbal aggression (Chart 2).
Violence was measured in numerous ways. Data were obtained using questionnaires, interviews or forms or through secondary data. The most commonly used instrument was the Scales of Tactics and Conflicts (CTS)/(CTS2) questionnaire, in four studies [10][11][12][13] (Chart 2).
The identification of factors associated with violence was organized into two axes, according to the outcome of each study included. Initially, the factors associated with any type of abuse were identified. Subsequently, factors associated with specific types of abuse (self-neglect, negligence, verbal abuse, emotional or psychological abuse, financial abuse, sexual and physical violence) were also investigated.
The factors associated with general violence were age, sex, marital status, education level, income, family arrangement, family relationship, social support, solitude, mental disorder, depression, suicide attempt, ADL dependence, cognitive function, chronic diseases, abuse of alcohol or drugs and poor bodily or oral hygiene (Chart 3).
Age was included as a factor in four studies. Two articles found that being under 70 years of age is a risk factor for abuse 10,14 . This diverged from the findings of two other articles that found this age to provide protection 12,15 . Three studies stated that women are more at risk of abuse [16][17][18] , while another came to the opposite conclusion 14 (Chart 3).
One study showed that those who live alone or with children are ten times more likely to be suspected of being abused 18 . Arguments and conflict with family members or friends were also found to pose a high risk of abuse 18,19 , and elderly individuals with some kind of mental disorder were nine times more likely to be subject to abuse than those without 17 (Chart 3).
Dependence on others for activities of daily living (ADL) was also found to be a risk factor in three studies 15,18,20 , with only one article diverging and arguing that lower ADL provided a 4% protection 14 . Alcohol abuse was considered significant for general violence, with an eight-fold increased risk for alcohol abusers 19 . Those with poor bodily or oral hygiene ran a 12-fold greater risk of abuse 18 (Chart 3).
As for the factors associated with specific types of abuse, age greater than 80 years was associated with the risk of mild, moderate and severe self-neglect 21 . Black men and black women were six and four times more likely to self-neglect, respectively 22 . Fewer years of schooling were found to be a risk factor in two studies 21,22 and lower income individuals were found to be five times more likely to self-neglect 22 (Chart 4).
Being separated or divorced doubled the risk of elder neglect, as did living below the poverty line, which was shown to be a risk factor in . Lilacs (24) . Pubmed (1,830) . Scopus (1,207 (1) identification Selection elegibility inclusion two studies 11,14 . Those who live with other family members were five times more likely to be neglected than those who live alone 23 and those who had depression twice as likely 24 (Chart 4).
The lowest economic level poses a threefold risk of verbal abuse and a very good family relationship was found to provide protection in 100% of cases 14 (Chart 4).
Four studies showed that being older is a protective factor for emotional or psychological violence 11,14,24,25 , and two studies indicated that being single, divorced or separated are risk factors 11,24 . This diverged from the findings of another survey, which found these factors to provide protection 25 . While a low level of schooling was found to be a protective factor in two studies 11,24 , one found it to be a risk 25 . The lowest economic level conferred a fourfold increased risk of elder abuse 14 , whereas those who had a diagnosis of depression were victims of emotional abuse seven times more often than those who did not 24 (Chart 4).
Financial abuse was three times greater in those aged over 85 years 23 , nine times higher in those who had the lowest economic level 14 and twice as likely in those who had depression 25 or a physical disability 24 (Chart 4).
One study showed that being single doubled the risk of sexual violence and tripled it for those with no income 26 . Another study found that poor social support increased the risk of this type of abuse fourfold 27  Lower economic level or feeling desperate was found to pose a fourfold greater risk of physical violence 14,28 and those who have a diagnosis of depression were six times more likely to be victims of such violence 24 (Chart 4).

Discussion
Elder abuse is a universal phenomenon. Although this is a relatively new topic, there are numerous published studies on the subject. The issue was first publicized in 1975 as "grandparent beating" in British magazines. It was seen as a social and policy problem in a population-based epidemiological survey that estimated the prevalence of such mistreatment in the United States in 1988 29 . In the 1980s and 1990s, scientific research and government action was undertaken in a number of countries. Elder abuse was initially identified in more developed countries, where most of the studies were conducted 3 .  In the present systematic review, the studies were conducted in countries in the Americas, Europe and Asia. No study came from Africa or Oceania. It has been shown that studies of violence in the elderly population are concentrated in countries with a higher level of human development (UNDP, 2015) and this may mask the reality of even higher rates of global violence than those already known. In another systematic review 30 concerning elder abuse, most of the primary studies had been conducted in developed countries, corroborating this finding.
Few scientific studies of elder abuse are based on strong evidence and most of the studies in-cluded were of cross-sectional design. This indicates the importance of the issue of abuse in academic and social circles. However, there is a need for studies with a higher level of evidence, since cross-sectional studies provide weaker evidence than those using other methodologies.
The studies used a variety of instruments to measure abuse and this hindered comparison of their findings. Most of the studies included reported flaws and the need for instruments adequate for each context. In this systematic review, the most common instrument was the Review Conflicts Tactics Scale (CTS)/(CTS2). This finding corroborates that of Espíndola & Blay 29 .
The CTS was not created specifically for the elderly population but is the oldest instrument, dating back to 1979, and also the most widely used. This can be explained by the fact that it meets the criteria of validity and reliability. CTS2 is a more up-to-date version that aims to correct some of the flaws in the original version 31 .
Of the factors associated with elder abuse, having a "lower income" was considered a highly significant risk factor for all types of abuse, general or specific. This indicates that the fewer financial resources the elderly have access to, the more vulnerable they are abuse.
Clusters of low-income or unemployed people tend to generate higher rates of housing instability, resulting in a deficit regarding the creation of common values and norms among individuals and the development of strong social bonds and support networks. Oversight is also compromised and this propagates conditions in which abuse can flourish, such as increased social marginalization and poor physical and mental health 6 .
Having a diagnosis of depression was also found to be a risk factor for almost all types of abuse. However, most studies were cross-sectional and thus did shed light on the direction of causality. There is, therefore, no way of inferring whether the elderly are more subject to violence because they are depressed or whether they are depressed as a consequence of the abuse.
Most studies have shown that elderly individuals who are women are more likely to experience general abuse. Illnesses prevalent in women, however, could be explained by a life-expectancy bias in cross-sectional studies, which do not take into account the higher life-expectancy of women and thereby overestimate the prevalence of abuse.
However, global estimates indicate that 30% of older women who have had a partner have been victims of physical and/or sexual violence at some point in their lives, with variations according to regions around the world. In Africa, the Eastern Mediterranean and Southeast Asia, approximately 37% of women have been abused by an intimate partner, followed by the Americas, where approximately 30% of women report some kind of life-threatening violence 6 .
Longevity in the elderly presented a heightened risk for self-neglect but constituted a protection against negligence, verbal, psychological, financial and physical abuse. This may be due to the difficulty reporting violence among such individuals, since the perpetrators are usually caregivers or people close to them. When considering the "age" factor, it should therefore also be borne in mind how much harder it is for an elderly person to report abuse, given all the natural limitations of age, not to mention dementia or associated physical and psychological incapacities. Self-neglect may therefore mask negligence.
Neglect, defined as a refusal or failure on the part of the caregiver to provide the necessary care, may be domestic or institutional and may generate physical, emotional, and social harm and trauma. People neglect the elderly because they feel they do not need care because, unlike children and adolescents, they are not developing psychologically and physiologically. However, this population has limitations inherent to the aging process, such poor hygiene and impaired functional and sensorial capacities 32 .
With regard to family relationships, "living with the family of a married child," "living alone or with children," "households with six or more residents," "having less family support," and "expressing frequent incidents of conflict with family or friends" were all risk factors for general abuse. "Living with other family members" was a risk factor for negligence. "Living with the family of a married child" and "living with a single child" were risk factors for financial abuse, while a "very bad family relationship" was a risk factor for neglect, and general, verbal, psychological and economic abuse.
According to the World Report on Violence and Health 3 , the elderly may be at greater risk of abuse when living with the people who care for them, as a result of lack of privacy for both or overcrowding in the home. This may generate conflicts within the family and older people with dementia can behave violently towards caregivers, thereby provoking violence in return.
The work overload that old age can entail, especially when the individual is dependent on others for activities of daily living (ADL) and instrumental activities of daily living (IADL) can also generate a higher rate of violence. This was seen in the present study, in which greater dependence on others for ADL and IADL were risk factors for general abuse and having difficulty with IADL was also a risk factor for financial abuse.
There is an interdependence between abuser and abused and the elderly are at greater risk of abuse when they are more dependent on others for daily activities, while aggressors , most of them young people, are generally more dependent on the elderly for housing and financial assistance, thereby creating a greater risk of abuse. Elderly people may be isolated due to physical or mental illnesses, as well as the loss of friends and family. This decreases the chances of social interaction 3 .
Social isolation may be a risk factor for abuse, as was seen in the present study in several variables that were found to increase the likelihood of general abuse. Social isolation can however also be construed as a consequence of abuse, as the elderly may be overwhelmed by the violence practiced against them and isolate themselves.
In the present systematic review, social factors were significant as well as those relating to health. This makes it clear that existing individual risks need solutions at the macro level, such as investment in improved living conditions, more social equality, better health and education for all, better quality of life and reduced prevalence of all types of elder abuse worldwide.
Elder abuse is a multifactorial phenomenon, with high complexity. It is complex, because distinct (economic, political, sociological, psychological, affective, mythological) components are inseparable constituents of the whole, and there is interdependence between abuse and the con-text in which it occurs, between the whole and the parts, and among the parts themselves 34 .
One limitation of this review was the paucity of studies dealing exclusively with abuse in the elderly. It nevertheless represents an important contribution to the study of violence against this population.
conclusion Elder abuse is a multifactorial phenomenon. In the present review, the following risk factors were identified: age, sex, marital status, educational level, income, family arrangement, family relationship, social support, solitude, mental disorder, depression, dependence on others for ADL and IADL, and others.
For this reason, abuse should not be seen in a partial, one-dimensional way. Its various dimensions should all be taken into account and the interdependence of all the associated factors should be recognized.

collaborations
MAB Santos is a principal researcher, participated in the preparation of the research, selection and analysis of data, writing of the article, submission and procedures of the article. RS Moreira participated in the guidance, preparation and correction of the writing of the article and approval of the final version. PF Faccio participated in the selection and analysis of the data. GC Gomes participated in the preparation and correction of the writing of the article. VL Silva participated in the orientation, preparation of the research, data analysis, preparation and correction of the writing of the article and approval of the final version.