Introduction

Elder abuse and neglect (EAN) is increasingly becoming a global health concern. The World Health Organization (WHO) has defined EAN as ‘a single or repeated act, or lack of appropriate action, occurring within any relationship where there is an expectation of trust which causes harm or distress to an older person’[1]. There are five subtypes of elder abuse–physical, psychological, financial, sexual and neglect [2]. Worldwide prevalence of EAN was estimated at 3.2% to 27.5% [3]. In Malaysia, one study reported a 12-month prevalence of 9.6% among urban elders [4] while another reported a lifetime prevalence of 8.1% among those in the rural area [5].

Numerous adverse effects of abuse in late life–across different health domains—have been documented. These include premature mortality [6, 7], depression [8], increased utilization of health services [9, 10], metabolic syndrome [11], musculoskeletal pain, suicidal ideation, anxiety, incontinence and gastro-intestinal symptoms [12]. Nevertheless, many health-related outcomes of EAN remain unexplored. A conceptual framework by Anetzberger on the health effects of elder abuse has categorized them as physical, psychological, behavioural and social [13].

Quality sleep is one of the key determinants of good health. Its importance is more prominent among older adults, as sleep disorders and sleeping difficulties become more pervasive with aging [14]. Poor sleep has been found not only to predict mortality [15, 16] and nursing home placement among older individuals [16], but it was also associated with obesity [17], cognitive decline [18] and diabetes [19]. Despite this, the relationship between abuse experience in late life and sleep quality has not been adequately explored. Olofsson et al reported that older Swedish victims of physical and psychological abuse were more likely to experience sleeping problems than those who were not abused [12]. Other than this, there has been no investigation to corroborate this finding and whether or not it is replicable across different populations remains a question. The issue of cross-cultural differences in the manifestations of EAN outcomes has been raised [20]. Culture is said to influence the expression of bodily symptoms [21], sleep habits and subjective appraisal of sleep [22]. As argued by Stranges et al, “Sleep habits are multifaceted and result from a complex interplay between genetics, environment, and social factors, as well as the presence of comorbidities. Factors contributing to sleep problems in older adults from low income countries may differ from those characteristic of Western societies”[22].

Sleep can be measured either objectively or subjectively. Objective measurement of sleep often employs polysomnography (PSG) and the Multiple Sleep Latency Test (MSLT), but these methods have been considered impractical for clinical screening and research [23]. The more common approach in research therefore has been the subjective measurement using questionnaires–the most widely used being the Pittsburgh Sleep Quality Index (PSQI) and Epworth Sleepiness Scale (ESS) [23]. In this study, we have employed the PSQI for its compatibility with our study objectives, that is to measure the overall sleep quality among those abused and not abused, and discriminate between ‘good’ and ‘poor’ sleepers [24, 25]. The ESS on the other hand has been designed more specifically for assessing daytime sleepiness [25, 26].

The PSQI is a nineteen-item questionnaire measuring seven components of sleep–subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction–yielding one global score [24]. Most studies posit a single factor structure of PSQI comprising the seven components scores. However, some studies have posited a two-factor model and a three-factor model [27, 28]. The two-factor model consists of sub-scales sleep efficiency (components three and four), and perceived sleep quality (the remaining components). The three-factor model consists of sub-scales sleep efficiency (components three and four), perceived sleep quality (components one, two and six), and daily disturbances (components five and seven). PSQI has been translated into various languages and is widely used in research studies [29]. Cross-cultural validation has also been conducted among Brazilian, Israeli, Nigerian, Italian, Korean and Arab samples [29–33]. The Malay version of PSQI was used in a number of studies [34–37], but validation was never attempted. In the current study all these models were explored to identify the most suitable factor structure for the Malaysian context.

We thus aim to: 1) validate the Malay version of PSQI among older Malaysians through confirmatory factor analyses (CFA) and reliability tests, and; 2) examine the association between EAN and sleep quality.

Methods

This study consisted of two phases. In Phase 1, we validated the Malay version of PSQI and in Phase 2, we examined the relationship between EAN and sleep quality. Assessment was carried out in a span of six months, from November 2013 to May 2014.

Phase 1: Validation of Malay PSQI.

Eight low-cost, subsidized government flats were randomly chosen from a list of all low-cost flats in the Klang Valley. A total of 239 residents aged 60 or more were then randomly selected and invited to join the study. The Malay version of PSQI was administered face-to-face by trained personnel two times, in a two-week interval.

The PSQI is a self-rated questionnaire comprising nineteen questions with an additional five questions rated by another person such as bed-partners or roommates. The five questions were not administered since they are usually used for clinical purposes and not included in the scoring system. According to the guidelines of Buysse et al. [24] the nineteen items were transformed into seven components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, use of sleeping medication, and daytime dysfunction. Each component score ranged from 0 to 3, where 0 indicated “no difficulty” and 3 indicated “severe difficulty”. All components were then summed to yield a global PSQI scores ranging from 0 to 21. Higher scores indicated poorer sleep. In our study, the validated Malay version retained three components (consisting of twelve items) and therefore had a score range of 0 to 9.

Phase 2: EAN and sleep quality.

Results

Phase 1: Validation of Malay PSQI.

Out of 239 older adults invited, 194 agreed to participate in the validation study. Response rate was 81.2%. Those found to have severe depression (GDS scores 12 or more) and cognitive impairment (MMSE scores 17 or less) were excluded. The final sample size was 183. Descriptive statistics of all PSQI components are shown in Appendix 1 (S1 Appendix). CFA was then run on the original models (one factor, two factor, three factor) consisting of all the seven components. Appendix 2 (S2 Appendix) demonstrates CFA model fit statistics, item loadings, and reliability values for the different PSQI models.

The overall model fit indices for the original models (one factor, two factor, three factor) were acceptable but the item loadings were poor. Components 3,4,6 and 7 were then dropped one by one and finally PSQI with three components (subjective sleep quality, sleep latency and sleep disturbances) was found appropriate, demonstrating good convergent validity. The final scale had acceptable Cronbach’s alpha and composite reliability, and significant ICC of 0.61 (S3 Appendix). The final Malay validated PSQI model is presented in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The final CFA model of the Malay validated PSQI.

https://doi.org/10.1371/journal.pone.0180222.g001

Phase II: EAN and sleep quality.

A total of 2118 individuals agreed to participate. Response rate was 84.9%. There was no difference between respondents and non-respondents with regards to socio-demographic characteristics. Screening of cognitive impairment using MMSE excluded 191 individuals, and 279 did not show up on the day of PSQI interview (unit non-response), yielding a final sample size of 1648. Those who did not undergo PSQI interview were older, while other characteristics did not differ. The sample comprised of 97.4% Malay, 60.2% women, and slightly more than half belonged to the younger age group (60–69). Missing data was found to be MAR (missing at random), and this was addressed by multiple imputation method.

The prevalence of EAN overall was 8.1% (n = 133). Financial abuse was the most common subtype, comprising 4.9% while psychological, physical, neglect and sexual abuse were 3.3%, 1.2%, 1.2% and 0.2% respectively. Unadjusted analyses showed that abuse had significant relationships with low income, depression and poor sleep quality. Mean PSQI scores for EAN victims were 3.80 (SD = 1.55), whereas for those not abused, mean scores were 3.41 (SD = 1.38). Table 1 illustrates the basic characteristics of our study respondents.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Basic characteristics of study respondents (N = 1648).

https://doi.org/10.1371/journal.pone.0180222.t001

Mean PSQI scores were highest in neglect (4.11, SD = 1.7), followed by physical (4.10, SD = 1.1), psychological (3.96, SD = 1.7), and financial abuse (3.60, SD = 1.4). Sexual abuse had to be excluded from analysis due to the extremely small number of cases (n = 5). A dose-response relationship was evident between clustering of abuse and PSQI scores; 3.41 (SD = 1.3), 3.50 (SD = 1.4) and 3.84 (SD = 1.3) for “no abuse”, “1 type of abuse” and “2 types or more”, respectively. Fig 2 demonstrates the graded relationship between clustering of abuse and PSQI scores.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Dose-response relationship between number of abuse and PSQI scores.

https://doi.org/10.1371/journal.pone.0180222.g002

Analyses using Generalized Linear Models were then performed to determine the relationship between independent variables and PSQI scores. There was no multi-collinearity between the any of the predictor variables. Six were found to be significantly associated with poor sleep (higher PSQI scores); 1) EAN–abuse victims were more likely to have poorer sleep than those not abused; 2) co-morbidities–those with multiple chronic diseases were more likely to have poor sleep than those with no or few diseases; 3) self-rated health; those who rated their health as “good” were more likely to have better sleep than those who rated their health as “poor”; 4) income; those with lower income had poorer sleep; 5) social support; those with lower social support were more likely to have higher PSQI scores, and; 6) gait speed; those with lower gait speed had poorer sleep. No interaction was found between abuse and other variables in affecting sleep quality. When similar analyses were run with overall EAN being replaced by the four subtypes–physical, psychological, financial and neglect–only neglect was associated significantly with decline in sleep quality.

Table 2 shows results of GLiM analyses to determine the associations between independent variables and PSQI scores, while Table 3 shows the associations between EAN subtypes and PSQI scores.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Relationship between abuse, socio -demographics, physical function, health-related and psychosocial variables with PSQI scores.

https://doi.org/10.1371/journal.pone.0180222.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Relationship between EAN subtypes and PSQI scores.

https://doi.org/10.1371/journal.pone.0180222.t003

Discussion

Our study showed that, overall, respondents with abuse experiences had significantly poorer sleep than those who were not abused. Unadjusted analyses showed associations between EAN and low income and depression (Table 1). This corroborated previous findings which reported low income and psychological or psychiatric problems as risk factors for elder abuse [43]. Besides, depression was also an established mental health outcome of EAN [44–46].

The graded relationship between clustering of abuse and PSQI scores strengthened the association between EAN and poor sleep–as abuse experience is multiplied, sleep quality declines further as demonstrated by the steady increase in PSQI scores. This can be explained by the link between stress–with EAN being a form of stressor–and sleep problems as reported by a number of studies: 1) high levels of emotional arousal as a result of adverse events were said to cause sleep difficulties [47]; 2) exposure to stress increased the risk of insomnia [48], and; 3) sleep recordings demonstrated that stress was associated with “shortened sleep, fragmentation, and possibly a reduction in sleep stages 3 and 4”, which then led to higher levels of cortisol, thus exacerbating stress [49]. Other forms of family violence–child abuse and intimate partner violence (IPV)–have all been shown to be associated with sleep disturbances [50, 51].

Among the subtypes of abuse, neglect had the strongest association with poor sleep. Similarly, mean PSQI scores were highest in neglect. The operational definition of neglect in this study included: 1) not being provided with support for basic activities of daily living (cooking, bathing, dressing, walking, climbing stairs, etc), and; 2) being deprived of safe and clean living conditions, medical attention or treatment, clean clothing, and adequate, nutritional food [38]. The dynamics between neglect and sleep problems may be explained by different pathways. First, a direct pathway–lack of access to proper shelter could indicate that the victim is made to live in a poor condition that prevented him from obtaining adequate, comfortable sleep. Second, an indirect pathway–deprivation of medical attention and nutritional food could result in various health problems or aggravate existing illnesses. As a consequence, sleep difficulties arise. Studies have shown that poor sleep among older individuals was associated with co-morbidities and health status [52, 53]. Similarly, limitations in activities of daily living–a component of neglect–was related to sleep disturbance [53]. Third, the overall effects of neglect can be greater than other individual subtypes of abuse, due to its multi-dimensional nature. Some components of neglect overlap with psychological and financial abuse; neglect victims inevitably suffer, in a subtle manner, from psychological distress and constraint resources. Fourth, neglect is chronic in nature–established by at least ten occurrences in our study [38]. Unlike other subtypes such as financial or physical abuse where a single episode of mistreatment was considered abuse, neglect victims tend to suffer longer, thus giving enough time for changes in sleep to manifest.

Other variables found to be significantly associated with sleep quality were self-rated health, co-morbidities, gait speed, income and social support. These findings have been substantiated by previous studies; older adults who rated their health positively, or had no or less co-morbidities reported better sleep in comparison with those with poor self-rated health and more chronic diseases [53, 54]. Low income had also been linked to poor sleep quality [55]. Similarly, physical function–represented by gait speed in this study–was associated with sleep difficulties. Dam et al reported a relationship between lower gait speed and greater sleep fragmentation, poorer sleep efficiency and sleep hypoxia among older American men [56], while Goldman et al found that older women with poor physical performance (including gait speed) had greater sleep disturbances [57]. Sleep quality among older adults was also said to be influenced by their social support and network [58, 59].

On the other hand, depression was not associated with poor sleep in our study, in contrast with prior findings which suggested a relationship between the two [60, 61]. However, a study on subjective sleep quality characteristics among older adults found that of all PSQI components, usage of sleeping medication was the only one associated with depression [62]. This component (use of sleeping medication) was among the four components dropped in our validated Malay PSQI. The relationship between depression and poor sleep in later life therefore may be more complex than the conventional assumption, thus warranting further investigation.

Given that our findings supported an association between EAN and poor sleep, current approach to managing abuse victims needs to be tailored accordingly. Healthcare providers should always be alerted to the possibility of sleeping problems when dealing with EAN victims, especially when there are elements of neglect. On top of that, detection of co-existing multiple chronic diseases, poor self-rated health and low social support–all of which can be easily identified–may be a simple yet effective technique to screen for sleeping problems among abuse victims, prior to confirmation with more sophisticated methods. Improvement of sleep quality therefore should be one of the aims while managing EAN victims. This can be done either through designation of specific treatment modalities by clinicians or a more general form of intervention by public health experts and policy-makers.

This study has several limitations. First, its cross-sectional design disallows causality. The possibility of the relationship between EAN and poor sleep in the opposite direction cannot be ruled out. Second, PSQI is a subjective measurement of sleep. Its accuracy may differ from objective methods such as polysomnography. It was reported that differences in results were found between the two methods, and that the correlation between self-reported and objectively-measured sleep could be biased [63, 64]. Third, our study excluded older adults with severe cognitive impairment, severe hearing impairment and inability to communicate independently. In addition, those who did not turn up for sleep assessment (through PSQI questionnaire) comprised the older group. This could have underestimated the prevalence of sleep disturbance and strength of association between abuse and sleep quality. Fourth, validation of the Malay PSQI was conducted among low-cost flat residents. This could have resulted in potential bias, as those from the higher income group were not included. However, the majority of our rural respondents in Phase II were classified as ‘low-income’ based on their monthly household earning, while the high-income group constituted only 3.5%. Nevertheless, our results need to be interpreted in the light of these constraints.

Supporting information

Acknowledgments

This study was funded by the University of Malaya Grand Challenge on Preventing Elder Abuse and Neglect Initiative (PEACE) (GC001D-14 HTM) and Postgraduate Research Grant (PG104-2014A). We would like to thank all members of Julius Centre University of Malaya (JCUM) and the Health State Department of Negeri Sembilan for their administrative support throughout the research period.