Quality of life and related factors of nursing home residents in Singapore

The study was approved by the Institutional Review Board of National University of Singapore (NUS) and conducted during 31 January 2012 to 8 February 2012. Participants were recruited from six local NHs run by VWOs; attempts to recruit residents from private NHs were not successful because no private NHs agreed to conduct the study in their premises. The inclusion criteria for the subjects were: 1) age ≥55 years; 2) residence in the NH for at least 3 months; 3) able to communicate views and opinions coherently; and 4) able to give informed consent to participate. NH residents who were uncommunicative or unable to respond meaningfully due to physical or cognitive impairment were excluded. A list of potential eligible NH residents was provided by a nurse manager from each NH to form the sampling frame. The residents’ ability to communicate was assessed using three screening questions: “What is your name?”; “What is your age?”; and “Where are you now?” If a resident answered irrelevantly to any of the questions (e.g. responded ‘very good’ to ‘What is your age?’) despite asking thrice, the resident would be assessed to be unable to communicate meaningfully and excluded. All eligible residents were informed both verbally and in writing, and informed consent was obtained from all participants.

Consenting NH residents were interviewed face-to-face by 16 pairs of NUS medical students. They had been trained to strictly adhere to the phrasing of the questions, to avoid excessive prompting of the residents, and to pay attention to non-verbal cues that may signify distress or withdrawal of consent, especially in residents with dementia who may not be able to express themselves adequately.

A questionnaire with various rating scales and questions was administered to the NH residents, with each interview session lasting between 15–30 min. The interviews were conducted in English, Mandarin, Malay, or one of the two main local dialects Hokkien and Cantonese. Multi-lingual translators were employed if the primary interviewer was not conversant with the participant’s language or dialect. The NH residents’ demographic (e.g. age, education) and clinical information (e.g. history of medical conditions, current medications) was retrieved from their medical records.

The rating scales used included the modified Minnesota questionnaire for assessing QOL, the Abbreviated Mental Test (AMT) for cognitive function, the Structural Clinical Interview (SCID) for depression, and the Katz Index of Independence in ADL.

The QOL scale used was designed based on the University of Minnesota study [20] which published a 66-item questionnaire examining 11 domains of QOL including comfort, security, functional competence, relationship, enjoyment, meaningful activities, dignity, privacy, autonomy, and spiritual well-being. The questionnaire has been well-validated for NH residents and has an abbreviated and validated version containing 34-item in 9 domains [20]. The abbreviated version was chosen to be adapted and tested in a pilot study, which was conducted on a total of 19 residents selected by the nursing directors of four NHs. The pilot study aimed to identify potential difficulties in questionnaire administration and distinguish the top 5 domains that were most relevant to local NH residents. This would help to shorten the length of the questionnaire and minimize burden to the residents while maintaining the adequacy of questionnaire for application to the local residents. The domains identified were security (79.6 %), dignity (77.8 %), food enjoyment (71.9 %), autonomy (70.2 %), and comfort (68.4 %). All the 18 items in the 5 domains (each domain has 3 or 4 items) were extracted from the questionnaire to generate a new questionnaire. Moreover, the items were revised based on the pilot study to minimize ambiguity and improve relevance to the local context. For example, ‘How often are you too cold here?’ was modified to ‘During your stay here, do you feel too hot or too cold?’ to better reflect the tropical climate in Singapore. The response scale was also modified from a 4-point Likert format to a 2-point format (1=‘mostly yes’, 2=‘mostly no’), both of which were recommended [20]. Meanwhile, a ‘not applicable’ option was also added to the response scale as some questions were found to be irrelevant for several NH residents in the pilot study.

The AMT is a 10-item scale for measuring the residents’ cognitive function [21]. It has a total score of 10 and the cut-off was set at the clinically validated score of 6/10. NH residents who have a score of ≤6 were assessed as having cognitive impairment.

The SCID which uses the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) criteria was chosen as the tool to assess depression in the study [22]. It is the gold standard diagnostic tool for measuring depression in the clinical setting and consists of 9 questions evaluating the presence of symptoms associated with depression. The cut-off value for major depression was set at ≥5/9. Minor depression was defined as the presence of 2 to 4 depressive symptoms.

The Katz Index of Independence in ADL was used to assess functional limitations of ADL in 6 aspects including bathing, dressing, toileting, transferring, continence, and feeding [23]. It adopted a dichotomous format, with 1 point for each ADL if the NH resident was independent (defined as requiring no assistance from the NH staff), and 0 point if the NH resident required any form of assistance. The cut-off value was determined at 3/6: those who were dependent in 0–3 ADLs and those who were independent in 4–6 ADLs. The cut-off value is also consistent with the definition of disability used in Eldershield which is a local insurance scheme providing basic financial protection for persons who need long term care in Singapore.

The QOL scale, AMT, SCID, and Katz were translated into Mandarin using forward-translation and back-translation which were done independently by different persons to ensure accuracy and appropriateness of the wording.

The questionnaire also included questions to determine whether the NH residents suffered from a lack of social contact, had a history of falls or being restrained, and had experienced difficulty in communication with staff in the NH.

Cronbach’s alpha, a measure of internal consistency, was calculated to assess the reliability of the five QOL scales as well as the AMT and Katz index of independence in ADL. A Cronbach’s alpha value ≥ 0.7 indicates a satisfactory level of reliability.

To identify predictors of NH residents’ QOL, 16 variables were examined. All of them were coded as categorical variables including age (<80 years vs. > = 80 years); gender; race (Chinese vs. non-Chinese); marital status (no partner vs. with partner); education (no or low education vs. higher education [secondary school and above]); religion (religion vs. no religion); length of stay (<=2 years vs. >2 years); depression (no depression, minor depression, and major depression), number of comorbidities (> = 3 vs. <3); cognitive functioning (AMT < =6 vs. AMT >6); functional impairment (ADL score < =3 vs. ADL score >3); social contact (no or lack of social contact vs. with social contact [one visitor visit at least once a month]); history of being restrained since admitted into NH (with vs. without); history of falls since admitted into NH (with vs. without); communication with staff (no difficulty vs. difficulty); and study sites (the six NHs). The majority was identified as predictors of QOL of NH residents previously. Sociodemographic factors such as education, marital status, and religion were also examined in prior studies [15, 16]; race was added in our study since Singapore is a multiethnic country. As NH staff tend to be more cautious with residents with a history of being restrained or falls and thus afford them less autonomy, the two factors were also examined. Study sites were explored as evidence has suggested that residents’ satisfaction with the NH was a significant predictor of high QOL [10].

The relationship between NH residents’ QOL and the variables was explored in each of the five domains of the QOL scale. Five multivariate regression models were used to identify factors significantly associated with the five domain scores, respectively. The domain scores were calculated by aggregating the item scores within the domain, ranging from 3 to 6 or 4 to 8. Reverse scoring was applied whenever appropriate to make higher scores consistently indicating better QOL. As the QOL data exhibited ceiling effects and heteroscedasticity, we adopted the censored least absolute deviations (CLAD) estimator [24] to estimate the coefficients of the regression model. The CLAD model does not require distributional assumptions or homoscedasticity assumption of residuals, and is robust to censoring [24]. Empirical evidence has suggested that it outperforms the OLS model and the Tobit model in the face of non-normality, heteroscedasticity, and censoring [25‐27].