In this cross-sectional study, our target population consisted of 1980 patients undergoing haemodialysis at 22 dialysis centers run by the National Kidney Foundation in Singapore (NKFS). The National Kidney Foundation provides subsidized haemodialysis to needy patients. Subsidy is offered to those who are unable to afford haemodialysis, as determined by financial assessment through a means test. Most of the patients of NKFS are of a lower socio-economic status. The patients were of Chinese, Malay and Indian ethnicity.
Participation in this study was voluntary and data was gathered from December 2006 through January 2007. For inclusion, patients had to be at least 21 years of age, have ESRD, and have been receiving haemodialysis (not peritoneal dialysis) at the National Kidney Foundation dialysis center for more than three months. In Singapore, the majority of patients are on haemodialysis (79% haemodialysis vs. 21% peritoneal dialysis) .
Trained nurses explained the study to the patients. Patients who volunteered to participate were recruited into the study. Written consent was obtained from participants and confidentiality of data was assured before the data was gathered. This study was approved by the Institutional Review Board of Singapore General Hospital.
The disease-specific instrument used in this study was the Kidney Disease Quality Of Life-Short Form (KDQOL-SF™) version 1.3, a self-report measure developed for individuals who have kidney disease and are on dialysis . The KDQOL-SF™ is available in English and was translated into Mandarin Chinese and Malay (Singapore version) by the KDQOL-SF™ group and RAND [15, 16]. The English version of the KDQOL-SF™ was used in surveying the Indian population, who mostly understood English. In this survey, very few participants (less than 10) completed the Chinese or Malay versions of the survey forms. In addition to providing translated versions of the KDQOL-SF™, the study provided trained nurses conversant in Chinese, Malay, Tamil and English to answer any queries from the participants.
The KDQOL-SF™ includes multi-item scales targeted at the particular health-related concerns of individuals who have kidney disease and are on dialysis. The instrument is composed of 36 general health items and 43 kidney-specific items. The items on general health are divided mainly between physical and mental health across eight sub-scales, with one item on overall health. The eight sub-scales are: Physical functioning, Role physical, Pain, General health, Emotional well-being, Role emotional, Social function and Energy/fatigue. Scoring algorithms given in the user manual  were used to calculate scores ranging from 0 to 100. The scores represent the percentage of total possible score achieved, with 100 representing the highest quality of life. The items ask about the patient's health and how the patient feels about his care. Items gather information regarding the patient's background such as gender, ethnicity, education, income, the number of days in their hospital stay, and the number of different prescription medications they were taking. This information is used to evaluate the care delivered and to enable a greater understanding of the effects of medical care on the health of patients . The KDQOL-SF™ was self-administered.
Treatment of Missing Data
Of the 1180 participants who completed the survey, 980 provided age, gender and race information, and this data was used in the analyses. Of these, 1.6% missed marking one item and 1.4% missed marking two items. Missing data for an item was substituted with a figure calculated by averaging the scores of the other items in the particular scale to which the missing item belonged.
The analysis was carried out using SPSS version 15 software. We first compared the sample demographic data with demographic data from the dialysis population listed in the Singapore Renal Registry, 2004  to determine whether the sample was representative of the full dialysis population in Singapore. We used Analysis of Variance (ANOVA) and the t-test to examine the differences.
We then used exploratory factor analysis to determine the basic structure of the KDQOL-SF™. This technique can be used to group independent latent variables (those which cannot be measured directly: i.e., subjective) into categories based on similar characteristics or behavior. We explored the unknown domains of the KDQOL-SF™ scores by dividing the characteristics/items into independent sources of variation (factors). Here we used a deductive approach by hypothesizing the existence of particular dimensions and assessing whether our data fit a factor structure identical to the structure found by previous researchers  (i.e., how well the measure represented the construct of interest [construct validity]). For selecting the number of factors, we used the criteria of the factor having an eigen value (which measures the amount of variation) greater than one. Varimax rotation (orthogonal rotation of quadrants) was used to control for certain influences (of items on the sub-scale) on the overall result. The rotated factors delineate a distinct cluster of relationships, while unrotated factors successively define the most general patterns of relationships in the data.
We used Cronbach's coefficient α to assess internal consistency reliability for the overall scale, and within individual sub-scales. Correlation coefficients were calculated to assess the strength of relationship between items within and outside each sub-scale. We also determined the mean and median of each sub-scale. We used Pearson Correlation (two tailed) to assess stronger relationships of items within scales and weaker relationships with items outside of the scale. We looked at the correlations between the overall health score and the Kidney disease-targeted scales of Symptoms, Effect of kidney disease, Burden of kidney disease, Work status, Cognitive function, Quality of social interaction, Sexual function, Sleep, Social support, Dialysis staff encouragement, and Patient satisfaction.
We also looked at two-tailed significance of correlation coefficients between scores on the eight sub-scales and age, income, and education to determine convergent and divergent validity. Considering that higher scores on the SF-36 scales indicate good quality of life, we hypothesized that the KDQOL-SF™ total score would be positively correlated with measures of self-rated health, and of socioeconomic status - represented by educational status. We expected the duration of dialysis to be positively correlated with health.