The effect of type 2 Diabetes Mellitus on health-related quality of life ( H R Q O L )

Improving the quality of life of all South Africans has become a major concern to health care practitioners, or­ ganisations and politicians. However, the paucity of local information on health-related quality of life (HRQOL) does not allow us to address this public health challenge. In order to rectify this deficiency and complement interna­ tional research, we undertook a study with 281 Type 2 Black diabetic patients and 437 controls, with no self-reported chronic conditions, to ascertain HRQOL. We used the SF-20 to measure functioning, general health, well­ being and bodily pain (HRQOL). It was hypothesised that diabetes mellitus significantly affects functioning, general health and well-being. Multiple analyses of covariance controlled for age, schooling, marital status, employment status and commodity ownership (a socio-economic meas­ ure). Patients were significantly more likely to report poorer role functioning, poorer general health and more pain than controls, providing partial support for the hy­ pothesis. Reliability (internal consistency) coefficients on the four multi-item SF-20 sub-scales ranged between 0.79 (well-being), 0.81 (general health), 0.83 (physical func­ tioning) and 0.94 (role functioning) for patients; for con­ trols these coefficients ranged between 0.70 (well-being), 0.78 (general health), 0.80 (physical functioning) and 0.90 (role functioning). Inter-correlations among the sub-scales were significant for patients and controls (p = 0.01). It was concluded that the SF-20 is a reliable instrument for measuring HRQOL in both patient and control samples, and diabetes mellitus has more impact on general health and level of pain than on well-being.


Introduction
World-wide, over 140 million people suffer from diabetes mellitus, with a projected increase to 300 million by the year 2025 (Preston, 1998).Type 2 (non-insulin-dependent) diabetes mellitus is the most common form of the disease in adults over 40 years, accounting for over 85% of all cases (Barceló, 1996).In South Africa, it has been "estimated that there are at least 1 million known diabetics and possibly up to an equal number who are currently undiagnosed" (Bonnici, Hough and Huddle, 1997, p. 440).South African studies in urban areas have re ported prevalence rates for black (African) persons of between 5% and 8% (Levitt, Katzenellenbogen. Bradshaw, Hoffman and Bonnici, 1993;Mollentze, Moore, Oosthuizen, Steyn, Steyn, Joubert, Muller and Weich, 1992;Omar, Seedat, Motala, Dyer and Becker, 1993), equating to approximately 1 million cases; a major public health challenge for South Africa.Diabetes mellitus has implications for physical and social func tioning, general health and well-being.Diabetes mellitus pa tients have been shown to be more susceptible to depression (Gavard, Lustman and Clouse, 1993), have poorer social inte gration (Lloyd, Robinson, Andrews, Elston and Fuller, 1993), and poorer functioning and general health than the general population (Stewart, Greenfield, Hays, Wells, Rogers, Berry, McGlynn and Ware, 1989).However, the relationship between diabetes mellitus and well-being is not a simple one.Studies in the United States (US) and Finland have shown no substantial differences between diabetic patients, patients with no chronic conditions, the general population and controls on well-being (Aalto, Uutela and Kangas, 1996;Stewart, Greenfield, Hays, Wells, Rogers, Berry, McGlynn and Ware, 1989).It is possible that the impact of the disease on well-being is mediated by diabetic complications.While two studies have reported that better glycaemic control was associated with better function ing and w ell-being (Nerenz, Repasky, W hitehouse and Kahkonen, 1992; Van der Does, de Neethling, Snoek, Kostense, Grootenhuis, Bouter and Heine, 1996), one study found no association betw een these two outcom es (W einberger, Kirkman, Samsa, Cowper, Shortliffe, Simel and Feussner, 1994).In South Africa, little attention has been paid to systematically assessing the functioning, general health and well-being of diabetic patients (health-related quality of life), despite improved quality of life being one of the treatment objectives (Bonnici, Hough and Huddle, 1997).Moreover, no local studies have compared the health-related quality of life (HRQOL) of diabetic patients and persons with no self-reported chronic conditions (controls).In order to rectify this deficiency and complement international research, we undertook a study to ascertain HRQOL in black diabetic patients and controls, with the hy pothesis that diabetes mellitus significantly affects function ing, general health and well-being (HRQOL).

M aterials and methods
The Questionnaire A structured questionnaire, with a consent form, was designed to obtain information on: demographic variables (gender, age, schooling, marital status and employment status); socio-economic variables (commodity ownership index); and health-re lated quality of life (HRQOL).The 20-item abbreviation of the Rand Medical Outcomes Study (SF-20) was used to measure HRQOL (Stewart, Hays and Ware, 1988).The 20 items form six scales: physical functioning, role functioning, social functioning, mental health, general health and bodily pain.The functioning sub-scales assess capacity to perform physical activities (walking, lifting, climbing), role limitations because of physical health problems, and social activity limitations due to health problems.Mental health is assessed by a 5-item affective measure of depression, positive affect, anxiety and psychological well-being.General health is measured by overall ratings of current health.The pain sub scale is defined in terms of the extent of bodily pain.Overall health status is a combination of the six health measures.Cut-off points for defining poor health have been developed for each of the scales.Poor physical and role functioning are defined as one or more limitations; poor social functioning as limitations a good bit of the time or more; poor mental health as a score of 67 or lower; poor general health as a score of 70 or lower; while for pain the cut-off point lies between mild and moderate pain (Stewart, Hays and Ware, 1988, p. 728).Scores on the six health measures and overall health status are trans formed linearly to 0-100, where 0 and 100 are assigned to the lowest and highest possible scores, respectively.With the exception of pain, high scores denote better functioning, gen eral health and mental health.High scores on the pain sub scale denote more pain.Reliability (internal consistency) coefficients for the four multi item scales ranged between 0.76 and 0.88 (general population); and between 0.83 and 0.87 (diabetic patients).Diabetes mellitus patients were significantly more likely (p <0.01) to rate their functioning and general health as poorer than patients with no chronic conditions and the general population (Stewart, Greenfield, Hays, Wells, Rogers, Berry, McGlynn and Ware, 1989).Scores on the sub-scales were associated with the de mographic characteristics of age, gender, education and in come.With the exception of mental health, older persons re ported poorer functioning and general health (p < 0.01) than younger persons; men reported better functioning than women (p < 0.01); and persons with more education and income tended to have better functioning and general health (p < 0.01) than persons with low education and income levels (Stewart, Hays and Ware, 1988).

Procedure
All Black patients attending a Diabetic Outpatients Clinic in Mamelodi for routine examination and treatment were asked to participate in the study.Only three patients refused to partici pate in the study.A multi-lingual trained Black female research assistant was employed to administer the questionnaire to the patients.In 1995,1,653 out of 6,933 households in an informal settlement in Soweto were visited to provide baseline data on health aware ness, health status and health needs of the community.Out of these 1,653 households, 500 were randomly selected to pro vide the comparison sample with the diabetic patients.The same questionnaire, as used with the diabetic patients, was administered to 487 black adult residents of this informal set tlement by 10 trained and paid black interviewers (residents of the area), who had been employed in previous annual studies in this informal settlement.As 50 respondents reported chronic conditions or disabilities, they were excluded from the analy sis; leaving a sample size of 437 controls.Ethical approval for the study was obtained from the Univer sity of Pretoria's Ethical Committee.The Nursing Service Man ager and staff at the hospital were consulted and fully informed about the study.The Health Committee and the councillors of the informal settlement approved the study.Informed consent was obtained from the patients and residents of the informal settlement.As measures of income are difficult to obtain, a Commodity Ownership Index was developed as an alternative SES meas ure (Westaway and Gumede, 2000).The Index consists of 10 items: telephone, car, television, refrigerator, separate freezer, stove, oven, microwave oven, radio and iron.Each item was scored 1 (yes) or 0 (no).Over 90% of patients and controls owned a stove, radio and iron.Since heterogeneous rather than homogeneous sample responses are required for index development (Nunnally, 1978), only 7 items were used for the index.Principal components analysis was used to develop a socio-economic score (SES) for each respondent, resulting in the equation:

Data Analysis
Data were analysed with the SPSS9, Windows package.Cronbach's alpha (1970), a measure of internal consistency, was estimated for the four multi-item SF-20 sub-scales (physi cal functioning, role functioning, mental health and general health).Reliability coefficients of >0.70 were regarded as satis factory, based on Nunnally's (1978) recommendation.
T tests and analysis of variance were used to ascertain the effects of gender, employment status, patient/control and marital status on HRQOL.Pearson correlation coefficients were used to ascertain relationships between age, schooling and HRQOL.Multiple analyses of covariance (MANCOVA), with Bonferroni corrections for multiple comparisons, controlled for demo graphic (age, gender, schooling, marital status and employ ment status) and socio-economic (commodity ownership) ef fects on HRQOL.

Dem ographic Inform ation
The questionnaire was administered to 281 patients (81 men and 200 women) and 487 controls (151 men and 336 women); 50 controls had a chronic disease/disability.These 50 were not included in the analysis.Analysis was conducted on 437 con trols (140 men and 297 women).As expected, patients were significantly older (p < 0.001), had less schooling (p < 0.001), were widowed (p < 0.001), were unemployed (p = 0.001) and owned more commodities (p < 0.001) than controls (Table 1).

Descriptive Statistics for H R Q O L
Mean scores, standard deviations and percentage scoring in the poor health range for the six sub-scales (physical function Older respondents had significantly poorer functioning and general health than younger respondents (p = 0.01); women had significantly poorer functioning (p = 0.05) but more pain than men (p < 0.01).Employed respondents had significantly better physical and role functioning, mental health and general health and less pain than unemployed respondents (p < 0.01).Widowed respondents had significantly poorer physical and role functioning and more pain than married or single respond ents (p < 0.01).SES was significantly related to general health (p < 0.01).These demographic and socio-economic findings were similar to those reported previously for Americans (Stewart, Greenfield, Hays, Wells, Rogers, Berry, McGlynn and Ware, 1989;Stewart. Hays and Ware, 1988).

Reliability and Inter-Correlations among the Health M easures
Reliability (internal consistency) coefficients for patients ranged between 0.78 (mental health), 0.81 (general health), 0.83 (physi cal functioning) and 0.94 (role functioning); for controls, these coefficients were 0.70 (mental health), 0.78 (general health), 0.80 (physical functioning) and 0.90 (role functioning).All relationships among the six health measures were statistically significant (p = 0.01) for both patients and controls (Table 3).For patients, there were substantial correlations among the functioning sub-scales and between mental health and general health.The relationship between mental health and general health was similar for controls.These findings suggested that mental health and general health are integral components of well-being.For controls, the so cial functioning item was mod estly related to the other health measures, suggesting that this item inadequately measured the social functioning of this sam ple (Table 3).

M ultiple Analyses of Covariance
M ANCOVA an aly ses, w ith Bonferroni corrections for mul tiple comparisons, controlled for age, gender, schooling, marital status, employment status and commodity ownership (Table 4).
Although controls scored higher than the patients on two of the functioning sub-scales, mental health, general health and health status, and lower on the pain sub-scale, significant differences were found between patients and controls on two functioning sub scales, general health, pain and health status.The findings on mental health were consistent

Discussion
tients and controls on the mental health sub-scale, consistent with previous research in the US and Finland (Aalto, Uutela and K angas, 1996;Stewart, Greenfield, Hays, W ells, R ogers, Berry, M cG lynn and Ware, 1989).However, only 26% of Finnish patients and 40% of healthy con trols reported good men tal health; very low lev els of well-being in com parison with US samples (78%) and our patients (62%) and controls (61%).Perhaps living in Finland is less conducive for well-being than living in the US or South Africa.Twenty per cent of our patients re ported good general health; slightly less than the 28% of Finn ish patients and considerably less than the 60% of US pa tients.Our patients had significantly higher levels of pain Table 3.
Intercorrelations among the health measures: patients (P) and controls (C controls provided support for the validity of the general health and well-being measures (Stewart, Hays and Ware, 1988).It was concluded that the SF-20 is a reliable and valid HRQOL measure; South African Black diabetic patients tend to have poorer health and more pain than Black controls; and Finnish diabetic patients, in comparison with South African and US patients, have the poorest HRQOL.

Table 1
Demographic information for type 2 diabetes mellitus patients and controls , social functioning, mental health, general health and pain) are shown in Table 2. Patients were signifi cantly more likely to report poorer physical and role function ing, poorer general health, more pain, but better social func tioning (p < 0.001) than controls.

Table 2 .
Descriptive statistics for the six health scales and percentage of patients (n = 28 1) and controls (n = 4 3 7) scoring in the poor health range