The mean global QOL score among patients in South Africa and Uganda who were recruited into this study was 2.81 out of a maximum of 5, and the mean Total QOL score was 17.32 (possible range 0 to 30). Overall, patients in this study exhibited relatively poor QOL in comparison to studies using the MVQOLI in the USA (Table 6). In particular, mean Interpersonal and Transcendent scores were significantly lower in our sample than in all previous studies using the MVQOLI both in the USA [43, 58] and Uganda,  indicating worse QOL in these domains. Our data go against the findings of a qualitative study in Kenya which found that patients' psychological, social and spiritual needs were met by family members, the community and faith groups,  and suggest that this is not always the case. The mean Transcendent score in our study (5.50) is especially low given the next lowest mean score found on this subscale was 14.10, reported by Steele et al in their study of home hospice cancer patients in the USA .
However, overall, patients in our study reported significantly better scores on the Symptom subscale than AIDS patients in Uganda and home hospice patients' in Steele et al's USA study . Patients in our study also had significantly better functional and psycho-spiritual wellbeing than AIDS patients in the Ugandan study . This may relate to the physical and psychological burden of living with AIDS and the fact that the patients in the Ugandan study were not receiving palliative care.
As in the Ugandan study, we found that the poorest QOL was reported in the function domain, followed by psycho-spiritual wellbeing (as measured by the Well being subscale), then physical symptoms. These domains were also the worst three in the US studies; however, in a different order: Byock and Merriman report poorest QOL in function, symptoms, and then psycho-spiritual well being; Steele et al report poorest QOL in psycho-spiritual well being, function, and then symptoms. However, in our study scores in the spiritual domain (measured by the Transcendent subscale) were also notably low, while in the other studies this was not the case [43, 44, 58].
Despite the pain and symptoms associated with incurable, progressive disease, physical comfort and being active were judged by participants in this study to be less important to QOL than close relationships, feeling at peace and having a sense of meaning in life, as shown by the relatively low scores on importance items 5 and 10, although these differences were not statistically significant. This is also reflected in the results of the correlation, which show that the Transcendent, Wellbeing and Interpersonal subscales are most highly correlated with Total QOL, with spiritual and psycho-spiritual domains reaching levels of strong correlation (r > 0.7). Our findings support other studies suggesting that physical symptoms and function are not as important to patients with life-limiting illness as other domains of QOL [60–63]. This finding could be due to the acceptance of discomfort and physical limitation and the re-evaluation of goals that can occur through 'response shift' [64, 65]. There is evidence that cancer patients tend to readjust expectations to fit their current health and functional status,  and that patients in palliative care shift their values away from self-enhancement (e.g. power) and towards self-transcendence (e.g. benevolence) . These findings may reflect coping processes in the face of the uncertainty of living with an incurable, progressive disease .
Owing to the lower scores on importance items 5 and 10, the symptoms and function domains were accorded less weight in calculating the Total QOL score than the non-physical domains. However, despite this the Symptoms and Function subscales showed poor mean scores. This suggests that even if response shift or adaptation is evident in the relative importance allocated to the five domains, patients in this study nevertheless experienced considerable suffering and reduction of their QOL owing to uncontrolled or unacceptable symptoms and physical function.
Close relationships, feeling at peace and having a sense of meaning in life were rated as highly important by patients (mean 4.13, 4.13 and 4.10 out of 5 respectively). This supports the findings of other studies regarding the importance of interpersonal and spiritual domains in incurable progressive disease [35, 61, 69–71].
Our findings also suggest that patients with HIV receiving palliative care in South Africa and Uganda may experience poorer QOL than patients with cancer. HIV-infected patients scored significantly worse than cancer patients on Well being, Transcendence and Total QOL. In the USA, a study by Sherman et al found that patients with AIDS had lower total QOL scores, and lower psychological QOL than patients with advanced cancer; however, patients with AIDS had higher physical QOL scores . In a study of 2,864 HIV-infected adults in the USA, Hays et al found that patients with AIDS had significantly worse physical functioning and emotional well being than patients with prostate cancer . Our findings also support those of other studies in sub-Saharan Africa documenting the psychological and spiritual burden of living with HIV  and the need for support in these areas [73–75].
Modified subscale scores based on factor analysis in this sample and reported elsewhere  suggest that symptom control/acceptance of symptoms and interpersonal well being are relatively good, while physical function and acceptance of physical limitations is poor. Unlike the subscales in the original tool, the modified subscale scores are not weighted according to the subjective importance of the domains, and this should be noted when interpreting the modified scores. For example, although overall patients rate highest on symptoms and interpersonal well being with respect to the modified scores, once the weighting aspect is taken into consideration the beneficial impact of symptom control/acceptance of symptoms falls: interpersonal well being, rather than symptoms, has a more positive effect on QOL when the original tool scoring system is used.
Finally, it is interesting that global QOL as assessed by item 1 of the MVQOLI ('How would you rate your overall QOL?') was poorly correlated with Total QOL (r = 0.37). A relatively low correlation between Total QOL and global QOL (r = 0.43) was also reported in the original validation of the tool in the USA , although this was higher (0.58) in the Ugandan study of advanced AIDS patients . In our study setting the poor correlation between global QOL and Total QOL could suggest that the Total QOL score measures a construct different to QOL as understood by study participants. One reason may be that the tool does not assess socio-economic factors which might be of particularly relevance to patients in a resource-constrained setting. Hunger and stigma, for example, are prevalent in this population [44, 45] and could adversely affect QOL in a way not assessed by the MVQOLI in its current format.
Our study may have overestimated QOL in patients receiving palliative care in sub-Saharan Africa. South Africa and Uganda, where this study was conducted, are widely recognised as the two African countries with the most advanced provision of palliative care . The services involved were not selected randomly; hence findings may not be generalisable across palliative care services in the two countries. Participating sites were some of the most well-established palliative care services in South Africa and Uganda; patients in this study are therefore likely to be receiving relatively high quality palliative care and may have higher QOL than patients at other services. In addition, patients were required to be well enough to participate in self-report data collection, which may bias our data against those with significant disease progression and nearing the end of life, who may have poorer QOL than patients in this sample. As our findings relate to patients receiving palliative care, it is worth noting that other people with life-limiting illness in sub-Saharan Africa have less access to care than this sample, and hence are likely to have poorer QOL . Further studies are required in other African settings.
Owing to resource constraints we were not able to use the best practice methods of double data entry and tool adaptation involving multiple translations, synthesis of translations, back translation, expert committee review and pretesting prior to psychometric testing of the MVQOLI . Two academic departments carried out two of the translations. To minimise inaccuracies, all translations were cross-checked by bilingual palliative care staff and difficulties in translation were resolved through discussion with the local research team. While the factor analysis suggested the tool might be psychometrically appropriate for measuring QOL in this context, the translations utilised in this study would benefit from further investigation and cognitive testing in the future. The interpretation of our findings is therefore subject to modification on the basis of further testing of the MVQOLI in South African and Ugandan populations in the future.
Another limitation relates to the small cancer sample (n = 50) used when comparing MVQOLI scores in cancer and HIV diagnoses. As data were collected as part of a larger study focussing on psychometrics, no power calculation was used to determine sample size for this survey; hence our comparison of QOL domains by diagnosis should be interpreted cautiously. However, our comparison showed highly significant differences between the HIV and cancer samples for three out of six variables, despite the limited sample size.
The heterogeneity of the services and the populations they care for also affects the generalisability of findings. For example, at site D patients were recruited relatively close to admission (mean 2.9 weeks, SD 5.1), while patients at site E had been under palliative care for a mean of 122.3 weeks (SD 100.6). To take account of these variations we limited our comparison of cancer and HIV sub-groups to those services where the majority of cancer patients were recruited. However, there are likely to be demographic differences between patients in the two groups (e.g. gender, age) which we were not able to take into account in our analyses, hence our comparative findings should be interpreted with caution.
Finally, we used parametric t-tests to check for the significance of differences between our results and those of other studies (Table 6), despite a lack of normality in the MVQOLI distributions in our study. However, this was the only comparison possible given the data reported in other studies.
Our findings suggest that patients receiving palliative care in South Africa and Uganda have poor QOL as measured by the MVQOLI. While physical function was the domain with the least positive impact on QOL in this study (Function subscale mean 0.21), psycho-spiritual well being had the next least positive contribution (Wellbeing subscale mean 2.59). These findings show that it is vital to identify and meet the psychological and spiritual care needs of patients, as well as to rigorously assess and treat pain and other symptoms, and support patients in coping with their functional limitations. This requires multi-dimensional palliative care provided by a multi-professional team. Palliative care providers in sub-Saharan Africa are directed to local guidance on supportive and palliative care , spiritual care , and counselling .
The significantly worse scores on Symptoms, Interpersonal and Transcendent subscales among HIV infected patients compared to cancer patients support other studies reporting high levels of physical and psychosocial symptoms [32, 80, 81] and spiritual concerns  in HIV.
Population level findings should be interpreted with caution clinically, as we found that relatively high mean subscale scores may mask a significant proportion of patients scoring poorly on individual items. Individual levels of QOL depend on a host of contextual factors, such as patient history, family and community relationships, and spiritual beliefs. The heterogeneity of patient experiences should therefore be taken into account by ensuring that multi-dimensional needs are assessed regularly and that care is patient-centred and tailored to individual patient and family needs.
On the basis of our findings we recommend the following areas for future research. First, research is needed to explore the QOL of patients receiving palliative care in sub-Saharan Africa in different settings and using different models of care. Research focussed on specific settings and population groups would enable demographic and clinical predictors of QOL to be identified. Second, our findings support the need for research into psychological and spiritual aspects of patients' illness experience, as these areas are highly important to patients, are rated relatively poorly as domains of QOL, and are currently neglected in palliative care research in Africa. Third, our finding that physical function as measured by the MVQOLI has a significant negative effect on patients' QOL warrants further research. For example, it would be useful to explore current occupational therapy, counselling and home care provision in sub-Saharan Africa and evaluate the effect of these interventions on the functional domain of patients' QOL. Finally, future studies are needed to examine and if necessary improve the translations of the MVQOLI used in this study, explore the comprehension of specific translations through cognitive interviewing, and conduct further testing of the tool's validity, reliability and appropriateness, including assessment of weighting and scoring systems and exploration of factor loadings in populations from different African countries.