HIV/AIDS affects patients' life expectancy and quality of life . Clinical trials have proved that antiretroviral therapy (ART) extends the number of survival years and reduces morbidity up to 85% [2,3]. As ART effectively transforms the management of HIV infection to that of a chronic disease , the impact of morbidity on a patient's quality of life increasingly becomes of concern to decision makers.
Health-related quality of life (HRQoL) measurement assesses people's perceptions of their health status in terms of symptoms, physical functioning, work, social activities and mental health. HRQoL is increasingly recognized as a key outcome in clinical trials, providing a comprehensive assessment of the individual's overall health . Whereas cost-effectiveness studies in industrialized countries are increasingly incorporating HRQoL to evaluate healthcare interventions , few have been performed in resource-poor countries and none related to the provision of ART in Uganda. The amount of work involved in culturally adjusting, validating and administering these tools can be costly and involve a lengthy process . Moreover, there are clear differences in the concept of ‘health’ between industrialized and other countries .
HRQoL tools include generic and specific instruments. Generic tools measure the overall impact of the intervention on the individual's HRQoL for any population and disease, whereas disease-specific tools evaluate particular stages of health deterioration. Generic measures consist of HRQoL indices and profiles. Whereas HRQoL profiles only measure health states within a single domain, HRQoL indices allow an overall valuation for each health state . Preference elicitation methods instruments, a category of HRQoL indices, permit a valuation of the potential outcome of an intervention according to relative values that individuals place on morbidity (quality of life) and mortality (quantity of life) outcomes. These utilities or preference values are obtained by administering one or more of three tools: visual analogue scale (VAS), time trade-off (TTO) and standard gamble (SG).
VAS is a scale in the form of a thermometer ranging in value from 0 to 1, representing the worst and the best possible health states; the respondent is asked to identify a point in the scale that represents their perception of a health state . In TTO individuals state the number of life years at which living in the best attainable health state would be equivalent to living a given number of years in the deteriorated health state; the ratio of the former to the latter is the TTO utility value . SG requires the respondent to determine the rate of risk of a treatment option with two possible outcomes, immediate painless death and an improved health state, that would make it equivalent to remaining in the deteriorated health state; the respective probability of success is the SG utility value .
This study aims to test the psychometric performance of preference elicitation methods to value health outcomes characteristic of HIV disease and treatment in a setting of constrained resources. This is an initial step towards measuring the benefits of competing health programmes to tackle HIV. HIV-positive individuals constitute a suitable population to answer these questions because they are likely to be more familiar than other populations with the health outcomes being investigated. The objective of this study was to obtain utility values for three predetermined HIV/AIDS health states using VAS, TTO and SG in HIV-infected Ugandan individuals, and to assess the usefulness of these tools for cost-effectiveness analyses in resource-constrained settings.
The study was conducted in Entebbe in the Wakiso district of Uganda. We recruited two groups of HIV-infected individuals: the first from the Development of AntiRetroviral Therapy in Africa (DART) trial before they started taking ART (the DART Cohort group), and the second from ART-naive HIV-infected individuals in the Entebbe Cohort (Entebbe Cohort group). Participants from this group are likely to differ from the DART group with respect to disease stage and CD4 cell counts at enrolment.
DART is an open-label randomized trial evaluating the management of ART in symptomatic HIV-infected adults in Uganda and Zimbabwe. It compares clinical monitoring only with laboratory plus clinical monitoring. This trial is being conducted in three centres and 1000 patients participate in each site; patients will be followed for up to 4 years and will initially receive zidovudine and lamivudine in combination (combivir) plus tenofovir or nevirapine or abacavir.
The Entebbe Cohort was established in 1995 as part of the collaborative work of the AIDS Support Organization and the Medical Research Council (MRC) Unit on AIDS in Uganda. HIV-infected individuals enrolled in the Entebbe Cohort receive initial general health examinations and investigations, including confirmatory HIV serology tests and CD4 cell counts, and are classified according to the WHO clinical staging system. Participants attend scheduled visits every 6 months.
Preference elicitation tools
Three predetermined health states for HIV/AIDS were constructed for the study, based on WHO stages 2, 3 and 4 for HIV-infected individuals: symptomatic HIV (SHI); minor AIDS-defining illness (MIADI); and major AIDS-defining illness (MAADI). The health states were intended to be as comprehensive as possible, including levels of physical, emotional and social functioning . Because ART does not cure, the choice of the reference comparison health state was ‘improved health’ state (IHS) and not ‘perfect health’. IHS intended to represent the expected major benefits of receiving ART. The extreme values used in this study were 0, for worst health state attainable, and 1, for best health state attainable. All health states were assumed to last for 10 years only, after which the individual would die.
Once constructed, the health states were reviewed by clinicians working in the HIV/AIDS field at the MRC Unit for AIDS in Uganda, and the Liverpool School of Tropical Medicine in the United Kingdom. Health states were amended on the basis of their comments, which were also incorporated in the focus group discussions involving HIV-infected individuals receiving ART (unpublished data). The results from the focus group discussions were used to revise the description of the health states evaluated through the preference elicitation tools.
Each individual assessed the three predetermined health states with VAS, TTO and SG. For VAS the interviewers used a yellow ruler of 100 cm, which allowed participants to indicate how they were feeling at the time of the interview and rate the health states for HIV/AIDS. This tool was also used as a warming up exercise to familiarize individuals with evaluation exercises. TTO and SG evaluate participant's willingness to trade-off time and risk to attain an improved health state [12–14]. TTO values were obtained by using a TTO board in which individuals were asked to express the amount of time that they were willing to give up in order to have a predefined improved health state instead of being in each one of the predetermined HIV/AIDS health states. TTO values were obtained once the participant had no preference between living in each predetermined HIV/AIDS health state for 10 years and living in the predetermined improved health state for a reduced number of years.
The SG chance board was used to assess the level of risk that an individual was willing to incur in return for an improved health state. The participants had the choice between staying at a predetermined HIV/AIDS health state or taking a gamble (a hypothetical drug) with two hypothetical possible outcomes: an improved health state with probability P or immediate painless death with probability 1−P. The utility weight was then obtained by varying the level of risk (from extremes i.e. 95% then varied to 5%, 90% then varied to 10% and so on) until the participant was indifferent to either option. For both options the number of prospective life years was the same (10 years), after which the individual would die.
Interviews were conducted face to face in the local language (Luganda) when participants were attending DART and the AIDS Support Organization clinics. Participants were asked to: rate his/her own health state using VAS; rank and evaluate the SHI, MIADI and MAADI health states with VAS; consider whether they would like to change their initial valuation and if so to provide their new valuation for own health state using VAS; evaluate SHI, MIADI and MAADI health states using TTO and SG relative to a predefined IHS using cartoon aids.
In addition, a socioeconomic questionnaire was designed to capture information on age, marital status, level of education, income and consumption expenditure.
Socioeconomic data were analysed using t-tests for continuous variables and chi-square tests for categorical proportions. The Mann–Whitney test was used to compare valuations of respondents' own health across groups and, for pre-determined HIV/AIDS health states, in pair-wise comparisons across states.
Each instrument was evaluated with respect to its psychometric properties. Practicality or feasibility was assessed through the mean time of administration, percentage of missing responses and ease of administration for interviewers and interviewees. Test–retest reliability was assessed using Spearman's rank coefficient by inviting a subgroup of 20 individuals from each group to return for another interview 2 weeks after their first. Empirical validity was assessed to see whether different instruments yield similar results by estimating the Spearman's rank correlation coefficient between the instruments. Construct validity was assessed through linear regression analysis to determine whether there was any relationship between the values obtained through VAS, TTO and SG and gender, age, CD4 cell counts or self-assessed health (measured with VAS). The construct validity analysis pooled data for SHI, MIADI and MADDI health states; the covariates were interacted by binary indicators variables (dummies) to control for the three different health states. We used standard t-tests (linear model) or Wald tests (in gamma models) on the coefficients of the interactions to test the hypothesis that the effect of the variable in question did not differ between groups.
Data collection forms were standardized and double entered into Access databases at the MRC statistics unit in Entebbe. The analysis was conducted using Stata 9 (Stata Corp., College Station, Texas, USA). None of the questionnaires carried the names of participants; instead, study numbers were used as personal identifiers. Data records were treated confidentially and were only available to the staff directly concerned with this research.
Ethical approval was obtained from the Uganda Virus Research Institute Science and Ethics Committee, the Uganda National Council for Science and Technology, and the Research Ethics Committee of the Liverpool School of Tropical Medicine.
We recruited 276 individuals from the DART trial and 159 from the Entebbe Cohort. The majority (64% DART, 76% Entebbe) were women. The mean age was similar: 36.5 and 36.6 years for the DART and Entebbe Cohorts, respectively. Whereas 90% of individuals from the DART group had received primary education, there was a non-significant trend towards a lower level of education in the Entebbe participants. The groups differed with respect to marital status; DART Cohort participants were more likely to be married and Entebbe Cohort participants were more likely to be widowed (P < 0.001). The average household size of the groups was similar, i.e. 4.8 for DART and 5.2 for Entebbe Cohort participants. A higher proportion of DART Cohort participants lived in peri-urban areas, whereas the majority of Entebbe Cohort participants lived in rural areas (P < 0.001), reflecting the DART trial requirement that participants live within a certain radius of Entebbe (see Table 1).
Both groups reported that their main activity was selling perishable goods at the local market. DART and Entebbe Cohort participants reported median family expenditures of 80 000 and 70 000 Uganda shillings, i.e. US$45 and US$39, respectively (exchange rates for 2005–2006, corresponding to the recruitment period, were used and obtained from the Central Bank of Uganda). The monthly income reported was significantly greater for DART Cohort than Entebbe Cohort participants, at a median of 100 000 and 50 000 Uganda shillings, i.e. US$56 and US$28, respectively (P = 0.002).
By dividing family expenditure by the number of people living in the household, we estimated the median monthly per capita family expenditure to be approximately 22 000 Uganda shillings (US$12) for the DART Cohort and 15 000 Uganda shillings (US$8) for the Entebbe Cohort group (P = 0.002). In total, 84% of the DART Cohort and 90% of the Entebbe Cohort participants were under the official international poverty line defined by the World Bank in 1990 of US$1 per capita per day [15,16].
Participants were asked about their employment status 12 months before their date of enrolment in this study. Fifty-seven per cent of the DART group and 28% of the Entebbe Cohort group, respectively, reported not having a paid job. The main reasons for not having a job were: non-availability of employment, in 107/158 (68%) and in 28/45 (62%) and ill health, in 20/158 (13%) and 12/45 (27%) of the cases for DART and Entebbe Cohort participants, respectively.
Those who had a paid job 12 months before study enrolment reported a mean monthly income of 80 000 Uganda shillings (US$45) and 60 000 Uganda shillings (US$33) for DART and Entebbe Cohort participants, respectively.
Preference elicitation results
A total of 267/276 (97%) and 150/159 (94%) individuals from the DART and Entebbe Cohorts agreed to participate in the preference elicitation exercise. Reasons for refusal included partial blindness, lack of time, language barrier, feeling distressed or unwell, incomprehension of the tools and religious beliefs.
For comparability with TTO and SG values, VAS scores were transformed, from a 0 to 100, to a 0 to 1 scale (Table 2). Participants' own health ranking was similar for both groups 0.50 versus 0.55 for the DART and Entebbe Cohorts, respectively (P = 0.04). Both groups valued their own health in the middle of the scale; when prompted to explain their valuation, the majority said that they were feeling ‘neither too sick nor too well’. Using VAS, participants in both groups were able to discriminate between the best, intermediate and worst predetermined HIV/AIDS states (P < 0.0001; see Table 2). After their evaluation of the predetermined HIV/AIDS states, most individuals from the DART Cohort group changed their initial personal VAS valuation. Although the mean value increased for both groups on reassessment, DART Cohort participants reported both a greater increase and overall higher personal values. This finding is unexpected: DART Cohort participants had lower CD4 cell counts (mean 76 cells/ml) and were expected to feel less well than the Entebbe Cohort participants (mean CD4 cell count 393 cells/ml).
Table 3 presents the results from the valuation of predetermined HIV/AIDS states by TTO. It shows that a representative individual from the DART Cohort group would be willing to give up 2.5 out of 10 years of life in order to improve his/her health status from the symptomatic HIV state. In comparison, the typical Entebbe Cohort respondent was willing to give up to 2.2 years. At the other extreme the average individual was willing to give up to 7.9 (DART) and 7.3 (Entebbe) years in MAADI. Participants from both groups were able to discriminate between the HIV/AIDS states with TTO (P < 0.0001; see Table 3).
Table 4 presents the results from the valuation of predetermined HIV/AIDS states by SG. Some individuals from both groups were unwilling to take any gamble and would always prefer the deteriorated health state to the risky alternative: SHI, DART Cohort 7/265 (3%), Entebbe Cohort 42/150 (28%); MIADI, DART Cohort 6/241 (2%), Entebbe Cohort 29/132 (19%); and MAADI, DART Cohort 6/238 (2%), Entebbe Cohort 19/112 (13%). The typical value given by a DART Cohort participant implies a willingness to take a gamble with a 50% (i.e. 1–0.50) chance of immediate death instead of facing a certain prospect of living for 10 years in the symptomatic HIV state. The corresponding value for an Entebbe Cohort participant was 49%. In the case of MAADI, DART and Entebbe Cohort participants would gamble with an 81% chance of sudden death. Participants from both groups were able to discriminate between the HIV/AIDS states with SG (P < 0.0001; see Table 4).
Psychometric performance results
It took fewer than 10 min to explain and administer VAS. TTO and SG took between 12 and 15 min. VAS was found to be the easiest tool to explain and administer, followed by TTO and SG. No missing values or problems were reported when using VAS or TTO and only two missing values for one MIADI and one MAADI for the DART Cohort group were recorded with SG.
Only 12 (60%) individuals from the Entebbe Cohort group who were invited for test–retest came back. All 20 participants invited from the DART Cohort returned. The results suggest that SG is less reliable than the TTO and VAS for both groups. The Spearman's rank correlation coefficients per instrument for both groups are: 0.71, 0.72 and 0.41 for VAS, TTO and SG for the DART Cohort group; for the Entebbe Cohort group the respective numbers were 0.83, 0.77 and 0.42.
In principle, one would expect TTO and SG to be more strongly associated than either of the two with VAS. The Spearman correlations, however, suggest that the values obtained by the different tools represent different constructs: TTO versus SG, DART Cohort (0.39) and Entebbe Cohort (0.21); TTO versus VAS, DART Cohort (0.61) and Entebbe Cohort (0.45) and SG versus VAS, DART Cohort (0.34) and Entebbe Cohort (0.26).
In this analysis valuations were modelled as a function of the covariates interacted with indicators used to distinguish between predetermined HIV/AIDS health states in a single regression. Differences in the effect of covariates on valuations across health states were therefore tested by conducting standard tests on the coefficients of interactions. See Table 5 for the main results of regression analyses.
Age, CD4 cell counts (not shown) and gender had no influence on the VAS and SG valuations of either group or on TTO valuations in the DART Cohort participants (P ≥ 0.098). Gender was associated with TTO valuations of the three predetermined HIV/AIDS health states for the Entebbe Cohort participants (P = 0.002); male participants were less willing to trade off time in exchange of an improved health state. No other covariates had any apparent effect on valuations of this group.
A participant's own health assessment was positively associated with VAS and TTO valuations of the hypothetical health states in both groups (P = 0.045 and P = 0.16 for the DART Cohort; P = 0.001 and P = 0.002 for the Entebbe Cohort) and was negatively associated with SG values in DART Cohort participants (P = 0.053). It was also negatively associated for the symptomatic HIV health state but positive for both MIADI and MAADI states with SG in Entebbe Cohort participants; these associations were statistically significant at P = 0.05.
This study shows that participants were able to discriminate between the three predetermined HIV/AIDS health states using SG, TTO and VAS, as shown by their valuations. All instruments showed good feasibility. Constructs measured by TTO and VAS are more closely related than either of the two is to SG. The analyses of construct validity appear to be acceptable and in line with other studies for SG, TTO and VAS valuations . In general, these valuations are independent of age and gender, while being (positively or negatively) associated with self-assessed health but not with the CD4 cell count. Only TTO valuations by Entebbe Cohort participants were influenced by gender.
The results obtained from TTO and VAS were in line with other empirical studies in industrialized countries [17,18]. Mean estimates derived from VAS were found to be lower than TTO. This finding is a typical result; because the VAS metric does not imply a trade-off between quantity and quality of life, as in TTO, or treatment risk and health reward, with SG, its values may be interpreted as lacking economic meaning . In particular, VAS represents a statement of relative standing in a best to worst possible measurement scale, whereas in the other two instruments the individual is asked to make a hypothetical decision involving an acceptable sacrifice in the form of treatment risk or a number of years of life in exchange for a better quality of life.
Our finding of mean SG values being lower than mean TTO values is atypical of studies conducted in populations from industrialized countries . The feasibility results presented here suggest that respondent burden is not a possible explanatory factor, despite the fact that SG questions invariably appeared last. On the other hand, the understanding of probabilities has proved difficult in preference elicitation studies in areas such as health, transport and environment [21,22]. How the interpretation of death risk interacted with individual's perceptions and information about treatment for a highly stigmatized disease remains an open question that will need to be addressed by further research.
Our choice of restricting the study population to HIV-positive individuals may cause reservation among those who take the view that resource allocation decisions should be based on valuations of health outcomes by a representative sample from the general population . Patients' values, such as those reported here, however, may inform resource allocation decisions, by addressing the limitations of valuation studies in the general population such as a lack of sensitivity of specific changes in health status, bias from omission of relevant health attributes, and bias against the conditions of people with disabilities . Arguably, the findings of cost-effectiveness analyses should be subject to tests for robustness by incorporating the relative valuations implicit in this study, because they reflect the views of the population groups that are likely to lose the most in cases in which an HIV programme more effective and more costly than existing programmes is not adopted.
The difficulties of measuring individuals' preferences arise in conceptualizing, measuring and obtaining comparative valuations of different levels of quality of life. No single preference elicitation method has been universally accepted as superior to the rest; theoretical debates are still taking place about the validity of all the tools tested in this study . Because of its format, VAS values are not commonly used to inform cost effectiveness analyses, although recent studies have sought to derive transformations to convert them to TTO and SG values . Therefore, the decision to use TTO or SG ‘need(s) to be informed by their respective performance on empirical grounds’ .
Although the absence of utility values for African populations might tempt researchers to consider using utility values derived from, say, UK or US general populations [26,27], meaningful cost-effectiveness analyses used to inform resource allocation decisions in resource-constrained settings should reflect the specific preferences of the population in each country. Moreover, the same generic item may refer to completely different constructs in a developed setting to those applicable to a resource-poor setting. To illustrate this, take mobility as an example: being bedridden in an industrialized setting is not comparable to being bedridden in Africa, where depending on household circumstances it may mean to be left without running water, no electricity and being unable to fetch firewood for cooking.
This type of research in Africa is in its infancy. Nevertheless, this assessment showed that VAS, TTO and SG have good psychometric properties, making them good candidates for use in resource-constrained settings. The valuations presented here, however, may be used for resource allocation in this setting, subject to adjustments to reflect the fact that these valuations were obtained with reference to an improved as opposed to a perfect health state. Further research that aims to evaluate and compare the HRQoL perceptions of healthy and HIV-infected individuals in Uganda and other African countries is necessary to generate the evidence base with which to inform resource allocation decisions in healthcare.
We thank all the patients and staff from the MRC/UVRI Uganda Research Unit on AIDS, Entebbe, Uganda: H Grosskurth, P Munderi, G Kabuye, D Nsibambi, R Kasirye, E Zalwango, M Nakazibwe, B Kikaire, G Nassuna, R Massa, K Fadhiru, M Namyalo, A Zalwango, L Generous, P Khauka, N Rutikarayo, W Nakahima, A Mugisha, J Todd, J Levin, S Muyingo, A Ruberantwari, P Kaleebu, D Yirrell, N Ndembi, F Lyagoba, P Hughes, M Aber, and the AIDS Support Organisation (TASO), Uganda: A Coutinho, B Etukoit for supporting this study. The authors are also grateful to Dr Ruben E Mujica Mota and to an anonymous reviewer for the helpful comments received on drafts of this paper.
Sponsorship: This study was funded by the HIV/AIDS and STI Knowledge Programme at the Liverpool School of Tropical Medicine.
Conflicts of interest: None.
1. Wua AW, Hays RD, Kelly S, Malitz F, Bozzette SA. Applications of the medical outcomes study health-related quality of life measures in HIV/AIDS. Qual Life Res 1997; 6:531–554.
2. Wachtel T, Piette J, Mor V, Stein M, Fleishman J, Carpenter C. Quality of life in persons with human immunodeficiency virus infection: measurement by the medical outcomes study instrument. Ann Intern Med 1992; 116:129–137.
3. Globe DR, Hays RD, Cunningham WE. Associations of clinical parameters with health-related quality of life in hospitalised persons with HIV diseases. AIDS Care 1999; 11:71–86.
4. Flepp M, Schiffer V, Weber R, Hirschel B. Modern anti-HIV therapy. Swiss Med Wkly 2001; 131:207–213.
5. Hobart JC, Freeman JA, Lamping DL. Physician and patient-oriented outcomes in progressive neurological disease: which to measure? Curr Opin Neurol 1996; 9:441–444.
6. US Department of Health and Human Services, FDA Center for Drug Evaluation and Research. Guidance for industry: patient-reported outcome measures: use in medical product development to support labeling claims: draft guidance. Health Qual Life Outcomes
7. Medina Lara A. Adaptation and assessment of health related quality of life (HRQoL) instruments for people living with HIV/AIDS in Uganda
. PhD Thesis. University of Liverpool; 2007. 295 pp.
8. The Euroqol Group. EuroQoL– a new facility for the measurement of health-related quality of life. Health Policy
9. Sen A. Health: perception versus observation. BMJ 2002; 324:860–861.
10. Wewers ME, Lowe NK. A critical review of visual analogue scales in the measurement of clinical phenomena. Res Nurs Health 1990; 13:227–236.
11. Torrance GW, Thomas WH, Sackett DL. A utility maximization model for evaluation on health care programs. Health Serv Res 1972; 7:118–133.
12. Torrance GW. Toward a utility theory foundation for health status index models. Health Serv Res 1976; 11:349–369.
13. Furlong W, Feeny D, Torrance G, Barr R, Horsman J. Guide to design and development of health-state utility instrumentation
. Working paper. Centre for Health Economics and Policy Analysis. Hamilton, Ontario: McMaster University; 1990. pp. 90–99.
14. Torrance GW, Feeny D. Utilities and quality-adjusted life years. Int J Technol Assess Health Care 1989; 5:559–575.
15. Ravallion M, Datt G, van de Walle D. Quantifying absolute poverty in the developing world. Rev Income Wealth 1991; 37:345–361.
16. World Bank. World development report 1990: poverty. New York: Oxford University Press; 1990.
17. Dolan P, Gudex C, Kind P, Williams A. Valuing health states: a comparison of methods. J Health Econ 1996; 15:209–231.
18. Brazier J, Green C, McCabe C, Stevens K. Use of visual analog scales in economic evaluation. Exp Rev Pharmacoecon Outcomes Res 2003; 3:293–302.
19. Bayoumi AM, Redelmeier DA. Economic methods for measuring the quality of life associated with HIV infection. Qual Life Res 1999; 8:471–480.
20. Brazier J, Deverill M, Green C, Harper R, Booth A. A review of the use of health status measures in economic evaluation. NHS R&D HTA programme. Health Technol Assess 1999; 3:174–184.
21. Loomes G, Moffatt P, Sugden R. A microeconometric test of alternative stochastic theories of risky choice. J Risk Uncertainty 2002; 24:103–130.
22. Hey JD. Experiments and the economics of individual decision making under risk and uncertainty
. York: University of York; 1995. p. 49.
23. Dolders MGT, Zeegers MPA, Groot W, Ament A. A meta-analysis demonstrates no significant differences between patients and populations preferences. J Clin Epidemiol 2006; 59:653–664.
24. Ubel PA, Richardson J, Menzel P. Societal value, the person trade-off, and the dilemma of whose values to measure for cost-effectiveness analysis. Health Econ 2000; 9:127–136.
25. Parkin D, Devlin N. Is there a case for using visual analogue scale valuations in cost–utility analysis? Health Econ 2006; 15:653–664.
26. Pitt J, Badri M. Wood R. Changes in quality of life in South African antiretroviral programme.
In: 15th International AIDS Conference
. Bangkok, Thailand, 12–16 July 2005 [Abstract MoPeD3782].
27. Badri M, Cleary S, Maartens G, Pitt J, Bekker LG, Orrell C, Wood R. When to initiate highly active antiretroviral therapy in sub-Saharan Africa? A South African cost-effectiveness study. Antivir Ther 2006; 11:63–72.