This study aimed to compare the accuracy of different adjustment methods for estimating the utility values of health states with multiple health conditions. We found that for scenario 2 (utility score for a specific condition measured for the subpopulation with only that condition), the estimates are less biased and more accurate, especially if additive and multiplicative methods are employed. However, we consider that scenario 1 (utility score for a specific condition measured for the subpopulation with that condition regardless of the presence of others) is more likely to occur in a real setting. In this case, the minimum and ADE methods could be considered, as they provided the best estimates. Additionally, according to our analysis, it is preferable to choose the mean utility of healthy individuals as the baseline utility rather than fix it to 1 (perfect health) for both scenarios. Overall, the estimates for health states with seven concomitant conditions were not satisfactory. This could be explained, at least in part, by the limited number of subjects in this subgroup.

Interestingly, these results contrast with some recommendations found in the literature. Fu and Kattan [14] concluded that the multiplicative method is not a good model and is outperformed by the minimum method for EQ-5D data from the US. However, other authors support the use of the multiplicative model for the estimation of joint health state utilities [15]. According to Janssen and Bonsel [16], the multiplicative method is highly accurate and superior to the additive method and should be used to adjust for multimorbidity. Ara and Brazier [7] compared the conventional methods (additive, multiplicative, and minimum) and found that if the baseline utility is fixed at the value 1 (perfect health), the additive method is the least accurate, whereas the multiplicative method produces the most accurate estimates.

Thompson et al. predicted health state utility values for individuals with one to four conditions in England [17]. According to their study, the multiplicative method was the best one for individuals with two simultaneous conditions. However, they do not recommend conventional nonparametric methods (additive, multiplicative, and minimum) for health states in which more than two health conditions coexist because of the possible bias in the predictions. In our study, however, the multiplicative approach produced good estimates for the utility values of patients with up to six simultaneous conditions.

The ADE model performed reasonably well for patients with two to four conditions in scenario 1, that is, when the utility values for a specific condition are measured in patients who have other concomitant diseases. Hu and Fu [5] reported that the ADE outperformed the three nonparametric methods for the estimation of joint health state utilities when two conditions are present.

Different approaches for estimating joint utilities are available, and there is not yet a broad consensus about the more appropriate one. Some studies showed that the additive and multiplicative methods tended to underestimate the utility values, regardless of the baseline used [7]. In contrast, the minimum and ADE methods could overestimate the utilities [7]. It is also argued that the minimum method could introduce a bias against older cohorts when applied in an economic model in conjunction with an age-adjusted baseline [12]. In turn, this would not be the case for the additive and multiplicative methods since both consider a utility detriment across ages [12].

In this context, it is reasonable to conclude that there is no ‘good fit for all’ method and the choice of the more appropriate one should rely on the characteristics of the available utility data. It is important to address that although accuracy is necessary, there are other criteria to be considered, especially whether the quality-of-life measures used to estimate the joint utilities are reliable [18]. Although there are absolute differences between the estimates obtained with the different methods, they are, in general, small. Further studies are necessary to assess the minimal important difference between the estimates [19] that would affect the results of the economic evaluations and, consequently, the decision-making process.

This study has some limitations. First, only mild self-declared conditions were reported. Additionally, the Brazilian valuation study [10] excluded individuals over the age of 64 years, primarily due to the high levels of illiteracy and cognitive impairments affecting this age group in Brazil. The high prevalence of dementia and other cognitive disabilities in this population could have made data collection even more challenging. However, age has been shown to influence an individual’s choice of quantity over quality of life. Therefore, the generalizability of the results for the elderly population should be further explored.

Nevertheless, the present study provides important insight into a more appropriate method for combining utilities in Brazil. The recommendations of the Brazilian guidelines for utility measures in health economic evaluations are based on the results of the present study. According to this guideline, the minimum method should be used if two simultaneous conditions are present since it produces good estimates in both scenarios and is simple to obtain, as no calculation is needed. For health states with more than two conditions, the multiplicative method is recommended if the utility of each single condition is obtained from individuals who also have other conditions (scenario 1), which is the most common scenario in our country. It is also recommended that the mean utility of the healthy population should be chosen as the baseline.