A meta-analysis demonstrates no significant differences between patient and population preferences

Abstract

Background and Objectives

To summarize and quantify mean differences between directly elicited patient and population health state evaluations (= preferences) and to identify factors explaining these differences.

Materials and Methods

Two meta-analyses of observational studies comparing directly elicited patient and population preferences for two stratified health state classifications: actual/hypothetical and hypothetical/hypothetical health states.

Results

Thirty-three articles comparing directly elicited patient and population preferences were included, yielding 78 independent preference estimates. These preference estimates served as input for the two stratified health state classifications. Data on health state assessments, elicitation methods, assessment method, and population characteristics was extracted by one reviewer, and checked by two other reviewers. These parameters were used to explain sources of heterogeneity. Overall, patients' actual health state preferences were not significantly higher than populations hypothetical health state preferences (summary mean difference [SMD] = −0.01, 95% confidence interval [CI] = −0.01, 0.03). Nor did preferences for hypothetical health states differ between patients and population (SMD −0.00, 95% CI = −0.02, 0.02). Most parameters substantially influenced the SMD, although the magnitude and direction differed for the two strata used (all P-values <.05).

Conclusions

The actual/hypothetical and hypothetical/hypothetical meta-analyses demonstrated no significant differences between patient and population preferences, suggesting that both can be used to allocate scarce resources.

Introduction

Cost-effectiveness and cost-utility analyses of health care interventions are increasingly used as instruments to allocate scarce resources in health care. The results of these analyses are summarized in Quality Adjusted Life Years (QALY) League Tables (QLTs). Whether these tables are appropriate decision-making tools has been extensively discussed in the literature [1], [2], [3]. One of the main shortcomings of these QLTs is that it is not always clear whether preferences are elicited from patients or from the general population. In addition, it is largely unknown whether it matters if one uses patient or population preferences [4], [5].

Preferences are quantitative expressions for certain health states and are used to calculate QALYs. They reflect the desirability of certain health states [6]. Methods for assigning preferences to health states may be choice based [e.g., Standard Gamble (SG), Time Trade-Off (TTO)] or nonchoice-based [e.g., Visual Analogue Scale (VAS), Rating Scale] [7]. The Health Utility Index (HUI), the Quality of Well-Being Scale (QWB), and the EuroQol-5D (EQ-5D) are multiattribute descriptive classification systems for generic health states with preference scores [7], [8], [9]. These instruments are suitable for economic evaluations [6], [10].

Preferences can be elicited from the patient him/herself. For some diseases, where the patient may be unable to answer questions (e.g., dementia or stroke) the judgments of physicians, other health care professionals or caregivers, so-called proxies, are used to elicit preferences. Since the Panel on Cost-Effectiveness recommended using the societal perspective for costs and effects, the general population can also be used [10]. However, the literature is ambiguous about the question whose preferences count [10], [11]. The conclusion that patient preferences are higher than [12], [13] or equal to [14], [15] population preferences is mostly derived from single articles or narrative reviews [15], [16]. These articles report mixed findings, and are confusing to those seeking guidance on decision making [17]. A meta-analytic approach may help to overcome this problem. To our knowledge, there has so far not been any attempt to summarize the literature quantitatively. The goal of our review was to analyze the difference between directly elicited patient preferences and directly elicited preferences from the general population. In the present study we applied a framework [18] to systematically identify the goals of the decision-making processes. We concentrated on the goal of guiding the decision-making process used to allocate resources. The relevant preferences were classified into the actual preferences of patients and the hypothetical preferences of the general population. The second goal of our review was to assess the bias of having experienced the disease, for which we compared the preferences of patients and the population for hypothetical health states.

Because preference studies are mostly observational studies, it is important to account for potential sources of heterogeneity [19]. As mentioned above, various methods can be used to elicit preferences and to explain the differences between patient and population preferences. Preferences can be attached to disease-specific or generic health states, or health state profiles. Hence, a health state assessment was included as a possible source of heterogeneity, as was the way preferences are assessed. External factors, such as gender, age, educational level, and type of disease (chronic, contagious, or risk factors for disease) were added [11], [15], [20], [21]. The third goal of this review was to assess the effects of different sources of heterogeneity on patient and population preferences.