Health State Valuation in Nigeria: An Insight into Discrete Choice Experiment-Based Preferences for Children and Adolescents from the Lenses of Adult Respondents.

Abstract

Background: Developing local value sets for child and adolescent health is critical for economic evaluation in Nigeria. Discrete choice experiments (DCEs) can elicit these preferences, yet data on choice behaviour from Nigerian populations are lacking.

Objectives: The aim of this study was to analyse DCE-based preferences for EQ-5D-Y-3L health states among Nigerian adult proxies, examining dimension-level drivers of choice, prevalence and correlates of preference reversals, and overall data quality.

Methods: A DCE was administered to 2699 adults across six Nigerian states, generating 40,485 choice task observations. Paired health states were constructed using an efficient design. The EuroQol Portable Valuation Technology (EQ-PVT) recorded choices and identified logical reversals. Analyses included descriptive statistics, mixed-effects logistic regression, and comparative tests examining differences between reversed and normal tasks.

Results: Reversals, which reflect respondent-level software orientation rather than task-level choice behaviour, occurred in 50.32% of tasks; severity, extreme-problem status, and demographics did not differ between EQ-PVT-reversed and normal tasks, confirming this measure as a data-entry artifact rather than evidence of decision strategy. A level-sum-score-based measure of substantive preference inconsistency, distinct from the EQ-PVT reversal, identified inconsistency in 10.7% of non-tied pairs. Pain/discomfort (β = - 1.544) and anxiety/depression (β = - 1.328) were the strongest predictors of choosing the better health state (p < 0.001).

Conclusions: This large-scale DCE provides foundational evidence for a Nigerian EQ-5D-Y-3L value set, confirming pain/discomfort and anxiety/depression as primary drivers of proxy preferences for child health. The EQ-PVT reversal rate reflects a balanced software-orientation assignment rather than flawed decision making; the true behavioural inconsistency rate (10.7% of non-tied pairs) was low and supports the feasibility of generating locally relevant preference data for health technology assessment in Nigeria.