OBJECTIVES There is a growing need to consider heterogeneity in cost-effectiveness analyses (CEA). To capture heterogeneity, subgroups analyses have been performed using various socio-demographic and clinical variables. However, the results of these subgroups CEA can be considered inequitable. Consequently, there is need to find new subgroups that can be used for decision-making. METHODS We explore whether subgroups defined by care-seeking behaviors are good candidates for CEA subgroup analysis. Care-seekers are defined as patients who received both an early diagnosis and an early treatment. We use data from the PLASA study, a French randomized controlled trial designed to reduce the rate of functional decline in Alzheimer’s disease: 1,131 patients were randomized in an intervention group and in a control group and were followed during a 2-year period. We use a sample selection model to explore whether the unobserved heterogeneity associated with the early diagnosis decision is correlated with the unobserved heterogeneity associated with the early treatment decision. We use a fixed-effect model to explore whether the rate of functional decline was lower within the care-seekers subgroup. RESULTS Our theoretically grounded selection model shows that the care-seeking behavior is associated with unobserved preferences, motivating the need to run subgroup analyses within a subgroup of care-seekers. Our fixed-effect model results show that on average, the clinical intervention was not effective. However, the intervention was effective within the subgroup of care-seekers. Care-seekers who received the intervention did not face a significant decline in their functional status over the 2-year study period. On the contrary, care-seekers in the control group lost on average 9 points of ADCS-ADL per year (p<0.01). CONCLUSIONS Stratifying CEAs by care-seekers subgroups seems relevant. Our analyses can be easily implemented by adding three questions in the clinical protocol.