Introduction:
It is estimated that tobacco causes more than 8 million deaths each year, of whom about 1.2 million occur due to exposure to second-hand smoke (SHS) ¹. Complete and accurate data on SHS exposure at a regional level would enable health authorities to plan context-dependent control interventions. However, due to limitations associated with sample size, national health surveys (NHS) do not allow for reliable prevalence estimates by sex and age group at a subnational level. Small-area estimation (SAE) methods could be a valid alternative to have meaningful prevalence at a subnational level ^2-6. The aim of this study is to apply a SAEmodel-based methodology to obtain reliable estimations of SHS exposure by sex and age group in the Autonomous Regions of Spain.

Material and Methods:
The units of analysis were 180 areas defined on the basis of Spain’s territorial division into Autonomous Regions, as well as sex and age group (15-34, 35-54, 55-64, 65-74, 75 years and over). In each area, we estimated the prevalence of exposed (Ex) and non-exposed (NEx) to SHS in 2017, as well as their coefficients of variation (CV), applying a weighted ratio estimator (direct estimator) and a multinomial logistic model with random area effects ⁷. The data source used for the SAE model was the Spanish NHS 2017.

Results:
The range of Ex prevalence was 1-35% in men and 2-36% in women. The group aged 15-34 years was the most exposed to SHS, with an associated prevalence of 23%. When SHS exposure was estimated using the small-area model, the precision of direct estimates greatly improved; the CV of Ex and NEx decreased by an average of 92%.

Conclusions:
This study proposes a methodology to obtain reliable estimates of SHS exposure in groups not covered in the design of a population health survey. The model applied is a good alternative to improve the precision of estimates at a detailed level, at a much lower cost than that involved in conducting large-scale surveys. Having such estimates directly after completion of a health survey would help characterize the prevalence of any risk factor with greater precision.

This work has not been presented in any other conference or disseminated.