Social Science & Medicine 54 (2002) 931–937 Predicting small area health-related behaviour: a comparison of multilevel synthetic estimation and local survey data Liz Twigg*, Graham Moon School of Social & Historical Studies, Institute for the Geography of Health, University of Portsmouth, Milldam, Burnaby Road, Portsmouth PO1 3AS, UK Abstract A recent paper in Social Science and Medicine (Twigg et al. 50 (2000) 1109) outlined an approach to the estimation of prevalences of small-area health-related behaviour using multilevel models. This paper compares results from the application of the multilevel approach with those derived using the more traditional strategy of the local ‘lifestyle’ survey. Estimations of smoking prevalence and high alcohol consumption are examined and critical assessments made of both estimation approaches. It is concluded that the alternative method is more suited towards the prediction of smoking rates as opposed to unsafe alcohol consumption. # 2002 Elsevier Science Ltd. All rights reserved. Keywords: Smoking; Drinking; Synthetic estimation; Multilevel models; Health survey for England The central importance of health related behaviour, such as smoking, diet, drinking and exercise, with regard to health outcome is now widely recognised. A number of key national and international policy documents have set out targets for either an increase or decrease in specific behaviours (WHO, 1981; DoH, 1992). Respon- sibility, for facilitating change or for providing counsel- ling services is usually left to organisations or agencies working at the local area or neighbourhood level (e.g. community alcohol teams and primary care health promotion clinics). These local organisations and agencies have a need for locally specific data on health-related behaviours to assist in the targeting of their services and to provide a yardstick against which to monitor any changes resulting from their actions. Such information however, is not routinely available at the local scale. Consequently, alternative strategies for the estimation of small area health-related behaviour have evolved. The main focus of this paper is to compare and contrast the results generated by two such strategies. We consider the most commonly used strategy } the local health survey } and a more recently devised approach based on multilevel models of national survey data (Twigg, Moon, & Jones, 2000). The paper is organised into three sections. The first section outlines the two strategies under discussion. A comparison of the estimates generated by the two methods is presented in the second section of the paper. We conclude with a brief discussion of the implications of the comparison in the final part of the paper. The two key themes that the paper pursues are the pervasive nature of uncertainty and the importance of geographical context in making estimates of small area health related behaviour. Background Although large scale, well-designed, annually re- peated, continuous surveys such as the General House- hold Survey (GHS) and the Health Survey for England (HSE), provide information on some of the more important health-related behaviours such as smoking, diet, exercise and alcohol consumption, they do not provide reliable estimates below the standard region or health authority level, respectively (Colhoun & Prescott Clarke, 1996; ONS, 1996). One of the most basic strategies for obtaining estimates for local geographies *Corresponding author. Tel.: 023-9284-2466; fax: 023-9284- 2512. E-mail address: liz.twigg@port.ac.uk (L. Twigg). 0277-9536/02/$-see front matter # 2002 Elsevier Science Ltd. All rights reserved. PII:S0277-9536(01)00065-X