Which is the best deprivation predictor of foetal and infant mortality rates? R Joyce 1 *, R Webb 2 , JL Peacock 1 and H Stirland 2 1 Department of Public Health Sciences, St George's Hospital Medical School, Cranmer Terrace, Tooting, London SW17 0RE; 2 South Thames West Perinatal Audit Team, Lower Ground Floor, Chiltern Court, 188 Baker Street, London NW1 5SD, UK This study investigates which, if any, population-based indicator of deprivation best predicts foetal and infant mortality rates in England. For the year 1995, the deprivation levels of 364 English Local Authorities were compared; using the three commonly used indicators, Jarman score, Townsend score and percentage unemployed. The predictive value of these for stillbirth, neonatal and infant mortality rates was then calculated. The three deprivation indicators were highly inter-correlated r  0.866 ± 0.924). For each mortality rate, the correlation with deprivation did not differ signi®cantly for the three indicators of deprivation. We conclude, when comparing these outcomes in different areas of England, that any of the three deprivation indicators may be used to adjust for deprivation. Public Health 2000) 114, 21±24. Keywords: deprivation; stillbirth; infant mortality; England Introduction Health Authorities, the new Primary Care Groups and NHS Trusts all need to be able to monitor their stillbirth and infant mortality rates. It is essential that their monitoring tools provide meaningful information. One of the methods of obtaining such information is to adjust these mortality rates for possible confounding factors. One of the most important predictors of stillbirth and infant mortality for which we need to adjust, is deprivation. As early as 1913, infant mortality rates in England and Wales were shown to be raised in lower compared to upper social class families. 1 This relationship is still evident. 2 Therefore, it is important to be able to adjust for the socio- economic deprivation of a population. This will provide information not only on the relative contribution of such deprivation to these mortality rates in populations, but also demonstrate variations in the mortality rates that are not accounted for by deprivation, and may have been `hidden' before adjustment. However, there has been no work until now to ®nd the best method for adjusting foetal and infant mortality rates for deprivation in a population. Many deprivation indica- tors have been developed, and some attempt has been made to compare them as predictors of other health outcomes. 3±5 There has, however, been no published literature showing the power of these indicators to predict perinatal mortality rates, and therefore no attempt to ®nd a `best indicator'. One paper only has shown a relationship with post- perinatal mortality and Jarman score. 6 We present here the results of a study, whose aims were to determine which of three commonly used deprivation indicators is the best predictor of stillbirth, neonatal and infant death rates; and hence the best indicator to use to adjust for deprivation in these populations. Materials and Methods Data Foetal and infant mortality rates were obtained, for each of the 366 English local authorities from the Public Health Common Dataset 7 for the year 1995. Deprivation indicators for the same local authorities were obtained, either from the Public Health Common Dataset or from the Department of Health, from information derived from the 1991 Census. The mortality rates obtained were those for stillbirth, neonatal and infant deaths de®ned as follows: 1. Stillbirth rate: babies born dead after 24 weeks gestation=babies born dead after 24 weeks gestation  live births)61000 2. Neonatal death rate: deaths occurring to live born infants under 28 days of life=live births61000 3. Infant deaths: deaths occurring under 1 y of life=live births61000. The deprivation indicators chosen were the Jarman score 8 obtained from the Public Health Common Dataset), Townsend score, 9 and the z score of the percentage of economically active population who were unemployed at the time of the Census both obtained from the Department of Health). Table 1 gives the components of the Townsend and Jarman scores, and their weightings. These particular deprivation measures were chosen because they are commonly used indicators that have been shown to compare well with other deprivation indicators in predicting health outcomes. 3,5 Townsend score is thought to be a good measure of material deprivation, whilst Jarman is considered a good measure of social deprivation. 10 Analysis Two local authorities, the Isles of Scilly and the City of London were excluded from the analysis. This is because the number of births were very small compared to the other local authorities 16 and 48 live births respectively), and hence their death rates were poorly estimated. *Correspondence: Dr R Joyce, Department of Public Health Sciences, St George's Hospital Medical School, Cranmer Terrace, London SW17 0RE, UK. Accepted 18 May 1999 Public Health 2000) 114, 21±24 ß R.I.P.H.H. 2000 www.nature.com/ph