A comparison of some simple methods to identify geographical areas with excess incidence of a rare disease such as childhood leukaemia

Six statistics are compared in a simulation study for their ability to iden tify geographical areas with a known excess incidence of a rare disease. Th e statistics are the standardized incidence ratio, the empirical Bayes meth od of Clayton and Kaldor, Poisson probability, a statistic based on the 'Br eslow T' test (BT) and two statistics based on the 'Potthoff-Whittinghill' test (PW) for extra-Poisson variance. Two alternative processes of clusteri ng are simulated in which high-risk locations could be caused by environmen tal sources or could be sites of microepidemics of an infectious agent cont ributing to a rare disease such as childhood leukaemia. The simulation proc esses use two parameters (proportion of cases found in clusters and mean cl uster size) which are varied to embrace a variety of situations. Real and a rtificial data sets of small area populations are considered. The most extr eme of the artificial sets has all areas of equal population size. The othe r data sets use the small census areas (municipalities) in Finland since th ese have extremely heterogeneous population size distribution. Subset selec tion allows examination of this variability. Receiver operator curve method ology is used to compare the efficacy of the statistics in identifying the cluster areas; statistics are compared for the proportion of true high-risk areas identified in the top 1 per cent and 10 per cent of ranked areas. On e of the PW statistics performed consistently well under all circumstances, although the results for the BT statistic were marginally better when only the top 1 per cent of ranked areas was considered. The standardized incide nce ratio performed consistently worst. Copyright (C) 1999 National Radiolo gical Protection Board.