Indonesia has set an ambitious target of reducing its maternal mortality rate to 125 maternal deaths per 100,000 live births by the year 2010. This poses formidable challenges in a geographically diverse country where the majority of births occur at home. One option for the Indonesian government in order to reduce its maternal mortality would be to increase rates of skilled assistance for home deliveries. In order to design appropriate policies to achieve this, it is imperative to understand the determinants of use of birth attendants by mothers delivering at home. We use the Andersen Behavioral Model as a theoretical framework to understand the determinants of the use of a trained provider, traditional birth attendant, or no trained assistance during home deliveries in Indonesia. The 1997 Indonesia Demographic and Health Survey (IDHS) was used, and data from the most recent home delivery was abstracted for analysis. Out of a total sample of 10,692 home deliveries, a majority (53%) used the services of a TBA, 40% had a doctor, nurse or midwife in attendance, and only 7% delivered with the help of family and/or friends. A multinomial logit model was used to predict determinants of use. Our results indicate that maternal education, religion, asset index quartile and number of antenatal visits are significant determinants among all choice sets.