Discontinuity Indices of Exclusive Breastfeeding Estimated by Probit Analysis of Current Status Data

This letter demonstrates the use of probit analysis as an optional method for estimating discontinuity indices in cross-sectional breast feeding surveys. Adjustment of a logistic function to current status data estimates discontinuity indices. The researchers analyzed data from a health center-based sample of 2411 infants aged 0-1 year living in Santo Andre in southwestern Brazil using a maximum likelihood iterative procedure to achieve a weighted linear regression of probits on average ages within each group. They found the cumulative discontinuity indices of exclusive breast feeding and 95% confidence intervals (CI) to be 51.3% (95% CI: 46.7-55.8) at 30 days old 64.6% (61.2-67.9) at 60 days 85.6% (83.2-87.7) at 120 days and 96% (94.4-97.2) respectively. These cumulative discontinuity indices in Santo Andre were like those in Havana Cuba. The Havana sample included children who either did or did not start exclusive breast feeding at birth. Partial discontinuity indices determine age intervals at which weaning is more frequent. In Santo Andre they were 27.3% for the age interval 30-60 days 33.3% for the age interval 60-120 days and 72.2% for the age intervals 120-180 days. In Havana the highest partial discontinuity index occurred during 2-4 months. In Santo Andre only 48.7% of children were exclusively breast fed at 30 days. The median duration of exclusive breast feeding was 28.9 days. These findings suggest different strategies of promoting exclusive breast feeding in Santo Andre and Havana even though the two cities had similar cumulative discontinuity indices. Two researchers conclude that probit analysis of cross sectional data estimate cumulative and partial discontinuity indices which help researchers to compare breast feeding patterns in different populations. Two potential advantages of probit analysis are: calculations are easy to do on programmable calculators in peripheral health services and confidence intervals are readily achieved for all estimates.