Regression analysis of “current status” life tables on duration of breastfeeding in Sri Lanka

Abstract

Provided that women report the dates of their children's births with reasonable accuracy, it is possible to derive good estimates of the duration of breastfeeding from women's breastfeeding status at the time of the interview. This paper illustrates the application of conventional regression techniques to the analysis of breastfeeding rates derived in this manner. Construction of current status rates is explained and a comparison between open interval, closed interval, and current status breastfeeding life tables is presented, indicating the extent of bias to which tables of the former types are open. Birth-weighted rates are used for WFS data from Sri Lanka; the variables entered into the regression equation include parity, educational level, residence, work experience since marriage and use of contraception since the birth. Contraception is not found to influence net breastfeeding rates in the 1st interval (1-16 months), although it is about as prevalent as in later intervals. The positive coefficients at intervals beyond the 1st also imply that contraceptive use is not a substitute for lactation in Sri Lanka or not a predominant one. Lifetime urban residence is associated with short durations of breastfeeding, and being an urban migrant is associated with intermediate durations relative to those of rural women. The effects of residence on breastfeeding are especially pronounced in the 1st interval. By parity as by contraception, differences in breastfeeding rates are not significant at short durations but become so with time as lower parity women reach pregnancy. Patterns by age are similar, but less sharp. Middle school attendance and work at home are both strongly associated with with lactation behavior, the former negatively and the latter to about an equal degree positively. Working outside the home seems not to influence breastfeeding to any great extent. In the multiple attribute regressions, middle schooling depresses breastfeeding durations about as strongly as lifetime urban residence relative to women who are rural and uneducated. Being at parity 5 or above or working at home almost as sharply increases durations. The use of contraception has slightly smaller positive effects. The patterns suggest that as Sri Lankan women become increasingly well educated and urban, and as family sizes decline, durations of breastfeeding will decline. Equally important however, even among better educated urban wives, breastfeeding continues longer than is typical of western countries.