A computer-generated randomization schedule was prepared in advance for the initial recruit- ment period, and before every extension period, by a statistician with no role in recruitment. The sequence was composed of randomly permuted blocks of 4 or 6. Group assignments were concealed in sequentially numbered opaque envelopes kept in a secure location. A research as- sistant regularly reviewed the list of admissions to identify all infants admitted to the hospital with jaundice. The research assistant completed eligibility screening, obtained written, in- formed consent, and allocated mothers to their study group.