Mothers of singleton infants were recruited shortly after delivery from 3 hospitals in the District of Columbia between August, 1995 and September, 1996. Two of the selected hospitals had a high prevalence of low-income patients, and enrollment at the third site was limited to nonprivate patients (patients cared for by staff physicians) to capture a sample of low-income, inner-city patients. Recruitment days were chosen systematically to ensure a representative sample of weekend and midweek sample days. Sampling was for a 24-hour period on the selected days.

Baseline interviews were conducted after delivery, usually during the postpartum hospital stay. Of the 452 baseline interviews, 417 (92%) were conducted in the hospital during the postpartum stay, and 94% were conducted within 2 days of delivery. All but 1 were completed within 2 weeks of delivery. Baseline interviews were conducted by trained research assistants and included a face-to-face interview and a short self-administered questionnaire. The face-to-face interview included questions about sociodemographic and psychological factors, including race, ethnicity, social support and health behaviors. The self-administered questionnaire included questions about potentially sensitive topics such as wantedness of the pregnancy, and household drug and alcohol abuse. Follow-up interviews were conducted when infants were 3–7 months of age and again when infants were 7–12 months of age. Of 780 follow-up interviews, 89% were face-to-face, and most were conducted in the mother's home.

At the 3–7 month follow-up interview, mothers were asked if they were currently breastfeeding, bottle feeding, or both. Women who were only bottle feeding were asked if they had ever breastfed their infant, and if so, for how many weeks or months. It was assumed that bottle feeding was interpreted by mothers to mean "formula" feeding or "milk" feeding and not the feeding of expressed breast milk by bottle. Mothers who answered that they had never breastfed their infants were asked why they chose not to breastfeed; mothers who had stopped breast feeding were asked to identify the reasons why they stopped. Choices were derived from the pretest for the 1995 National Survey of Family Growth 19 and included "I was ill/weak," "baby was ill/weak," "Nipple and/or breast problems (e.g., pain, dryness)," "I produced insufficient milk," "Schedule difficult (e.g. my job or baby's eating schedule)," "Baby refused," "I didn't want to pass dangerous things through breast milk (e.g., medications, cigarette smoke, alcohol, drugs)," "I was never taught or didn't know how," "I didn't want to," or "Other reason." Women were allowed to select as many choices as applied.

At the 7–12 month follow-up interview, mothers were asked again if they were currently breastfeeding, bottle feeding, or both. Mothers who had quit breastfeeding were asked why they stopped, using the same selection of reasons as previously described.

Information abstracted from maternal medical records included parity; type of delivery; gestational age at delivery; use of medications at the time of admission for delivery, use of alcohol, tobacco, or drugs of abuse during pregnancy; presence of selected medical conditions during pregnancy (including HIV/AIDS); and discharge ICD9 codes. Information abstracted from infant medical records included infant birth weight, sex, and discharge ICD9 codes.

The definition of a documented contraindication to breastfeeding was based on a review of recent literature addressing contraindications to breastfeeding for North American women 151617 and included one or more of the following: maternal HIV infection, active tuberculosis infection, or infection with human T-cell lymphotropic virus type I or type II; illicit drug abuse/drug dependence (opioids, cocaine, amphetamines, barbiturates, and hallucinogens); or maternal use of certain medications (radiopharmaceuticals at therapeutic doses, chloramphenicol, bromocriptine, cyclosporin, methadone, cyclophosphamide, doxorubicin, ergotamine, lithium, methotrexate, phenorbarbital, and phenindione). Women with contraindications were identified in the following manner. Maternal medical records were reviewed for use of the medications listed above. Maternal discharge ICD9 codes were searched for documentation of active tuberculosis (ICD9 codes 010–018) and human T-cell lymphotrophic virus (ICD9 codes 079.51 and 079.52). Maternal medical records and discharge ICD9 codes were searched for documentation of HIV/AIDS (ICD9 code 042 or V08), and drug abuse/drug dependence (ICD9 codes 304.0–304.2, 304.4–304.9, 305.1–305.3–305.7, 648.30, 760.71–760.73, 760.75). It was assumed that women with ICD9 codes for unspecified drug dependence (304.8,304.9, and 648.30) were using drugs included in the list of breastfeeding contraindications. Infant discharge ICD9 codes were reviewed for documentation of infant drug withdrawal (ICD9 code 779.5). Women with evidence of any breastfeeding contraindication from any source were considered to have a "documented" contraindication to breastfeeding. Women who reported in follow-up interviews that they did not initiate breastfeeding because they "didn't want to pass dangerous things" but who had no documentation of a contraindication were considered to have a "perceived" contraindication to breastfeeding. Women responding that they "didn't want to pass dangerous things" were not asked to name the substance or infection about which they were concerned, so it is unknown how many women with perceived contraindications had otherwise unidentified true contraindications.

Duration of breastfeeding was based on the number of weeks or months mothers reported breastfeeding at the 3–7 month interview (for those women who had quit before the 3–7 month follow-up interview) or at the 7–12 month interview (for those still breastfeeding at the time of the 3–7 month follow-up interview). Duration of breastfeeding was determined without consideration of whether or not mothers were combining bottle feeding with breastfeeding.

Of the 1802 mothers delivering on the sampled days, 518 were eligible for enrollment in the study. The most common reasons for ineligibility were residency outside the District of Columbia, (n = 704), being a private patient at site C (n = 373), and birth weight less than 1800 grams (n = 91). Four hundred fifty-two eligible mothers (87%) gave informed consent and were successfully enrolled. Of these, 395 women (87%) were successfully contacted for follow up at 3–7 months (and thus were interviewed about breastfeeding initiation) and 394 reported that they were still living with their infants since being discharged from the hospital. Information on breastfeeding initiation was missing for 1 woman, leaving 393 women for analysis of initiation rates. Three hundred sixty-nine women (82%) were successfully contacted for an additional follow-up interview at 7–12 months. Characteristics of the study population are detailed in Table 1.

Based on information abstracted from medical records, women who were eligible but not enrolled or who were enrolled but later excluded from the analysis (n = 125) were slightly more likely to have had one or more previous live births or to have had a vaginal delivery than women included in the final analysis (n = 393) (75% vs. 66% and 84% vs 77%, respectively). The two groups were not markedly different, however, with respect to race, age, marital status, type of medical insurance, mother's employment status, WIC enrollment, birth weight, gestational age at delivery, infant sex, or delivery hospital.

Of the 393 women seen for the 3–7 month interview, 201 (51%) initiated breastfeeding and 192 never breastfed. Sixty-one women (16% of the study population) had one or more documented contraindication to breastfeeding and of these, only 8 (13%) initiated breastfeeding. In contrast, among the 332 women without contraindications, 193 (58%) initiated breastfeeding. Of the 61 women with contraindications, 75% used cocaine, 28% had HIV infection, 5% had tuberculosis, 5% used PCP, and 3% used heroin or methadone. Ninety-four percent of women with documented contraindications had either HIV infection, used cocaine, or both.

Breastfeeding initiation rates by potential predictors of interest are summarized in Table 2. In adjusted analysis, factors found to be independently associated with initiation of breastfeeding (p < 0.05) were mother foreign-born, supportive friend available (both positively associated); presence of a documented contraindication, maternal tobacco use, and supportive relative available (all negatively associated). The two factors most strongly associated with breastfeeding initiation were mother foreign-born (AOR, 4.90; 95% CI, 2.38–10.10), and the presence of a documented contraindication (AOR, 0.19; 95% CI, 0.08–0.47) (Table 3).

Of the 192 women who never breastfed, the most common reasons reported for not breastfeeding were, "I did not want to" (48%), "I didn't want to pass dangerous things" (25%), and "Other reason" (11%). Less common were "Nipple and/or breast problems" (9%), "Schedule difficult" (4%), "I was ill/weak" (4%), "Baby refused" (3%), "I produced insufficient milk" (3%), "Baby was ill/weak" (1%), or "I was never taught" (1%).

Of the 48 women who didn't breastfeed because they "did not want to pass dangerous things," only 22 (46%) had a documented contraindication to breastfeeding. Conversely, of the 53 women with a documented breastfeeding contraindication who did not breastfeed, 22 (42%) reported that they "did not want to pass dangerous things." In all, 87 women (22% of those seen for 3–7 month follow-up) had a documented and/or perceived breastfeeding contraindication, and of the 192 women who did not breastfeed, 41% had a documented and/or perceived contraindication.

Out of 201 women who initiated breastfeeding, 3 had incomplete follow-up information. Of the remaining 198 women, 77% had quit breastfeeding by the 3–7 month follow-up interview and 91% had quit by the 7–12 month follow-up interview. Twenty-one percent of initiators quit in the first 2 weeks, 30 percent between 2 weeks and 2 months, 28 percent between 2 months and 4 months, and 21 percent breastfed 4 months or more.

Factors potentially associated with discontinuation of breastfeeding (p < 0.20) were identified in unadjusted survival analysis. Potentially protective factors identified included increasing maternal age, Hispanic ethnicity, high school education or greater, mother employed in the last 12 months, married or with a partner, one or more previous live births, vaginal delivery, pregnancy wanted, mother foreign-born, and infant birth weight ≥ 2500 grams; potential risk factors included presence of a documented contraindication, feeding choice discussed with a doctor or nurse, a grandmother living in the home, and supportive relative available. In adjusted analyses, factors independently associated with discontinuation of breastfeeding (p < 0.05) were mother foreign-born, increasing maternal age, and infant birth weight ≥ 2500 grams (all were protective against discontinuation) (Table 3). Presence of a contraindication was not associated with continuation in adjusted analysis. We found no significant interaction between age and ethnicity or between education and ethnicity.

Information on why women stopped breastfeeding was available for 181 of the 201 women who initiated breastfeeding (the reason for stopping was unknown for the 17 women who were still breastfeeding at the second follow-up and for the 3 women whose follow-up information was incomplete). The most frequently reported reasons for discontinuing were "Baby refused" (24%), "I produced insufficient milk" (23%), "Other reason" (21%) "Nipple and/or breast problems" (17%), and "Schedule difficult" (16%). Less common were, "I didn't want to" (7%), "I was ill/weak" (3%), "I didn't want to pass dangerous things" (4%), or "I was never taught" (0%). Reasons for discontinuation varied by duration of breastfeeding. Among women who breastfed less than 2 weeks, "Nipple and/or breast problems" was the most common reason cited, while among women who breastfed 4 months or more, "Schedule difficult"was most common (Table 4).

The rate of breastfeeding initiation in our study (51%) was slightly lower than rates described in two recently published reports that were drawn from population-based samples of women living in the United States. The breastfeeding initiation rates reported in these studies were 57% for 1993 24 and for 1995 12. We did not observe significant associations between breastfeeding and several factors previously reported to be predictive of breastfeeding initiation and/or duration, such as maternal ethnicity, education, socioeconomic status, employment status, marital status, parity, and type of delivery 121325. This may reflect our economically homogeneous study group. We did find a strong association between breastfeeding initiation and maternal foreign-birth, and between breastfeeding duration and maternal age, foreign-birth, and infant birth weight, which have been previously described 26272829303132. The association between maternal foreign birth and breastfeeding initiation and breastfeeding duration underscores the importance of cultural forces in infant feeding method-decisions. Interestingly, we also noted a negative association between breastfeeding initiation and the availability of a supportive relative and a positive association with the availability of a supportive friend. It may be that relatives discourage breastfeeding while friends are more encouraging. The attitude toward breastfeeding of the father or partner, which has been found to play an important role in infant feeding decisions 33, was not evaluated in this study.

The most common reason reported for not breastfeeding was "I didn't want to." This finding is consistent with a recently published study of breastfeeding initiation in which, in a national sample of women, the most common reason reported for not breastfeeding was "preferred to bottle feed" 12. However, in our study, the second-most common reason women reported for not breastfeeding was "I didn't want to pass dangerous things." In this high-risk population, "concern about passing dangerous things" appears to have played an important role in infant feeding method decisions.

It is not clear why some women with contraindications to breastfeeding did not express concern about passing dangerous things through breast milk when asked why they chose not to breastfeed. It is possible that these women had other more pressing obstacles to breastfeeding or that they were unaware that breastfeeding was contraindicated. Alternatively, in some cases the contraindication may have been resolved by the time of delivery.

Breastfeeding drop-off rates were high in our study-only 49% of women who initiated breastfeeding continued 2 months or more and only 21% continued 4 months or more. This is in contrast to the finding that 57% of initiators breastfeed 10 or more weeks in a population-based sample of women 24. Barriers to continued breastfeeding identified in our study included baby's refusal, a perception of insufficient milk production, and nipple and/or breast problems, which is also consistent with previous studies 293435. Contrary to what we observed in our analysis of breastfeeding initiation, we did not find that concern about passing dangerous things was important in decisions to discontinue breastfeeding.

Little is known about how women's perceptions of contraindications such as the use of certain medications and the presence of illness in the mother or infant influence infant feeding method decisions. In our study, we could find no documentation of any breastfeeding contraindications in over half of the women who reported that they didn't breastfeed because they "did not want to pass dangerous things." While these women may have had true (but undocumented) breastfeeding contraindication, this finding raises the possibility that misperceptions regarding breastfeeding contraindications may have contributed to low breastfeeding initiation rates in our study population. Additional studies are needed to determine to what extent misperceptions regarding breastfeeding contraindications contribute to infant feeding method decisions.

The primary strengths of this cohort study are the inclusion of information on breastfeeding contraindications and maternal perceptions of contraindications, detailed information on other potential risk factors for not breastfeeding, and estimates of duration of breastfeeding. Perhaps the most important limitation is its generalizability. Although women were selected to represent District of Columbia residents of lower socioecomonic status, our sampling scheme was not population-based. Because the prevalence of contraindications to breastfeeding such as HIV infection and substance abuse can vary widely between communities, our findings likely have the most relevance for high-risk communities. An additional limitation is that estimates of duration of breastfeeding were obtained retrospectively from interviews conducted at 3–7 months and 7–12 months postpartum, and so mothers may not have reported breastfeeding duration accurately. Finally, women in our study who did initiate breastfeeding were not asked if they were concerned about "passing dangerous things." Therefore, we were not able to directly evaluate the role of perceived contraindications as a predictor of breastfeeding practices.