|Home | About | Journals | Submit | Contact Us | Français|
To analyze the relationship between perinatal drug/alcohol use and maternal, fetal, neonatal, and postneonatal mortality.
Linked California discharge, birth and death certificate data from 1991–1998 were used to identify drug/alcohol-diagnosed births. Mortality relative risk (RR) ratios were calculated and logistic models were generated for mortality outcomes.
Among 4,536,701 birth records, 1.20% contained drug/alcohol discharge diagnostic codes (n = 54,290). The unadjusted RRs for maternal (RR = 2.7), fetal (RR = 1.3), neonatal (RR = 2.4), and postneonatal (RR = 4.3) mortality were increased for drug/alcohol-diagnosed births. After controlling for potential confounding, the odds of maternal death for cocaine use (OR = 2.15) remained significant as did amphetamine (OR = 1.77), cocaine (OR = 1.43), polydrug (OR = 2.01) and other drug/alcohol use (OR =1.79) for postneonatal mortality.
The association of cocaine use with maternal mortality and any drug/alcohol use with postneonatal mortality supports screening and identifying women using illicit drugs and alcohol during pregnancy. Increased collaboration with drug treatment programs and closer follow-up for drug-using women and their children may improve mortality outcomes.
There is a strong desire to improve the health of pregnant women and their infants by identifying factors that place them at greater risk and providing interventions to improve their outcomes. The use of illicit drugs and alcohol during pregnancy has been associated with both maternal and infant morbidity. Women who use illicit drugs during pregnancy have higher risks of placental abruption, premature rupture of membranes,1 and polyhydramnios,2 and their infants have higher rates of prematurity and low birth weight.3–7 These health outcomes, by themselves, have been associated with increased morbidity and mortality.8–12
Although mortality is only one measure of the health of pregnant women and their infants, it represents an important indicator. A national priority has been established to reduce maternal and infant mortality, as described in the Healthy People 2010 objectives.13 Studies on infant mortality have reported increased rates by race/ethnicity, birth weight, and geographic area.14,15 Others have analyzed trends in pregnancy-related maternal mortality and have found increasing rates with maternal age, for African-American women, and with increasing birth order.16
However, few studies have analyzed the contribution of drug and alcohol use on mortality. In a meta-analysis, Hulse et al.17 found a six-fold increase in the risk of neonatal mortality among women who used heroin late into pregnancy before entering methadone treatment. In an analysis of 179 births with five neonatal and 6 postneonatal deaths, the neonatal mortality rate for infants of drug-dependent mothers was 27.9/1000 births compared to 6.7/1000 in nondependent births; similarly, the postneonatal mortality rate was 61.5/1000 compared to 12.0/1000 in nondependent births.18 Another study reported no increased risk of postneonatal mortality or incidence of Sudden Infant Death Syndrome (SIDS) in cocaine-, opiate- or cannabinoid-exposed infants in the first 2 years of life.19 Researchers have also looked at the association between parental drug and alcohol use and child maltreatment, which demonstrated an increase in morbidity and mortality in children.20,21 In addition, alcohol use during pregnancy has been associated with increased morbidity and mortality in infants and children.21
Since mortality is a rare outcome, many studies suffer from small sample sizes and therefore, analyses may under- or overestimate the impact of drug use on mortality. The purpose of this study was to address this gap in the literature by analyzing the relationship between drug/alcohol use during pregnancy and perinatal mortality, utilizing a large population sample and a data set which combines the use of the birth certificate and hospital discharge data. Our analyses addressed whether drug and alcohol use during pregnancy was associated with an increased risk of mortality for the mother and the infant, and whether this relationship remained after adjusting for potential confounding variables.
A single data set was obtained from the California Office of Statewide Health Planning and Development (OSHPD), which included all births in California from 1991 to 1998. This data set combined information reported to the state about the mother and newborn from Patient Discharge data and the Birth Cohort file. The Birth Cohort file combines information from the Birth Statistical Master file, the Death Statistical Master file, and the Fetal Death Master file. The probabilistic methods used to link these data sets are described elsewhere22 Study procedures were approved by Institutional Review Boards from the State of California Health and Human Services Agency and the University of California, San Francisco. No individual identifying information was obtained or used in these analyses.
Patient Discharge Data included the demographics of the patient, dates of hospital stay, and medical diagnoses. The Healthcare Information Division establishes standard methods for collecting and reporting these data, and edits the information submitted. The data in this file were linked to the Birth Cohort file for all years except 1998. For 1998, data were linked with the individual birth, death and fetal death files separately to include as much information as possible. The combined data set contained information from birth and through the first year of life to record any deaths that occurred.
Drug and/or alcohol use by the mother during pregnancy was identified using discharge diagnostic codes. In all, 18 International Classification of Diseases, 9th Revision Clinical Modification (ICD-9 CM)23 codes that are associated with drug/alcohol abuse or dependency were used to identify women (e.g. cocaine dependence, cocaine use) and five codes for infant exposure (e.g. cocaine, narcotic). Infant drug categories excluded anesthetic and analgesic drugs administered during labor and delivery. The drug codes were combined for the mother and infant in order to identify drug/alcohol diagnoses per dyad (e.g. cocaine diagnosed in mother and infant equals cocaine). Drug categories included in the analysis were amphetamine, cocaine, polydrug, and others. Polydrug use was defined as any combination of two or more drugs (including alcohol) in either the mother or infant or different drugs diagnosed in the mother and infant (e.g. cocaine diagnosed in mother, narcotic diagnosed in infant). The other drug category included alcohol, barbiturates, cannabis, hallucinogens, opioids, and nonspecific drugs of dependence. These substances were combined for analysis because of the small number of deaths, for each substance. Tobacco use, as reported on the birth certificate as a complication of pregnancy, was included in the analysis as a separate variable.
A race/ethnicity variable was created by combining race with Hispanic ethnicity. Pacific Islanders, Native Americans/Eskimos, Asians and others were combined into one category in the analysis because of the small number of deaths. Maternal age was categorized as ≤ 19 years, 20 to 34 years, and ≥ 35 years. Educational level was divided into three categories: less than or equal to 11 years, high school graduate, or 13 or more years.
Pregnancy-related variables included the month prenatal care began (grouped by trimester) and the number of prenatal visits. Birth weight was measured as a continuous variable as well as a categorical variable (low birth weight, ≤2499g, and normal birth weight, ≥2500g). Maternal death, coded on the birth certificate as a complication of delivery, was defined as a death which occurred within 72 hours of delivery. Fetal deaths were defined as deaths which occurred after 22 weeks gestation or a birth weight >400 g. Neonatal deaths were those which occurred within the first 28 days of life, and postneonatal deaths were those that occurred within the first year of life.
The total data set consisted of 4,536,701 birth records. A total of 54,290 births were identified over the 8-year period in which drug/alcohol use was noted in either the maternal, infant, or both sets of diagnostic codes. The percentage of births with drug and/or alcohol diagnostic codes each year in California ranged from 0.95% of births in 1991 to a high of 1.43% in 1994. This translates into a rate of 9.5 drug/alcohol diagnosed births per 1000 births in 1991 to a high of 14.3 births in 1994. The overall rate of drug/alcohol diagnosed births for the 8 years was 12 per 1000 births, or 1.2%.
Cocaine, defined as the only drug identified in mother and/or infant, was present in 30% of births, followed by opioids (21%), polydrug use (13%), and amphetamines (11%). Alcohol dependence/abuse was the only diagnosis in 4% of the deliveries and alcohol plus drug use was identified in 5% of deliveries. Tobacco use was noted on the birth certificate in 2% of the women who were not identified as having used drugs and/or alcohol during pregnancy but in 17% where drug and/or alcohol use was identified by diagnostic codes.
There were a total of 1944 maternal deaths, 28,011 fetal deaths, 19,136 neonatal deaths, and 10,926 postneonatal deaths during these 8 years. There were 62 maternal deaths, 430 fetal deaths, 529 neonatal deaths, and 536 postneonatal deaths for drug- and alcohol-exposed births.
As the large sample size would make even trivial differences statistically significant, we relied on point estimation and confidence intervals (CI) to analyze the data. To assess the relative risk (RR) of maternal, fetal, neonatal, and postneonatal deaths, the following rate calculations and CI (95%) were computed: Incidence of deaths in nondiagnosed births (Ind) = # deaths in nondiagnosed/# of births in nondiagnosed; incidence of deaths in diagnosed births (Id) = # deaths in diagnosed/# diagnosed births, and unadjusted risk ratio = Id/Ind. The RRs of maternal, fetal, neonatal, and postneonatal death were then further stratified by race/ethnicity, maternal age, use of tobacco during pregnancy, trimester prenatal care began, and birth weight group. Logistic regression models were estimated to explore the relationship between all potential confounding variables and mortality outcomes. Variables included in the multivariate models were those factors thought to be associated with the outcomes, either from the literature or from clinical experience.
The demographic descriptors for nondiagnosed and drug/alcohol-diagnosed births are reported in Table 1. African Americans and White non-Hispanics are disproportionately represented in the drug/alcohol-diagnosed group. Similarly, Asians, White Hispanics, and other race/ethnic groups are under-represented in the drug/alcohol-diagnosed group relative to their proportions in the nondiagnosed group. However, 39% of all drug/alcohol-diagnosed births were to White non-Hispanics, with this group representing the largest proportion of drug/alcohol-diagnosed births. A higher proportion of drug/alcohol-diagnosed births were to mothers who had 12 years or less years of education. Mean maternal age was similar in both groups. Drug/alcohol-diagnosed births were more likely paid for by Medicaid.
In further comparisons (not in table), a higher proportion of women delivering drug/alcohol-diagnosed infants had no prenatal care (14 vs 1%), or began prenatal care in the second (28 vs 17%) or third trimester (10 vs 4%); consequently, they also had fewer prenatal visits (7.2 vs 11.1, t = 164.09, p<0.01). The average length of gestation was similar with exposed infants having a mean gestation of approximately 8.9 months compared to 9.2 months for the nonexposed. In drug/alcohol-diagnosed births, there was a higher proportion of very low birth weight (defined as ≤ 1499 g; 4% vs 1%) and low birth weight (defined as 1500 to 2499 g; 20 vs 5%) infants. The mean birth weight for the nondiagnosed group was 3360 g (SD = 591.7) and for the diagnosed group, 2905 g (SD = 709.8).
The unadjusted and stratified risk ratios comparing mortality outcomes between nondiagnosed births and drug/alcohol-diagnosed births appear in Tables 2 and and3.3. Neonatal and postneonatal death rates are calculated based on the rate per 1000 live births. The total number of deaths used in the calculations is also reported.
As shown in the second row of Table 2, the unadjusted RR of maternal death is 2.7 times greater if drug/alcohol use is diagnosed. However, the RR of maternal death is different when stratified by potential confounding variables. The increased risk of maternal death is less among drug/alcohol users who smoke (RR = 2.2). The risk of maternal death is significantly greater for White Hispanics (RR = 3.4) and women aged 20 to 34 years (RR = 3.2) with a drug/alcohol diagnosis. However, the RR of maternal death is less for drug/alcohol-diagnosed births if African American (RR = 2.1), and White non-Hispanic (RR = 2.1), or if looking at when the trimester prenatal care began (ranging from RR = 1.8 for no prenatal care to RR = 2.2 for initiation of care in the second trimester). Although the RR of maternal death was significantly elevated for other ethnicity groups, the CI was wider because of the smaller number of deaths.
The unadjusted RR of fetal death is 30% higher (RR = 1.3) for births with a drug/alcohol diagnosis during pregnancy than for nondiagnosed births (Table 2). Stratifying by race/ethnicity, maternal age, tobacco use, trimester prenatal care began, and birth weight demonstrated confounding of the fetal mortality risk. There was an increased risk of fetal death among drug/alcohol-diagnosed births as maternal age increased (RR = 1.2, 1.6, 1.9, respectively), with late initiation of prenatal care (RR = 1.5) and among infants born with birth weights greater than 2500 g (RR = 1.7). The risk of fetal death was less for African Americans (RR = 0.7) and other ethnicities (RR = 0.7) who had a drug/alcohol diagnosis. Tobacco use and birth weight also produced lower risks of fetal death (RR = 0.7 for tobacco use, RR = 0.4 for low birth weight) for drug/alcohol-diagnosed births.
The unadjusted RR of neonatal death was 2.4 times higher among drug/alcohol diagnosed-births (Table 3). This relationship varied by race/ethnicity, maternal age, tobacco use, trimester prenatal care began, and birth weight group. An increased risk of neonatal death with drug/alcohol-diagnosed births, after stratifying, was noted for White non-Hispanics (RR = 2.5), and maternal age group 20 to 34 years (RR = 2.7).
The unadjusted RR of postneonatal death was 4.3 times greater among drug/alcohol-diagnosed births (Table 3). This relationship also varied by all stratified variables. The variables with an increased risk of postneonatal death for drug/alcohol-diagnosed births, once stratified, were for Other race/ethnicity (RR = 4.8), women aged 20 to 34 years (RR = 4.7), and ≥35 years (RR = 5.3), women who had not used tobacco during pregnancy (RR = 4.4), and women who began prenatal care in the first trimester (RR = 4.4).
Odd ratios were calculated to examine the relationship between drug use and mortality (Table 4). The odds of maternal death were significantly increased for all drug categories compared to no drug use during pregnancy. The odds of fetal death were significantly increased only for amphetamine use and polydrug use, whereas the odds of neonatal and postneonatal mortality were significantly increased for all drug categories.
To assess whether drug/alcohol-diagnosed births were at higher risk for maternal, fetal, neonatal, and postneonatal mortality, given the results of the univariate analyses, four models were tested using multivariate logistic regression in order to adjust for potential confounding. Variables included in the models were not highly correlated (Pearson correlation coefficients ranging from r = 0.001 to 0.45). The highest correlations were between health insurance coverage and age, ethnicity/race, and educational level, therefore, health insurance coverage was not included in the analyses (see Table 5).
In the first model, the odds of maternal death were doubled (odds ratio (OR) = 2.15) for women who used cocaine during pregnancy, by 92% for women who smoked tobacco, and three times more likely if women received no prenatal care, even after controlling for other factors. Race/ethnicity, maternal age group, and education level were not significant once all other factors were included in the model. In the second model, the odds of fetal death were significantly increased for women aged 35 years or greater (OR = 1.29), African American (OR = 1.12), White Hispanic (OR= 1.06), less than 12 years education (OR = 1.11), and no prenatal care (OR = 2.18), even after controlling for other variables. Low birth weight infants were 34 times more likely to have a fetal death (OR = 34.7), even after controlling for other factors. However, drug use, adolescents, other race/ethnicities, and tobacco use have a significant protective effect against fetal death, once controlled for other variables.
In the third model, only four variables significantly increased the odds of neonatal death, after controlling for all other variables: African American (OR = 1.16), less than 12 years education (OR = 1.13), no prenatal care (OR = 2.25), and low birth weight (OR = 48.3). Other ethnicities, cocaine, polydrug, and other drugs/alcohol were significant but had a protective effect. Maternal age group, White Hispanic ethnicity, amphetamine, and tobacco use were not significant. In the fourth model, the odds of postneonatal death were significantly increased for adolescents (OR = 1.33), African Americans (OR = 1.60), less than 12 years of education (OR = 1.40), tobacco use during pregnancy (OR = 1.80), no prenatal care (OR = 1.85), low birth weight (OR = 6.34), and for all drug categories, even after controlling for all factors. For women aged 35 years or greater (OR = 0.89) and White Hispanic women (OR = 0.81), there was a significant protective effect.
Results suggest that illicit drug and alcohol use during pregnancy may be independently associated with increased risk of postneonatal mortality and that cocaine use may be independently associated with increased risk of maternal mortality. Drug and alcohol use was not associated with an increased risk of fetal or neonatal death. Low birth weight and lack of prenatal care are significantly associated with mortality, whereas tobacco use is only associated with increases in maternal and postneonatal death in our study.
Few studies have been able to study drug use and maternal mortality, but there have been reports of maternal death due to cocaine intoxication.24 Studies of perinatal mortality related to cocaine use have often found a protective effect in subgroups25,26 or comparable survival rates.27 Other studies continue to show a relationship between drug use and increased risk of perinatal mortality.4,28 The relationship between drug use and fetal and neonatal mortality may be related to improvements in the medical management of high-risk mothers and infants in hospitals, a factor which was not analyzed in this study.
The apparent protective affect of smoking noted in our findings reflects a paradoxical relationship noted previously.29 Smoking has been shown to decrease birth weight30 and yet, some studies have shown that among smaller infants, mortality rates are lower for smokers than nonsmokers.31 Others have found that the effect of tobacco on birth weight is the mechanism by which it affects mortality.32 The relationship between drug and tobacco use on mortality may require further research concerning the contributory role of birthweight.
Although our results suggest an association between drug use and maternal and postneonatal mortality, it is likely that drug/alcohol use is a surrogate for a constellation of many factors which may influence mortality, for example, decreased utilization of prenatal and pediatric care, poor or unstable housing, poor nutrition and exposure to violence.3 Consistent with our findings, others have noted a relationship between maternal drug and alcohol use during pregnancy and increased postneonatal mortality.18,20,21
A limitation of this study was the reliance on diagnostic codes, which may underestimate the number of drug- and alcohol-exposed births. In a large 1992 drug screening study, 11.35% of women presenting in California hospitals for delivery tested positive for illicit drugs or alcohol.33 This is approximately 10 times greater than the percentages reported in this study. The diagnostic code is dependent on the provider identifying the diagnosis in the medical record. It is possible that this data set identifies only the most severe cases where there is a drug and/or alcohol problem and therefore overestimates the relationship between mortality and drug/alcohol use.
A further limitation is the accuracy of the information contained on the birth certificate, which identifies many of the health and behavioral risk factors related to pregnancy (e.g. maternal death, tobacco use). The risks associated with increased mortality may be biased because of misclassification leading to an over- or underestimation of the relationship. However, birth certificates provide a valuable source for monitoring the health of women and infants. Improved documentation of maternal risk factors (e.g. tobacco use, drug use, alcohol use) on birth certificates would increase the value of this information for identifying risks and providing interventions.
Findings support the importance of identifying women with substance use problems during pregnancy and of providing interventions that encourage treatment in order to prevent negative outcomes for the women and their infants. Kogan et al.34 studied the advice given to 8310 women during pregnancy and found that a large percentage of women did not receive any counseling about smoking, drinking, or illicit drug use. Another study noted that physicians were unaware of substance abuse problems in 45% of their clients who were presenting at drug treatment programs.35 Physician satisfaction with dealing with substance using clients is often influenced by negative experiences;36 however, these can be countered with improved training in methods to assess and intervene with this population.
In addition, gaps in interorganizational collaboration may limit the comprehensive care that is needed to address the multiple problems that often face women. Relationships between pediatric and obstetrical providers and drug treatment programs can be strengthened. Providers can offer support and monitoring through links with public health nursing in order to ensure the health and safety of the women and their children. Providers may be more willing to identify drug and alcohol problems if there are improved connections with services that can address the needs of these women and infants.
Illicit drug and alcohol use among pregnant women warrants continued efforts to track the prevalence of the problem and to continue to develop innovative approaches to improve the health and well being of women, children, and families. Although more research is needed to study the direct effect of illicit drugs on pregnancy and children, it appears that maternal drug and alcohol use is associated with higher mortality rates for pregnant women and infants.
Support was provided by Jan Morgan and Beate Danielsen from the California Office of Statewide Health Planning and Development. This study was also funded by the California Policy Research Center, the National Institute on Drug Abuse San Francisco Treatment Research Center (Center Grant P50 DA09253), and Treatment on Demand Grant (R01 DA12221).