Ninth Annual Maternal and Child Health Epidemiology Conference / December 10-12, 2003
Increasing Infant Mortality Among Very Low Birth Weight Infants - Delaware 1994-2000
MARCI L. DREES: Thank you very much and good afternoon. This graph shows Delaware’s infant mortality rates in blue compared to the U.S. average in yellow. Both are calculated as three-year moving averages. Throughout the 1980s, Delaware had higher than average infant mortality; but this improved during the early ‘90s, reaching the national average around 1993. However, beginning in 1996, while the U.S. average continued to improved, Delaware’s IMR again worsened. For the most recent period, 1999 to 2001, Delaware’s IMR was 9.2 compared to the U.S. average of 6.9. Delaware now ranks sixth highest nationally for infant mortality. To address this problem, we investigated three hypotheses of the reason for the increase: that it was due to an increase in plurality, that it was due to an increase in the number of low birth weight infants, or that it was due to an increase in birth weight-specific mortality. We analyzed Delaware’s linked birth/death certificate cohort data from 1994 to 2001. This database includes all live births to mothers residing in Delaware and is quite complete. We compared two three-year time periods. As baseline, we used 1994 to ’96, when Delaware’s IMR was at its lowest rate of 7.1. The most recent three-year period for which data is available is 1999 to 2001, when the IMR was 8.8. There were approximately 30,000 live births and 250 infant deaths during both of these time periods. We adjusted the IMR for plurality and for birth weight distributions. Stratified analyses examined IMR trends within plurality and birth weight subgroups and according to maternal characteristics.
This table shows the change over time in plurality and birth weight distributions. Rates of twin births increased from 2.8 to 3.4 percent. Rates of low birth weight also increased. Very low birth weight increased from 1.7 to 1.9 percent. This graph shows Delaware’s unadjusted infant mortality rate. Because of concern about the rising multiple birth rate, we wanted first to make sure that Delaware’s increase in IMR couldn’t simply be explained by the increasing frequency of multiple births. Therefore, we directly adjusted the rate for plurality distribution, assuming that the multiple birth rate had remained the same, as in ’94 to ’96, and found very little change. This indicates that the increasing IMR is not due to the rising incidence of multiple births. To evaluate our second hypothesis that the increased IMR was due to an increase in low birth weight, we then performed a similar adjustment for birth weight distribution. If the rates of low birth weight had remained the same, as in ’94 to ’96, the IMR still would have risen from 7.1 to 8.2. The increasing frequency of low birth weight only accounted for approximately one-third of the increase in IMR. We then turned to our third hypothesis, birth weight-specific mortality, to attempt to explain the remaining two-thirds of the increase.
The birth weight categories are shown on the X-axis, and deaths per thousand live births on the Y-axis. The ’94 to ’96 period is shown on the left in yellow, and in the ’99 to 2001 period on the right in orange. In this and all the following slides, the asterisk denotes a P-value of less than .05 for the trend between the two time periods. Mortality rates for normal weight and moderately low birth weight infants showed no significant change, but rates for very low birth weight infants rose 33 percent from 235 to 312 per thousand. We then wanted to take this trend into account for our adjusted infant mortality rates. This slide again shows our previous univariate adjustments. In order to evaluate together the increasing frequency of low birth weight and the increased mortality among very low birth weight infants, we kept the adjustment for birth weight distribution, again shown in blue, but then also assumed that the IMR among very low birth weight infants had remained a 235 per thousand and found that when both these trends were taken into account, the overall infant mortality rate remained stable. This suggests that it is both the increasing frequency of low birth weight, but particularly the increased mortality among very low birth weight infants that is causing Delaware’s IMR to increase. We then examined these very low birth weight infants more closely. We first looked at plurality distribution and found that all plurality subgroups had increasing mortality.
Singleton infants had a 14 percent increase in mortality while twins had a 96 percent increase. Triplet and higher order births had an increase of over 450 percent. Very low birth weight and early gestational age are highly inter-correlated. We found that all of the very low birth weight infants who died were born at less than 37 weeks gestation, and the increasing mortality occurred in those that were born at less than 28 weeks. We then analyzed the infant mortality trends according to maternal characteristics. First, we looked at maternal age and found that infant mortality increased among all maternal age categories. Women 30 years of age or older had the only significant increase of 109 percent. Analysis by mothers’ marital status revealed that single women’s infants’ mortality did not change significantly, but the mortality rates of infants of married women rose 115 percent. We found that the mortality rates for infants of all maternal education levels increased, but the rate of increase was greatest for women with the highest education level. Among women with at least one year of college education, infant mortality rates rose by 84 percent.
This graph shows the geographic differences among Delaware’s three counties and the city of Wilmington. Infant mortality rates in Kent and Sussex counties, which are primarily rural, and within Wilmington, which is the only urban area in the state, did not change significantly. However, in the remainder of Newcastle County excluding the city of Wilmington, infant mortality increased 138 percent. This region is primarily suburban. This graph shows the source of payment for medical care divided into self-pay, Medicaid, and private insurance. Infants of privately insured women demonstrated a 72 percent increase in mortality. We looked at whether or not mothers had initiated prenatal care in the first trimester as a gauge of health care access. The mortality rate for infants of mothers who started care early--on the right--increased by 39 percent, whereas the IMR for those who didn’t showed no significant change. To summarize, analysis of maternal risk factors suggested greater increases in mortality rates among infants born to women of higher socioeconomic status. While we have more direct measure of income in our data, the characteristics that I’ve described tend to correlate with higher SES.
The next question, of course, is what is the underlying cause of this increased mortality? Given that the affected population appears to be more affluent, one concern has been that this increase is related to the national trend toward delayed childbearing and the resultant problems with infertility and risks of treatment for infertility. Changes in perinatal practices may affect some subpopulations more negatively than others. For example, if more inductions of labor are occurring among certain groups of women, that might affect infant outcomes. Another possibility is that there has been a shift from fetal mortality to neonatal mortality. This could be either due to a change of classification or in fetal survival. Because of the concern about increased multiple birth rates, early on the question was raised of whether assisted reproductive technology, or ART, was responsible. Aside from the risk of multiple births, ART has been associated with low birth weight, preterm delivery, maternal complications, and other risks. While as I discussed earlier, we found that the rising multiple birth rate itself was not responsible, given the increased mortality among very low birth weight multiple births and the maternal demographics I described, we felt that it was important to try to address this issue. Therefore, we looked at data from the U.S. Registry of ART Procedures.
Similar to the rest of the U.S., there was a 13 percent increase in ART procedures performed in Delaware from 1996 to 2000, and use also increased in surrounding states. ART success rates also increased, resulting in the births of more ART-conceived infants. For babies born in 2000, we estimate that one percent of Delaware’s birth cohort overall and 7.5 percent of very low birth weight infants were conceived with ART. Of multiple births, 56 percent would be expected from national conception. This estimate was based on maternal age-specific multiple birth rates from 1971, prior to the widespread use of ART and ovulation-inducing drugs. Based on births reported to the ART registry, 16.8 percent of Delaware’s multiple births were due to ART. The remember were likely due to other infertility treatments. These estimates are also similar to the nation as a whole. Of course, ART is not the only infertility-related issue that could be contributing. Other infertility treatments, such as ovarian hyper-stimulation alone, have been linked to low birth weight and, of course, multiple births. Unfortunately, the use of these medications is not well monitored or well studied. The underlying biology of women who are infertile may likely also be contributing. Some of these women may simply be less capable of maintaining a healthy pregnancy.
This is supported by evidence of higher rates of low birth weights among sub-fertile couples who take longer than one year to conceive but do not undergo any infertility treatment. To examine the issue of fetal survival and classification, we examined the birth weight distribution in more detail. In this graph, birth weight categories are shown on the X-axis and percent of all births on the Y-axis. We wanted to make sure that the increase in mortality wasn’t simply caused by an increase in the tiniest, least-viable infants of less than 500 grams. As you can see, increases in all three birth weight subgroups comprised the total increase in very low birth weight births of 0.2 percent. We then looked at the mortality rates among these detailed birth weight categories. IMR increased among all three very low birth weight subpopulations. Although less than 500-gram infants make up a tiny proportion of all births from Delaware, their extremely high, and increasing, mortality may be contributing to the overall increase in IMR. We also examined when these infants were dying. This graph shows the proportion of Delaware’s infant deaths, regardless of birth weight, on the Y-axis, grouped by age at death on the X-axis. As you can see, an increasing proportion of infants died within their first week of life. In fact, this increase is nearly all accounted for by deaths on days zero or one of life.
This suggests either a change in fetal death classification or in severity of illness at the time of birth. We then examined fetal mortality rates. Unfortunately, we were unable to examine Delaware’s fetal death rate data directly because of poor data quality. However, we were able to examine fetal deaths from one hospital, Christiana, which accounts for around half of all births to Delaware mothers. This hospital is also the main referral center for high-risk obstetrics and has the only high-level neonatal care unit in the state. It is also the primary hospital surveying Newcastle County and the city of Wilmington. This graph shows infant and fetal mortality rates for Delaware residents born at Christiana. The hospital’s infant mortality rate, in yellow, has been increasing in parallel to that of Delaware overall. Concurrently to the increase in infant mortality, there has been a decrease in fetal mortality. This also suggests that there has either been changes in fetal death classification and reporting or an improvement in fetal survival but with subsequent neonatal death. This study is subject to at least four limitations. Delaware is a small state, and some trends must be interpreted with caution because of very few infant deaths among certain subpopulations. Our data is limited to that which is recorded on the birth or death certificate and factors such as ART use are not recorded. Therefore, we were unable to link ART to infant mortality directly.
Women’s fertility history as well as other infertility treatments are also not recorded on the birth certificate, and thus we were unable to assess their effect. And while timing of initiation of prenatal care could be evaluated, the quality and content of perinatal care could not; and therefore, we could not assess the potential role of changing practices among different groups of women. In conclusion, Delaware’s increasing infant mortality was primarily due to increasing mortality among very low birth weight infants and secondarily to increasing frequency of low birth weight births. The maternal profile suggested that infants born to women of higher socioeconomic status had the largest increase. It is likely that the etiology behind the increased mortality among very low birth weight infants in Delaware is multifactorial. Supplemental data from an ART registry was consistent with the hypothesis that infertility and/or infertility treatments may be contributing to the increase, but further study is needed to directly assess the contribution. Evidence suggests that changes in the classification and reporting of fetal deaths have also contributed to the increase in neonatal mortality, and improvement in fetal survival may also be playing a role. There continues to be political pressure not only to find definitive answers to this recent infant mortality trend that I’ve described today, but also to address the long-standing high infant mortality rates among Delaware’s high-risk groups: women who are young, single, Black, poor, and uninsured.
The Delaware Division of Public Health together with partners in the state Perinatal Board and CDC’s Division of Reproductive Health will continue research to determine the best ways to investigate this complex problem. Current plans include more detailed analysis of state vital statistics data as well as other data sources such as hospital discharge data to look at issues such as maternal complications and pre-pregnancy health. Christiana Hospital is conducting independent research looking at illness severity among their birth cohort. And another consideration is to add method of conception to the birth certificate so that you can directly assess the impact of infertility treatment on infant mortality. This, however, will require an act of the Delaware State Legislature. I’d like to thank all the partners who collaborated with this project. Thank you very much.