The Future of Children, Princeton - Brookings: Providing research and analysis to promote effective policies and programs for children.

If birth weight is a valid indicator of the intrauterine experience and thus, indirectly, of the effectiveness of prenatal care services, infant mortality (death within the first year of life) is an indicator of the quality and impact of obstetric and neonatal care services. Because, in countries like the United States, most children who die in the first year of life do so in the first month, those factors that influence the birth process and the immediate postnatal condition of the baby may be expected to have a discernible impact on the infant mortality rate. To appreciate what the receipt of obstetric and neonatal care means for the health status of children, we begin with a look at infant mortality.

Much progress has been made in reducing the U.S. infant mortality rate during the twentieth century. It fell from 100 deaths per 1,000 live births in 1915 to 9.1 deaths per 1,000 live births in 1990.²¹ Improvements in medical care and advances in medical technology played an important role in this process, particularly the introduction of sulfonamide and other anti-microbial drugs in the period from 1935 to 1950²² and advances in neonatology since the late 1960s.^23,24 Until recently, the only interruption in the downward trend in infant mortality occurred in the late 1950s and early 1960s.²⁵

There are two components of infant mortality: neonatal mortality and postneonatal mortality. The former refers to deaths of infants before the first 28 days of life, while the latter refers to deaths of infants between the ages of 28 and 365 days. Neonatal deaths are usually caused by congenital abnormalities, conditions associated with prematurity, and complications of delivery; while postneonatal mortality results primarily from sudden infant death syndrome (SIDS), infectious diseases, and accidents. In the early 1900s, approximately 70% of all infant deaths occurred in the postneonatal period. By 1989, only 36% of infant deaths occurred in the postneonatal period. Unlike earlier in this century, recent trends in the infant mortality rate have been dominated by declines in the neonatal mortality rate. Today, the mortality rate (that is, the chance of dying) for a baby's first month of life is more than twice as high as the mortality rate during the remainder of the first year.

Neonatal mortality, in turn, is highly negatively correlated with birth weight so that the birth weight distribution of a population is a major determinant of its neonatal mortality rate. Those populations with higher proportions of low (less than 2,500 grams) and very low (less than 1,500 grams) birth weight babies have higher rates of neonatal mortality. Once the distribution of birth weights in a population is taken into account, the mortality rates specific to each birth weight category are determinants of overall neonatal mortality. By multiplying the proportion of births in each birth weight group by the birth weight specific mortality rate for that group, one arrives at the overall neonatal mortality rate for a population.

Despite the rapid decline in infant mortality, substantial differences in this outcome and in its most proximate determinant—low birth weight—have persisted between different groups over time.^26,27 The most notable of these is the excess mortality of black babies who presently experience an infant mortality rate twice that of their white counterparts. In addition, the black low birth weight rate has typically been more than twice as large as the white rate. Among the Hispanic population, Puerto Rican women give birth to LBW infants at approximately three-fourths the rate of blacks, while the low birth weight rates of other Hispanics are similar to those of white non-Hispanics.²⁸

In addition, the U.S. infant mortality rate remains higher than those of a number of other developed countries even when the rate is limited to whites.¹⁷ Finally, Marks and others report that in 1980 "differences between the States of the U.S. in infant . . . mortality are greater than those between the U.S. and the countries of Scandinavia [with the lowest infant mortality]."²⁹

These facts have led some observers to conclude that U.S. infant mortality policy has reached a crossroad.³⁰ Two decades of dramatic declines in infant mortality have failed to improve the survival prospects of black infants relative to white infants or of white infants born in the United States relative to those born in other developed countries. Moreover, during the decade of the 1980s, tentative evidence emerged that the period of rapid decline may have ended. The fall in mortality slowed to an average year-to-year reduction of 2.5%,³¹ and the incidence of low birth weight stopped declining. In 1984 the fraction of LBW births reached an all-time low for both races: 5.59% for whites and 12.36% for blacks. The corresponding figures in 1989 were 5.7% for whites and 13.5% for blacks.^16,32

The infant mortality rate in 1990 was 6% less than the rate in 1989.³³ The introduction of treatment with artificial surfactant, which helps prevent deaths of premature infants from respiratory distress syndrome, and the expansions in Medicaid coverage for prenatal care may have contributed to this decline.³¹ Obviously it is too early to determine whether this change represents a reversal of the deceleration in the downward trend during the 1980s. It is clear, however, that three important 1990 infant health goals of the Public Health Service as set forth in The 1990 Health Objectives for the Nation: A Midcourse Review have not been met.¹⁰ These are an infant mortality rate no higher than 12 per 1,000 live births for any minority group (the black rate was 17.6 per 1,000 live births in 1988), a low birth weight rate of 5% of all infants, and a low birth weight rate of 9% for any minority group.

Where a child is born, how that child is delivered, and who attends the delivery may all be important determinants of neonatal outcome under certain circumstances. The volume of obstetrical procedures in the hospital where birth occurs, for example, has been shown to be related to survivorship in low birth weight babies though not in term, normal birth weight babies.³⁴ Recent trends indicate that increasing numbers of infants in the United States are being born in the hospital. For white infants the percentage of babies born in a hospital rose from 96.6% in 1975 to 98.7% in 1989, while for nonwhite infants the comparable percentages were 94.6% in 1975 and 99% in 1989.³⁵

In addition to in-hospital deliveries, cesarean section rates have also been on the rise since the middle of the 1960s. In 1965 the chances of being born via cesarean section were about 1 in 20. By 1989 the likelihood had increased to 1 in 4. Most authors attribute this rise to the increase in electronic fetal monitoring, changes in the legal atmosphere surrounding obstetrical care, and the policy of discouraging spontaneous vaginal deliveries in women who have undergone previous cesarean sections.³⁶ The increase in these procedures has not been shown to have any positive contribution to neonatal outcome.³⁷

As we noted earlier, the most important component of infant mortality in the United States today is the death rate of babies less than 28 days old. This rate, in turn, depends upon the distribution of birth weights in a population and the specific death rate associated with each birth weight category.

In the first part of this paper we discussed some of the factors that influence the birth weight distribution of a population, particularly the provision of adequate prenatal care for pregnant women. We now turn our attention to developments that have influenced birth weight specific mortality rates. The most important of these has been the expansion in the availability of neonatal intensive care services.

Over the past 2 decades in the United States, the single most significant trend in the delivery of neonatal health services has been the emergence of regionalization as a guiding principle in the provision of care for high-risk mothers and newborn babies.³⁸ Many hospitals began to introduce intensive care units for newborns during the decade of the 1960s, but it was not until the early 1970s that a comprehensive set of guidelines was developed to organize the distribution of these services within given geographic areas by incremental levels of intensity.

The 1977 National Foundation—March of Dimes recommendations,³⁹ subsequently refined by the American Academy of Pediatrics and the American College of Obstetricians and Gynecologists,⁴⁰ include three levels of infant care: (1) Level I nurseries to provide routine newborn care; (2) Level II nurseries, manned by board certified pediatricians, to provide care to seriously ill infants with life-threatening conditions for finite periods and to maintain ongoing liaisons with (3) Level III nurseries which serve as regional referral centers offering the most sophisticated neonatal care including specialized respiratory support systems, pediatric surgery services, and infant transport teams capable of delivering seriously ill neonates to and from the hospital.

In practice there has been some blurring of the distinction between Level II and Level III nurseries as the competition for patients has intensified in recent years.⁴¹ Nevertheless, specific standards have arisen for adequate staffing, physical design, ancillary services, and transport facilities appropriate to the most advanced levels of neonatal care.^40,42

During the decade of the 1980s, the number of Neonatal Intensive Care Units (NICUs)—combined Level II and Level III units—in U.S. hospitals has risen dramatically (see Figure 3). While the total number of hospitals, as surveyed by the American Hospital Association in their annual survey, rose 14.33% from 1979 to 1988, the number of NICUs climbed 44.33% from 485 to 700 and the number of NICU beds increased 66.97% from 6,591 to 11,005 during this period.⁴³

Although no national data bases exist which provide information on the utilization rates of these NICUs, the Office of Technology Assessment (1987) estimated that there were between 150,000 and 200,000 admissions to NICUs annually in the mid-1980s.⁴⁴ In recent years, anecdotal evidence suggests that a greater and greater proportion of NICU admissions have gone to infants weighing less than 1,500 grams while the proportion of admissions of moderately low birth weight infants has declined.⁴⁴ One study from Cleveland reported increases in the numbers of extremely low birth weight admissions (babies born weighing less than 751 grams) from the early to the late 1980s of 30%.⁴⁵ Whether these trends result from recent tendencies to focus more aggressive resuscitative efforts on smaller and smaller neonates remains to be fully elucidated. Whatever the origin of this trend to lower birth weight NICU admissions, this development inevitably forces policymakers to confront the question of how much improvement in health outcomes can reasonably be anticipated from such high tech attention to smaller and smaller babies.

In view of the rapid growth in the number of NICUs during the past 15 years, the substantial expense attendant upon caring for newborns in such environments, and the trend to lower birth weight NICU admissions, it was inevitable that questions would arise as to whether the expenditure of such prodigious resources was worthwhile. (See the Lewit and Monheit article in this journal issue.) Clinicians, epidemiologists, and economists have all addressed this issue, with each group approaching it from the vantage point of its own particular discipline.

As practitioners, neonatologists have been anxious to demonstrate that the specific content of care provided by NICUs makes a difference in the outcome of patients treated there. To make this case, investigators have applied two different methods. One approach compares the experiences of successive cohorts of infants treated at the same institution during two time periods to show how the development of new technologies has improved birth weight specific survival rates for all but the very smallest babies.^45,46 Increasing willingness to refer high-risk pregnancies to centers with neonatal intensive care units, the antenatal use of steroids to hasten fetal lung development, improvements in techniques to assist breathing artificially, and the ability to provide adequate intravenous nutrition to extremely immature infants have all been implicated in the trend toward increased birth weight specific survival over time.

A second type of study makes use of cross-sectional comparisons among several NICUs to demonstrate that different applications of similar technologies may also affect outcomes.^47,49 After controlling for the potentially confounding influences of race, birth weight, and gender, these investigations were still able to demonstrate significant intercenter differences in various outcome measures. The results of these studies imply that appropriate application of assisted mechanical ventilation in NICUs can significantly reduce morbidity and mortality among immature newborns.

Epidemiologists have looked at the impact of neonatal intensive care on the mortality rates of entire populations and have come to the conclusion that new neonatal technologies, properly distributed, provide significant improvements in this measure of health status. For example, an early Canadian study covering the period 1967 to 1974 documented a 43% decrease in the neonatal mortality rate for the province of Quebec coincident with several important institutional developments including (1) an increase in the proportion of sick and premature newborns cared for in referral centers; (2) the establishment of perinatal intensive care units in two university hospitals; and (3) the regionalization of the high-risk obstetric service associated with McGill University, where fully 14% of the province's births occurred.⁵⁰ The authors of this study were able to show that the improvement resulted from decreases in mortality rates within birth weight categories, not from any dramatic shift in the birth weight distribution itself, and that the diffusion of specialized care both to high-risk mothers and their newborn infants played a significant role in the observed decline in Quebec's neonatal mortality rates. Subsequently, Lee and his colleagues arrived at similar conclusions using data from the United States for the period between 1950 and 1975.⁵¹

When considering the effectiveness of neonatal intensive care, economists have used their particular skills to accomplish two functions: (1) to estimate the magnitude of improvement in outcome associated with neonatal intensive care and (2) to construct cost-benefit or cost-effectiveness analyses of these services.

As an example of the first type of contribution, Harris reported that an increase in the annual volume of deliveries in the hospital of birth lowered the probability of neonatal death (including late fetal death) among black babies in Massachusetts in 1975–1976.⁵² A 10% increase in the annual volume of deliveries was found to lower the death rate by approximately 5%.⁵³ A similar approach using data for white infants born in Michigan in 1984 found that birth in a hospital with a neonatal intensive care unit lowered the probability of a late fetal or neonatal death by approximately 30%.

This type of analysis has been generalized to examine the impact of the use of neonatal intensive care services on neonatal mortality using data that cover almost the entire U.S. population. These studies show that a 10% increase in neonatal intensive care use was found to lower the mortality rate of white low birth weight infants by 2%. For black infants, a 10% increase in use lowered the death rate by 3%. These findings indicate how much improvement in mortality can be expected from increases in the availability of neonatal intensive care to low birth weight babies.⁵⁴

Faced with limited resources and the need to make decisions about the implementation of costly but effective programs to improve infant health, policymakers have increasingly relied on the tools of cost-benefit and cost-effectiveness analysis to help determine whether the observed improvements in health are worth the costs expended to achieve them. Yet these techniques are not easy to apply.⁵⁵ First, because traditional accounting practices within hospitals use revenue from one source to offset shortfalls elsewhere (a practice known as cost shifting), dollar costs of neonatal intensive care are often difficult to determine accurately. Furthermore, costs of the long-term care required by some severely disabled NICU survivors are also difficult to estimate. On the benefits side, the probabilities of various outcomes are not easily determined and vary substantially among institutions. Moreover, to perform a full cost-benefit analysis, it is necessary to assign a monetary value to survival with and without different degrees of disability. This often proves an intractable problem. As a result of these difficulties, the conclusions drawn from cost-benefit studies must be regarded with caution.

Despite these and other obstacles, however, attempts have been made to compare the costs and benefits of neonatal intensive care in monetary terms. Budetti and his colleagues concluded that, while NICU care for babies weighing less than 1,500 grams was marginally cost-effective given the data available to them at the time of their analysis, such care for infants weighing less than 1,000 grams was not.⁵⁵ Because of the potentially large expenditures required to care for severely disabled NICU survivors, as the numbers of such survivors increased in lower birth weight groups, Budetti and his colleagues found that the costs of providing this type of care began to substantially outweigh the benefits. Other investigators have reached similar conclusions.⁵⁶

In a very careful analysis conducted in Hamilton-Wentworth County in Ontario, Boyle and others used cost-benefit, cost-effectiveness, and cost-utility analysis to evaluate NICU care.⁵⁷ In the last of these approaches, an attempt is made to adjust life-years gained by a factor which takes into account the quality of life for individuals who survive with disabilities. As in previous studies, the findings were birth weight dependent. For infants in the 1,000- to 1,499-gram range, the cost of neonatal intensive care was found to be $59,500 (1978 Canadian dollars) per survivor, $2,900 per life-year gained, or $3,200 per quality-adjusted life-year gained. The corresponding figures for infants in the 500- to 999-gram category were $102,500, $9,300, and $22,400. The authors concluded that "by every economic measure neonatal intensive care for infants weighing 1,000 grams to 1,499 grams is superior to neonatal intensive care for infants weighing 500 grams to 999 grams." Neonatal intensive care for the lower birth weight group appeared to generate a net economic loss in all scenarios analyzed. Yet, when the analysis focused on clinical outcomes, it yielded a different conclusion. Here it was found that neonatal intensive care for infants weighing 750 grams to 999 grams at birth resulted in the largest survival gain for any subgroup. The apparent inconsistency introduced by evaluating neonatal intensive care from these alternative perspectives highlights the difficulties attendant to decision making regarding high-cost, high technology medical services.

As authors of these and other studies are quick to point out, cost-benefit analyses can help identify those programs that not only provide a health benefit to participants but also pay for themselves in an economic sense. Programs that are considered cost-beneficial produce not only a survivor for the cost outlay, but also the future economic productivity of that survivor. Because this situation is akin to a man's going into a store to buy a loaf of bread and coming out with not only the bread but also more money than when he went in, it is a very restrictive criterion by which to judge a medical intervention. More pertinent questions to pose, especially if costs outweigh the benefits in dollars generated by a program, are (1) whether society deems the net costs worth undertaking in view of the health benefits produced by the program under consideration and/or (2) whether alternative, less costly strategies exist to achieve similar health outcomes.

Knowing that the economic costs of caring for very low birth weight infants (those weighing less than 1,000 grams) outweigh the economic benefits forces policymakers to decide either that these expenditures are worthwhile despite the costs or that alternative strategies for saving infants' lives need to be entertained.

A recent study by Joyce, Corman, and Grossman speaks directly to this dilemma.⁵⁸ They found that expansion of prenatal care is a more cost-effective method of saving additional infant lives than expansion of neonatal intensive care. For whites, the cost of saving an additional life by expanding the number of women who receive prenatal care in the first trimester is $31 per additional life saved in 1984 dollars. The cost of saving an additional life by expanding the number of low birth weight babies who receive neonatal intensive care is $2,834 per additional life saved. The comparable figures for black women and infants are $39 in the case of prenatal care and $2,150 in the case of neonatal intensive care.⁵⁹ These cost-effectiveness ratios although suggestive are far from definitive because they do not consider improvements in neonatal intensive care technology. As with many policy alternatives, changes in technology, particularly in rapidly developing fields of inquiry, may be expected to shift the advantage of one or another approach to a given problem over time.

Faced with the desire to improve the health of newborns and rapidly rising costs for the care of sick infants, the policymakers' dilemma is that not all technological innovations with potential clinical application may be cost saving. Two examples of recent improvements in NICU respiratory care, artificial surfactant and extra-corporeal membrane oxygenation (ECMO), illustrate this point.

During the first week of life, many premature infants are faced with the problem of breathing effectively with immature lungs. When clinically significant, this problem is called respiratory distress syndrome (RDS). RDS results from the failure of the premature infants' lungs to produce surfactant, a substance that allows the small air sacs in the lungs of premature infants to remain inflated during all phases of respiration. Providing premature newborns with exogenous surfactant ameliorates RDS. Since the efficacy of exogenous surfactant was demonstrated in 1980, it has been synthesized commercially and is now instilled in the windpipes of newly born premature infants in an attempt either to prevent or to treat RDS and thereby decrease the accompanying morbidity and mortality.

The early experience with artificial surfactant has been favorable.^60-68 At a wholesale cost of approximately $900 per treatment for a neonate, it is relatively inexpensive and can be administered at virtually any Level II or Level III nursery without special equipment or monitoring. A recent analysis from England suggested that the use of surfactant was cost-effective in very low birth weight infants,⁶⁹ and there has been speculation that the dissemination of the use of this material was responsible for the decline in infant mortality witnessed in 1991.⁷⁰

The second technology mentioned, ECMO, is a technique for bypassing the newborns' lungs altogether by placing the child on a machine similar to one used during open heart surgery. Not indicated for routine use in premature babies with RDS, ECMO is reserved for desperately ill, full-term babies whose condition is unresponsive to the conventional treatment of respiratory assistance. Very few newborns are candidates for ECMO so the potential effect of ECMO on neonatal mortality is limited. Nonetheless there appear to be more centers in the United States with the capability to perform ECMO than are warranted by the number of infants who would benefit from the therapy.⁷¹ Accordingly, there is concern that the therapy may be used inappropriately and that costs associated with this technique may be difficult to contain. As Southgate, who studied the use of ECMO in Georgia concluded, "Short-sighted proliferation of ECMO units will almost certainly result in an increased consumption of shrinking health care dollars, as well as the unnecessary duplication of services. . . ."⁷²

Careful analysis of both the cost and the clinical efficacy of new technologies would be helpful in guiding the allocation of resources so as to optimize both infant survival and the ultimate health of survivors.