8
The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems 1n four states Steven Gortmaker, Ph.D., Arthur Sobol, M.A., Colleen Clark, M.P.H., Deborah K. Walker, Ed.D., and Arline Geronimus, A.B. Boston, Massachusetts This study estimates differentials in survival among very low-birth weight infants according to hospital of birth, and seeks to determine importance of birth at high-technology centers versus birth at other urban or rural hospitals. Data from four states for 1978 and 1979 were used to estimate survival curves for the first 24 hours of life by type of hospital at birth, birth weight, and race. Significant (p <0.0001) differences in survival by type of hospital for both races at birth weights of 1 000 to 1500 gm were observed. Smaller disparities were seen at birth weights of 750 to 1000 gm. Differentials in survival by hospital setting emerged in the first few hours after birth, underscoring the effectiveness of neonatal intensive care units in reducing infant mortality and the importance of maternal transport. Differentials persisted throughout the neonatal and postneonatal periods, although differences were attenuated. Prenatal assessment and provider and institutional cooperation can contribute to lowered mortality for high-risk infants and mothers. (AM J 0BSTET GYNECOL 1985;152:517-24.) Key words: Very low birth weight, neonatal mortality, neonatal intensive care, survival curves Population-based data in the United States demon- strate that birth weight-specific neonatal mortality has been decreasing in recent decades, with the most dra- matic successes being reported in the survival of very low-birth weight infants (<1501 gm). Data from a va- riety of regional centers providing neonatal intensive care indicate that the neonatal survival of these infants, that is, survival to 28 days after birth, has increased from 38% in the 1960s to 50% in the early 1970s and 69% in the late 1970s.' However, the benefits of ad- vances in technology are not being provided equitably to all very low-birth weight infants. For example, in 1976 to 1978, in the New York metropolitan area, only 34% of very low-birth weight infants were born at the centers providing such care! In Iowa, in 1978, only 22% of very low-birth weight infants were born in that state's one regional center.' There are many possible explanations for this lack of successful development of regionalized perinatal systems. These include barriers From the Department of Behavioral Sciences, Harvard School of Public Health. Supported by Grant 05146 from the Robert Wood johnson Foun- dation, Princeton, New Jersey. The opinions expressed are those of the authors and should not be attributed to the institutions or individuals mentioned. Received for publication july 12, 1984; revised January 30, 1985; accepted February 18, 1985. Reprint requests: Steven L. Gortmaker, Ph.D., Associate Professor, Department of Behavioral Sciences, Harvard School of Public Health. 677 Huntinl(ton Ave.. Boston. MA 02115. at the individual, provider, and institutional levels. In addition, there is considerable discussion of the cost- benefits of the new technology, particularly for infants of 500 to 1000 gm.' This study addresses three questions relevant to the discussion of what constitutes optimal use of neonatal intensive care technology for very low-birth weight in- fants. First, does survival of these infants vary by the technologic sophistication of the hospital of birth, and if so, how quickly after birth do these differences emerge? Second, do differences persist throughout the first year oflife, or is mortality only postponed by access to advanced technology? Third, is neonatal intensive care technology equally beneficial for infants of differ- ent races and birth weights? We measured survival by single hours following birth to identify the times of greatest risk differential and to note the speed with which the differentials emerge. Data from four states from 2 years were used, affording us the ability to ana- lyze various subgroups of infants and to observe state- wide systems of care. Methods Linked birth and death certificate data. Linked birth and death certificate data from Louisiana, Ohio, Ten- nessee, and Washington, from 1978 and 1979, were used in the study. A total population of 46,089 single and 7859 plural births, with infants <2501 gm in birth weight, were available for analysis; 9021 of these infants 517

The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

  • Upload
    arline

  • View
    213

  • Download
    1

Embed Size (px)

Citation preview

Page 1: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems 1n four states

Steven Gortmaker, Ph.D., Arthur Sobol, M.A., Colleen Clark, M.P.H., Deborah K. Walker, Ed.D., and Arline Geronimus, A.B.

Boston, Massachusetts

This study estimates differentials in survival among very low-birth weight infants according to hospital of birth, and seeks to determine importance of birth at high-technology centers versus birth at other urban or rural hospitals. Data from four states for 1978 and 1979 were used to estimate survival curves for the first 24 hours of life by type of hospital at birth, birth weight, and race. Significant (p <0.0001) differences in

survival by type of hospital for both races at birth weights of 1 000 to 1500 gm were observed. Smaller

disparities were seen at birth weights of 750 to 1000 gm. Differentials in survival by hospital setting emerged in the first few hours after birth, underscoring the effectiveness of neonatal intensive care units in reducing infant mortality and the importance of maternal transport. Differentials persisted throughout the

neonatal and postneonatal periods, although differences were attenuated. Prenatal assessment and provider and institutional cooperation can contribute to lowered mortality for high-risk infants and mothers.

(AM J 0BSTET GYNECOL 1985;152:517-24.)

Key words: Very low birth weight, neonatal mortality, neonatal intensive care, survival curves

Population-based data in the United States demon­strate that birth weight-specific neonatal mortality has been decreasing in recent decades, with the most dra­matic successes being reported in the survival of very low-birth weight infants (<1501 gm). Data from a va­riety of regional centers providing neonatal intensive care indicate that the neonatal survival of these infants, that is, survival to 28 days after birth, has increased from 38% in the 1960s to 50% in the early 1970s and 69% in the late 1970s.' However, the benefits of ad­vances in technology are not being provided equitably to all very low-birth weight infants. For example, in 1976 to 1978, in the New York metropolitan area, only 34% of very low-birth weight infants were born at the centers providing such care! In Iowa, in 1978, only 22% of very low-birth weight infants were born in that state's one regional center.' There are many possible explanations for this lack of successful development of regionalized perinatal systems. These include barriers

From the Department of Behavioral Sciences, Harvard School of Public Health.

Supported by Grant 05146 from the Robert Wood johnson Foun­dation, Princeton, New Jersey.

The opinions expressed are those of the authors and should not be attributed to the institutions or individuals mentioned.

Received for publication july 12, 1984; revised January 30, 1985; accepted February 18, 1985.

Reprint requests: Steven L. Gortmaker, Ph.D., Associate Professor, Department of Behavioral Sciences, Harvard School of Public Health. 677 Huntinl(ton Ave .. Boston. MA 02115.

at the individual, provider, and institutional levels. In addition, there is considerable discussion of the cost­benefits of the new technology, particularly for infants of 500 to 1000 gm.'

This study addresses three questions relevant to the discussion of what constitutes optimal use of neonatal intensive care technology for very low-birth weight in­fants. First, does survival of these infants vary by the technologic sophistication of the hospital of birth, and if so, how quickly after birth do these differences emerge? Second, do differences persist throughout the first year oflife, or is mortality only postponed by access to advanced technology? Third, is neonatal intensive care technology equally beneficial for infants of differ­ent races and birth weights? We measured survival by single hours following birth to identify the times of greatest risk differential and to note the speed with which the differentials emerge. Data from four states from 2 years were used, affording us the ability to ana­lyze various subgroups of infants and to observe state­wide systems of care.

Methods

Linked birth and death certificate data. Linked birth and death certificate data from Louisiana, Ohio, Ten­nessee, and Washington, from 1978 and 1979, were used in the study. A total population of 46,089 single and 7859 plural births, with infants <2501 gm in birth weight, were available for analysis; 9021 of these infants

517

Page 2: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

518 Gortmaker et al. July I, 1985 Am J Obstet Gynecol

Table I. Hospital of birth of infants < 1500 gm, by race, urban/rural residence, and state, 1978 and 1979

Race/residence and hospital level at birth Louisiana

White, urban (n) 536 Level III (%) 27 Other urban(%) 71 Rural(%) 1

Black, urban (n) 894 Level III (%) 59 Other urban(%) 41 Rural(%)

White, rural (n) 342 Level III(%) 14 Other urban (%) 34 Rural(%) 52

Black, rural (n) 460 Level III (%) 31 Other urban(%) 28 Rural(%) 41

weighed <1501 gm. Comparable data available for analysis across the four states included date of infant's birth and death, time of death, hospital of birth, birth weight, race, date of mother's last menstrual period, and county of residence. Complete data on all items except last menstrual period were available for 99% of the births in this sample; data on last menstrual period were available for 88% of the records. Some codes were not available across states in these linked files; for ex­ample, birth weights of < 1000 gm were coded as one category in Ohio and left as a continuous variable in other states. None of these omissions seriously affected the analysis.

Level of hospital of birth. There are no uniform criteria for neonatal intensive care units. The Com­mittee on Perinatal Health, formed by representatives of the American College of Obstetricians and Gyne­cologists, the American Medical Association, the Amer­ican Academy of Pediatrics, and the American Acad­emy of Family Physicians in 1976, produced guidelines for the regionalization of maternal and perinatal health services that include criteria for level I, II, and III hospitals. They write, "These guidelines are for plan­ning only and are not to be taken as standards of care. They have not been written as absolutes or regulations to be uniformly applied throughout the country."5 The "Standards and Recommendations for Hospital Care of Newborn Infants" of the American Academy of Pe­diatrics includes a similar caveat. Some states and other jurisdictions have designated criteria to use in evalu­ating maternity services; the Bureau of Maternity Ser­vices and Family Planning of the New York City Health Department, for example, evaluated each hospital's perinatal service and assigned a level between 1976 and 1978! However, none of the states in this study has established legally enforceable standards.

State

Ohio Tennessee Washington

2059 643 201 26 40 42 72 56 47 2 4 11

1081 576 10 35 67 40 65 33 60

1 543 510 644

13 38 44 10 16 43 77 47 13

116 186 56 11 45 61 12 10 39 77 45

For this study we have neither a set of standards for the hospitals in each state nor the capability of scoring the hospitals by level independently. Instead we have had to rely on the consensus in each state on what the level III neonatal intensive care units are. All hospitals not designated level III are grouped together as level 1111 urban, within a Standard Metropolitan Statistical Area (SMSA), or level 1111 rural, outside an SMSA. To identify the level III neonatal intensive care units, we have relied on information provided to us by the state health departments, heads of state university perinatal units, and others working in maternal and infant care. In no case did we encounter disagreement among our informants. Of the four states, the designations were least clear in Washington, where a number of hospitals were identified as level II+ . Such hospitals might be ones lacking capability only in some kinds of specialized pediatric surgery or might be so far from the University Hospital in Seattle that they act as local level III referral centers, although from the point of view of the uni­versity they are not quite level Ill. Another gray area is where a hospital at a distance from Seattle might have had a neonatalogist at some times and not at others. In sum, our definition of a level III neonatal intensive care unit is very conservative. Given that it is very difficult to identify even these highest level centers, we made no attempt whatsoever to distinguish between level I and level II perinatal services.

Data analysis. Infant mortality rates and survival curves are estimated for race and birth weight-specific groups, as well as for types of hospital of birth as de­scribed above. Race and birth weight are thus con­trolled in these analyses because these variables are ma­jor predictors of infant mortality, and their distribution varies substantially among hospitals. In our analyses, we categorize births by 500 gm weight categories. An

Page 3: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

Volume 152 Number 5

0

o Level ID Centers

• Other -Urban Hospitals

o Other -Rural Hospitals

Fig. 1. Cumulative survival of white infants, 1000 to 1500 gm, first 96 hours after birth in four states, 1978 to 1979.

empirical analysis of the relationship between neonatal mortality and birth weight demonstrated that finer weight groupings add little explanatory power, except for the 500 to 1000 gm range, which is broken into two 250 gm intervals. Two other analytic techniques are used to control for additional variables that might otherwise bias estimates of the relationships between hospital type at birth and subsequent death: multivar­iate logistic regression and the proportional hazards model.6 Further variables controlled in these analyses include length of gestation and whether the birth was plural or singular. A previous analysis of institutional differences in infant mortality among low-birth weight infants found that once birth weight, gestation, race, and sex were controlled, other potentially confounding variables exerted little statistically significant influence on neonatal mortality. 7 Survival curves are estimated with the Statistical Package for the Social Sciences sur­vival program. 8

Results Selection of hospital of birth. Data from the four

states under study indicated wide variation in the extent to which infants of very low birth weight are born in specialized centers (Table I). Only a fraction of the very low-birth weight births in these states took place in

0

Survival of very low-birth weight infants 519

o Level ID Centers

• Olher -Urban Hospitals

o Other -Rural Hospitals

Fig. 2. Cumulative survival of black infants, 1000 to 1500 gm, first 96 hours after birth in four states, 1978 to 1979.

specialized centers. The smallest proportion of very low-birth weight infants born in level III centers oc­curred among whites in Louisiana and Ohio; overall, 23% of white very low-birth weight infants were born in regional centers in these states. Black infants, in gen­eral, had better access to these specialized services. In Washington and Tennessee more than 50% of the black very low-birth weight infants were born in these cen­ters. The lowest percentages of births occurring in re­gionalized centers, as one would expect, were to resi­dents of rural areas, but here again, black infants were more likely than white ones to be born in specialized centers.

Estimating the impact that the place of birth had on the survival chances of such high-risk infants was con­strained by the obvious fact that pregnant mothers and hence births were not randomly assigned to the various types of institutions. Some high-risk fetuses were iden­tified during the prenatal period, and maternal trans­fers took place with the fetus at risk in utero. Maternal transfers were also possible during premature labor. Thus some hospitals should have been delivering more very low-birth weight and more high·risk infants than others. Other data not shown in Table I indicate that this sort of selection was operating in each of these states. The proportion of 1000 to 1500 gm births taking

Page 4: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

520 Gortmaker et al.

1.0

0.9

::.... 0.8 h. ....... :::::! Q::)

~ ~ Q_:

-...J § :;;: ~ 0 Level ill Centers V) . Other -Urban Hospitals

~ 0 Other -Rural Hospitals j::::: "'{ 0.3 ;5 ~ '-' 0.2

0.1

0

Fig. 3. Cumulative survival of white infants, 750 to 1000 gm, first 96 hours after birth in four states, 1978 to 1979.

place in level III centers was larger than the proportion of 2500 + gm births taking place in these centers. In addition, transfers could take place soon after delivery. High-risk infants born in a level I hospital who were strong enough to survive the first hazardous hours after birth could soon be enjoying all the facilities of a level III hospital. Our analysis therefore leads to minimal estimates of the impact of neonatal intensive care on the survival of all very low-birth weight infants.

Survival by hour after birth. We took two approaches to estimating the effect of place of birth on subsequent survival in the present analysis. We first examined the time period immediately following birth (the first 96 hours) in order to analyze closely the influences on mortality that were related to the hospital of birth. Dif­ferences in survival that emerged during this period for specific groups of infants at risk could be reasonably attributed, we believe, to the level of technology and skills available at the hospital of birth. We then looked at the survival experience of infants throughout their first year after birth in order to see the longer-term impact of differences in the perinatal care received and to test the hypothesis that specialized care may simply postpone the death of high-risk infants to the post­neonatal period. 9

Survival curves by hospital of birth were estimated for 10 different birth weight/race groupings. No sta-

::.... h. ....... :::::! Q::)

95 ~ Q_:

-...J § :;;: gs V)

~ j::::: "'< ;5 ~ '-'

July I, 1985 Am J Obstet Gynecol

1.0-----------------------,

0.1

0

o Level ill Centers

• Other -Urban Hospitals o Other - Rural Hospitals

Fig. 4. Cumulative survival of black infants, 750 to 1000 gm, first 96 hours after birth in four states, 1978 to 1979.

tistically significant differences in survival were noted for either black or white infants of 2000 to 2500 or I500 to 2000 gm. In contrast, significant differentials in early survival did appear among infants of <I500 gm at birth. Figs. I and 2 plot the 96-hour cumulative survival of 1000 to I500 gm white and black infants by their hospital of birth. In these graphs, the differentials in survival between births occurring in the different hospital types emerged mainly during the first few hours after birth. The curves were relatively smooth, with the exception of a sharp drop at I2 hours. Ap­parently, a number of deaths occurring around this point have been rounded off to I2 hours. The curves did not cross over time, however, and we did not con­sider this a problem.

The survival of white and black infants of 1000 to I500 gm born in level III centers was significantly greater than that of infants born in rural hospitals (p < 0.000 I). White infants born in level III centers also experienced significantly greater 4-day survival than white infants born in other urban hospitals (p = 0.0001), although the same was not true for black in­fants (p = O.I2). In all these comparisons, however, survival among infants born in the level III centers was highest.

Infants < 1000 gm born in level III centers also ex­perienced significantly better survival than infants born

Page 5: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

Volume 152 Number 5

Survival of very low-birth weight infants 521

Table II. Mortality of very low-birth weight infants by level of hospital of birth and race in four states, 1978 and 1979

No. of deaths/ 1000 live births

Early neonatal death Neonatal death Postneonatal death Infant death (0-4 days) (0-28 days) (28 days-] yr) (birth-1 yr)

750-1000 gm* White infants

Level III (226) 358 486 58 544 Other urban (296) 490t 588t 44 632t Rural (135) 519t 608t 7t 615t

Black infants Level III (289) 318 481 42 522 Other urban (146) 315 486 48 534 Rural (57) 386 404 88 491

1000-1500 gm White infants

Level III (1 020) 113 162 30 192 Other urban (1557) 171§ 232§ 31 263§ Rural (543) 190§ 250§ 26 276§

Black infants Level III (84 7) 73 111 32 143

· Other urban (718) 95 142 28 170 Rural (189) 180§ 233§ 21 254§

Numbers in parentheses are counts of live births. *Data from Ohio not available; analysis of births < 1000 gm from all four states, however, revealed very similar results. tDifference between level III and this category significant at p < 0.01. tDifference between level III and this category significant at p < 0.05. §Difference between level III and this category significant at p < 0.001.

in either other urban or rural hospitals. A further anal­

ysis by 250 gm intervals revealed no significant differ­entials in survival among the 500 to 750 gm births but substantial differentials by hospital group among white and black births of 750 to 1000 gm. These survival curves are plotted as Figs. 3 and 4. (Data from Ohio have been omitted from these figures because 250 gm breakdowns of the Ohio data were not possible.) Among the white births, the differences between the survival of infants born in level III versus either of the other hospital groups were significant at p < 0.00 1. Among the black births, the differences were not sta­tistically significant (p = 0.84 and p = 0.10). The black 750 to 1000 gm infants born in rural hospitals expe­rienced initially much greater mortality, but this dif­ference was almost gone by the end of 4 days. The samples of black births in rural hospitals was relatively small, however, and thus statistical insignificance could

be expected. These curves illustrate the high risk of mortality ex­

perienced during the first few hours after birth and thus indicate the constraints under which the delivery of optimal perinatal care operates. Any deterioration which takes place early in the life of a high-risk newborn infant is probably an additional risk factor later on, no

matter what level and expertise of care are subsequently available. Within the first few hours, in Figs. 1 to 4, curves for both black and white infants born in the level III centers diverged sharply from the curves of infants

not born in these centers. The largest differentials ap­peared in the contrasts between infants born in level III centers and those born in rural hospitals. A look at the hazard rates estimated from these survival data con­firms the substantial differences in risk experienced by infants during the first few hours after birth. White infants of 1000 to 1500 gm experienced mortality of 0.9% per hour during the first 3 hours if the birth occurred in a level III center; white infants of similar birth weight born in rural hospitals experienced mor­tality of 1.8% per hour during this time period. Among black infants weighing 1000 to 1500 gm, the hazard rate was 0.5% per hour if the birth occurred in a level III center and 2.6% per hour if in a rural hospital.

Survival of black infants versus white infants. As we have already noted, there were substantial differ­ences in birth weight-specific survival between white and black infants. Many of the differentials in early survival by race were of a magnitude similar to the differentials noted by hospital of birth. In general, for all birth weight categories, survival during the first 96 hours was greater among black than among white in­fants at the same level of hospital care. Racial differ­entials in birth weight-specific survival have long been noted' 0

• 11 and thus were not surprising. We could have

combined these different curves in our analysis and statistically controlled for race, but we believed that the separate curves were useful descriptive data in their own right. Furthermore, some expected differentials

Page 6: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

522 Gortmaker et al. July I, 1985 Am J Obstet Gynecol

Table III. Estimated relative risks of death by birth weight, level of hospital of birth, and race in four states, 1978 and 1979

Estimated relative risk

Early neonatal period Neonatal period Postneonatal period Infancy (0-4 days) (0-28 days) (28 days-] yr) (birth-1 yr)

750-1000 gm* White infants

Level III 1.00 1.00 1.00 1.00 Other urban 1.36t 1.2U 0.76 1.16t Rural 1.45t 1.25t O.l2t 1.13t

Black infants Level III 1.00 1.00 1.00 1.00 Other urban 0.99 1.01 1.14 1.02 Rural 1.21 0.84 2.10 0.94

l 000-1500 gm White infants

Level III 1.00 1.00 1.00 1.00 Other urban 1.51 § 1.43§ 1.03 1.37§ Rural 1.68§ 1.54§ 0.87 1.44§

Black infants Level III 1.00 1.00 1.00 1.00 Other urban 1.30 1.28 0.88 1.19 Rural 2.47§ 2.10§ 0.66 1.78§

*Data from Ohio not available; analysis of births <1000 gm from all four states, however, revealed very similar results. tDifference between level III and this category significant at p < 0.01. tDifference between level III and this category significant at p < 0.05. §Difference between level III and this category significant at p < 0.001.

by race did not emerge as statistically significant. For example, among 750 to 1000 gm infants born in level III centers, no statistically significant differences in sur­vival appeared (although survival of black infants was slightly higher). Among 1000 to 1500 gm births, black infants born in rural hospitals did not fare significantly better than white infants born in rural hospitals. Be­cause of the differences in birth weight-specific sur­vival by race, it was important in comparing outcomes among hospitals or geographic areas to control for race as well as for birth weight distribution.

Controlling for other variables. We have already mentioned other variables that have been related to infant survival independently of birth weight and race, such as gestation and plural or singular birth. When we analyzed the data after controlling for these vari­ables we found that the differentials described above by hospital of birth remained essentially unchanged after these controls were introduced as co-variates into proportional hazards models. No differentials in sur­vival for singular versus plural births were noted once birth weight was controlled. This was different from what Williams et al. 12 found in California; further in­vestigation of this question may be in order. Some dif­ferentials in survival did emerge among states, even after birth weight, race, and hospital type were con­trolled. These, however, were generally small.

Neonatal and postneonatal survival. The mortality of these 750 to 1000 gm infants and 1000 to 1500 gm

infants throughout their first year of life is summarized in Table II. The relative risks associated with the dif­ferent types of hospitals are summarized in Table Ill. The largest relative risks appeared for infants of 1000 to 1500 gm. Generally small and inconsistent differ­entials appeared for 750 to 1000 gm births. The data in both of these tables indicated that the initial early differentials in risk of neonatal mortality by hospital of birth persisted throughout the first year of life. While some greater risks of postneonatal mortality appeared among infants born in level III centers, these did little to attenuate risks by hospital of birth which emerged during the first few days. Infants born in level III cen­ters, in general, experienced the best survival, while infants born in the rural hospitals experienced the worst survival. In general, black infants had better sur­vival rates than white infants of the same birth weight and hospital of birth.

Comment

We have demonstrated significant differentials in the survival of infants of similar race and birth weight, depending on their hospital of birth. These differen­tials appear among infants weighing 750 to 1500 gm at birth, for both black and white infants (in only three states for 750 to 1000 gm births). The results are con­sistent with those of other studies of the survival of very low-birth weight infants, including studies that are in­stitutionally based and those that are population

Page 7: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

Volume 152 Number 5

based. 13 The neonatal survival rate of 500 to 1500 gm infants born in level III centers in the three states with these birth weight data was 68% for both black and white infants,* remarkably close to the 69% noted in a recent literature review of the experience of region­alized centers. 1 The advantages in survival we dem­onstrated among infants born in level III centers versus those born in other hospitals located in metropolitan areas are also consistent with the results of two previous population-based studies!· 3 The relative risks associ­ated with very low-birth weight infants born in the rural hospitals in this study were often higher than those reported for the urban hospitals and were sub­stantially higher than those reported for births taking place in level III centers.

Our analysis of survival by hour following birth il­lustrates the rapidity with which these mortality dif­ferentials emerge. This corroborates the most recent work done by Paneth et a!. 14 The advantages of ma­ternal as opposed to infant transport for very low-birth weight infants become obvious. Even if excellent infant transport is available in rural areas, the hours involved place very low-birth weight infants at risk. Our data show that the neonatal mortality risk differentials by hospital of birth are sustained throughout the first year of life and are reflected in infant mortality rates. While some deaths are merely postponed into the postneo­natal period by birth at level III centers, more often death is prevented.

Finally, it needs to be emphasized that the numbers of handicapped very low-birth weight survivors is likely to be small. The infants whose cost-benefit status seems most questionable at this point are those weighing 800 to 1000 gm.• These constitute about 0.4% of all live births in the United States. If we assume that !-year survival is 37% and that the rate of handicaps among survivors is 40% (twice what Bennett et al. 15 found in a recent study), the number of handicapped individuals that could be maximally attributed to improved care is still only 0.6/1000 live births. As long as any new tech­nique used in the future is carefully evaluated, provid­ing intensive services at the time of birth more equitably and to larger numbers of infants at risk should improve both the survival prospects and the chances for a healthy life for many children.

Our results also illustrate the difficulty of evaluating changing neonatal intensive care unit technologies be­cause of the very different survival patterns observed for infants born in different level hospitals and for

*While the overall percentage of survival is similar for black and white infants, birth weight-specific survival for 500 to 750 and 1000 to 1500 gm is greater for black births than for white births. This differential is counterbalanced by the lighter birth weight distribution of the black infants.

Survival of very low-birth weight infants 523

infants of differing birth weight, gestation, and race. The differentials by race are large and indicate that black very low-birth weight infants have better survival than white infants of similar birth weight. Among in­fants of 1000 to 1500 gm born in level III centers in these four states, white infants appeared to experience neonatal mortality almost 50% greater than that of non­white infants. Adequate evaluation of neonatal inten­sive care units thus requires the estimation of separate rates by race.

Although there is no agreement on what would con­stitute the optimal system of perinatal care, it is clear from our study that very low-birth weight infants ben­efit greatly from being born in a level III perinatal center. Unless pregnant women are enrolled in a system of prenatal care that includes identifying high-risk pregnancies and premature labor, maternal transfers cannot easily take place. The key to improving the pres­ent system of care for very low-birth weight infants lies in improving prenatal care for all pregnant women. An added benefit of a comprehensive and universally ac­cessible system of prenatal care is that good care com­bined with techniques for early detection and preven­tion of premature labor holds the promise of reducing the incidence of low birth weight and thereby the de­mand for high-cost, high-technology neonatal intensive care. 16

We thank jeanne Huba, Rural Infant Care Program, Ohio; Nancy Julian, Rural Infant Care Program, Lou­isiana (Tulane); Patrick Sweeney, M.D., Rural Infant Care Program, Tennessee; and David Woodrum, M.D., University of Washington, for their help in identifying the level III neonatal intensive care units in their re­spective states. We also thank the four state health de­partments for making the data tapes available.

REFERENCES

1. Philip A, Little G, Polivy D, Lucey]. Neonatal mortality risk for the eighties: the importance of birth weight/ges­tational age groups. Pediatrics 1981;68:122-30.

2. Paneth N, Kiely JL, Wallenstein S, Marcus M, Pakter J, Susser M. Newborn intensive care and neonatal mortality in low-birth-weight infants. N EnglJ Med 1982;307:149-55.

3. Hein HA. Evaluation of a rural perinatal care system. Pediatrics 1980;66:540-6.

4. Boyle MH, Torrance GW, Sinclair JC, Horwood SP. Eco­nomic evaluation of neonatal intensive care of very-low­birth-weight infants. N EnglJ Med 1983;308:1330-7.

5. March of Dimes, National Foundation-Committee on Perinatal Health. Toward improving the outcome of preg­nancy. Recommendations for the regional development of maternal and perinatal health services. White Plains, New York: National Foundation-March of Dimes 1976.

6. Anderson S, Auquier A, Hauck WW, Oakes D, Vandaele W, Weisberg HI. Statistical methods for comparative stud­ies. New York: John Wiley & Sons, 1980.

7. Paneth N, Wallenstein S, Kiely J, Susser M. Social class indicators and mortality in low-birthweight infants. Am J Epidemioll982; 116:364-75.

Page 8: The survival of very low-birth weight infants by level of hospital of birth: A population study of perinatal systems in four states

Gortmaker et al.

8. Hull CH, Nie NH. SPSS update. New procedures and facilities for releases 7 and 8. 1979.

9. Hack M, Merkatz IR, Jones PK, Fanaroff AA. Changing trends of neonatal and postneonatal deaths in very­low-birth-weight infants. AM j 0BSTET GYNECOL 1980; 137:797-800.

10. Shapiro S, Schlesinger E, Nesbitt R. Infant, perinatal, ma­ternal, and childhood mortality in the United States. Cam­bridge, Massachusetts: Harvard University Press, 1968.

11. North AF, MacDonald HM. Why are neonatal mortality rates lower in small black infants than in white infants in similar birth weight? J Pediatr 1977;90:809-10.

12. Williams RL, Creasy RK, Cunningham GC, Hawes WE, Norris FD, Tashiro M. Fetal growth and perinatal viability in California. Obstet Gynecol 1982;59:624-32.

July I, 1985 Am J Obstet Gynecol

13. Cordero L, Backes CR, Zuspan FP. Very low-birth weight infant. I. Influence of place of birth on survival. AM J 0BSTET GYNECOL 1982;143:533-7.

14. Paneth N et al. Age at death used to assess the effect of interhospital transfer of newborns. Pediatrics 1984; 73:854-61.

15. Bennett FC, Robinson NM, Sells CJ. Growth and devel­opment of infants weighing less than 800 grams at birth. Pediatrics 1983;71:319-23.

16. Herron MA, Katz M, Creasy RK. Evaluation of a preterm birth prevention program: preliminary report. AM JOB­STET GYNECOL 1982;59:452.

The Dublin randomized controlled trial of intrapartum fetal heart rate monitoring

Dermot MacDonald, Adrian Grant, Margaret Sheridan-Pereira, Peter Boylan, and lain Chalmers

Dublin, Ireland, and Oxford, England

In a randomized controlled trial involving 12,964 women, a policy of continuous electronic intrapartum fetal heart monitoring was compared with an alternative policy of intermittent auscultation, both policies

including an option to measure fetal scalp blood pH. Women allocated to electronic fetal heart monitoring

had shorter labors and received less analgesia. The caesarean delivery rates were 2.4% for electronic fetal heart monitoring and 2.2% for intermittent auscultation but this small difference arose from the

identification of nearly twice as many fetuses with low scalp pH (<7.20) in the electronic fetal heart monitoring group. The forceps delivery rate was 8.2% in the electronic fetal heart monitoring group compared with 6.3% in the intermittent auscultation group, and this excess was explained by more

instrumental deliveries prompted by fetal heart rate abnormalities. There were 14 stillbirths and neonatal deaths in each group, with a similar distribution of causes. There were no apparent differences in the rates of low Apgar scores, need for resuscitation, or transfer to the special care nursery. Cases of neonatal

seizures and persistent abnormal neurological signs followed by survival were twice as frequent in the intermittent auscultation group, and this differential effect was related to duration of labor. Follow-up at 1 year of babies who survived neonatal seizures revealed three clearly abnormal infants in each group. The implications of these findings for both theory and practice are discussed. (AM J OBSTET GYNECOL

1985;152:524-39.)

Key words: Fetal monitoring, labor, random allocation, delivery, newborn infant

During the past decade there have been dramatic changes in the methods used to assess the condition of

From the National Maternity Hospital, Dublin, and the National Perinatal Epidemiology Unit, Radcliffe Infirmary, Oxford.

Generous financial support was received from the Medical Research Council of Ireland, the National Maternity Hospital Research Fund, the Wellcome Trust, and the Department of Health and Social Security, which supports the National Perinatal Epidemiol­ogy Unit.

Received for publication October 22, 1984; revised and accepted March 27, 1985.

Reprint requests: Dr. Dermot MacDonald, National Maternity Hos­pital, Holies St., Dublin 2, Ireland.

524

the fetus during labor. Intermittent auscultation has increasingly been displaced by continuous recording of the fetal heart rate. This widespread trend has been associated to a variable extent with the use of fetal scalp blood sampling to assess fetal acid-base status when the fetal heart rate pattern causes concern.

Controversy continues about the relative merits of these different methods of intrapartum fetal assess­ment. Although there is some evidence of a consensus that the use of the more intensive methods of intra­partum monitoring is appropriate when the fetus is deemed to be at high risk, there is no such agreement