|
|
| |||||||||||||||||||||||||||||||
Background The administration of surfactant decreased mortality, morbidity, and costs of care for very-low-birth-weight infants in clinical trials. The extent to which these benefits can be achieved in the usual clinical settings is not known.
Methods We analyzed clinical and financial data obtained from 1985 to 1990 at 14 perinatal centers in the United States on 5629 neonates weighing 500 to 1500 g. The infants were divided into groups according to whether they were born before or after surfactant was introduced into clinical practice. Regression models controlling for race, sex, and birth weight were used to assess mortality, morbidity, and use of resources. Mortality rates specific for these variables were projected to the nation as a whole with reference to the 1985 U.S. birth cohort.
Results The odds of death in the hospital for very-low-birth-weight infants were reduced by 30 percent after surfactant was introduced. Among infants with bronchopulmonary dysplasia, mortality declined 40 percent. Projections of mortality nationwide declined 5 percent. Eighty percent of the decline in the U.S. infant mortality rate between 1989 and 1990 could be attributed solely to the use of surfactant. Among the survivors, the overall odds of morbidity did not change, whether or not we adjusted for changes in race, sex, and birth weight. The odds of respiratory distress syndrome and pulmonary interstitial emphysema among the survivors declined by 20 percent and 40 percent, respectively, with surfactant. Inflation-adjusted charges per survivor declined by 10 percent, or $5,800, whereas the cost of care for each infant who died declined by 31 percent, or $4,400.
Conclusions The introduction of surfactant has led to decreased mortality and morbidity in very-low-birth-weight infants and to decreased use of resources both for infants who survive and for those who die.
By the last quarter of 1989, two formulations of surfactant (Exosurf and Survanta) had been designated investigational new drugs for use in treatment by the Food and Drug Administration. In 1991, the National Center for Health Statistics reported that infant mortality had declined from 9.7 per 1000 in 1989 to 9.1 per 1000 in 1990, a decrease of 6 percent5; the center estimated that as much as half this decline may have been due to the use of surfactant therapy6. Beyond considerations of morbidity and mortality, the impact of surfactant use on the cost of care is not well understood. The samples used in some studies, though they indicated a savings with surfactant use, were small7,8,9. We undertook an epidemiologic study during the years before and after surfactant entered the market to determine whether surfactant, once widely introduced, had an effect on mortality, morbidity, and resource use.
Methods
Data on Neonates
We used data from 1985 through 1991 from 14 hospitals that provide data to the National Perinatal Information Center, a nonprofit research and information center. Included in the data base are hospital identifiers; admission and discharge dates; the age, sex, race, and birth weight of the infant; the payer; diagnosis codes; procedure codes; room and board charges; charges for the intensive care nursery; charges for intermediate care; and ancillary charges, including laboratory, radiology, pharmacy, and surgery charges. Pharmacy charges are aggregated, so it is not possible to determine which newborns received surfactant.
The National Perinatal Information Center performs a cost analysis of the Medicare Cost Report for each member hospital. This analysis allows charge data to be converted to cost data at the level of the individual patient. Thus, each record includes calculations of cost in the same categories as those for charges. The resulting computations of cost allow comparisons between hospitals.
A total of 7318 very-low-birth-weight singleton neonates weighing 500 to 1500 g and born in the 14 hospitals were identified from the records. Duplicate records were removed from the file, as were those of neonates sent home after a stay of less than four days, which indicates an error in the birth-weight data, leaving 7081 neonates remaining in the study.
Characteristics of the Hospital
In 1992, the 14 hospitals had an average of 6 neonatologists, 3 maternal-fetal specialists, 34 beds in the neonatal intensive care unit, and 6 intermediate care beds. From 1986 through 1991, these hospitals increased the number of beds in the neonatal intensive care unit by 25 percent and the number of days spent in that unit by 10 percent, with no changes in the number of births.
Information on the services available, the timing of the introduction of surfactant, the use of other new pharmaceuticals and forms of technology, and the occurrence of epidemics of infectious disease was obtained by telephone from the director of each neonatal intensive care unit. We documented that the hospitals were perinatal referral centers that retained infants of the lowest birth weight, accepted neonatal and maternal transfers, and operated organized transport systems among referring hospitals. All the hospitals provided intravenous infusions, hyperalimentation, mechanical ventilation, and monitoring of central venous pressure. Each was a tertiary perinatal center, as defined by the Committee on Perinatal Health10. We found no epidemics in any hospital and no major new forms of technology, drugs other than surfactant, or methods of patient care introduced during the study period. The antenatal administration of glucocorticoids had been introduced in each hospital before 1985.
Study Design
The periods before and after the introduction of surfactant were compared. From 1985 to 1990, seven hospitals participated in clinical trials of surfactant administration, and 20 percent of the 7081 study infants were discharged from January 1986 through October 1989. To ensure that an infant's participation in clinical trials would not affect the results, periods before and after the use of surfactant were established for each hospital, with the periods of the clinical trials excluded. This left 3922 neonates in the before-surfactant period and 1707 in the after-surfactant period, for a total of 5629 infants.
The periods before and after the introduction of surfactant were similar among the 14 hospitals. The before-surfactant period ended in 1988 or 1989 in 11 hospitals, in 1987 in 2 hospitals, and in 1986 in 1 hospital. Table 1 shows the distribution of newborns according to sex, race, payer, and birth weight during the periods before and after surfactant use. The only change was a 16 percent increase after the introduction of surfactant in the number of newborns in the group without private insurance.
|
The analysis of mortality included in-hospital death rates for each year of the study, in-hospital death rates and crude odds ratios of death in four 250-g weight categories before and after the introduction of surfactant and for all infants in all birth-weight categories, multivariate logistic-regression analyses at the patient level, and projections of mortality at the national level.
We designed a logistic-regression model in which the risk of dying in the hospital was seen as a function of the infant's sex, race, birth weight, and period of birth (before or after the introduction of surfactant). The model quantified the odds of death, with control for independent variables. In addition to examining overall in-hospital mortality (death at any time after birth during the hospital stay), we used the same model to analyze in-hospital mortality in the first 28 days of life and among infants with bronchopulmonary dysplasia. Two modifications in the model were tested. In the first, insurance status was examined by comparing privately insured newborns with uninsured and Medicaid-dependent newborns. In the second, to explore whether mortality declined over time independently of the availability of surfactant, we added date of birth to the model in 24 quarter-year intervals.
We projected the change in infant mortality to the national level by applying the mortality rates specific to birth weight, race, and sex before and after the introduction of surfactant in the 14 study hospitals to the 1985 U.S. cohort of newborns weighing 500 to 1500 g at birth.
Analysis of Morbidity
The analysis of morbidity focused on the surviving infants. Although data obtained from discharge abstracts are an imperfect measure of morbidity, they provided a frame of reference for assessing whether the survivors were sicker after the introduction of surfactant than before. The key question was, Has morbidity among survivors risen, with and without control for a downward shift in birth weight? In the period before surfactant use, 32 percent of surviving very-low-birth-weight newborns weighed 500 to 999 g, as compared with 38 percent in the period after the introduction of surfactant -- a 13 percent increase in newborns in the lower half of the 500-to-1500-g weight category.
The analysis of morbidity examined the crude odds ratios of a diagnosis of respiratory distress syndrome, interstitial emphysema, bronchopulmonary dysplasia, intraventricular hemorrhage, patent ductus arteriosus, necrotizing enterocolitis, and septicemia (as defined in the International Classification of Diseases, 9th Revision, Clinical Modification11) before and after the introduction of surfactant. In addition, all the morbid conditions except septicemia were examined together to determine the overall effect of surfactant on morbidity. Finally, all the morbid conditions, including septicemia, were analyzed together. The crude odds ratio was calculated by computing the ratio of the number of newborns with the diagnosis (or diagnoses) in the post-surfactant period to the number without, and then dividing this ratio by the same ratio for the pre-surfactant period. The adjusted odds ratios for the same dependent dichotomous variables for morbidity were calculated with the logistic-regression model used in the mortality analysis.
Use of Resources
The analysis of resource use assessed the effect of the availability of surfactant on charges, cost, and length of stay. In earlier work, resource use among infants who died was found to be much lower than that among survivors12. In this study, the mean charges for survivors and decedents were $58,874 and $24,691, respectively. Therefore, survivors and decedents were analyzed separately. We examined four dependent measures of resource use for a given infant: total charges, total cost, total ancillary charges, and length of stay. Data on charges and costs were adjusted to 1985 levels with use of inflation factors for health care costs obtained from Data Resources, Inc. Total charges were defined as the sum of the charges posted on the patient's hospital bill. Total cost was the sum of the costs of the individual elements, as calculated from the charge data and the Medicare Cost Report. Ancillary charges were the sum of the charges for pharmacy, laboratory, radiology, surgery, and other services.
We used a regression model in which resource use was a function of sex, race, birth weight, date of birth expressed as a quarter-year interval, and time of birth in relation to the introduction of surfactant. Quarter-year of birth was included in the model both to correct for gradual changes in inflation that were not accounted for by adjustments to cost and charges and to control for possible increases in the intensity of service over time.
Results
Mortality
The trend in mortality among infants weighing 500 to 1500 g at birth in the 14 hospitals is shown in Figure 1. There was a clear decline in the hospital mortality rate in 1990, after the introduction of surfactant. Table 2 shows that in-hospital mortality declined from 24 percent to 20 percent and that the crude odds ratio of death was four fifths of the earlier rate (P<0.001). The odds of dying were 70 percent of what they had been during the pre-surfactant period for infants weighing 500 to 749 g (P = 0.005), 60 percent of what they had been for infants weighing 750 to 999 g (P<0.001), and 70 percent of what they had been for infants weighing 1000 to 1249 g (P = 0.02), with no change for infants weighing 1250 to 1500 g.
|
|
|
On the basis of the national data, we estimated that if surfactant had been available in 1985, 1936 additional very-low-birth-weight newborns could have survived. A reduction of this magnitude from the 39,145 infant deaths in 1985 would have lowered the infant-mortality rate from 10.4 to 9.9 per 1000, a 5 percent decline, or 80 percent of the 6 percent decline between 1989 and 1990.
Morbidity among Survivors
When the seven types of morbidity were analyzed together (Table 4), neither the crude nor the adjusted odds ratios changed significantly. That is, overall morbidity among survivors was not higher in the post-surfactant period than in the pre-surfactant period. In individual analyses controlling for birth weight, sex, and race, the incidence of respiratory distress syndrome and interstitial emphysema declined; the incidence of bronchopulmonary dysplasia, patent ductus arteriosus, and septicemia did not change; and the incidence of necrotizing enterocolitis and intraventricular hemorrhage increased.
|
Results of Analyses of Resource Use
Adjusted total charges for survivors (Table 5) declined approximately 10 percent (P = 0.03), or about $5,800 per infant, after the introduction of surfactant. Adjusted total costs declined 8 percent. Adjusted ancillary charges, which include charges for any surfactant administered, declined slightly but not significantly. There was no significant change in the length of stay.
|
Taken together, these analyses indicate that resource use declined after the introduction of surfactant, with insignificant changes in the length of stay; furthermore, the decline occurred among both the survivors and the decedents.
Discussion
Although we could not identify the patients who received surfactant, we demonstrate that the availability of surfactant had a substantial effect on very-low-birth-weight infants born in 14 study hospitals. Our survey of directors of neonatal intensive care units indicated that the introduction of surfactant was the only major change in care during the study period. The availability of surfactant not only improved morbidity and mortality but also decreased the use of resources. After we controlled for birth weight, sex, and race, the availability of surfactant resulted in a 30 percent decline in hospital mortality in very-low-birth-weight infants. Among newborns with a diagnosis of bronchopulmonary dysplasia, the decline (40 percent) was even more dramatic.
The decline in mortality in infants given surfactant for either prophylaxis or rescue (after the onset of symptoms) in clinical trials in Europe and America13,14,15,16 is confirmed by our results, which were based on an analysis of mortality at times when clinical trials were not being conducted at the study hospitals. Because our results reflect the experience in tertiary perinatal facilities, they cannot be considered to apply to the nation without caveats. Still, most very-low-birth-weight infants are born in tertiary facilities. Using standard adjustment techniques, we estimated that surfactant, if available, could have caused a 5 percent decline in the U.S. infant mortality rate in 1985, a decline that is about 80 percent of the 6 percent decline between 1989 and 1990. In all likelihood, therefore, much of the actual decline was due to the use of surfactant.
The status of the infants who survive is especially important. In our analysis, the odds of having a diagnosis of both respiratory distress syndrome and interstitial emphysema declined, whether or not adjustments were made in the survivor group. The odds of septicemia did not change. The odds of bronchopulmonary dysplasia and patent ductus arteriosus increased slightly when no adjustments were made, but with adjustments in birth weight and other factors they did not change. Most important, in both analyses there was no change in the odds of having any one morbid event. With reimbursement based on diagnosis-related groups, one would expect more recent hospital records to document morbidity more thoroughly, making the decrease in morbidity all the more remarkable. These results are generally similar to those in the reports of the clinical trials of surfactant13,14,15,16,17,18,19,20.
Our findings of an increase in the odds of intraventricular hemorrhage and necrotizing enterocolitis, points on which there is some ambiguity in the literature,2,13,15,19,20,21 suggest that surfactant may be changing the mix of complications among survivors. Some infants survive who would have died; perhaps these infants have different complications than those who survived without surfactant.
Two questions are consistently raised about the introduction of new treatments. First, did resource use for similar patients increase or decrease? Second, did social costs increase or decrease? These issues are especially difficult to assess when the treatment itself affects the case mix of the survivor cohort. The introduction of surfactant had such an effect: the survivors included a larger group of lower-weight newborns. Taking these and other salient factors, such as inflation, race, and sex, into account, we found that resource use declined. As measured by total charges, it declined by about 10 percent among survivors; for infants who died, the cost per infant was reduced by about 30 percent. The lower costs for those who died imply that the availability of surfactant has not led to major efforts to extend life in moribund newborns. Overall, our findings regarding resource use corroborate the results of smaller studies comparing treatment and nontreatment groups. Maniscalco et al.7 found that the average charges for survivors were about 13 percent less and ancillary charges an average of 25 percent less for all patients.
If inflation is taken into account, our analysis implies that the very-low-birth-weight infant who survives in the era of surfactant use will incur hospitalization charges about 10 percent lower than would have been incurred earlier. For the nation, this translates into a savings of about $200 million per year for all survivors. This savings must be offset by the increase in the number of very-low-birth-weight infants born (an increase of 1300 between 1985 and 1989), as well as by the higher costs for newborns who survive who otherwise would have died. Calculations that take both factors into account, based on the regressions presented, suggest a national savings of roughly $90 million.
Supported under a contract (N01-H0-29024) from the National Heart, Lung, and Blood Institute.
We are indebted to the directors of neonatal intensive care units who responded to our surveys; to the hospitals for their continuous participation in the National Perinatal Information Center; and to Ms. Janet Muri, Dr. John Pezzullo, Dr. Carl Roth, and Mr. David Gagnon.
Source Information
From the National Perinatal Information Center, Providence, R.I. (R.M.S., A.M.L., R.J.K.), and the Division of Neonatology, Columbia Hospital for Women, Washington, D.C. (J.W.S.).
Address reprint requests to Ms. Schwartz at the National Perinatal Information Center, One State St., Suite 102, Providence, RI 02908.
References
The following hospitals participated in the study: Bronson Methodist Hospital, Kalamazoo, Mich.; Children's Hospital of Buffalo, Buffalo, N.Y.; Kapiolani Medical Center for Women and Children, Honolulu, Hawaii; F.G. McGaw Hospital, Loyola University, Maywood, Ill.; Magee-Women's Hospital, Pittsburgh; Metro-Health Medical Center, Cleveland; Montefiore Medical Center, Bronx, N.Y.; Rockford Memorial Hospital, Rockford, Ill.; Self Memorial Hospital, Greenwood, S.C.; St. Margaret's Hospital for Women, Dorchester, Mass.; State University Hospital, Brooklyn, N.Y.; University of Chicago Hospital, Chicago; Women and Infants' Hospital of Rhode Island, Providence; and Yale-New Haven Hospital, New Haven, Conn.
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
