Are there health risks in using risking systems? the case of perinatal risk assessment

297

Health Policy, 7 (1987) 297-307 Elsevier HPE 00147

Are there health risks in using risking systems? The case of perinatal risk assessment* Miriam Orleans and Albert D. Haverkamp Department of Preventive Medicine and Biometrics, Health Sciences Center, Denver, Colorado, U.S.A. Accepted

5 February

University

of Colorado

1987

Summary Health risk management is receiving widespread public and professional attention. Relationships between risks and health effects are often assumed to be real and taken for granted. While health risk assessments are increasingly conducted, they are sometimes not well grounded and frequently poorly studied. Knowledge of the sensitivity, specificity and predictive value of risk factors in clearly identified target populations is required if appropriate interventions are to be designed. ‘Risking systems’ are often comprised of risk factors which have little predictive usefulness. In the field of perinatal medicine, the identification of risk factors has a long history. The changing risk status of pregnant women during the prenatal period, in labor, and at the time of delivery poses problems for the researcher and clinician who rely on risking systems to characterize the likelihood of adverse events. Currently used risking systems, antepartum and intrapartum, are not sufficiently robust for this task. The assignment of a high risk status often results in overcalling a problem and a cascade of technological interventions may follow. In the current climate of ‘managing’ pregnancies, a thoughtful review of the quality of risking systems is required. Are we being over-risked and over-served? Perinatal

risk assessment;

Perinatal technology

*Revised version of a paper presented at the 1 lth Joint European Health Policy Forum - World Health Organization meeting on ‘The Assessment of Risks Related to Health.Technology’ held in Brussels on 10-l 1 September 1986. Address for correspondence: Professor and Biometrics, University of Colorado Avenue, Denver, CO 80262, U.S.A. 016%8510/87/$03.500

1987 Elsevier

Miriam Health

Science

Orleans, Sciences

Publishers

Ph.D., Center,

Department of Preventive Medicine Campus Box C245, 4200 East Ninth

B.V. (Biomedical

Division)

298

Introduction The methods of health risk assessment are being critically reviewed and evaluated in the United States, as in many western countries, in response to the rapid and widespread dissemination of health promotion and disease prevention strategies. A recent report of the National Center for Health Services Research (NCHSR) [l] in forms us that more than 600 organizations in the United States are now using the health risk appraisal methods developed by the Centers for Disease Control for health education purposes or to stimulate health behavior change. In addition, more than 40 vendors, public and private, are making some version of these methods available to the public. In stressing the importance of a sound scientific basis for recommended behavioral changes, the report concludes that many risk factors are poorly studied and understood. In many instances, the sensitivity, specificity and predictive value of risk screening systems have not been determined. Their ability to inform us about the association of particular risk factors with good or bad outcomes is often questionable. However, the public is told to run, then rtot to run, but to walk briskly; women are told to fight osteoporosis with large calcium supplements, then not to do this because studies show no improvements in bone density; to use only vegetable oils in cooking, then shortly thereafter, to replace them with fish oils. It is unfortunate that there is not yet a NCHSR companion report that discusses the ways in which risking systems are used in medical diagnosis and treatment, their quality, predictive value, and their efficacy in the light of their widespread dissemination. It would be an enormously difficult undertaking, since it would need to consider risking systems in almost every specialty field in medicine. The benefit might well be worth the effort. One discipline that has dealt with risk assessment and management with great intensity and for a long period of time is perinatal medicine. The use of technology to manage and thereby reduce risk during birth is probably as old as written history. Specific rituals, medicines, persons, and beliefs are associated with birth in almost all cultures. Different cultures have ‘stylized’ birth in various ways in efforts to reduce the accompanying uncertainty. Rituals have been used to speed labor, to reduce pain, and to ensure safe deliveries. For example, slow progress in labor was, and is often today, seen as ‘abnormal progress’. ‘Abnormal progress’ is seen as a risk to both infant and mother. Slow progess, for many practitioners, is treated as an intrapartum ‘risk factor,’ and the response of managers is to speed up the labor and/or to deliver the baby through operative methods. There are many examples of the view that birth is a time of vulnerability for both mother and infant. helen Marieskind in her book, Women in the Health System [2], provides an example. During the long labor of Beatrice d’Este on New Year’s Day in 1494, birth attendants suggested many remedies for her severe labor pains - midwives suggested that she swallow an egg-white mixed with pieces of scarlet silk; that she sit over a pot of boiling water, tie the duke’s cap to her abdomen, swallow a potion of hot spirits cooked with stubs of deer’s antlers and cochineal, etc. The court doc-

299

tor said he could do nothing, but a second medicus proposed three ounces of river snails, mixed with muskat nuts and red coral; a third said to add cow’s dung, while a fourth advocated phlebotomy. However, this was not possible because Mars was in the constellation of Cancer, a contraindication. Her husband said they were ail fools and went to pray in the family chapel. Beatrice died. Had she lived in the eighteenth century, when Cesarean section was first being used, she might have had one, but then too she might have died. Until the fifties in this century, mortality rates from C-sections were high. The transfer of birth from homes to hospitals in the late 19th century in the United States, as in many western countries, resulted in increased physician involvement in labor and delivery as well as an increase in knowledge about pregnancy and its risks. Enhanced understanding seemed, then as now, to precipitate greater effort to control the process, in order to reduce the uncertainty and thereby to improve the outcome. This wish to better understand pregnancy and its risks is relevant for our own understanding of the current widespread use of ultrasound, amniocentesis, chorionic villus sampling, and a-feto-protein testing to monitor the growth and assess the normality of the fetus. Intrapartum surveillance by means of electronic fetal monitoring, contraction stress testing, and fetal-movement monitoring are also widely disseminated. The first X-rays of pregnant women occurred in the 1890s. These opportunities to see the fetus were regarded at first as being of scientific value, then quickly as being of clinical value. By the 192Os, the diagnosis of pregnancy at 6-10 weeks was being recommended as a useful method for assessing gestational age of the fetus. By the mid-1930s in Britain and the United States, it was believed that no reputable obstetric service could responsibly function without routinely X-raying all patients (at that time, a one-hour exposure was required). By 1954 in one London hospital, 67% of all maternity patients were X-rayed. However, in 1956 a British epidemiologist, Dr. Alice Stewart, published a report associating antenatal X-ray with later childhood cancer and obstetricians began to be more cautious in their use of this diagnostic procedure [3] By then, another antenatal assessment technology was being developed. The first obstetric experiments with ultrasonography took place in the mid-50s. By 1963, it was being used by clinicians to measure the size of fetal heads and to monitor fetal growth. In London, by 1974, the hospital that had X-rayed 67% of pregnancies, now used ultrasound on 62% [4]. An earlier diagnostic strategy, auscultation, has been with us for more than 200 years. In the 17OOs, signs of life or death of the fetus were identified by feeling its pulse in the presenting head, foot or umbilical cord. It was described by DeKergaradec in 1821 as an important diagnositic procedure. A specific definition of fetal distress was provided by Von Winckle in 1893, as being a fetal heart rate (FHR) of >160 or <120. Through the first half of this century, there were a number of efforts to improve the precision with which fetal heart beats were recorded. The first vaginal electrode was used to obtain a fetal EKG tracing in 1906. In 1957, Hon reported the successful use of an abdominal Doppler and by 1960, he had developed a monitor. By 1972, he had developed a spiral electrode, which passes

300

through the cervix and is clipped to the head of the fetus. Whether external or internal, electronic fetal monitoring (EFM) is based on the continuous recording on polygraph of both the FHR and uterine contractions. By 1969, both electronic fetal monitoring by an external indirect ultrasonic Doppler and direct internal EKG monitoring began to spread rapidly. By 1972, there were an estimated 1000 EFM systems in use in the United States. Today monitors are available in almost every obstetrical unit in the United States [5]. The questions that one quite naturally raises in thinking about the constantly growing antepartum and intrapartum risk assessment technologies are: What have we gained? How have we benefited? Are we occupied with real and important risks? Have we altered the course of disease? Do we manage our problems differently because we identify the risks so much more clearly? Have we truly reduced uncertainty? Have there been - as with the use of fetal X-ray - any unforeseen consequences? In a recent article in the New England Journal of Medicine, the words ‘cascade effect’ were used to describe the use of technology in labor and delivery. In biology, ‘cascade’ refers to a process that once started, proceeds stepwise to a seemingly inevitable conclusion. Mold and Stein write [6]: ‘Women with uncomplicated pregnancies are admitted to the hospital and attached to uterine-contraction monitors and fetal heart-rate monitors. It is assumed that the risk of untoward effects associated with these devices is less important than the benefits of being able to detect an otherwise unrecognized problem. However, use of the monitors requires that the women be relatively inactive in bed, and it may in some cases increase their level of anxiety. There is also a marked incidence of false positive monitor readings. Labor may be slowed somewhat by the combination of inactivity and anxiety, and this retardation may in turn lead to interventions to speed up labor. One such intervention is artificial rupture of the membranes - another ‘benign procedure’ used to speed labor and detect meconium-stained amniotic fluid. However, when labor is accelerated, the pain of the contractions is also increased, and pain medication or anesthesia may be requested by the patient or suggested by the physician. Also, the lack of amniotic fluid cushioning the baby’s head in the pelvic canal has been shown to increase the pressure within the baby’s cranial vault, which may lead to some abnormal readings on the fetal heart-rate monitor. These, of course, may trigger further interventions. It is not hard to see that at least in some cases, the cascade of benign interventions to which physicians subject many women can lead to complications that lead to further interventions and more complications, and end in some final intervention - usually a Cesarean section - that would not have occurred had not the cascade been set in motion.’ The effort to learn more about risks in order to select appropriate management strategies and improve outcomes has led to an explosion of diagnostic procedures, whose consequences are not always well understood and are not always benign. In thinking about the opportunities for improving perinatal outcomes, it is useful to consider the number of infants who might be helped, or the amount of risk re-

301

duction or benefit that might result from an optimal use of diagnostic technology. For example, in the 1970s advocates suggested that the use of EFM could significantly reduce intrapartum deaths or still-births, mental retardation, and cerebral palsy. One dreaded intrapartum risk factor was hypoxia, an indicator of fetal distress, mortality and morbidity. Let us review the evidence.

The risk of intrapartum

death

In the United States Collaborative Perinatal Study (1979), the total incidence of intrapartum fetal deaths was four per 1000 live births. For term-sized fetuses weighing more than 2500 grams, the incidence was 1.5 per 1000 [7]. Prematurity was identified as a significant risk factor, since 62% of intrapartum deaths occurred among low birthweight infants, i.e., those weighing less than 2500 grams. However, prematurity occurs in only 8.2% of all births [8]. More than 40 risk factors have been identified as being associated with low birthweight alone [9]. While the causes of pre-term labor are not fully understood, some of the risk factors have been identified. The risk factors pertain mainly to maternal characteristics or problems of the developing fetus that are only infrequently reversible. The use of EFM during the labor rarely informs the practitioner in such a way that he or she can increase the weight of the infant by removing or managing the risk. The benefit of EFM to full-term, low-risk mothers has not been demonstrated. Thompson has calculated that to do a study of the difference in perinatal death rates between full-term, low-risk monitored and unmonitored patients, 180000 patients would be required in each group [lo]. The very size of the required groups casts doubt on the need for universal use of EFM, since obviously the problems are of very low incidence. Further, the possible adverse effects resulting from the use of EFM also warrant consideration.

The risk of cerebral palsy Cerebral palsy, a central nervous system disorder, has been attributed to perinatal hypoxia. The rate of cerebral palsy is 2.5 per 1000 in school-aged children, a fairly consistent finding in western industrialized countries. It has been estimated that from 20% to 40% (a recent estimate is 10%) of the 2.5 cases per 1000 children are the result of intrapartum hypoxia [ll]. Factors other than hypoxia are now known to contribute more importantly to cerebral palsy. A recent report by Nelson and Ellenberg [12] indicated that maternal mental retardation, birthweight below 2001 g, and breech presentation are the principal risk factors. Although 2% of the children with cerebral palsy had at least one of three markers suggestive of asphyxia, over 90% of those cases also had major congenital or other intrinsic defects that might have contributed to their cerebral palsy. In the course of their study of 51285 pregnancies, these researchers were able to obtain information about 45559 children who were 7 years of age.

302

One hundred and eightly-nine of these children were found to have cerebral palsy, and among them 118 (0.2%) were to varying degrees mentally retarded. This research suggests that asphyxia did not play the important role attributed to it in earlier writings.

The risk of mental retardation

following

hypoxia

Surveys of severe mental retardation (IQs of less than 50) reveal that it occurs among 34 per 1000 children in the United States (Table 1). Although some obstetrical authorities have believed that half the cases of mental retardation could be prevented by the use of electronic fetal monitoring, events during labor and delivery (of which intrapartum hypoxia is only one) cause only about 8% of the total cases of severe mental retardation (Table 2). Among full-term infants, even the most severely asphyxiated infants appear to develop normally. About 90% of black and 95% of white surviving children in the Collaborative Perinatal Study who had 5-min APGAR scores of 3 or less were found to be of normal intelligence several years later. The National Institute of Child Health and Human Development Task Force estimated the maximum number of cases of cerebral palsy and severe mental retardation potentially preventable through the universal use of EFM to be one per 1000 live births across all categories of.risk.

Reactions to signs of intrapartum

risk

In 8 randomized controlled trials of the use of EFM, conducted in several European countries, the United States and Austrialia, there was no consequent reduction in perinatal death [13-201. However, in many of these studies operative

Table 1 Some estimates given age

of prevalence of severe and mild mental retardation

per 1000 population

Location

Group

Rate/l000 severe

Rate/1000 mild

Oregon, U.S. (1962) Aberdeen, Scotland (1970) Onondaga, New York (1955) Middlesex, U.K. (1962) Quebec, Canada (1973) Netherlands (1976) Isle of Wight, U.K. (1970)

12-14 8-10 5-7 10-14 10 19 9-14

3.3 3.7 3.6 3.6 3.8 3.7

30.3 23.7

Source: Antenatal Diagnosis, Report of a Consensus Conference, National Institute and Human Development, NIH, Bethesda, Maryland, U.S. Department of HEW, No. 80-1973. 1979.

of a

31.0 25.3 of Child Health NIH Publication

303 Table 2 Estimated current distribution oped countries

(%.) of selected causes of severe mental retardation

Distribution

Cause

Chromosomal Congenital Genetic

in devel-

anomaly malformation

metabolic

syndromes

with recurrence

risks

errors

Prenatal Perinatal causes: hypoglycemia,

birth, trauma, hypoxia, intracranial hemorrhage

hyperbilirubinemia,

Infections: Prenatal Perinatal Postnatal

Reported

Inferred

36

36

20

27

7

8

8

8

8

9

2 2 2

4 5 3

Source: Antenatal Diagnosis, Report of a Consensus Conference, National Institute and Human Development, NIH, Bethesda, Maryland, U.S. Department of HEW, No. 80-1973, 1979.

of Child Health NIH Publication

deliveries increased among those who were electronically monitored. In the United States, C-sections increased; in Ireland, there were more forceps deliveries; in other countries, more vacuum extractions (Table 3). The costs of these procedures are surely of concern, but the health risks are even more important because, for example, C-section carries its own risks to mothers. Primary C-sections are almost always followed by second C-sections. C-sections are more often accompanied by increased rates of infection than are vaginal deliveries. In the summer of 1986, an Table 3 C-section rates in 8 randomized

EFM trials AUS (%)

EFM (%)

EFMIFBS (%)

Selective monitoring (%)

Haverkamp [13] (n=483) Renou [14] (n=350) Haverkamp [15] (n=695) Kelso [16] (n=504) Wood [ll] (n=828) Neldam [8] (n=968) McDonald [19] (n=12,964) Leveno [20] (n=34,995) * Vacuum **Forceps

7 14 6 4 2 3.6 2.2

extraction was significantly deliveries were significantly

14 _ 18 9

_ 22 12

_ _

4 5.7 2.4 11.0 more frequent more frequent

10.2 in EFM group (P = 0.003). (P < 0.0001) in EFM group.

P < 0.01 P < 0.05 P < 0.01 P < 0.05 P = NS P=NS* P=NS** P < 0.05

304

NIH Consensus Conference reported a small but increased risk of deep venous thrombosis among women undergoing C-section [21]. In 1978, when United States C-section rates were lower than they are today, an Office of Technology Assessment report indicated that nearly 3 times as many women who are electronically monitored have C-sections as those monitored by stethoscopes. The cost of delivery was at that time increased from about $700 to $3000, adding an estimated $175 million to the national health bill. This estimate excluded the cost of any morbidity or mortality related to the C-section itself [22]. These costs result from the cost of caring for the increased incidence of respiratory distress syndrome, infection, the less frequent complications of herniation, and thrombophlebitis. Direct and indirect costs of delivering EFM services, C-sections, and various outcomes were estimated by Banta and Thacker [23]. The rarely occurring deaths of mothers (30 deaths per 96500 C-sections) between the ages of 30 and 34 would cost the United States $114,057 for each death. They write: ‘The estimate of total cost of $411 million per year compares to $80 million spent annually for all public and private childhood immunization programs.’ The management of what are held to be intrapartum risk factors has contributed to the creation of yet other risks, those of operative deliveries. The risks associated with C-sections are indeed believed to be more serious than those of vaginal operative deliveries. In their review of obstetrical interventions in Europe, Bergsjo et al. [24] have questioned the medical justification for the rise in C-section rates. The variation in perinatal mortality rates observed in European country data did not appear to be attributable to operative deliveries. Perinatal death was not reduced by these interventions. Costs were not discussed. Thus far we have chosen to emphasize the management of intrapartum risk factors, rather than those assessed during prenatal screening and antepartum care. How valid are these risking systems that are constructed for diagnostic and for routine screening purposes? Let us look briefly at the content of risking systems to see how well they do what they are intended to do. There are more than 20 risking systems in use at the present time. A composite risking system is shown in Table 4. Risking systems are used in order to provide a rational basis for identifying differential prophylactic or therapeutic interventions that may be required during the course of the pregnancy, labor or delivery. They are not always robust. There are varying estimates in the literature, but most obstetrical researchers would agree Table 4 Obstetrical (A) (B) (C) (D) (E) (F) (G)

risking

Mother’s social, educational, and other demographic Past medical and surgical history Past obstetrical history Present health status Present obstetrical history Present labor and intrapartum evaluation Gestational age of the fetus

characteristics

305 Table 5 The ability of risk assessment

methods to predict neonatal complications Sensitivity

Specificity

False

(“ro)

(%)

(%)

+

False (%)

-

Classified as low risk but with problem (%)

Hobel ( 1979)26 Prenatal factors Prenatal & intrapartum Edwards (1979)27 Winters (1979)28 Source:

Selwyn,

B.J.,

factors

54 94 75 52

72 28 63 70

in Research Issues in the Assessment

28 72 37 38

of Birth Settings

46 6 25 38

12 6 12 39

[25] pp. 156-160.

that about a third of the problems that can arise during a pregnancy, actually arise at the time of delivery, during the third stage of labor, and are not predictable by the systems presently in use. It is obvious that when risking systems work well there will be few false negatives and false positives. The sensitivity of a risking system indicates its ability to correctly identify patients with a given problem or condition. A high false negative rate is one measure of predictive inaccuracy and the insensitivity of the risk assessment method. Specificity is an indication of the instrument’s ability to identify those mothers without a problem. Thus, a high false positive rate is a measure of the non-specificity of the test, or the number of mothers inappropriately called ‘high risk’. Many mothers who are inappropriately allocated to a ‘high risk’ category may be needlessly subjected to intervention, while those who are inappropriately called ‘low risk’ may not receive needed care. Accuracy is critical and is only achieved when those who are called ‘high risk’ are actually those who experience complications and when those called ‘low risk’ are indeed free of problems. Since many risk factors are successfully managed during pregnancy (e.g. the control of gestational diabetes or maternal infection), a risking system which is used periodically throughout the pregnancy may result in lowered or improved scores as to the mother’s risk status. On the other hand, scores may increase as conditions arise which suggest later problems. Not all risk assessment systems are used throughout the course of the pregnancy. Some risking systems are employed during the first prenatal visit and then set aside. The comparison of ‘scores’, therefore, is not always possible. One needs to know not only the system used, but the frequency of its use, and the time period in the course of the pregnancy that a score was assigned. Further, some are not intended for use during the intrapartum period. In Table 5 we can see that for three commonly used and well-respected risking systems, there are problems of accuracy. The Committee on Assessing Alternative Birth Settings [25] has discussed the assets, liabilities and potential traps that await the well-intentioned practitioner. The ability of most risking systems to predict perinatal death is greater than their ability to predict neonatal complications [26--281. Many of the problems of labor defy prediction and previously established

306

risk status, high or low, does not help the physician to deal with a difficult breech or maternal hemorrhage. At the present time, with high sensitivity and low specificity, women may often be treated as being at high risk, when in fact they are not. There is a tendency to overcall problems, to intervene, to manage a pregnancy at greater expense, sometimes only a monetary expense, but sometimes creating health risks along the way. We must accept the caveat that we will never arrive at the perfect risking system. In the absence of full predictive power, we must be careful not to treat all pregnancies as high-risk pregnancies because of the uncertainty of outcome. It is important to turn our attention and research focus on this problem, if managers of risk, the maternal health care providers, are to have a sound basis for their decisions. It is sometimes disappointing to an epidemiologist that hospital administrators, health planners, insurers, and even obstetricians and midwives do not press harder for information, asking: is that risk really a risk? Is it more harmful to manage it than to ignore it? And, finally, how much does it cost us to treat the risks that may be fictional, or more gently, hypothetical rather than demonstrated?

References 1 Beery, W., Schoenbach, V.G., Wagner, E.H. et al., Health risk appraisal: methods and programs, with Annotated Bibliography, NCHSR, DHHS Publication No. (PHS) 86-3396 (1986) Forward. 2 Marieskind, H. Women in the Health System, C.V. Mosby, St. Louis, MO, 1980, pp. 247-248. 3 Oakley, A., History of technology in birth, paper presented at Interregional Conference on Appropriate Technology for Birth, AMRO/EURO, Fortaleza, Brazil, April, 1985 (ATB/SC 4.4). 4. For a comprehensive treatment of this subject, see Oakley, A., The Captured Womb: A History of the Medical Care of Pregnant Women, Oxford, Blackwells, 1984. 4 Oakley, A., op. cit., 5. 5 Banta, H.D. and Thacker, S.B., Costs and benefits of electronic fetal monitoring: a review of the literature, NCHSR, DHHS Publication No. (PHS) 79-3245 (1979) 1. 6 Mold, J.W. and Stein, H.F., The cascade effect in the clinical care of patients, New England Journal of Medicine, 314:8, Feb. 20 (1986) 512-514. 7 Lilien, A., Term intrapartum fetal death rates, American Journal of Obstetrics and Gynecology, 107 (1970) 595-603. 8 Chase, H.C. Perinatal mortality: overview and current trends, Clinical Perinatology, 1 (1974) 3-17. 9 Committee to study the prevention of low birthweight, Institute of Medicine, Prevention of Low Birthweight, National Academy Press, Washington, DC, 1985, p. 51. 10 Thompson, M. and Cohen, A.B., Uncertainty concerning electronic fetal monitoring, working draft, April, 1980. 11 Antenatal diagnosis: report of a consensus conference, National Institute of Child Health and Human Development, NIH, Bethesda, Maryland, U.S. Department of Health, Education, and Welfare, Publication No. 79-1973 (1979). 12 Nelson, K.B. and Ellenberg, J.H., Antecedents of cerebral palsy: multivariate analysis of risk, New England Journal of Medicine, 315:2, 81-86. 13 Haverkamp, A.D., Thompson, H.E., McFee, J.G. et al., The evaluation of continuous fetal heart rate monitoring in high-risk pregnancy, American Journal of Obstetrics and Gynecology, 125 (1976) 31&320. 14 Renou, R., Chang, A., Anderson, I. et al., Controlled trial of fetal intensive care, American Journal of Obstetrics and Gynecology, 126 (1976) 470-476.

307 15 Haverkamp, A.D., Orleans, M., Langendoerfer, S. et al., A controlled trial of the differential effects of intrapartum fetal monitoring, American Journal of Obstetrics and Gynecology, 154 (1979) 399-408. 16 Kelso, A.M., Parsons, R.J., Lawrence, G.F. et al., An assessment of continuous fetal heart rate monitoring in labor, American Journal of Obstetrics and Gynecology, 131 (1978) 526532. 17 Wood, R., Renou, R. et al., A controlled trial of fetal heart rate monitoring in a low-risk population, American Journal of Obstetrics and Gynecology, 141 (1981) 527-534. 18 Neldam, S., Osler, M., Hansen, P.D. et al., A controlled trial of the fetal heart rate monitoring during labor in a combined low- and high-risk population, in press, European Journal of Obstetrics and Gynecology, 1986. 19 McDonald, D., Grant, A., Sheridan-Pereira, M. et al., The Dublin randomized controlled trial of intrapartum fetal heart rate monitoring, American Journal of Obstetrics and Gynecology, 152 (1985) 524-539. 20 Leveno, K.J., Cunningham, F.G., Nelson, S. et al., A prospective comparison of selective and universal electronic fetal monitoring in 34,995 pregnancies, New England Journal of Medicine, 315:lO (1986) 615-619. 21 Prevention of venous thrombosis and pulmonary embolism, National Institutes of Health, Consensus Development Conference Statement, 6:2 (1986) 5. 22 Assessing, the efficacy and safety of medical technologies, Office of Technology Assessment, Congress of the U.S. (1978) 41. 23 Banta, H.D. and Thacker, S.B., Costs and benefits of electronic fetal monitoring: a review of the literature, National Center for Health Services Research (1979) DHEW Publication No. (PHS) 793245, 16-17. 24 Bergsjo, P., Schmidt, E. and Pusch, D., Differences in the reported frequencies of some obstetrical interventions in Europe, British Journal of Obstetrics and Gynecology, 90 (1983) 628-632. 25 Committee on assessing alternative birth settings, Institute of Medicine and National Research Council, Research Issues in the Assessment of Birth Settings, National Academy Press, 1982; see especially Chapter 3, Risk assessment, pp. 45-54 and the excellent chapter by Selwyn, B.J., Review of obstetrical risk assessment methods, pp. 149-170. 26 Hobel, C.J., Assessment of the high risk fetus, Clinics in Obstetrics and Gynecology, 6 (1979) 367-377. 27 Edwards, L.E., Barrada, M.I. et al., A simplified antepartum risk-scoring system, Obstetrics and Gynecology, 54 (1979) 237-240. 28 Winters, S., Itzkowitz, S. et al., Prenatal risk assessment: an evaluation of the Hobel record in a Mount Sinai clinic population, Mount Sinai Journal of Medicine, 46 (1979) 424-427.