Click for next page ( 33


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 32
2 Research Strategies for Assessing Childbirth Settings Tbis chapter reviews general research designs and indicates the partic- ular strengths and weaknesses of several of th•• Investigators should consult available texts on research design for .ore exhaustive treat- aents of research aethods (e.g., Callpbell and Cook, 1979r Callpbell and Stanley, 1963r cochran and cox, 1957r Ricks, 1973r Spector, 198lr Susser, 1973 r Winer, 1971) • (Appendix D discusses sa.e of the aethocJo- logical issues concerning the assessaent of psychological variables and identifies a number of gaps in inforaation.) Whatever design is used, the committee believes that assessaent of the safety and efficacy of birth settings should receive priority in research. Also iaportant is an assessaent of the psychological benefits of one birth setting versus another. A research strategy consists of three el...ntsa the research design (randOIDized or nonrandOIDized), the type of data collection (prospective or retrospective), and the aethods of analysis. Tbe first two eleaents are discussed in this chapter. Tbe choice of the strategy reflects the scientific questions to be answered and the extent to which an investi- gator can intervene in a continuing process. Several strategies exist for planning experi..ntal and observational investigations, ranging frOID designs in which investigators have control of .oat of the variables to designs in which the investigator cannot specify all of the conditions. Often the choice of a study design is dictated by the circuastances in which data are collected. Ordinarily the data aay be collected pros- pectively or retrospectively. The term •prospective• indicates that data will be collected specifically for the purposes of the study in questionr the term •retrospective• ..ana that the data for the study will be obtained frOID one or more existing data sets. One of the main distinctions between prospective and retrospective data-gathering sys- tems is that the prospective data collection can be specially designed and t.plemented to relate directly to specific hypotheses. In con- trast, retrospective studies attempt to make use of data already collected, generally for other purposes. Nearly all phenomena have variables1 that affect the outcomes (see Chapter 4). Por exDple, birth weight, social class, JDOther's I I 1A variable is a characteristic whose value can change frOID subject I to subject. Anything that can be measured, counted, weighed, or I 32 I I I J

OCR for page 32
33 age, and parity are tmportant variables affecting perinatal mortality (see Table 1). These factors must be accounted for when comparisons are made of subjects, to reduce the effect of prognostic factor bias. Biases can be reduced by the specific study designs (Cohen and COhen, 1975r England, 1975r Hayden et al., 1982r Lancaster, 1974r Lilienfeld and Lilienfeld, 1980) and by statistical methods, such as regression models, logistic and loglinear models, and proportional hazard models (Breslow and Day, 1980r Draper and Smith, 1966r Marsden, 1981). DESCRIPTIVE AND OBSERVATIONAL STUDIES In some situations all the conditions under which the study is to be conducted cannot be specified. Nevertheless, it may be possible to learn much about a process by using an organized system of data collection. The committee distinguished between descriptive studies that rely on available data and observational studies where new data is collected. The aoat common types of descriptive studies in maternity care are baaed on collections of vital statistics undertaken by federal and state health agencies. These vital statistics are important in documenting trends and in supplying ideas for further investigations. The application of statistical methodology to vital statistics can result in the identification of tmportant associations between popula- tion characteristics and the occurrence of disease. However, because moat collections of descriptive data are not organized to answer speci- fic scientific questions, caution must be exercised in their use. Por ex.-ple, if the data come from many different states, care must be taken to ensure that the same definitions are used by all states. Neverthe- less, studies baaed on descriptive data have been and will continue to be important sources of information in the social and health sciences (Williaaa, 1979r Williams and Hawes, 1979r Williaaa and Chen, 1982). Proper attention to data quality and to the inclusion of potentially relevant variables is necessary for adequate research design. In contrast to a descriptive study that relies on vital statistics, an observational study is one in which an organized system of data col- lection is introduced to examine aoae specific features of the phenome- non under study. Por ex.-ple, Lubic (1980) determined the outcoae of clients in a freestanding birth center in New York City. Table 2 traces a population of 1,965 women from their initial appearance at an orienta- tion session at a freestanding birth center to delivery by 455 of these women at the center. Careful prospective but uncontrolled observational studies of this type can make valuable contributions to understanding aspects of a birth setting and ita problema. scored--a property, a characteristic, an event, an effect, an object-- may vary from subject to subject in the same group or in the same sub- ject at different tt.ea and under different circumstances. (See Chapter 4 for aore details.)

OCR for page 32
34 TABLE 1 The Effect of Prognostic Factors on Perinatal Mortality Maximum/ Factor Minimum!. Comparison Groups Birth weight 23 Less than 2.5 kgr greater than 2.5 kg Social class 5 Unsupported mothersr social class 1 Age of mother 2 35+r 20-24 Parity 2 4+~ 1 ~tio of perinatal mortality rates by extremes of values within variables. SOURCE: Chamberlain et al., 1975. RANDOMIZED EXPERIMENTAL DESIGNS When comparing alternative treatments or methods in a controlled study, it may be possible to assign the treatments or methods such that each subject has the same opportunity to receive any of the treatments under investigation. Assignment of treatments or methods to subjects is usu- ally accomplished by the mechanism of randomization (Zelen, 1974). The role of randomization is to make the groups receiving the different treatments •alike on the average.• Because any known or unknown factors that may affect the outcome are distributed randomly, interpretation of the outcome is usually unambiguous (Byar et al., 1976). Such experi- ments are called randomized clinical trials (Gore, 198la,b,c,d). Bow- ever, they sometimes present difficulties in execution. These diffi- culties may involve ethical issues associated with choosing treatments by chance, complicated logistics introduced by the randomization mech- anism, the need for patient consent when human beings are involved, and the unexpected refusal of patients or physicians to agree to receive or administer the randomly assigned treatment. Sometimes these problems can be anticipated and minimized by the experimental design (Gehan and Frereich, 1974r Simon, 1979). ~e use of randomized allocation rules to compare different birth sites may be difficult if it is necessary for the prospective mother to be assigned to a center or hospital that is different from the preferred place of delivery. It does not seem likely that a prospective mother or her physician would consent to enter a study wherein delivery would take place in a site that neither she nor her physician would prefer. Nevertheless, because randomization is an optimum way to study interven- tions, opportunities for randomization of women to different birthplaces should be sought. Examples of such opportunities are (1) when a pro- spective mother is of a divided mind about two different birth sites or (2) when the women choosing home births may be willing to be randomly assigned to a home-like birth room in a hospital. Use of randomization may also be feasible when different technol-

OCR for page 32
35 TABLE 2 Patient Outcomes and Flow of Patients at a Freestanding Birth Center Number of mothers who appeared for physical screening: 1,166. (101 women deemed ineligible for participation in study.)~ Number of eligible patients: 1,065. (167 women were still pregnant and awaiting delivery in the program at time of report publication' 42 women had spontaneous abortions, and 109 women withdrew or transferred.) Outcomes of the 747 eligible patients: 189 women were antepartum transfers. Reasons for antepartum transfer included obstetrical problems (134 cases), such as ruptured membranes with no labor in 12 hours, nonvertex presentation, premature labor, and post datism, ~ pathophysiologic problems (40 cases), and circumstantial (15 cases). Two families whose labors were managed entirely in hospitals experienced neonatal death. 103 women were intrapartum transfers. Reasons for intrapartum transfer included delay in labor (56 cases), hypertension (16 cases), meconium staining (14 cases), prolonged second stage of labor (10 cases), nonvertex presentation (5 cases), fetal brachycardia (4 cases), and no fetal heart tones (2 cases). Of these women, 27 had cesarean sections. 455 women gave birth at the center. Patient transfers among the 455 freestanding birth center deliveries: 1 women were postpartum transfers. Reasons for transfer included retained placenta (2 cases), irregular vital signs (2 cases), labial hematoma (1 case), inspection and repair under general anesthesia (1 case), and hypertension (1 case). 11 infants were transferred to the hospital. Reasons for transfer included mild respiratory distress (6 cases), birth weight less than 2,500 grams (3 cases), appearance of clinical postmaturity (1 case), and question of sepsis (1 case). One infant experienced sudden death, in the second day of life at home. !Reasons included pathophysiological problems determined by examination (41 cases) or by history (35 cases), nonpathological problems (23 cases), and circumstantial or not specified (2 cases). SOURCE: Lubic, 1980. ogies or methods are being studied within the same birth center. There are a large number of studies that have effectively employed such de- signs for study of a particular maternity care practice. For example, randomized clinical trials have been conducted to examine the effect of such variables as the position of the mother during delivery (Humphrey et al., 1973), the presence of a supportive lay person during delivery (Sosa et al., 1980), use of electronic fetal monitoring (Haverkamp et

OCR for page 32
36 al., 1976, 19791 Kelso et al., 1978r Renou et al., 19761 Langendoerfer et al., 1980), whether the mother received •extra contact• with the infant (Kennell et al., 1974r Ringler et al., 1975) or •rooming in• (Greenberg et al., 1973), the timing of initial contact between infant and mother after birth (Bales et al., 1977), and whether the initial contact was with a wrapped infant or was •akin-to-akin• (CUrry, 1979). One study compared the Leboyer method (an approach to birth that emp~s a specific technique to minimize a neonate's first separation experi- ence) with a control group of mothers giving birth in the same hospital without this method (Nelson et al., 1980). MATCHED GROUPS In many investigations where randomization is not possible, the use of the matching method can reduce or eliminate prognostic factor biases. The subjects assigned to the treatment group are •matched• to nontreated control subjects individually in terms of prognostic factors. A vari- ant of matched groups that is widely used in epidemiological studies is the retrospective case-control study (Hayden et al., 1982). These are especially useful when attempting to associate the occurrence of a rare disease with a causal factor. There are many variants of matched group and case-control designs, all of which share the objectives of reducing biases arising from prognostic factors. Although the most c01a10n use of matched groups is in instances for which the data of both groups is retrospective, it is also possible to conduct a study in which the matching is done initially and the data collection is prospective. One problem with matching is that it is difficult to match on more than a few variables unless one bas a large pool of control patients. However, this problem may be resolved through the use of statistical procedures. Prospective studies using carefully matched groups of women who deliver in different settings could be used for assessing birth set- tings. Both selection bias and bias in obtaining information would need to be considered by researchers when matching groups of women. However, the committee recognizes that is difficult, perhaps impossible, to eliminate bias completely when using this research approach. women who self-select nontraditional birth care services may have characteristics that are different fr011 other women, and the difference may not be accounted for when matched on demographic and health/obstetrical history variables classically associated with outcomes of pregnancy. Almost all prospective or retrospective studies will have to rely on data col- lected by those providing care to the subject mothers. Differences in training of the care providers, as well as different conceptual ap- proaches to childbirth, will affect tbe data collected. Despite these problems, studies with rigorous prospective monitoring of planned deliveries in different sites could begin to provide essential data on the safety of care. Furthermore, such studies are less costly than some other approaches. Study of psychological variables using this approach could also begin to provide information on the benefits of different settings. At least two prospective studies are in progress (see Appendix A).

OCR for page 32
37 case-Control Studies In case-control studies (Hayden et al., 1982), a group of individuals, all of whom were subject to the event under study, are matched with a control group chosen from a pool of individuals who did not experience the event. One or more individuals in the control group are matched with each case on known prognostic variables (age, sex, etc.). Analysis is made of the frequency of the hypothesized causal factor among cases and controls. If planned nonhospital delivery is found more frequently among cases than among controls, for example, this may be taken as evi- dence for a differential effect of place of delivery on the adverse event. case-control studies may be one of the least costly ways of examin- ing factors associated with events of low frequency. However, findings may be distorted by bias in the way cases are selected or by the way information is obtained or collected. Both of these biases are especi- ally problematic in research on birth settings where patients themselves select specific settings and when different providers (who have their own biases) collect the information. case-control studies should be viewed, for general assessment pur- poses, as a secondary option. However, this method may be useful for certain unusual outcomes. Por example, if an outbreak of staphyloccocal infection in neonates that results in hospitalization is recognized in a community, it could be useful to match such cases to infants free of the disorder and compare the birth locations of each group. SURVEILLANCE METHODS Another promising research strategy is to have a surveillance mechanism that monitors adverse events as they occur, so that corrective action can be taken. In its simplest form, survei-llance Mans maintaining a count of certain predesignated events (Rutstein et al., 1976). This count may, in itself, be of interest, but more usually it serves as the starting point for more intensive investigation. Its greatest utility is in situations in which the presence of a single adverse event man- dates a chain of public health activities, as, for example, in the presence of an infectious disease. An example of surveillance is the Abortion Surveillance Program of the Centers for Disease Control (Centers for Disease Control, 1979, 1980J Cates, 1982J cates et al., 1978). Most state health departments require the reporting of an induced abortion. Hence, with the use of an appropriate denominator population (e.g., the number of women in cer- tain age groups), it is possible to calculate the relative frequency and characteristics of induced abortion in the United States overall, as well as by geographic area. Similarly, vital statistics data can be used to obtain counts of maternal deaths. When combined, these two frequencies can be used to obtain maternal mortality rates associated with abortions of different kinds, at different gestations, etc. In a direct analogy to birth practices, the Abortion Surveillance Branch of the Centers for Disease Control has been able to examine the effect on

OCR for page 32
38 maternal mortality of abortions performed in and out of hospitals (Grimes et al., 1978). Special studies often are added onto the routinely collected data of surveillance studies to examine, for example, the specific circum- stances surrounding maternal death (cates and Jordan, 1979) or the effect of abortion-restricting legislation on abortion-related complications (Cates et al., 1979). Another use of the surveillance mechanism is to detect geographic and temporal clusters of adverse events. These can often point the way to specific problems in the system under surveillance (Centers for Disease Control, 1980). ASSESSING ADVERSE EVENTS Comparisons of different birthplaces, birth practices, maternity care providers, and populations all rest on a system of measuring the fre- quency of adverse events. Moreover, a judgment must be made as to whether the adverse event could have been avoided through some inter- vention or change in the birth setting. The suggestions that follow are frameworks for these kinds of comparisons, the method of choice largely depends on the nature of the available data. FOr example, in the absence of denominator data (a count of all deliveries at several delivery sites), case-control methodologies or perinatal audits may be the only feasible options. Given the availability of denominator data, a range of opportunities for assessment appears. The discussion that follows deals with the identification of adverse events and their use in evaluating birth settings. A convenient and practical way of classifying adverse events is by three categories of data that document the events: adverse events doc- umentable through vital statistics data alone, adverse events requiring the collection of special data, and adverse events whose definitions are based on expert opinion. Use of Vital Statistics The routine recording of births and deaths in all states of the United States can serve as a useful starting point for analysis of risks to mothers and infants as mediated by place of delivery and care provider. Events universally recognized as adverse and routinely recorded on vital certificates include maternal and perinatal deaths. In most reporting areas in the United States, documentation of low birth weight or preterm delivery is possible with reasonable accuracy. In some reporting areas, low Apgar scores and complications of pregnancy and labor are recorded. Although not all of these events can be regarded as avoidable, their presence in a planned nonhospital delivery may reflect a failure of the screening process (see Chapter 3). The vital record data can be used to refine the definition of an adverse event, so as to obtain a better sense of the need to follow up the event to find an assignable cause. FOr example, death from labor asphyxia in a term baby weighing more than 2,500 grams with no reported

OCR for page 32
39 anomalies might be considered evidence of lack of optimum application of available resources. It is unlikely, however, that vital statistics data by themselves would be sufficient for a rigorous and fair analysis of rates of avoid- able adverse events. At best, they can provide early warning signals, the starting point for more detailed study in a system of surveillance. The use of birth weight standardization for purposes of comparing peri- natal mortality rates is appealing (Paneth, 1982), but it is likely that the number of deaths found in comparisons of birth locations will be too small to make standardized rates meaningful. Maximum utility of vital certificates will be achieved only if circumstances of delivery are clearly definable. Collection of Relevant Special Data The limitations of vital statistics data argue for considering proce- dures for systematically obtaining data on the circumstances of delivery and the postnatal complications of mothers and children. Birth certifi- cates are completed as close to the time of birth as possible, and later morbidity data for mother or child cannot be obtained from such a source. Birth certificates are a poor source of information on congeni- tal malformations because many of these disorders are not manifest at the time the certificate is filled out. One way to monitor complications is to monitor adverse events in a community at large, for example, hospitalizations for infections of infants in the first three months of life. These complications could be linked with birth settings. No present surveillance mechanism exists for such phenomena, but hospital discharge summary data could be used as a basis for such surveillance, as it is for congenital malformations (Edmonds et al., 1981). Use of Expert Opinion In this system, adverse events (deaths, serious illness, etc.) would initially be signaled by examination of vital statistics data. The events would be reviewed by a panel of experts to determine the degree of preventability of the event in question. Their assessment would be based on a review of the medical chart, autopsy report, laboratory find- ings, and any other pertinent information. The initial signaling event would be agreed upon in advance, but the assessment of preventability would be based on expert judgment (Rutstein et al., 1976). Considerable experience has been accumulated in this method of assessment, initially with maternal mortality committees (Grimes and Cates, 1977) and, more recently, with groups performing •perinatal audits• (Mersey Region WOrk- ing Party on Perinatal Mortality, 1982). One advantage of monitoring adverse events by such a procedure is that they may be interpretable without reference to a denominator pop- ulation. The simple presence of any preventable adverse event may be taken as evidence of the need for improvement in that birth location, regardless of the number of such events.

OCR for page 32
40 Adverse Events that Reflect Failure of Risk Prediction (Screening) A general assumption underlying the discussion of alternate birth set- tings is that planned nonhoapital deliveries will be carefully screened beforehand to select only low-risk mothers. Thus, certain kinds of deliveries at planned nonhoapital locations attest to a failure of the screening process, e.g., ·1ow birth weight newborns, and deliveries by mothers with hypertension or diabetes. These kinds of events are easily detectable using moat present birth certificate systems, as long as the planned delivery site can be ascertained. certain methods of obtaining information on adverse events have been detailed above. However, the relationship of these adverse events to place of delivery must be baaed on calculation or estimation of rates for such events at various delivery locations. Direct calculation of rates requires data on the total number of deliveries from which these adverse events arise, i.e., data on the denominator population. As discussed in Chapter 1, available vital statistics do not give a reliable count of the number of deliveries in the different delivery schemes because state birth certificates generally do not contain pro- vision for place of intended delivery. Moat certificates do note whether the delivery was in a hoapitalJ however, the hospital category includes births at freestanding birth centers and deliveries recorded as nonhoapital ones include unplanned deliveries in many diverse loca- tions (taxi, home, street). These vital statistics are biased toward prematurity, because the frequency of precipitate deliveries is in- versely related to gestational age. Unplanned nonhoapital delivery also may be related to lack of prenatal care andVor nonacknowledgwent of pregnancy. For all of these reasons, perinatal mortality rates for non- hospital deliveries are invariably higher than for hospital deliveries (Burnett et al., 1980). Assessment of rates of adverse events by birthplace would be greatly assisted by the incorporation into birth certificates of data that would clearly distinguish planned from unplanned nonhoapital deliveries. A strong case can be made for recording planned nonboapital deliveries as a data item on all birth certificates. The data would be especially useful when combined with data on the training of the attendant at delivery, a variable now recorded in many states. If this information were available, it would be relatively simple to combine counts of adverse events, however determined, with the population at risk and thus to generate rates for such events at different delivery places. Such recordkeeping would allow for another objective of aurveillancea monitoring of the frequency and characteristics of planned nonhoa- pital deliveries, which at this time is not possible. COOPERATIVE REGISTRIES Another approach to obtaining data for the evaluation of different birth settings is to organize groups of hospital and nonhoapital birth centers that will cooperate in submitting data to a central collection center. The aim of such a cooperative registry would be to collect

OCR for page 32
41 uniform infor.ation on all births. Eventually, a data base could be built that might answer questions on the quantitative aspects of birth practices. Because all states require the submission of birth informa- tion, both hospital and nonhospital settings are familiar with routine recordkeeping functions. Expanding the routine recordkeeping activities can provide a research data base that bas uniform definitions and ade- quate quality-control checks. Information from a variety of birth cen- ters could be collected, which would permit an evaluation of various birth settings. Furthermore, the data could include tmportant prog- nostic factors, so that subsets within the population of mothers and infants could be compared properly. T.he drawbacks of this approach are that very large samples would be required to study rare events of interest. A registry would require much planning and standardization of collection procedures, which would be costly. Also, it would take a long tt.e to accumulate and analyze the data. And since birth prac- tices are changing so rapidly, the results might be outdated by the tt.e they became available. T.he formation of a cooperative registry would require selecting a range of birth settings. In effect, a cooperative registry was the aethod of the Collaborative Perinatal Project of the National Institute of Neurologic and Communicable Diseases and Stroke (NINCDS), which studied 50,000 pregnancies at 12 institutions between 1959 and 1966 (Niswander and Gordon, 1972). Another such effort is the Obstetric Statistical Cooperative formed by several Brooklyn hospitals in 1950 that has now accumulated a sufficiently large population so that rare aalfor.ations can be studied (Stein et al., 1982). SUMMARY T.he coaaittee concludes there are a number of different research designs that can be used to study alternative birth settings. T.he frequency of the outcome chosen for study will determine to a large degree the re- search strategy. A very rare event will require very large sample sizes and may only be feasible to study by a surveillance, registry, or case- control approach. Because of continued controversy and the growing num- ber of different birth settings, the safety and efficacy of these set- tings is a high-priority matter for research. Recommended research designs or methods for collecting data include randomized clinical trials wherever possible, matched groups or cohort studies of low-risk women delivering in different settings, and surveillance of live births and their complications together with special data collection and methods for evaluating adverse events. Other possible approaches are establishing a registry in order to collect data for evaluating mater- nity care in a number of different institutions and settings and case-control studies of adverse events.

OCR for page 32
42 REFERENCES Breslow, N. B., and N. E. Day. 1980. Statistical Methods of Cancer Research. Lyon, France: International Agency for Research on cancer. Burnett, c. III, J. A. Jones, J. Rooks, c. Chen, c. w. Tyler, and c. A. Miller. 1980. Home delivery and neonatal mortality in North Carolina. Journal of the American Medical Association 244:2741-2745. Byar, D. P., R. M. Simon, w. T. Freudwald, J. J. Schlesselman, D. L. DeMets, J. H. Ellenberg, M. H. Gail, and J. H. Ware. 1976. Random- ized clinical trials: Perspectives on some recent ideas. New England Journal of Medicine 295:74-80. Campbell, D. T., and T. D. Cook. 1979. Quasi-Bxperiaentation: Design and analysis for field settings. Chicago: Rand McNally. Campbell, D. T., and J. c. Stanley. 1963. Experimental and Quasi- Experimental Designs for Research. Chicago: Rand McNally. Cates, w., Jr. 1982. Legal abortion: The public health record. Science 215:1586-1590. Cates, w., Jr., J. c. Smith, R. w. Rochat, J. B. Patterson, and A. Dolman. 1978. Assessment of surveillance and vital statistics data for monitoring abortion mortality, United States, 1972-1975. American Journal of Epidemiology 108:200-206. Cates, w., Jr., and H. F. Jordan. 1979. Sudden collapse and death of women obtaining abortions with prostaglandin F2a. American Journal of Obstetrics and Gynecology 133:398-400. Cates, W., Jr., A. M. Kimball, J. Gold, G. L. Rubin, J. C. Smith, R. W. Rochat, and c. w. Tyler, Jr. 1979. The health impact of restricting public funds for abortion. American Journal of Public Health 69: 945-947. Centers for Disease Control. 1979. Abortion Surveillance, 1978. Atlanta, Ga.: Centers for Disease Control. Centers for Disease Control. 1980. Congenital Malformations Surveillance Report, April 1978-March 1979. Atlanta, Ga.: Centers for Disease Control. Chamberlain, R., G. Chamberlain, B. Howlett, and A. Claireux. 1975. British Births, 1970. London: William Heinemann Medical Books. Cochran, w. G., and G. M. Cox. 1957. Experimental Designs. New York: John Wiley & SOns. Cohen, J., and P. Cohen. 1975. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. Hillsdale, N.J.: Lawrence Erlbaum Associates. Curry, M. A. 1979. Contact during the first hour with the wrapped or naked newborn: Effect on maternal attachment behaviors at 36 hours and three months. Birth and the Family Journal 6:227-235. Draper, N., and H. Smith. 1966. Applied Regression Analysis. New York: John Wiley & Sons. Edmonds, L. D., P. M. Layde, L. M. James, J. w. Flynt, J. D. Erickson, G. D. Oakley. 1981. Congenital malformations surveillance: Two American systems. International Journal of Epidemiology 10:247-252. England, J. M. 1975. Medical Research: A Statistics and Epidemiological Approach. Edinburg: Churchill Livingstone. __ __.

OCR for page 32
43 Gehan, B. A., and B. J. Frereich. 1974. Non-randoainized controls in cancer clinical trials. New England Journal of Medicine 2901198-203. Gore, s. M. 198la. Assessing clinical trials--design I. British Medical Journal 28211780-1781. Gore, s. M. 198lb. Assessing clinical trials--design II. British Medical Journal 28211861-1863. Gore, s. M. 198lc. Assessing clinical trials--restricted randaaization. British Medical Journal 28212114-2117. Gore, s. M. 198ld. Assessing clinical trials--why randaaize? British Medical Journal 28211958-1960. Greenberg, M., I. Rosenberg, and J. Lind. 1973. Pirst mothers raa.ing-in with their newborna1 Ita impact upon the mther. American Journal of Orthopsychiatry 431783-788. Griaea, D. A., and w. Cates, Jr. 1977. The impact of state maternal mrtality coamitteea on aaternal deaths in the United States. American Journal of Public Health 671830-833. Grt.ea, D. A., w. Cates, Jr., and c. w. Tyler, Jr. 1978. eo.parative risk of death fro. legally induced abortions in hospitals and nonboapital facilities. Obstetrics and Gynecology 511323-326. Bales, D. J., B. Lozoff, R. Soaa, amd J. B. Kennell. 1977. Defining the liaita of the aaternal sensitive period. Developmental Medicine and Child Neurology 191454-461. Haverkamp, A. D., B. B. Tboapaon, J. G. McFee, and c. Cetrulo. 1976. The evaluation of continuous fetal heart rate mnitoring in high- r iak pregnancy. Aaer ican Journal of Obatetr ica and Gynecology 1251310-320. Haverkamp, A. D., M. Orleans, s. Largendoerfer, J. McFee, J. Murphy, and B. B. Thompson. 1979. A controlled trial of the differential effects of intrapartum fetal monitoring. American Journal of Obstetrics and Gynecology 1341399-412. Hayden, G. G., M. s. Kraaer, and R. I. Horwitz. 1982. The case-control study. Journal of the American Medical Association 2471326-331. Hicks, c. R. 1973. Fundamental Concepts in the Design of Bxperiaenta, 2d ed. New York1 Bolt, Rinehart, and Winston. Buaphrey, M., D. Bounalow, and s. Morgan. 1973. T.he influence of aaternal posture at birth on the fetus. Journal of the Obstetrics and Gynaecology of the British Commonwealth 8011075-1080. Kelso, I. M., R. J. Parsons, G. P. Lawrence, s. s. Arora, D. K. Bdaonds, and I. D. Cooke. 1978. An aaaeaaaent of continuous fetal heart rate monitoring during labor1 A randaaized trial. American Journal of Obstetrics and Gynecology 1311526-532. Kennell, J. B., R. Jerauld, B. WOlfe, D. Chesler, N.c. Kreger, w. McAlpine, M. Staffa, and M. B. Klaus. 1974. Maternal behavior a year after early and extended poat-partua contact. Developmental Medicine and Child Neurology 161172-179. Lancaster, B. o. 1974. An Introduction to Medical Statistics. New York1 John Wiley & Sons. Langendoerfer, s., A. D. Haverkamp, J. Murphy, K. D. Novick, M. Orleans, P. Pacoaa, and w. van Doorninck. 1980. Pediatric follow-up of a randaaized controlled trial of intrapartum fetal monitoring techniques. Journal of Pediatrics 971103-107.

OCR for page 32
44 Lilienfeld, A.M., and D. E. Lilienfeld. 1980. POundations of Epidemiology. New York1 Oxford University Press. Lubic, R. w. 1980. Evaluation of an out-of-hospital aaternity center for low-risk patients • .!!!. Health Policy and Nursing Practice, Linda B. Aiken, ed. New York1 McGraw-Bill. Marsden, P. v. 1981. Linear Models in Social Research. Beverly Bills, Calif.: Sage. Mersey Region Working Party on Perinatal Mortality. 1982. COnfidential inquiry into perinatal deaths in the llersey region. Lancet 11491-494. Nelson, N. M., M. w. Enkin, s. Saigal, K. J. Bennett, R. Miler, and D. L. Sackett. 1980. A randomized clinical trial of the Lebofer approach to childbirth. New England Journal of Medicine 3021655-660. Niswander, K. R., and M. Gordon. 1972. 'J.be Women and 'J.beir Pregnancies. Philadelphia1 w. B. Saunders. Paneth, N. 1982. Infant mortality r•:xamined. Journal of the Aller ican Medical Association 24711027-1028. Renou, P., A. Chang, I. Anderson, and c. WOOd. 1976. COntrolled trial of fetal intensive care. AMrican Journal of Obstetrics and Gynecology 1261470-476. Ringler, N. M., J. B. Kennell, R. Jarvella, B. J. Navojosky, and M. B. Klaus. 1975. Mother-t~hild speech at 2 years--effects of early postnatal contact. Journal of Pediatrics 861141-144. Rutstein, D. c., w. Berenberg, T. c. Chalmers, c. G. Child III, A. P. Pishllan, and E. B. Perrin. 1976. Measuring the quality of aedical care1 A clinical •thod. New England Journal of Medicine 2941 582-588. Simon, R. 1979. Restricted randomization designs in clinical trials. Biometrics 35a503-512. Soaa, R., J. Kennell, M. Klaus, s. Robertson, and J. Urrutia. 1980. The effect of a supportive companion on perinatal problems, length of labor, and mother-infant interaction. New England Journal of Medicine 3031597-600. Spector, P. 1981. Research Designs. Beverly Bills, calif.l Sage. Stein, s. c., J. G. Peldman, M. Priedlander, and R. J. Klein. 1982. Is mening~elocele a disappearing disease? Pediatrics 691511-514. Susser, M. w. 1973. causal 'J.binking in the Health Sciences• Concepts and Strategies. New Yorka Oxford University Preas. Williams, R. L. 1979. Measuring the effectiveness of perinatal aedical care. Medical care 17195-110. Williams, R. L., and Chen. 1982. Identifying the sources of the recent decline in perinatal DK>rtality rates in california. New England Journal of Medicine 306a207-214. Williams, R. L., and w. E. Hawes. 1979. Cesarean section, fetal monitoring, and perinatal mortality in california. Aaerican Journal of Public Health 69•864-870. Winer, B. J. 1971. Statistical Principles in Experimental Designs, 2d ed. New York• McGraw-Bill. Zelen, M. 1974. The randomization and stratification of patients to clinical trials. Journal of Chronic Diseases 271365-375.