Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 Variables Variables important in choosing a study population were described in Chapter 3 in connection with obstetric risk instruments. Furthermore, Chapter 2 pointed out how prognostic factors had to be carefully matched in research designs to ensure reliable conclusions. This chapter de- scribes variables of interest to researchers assessing birth settings. It also reviews methodological issues that arise with the measurement of variables and the collection of data. The committee was able to develop a list of variables that might be considered in the study of birth settings and to develop several approaches for their conceptual- ization. However, it became clear that more research is needed to develop accurate outcome measures other than mortality. The single exception may be Apgar (1953) scores, which can be easily and readily measured at the time of delivery. SOME GENERAL OBJECTIVES IN MEASUREMENT A variable is a characteristic whose value can vary from subject to subject. Anything that can be measured, counted, weighted, or scored-- a property, a characteristic, an event, an effect, an object--may vary in value from subject to subject in the same group, or for the same subject at different times and under different circumstances. Examples of variables include things like age, race, blood pressure, weight, and Apgar scores. variables may be quantitative or qualitative, their mea- surements will yield either discrete or continuous score values. The strength and magnitude of the relationships among variables of interest are used by investigators to describe and understand a problem as well as to draw inferences and conclusions. Independent variables are those thought to exert an influence on some outcome. Dependent variables show an effect or a change when the independent variables are manipulated. Variables can have different functions depending on what the investigator wishes to study. One investigator's dependent variable may be regarded by another investi- gator as an independent variable, according to each hypothesized chain of effects. In research, it often occurs that other factors, e.g., background, intervening, or confounding variables, exert an influence on the relationship between an independent and a dependent variable. 55
56 Their potential for influencing any observed associations should be kept in mind (Rosenberg, 1968). Attention will be given in this section to issues important in mea- surement, including assurance of reliability and validity, standardi- zation of measurement, and selection of variables. More detailed and extensive consideration of these general issues is available in many texts on research methodology and statistics, such as those by Campbell and Fiske (1959), Duncan et al. (1977), England (1975), Kerlinger (1973), Lancaster (1974), Miller (1981), Rosenberg (1968), and Selltiz et al. (1976). STANDARDIZATION OF MEASUREMENTS To study differences among birth settings, care should be taken to stan- dardize measurements so that experimental conditions are similar for all groups. With standardized conditions, the effects of extraneous variables are apt to be cancelled out. Standardization implies that written procedures exist for making measurements in the same fashion every time by all investigators. The investigator should describe how variables will be measured, the nature and use of the equipment for measurement, how data is recorded, the execution of the study, and the skills and training of staff (see Appendix E). Investigators may either follow established procedures or create new ones, but clear and exact explanations of what will be done will help ensure that the study is carried out as intended and that it can be replicated by others. Same variables have uniform definitions, e.g., age, educational level, and birth weight. If, however, such variables are defined in a different manner from their definition in existing literature, explana- tions must be given to support the need for this difference, because comparison of data between the proposed research and existing litera- ture might be made more difficult. Often, commonly used variables are assumed to be defined similarly, but this assumption may be misleading. Parity, for example, can be defined either to include only live births or to include all births, live or stillborn. The differing definitions could lead to differing results. The need to state criteria for vari- able definition is particularly great for newly created variables. RELEVANT EXAMPLES OF PROGNOSTIC VARIABLES Race, maternal age, parity, socioeconomic status, and medical obstetric risk level are well known influences on the outcome of pregnancy. Other variables, such as personality characteristics, attitudes, health behav- ior, and beliefs may also be important determinants of outcome. For example, Fullerton (1981) found a more positive attitude toward choices in childbirth and a greater desire to control their own health care among women who chose a nonhospital birth experience than among those who chose to deliver in the hospital. Specific psychosocial variables
57 relevant to pregnancy outcomes have not yet been identifiedJ studies to develop reliable and valid measures for these characteristics are needed (Yang, 1981). Variables known to affect outcomes may be used in selecting com- parison groups. In some research designs, study and comparison groups should be similar on known sociodemographic, medical-obstetric, and psychological characteristics. Selection bias can affect study results because individuals with specific characteristics may be included in the study while those without the characteristics are excluded. Self- selection is one of the most difficult problems in research on alternative birth settings and may be hard to overcome. RELEVANT EXAMPLES OF OUTCOME OR DEPENDENT VARIABLES Pregnancy outcome traditionally was measured by late fetal, neonatal, and maternal mortality. But those events have become so rare (with rates generally lower than 5 percent) that they no longer can be the only measure of quality of care. Morbidity is becoming a more frequent measure of pregnancy outcome. Morbidity is reported either as a cumu- lative score reflecting the total number of morbid conditions in the mother or infant, or as the incidence of selected individual morbid events. Some types of morbidity, such as infections in mother or infant, birth injuries, neonatal asphyxia, or excessive jaundice, can reflect the quality of care. Others, such as the incidence of prematurity and some congenital malformations, are largely beyond our present ability to prevent. Indicators of potentially dangerous morbidity may vary between two institutions primarily because of the availability of tests to mea3ure those conditions rather than because of varying quality of care or varying incidence. For example, hypoglycemia and hypocalcemia in the neonate may be related to the quality of care insofar as they could have been prevented or detected and properly treated. However, their incidence may also reflect a more active approach to neonatal man- agement or the more ready availability and use of laboratory investiga- tions. Because of differences in access to diagnostic procedures, care must be taken to evaluate whether reported selected conditions are truly different in incidence or only reflect laboratory testing. Because interest has shifted to the effects of maternal and peri- natal care on psychosocial parameters, efforts are needed to develop good outcome measures. Some short- and long-term possible topics for study include parent satisfaction with the birth experience, the quality of bonding established between parents and infantsJ â¢parentingâ¢ abilityJ and the emotional, intellectual, and physical development of the infant. It is not known at present whether any relation exists between maternity care and these or other similar outcomes. Hypotheses in regard to such associations need to be developed and tested, and appropriate measure- ments have to be developed. If such relationships are demonstrated, these outcomes could be used as complements to morbidity for evaluating the quality of maternity care.
58 TIME AS A VARIABLE In research involving pregnancy and childbirth, the recognition of time as a variable deserves special attention. Variables such as risk as- signment, presence of anxiety, or high blood pressure may change over the course of pregnancy and childbirth. Gestational age can act as a potential confounder to the number of prenatal visits. Study objectives reflecting an awareness of these temporal factors may require long-term follow-up to assess outcomes. VARIABLES OF PLACE 1 PRACTICE 1 PROVIDER, AND RECIPIENT Variables can be categorized according to the topics or concepts ad- dressed as part of the research. One grouping pertinent to childbirth settings consists of four categoriesa place, practice, provider, and recipient of care. Place of birth variables describe the building, sur- roundings, abDosphere, equipment, and supplies that make up the environ- ment where birth occurs. Provider variables describe physical, psycho- logical, professional or technical training, and social aspects of the persons who give care to the childbearing family. Practice variables describe the organization, policy, and activities occurring in the set- ting. Examples of practices would include whether episiotomies are performed, whether fathers are allowed in the delivery roam, the client's length of stay, and the extensiveness of childbirth education. Variables for recipients of care could include aspects of the biology, demography, or psychology of the study group. 'l'he fourfold categoriza- tion of variables here is meant only to be illustrative, not exhaustive. SELECTION OF PLACE VARIABLES 'l'he variables chosen to describe the place of birth follow from the study objectives or hypotheses, the design of the study, and its loca- tion. 'l'he physical surroundings and abDosphere of the birth site can affect how individuals or groups react to their experiences. Variables might include size of building and roams, interior design, availability of parking, the client's perception of the atmosphere, cleanliness, and staff behavior. The presence or absence of the equipment and supplies used in child- birth could be recorded and quantified. These facts usually determine the complexity of cases or the emergencies that can be handled at the facility. For childbearing families, this information can serve to alleviate or to produce anxiety. For researchers, information on the amount and type of equipment is most useful when there are accompanying data pertaining to practice. The geographic location of the birth place in the community can pro- vide information about distance from backup facilities, residential areas, and neighborhood ambiance. Variability in access to birth loca- tions between and within studied groups might suggest explanations for subtle differences in outcomes. -~-
59 Some provision should be made to measure factors known to affect use of services, e.g., distance to service, transportation availability, and fee scales, because these factors affect behavior toward a service. Every service appeals to a particular clientele, and these preferences should be documented. Use of a certain facility may be limited to indi- viduals with special characteristics (such as high-risk or low-risk mothers). Such documentation by the investigator is important because it greatly influences the ability to make valid comparisons. SELECTION OF PROVIDER VARIABLES The physical, psychological, social, and behavioral aspects of providers are important factors in the interaction between the service and the client (Danziger, 1978). Provider training, skills, and experience affect both the care of the clients and the basic philosophy with which providers approach clients. For example, information on provider gen- der, ethnic group, social class, level of support to the mother, and disposition may be as important to collect and analyze as information on number of years of provider training and on number of deliveries attended. Selection of variables will depend on the types of providers chosen for study. SELECTION OF PRACTICE VARIABLES Choice of childbirth practice variables to be included will depend on the study, the place, and the provider. Generally, information on such activities and on aspects of care given should be collected. It is possible that some of these activities will be manipulated as part of the study design, but other variables may impinge on the manipulated activities and should be described as well. The policies and organization of care can be determinants of the population that chooses to come to a place of birth. Differences and similarities in such matters as admission and discharge policies, organized referral and transport facilities, or hours of work for staff should be documented. The process of care may be related to characteristics of the study population as well as to outcomes. Specific aspects of maternity care that are likely to influence outcome regardless of the setting are especially useful measures. Such practices might include electronic fetal monitoring for low-risk women as compared with high-risk women, routine cesarean section following a previous birth by cesarean section, use of anesthesia or analgesics, routine delivery of a breech by cesar- ean section, and the routine use of episiotomy. Newborn practices might also be of interest: extent of the encouragement of breastfeeding, rooming in, duration of postdelivery stay, parental contact after birth, and bathing procedures and other anti-infective measures such as use and timing of silver nitrate eye drops. Procedures of care found to be beneficial to women and their infants might eventually be implemented in all settings and those found harmful eliminated.
60 SELECTION OF POPULATION VARIABLES Three categories of population variables are used in this discussion: sociodemographic, biological, and psychological variables. Besides serving as variables, they can help define population characteristics crucial to selection of study groups. Selection of Sociodemographic Measures Social and demographic factors such as maternal age, race, income level, or education have a pervasive influence in pregnancy and childbirth. Their effect is twofold: independent effect on outcomes (irrespective of birthplace) and effect on choice of birth setting and therefore indirect effect on outcome. They can also have important moderating effects on other variables of more direct interest to the investigator. For sociological and anthropological considerations of childbirth, see, for example, De Vries (1981), Jordan (1978), and Macintyr (1977). Selection of Biological Variables and Sample Size Age, infant gender, and obstetrical history are biological factors frequently selected as influential or independent variables. Mortality has been the major biological variable studied as an outcome in research on birth settings. Morbidity and birth weight should also be considered as outcome variables for study. Selection of variables is affected by the available population size and by the sample size required to use the variable reliably. Mortality The 1980 infant mortality rate in the United States was approximately 12.5/1,000 live births, and the maternal mortality rate was approximately 6.9/100,000 live births (National Center for Health Statistics, 1981). The rates indicate a low incidence of mortality in the population, and there is an even lower incidence in settings that select low-risk patients. Therefore, the size of study groups has to be extremely large to use mortality data alone as an outcome. Antepartum deaths are not influenced by the place of delivery, al- though prenatal care may have some effect. Stillbirth rates are, for the most part, composed of deaths before labor. Unfortunately, little is known about the epidemiology of intrapartum fetal deaths because few areas record it as separate from fetal deaths in general. Thus, overall still- birth rates are poor indicators of obstetrical quality during delivery, but might be a useful measure of prenatal care. Late intrapartum death rates show promise as being useful indicators of obstetric care. Some components of the perinatal death rate are only slightly influ- enced by medical care. These include many deaths due to congenital anomalies, deaths in infants whose birth weights are less than 750 grams, and deaths in the first few months of life due to Sudden Infant Death Syndrome.
61 Finally, the principal associations among perinatal mortality, birth weight, social class, age of mother, and parity are not well understood. Birth weight, which is the most important known determi- nant in perinatal outcome, is seemingly resistant to medical interven- tion (Chalmers and Adelstein, 1981J Paul et al., 1979J Sinclair et al., 1981J Stewart et al., 1981). Although the neonatal mortality (death to the infant in the first 28 days after birth) rate is low (even when congenital anomalies are included), there are certain advantages associated with using neonatal mortality as an outcome measure. For instance, neonatal mortality is a finite event with an existing system in place to record its occurrence. For neonatal data to be meaningfully evaluated, the following types of variables should be studied: a) birth weight and gestational age, b) age at death in minutes, hours, or days, and c) diagnoses of congenital anomalies and other conditions identified as to prepartum, peripartum, or postpartum etiology. Measures associated with maternal mortality are even more prob- lematic than those of neonatal mortality. There is difficulty in obtaining a sufficiently large study sample because of the low maternal death rate. Furthermore, although there is an extensive literature that attempts to identify maternal factors that increase the risk of death or damage to the fetus, there are few reports that identify risks of delivery to the mother. Morbidity In discussing measurements of morbidity, it should be noted that there are certain disadvantages associated with biological measures of neonatal morbidity, extrauterine adjustment, and other physiological factors, especially when these measures are used alone. First, the specificity of diagnostic criteria may be poor. Second, it is difficult to isolate these outcomes from their interactions with other processes or events. Finally, there has been a notable lack of systematic basic research on most of these outcome measures, with Apgar scores and some diseases constituting possible exceptions. Nevertheless, many factors available for study may have linkages with well-studied and well-recorded factors such as birth weight and gestational age. Most of these outcome measures do not require a pro- fessional observer. Some useful measures could include Apgar scores and the presence of some abnormality. Other factors to be recorded might include extrauterine adjustment and physiological processes such as body temperature, time and details of the first feeding, weight (including time to regain birth weight), neurobehavioral status (Brazelton 1973), and laboratory data such as bilirubin level and bacterial colonization, e.g., of the intestinal tract. Although there have been few attempts to identify factors that might predict poor maternal outcome, certain medical or obstetrical mishaps can be termed â¢poor maternal outcome.â¢ TO underscore the importance of comprehensive demographic data, some of these events occur with reduced frequency in certain groups of women. Maternal biologicai variables could include the following:
62 â¢ Use of oxytocin â¢ Use of analgesia and anesthesia â¢ Use of forceps â¢ Failure to progress in labor â¢ Fetal presentation â¢ Need for cesarean section â¢ Episiotomy â¢ Hypertension in labor â¢ Uterine dysfunction â¢ Retained placenta â¢ Laceration â¢ Blood loss â¢ Infection, such as mastitis, cystitis, pylonephritis â¢ Amnionitis â¢ Endometritis â¢ Thrombophlebitis Selection of Psychological Variables A wide range of psychological hypotheses can be studied in research on childbirth settings. Methods that are productive for obtaining infor- mation on psychological processes can be combined with methods used for obtaining information on physiological processes (Trause et al., 1981). The opportunities for research on psychological aspects of child- birth settings are numerous (Chalmers, 1982). Appendix D describes some of the opportunities at length. In particular Table 1 of that appendix indicates the many areas for which no information exists on psychologi- cal aspects of family membersâ¢ experiences related to childbearing. SOUBCES OF DATA FOR STUDY OF VARIABLES AND OU'l'COMES Vital statistics, medical records, and large-scale surveys are sources of data useful for analysis of events that occur infrequently in the population. Some modifications of vital and medical records would enhance our ability to answer questions about childbirth settings. Retrospective studies would be greatly facilitated by several changes in vital records. At this time it is impossible to link birth and death certificates on a nationwide basis, though this is done routinely in many states (Burnett et al., 19807 Rindfuss et al., 19781 Williams, 19797 Williams and Chen 19827 Williams and Hawes, 1979). Such linking of records would be very useful for research on birth settings (Fedrick and Yudkin, 1976). Information on the actual place of birth, the attendant actually managing the birth, and whether the birth was planned to occur at that location must be added before we can determine the numbers of births taking place out of the hospital and who manages the birth. In the meantime the interpretations of results derived from vital records must be made carefully. Information is frequently abstracted from medical records (Chng et al., 19807 Ballet al., 19807 McNay et al., 1977). Obstetric-medical
63 history may be used to assess a woman's likelihood of experiencing adverse problems during pregnancy or the peripartum period. However, medical records are typically designed not for research but rather to facilitate diagnosis and treatment. These records are not standard- ized as are interviews or questionnaires. A notable exception is the Bobel record system (Bobel, 1976r SOkol et al., 1977), which serves as the clinical document and as a research document. Researchers employ- ing medical records will have to decide how reliability and validity of information contained in the record is to be assessed (see Dambrosia and Ellenberg, 1980r Institute of Medicine, 1977, 1980). Investigators can choose to use interviews, observation, physiologi- cal indices of behavior, archival records, or some combination of these. Some questions, such as women's attitudes toward pregnancy or expecta- tions about delivery, can be answered only by self-reports in question- naires or interviews. In other cases several different procedures of data collection may be feasible. FOr example, an investigator inter- ested in drug use during labor will have to decide whether to rely on observation, interview data, medical records, or some combination of these. In selecting a particular data collection procedure, the inves- tigator should be able to explain the advantages and disadvantages of the alternatives and should provide a rationale for the procedure selected. The use of existing data sets may limit the investigator's choice of variables for study. The Bobel record, for example, contains few psychosocial indicators. Thus, data to answer questions about psycho- social events may have to come from new research studies. However, the increasing levels of multidisciplinary collaboration among biomedical, behavioral, and social scientists offer promise that the existing obstacles to producing a scientific literature on childbirth settings can be overcome. REFERENCES Apgar, v. A. 1953. A proposal for a new method of evaluation of the newborn infant. Current Researches in Anesthesia and Analgesia 32a260-267. Brazelton, T. B. 1973. Neonatal Behavioral Assessment Scale. Philadelphiaa Lippincott. Burnett, c. A., J. A. Jones, J. Rooks, c. B. Chen, c. w. Tyler, and c. A. Miller. 1980. Home delivery and neonatal mortality in North Carolina. Journal of the American Medical Association 244a2741-2745. campbell, D. J., and D. w. Fiske. 1959. Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin 56a81-105. Chalmers, B. 1982. Psychological aspects of pregnancya Some thoughts for the eighties. Social Science and Medicine 16a323-331. Chalmers, I., and A. M. Adelstein. 1981. Improving the quality of perinatal statistics. Lancet 2a640.
64 Chng, P. K., M. H. Hall, and I. MacGillivray. 1980. An audit of antenatal care: ~e value of the first antenatal visit. British Medical Journal 281:1184,1186. Dambrosia, J. M., and J. B. Ellenberg. 1980. Statistical considerations for a medical data base. Biometrics 36:323-332. Danziger, s. K. 1978. The uses of expertise in doctor-patient encounters during pregnancy. Social Science and Medicine 12:359-367. DeVries, R. G. 1981. Birth and the death: Social construction at the roles of existence. Social Forces 59:1074-1093. Duncan, R. c., R. G. Knapp, and M. c. Miller. 1977. Introductory Bio- statistics for the Health Sciences. New York: John Wiley & Sons. England, J. M. 1975. Medical research: A Statistical and Epidemio- logical Approach. Edinburgha Churchill Livingstone. Fedrick, J., and P. Yudkin. 1976. Obstetric practice in the Oxford Record Linkage Study area. British Medical Journal la738-740. Fullerton, J. D. T. 1981. The choice of in- or out-of-hospital birth environment as related to selected issues of control. Ph.D. thesis. Temple University. Hall, M. H., P. K. Chng, P.K., and I. MacGillivray. 1980. Is routine antenatal care worthwhile? Lancet 2:78-80. Robel, c. J. 1976. Recognition of the high-risk pregnancy woman. ~ Management of the high risk pregnancy, w. N. Spellacy, ed. Baltimore: University Park Press. Institute of Medicine. 1977. Reliability of Hospital Discharge Abstracts. Washington, D.C.: National Academy of Sciences. Institute of Medicine. 1980. Reliability of National Hospital Discharge Survey Data. Washington, D.C.: National Academy Press. Jordan, B. 1978. Birth in Four CUltures: A Cross-Cultural Investigation of Childbirth in Yucatan, Bolland, Sweden, and the u.s. st. Alban's, Vt.: Eden's Press Women's Publications. Kerlinger, F. N. 1973. Foundations of Behavioral Research, 2nd edition. New York: Holt, Rinehart, and Winston. Lancaster, H. o. 1974. An Introduction to Medical Statistics. New York: John Wiley & Sons. Macintyr, s. 1977. The management of childbirth: A review of socio- logical research issues. Social Science and Medicine 11:477-484. McNay, M. B., G. M. Mcilvaine, P. w. Bowie, and M. c. MacNaughton. 1977. Perinatal deathsa Analysis by clinical cause of assess value of induction of labor. British Medical Journal la347-350. Miller, s. L. 1981. Introductory Statistics for Dentistry and Medicine. Reston, Virginia: Reston Publishing Company. National Center for Health Statistics. 1981. Annual summary of births, deaths, marriages, and divorcesa United States, 1980. Monthly Vital Statistics Report 29(13). Paul, R. H., K. s. Koh, and A. H. Monfared. 1979. Obstetric factors influencing outcome in infants weighing from 1,000 to 1,500 grams. American Journal of Obstetrics and Gynecology 133:503-508. Rindfuss, R. R., s. L. Gortmaker, and J. L. Ladinsky. 1978. Elective induction and stimulation of labor and the health of the infant. American Journal of Public Health 68:872-877.
65 Rosenberg, M. 1968. The Logic of Survey Analysis. New York1 Basic Books. Selltiz, c., L. s. Wrightsman, and s. w. Cook. 1976. Research Methods in Social Relations (3rd ed.). New Yorka Holt, Rinehart, and Winston. Sinclair, J. c., G. w. Lorrance, M. Boyle, s. P. Horwood, s. Saigal. 1981. Evaluating neonatal intensive care. Lancet 2al052. Sokol, R. J., M.G. Rosen, J. Stojkow, and L. Chik. 1977. Clinical application of high-risk scoring on an obstetric service. American Journal of Obstetrics and Gynecology 128:652-661. Stewart, A. L., E. o. R. Reynolds, and A. P. Lipscomb. 1981. Outcome for infants of very low birthweight: Survey of world literature. Lancet 1:1038-1041. Trause, M. A., D. Voos, c. Rudd, M. Klaus, J. Kennel, and M. Boslett. 1981. Separation for childbirth& The effects on the sibling. Child Psychiatry and Human Development 12:32-39. Williams, R. L. 1979. Measuring the effectiveness of perinatal medical care. Medical care 17:95-110. Williams, R. L., and P. M. Chen. 1982. Identifying the sources of the recent decline in perinatal mortality rates in California. New England Journal of Medicine 3061207-214. Williams, R. L., and w. E. Hawes. 1979. Cesarean section, fetal monitoring, and perinatal mortality in California. American Journal of Public Health 691864-870. Yang, R. K. 1981. Maternal attitudes during pregnancy and medication during labor and delivery: Methodological considerations. ~ Newborns and Parents, v. L. Smeriglio, ed. Hillsdale, N.J.a Lawrence Erlbaum.