Research Issues in the Assessment of Birth Settings: Report of a Study (1982)

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.

APPENDIX F Vital Statistics and Nonhospital Births: A Mortality Study of Infants Born Out of Hospitals in Oregon Nancy Clarke in consultation with Anita B. Bennetts Delivering infants in settings other than hospitals became increasingly common in Oregon during the late 1970s. In 1974, only 1.5 percent of all births took place in freestanding clinics, doctors• offices, homes, and other nonhospital addresses, compared with 3.9 percent by 1979 (Oregon Center for Health Statistics, 1981, and unpublished data, 1981). The number of births not attended by a physician also increased. In 1974, 1.2 percent of all Oregon births were attended by a lay midwife, certified nurse midwife, chiropractor, naturopath, relative, friend, or other person, compared with, 2.4 percent in 1979 (Oregon Center for Health Statistics, 1981, and unpublished data, 1981). There has been a siailar though less pronounced trend for the United States as a whole. The proportion of births attended by midwives increased fro. 1.2 percent to 1.6 percent between 1977 and 1979 (National Center for Health Statis- tics, 1981). These changes in birth sites and delivery attendants have stimulated an interest in the safety of nonhospital births. It is extremely difficult to assess the relative safety of births occurring in various settings with different providers. Definitive assessments cannot be made until results have been obtained from pro- spective studies that can control for maternal risks, demographic and social characteristics, and intended delivery sites, and that can assess outcoaes in terms of .orbidity for both mothers and infants. An impor- tant preliminary step in designing these studies is to review existing data on births and subsequent deaths for infants born elsewhere than in a hospital. Mortality rates provide only crude indicators for measur- ing birth outcomes, and retrospective studies using data collected for entirely different purposes introduce many measurement problems. Con- cluding that a causal relationship exists when mortality rates vary between subgroups is inappropriate. Nevertheless, vital statistics provide a relatively inexpensive means for generating hypotheses about providers, sites, and populations for further study. Also, the ca.- prehensive coverage of vital data plays a crucial role in emphasizing the diversity of the providers and sites that must be included in the description of nonhospital births. Finally, vital statistics can be used to identify populations with excessive mortality, thereby serving as an important tool for those interested in promoting public health. 171 

U2 Shy et al. (1980) used vital statistics to exaaine differences in infant mortality outcomes by site. They found that the infant mor- tality rates for freestanding birth center deliveries were lower than those for all Washington State residents and that home delivery mortal- ity rates were higher than state resident figures. The authors cau- tioned that biases are built into such comparisons because low-risk pregnancies should have lower mortality rates. They reca.mended that prospective studies should be based on the mothers' intention to have a nonhospital delivery. In this paper, vital records are used to examine the providers of maternity care. The following pages describe the variation in neonatal and infant .ortality for births occurring in all nonhospital settings by the category of attendant indicated on birth certificates. There also follows a discussion of the context for interpreting this variation. NOHBOSPITAL BIRTHS IN OREGON The nonhospital birth experiences vary by state. Regulations concern- ing the births and who aay attend them are not the same, and the popu- lations choosing a nonhospital setting may differ. A review of Ding- ley's published data concerning Oregon's nonhospital births (Dingley, 1977, 1979) is relevant for an understanding of the mortality rates presented in this paper. Since 1977, approxiaately 4 percent of the live births in Oregon have occurred in a setting other than a hospital. Oregon law prohibits ~ay persons from performing episiotomies and administering medications, but no other limitations concerning birth attendants exist. Dingley's descriptions of Oregon nonhospital births for 1976 and 1977 indicated that the parents tended to be better educated than the parents selecting hospital births. There were relatively fewer teenage .others, fewer first births, and fewer immature and low birth weight babies than for all births to Oregon residents. This suggested a de.ograpbic profile that could favor the delivery of healthy infants. When nonhospital births were categorized by the type of attendant, however, sa.e indicators of low-risk pregnancies did not apply to all categories. For births attended by fathers, mothers, other relatives, friends, helpers, Followers of Christ, and other attendants, excluding licensed professionals and midwives, a high percentage of .others had achieved less than a twelfth-grade education and had received no pre- natal care. Dingley expressed concern that approxtaately 15 percent of all the mothers who delivered out of hospital had a history of previous fetal deaths, a high-risk indicator. Her findings concerning .ortality outcomes for 1976 were inconsistent with those of 1977. These studies noted two subcultures with a large nuaber of nonbos- pital deliveriess a caa.unity of Old Believers that had . .igrated from Russia, and a religious ooaaunity called Followers of Christ. These two groups constituted approxiaately one-fourth of the deliveries in the •other and No Attendant• category. 

173 MBTIIOD Tbe 1975-1979 birth and death recorda for all infants born out of hOs- pital in Oregon were exDined. Those classified as •born en route• were excluded because they were assu.ed to have been intended hospital deliveries. Certainly other births were meant to have occurred in a hospital but have been included in this analysis because they cannot be identified. Tbe delivery attendants for all infants weighing 1,500 graas or less were contacted to assure that death reporting vas ca.plete. State and county staffs have done extensive field work in an atte.pt to ensure as close to 100 percent coverage of nonhospital births as possible. In this study, nonhospital births have been defined as all deliveries that occurred in locations other than a hospital. The birth attendants indicated on the certificates were classified according to the following categoriesa Other and Licensed Attendant Midwife No Attendant Medical doctors Certified nurse Relatives Osteopaths midwives Friends Naturopath& Lay aidwives who Helpers Chiropractors identify thea- Followers of Christ Registered nurses selves as such Old Believers a.ergency -.dical Unknown attendants personnel No attendant These categories are certainly not ideal because there are significant differences in training and orientation, for exa.ple, between a naturo- path and a pbysician or between a lay aidwife and a certified nurse •idwife (CMM). However, coding practices in 1975 and 1978 did not aake finer distinctions.' The necessity of combining five years of data in order to provide large enough nuabers for statistical reliability prohibits the use of aore refined categories. A birth was attributed to a lay aidwife if the attendant si.ply identified herself as such or if her name appeared on a sufficient nuaber of certificates for the birth certificate coders to recognize her name and classify her as a lay aidwife. In recent years nearly all lay midwives have been identified. However, as Dingley pointed out, there are aany aidwives, particularly those in traditional and religious camaunities, who do not sign certificates, preferring to have the 1 Data from 1976, 1977, and 1979 indicate that the aidwife category consists of 57 percent lay midwives and 43 percent eRMa. The pro- portion of births attended by lay midwives has been increasing. In fact, 66 percent of the 1979 births in the aidwife category were delivered by lay aidwivea. No differences in the death rates for infants delivered by the two types of aidwives are apparent in the three years for which aore detailed inforaation is available. 

174 TABLE 1 Selected Oregon and u.s. Honhospital Births and Subsequent Death Rates (per thousand), 1970-1979 Nonhospital Nonhospital Births Occurring Births Occurring All Births to in Ore«Jon in Ore«Jon Oregon Residents All u.s. Births Cate«Jory 1970-1974 1975-1979 1975-1979 1975-1979!. Births 2,224 6,398 186,187 16,444,897 Neonatal deaths 58 48 1,475 165,696 Infant deaths 81 89 2,351 236,710 Neonatal 26.1 7.5 7.9 10.1 death rate Infant death 36.4 13.9 12.6 14.4 rate !Includes provisional estimates for 1979. father or another relative sign as attendant (Dingley, 1977, 1979). These cannot be identified and have been classified in the •other and No Attendant• category. This analysis consists of cross-tabulations of the rates, maternal characteristics, and causes of death by attendant. caution is war- ranted when interpreting these results because of the small sample used, the low probability of neonatal and infant deaths, and the inaccuracies in recording and lack of refinement in categorizing data on type of attendant. RESULTS Infant and Neonatal Death Rates for All Nonhospital Births Table 1 compares the figures for nonhospital births occurring in Oregon during the past decade (1970-1974 and 1975-1979) to the 1975-1979 rates for all Oregon residents and to the u.s. rates for the same period. Although the number of nonhospital births has nearly tripled, the number of neonatal deaths (deaths in the first 28 days) and infant deaths (deaths in the first year) has stayed approximately the same. As a result, the rates for both infant and neonatal deaths are con- siderably less than half of what they were earlier in the decade. This dramatic drop in the rates indicates that the year for which data was collected aust be considered when evaluating and comparing research findings. For the last five years, there have been only small differences between the death rates for nonhospital births and those for all residents of Oregon (see Figure 1). The infant death rate for non- hospital births is approximately one point higher (13.9 compared with 12.6). The neonatal death rate, which is a better indicator of problems associated with pregnancy and delivery, is slightly lower (7.5 compared with 7.9). Because of the small numbers involved, these differences are not statistically significant (p <.OS). Both rates are lower than the u.s. figures for all births. 

175 NEONATAL ~Licenaed (I) 26 Ell CNMs end Lay Midwives (I) 26 % % t- ~ Other end No Att•ndllnt t- a: a: CD 20 111111 All S~• R•idlntl iii 20 ~ ~ a: w a.. 16 - a: w a.. 16 (I) (I) :J: :J: t- 10 t- 10 < w < w c c 6 6 0 0 L- Out of Hospital ___. All L- Out of Hospital ___. All Residents Residents FIGURE 1 Neonatal and infant death rates for Oregon residents and for nonhospital births by attendant, 1975-1979. Infant and Neonatal Death Rates, by Attendant Marked differences in the reported figures by attendant are apparent in Table 2. Although the •other and Ro Attendant• category accounted for only 28 percent of the nonhospital deliveries, this group contributed to .ore than one-half of the neonatal deaths and nearly the same frac- tion of the infant deaths. If the extreaely small infants are eliai- nated and only the infants who weighed more than 2,500 grams at birth are considered, the differences become even more pronounced. The neonatal death rate of 13.9 for the •other and No Attendant• deliveries is more than four times higher than the rate for lay aidwives and certified nurse midwives and twice the rate for the licensed aedical professionals. These differences are statistically significant (p < .OS). Although the midwife rate is one-half the licensed rate, this difference is not significant. Differences in the infant death rates show a pattern similar to that of the neonatal ratesJ the •other and Ro Attendant• infant death rate is nearly twice the rate for the licensed attendants. DISCUSSION Extreme caution aust be exercised when interpreting these rates. With vital records, a difference in mortality rates among settings may have little to do with the safety of a planned delivery in a nonhospital setting or with any particular attendant. The limitations associated 

176 TABLE 2 Nonhospital Births in Oregon and Subsequent Deaths by Type of Attendant, Birth Weight, and Age at Death, 1975-1979 Licensed Other and Ro Total Catec)ory Attendants Midwives!. Attendants Nonhoapital Births 3,006 (471) 1,597 (25,) 1,795 (281) 6,398 (100,) Neonatal deathsa 18 (371) 5 (10') 25 (52') 48 (100,) Unknown weiCJht 0 0 2 2 2,500 9raaa 8 1 7 16 2,500 9raaa 10 (33') 5 (131) 16 (53') 30 (100,) Infant death~a 38 (43,) 8 ( 91) 43 (48,) 89 (100') Unknown weiCJht 0 0 2 2 2,500 9raas 12 1 7 20 2,500 9raaa 26 (391) 7 (10') 34 (511) 67 (100') Neonatal death rate£a All deaths 6.0 3.1 13.9 7.5 Infant death rate£a All deaths 12.6 5.0 24.0 13.9 !Includes CHMa and self-identified lay midwives. ~ine infanta delivered by nonprofessional attendants had an unknown birthweiCJht. The seven who died durinCJ the poatneonatal period have been added to the >2,500 catec)ory because all of th• lived 110re than one 110nth, none had any indication of beinCJ pr...ture, and all died of causes not related to pr...turity, preC)nancy, delivery, or perinatal conditions. 2.lter thousand. with interpreting these rates are discussed in ter.a of ..asureaent biases, risk factors, and causes of death--factors that may result in a misplaced emphasis on the attendants rather than on the populations served by those attendants. Reporting Bias The collection and coding of information from vital records introduces a number of biases in studies concerned with evaluating outco.es associated with nonhospital births. Pirat, mortality ia only a crude indicator of unsatisfactory pregnancy outca.ea, and large populations are required to produce statistically significant ratea. Pive years of birth and death data for Oregon do not provide a sufficient population for adequately detailed comparisons. Second, the categories of attend- ants may be misleading. Por example, although the neonatal death rate for all licensed attendants combined is 6.0 per 1,000 births, the rate for naturopath& and physicians may be vastly different. Furthermore, because the tera lay midwife bas no official definition in Oregon tbia category undoubtedly includes people with quite different akilla and practices. A third factor to be considered ia the possibility of incomplete registration. Although there are no means for aaaeaaing 

177 underreporting of deaths, the severity of the consequences for failure to report a death suggests that this is probably a rare event. However, nonhospital births are known to be underreported because these births are soaeti..s registered after the children.are more than one year old. This difference could result in exaggerated mortality rates for one group when compared with another. A last set of i~rtant considerations concerns the differences between the reported and the intended site and provider. A midwife with appropriate physician and/or hospital backup may consult a phy- sician as well as transfer a patient to a hospital in the event that complications arise during labor and delivery. · But the birth record contains only information on the final site and provider. Therefore, the vital records identify many complicated deliveries as hospital- based or physician-attended when the birth was actually planned to be nonhospital with a aidwife attendant. A bias in the opposite direction also exists because some of the nonhospital births may be deliveries for mothers who were unable to obtain medical assistance quickly when labor began. The possibility also exists that a father or other rela- tive is asked to sign as attendant when a delivery by any attendant goes awry. Bias Due to Variation in Risk Status The reporting biases in the mortality rates presented in this paper are •inor compared to the biases introduced in these rates by variations a.ong the populations. The medical, social, and demographic risks pre- sented to the different categories of attendants undoubtedly account for much of the variation in the rates. Birth certificates contain only limited information about maternal risks, and data are only available for 1976, 1977, and 1979. Nevertheless, the differences in the charac- teristics of mothers in the three attendant categories emphasize the necessity of considering these variables. Data concerning parental educational attainment and mothers' prenatal care show a pronounced disadvantage for mothers in the •ather and No Attendant• category. An examination of prenatal-care history makes it apparent that the births in the third attendant category do not represent the ideal of well-screened mothers anticipating normal deliveries (Figure 2). In the three years for which information is available, more than one-fourth (28 percent) of the mothers without a licensed attendant or a midwife bad received no prenatal care. For the same period, the figure is less than 1 percent for the mothers in the licensed attendant and midwife categories. The mother's lack of education is another characteristic shown to be associated with higher infant and neonatal mortality. Of those who answered the education question on the certificates during the three- year period, slightly more than one-third of all mothers who delivered out of hospital reported exactly 12 years of education. However, 29 percent of the mothers who delivered without a licensed attendant or a •idvife had less than a high-school education, coapared with 13 percent 

178 30 ~Licensed 25 111111 CNMs and Lay Midwives en ~ Other and No Att8ndant I= ~ 20 ID IL. 0 15 1- z w u ~ w 10 1:1.. 5 + + L- No Prenatal Care ....J Mother< 12 Years Education PIGURE 2 Percent of births in aaternal risk category for births occurring out of hospital, by type of attendant, 1976, 1977, 1972. of those with a licensed attendant and 9 percent of those with a aid- wife. Although more than one-half (53 percent) of the mothers attended by licensed professionals and midwives had some college education, only 36 percent of the •other and No Attendant• category reported this level of education. Pathers showed patterns of educational attainment that were similar to those of the mothers. Other data concerning varying numbers of high-risk .others bY the type of attendant are ambiguous. Although mothers with a reported complication of pregnancy accounted for 2 percent of each nonhospital attendant category (compared with 5 percent for all state resident births), it is doubtful that such problema as anemia and Rh incompati- bility, which are common for medically attended births, would be diag- nosed for the large number of mothers with no prenatal care and without a licensed attendant or a midwife. The risk factors concerning mater- nal age and pregnancy history did not vary by type of attendant. The large differences in the educational attainment and prenatal care of the populations served by the three attendant categories used in this analysis are an indication that other medical, social, and demographic characteristics must vary as well. Causes of Death The above discussion of maternal risks demonstrates a need to consider not only the providers of care but also the populations served. An 

179 T~ 3 Ronhoapital Births in Oregon and Subsequent Deaths of Infanta with Birth Weights Higher than 2,500 Graaa, by Type of Attendant, Age, and Cause of Death, 1975-1979 Licensed Other and Ho Total catec:Jory Attendanta Midwives!. Attendants Honhospital Births 3,006 1,597 1,795 6,398 Deaths by cause Pregnancy, delivery and per ina tal conditions Neonatal 6 1 8 15 Infant 7 1 8 16 Congenital anoaalies Neonatal 3 1 2 6 Infant 7 1 5 13 other causesa.2 Neonatal 1 2 6 9 Infant 12 5 21 38 Sudden Infant Death Syndraae g£ ~ 1~ 22 External causes (IIOtor vehicle, drowning, acci- dent, assault, undetermined) 1 s!t 6 PneUIIOI\ia and upper respira- tory tract infection 3 3 Meningitis 1 1 2 Malignant neoplasas 2 2 Septicemia 1 1 Skin infection 1£. 1 Intestinal obstruction 1 1 Infant death rate (per thousand) for other causes 4.0 3.1 11.7 5.9 Aincludes CMMs and lay midwives. 2rroa categories 000-739 and 780-999 in World Health Organization, 1977. £Includes one neonatal death. ~Includes two neonatal deaths. ~Includes three neonatal deaths. examination of the causes of death for infanta born out of hospital provides further evidence that addressing the attendants rather than the population served can be misleading. Table 3 and Figure 3 present the causes of death for infants who were born out of hospital and who weighed at least 2,501 graaa at birth. 

180 35 ~ Pregnancy Delivery end Perinatal Conditions 30 1111111111 Congenital Anomalies 25 llllllmllll Other eeu- (SIDS, external ceu... pneumonia, (I) meningitis, malignant neaplans, :I: 1- skin infection, intestinal obltructionl <( w 20 c IL. 0 a: 15 w a:l ~ ~ z 10 5 0 ,, .. c: • c: c: ... :: II II e<( .J:. 0 c5z PIGURE 3 Infant deaths for births occurring out of hospital, by cause and type of attendant, 1975-1979. More than one-half of all the deaths of infanta born out of hos- pital were attributed to causes not directly related to pregnancy, delivery, perinatal conditions, or congenital anoaaliea. These other causes account for a large proportion of the differences in mortality rates between attendant categories. Nearly one-half of the neonatal deaths and almost two-thirds of all infant deaths in the •Other and No Attendant• category were attributed to causes such as Sudden Infant Death Syndrome (SIDS), external causes, and other diseases. This COB- pares with one-sixth of the neonates and approximately one-half of the infanta in the licensed attendant category. This is important because the number of infant deaths atributed to other causes was more than twice the number attributed to pregnancy, delivery, and perinatal con- ditions. The 21 infant deaths due to other causes in the •other and No Attendant• category translates to a cause-specific rate of 11.7 per 1,000 births, which is significantly different fro. the 4.0 per 1,000 births rate for licensed attendants (p <.OS). Differences in the rates due to perinatal conditions and congenital ana.aliea are not significant. 

181 The list of other causes emphasizes the need to consider the entire social and health care environment of the population served by the •Other and No Attendant• providers. In nearly one-half (10) of the cases of other causes for other attendants, the state medical examiner could find no sign of disease and attributed death to SIDS. Another one-fourth (5) of the cases were due to external causes such as automobile accidents, drowning, and homicides. Three infants (not neonates) died of pneumonia and upper respiratory tract infections. Meningitis, septicemia, and skin infection each resulted in one death. CONCLUSIONS State vital statistics can be used as a basis for public health efforts to improve the outcome of nonhoapital births. Such statistics can identify problem areas and suggest hypotheses for further study. Im- pedimenta to using the data include incomplete and missing informa- tion on such ite. . as delivery site and personnel, reporting bias, and variation )n risk factors associated with the population. The data for Oregon indicate that more than one-half of the infant deaths associated with nonhospital births occurred in a population that delivered without the aid of a licensed attendant or a midwife. This s ... group had a poor prenatal-care history, a large proportion of par- ents with leas than a twelfth-grade education, and, probably, a poor medical and demographic risk profile as well. In Oregon, attempts to influence the outcomes of nonhospital births will require much more investigation of those deliveries attended by relatives, friends, helpers, Followers of Christ, Old Believers, unknown attendants, and those with no attendant at all. RBFBRBNCBS Dingley, B. 1977. Birthplace alternatives. Oregon Health Bulletin 55(10):1-4. Dingley, B. 1979. Birthplace and attendants, Oregon's alternative experience, 1977. Women and Health 4:239-253. National Center for Health Statistics. 1981. Advance report of final natality statistics, 1979. Monthly Vital Statistics Report 30(6, Supplement 2). Oregon Center for Health Statistics. 1981. Oregon Vital Statistics. Portland: State Department of Ruman Resourc~s. Shy K. K., F. Frost, and J. Ullom. 1980. Out-of-hospital delivery in Washington State 1975-1977. American Journal of Obstetrics and Gynecology 137:547-552. World Health Organization. 1977. International Classification of Diseases, Ninth Revision. Geneva: World Health Organization. 

• --