Vol.21, No. 2 Printed in Great Britain
International Journal of Epidemiology © International Epidemiological Association 1992
Life Table Analysis of Infant Mortality and Feto-lnfant Mortality Distributed on Causes of Death in Denmark 1983-1987 FINN B0RLUM KRISTENSEN* AND FLEMMING MACj
Perinatal and infant mortality rates are standard measures in vital statistics and epidemiology. One shortcoming of these measures is the compilation, into a single rate, of deaths which occur over a considerable period of time (e.g. the first year of life). Graphic presentation of life tables may remedy the loss of information brought about by the use of standard rates. Many countries now have national or regional birth registers. Where integrated birth registration of all live births, stillbirths and deaths within the first year of life occurs straightforward life table methodology can be applied to data on gestational age and duration of life. Survival of the population which is covered by the register can be expressed not only for liveborn babies but for unborn babies also. If information on underlying cause of death is integrated in the registration it is
possible to study the distribution of the mortality which is recorded during the follow-up period. The following presents some results of life table analyses of data in the Danish Medical Birth Register (MBR) which covers all births to residents in Denmark. We have aimed at a direct graphic presentation of the survival function of the 1983-1987 national birth cohorts with mortality distributed on groups of underlying causes of death. This combined approach was explored to display better the periods of follow-up in which functionally related causes of deaths contributed to perinatal and infant mortality. SUBJECTS AND METHODS The data comprises 263 322 singletons of the 1983-1987 Danish birth cohorts (262 159 liveborn and 1163 stillborn babies). Through record linkage between the national birth registration and the registration of causes of deaths, causes of stillbirths and deaths during the first year of life were routinely recorded in the MBR.1 Gestational age was estimated
•Department of General Practice, University of Copenhagen, Juliane Mariesvej 18, DK 2100 Copenhagen 0, Denmark. tDepartmem of Medical Statistics, National Board of Health, Copenhagen, Denmark.
320
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
Kristensen F B (Department of General Practice, University of Copenhagen, Juliane Mariesvej 18, DK 2100, Copenhagen 0, Denmark) and Mac F. Life table analysis of infant mortality and feto-infant mortality distributed on causes of death in Denmark 1983-1987. International Journal of Epidemiology 1992; 21: 320-323 The survival function of 263322 singletons of the 1983-1987 Danish birth cohorts (262159 liveborn and 1163 stillborn babies) with mortality distributed on functional groups of underlying causes of death is presented in two graphic forms on the basis of Kaplan-Meier estimates. About half of all first-day deaths in liveborn babies occurred during the first 4 hours. More than half of all first-week deaths happened during the first day. More than half of all deaths from 31 weeks to 76 weeks after the first day of the last menstrual period (LMP) were either fetal deaths prior to the onset of labour unexplained by fetal factors or unexplained sudden infant death syndrome (SIDS) deaths. Graphic presentation of feto-infant mortality distributed on functional cause-of-death groups improved the expression of the relative contribution and timing of the different causes of deaths. Despite a high autopsy rate and a uniform coding practice the distribution on causes of deaths from register data should be interpreted with caution. Full use of the feto-infant approach is only achieved with data which include late abortions. Thus the feto-infant approach is of special interest in countries which have registers of live births and fetal deaths from week 22 of gestation or earlier. The integration of the life table approach and analysis of underlying causes of deaths should be further explored as a way of utilizing vital statistic databases for the evaluation of perinatal care.
LIFE TABLE ANALYSIS OF INFANT MORTALITY
TABLE 1 International Collaborative Effort (ICE) functional classification of infant deaths.2 ICE ICE ICE ICE ICE ICE ICE ICE
1. 2. 3. 4. 5. 6. 7. 8.
Congenital anomalies Asphyxia-related conditons Immalurity-related conditions Infections Sudden infant death Deaths due to external causes Specific conditions other than above Remaining causes
All stillbirths and infant deaths were divided into six groups of a clinico-pathological classification of causes of infant deaths and stillbirths which has been developed in Denmark.3 This classification focussed on fetal and neonatal factors.4 It was based on information in the MBR on underlying cause of death, and in some cases also on time of death, gestational age, and time of fetal death in relation to onset of labour. The classification of feto-infant deaths is shown in Table 2. Group FI 2 included deaths which were not explained by fetal factors and deaths which were unexplained from an obstetric point of view. Overall survival as well as survival from each of the competing causes-of-death groups was studied using the product limit (Kaplan-Meier) method.5 In the life table analysis of each competing causes-of-death
TABLE 2 Clinico-pathological classification of underlying causes of feto-infant deaths based on fetal and neonatal factors. FI 1. Congenital anomaly: Deaths that had their origin at conception or during embryogenesis. FI 2. Unexplained death prior to the onset of labour: Deaths due to asphyxia and other unexplained antepartum deaths (e.g. premature rupture of membranes, placental abruption, placenta praevia, placenta! insufficiency and cause unknown). FI 3. lntrapartum events: Deaths before or after birth due to events occurring during labour. FI 4. Conditions consequent upon preterm birth: Deaths due to pulmonary immaturity, hyaline membrane disease, cerebral haemorrhage or infarction in preterm babies. FI 5. Other specific conditions: All other deaths due to intracranial haemorrhage. Deaths due to isoimmunization, or infection, and deaths due to other specific miscellaneous conditions including unequivocal cot death. FI 6. Unclassified deaths: Inadequately documented deaths.
group, babies who died from other causes were censored. The relatively small group of babies who emigrated during the follow-up period could not be censored. In the survival analysis of 262159 liveborn singletons Kaplan-Meier estimates were computed at completion of each hour during the first day, each day during the first week and each subsequent week of follow up. The feto-infant survival analysis was based on 261 722 single fetuses and neonates with known gestational age. In order to eliminate potential bias from underreported stillbirths with low gestational age who had been misclassified as abortions, feto-infant survival analysis was restricted to unborn fetuses and live babies, who had survived 31 completed weeks after LMP in utero or as very preterm live infants. Time of death in relation to LMP in liveborn babies was calculated as the sum of gestational age at birth and duration of life in completed weeks. Consequently survival was expressed in completed weeks after LMP. As live babies are only followed up for 1 year in the birth register, feto-infant follow-up was truncated at 76 weeks after LMP. This was done to avoid bias from loss to follow-up of prematurely born babies with gestational age 24 weeks or more, who had survived 52 weeks after birth. RESULTS Infant Mortality Please recognize that the horizontal scale in Figure 1 changes twice. The first 24 data points after birth (B) indicate 0-23 completed hours of the first day of life, the next 6 points 1-6 completed days of the first week and the rest of the points each completed week of the
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
on the basis of history or ultrasound and was defined as the number of completed weeks after last menstrual period (LMP). Gestational age at fetal death was equal to gestational age at birth. Survival of liveborn babies was expressed in completed hours during the first 3 days of life, in completed days during the rest of the first week and in completed weeks during the following 3 weeks. Date of death was also recorded. Only late fetal deaths (28 completed weeks of gestation, stillbirths) were recorded in the MBR, while live births were recorded irrespective of gestational age. Causes of death from certificates of death were coded using the International Classification of Diseases (ICD) according to WHO guidelines in the Department of Medical Statistics of The National Board of Health by three clerks under medical supervision. Coding practice was checked in the department on a regular basis. More than 80% of stillbirths and 70% of neonatal deaths underwent autopsy during the period. All infant deaths were divided into eight functional groups published by the International Collaborative Effort on Perinatal and Infant Mortality (ICE) on the basis of underlying cause of death.2 The groups are shown in Table 1.
321
322
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
first year of life. Deaths within the first 59 minutes of life survived 0 completed hours. The vertical scale indicates the fraction of survivors. The lowest line in Figure 1 depicts the survival function of the 1983-1987 cohorts of liveborn babies. About half of all first day deaths occurred during the first 4 hours. More than half of all first week deaths happened during the first day.
FIGURE 1 Survival function for liveborn singletons. Mortality divided into eight groups of the ICE functional classification of infant deaths.
The contribution of each cause-of-death group to mortality up to a given point is indicated by the distance between each line and the line immediately above it. When the risk of death from immaturityrelated conditions in Figure 1 decreased substantially after 3 completed weeks, sudden infant death syndrome (SIDS) became the largest risk until 5 months. Only after that did the risk of infant death become very low. Feto-Infant Mortality About half of all perinatal deaths in Denmark are stillbirths. All deaths in the perinatal period and deaths related to perinatal events are important for the evaluation of the health care system be they stillbirths or infant deaths. When follow-up starts at birth no stillbirths are included and the members of the cohort have very different ages of development ranging from about 24 weeks to about 43 weeks after LMP. Consequently, we explored how survival analyses of fetoinfant cohorts on the basis of gestational age could provide insights for the study of perinatal losses which cohort studies of mortality in liveborn babies alone could not provide. In the survival analysis which is presented in Figure 2 stillbirths after 31 completed weeks of gestation were included in the population at risk together with all un-
FIGURE 2 Survival function for fetuses and babies who had survived 31 completed weeks after last menstrual period. Mortality divided into six groups of a clinico-pathological classification of stillbirths and infant deaths based on fetal and infant factors.
DISCUSSION Graphic presentation of feto-infant mortality distributed on functional cause-of-death groups improved the expression of the relative contribution and timing of the different causes of deaths. The results suggest that SIDS contributed just as much to infant mortality as did immaturity-related conditions. Fetal deaths prior to the onset of labour and SIDS deaths are major challenges to contemporary obstetric and paediatric care. The analysis was based on routine data and the allocation of individual cases to the causes-of-death
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
born babies and very immature liveborn infants who were still alive 31 completed weeks after LMP. Nearly all of the babies were unborn at the start of follow-up. After 10-11 completed weeks of follow-up practically all babies had been born. The lowest line in Figure 2 shows the survival function of the 1983-1987 cohorts of single live fetuses and infants from 31 weeks to 76 weeks after LMP. At the end of the 44 weeks period of follow-up more than 99% of the initial population at risk were still alive. During the period in which the babies were born the survival curve was relatively steep because stillbirths and a large proportion of the neonatal deaths occurred during that period. More than half of all deaths from 31 weeks to 76 weeks after LMP were either fetal deaths prior to the onset of labour unexplained by fetal factors or unexplained SIDS deaths. Virtually all deaths due to the consequences of preterm birth occurred during the first 10-12 weeks of follow-up.
323
LIFE TABLE ANALYSIS OF INFANT MORTALITY
The integration of the life table approach and analysis of underlying causes of death should be further explored as a way of utilizing vital statistic databases for the evaluation of perinatal care. Analytical studies using proportional hazard regression for the study of differences between groups expand the life table approach and should be encouraged.12 The feto-infant approach is of special interest in countries which have registers of live births and fetal deaths from week 22 or earlier.
REFERENCES 1
Knudsen L B, Kristensen F B. Monitoring perinatal mortality and perinatal care with a national register. Content and usage of the Danish Medical Birth Register. Comm Med 1986; 8: 29-36. 2 Cole S, Hartford R B, Bergsjp P, McCarthy B. International Collaborative Effort (ICE) on Birth Weight, Plurality, Perinatal, and Infant Mortality III: A method of grouping underlying causes of death to aid international comparisons. Acta Obstel Gynecol Scand 1989; 68: 113-17. 3 Kristensen F B. A short list of causes of death in the perinatal period based on principles published by Hey, Lloyd and Wigglesworth. Appendix In: Sundhedsstyrelsen. Medicinsk fodsels- og misdannelsesstatistik 1986. Vitalstatistik 1988; 1:23 (Copenhagen). 4 Hey E N, Lloyd D J, Wigglesworth J S. Classifying perinatal death. Fetal and neonatal factors. Br J Obstet Gynaecol 1986; 93: 1213-23. 5 Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 457-81. 6 Andersen K V, Helweg-Larsen K, Lange A P. The validity of the perinatal classification by the Danish Health Department. UgeskrLteger 1991; 153: 1575-77. 7 Shapiro S, Jones E W, Densen P M. A life table of pregnancy terminations and correlates of fetal loss. Milbank Mem Fund Q 1962; 40: 7-45. 8 French F E, Bierman J M. Probabilities of fetal mortality. Public Health Rep 1962; 77: 835-47. ' Mellin G W. Fetal life tables. A means of establishing perinatal rates of risk. JAMA 1962; 180: 91-94. 10 Bakketeig L S, Seigel D G, Sternthal P M. A fetal-infant life table based on single births in Norway, 1967-1973. Am J Epidemiol 1978; 107: 216-25. ' ' International Federation of Gynecology and Obstetrics. Report of the Committee following a workshop on Monitoring and Reporting Perinatal Mortality and Morbidity March, 1982. London: The Chameleon Press Ltd. 12 Cox D R. Regression models and life tables (including discussion). J Roy Stat Soc B 1972; 34: 187-220.
(Revised version received September 1991)
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
groups was done with a computer algorithm from ICD codes in the MBR. Despite a high autopsy rate and a uniform coding practice the distribution on causes of death should be interpreted with as much caution as is normally called for when drawing conclusions from routine mortality statistics. A recent study of all 282 perinatal and neonatal deaths in three counties in 1985-1986 evaluated the feto-infant classification in Table 2. Experts classified deaths on the basis of all available record material and compared these results with the algorithm's classification of the same deaths from routine data in the MBR.6 The algorithm was found to be relatively invulnerable to misclassifications at ICD code level as 87% of deaths were correctly classified. Two fetal life table analyses were published in 1962.7-8 Taking the point of view that birth can be looked upon as an event in life, not the beginning of it, Mellin extended the fetal life table concept to gestational week specific life tables for the neonatal period.9 A study of the Norwegian Birth Register 1967-1973 provided information on the probabilities of fetal, neonatal, and postneonatal death for each subsequent week after 16 weeks of gestation.10 The report recommended the feto-infant approach as an extension of vital statistics of value in monitoring health changes and in comparing perinatal mortality between populations. Our study was such an extension, which took causes of death into consideration. Feto-infant follow-up from 31 weeks after LMP rather than from 28 weeks or earlier was chosen in order to avoid bias from potential underreporting to MBR of some stillbirths misclassified as late abortions at around 28 weeks of gestation. Very immature liveborn babies had already died at the beginning of follow-up. Consequently, group FI 4 (consequences of preterm birth) in Figure 2 was relatively small when compared with immaturity-related conditions (ICE 3) in Figure 1. As a consequence of this conservative approach deaths in very prematuare babies and stillbirths before 31 weeks (one-third of all neonatal deaths and one-fifth of all stillbirths) were not depicted in Figure 2. This is a serious limitation to the feto-infant approach to data which do not include late abortions. Preferably registration of fetal deaths and live births should start at 22 weeks or earlier." In Denmark, this would require a change in the birth registration system.
International Journal of Epidemiology © International Epidemiological Association 1992
Life Table Analysis of Infant Mortality and Feto-lnfant Mortality Distributed on Causes of Death in Denmark 1983-1987 FINN B0RLUM KRISTENSEN* AND FLEMMING MACj
Perinatal and infant mortality rates are standard measures in vital statistics and epidemiology. One shortcoming of these measures is the compilation, into a single rate, of deaths which occur over a considerable period of time (e.g. the first year of life). Graphic presentation of life tables may remedy the loss of information brought about by the use of standard rates. Many countries now have national or regional birth registers. Where integrated birth registration of all live births, stillbirths and deaths within the first year of life occurs straightforward life table methodology can be applied to data on gestational age and duration of life. Survival of the population which is covered by the register can be expressed not only for liveborn babies but for unborn babies also. If information on underlying cause of death is integrated in the registration it is
possible to study the distribution of the mortality which is recorded during the follow-up period. The following presents some results of life table analyses of data in the Danish Medical Birth Register (MBR) which covers all births to residents in Denmark. We have aimed at a direct graphic presentation of the survival function of the 1983-1987 national birth cohorts with mortality distributed on groups of underlying causes of death. This combined approach was explored to display better the periods of follow-up in which functionally related causes of deaths contributed to perinatal and infant mortality. SUBJECTS AND METHODS The data comprises 263 322 singletons of the 1983-1987 Danish birth cohorts (262 159 liveborn and 1163 stillborn babies). Through record linkage between the national birth registration and the registration of causes of deaths, causes of stillbirths and deaths during the first year of life were routinely recorded in the MBR.1 Gestational age was estimated
•Department of General Practice, University of Copenhagen, Juliane Mariesvej 18, DK 2100 Copenhagen 0, Denmark. tDepartmem of Medical Statistics, National Board of Health, Copenhagen, Denmark.
320
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
Kristensen F B (Department of General Practice, University of Copenhagen, Juliane Mariesvej 18, DK 2100, Copenhagen 0, Denmark) and Mac F. Life table analysis of infant mortality and feto-infant mortality distributed on causes of death in Denmark 1983-1987. International Journal of Epidemiology 1992; 21: 320-323 The survival function of 263322 singletons of the 1983-1987 Danish birth cohorts (262159 liveborn and 1163 stillborn babies) with mortality distributed on functional groups of underlying causes of death is presented in two graphic forms on the basis of Kaplan-Meier estimates. About half of all first-day deaths in liveborn babies occurred during the first 4 hours. More than half of all first-week deaths happened during the first day. More than half of all deaths from 31 weeks to 76 weeks after the first day of the last menstrual period (LMP) were either fetal deaths prior to the onset of labour unexplained by fetal factors or unexplained sudden infant death syndrome (SIDS) deaths. Graphic presentation of feto-infant mortality distributed on functional cause-of-death groups improved the expression of the relative contribution and timing of the different causes of deaths. Despite a high autopsy rate and a uniform coding practice the distribution on causes of deaths from register data should be interpreted with caution. Full use of the feto-infant approach is only achieved with data which include late abortions. Thus the feto-infant approach is of special interest in countries which have registers of live births and fetal deaths from week 22 of gestation or earlier. The integration of the life table approach and analysis of underlying causes of deaths should be further explored as a way of utilizing vital statistic databases for the evaluation of perinatal care.
LIFE TABLE ANALYSIS OF INFANT MORTALITY
TABLE 1 International Collaborative Effort (ICE) functional classification of infant deaths.2 ICE ICE ICE ICE ICE ICE ICE ICE
1. 2. 3. 4. 5. 6. 7. 8.
Congenital anomalies Asphyxia-related conditons Immalurity-related conditions Infections Sudden infant death Deaths due to external causes Specific conditions other than above Remaining causes
All stillbirths and infant deaths were divided into six groups of a clinico-pathological classification of causes of infant deaths and stillbirths which has been developed in Denmark.3 This classification focussed on fetal and neonatal factors.4 It was based on information in the MBR on underlying cause of death, and in some cases also on time of death, gestational age, and time of fetal death in relation to onset of labour. The classification of feto-infant deaths is shown in Table 2. Group FI 2 included deaths which were not explained by fetal factors and deaths which were unexplained from an obstetric point of view. Overall survival as well as survival from each of the competing causes-of-death groups was studied using the product limit (Kaplan-Meier) method.5 In the life table analysis of each competing causes-of-death
TABLE 2 Clinico-pathological classification of underlying causes of feto-infant deaths based on fetal and neonatal factors. FI 1. Congenital anomaly: Deaths that had their origin at conception or during embryogenesis. FI 2. Unexplained death prior to the onset of labour: Deaths due to asphyxia and other unexplained antepartum deaths (e.g. premature rupture of membranes, placental abruption, placenta praevia, placenta! insufficiency and cause unknown). FI 3. lntrapartum events: Deaths before or after birth due to events occurring during labour. FI 4. Conditions consequent upon preterm birth: Deaths due to pulmonary immaturity, hyaline membrane disease, cerebral haemorrhage or infarction in preterm babies. FI 5. Other specific conditions: All other deaths due to intracranial haemorrhage. Deaths due to isoimmunization, or infection, and deaths due to other specific miscellaneous conditions including unequivocal cot death. FI 6. Unclassified deaths: Inadequately documented deaths.
group, babies who died from other causes were censored. The relatively small group of babies who emigrated during the follow-up period could not be censored. In the survival analysis of 262159 liveborn singletons Kaplan-Meier estimates were computed at completion of each hour during the first day, each day during the first week and each subsequent week of follow up. The feto-infant survival analysis was based on 261 722 single fetuses and neonates with known gestational age. In order to eliminate potential bias from underreported stillbirths with low gestational age who had been misclassified as abortions, feto-infant survival analysis was restricted to unborn fetuses and live babies, who had survived 31 completed weeks after LMP in utero or as very preterm live infants. Time of death in relation to LMP in liveborn babies was calculated as the sum of gestational age at birth and duration of life in completed weeks. Consequently survival was expressed in completed weeks after LMP. As live babies are only followed up for 1 year in the birth register, feto-infant follow-up was truncated at 76 weeks after LMP. This was done to avoid bias from loss to follow-up of prematurely born babies with gestational age 24 weeks or more, who had survived 52 weeks after birth. RESULTS Infant Mortality Please recognize that the horizontal scale in Figure 1 changes twice. The first 24 data points after birth (B) indicate 0-23 completed hours of the first day of life, the next 6 points 1-6 completed days of the first week and the rest of the points each completed week of the
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
on the basis of history or ultrasound and was defined as the number of completed weeks after last menstrual period (LMP). Gestational age at fetal death was equal to gestational age at birth. Survival of liveborn babies was expressed in completed hours during the first 3 days of life, in completed days during the rest of the first week and in completed weeks during the following 3 weeks. Date of death was also recorded. Only late fetal deaths (28 completed weeks of gestation, stillbirths) were recorded in the MBR, while live births were recorded irrespective of gestational age. Causes of death from certificates of death were coded using the International Classification of Diseases (ICD) according to WHO guidelines in the Department of Medical Statistics of The National Board of Health by three clerks under medical supervision. Coding practice was checked in the department on a regular basis. More than 80% of stillbirths and 70% of neonatal deaths underwent autopsy during the period. All infant deaths were divided into eight functional groups published by the International Collaborative Effort on Perinatal and Infant Mortality (ICE) on the basis of underlying cause of death.2 The groups are shown in Table 1.
321
322
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
first year of life. Deaths within the first 59 minutes of life survived 0 completed hours. The vertical scale indicates the fraction of survivors. The lowest line in Figure 1 depicts the survival function of the 1983-1987 cohorts of liveborn babies. About half of all first day deaths occurred during the first 4 hours. More than half of all first week deaths happened during the first day.
FIGURE 1 Survival function for liveborn singletons. Mortality divided into eight groups of the ICE functional classification of infant deaths.
The contribution of each cause-of-death group to mortality up to a given point is indicated by the distance between each line and the line immediately above it. When the risk of death from immaturityrelated conditions in Figure 1 decreased substantially after 3 completed weeks, sudden infant death syndrome (SIDS) became the largest risk until 5 months. Only after that did the risk of infant death become very low. Feto-Infant Mortality About half of all perinatal deaths in Denmark are stillbirths. All deaths in the perinatal period and deaths related to perinatal events are important for the evaluation of the health care system be they stillbirths or infant deaths. When follow-up starts at birth no stillbirths are included and the members of the cohort have very different ages of development ranging from about 24 weeks to about 43 weeks after LMP. Consequently, we explored how survival analyses of fetoinfant cohorts on the basis of gestational age could provide insights for the study of perinatal losses which cohort studies of mortality in liveborn babies alone could not provide. In the survival analysis which is presented in Figure 2 stillbirths after 31 completed weeks of gestation were included in the population at risk together with all un-
FIGURE 2 Survival function for fetuses and babies who had survived 31 completed weeks after last menstrual period. Mortality divided into six groups of a clinico-pathological classification of stillbirths and infant deaths based on fetal and infant factors.
DISCUSSION Graphic presentation of feto-infant mortality distributed on functional cause-of-death groups improved the expression of the relative contribution and timing of the different causes of deaths. The results suggest that SIDS contributed just as much to infant mortality as did immaturity-related conditions. Fetal deaths prior to the onset of labour and SIDS deaths are major challenges to contemporary obstetric and paediatric care. The analysis was based on routine data and the allocation of individual cases to the causes-of-death
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
born babies and very immature liveborn infants who were still alive 31 completed weeks after LMP. Nearly all of the babies were unborn at the start of follow-up. After 10-11 completed weeks of follow-up practically all babies had been born. The lowest line in Figure 2 shows the survival function of the 1983-1987 cohorts of single live fetuses and infants from 31 weeks to 76 weeks after LMP. At the end of the 44 weeks period of follow-up more than 99% of the initial population at risk were still alive. During the period in which the babies were born the survival curve was relatively steep because stillbirths and a large proportion of the neonatal deaths occurred during that period. More than half of all deaths from 31 weeks to 76 weeks after LMP were either fetal deaths prior to the onset of labour unexplained by fetal factors or unexplained SIDS deaths. Virtually all deaths due to the consequences of preterm birth occurred during the first 10-12 weeks of follow-up.
323
LIFE TABLE ANALYSIS OF INFANT MORTALITY
The integration of the life table approach and analysis of underlying causes of death should be further explored as a way of utilizing vital statistic databases for the evaluation of perinatal care. Analytical studies using proportional hazard regression for the study of differences between groups expand the life table approach and should be encouraged.12 The feto-infant approach is of special interest in countries which have registers of live births and fetal deaths from week 22 or earlier.
REFERENCES 1
Knudsen L B, Kristensen F B. Monitoring perinatal mortality and perinatal care with a national register. Content and usage of the Danish Medical Birth Register. Comm Med 1986; 8: 29-36. 2 Cole S, Hartford R B, Bergsjp P, McCarthy B. International Collaborative Effort (ICE) on Birth Weight, Plurality, Perinatal, and Infant Mortality III: A method of grouping underlying causes of death to aid international comparisons. Acta Obstel Gynecol Scand 1989; 68: 113-17. 3 Kristensen F B. A short list of causes of death in the perinatal period based on principles published by Hey, Lloyd and Wigglesworth. Appendix In: Sundhedsstyrelsen. Medicinsk fodsels- og misdannelsesstatistik 1986. Vitalstatistik 1988; 1:23 (Copenhagen). 4 Hey E N, Lloyd D J, Wigglesworth J S. Classifying perinatal death. Fetal and neonatal factors. Br J Obstet Gynaecol 1986; 93: 1213-23. 5 Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 457-81. 6 Andersen K V, Helweg-Larsen K, Lange A P. The validity of the perinatal classification by the Danish Health Department. UgeskrLteger 1991; 153: 1575-77. 7 Shapiro S, Jones E W, Densen P M. A life table of pregnancy terminations and correlates of fetal loss. Milbank Mem Fund Q 1962; 40: 7-45. 8 French F E, Bierman J M. Probabilities of fetal mortality. Public Health Rep 1962; 77: 835-47. ' Mellin G W. Fetal life tables. A means of establishing perinatal rates of risk. JAMA 1962; 180: 91-94. 10 Bakketeig L S, Seigel D G, Sternthal P M. A fetal-infant life table based on single births in Norway, 1967-1973. Am J Epidemiol 1978; 107: 216-25. ' ' International Federation of Gynecology and Obstetrics. Report of the Committee following a workshop on Monitoring and Reporting Perinatal Mortality and Morbidity March, 1982. London: The Chameleon Press Ltd. 12 Cox D R. Regression models and life tables (including discussion). J Roy Stat Soc B 1972; 34: 187-220.
(Revised version received September 1991)
Downloaded from http://ije.oxfordjournals.org/ at University of Victoria on August 14, 2015
groups was done with a computer algorithm from ICD codes in the MBR. Despite a high autopsy rate and a uniform coding practice the distribution on causes of death should be interpreted with as much caution as is normally called for when drawing conclusions from routine mortality statistics. A recent study of all 282 perinatal and neonatal deaths in three counties in 1985-1986 evaluated the feto-infant classification in Table 2. Experts classified deaths on the basis of all available record material and compared these results with the algorithm's classification of the same deaths from routine data in the MBR.6 The algorithm was found to be relatively invulnerable to misclassifications at ICD code level as 87% of deaths were correctly classified. Two fetal life table analyses were published in 1962.7-8 Taking the point of view that birth can be looked upon as an event in life, not the beginning of it, Mellin extended the fetal life table concept to gestational week specific life tables for the neonatal period.9 A study of the Norwegian Birth Register 1967-1973 provided information on the probabilities of fetal, neonatal, and postneonatal death for each subsequent week after 16 weeks of gestation.10 The report recommended the feto-infant approach as an extension of vital statistics of value in monitoring health changes and in comparing perinatal mortality between populations. Our study was such an extension, which took causes of death into consideration. Feto-infant follow-up from 31 weeks after LMP rather than from 28 weeks or earlier was chosen in order to avoid bias from potential underreporting to MBR of some stillbirths misclassified as late abortions at around 28 weeks of gestation. Very immature liveborn babies had already died at the beginning of follow-up. Consequently, group FI 4 (consequences of preterm birth) in Figure 2 was relatively small when compared with immaturity-related conditions (ICE 3) in Figure 1. As a consequence of this conservative approach deaths in very prematuare babies and stillbirths before 31 weeks (one-third of all neonatal deaths and one-fifth of all stillbirths) were not depicted in Figure 2. This is a serious limitation to the feto-infant approach to data which do not include late abortions. Preferably registration of fetal deaths and live births should start at 22 weeks or earlier." In Denmark, this would require a change in the birth registration system.
