Pregnancy-associated deaths in Finland 1987-1994 - definition problems and benefits of record linkage

Acta Obstetricia et Gynecologica Scandinavica

ISSN 0001-6349

ORIGINAL ARTICLE

Pregnancy-associated deaths in Finland 1987-1994 - definition problems and benefits of record linkage MIKA GISSLER’, RIITTA KAUPPILA~, JOUNI MEFIILAINEN”~, HENRI TOUKOMAA1’3 AND ELINA HEMMINKI’

From ’Stakes, (the National Research and Development Centre for Welfare and Health), the 2Provincial State Office of Hame, Department of Social Affairs and Health, and the 3University of Helsinki, Department of Public Health, Helsinki, Finland

Acta Ohslet Gynecol Scand 1997; 16: 651-657. 0 Acta Obstet Gynecol Scand 1997

Background. Our aim was to study the impact of record linkage and different classification principles on maternal mortality rate. Methods. The death certificates of all fertile-aged women who died in 1987-94 in Finland (n=9,192) were linked to the Birth, Abortion, and Hospital Discharge Registers (n=513,472 births, 93.807 induced abortions, and 71,701 other ended pregnancies) to identify the women who had been pregnant during their last year of life. All deaths that occurred up to 1 year after the end of pregnancy were classified according to their connection to pregnancy. Results. In total, 281 qualifying deaths were found. Only in 22’% of the death certificates was the pregnancy or its end mentioned. The mortality rate was 41 per 100.000 registered ended pregnancies (27 for births, 48 for miscarriages or ectopic pregnancies, and 101 for abortions). The maternal mortality rate depended greatly on which of these 281 cases were defined as maternal deaths. The early maternal mortality rate varied between 5.6 and 6.8 per 100,000 live births. and the late maternal mortality rate between 0.6 and 2.5 depending on the definition used. The classification of other than direct maternal deaths was ambiguous, especially in case of late cancers, cardio- and cerebrovascular diseases, and early suicides. The official Finnish figure for early maternal mortality (6.0/100,000 live births) seems to be a good estimate, although only 65% of individual deaths were unambiguously classified. Conclusions. Register linkage is necessary to identify late maternal deaths and pregnancy- associated deaths. The current official classification of maternal deaths as indirect, direct and fortuitous is arbitrary and allows much variation in defining a maternal death.

Key words: classification; maternal mortality; pregnancy-associated death; register linkage

Submitted 22 November, 1996 Accepted 21 January, 1997

Many international comparisons have used maternal mortality as an indicator of the quality of health services. It has been estimated that globally more than 500,000 women die every year from causes related to pregnancy and childbirth (1). This estimate is, however, very rough, mainly be-

Abhreviutions: ACOG: American College of Obstetricians and Gynecologists; CDC: Center for Disease Control; HDR: Hospital Discharge Register; ICD: International Classification of Diseases; MBR: Medical Birth Register; WHO: World Health Organization.

cause of a lack of routine data collection in devel- oping countries where over 99% of all maternal deaths occur (2, 3).

Also in the developed countries, the usefulness of maternal mortality statistics has been ques- tioned due to underreporting and difficulties in defining a maternal death. In Georgia (U.S.A.), 27% of maternal deaths were not reported (4). In the United Kingdom, a 38% higher maternal mortality rate was found by a confidential inquiry compared to the official rate (5) . In Flanders (Bel-

0 Acta Obstet Gynecol Scand 76 (1997)

652 M. Gissler r t al.

gium) and in France, 48'% and 56% of maternal deaths were not included in the official statistics (6, 7). Furthermore, there is no consensus concerning which cases should be included as maternal deaths. Problematic are, for example, some cancers, stroke, asthma, liver cirrhosis, pneumonia with in- fluenLa, anorexia nervosa, and many violent deaths, such as suicide, homicide, and accidents (6, 8-1 1).

This study had two purposes. First, our aim was to study, how completely maternal deaths were re- corded in Finland between 1987-1994 by using record linkage of national registers to identify all deaths occurring within a year of a completed pregnancy without LI priori definition of the case as a maternal death or not. The second aim was to study the impact of different classification principles on the maternal mortality rate.

Material and methods

According to the International Classification of Diseases (ICD, 9th and 10th revisions), a maternal death is the death of a woman while pregnant or within 42 days of termination of a pregnancy, irrespective of the duration and the site of the pregnancy, caused by a disease related to (direct cause) or aggravated by the pregnancy or its management (indirect cause). Accidental or incidental deaths are not included (12, 13).

In the 10th revision of the ICD, two new con- cepts were introduccd. The term late maternal death indicates the death of a woman from direct or indirect obstetric causes more than 42 days but less than one year after the termination of the pregnancy. A pregnancy-related death was defined to be the death of a woman while pregnant or within 42 days of the termination of the pregnancy, irrespective of the cause of death ( 1 3).

Furthermore, in the United States, the Center for Disease Control (CDC) and the American Col- lege of Obstetricians and Gynecologists (ACOG)

Table I . Definitions related to maternal death by causality and timing'

Causality Timing (days)

Direct Indirect Fortuitous 0-42 43-365

Maternal death' X X - x - Late maternal death X X - - x Pregnancy-related death X X X x - Preg nancy-associated death X X X x x

' X-included, -:excluded. In our text, the unofficial term of early maternal death is also used to clarify the definitions.

0 Actu Ohstet Gynecol Sctrrrd 76 ( I 997)

have recently introduced the definition pregnancy- associated death which includes all deaths of pregnant women or within one year of the termination of the pregnancy, irrespective of the cause of death or site of pregnancy (14). The definitions used in this study are clarified in Table 1.

The Finnish Death Cause Register is based on death certificates written either by the physician who has been taking care of the patient, or in case of a violent death or an unclear death, by the ob- ducting physician. All death certificates are check- ed by a physician in the provincial government or health care office, and by medical experts at Stat- istics Finland.

The Medical Birth Register (MBR) was started in 1987 to collect inforniation on livebirths and stillbirths weighing more than 500 g, or having a gestation length of 22 weeks or more (15, 16). After individual linkage with Birth and Death Cer- tificates, the MBR is complete (17). Information on abortions has been collected since 1950, and the Abortion Register was computerized in 1977. According to a data quality study, more than 99% of induced abortions are reported in the register (18). The Hospital Discharge Register (HDR) is based on inpatient care episodes. According to a data quality study in 1986, 9.5%) of hospitalizations were registered, and 97% of main diagnoses concerning pregnancy, birth, and puerperium were correctly reported (19). A death during pregnancy or after a miscarriage is found in the HDR only if the woman was hospitalized.

The basic data consist of information on all the deaths of 15-49 year-old women in Finland between 1987-94 (n=9,192 deaths, of which 6,029 were natural deaths) (20). All deaths were individu- ally linked to the MBR (n=513,472 between 1987- 94), to the Abortion Register (n=107,126 between 1986-941, and to the HDR (n=9,882,254 hospitalizations including 57,184 miscarriages and 1 4 s 17 ectopic pregnancies between 1986-94) to find evi- dence of a pregnancy up to one year before the death. All together, 284 deaths were found, but three occurring abroad were excluded because of missing information on cause of death.

In further analyses, a forensic pathologist (RK) reviewed and reclassified the death certificates of all remaining 281 cases according to the underlying cause of death. Since it turned out to be difficult to define whether some deaths were indirect maternal deaths or fortuitous deaths, a separate class was created to estimate their impact on the maternal mortality figures (see also Table 111). The main cause of death mentioned in the death certificate was changed for eight cases: two deaths having drug dependence as an underlying cause were reclassified either as an unknown violent death or an

Pregnancy-assoc~~ited deaths in Fintund 653

Table II. Pregnancy-associated mortality per 100,000 cases and age-adjusted odds ratios by the type of end of pregnancy compared to other women, Fin- land 1987-1 994

accident, one death because of conditions of the nature was reclassified as a suicide, and five unknown violent deaths were classified either as an accident (n=3) or a suicide (n=2).

In the official Finnish maternal mortality figures, all violent deaths were excluded. We made two hypothetical calculations to estimate the impact of such deaths: a) half of the suicides and homicides during pregnancy or within 42 days were arbitrarily defined as indirect maternal deaths and included as early maternal deaths, and b) 10% of suicides and homicides after 42 days were classified as late indirect maternal deaths.

The maternal mortality rate was counted per 100,000 live births, as recommended by WHO (1 3 ) . The pregnancy-associated mortality rate was counted by the type of end of pregnancy (per 100,000 births, 100,000 induced abortions, and 100,000 miscarriages and ectopic pregnancies). Age-adjusted mortality figures were calculated using the age distribution of registered pregnancies. Age-adjusted odds ratios (OR) were calculated for the main causes of death based on our classification by the type of pregnancy, using deaths without a pregnancy as a standard. The mean of the minimum and the maximum number of maternal deaths resulting from using different definitions was used as the best estimate of maternal deaths.

Results Completeness

According to the official statistics, 31 deaths were defined as early maternal deaths. With the register linkage, we found in total 281 pregnancy-associated deaths, of which 78 occurred within 42 days of the end of the pregnancy. The pregnancy-associated mortality rate was 41.4 per 100,000 registered pregnancies. The risk for a death after a birth was notably lower than that for women without a completed pregnancy during the last year of living. After an abortion, the mortality risk was increased for accidents, suicides, and homicides (Table 11). Our data also included 20 deaths occurring during pregnancy, giving a mortality rate of 2.9 per 100,000 pregnancies. This, however, is likely to be an underestimation.

A pregnancy was mentioned in 22% of the death certificates (for 55% of deaths within 42 days, and 9% of those occurring later). These proportions were considerably higher for natural deaths (72% and 14%) than for violent deaths (20% and 6'Yo).

Impact oj' clussiJicution

On the basis of death certificates, it was easy to distinguish between direct and indirect maternal

End of pregnancy No

Birth Miscarriage Abortion pregnancy1

Number of deaths

Mortality: Crude, total Age-adjusted, total

OR2: Total mortality

Natural deaths

Accidents

Suicides

Homicides

137 40

26.7 47.8 29.4 51.3

0.50 0.87 (0.32-0.78) (0.60-1.27)

0.49 0.43 (0.27-0.89) (0.23-0.80)

0.49 1.40 (0.18-1.33) (0.66-2.98)

0.57 I .44 (0.22-1.48) (0.68-3.05)

0.31 1.82 (0.02-4.42) (0.36-9.10)

84

100.5 103.2

1.76 (1.27-2.42)

0.80 (0.48-1.33)

2.08 (1.03-4.20)

3.68 (1.92-7.04)

4.33 (1.03-18.2)

8931

91.6 58.8

1 .o

1 .o

1 .o

1 .o

1 .o ~ _ _ _ _ _ _ _ _

Women aged 15-49 not having a completed pregnancy during their last year of living, including 20 deaths of pregnant women Age-adjusted odds ratio of mortality after birth, miscarriage or abortion compared to mortality of other women (95% confidence intervals in paren- theses)

deaths, but some deaths could be classified either as an indirect maternal death or a fortuitous death depending on the timing and previous medical conditions of the woman (Table 111). Problems were found especially for some cancers (i.e., stom- ach, liver, skin, breast, and ovary), cardiomyopathy, occlusion of cerebral arteries, thrombophle- bitis, and late natural deaths in general.

We classified 26 deaths as direct maternal deaths (23% occurred during pregnancy, 58Y0 after a birth, 4% after an induced abortion, and 15% after a miscarriage). The most common reasons for the 24 early direct maternal deaths were pulmonary embolism (n=6), ectopic pregnancy (n=3), ante- partum hemorrhage, abruptio placentae, and placenta previa (n=3), and venous complications in pregnancy and puerperium (n=3). The two late direct maternal deaths were caused by cancer of the placenta and by cardiomyopathy. The five early indirect maternal deaths were caused by cardiovascu- lar problems (n= 3), hypertension, and hepat- opathy. The only late indirect maternal death was caused by breast cancer.

Six early and ten late deaths were classified in an uncertain category reflecting indecision about whether the death was an indirect maternal death or a fortuitous death. Ten of these deaths (one early and nine late) were caused by cancers, and three (two early and one late) by cardio- and cerebrovascular accidents. The three remaining deaths were caused by virus-associated hemophagocytic

0 Actu Obstet Gynerol Scand 76 (1997)

654 M . Gissler. ct ul.

Table Ill Our classification of pregnancy-associated deaths by timing, Finland 1987-1994

n %

Early’ Late Early Late 0-42 d 43-365 d Total2 0-42 d 43-365 d Total

Direct obstetric death 24 2 26 31 1 9 Indirect obstetric death 5 1 6 6 0 2 Maternal or fortuitous natural death 6 10 16 8 5 6 Fortuitous natural death 18 59 77 23 29 27 5 Unclear natural cause 2 2 0 1 1 Accident 10 47 57 13 23 20 Suicide 12 65 77 15 32 27 5 Homicide 2 12 14 3 6 5 Unclear violent cause 1 5 6 1 2 2

Total 78 203 281 100 99 100

-

Pregnancy-related deaths. Pregnancy-associated deaths

syndrome, by streptococcal infection and by an unclear cause.

Table 1V illustrates the differences in how deaths were categorized using our classification system and the official classification system. There were 40 cases defiiicd as an early maternal death either by us or in the official statistics, but joint classification occurred in only 65% of cases (n=26).

Three deaths (one embolism and two aneu- rysms) were classified as maternal deaths by Stat- istics Finland, but as fortuitous deaths by us. In these cases, the diagnosis of the original cause of death had been changed to a maternal death by Statistics Finland. One direct obstetrical death was not included in the official statistics because it occurred with a nonresident foreigner, and nonresident foreigners are excluded from the official statistics. Two deaths not included in the official maternal mortality statistics were classified by us as indirect maternal deaths and six deaths as possible maternal deaths. Two cases included in the official statistics were not found in our data: One death

Table IV. The classification of pregnancy-related deaths [n=78) by their in- clusion in the official maternal statistics, Finland 1987-1994

Maternal death in official statistics

Our classification

Direct obstetric death Indirect obstetric death Maternal or fortuitous natural death Fortuitous natural death Violent death [accident, suicide, or homicide) Not included in our data

Total

Yes

23 3 0 3 0 2’

31

No

1 2 6

15 25

49

-

concerned a Finnish woman living permanently abroad, and the other death concerned a woman who was unconscious two years after giving birth due to birth complications.

When using different classifications, the number of early maternal deaths varied between 29 and 35, while the official number of maternal deaths was 31 (Table V). The classification of late maternal deaths was more difficult: We found at least three, but up to 13 cases, which might have been classified as late maternal deaths. Including late maternal deaths in maternal mortality figures increased the maternal mortality rate by 30%. Fur- thermore, including a percentage of suicides and homicides as maternal deaths increased the early maternal mortality rate by 23%, and the total maternal mortality rate by 38%.

Discussion

Different methods have been used to increase the validity of maternal mortality statistics (21). Local voluntary confidential inquiries were started in Scotland and Australia already in the 1930’s to get more accurate information on maternal deaths (2, 22). In Flanders (Belgium) a separate question on maternal mortality is included in voluntary birth registration (7). lntroduction of a revised death certificate with an item on pregnancy within the past year improved maternal mortality surveillance in Puerto Rico (23). In Germany, the active role of state health authorities in finding maternal deaths has been suggested to improve the mortality statistics (24). An extreme effort to include all ina- ternal deaths in the official figures is found in Aus- tria, where even newspaper articles have been used to identify violent and accidental deaths (25).

In our data, pregnancy was mentioned in the

Pregnancy-associated deaths in Finland 655

Table V. Maternal mortality by the official classification and by our classification, Finland 1987-1994

Definition

Maternal deaths in the official statistics

Excluding violent deaths: Early maternal deaths Early+late maternal deaths Pregnancy-related deaths 5 4 2 days Pregnancy-associated deaths 5 1 year

Including some violent deaths: Early maternal deaths2 Earlyilate maternal deaths3

n Per 100,000 live births

Best estimate Range’

31

32 (29-35) 40 (32-48) 78

281

39 (36-42) 55 (47-63)

Best estimate flange’

6.0

6.2 (5.7-6.8) 7.8 (6.2-9.4)

15.2 54.8

7.6 (7.0-8.2) 10.8 (9.2-12.3)

Range=minimum and maximum number of cases or rate when using different definitions. 50% of suicides and homicides defined as maternal deaths. 50% of suicides and homicides before 42 days and 10% after 42 days defined as maternal deaths

death certificate in less than half of natural deaths, and in every tenth violent pregnancy-associated death. This makes data linkage of the Death Cause Register to other registers necessary when studying pregnancy-associated deaths. In other countries having poor coverage of maternal mortality statistics, this kind of data linkage could improve the accuracy of maternal mortality figures. However, to be easy this kind of linkage requires a uniform system of person identification numbers, like the ones used in the Nordic countries (26). In addition, the usefulness of the record linkage depends on the quality of the data in registers on reproduction. In Finland, these registers have been shown to be good, but the situation can be different in other countries, as has been shown for Denmark (27-28).

Some earlier studies have used birth register or birth certificate data to improve maternal mortality statistics. In Georgia (U.S.A.), a record linkage revealed eight new maternal deaths, increasing the maternal mortality figure by 36% (4). In New York City, only one new pregnancy-related death was found by using a record linkage, but 13 new cases were found when using other methods (9). In Sweden, data linkage of death certificates and the Medical Birth Register for 1980-1988 did not re- veal any new direct maternal deaths, but did un- cover six pregnancy-associated deaths (1 1). In our study, we also used the Hospital Discharge Regis- ter and the Abortion Register to find additional pregnancy-associated deaths, which doubled the total number of deaths. Three maternal deaths were identified with the data linkage.

Our data linkage showed noticeable differences in the pregnancy-associated mortality by the type of pregnancy. The age-adjusted risk for a violent death (accident, suicide, or homicide) was increased for women with a recent abortion compared to other women, probably because of factors

related to social class and lifestyle (29). Further- more, the age-adjusted risk for a natural death was decreased after a birth or a miscarriage compared to that for women without a recent completed pregnancy. This may be explained by the fact that the women capable of and willing to have children are healthier than women in general.

Our study covered deaths after births and abortions well, but it may underestimate the real number of pregnancy-associated deaths after miscarriage and deaths occurring during pregnancy. These deaths were found only if the woman was hospitalized, or if the pregnancy was mentioned in the death certificate. The mortality rate within one year after the end of a pregnancy was 38.5 (=41.4- 2.9) per 100,000 pregnancies. Assuming that the mortality risk level during pregnancy is similar, the mortality rate for pregnant women should be 24.4 (adjusted by gestation length) instead of the found 2.9 per 100,000 pregnancies, suggesting that many deaths of pregnant women are missing from our data. This estimate on missing cases may, however, be too high, because the risk of death may considerably vary during the course of a pregnancy and after its end.

It is problematic to determine causality between a pregnancy and a natural death, and therefore groups for borderline cases were introduced in our classification. This especially concerns some hor- mone-related cancers, and cardio- and cerebrovascular diseases. Including or excluding this borderline group caused a noticeable variation in the maternal death rate estimates. Once a death was defined as an early maternal death, it was easy to classify it as a direct or an indirect maternal death. The proportion of direct maternal deaths varied between 69%1-81% of early and 53%-82% of all maternal deaths depending on the definition used. These rates equal those reported for Australia (lo),

0 Actri Obstet G p e r o l Srand 76 (1997)

656 M . G'isslcr et ril.

Austria (25, 30), Great Britain (lo), and Sweden (11).

All the violent deaths were classified separately bccause the connection with pregnancy was difficult to clarify afterwards based on death certificates. Aftcr making hypothetical calculations including some of the suicides and homicides as indirect maternal deaths, the number of early and late maternal deaths increased considerably. Therefore, it might be useful to give the number of pregnancy-associated violent deaths together with the traditional maternal mortality figures.

The official rate for early Finnish maternal inor- tality was found to be a good estimate with a maximum error of 13%, but the classification of individual cases was ambiguous. The new ICD-10 definitions were not being used in Finland at the time of the study, and therefore no official rates on pregnancy-related or late maternal deaths were given. Our experience suggests that it is difficult to define a late maternal death, a conclusion which can also be reached by comparing the results of previous studies: The proportion of late maternal deaths among all maternal dcaths was estimated to be 11% in the U S A . ( X ) , but 55% in one Australian hospital (31). The old upper tiinelimit of 42 days for maternal mortality has its origins in religious and cultural practices (8), but because of the improved effcctiveness of lifesupporting technology, this limit has become meaningless. The ICD- 10 classification tried to solve this problem, but the concept of late maternal deaths may just compli- cate the collection of maternal mortality data by resulting in disconcordant practices in defining maternal deaths.

The coverage of maternal mortality figures can be improved by using several methods, and data linkage seems to be very valuable. Tn addition, our new concept of pregnancy-associated deaths and their new classification inight give more useful and comparable data. However, international comparisons may still be difficult because registration sys- tems and their completeness vary between countries, and because many countries do not have hcalth registers or they cannot be used in data linkages. An attempt to solve the problems of varying national classifications of maternal death was made when a European research group was formed in 1994 to study the data collection pro- cesseb for and definitions of maternal deaths in Europe; the results of this work will be published in the near future (32).

Acknowledgments

This study is associated with the MOMS-project (Frequency and Risk Factors of Mntern;il Morbidity and Mortality

0 Aclir O h l e t (+iecol Sccirid 76 (1997)

'Avoidable' Deaths and Evaluation of Care). which is financed by the Commission of the European Communities. The project is led by Marie-Helknc Bouvier-Colle, who IS with INSERM. Paris (France).

References

1. WHO. New estimates of maternal mortality. Weekly Epidc-

2. Walters WA. Maternal mortality. Med 5 Aust 1989; 151:

3. Royston E, Abouzahr C. Measuring iiiaternal mortality. Br

miological Record 1991; 66: 345-8.

615-1 6.

J Obstet Gynaecol 1992: 99: 540-3. 4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14

15

Rubin GL,'Peterson HB, Rochat RW, McCarthy BJ. Terry JS. Maternal death after cesarean section in Georgia. Am J Obstet Gynecol 1981; 139: 681-5. Department of Health, Welsh Office. Scottish Office Home and Health Department, Department of Health tiid Social Services, Northern Ireland: Report on Confidential Enquir- ies into Maternal Deaths in 1988-1990. London, 1994. Bouvier-Cok M-H, Varnoux N, Costes P, Hatton E Rea- sons for the underreporting of maternal mortality in France, as indicated by a survey of all deaths among woinen of childbearing age. Int J Epideiiiiol 1991; 20: 717- 21. Vleugcls A, Bekaert A. The Flemish Centre for the Study of Perinatal Epidemiology and its Rcgistry. Quality Assur- ance in Health Care 1992; 4: 115-24. Fortney JA. lmplicatioiis of the ICD-10 definitions related t o death in pregnancy, childbirth or the puerperium. Rapp Trimest Statis Sanit Mond 1990; 43: 246-8. Allen MH, Chavkin W, Marinoff J. Ascertainment of maternal deaths in New York City. Am J Public Health 1991:

Beischcr NA. Maternal mortality in Australia. Aust NZ Obstct Gynaecol 1992; 32: 183-5. Hogberg U, Innala E, Sandstrein A. Maternal mortality in

Sweden 19x0-1988. Obstet Gynecol 1994; 84: 240- 4. WHO: International Statistical Classification of Diseases and Related Health Problems. 9th revision. Vol. 1 . World Health Organisation. Geneva, 1978. WHO: International Statistical Classification of Diseases and Relilted Health Problcms. 10th revision. Vol. 1 . World Health Organisation. Geneva, 1992. Wilcox LS, Marks JS. From data to action. CDC's public health surveillance for women, inthnts and children. U.S. Department of Health and Human Services, 1995. Teperi J. A multi-method approach to the assessment of data quality in the Finnish Medical Birth Registry. J Epidc- miol Coinniunity Health 1993; 47: 242-7.

8 1 : 380-2.

16. Gissler M, Teperi J, Hemminki E, Merilainen J. Data quality after re-structuring a nationwide iiiedical birth registry. Scand J Soc Med 1995: 23: 75-80.

17. MerilZinen .I. Gissler M, Heiiiiiiiiiki E, Teperi J. Finnish perinatal statistics 1993. Stakes Tilastotiedote 1995; I X .

18. Gissler M, Ulander V-M, Heniininki E, Rasimus A. Declin- ing induced abortion rate in Finland: data-quality of the Abortion Register. Int J Epidemiol 1996; 25: 376-80.

19. Keskimgki I , Aro S. Accuracy of data on diagnoses. pro- cedures and accidents in the Finnish Hospital Discharge Register. Int J Health Sciences 1991; 2: 15-21.

20. Statistics Finland: Death Causes 1987-1 994. Finnish Of- ficial Statistics: Health 19x9 -1996. Helsinki, 1989-1996.

21. Atrash HK, Alexaiidcr S. Berg CJ. Maternal mortality in developed countries. Obstct Gynecol 1995; 86: 700-5.

22. WHO: Maternal and perinatal deaths (1981-1985). Weekly Epidemiological Record 1990; 17: 130-1.

23. Coiiies A, Navarro A. Conde J, Blasiiii I, Adamsons K .

Misreporting of maternal mortality in Puerto Rico. Bol Asoc Med PR 1990; 82: 343-6.

24. Welsch H. Mutterliche Mortalitat. Arch Gynecol Obstet 1989; 245; 321-8.

25. Beck A, Vutuc C, Fried1 H. Muttersterbef3lle in Osterreich 1975-1989. Gynakol Rundsch 1991; 31: 166--9.

26. Gissler M, Louhiala P, Hemminki E. Nordic medical birth registers in epidemiological research. Eur J Epidemiol,

27. Schmidt L, Damsgaard MT, Nilsen JM. The validity of some abortion diagnoses in the Danish Hospital Discharge Register. Ugeskr Laeger 1989; 151: 3478-82. (Danish, Eng- lish summary).

28. Andersen KV, Helwig-Larsen K, Lange AP The validity of the perinatal classification by the Danish Health Depart- ment. Ugeskr Laeger 1991; 153: 155-7. (Danish, English summary).

29. Gissler M, Hemminki E, Lonnqvist J. Pregnancy-associated suicides in Finland 1987-1994. BMJ, 1996: 313: 14314.

1997; 13: 169-75.

Pregnancy-associated deaths in Finland 651

30. Heidegger H, Krone HA. Zur Frage der mutterlichen Mor- talitat, ihren Definition und Erfassung. Gynakol Geburts- hilfliche Rundsch 1989; 29: 323-5.

31. Henry OA, Sheedy MT, Beisher NA. When is a maternal death a maternal death? A review of maternal deaths at the Mercy Maternity Hospital, Melbourne. Med J Aust 1989; 151: 628-31.

32. Bouvier-Colle M-H, Varnoux N, Breart G. Les morts ma- ternelles en France. INSERM: Analyses et prospective. Paris, 1994.

Address for correspondence: Mika Gissler, M. SOC. Sc. Stakes P.O. Box 220 0053 1 Helsinki Finland

0 Actu Ohstet Gynecol Scund 76 (1997)

Documents

Pregnancy-associated deaths in Finland 1987-1994 - definition problems and benefits of record linkage