The Measurement of Maternal Mortality IRVINE LOUDON r

^ HERE are, of course, problems in the interpretation of all historical mortalities, but those surrounding maternal mortality show some unusual and rather complicated features, and they illustrate some interesting points concerning what one might call the "mechanics" of death registration in the past. For the most part, I deal here with the Registrar General's statistics for England and Wales. The easiest way to show a trend in mortality is to display the total number of deaths. Figure i shows total maternal deaths (and also total births) in each decade from 1850 to 1980, indexed to 100 in 1850. In round figures there were between 8 and 9 maternal deaths per day in 1850, rising to a peak of 13 per day in the 1890s and falling to less than 1 per week in the 1980s. The shape of this graph— the initial rise followed by the steep fall from the end of the nineteenth century—resembles the trend in infant mortality, which perhaps is not surprising. We think of mothers and infants as closely linked and expect their deaths rates to be shaped by the same sort of determinants of mortality. Showing total deaths, however, is acceptable for comparing something like road deaths over two or three years, but it is a misleading way of showing maternal deaths over a period of 140 years, for it takes no account of changes in the population at risk, which, in the case of maternal deaths, is women of childbearing age (15-44). Figure 2 (also indexed to 100 in 1850) corrects for this factor. The figure shows a huge change in the population of women aged 15—44 a n d also shows the maternal death rate expressed as the number of maternal deaths per 1,000 women aged 15—44. The French, incidentally, were fond of showing maternal mortality in this way. Once again we have © 1999 OXFORD UNIVERSITY PRESS ISSN 0 0 2 2 - 5 0 4 5 VOLUME 54

[ 312]

PAGES 3 12 TO 329

London

:

Measurement of Maternal Mortality

313

Total Births

100

. Total Maternal Deaths

S

Values for 1850 = 100

Fig. 1. Total births and total maternal deaths, England and Wales, 1851—1860 to 1971—1980. Source: Registrar General for England and Wales, Decennial Supplements (London: Her Majesty's Stationery Office).

a graph of maternal deaths showing an initial rise and a steep fall, similar to Figure 1. Figure 2, however, also shows the fall in the crude birth rate. This should alert us to the fact that the population at risk is not all women aged 15-44 but only women during the "puerperal state," which can be defined as during pregnancy, labor, and the postnatal period. It is obvious that, other things being equal, if women have fewer babies, there will be fewer maternal deaths. Thus the correct denominator is not the population of women of childbearing age, but deliveries or births. The definition of the maternal mortality rate (MMR) is the number of maternal deaths per 1,000 (or 10,000 or 100,000) births. Put simply, the MMR measures the cost in maternal deaths of producing 1,000 (or 10,000 or 100,000) babies. Figures 3 and 4 show the trend in the MMR. Figure 3 is

314 Journal of the History of Medicine :

Vol. 54, April iggg

200 -

100

Birth rate

\ ( Maternal Death Rate

1

1

1

1

Values for 1850 = 100 Fig. 2. Birth and maternal death rates for population of women aged 15—44, England and Wales, 1851—1860 to 1971-1980. Source: Registrar General for England and Wales, Decennial Supplements (London: Her Majesty's Stationery Office).

Loudon

:

Measurement of Maternal Mortality

315

60 50 40 30 20 10 0

Fig. 3. Fig. 3. Annual rates of maternal mortality, England and Wales, 1850 to 1970. Source: Registrar General for England and Wales, Decennial Supplements (London: Her Majesty's Stationery Office).

based on annual rates. Figure 4 is based on five-year averages, and the vertical axis is logarithmic to demonstrate comparable rates of decline. This is a vivid example of the importance of choosing the correct denominator, for the shape of the graph is now quite different. Instead of falling steeply from the 1890s, the MMR stayed on a high plateau (albeit a plateau with spikes on it, deliberately smoothed in Fig. 4) from 1850 to the mid-i93os. Then it declined steeply, the first year of the fall being 1937, and continued to fall at a remarkably constant rate until the present. This is an extraordinary trend. The decline in infant mortality began around 1890-1900 in most Western countries. The received wisdom is that the decline was probably due to better standards of living, better diet and housing, improvements in hygiene, and so on. Surely, the same factors should have reduced the MMR? In fact, the MMR should have fallen even more steeply than the infant mortality rate because of additional factors such as the introduction of antisepsis and asepsis (c. 1890), which dramatically lowered the mortality in lying-in hospitals; the training of midwives (from 1902 in England); the introduction of antenatal/prenatal care (roughly from the 1920s);

316 Journal of the History of Medicine 1850

1900

Vol. 54, April 1930 1940

1980

80 70 50 40

30 .2 20 15

o 10

i of

lL

Fig. 4. Maternal mortality rates (five-year averages), logarithmic scale, England and Wales, 1850 to 1980. Source: Registrar General for England and Wales, Decennial Supplements (London: Her Majesty's Stationery Office).

and supposedly better medical education. Yet, instead of a decline there is a plateau; what is more, it was a gently rising plateau from 1900 to the mid-i93Os, the very period -when the additional factors came into operation. It was these paradoxical features that awakened my interest in maternal mortality more than a decade ago. Initially I refused to believe the graph was accurate. How could one possibly explain the plateau of mortality from 1850 to the mid-i93Os, which, if correct, suggested that for those aged 65 or over in 1999, the risk of their mothers dying in childbirth when they were born was virtually the same as it had been in the 1850s? Further, what caused the steep decline from 1937 to the present? I thought perhaps that the trend shown in Figures 3 and 4 was a statistical artifact. Not so. Although there are many confounding factors, some slight, others quite considerable, the trend shown in

London

:

Measurement of Maternal Mortality

317

Figures 3 and 4 is, broadly speaking, accurate. But there are still considerable difficulties, especially when it comes to international comparisons. THE CORRECT DENOMINATOR

I have said the correct denominator is deliveries or births. In fact, this is not quite true. Because some women died without delivering during pregnancy, the correct denominator is pregnancies rather than births. Unfortunately, for England and Wales, the number of pregnancies that end in abortion is (and always will be) unknown, so births have to be used instead. Then there is the question of multiple births. Twins are one delivery but two births. Because twins occur on average in one in 80 births, triplets in one in 802, and quadruplets in one in 803 (proof that God is a mathematician?), one can calculate the probable number of deliveries for any given number of births. But, as shown below, it is scarcely worth the trouble because it makes so little difference. A greater problem is stillbirths. Stillbirths were not registered in England and Wales until 1929. The M M R was therefore based on live births before that date and total births (live births plus stillbirths) afterward. This makes some difference but not very much. An even greater problem is the wide variety of definitions of stillbirths in various countries, especially in the United States and France. It has been shown, however, that in practice this makes little difference to international comparisons.1 The extent of the deviation due to these factors can be shown by a simple calculation. For this I have arbitrarily chosen England and Wales in the year 1925. In 1925, the official published maternal mortality rate per 10,000 live births was 40.80. The maternal mortality rate if allowance had been made for stillbirths, would have been 38.76, and for stillbirths and multiple births 39.60. The difference between an M M R of 40.80 and 39.60 lies well within the range of random variation (see Table 3). THE DEFINITION OF MATERNAL DEATHS

Defining a maternal death would appear to be easy, but it is not. If a woman dies of a cause directly related to childbirth (puerperal fever 1. Elizabeth Tandy, Comparability of Maternal Mortality Rates in the United States and Certain

Foreign Countries, Children's Bureau Publications, no. 229, U.S. Dept. of Labor (Washington, D.C.: Government Printing Office, 1935).

318

Journal of the History of Medicine

:

Vol. 54, April lggg

or obstetric hemorrhage, for example), it is clearly a maternal death. Such deaths used to be called "true" maternal deaths; they are now called "direct." If a woman who happened to be pregnant was run over and killed by a horse and cart or automobile, it would be silly to include the case as a maternal death; it would be an accidental death. The fact of pregnancy, however, may be noted. Such deaths are now recorded as "fortuitous [maternal] deaths."2 But what if a woman died during pregnancy, in labor, or after delivery from influenza, heart disease, smallpox, or tuberculosis? Such deaths, now called "indirect maternal deaths," used to be called "associated maternal deaths." Should they be included as maternal deaths in the published records of a nation's MMR, or should they be excluded? (They can always be counted and delegated to a footnote for comparative purposes.) According to one view, pregnancy was a delicate state which weakened the resistance of women and made them more liable to die from the general diseases that afflict the human race. In this sense they were true maternal deaths and should be counted as such. Countries in which this view prevailed (United States, Denmark, Australia, and Scotland) therefore included associated deaths in their calculations of maternal mortality.3 Statistics for these countries showed a high peak of maternal mortality in 1918—1919 corresponding to the pandemic of influenza. In other countries it was held that pregnancy is a healthy state in which women are no more liable to die of incidental disease than they would have been if not pregnant. This was the English view, and in 1933 the Registrar General for England and Wales appeared to prove the point. He calculated the maternal deaths due to various diseases such as influenza, pneumonia, heart disease, and so on, and compared the result with the death rate from the same diseases in a matched cohort of women of the same age structure and social class who were not pregnant or had not recently delivered. There was no 2. Report on Confidential Enquiries into Maternal Deaths in England and Wales, Report on Health and Social Subjects 29, Department of Health and Social Security (London: Her Majesty's Stationery Office, 1986). 3. Robert M. Woodbury, Maternal Mortality. Tlie Risk of Death in Childbirth and from All the Diseases Caused by Pregnancy and Confinement, Children's Bureau Publications, no. 152, U.S. Dept. of Labor (Washington, D.C.: Government Printing Office, 1926), p. 32.

Loudon : Measurement of Maternal Mortality

319

difference. He concluded that "this group of deaths [i.e., associated deaths] should not be laid at the door of childbearing."4 In fact, recent work has shown that childbearing women have reduced immunity. They are more likely to die of infective disorders during the process of childbearing. Technically, the Registrar General was wrong in 1933, but the point at issue is the effect on international comparisons of maternal mortality. How much were comparative rates affected by the fact that some countries published associated deaths as maternal deaths while others did not? In the United States, which, during the first half of this century, had the highest national rate of maternal mortality in the Western world, it was often said that international comparisons were meaningless precisely because of this factor.5 It was all a question of the inclusion or exclusion of "associated deaths." Were they right? Elizabeth Tandy, a member of the U.S. Children's Bureau, tackled this problem in 1935. She collected a number of case histories —a mixture of true (direct) and associated (indirect) maternal deaths — and sent them to the vital statistics authorities in various countries. She asked them to see how many of these deaths they would have assigned to maternal deaths according to their own rules. The results can be seen in Table 1. Then she calculated what the mortality rates for each country would have been if they had all used the method of assigning maternal deaths used by the United States. The results are shown in Table 2.6 Certainly there is a difference between columns "1927 (A)" and "1927 (B)," but the United States still occupied the unenviable position of having the worst MMR in the world. But there is more to this than meets the eye. As Table 1 shows, the difference between countries that included associated deaths and those that did not was not clear cut. Countries differed in which associated deaths they included. It is important to remember that in spite of guidance in the form of international publications on the classification of causes of death, each country was to some extent 4. Annual Report of the Registrar General for England and Wales for 1933 (London: His

Majesty's Stationery Office, 1934), pp. 97-8. 5. Woodbury, (n. 3) Maternal mortality, p. 57; SJ. Baker, "Maternal mortality in the United States,"/ Am. Med. Assoc, 1927, 8g, 2016-17; J-G. Marmol, A. L. Scriggins, and R. F. Vollman, "History of the maternal mortality study committees in the United States," Obstet. Gynaecoi, 1969, 34, 123-38. 6. Tandy, (n. 1) Comparability of Maternal Mortality.

320

Journal of the History of Medicine

:

Vol. 54, April iggg

TABLE 1

The Assignment of 447 Maternal Deaths according to the Methodology Used in Various Countries for the Classification of Maternal Deaths Puerperal causes (%)

Country

Denmark USA

New Zealand Australia Scotland Netherlands France Sweden England and Wales Norway

Non-puerperal causes (%)

99-4 92.9 92.8 92.7 92.3 91.9 82.7 80.5 78.7 76.9

0.6 7-i 7-2

7-3 7-7 8.1

17-3 19-5 21.3 23-1

Source: Elizabeth C. Tandy, Comparability of Maternal Mortality Rates in the United States and Certain Foreign Countries, Children's

Bureau Publication, no. 229 (Washington, D.C.: Government Printing Office, 1935).

guided by its own conventions, its own "house rules." The same applied to individual American states. Some of the published statistics of maternal mortality included only true or direct maternal deaths, others included associated deaths as well. In England and Wales, inclusion of associated deaths would have raised the MMR by between 15 and 20 percent. Note that Scotland and Australia used the same methodology as the United States, whereas Denmark appeared to include even more associated deaths than Scotland and the United States. REGISTRATION EFFECTS

Changes in the methodology of death registration had only minor effects on the records of maternal mortality. In England and Wales the Births and Deaths Registration Act of 1874 made the certification of the cause of deaths by doctors compulsory in England and Wales

London

:

Measurement of Maternal Mortality

321

TABLE 2

The Effect of Differences in Methodology on the Reported Maternal Mortality Rates (MMR) in Certain Countries, 1927 Maternal deaths per 10,000 births"

Country USA

Scodand Belgium Australia New Zealand England and Wales Norway The Netherlands Sweden Denmark

b

19io

1920

1927 (A)' 1927 (Bf

69 57

78 62

65 64

52

61

61

51

59 49

27

50 65 44 35

25

24

25

45 36

65 64 —

41

59 59 48

24

28

29 28

30

26 24.

31

20

32

"To nearest whole number. TData for 1910 and 1920 are included to demonstrate the constancy of the rank order. c Column 1927 (A) shows the actual MMR published by each country in 1927. In column 1927 (B) the values represent the mortality rate which would have been recorded in each country if the method of assigning maternal deaths which was used in the U.S. had been applied uniformly. Source: The published vital statistics of various countries and Children's Bureau Publications, U.S. Department of Labor, Washington, D.C.: Grace Meigs, Maternal Mortalityfrom All Conditions Connected with Childbirth in the United States and Certain Other Countries. Children's Bureau Publications, no. 19 (1917); R. M. Woodbury, Maternal Mortality: The Risk of Death in Childbirth and from All the Diseases Caused by Pregnancy and Confinement, ibid., no. 152 (1926); Elizabeth C. Tandy, Comparability of Maternal Mortality Rates in the United States and Certain Foreign Countries, ibid., no. 229 (1935).

and allowed a longer time for the registration of births, but this made little difference to the registration of maternal deaths. The introduction ofvarious editions of the International Classification of Diseases (ICD) was more important. Britain adopted this system in 1911 and it had one important effect. Before 1911 deaths due to toxemia of pregnancy (but not deaths due to eclampsia) were allocated to diseases of the kidney. They are lost to posterity in a "dustbin" category which contained an undifferentiated collection of actual and

322 Journal of the History of Medicine :

The Periods in which Mates were admitted to (he death registration area of the USA States admitted before 1000 States admitted between 1900 and 1915 I

I States admitted between 1916 and 1923

|

| States admitted between 1924 and 1933

Vol. 54, April iggg

i Miss. ! Conn. >NJ • Del. I Md. . W. Vi. :DC

Fig. 5. The periods in which states were admitted to the United States Death Registration Area. Source: Maternal Mortality in Fifteen States, Children's Bureau Publication, no. 213, U.S. Department of Labor (Washington, D.C.: Government Printing Office, 1934).

supposed renal disease. From 1911 deaths from toxemia were placed where they belonged, in with the true or direct maternal deaths. But, again, the addition made relatively little difference.7 In the United States the situation was confused by the slow introduction of the death (and the birth) registration areas to which states were admitted when their statistics were judged to have reached a high degree of reliability (see Fig. 5). Nine states (or large cities) were admitted before 1900, mostly in New England. Between 1900 and 1915 another 18 states were admitted, and the total of 27 states, known as the 1915 death registration area, was often used for national statistics. Constructing the time trend of maternal mortality in the 7. Irvine Loudon, Death in Childbirth: An International Study of Maternal Care and Maternal

Mortality 1800-1950 (Oxford: Oxford University Press, 1992), pp. 19-39, discusses at greater length the points raised in this paper.

London

:

Measurement of Maternal Mortality

323

Miunul Deaths per 10.000 Births 50 or more

I*^M 30-9

Fig. 6. Interstate variations in maternal mortality rates in terms of four categories of maternal mortality, 1938 to 1940. Source: Births, Infant Mortality, and Maternal Mortality, Children's Bureau Publication, no. 288, U.S. Dept. of Labor (Washington, D.C.: Government Printing Office, 1945).

United States between, say, 1915 and 1935 was tricky. Some used the 1915 death registration area and ignored later additions in order to achieve consistency. The trouble with this approach was that the states added last were mostly in the South which were precisely the states with the highest level of maternal mortality (see Figs. 5 and 6). Leaving them out made the national level of maternal mortality lower than it really was. The complications of assessing American national trends against the background of two moving targets, the death and birth registration areas, are dealt with admirably by Robert Morse Woodbury in Maternal Mortality.9 One of his important conclusions was that although maternal deaths were under-registered, there was a corresponding under-registration of births. The two faults tended to cancel each other out in the calculation of maternal mortality. 8. Woodbury, (n. 3) Maternal Mortality, pp. 45-56.

324 Journal of the History of Medicine :

Vol. 54, April iggg

TABLE 3

The Range of Random Variation (Critical Range of Variation) in Maternal Mortality Rates in Relation to the Number of Births, assuming a Rate of Maternal Mortality of 50 Maternal Deaths per 10,000 Births Assumed Annual births 1000 5000

Maternal deaths

rate of

Critical

maternal mortality

range of

5

5O

25

5O

10,000

50

50

25,000

125

50

50,000

250

5O

100,000

500

50

variation

Regions with annual births close to these values during the lgjos

Doncaster, Darlington Bristol, Newcastle upon Tyne 35-65 Manchester, Amsterdam 40-60 Lancashire, Middlesex, New Zealand 43-57 Paris 45-55 Sweden, Belgium, Australia 5-95

30-70

The critical range of variation shows the values for the maternal mortality that could occur through chance if the true value were 50 per 10,000 births. Thus, if it had been possible to show over a large number of years that the true maternal mortality rate in Doncaster was exactly 50 per 10,000 births, in any given year the actual rate could vary by chance alone from 5 per 10,000 to 95 per 10,000 because the total annual number of births was only 1,000.

One final point will be so obvious to demographers that it scarcely needs to be made. In the study of maternal mortality, one needs accurate records of a large number of deliveries to achieve statistical significance. This is illustrated by Table 3. HIDDEN MATERNAL DEATHS

The worst problem in assessing maternal deaths is that of "hidden maternal deaths."9 The accuracy of vital statistics is determined by the quality of those who certify causes of death. The "old women searchers" and the London Bills of Mortality is an example. As for doctors in the nineteenth and twentieth centuries, there are in medicine as in every profession ignorant, careless, and dishonest practitioners as well as well-informed, careful, and honest ones. The 9. Loudon, (n. 7) Death in Childbirth, pp. 519-27.

London : Measurement of Maternal Mortality

325

former played a significant role in the statistics of maternal mortality by filling out death certificates in such a way that maternal deaths were allocated to nonmaternal categories, creating the "missing" or, as I prefer to call them, the "hidden" maternal deaths. Sometimes this was mere carelessness; more often, I suspect, it was deliberate. The hidden deaths were almost always deaths from puerperal fever for the simple reason that a doctor (or midwife) whose patient died of puerperal fever was liable to be blamed for the death whether justified or not. No doctors certified a death as due to puerperal fever unless they felt compelled to do so, and many submitted to the temptation of cover-up. As a result deaths from puerperal fever were never exaggerated; they were virtually always an underestimate. Hidden maternal deaths were created by doctors developing certain well-recognized strategies while ensuring they could not be accused of gross lies and distortion, for death certificates were seen by relatives of the deceased and also by the local Registrar of Births, Deaths and Marriages who would detect a gross lie. One way to hide a maternal death due to puerperal fever was to invoke multiple causes and relegate puerperal fever to a secondary position. For example, if a mother had a slight postpartum hemorrhage and died a week later of puerperal fever, "hemorrhage" could be put as the primary cause on the death certificate, and "puerperal fever" as the secondary cause. Only the primary cause appeared in the published statistics, and the death remained in the broad category of maternal mortality. The mode of death in puerperal fever was either peritonitis or septicemia or both. Indeed, puerperal peritonitis and puerperal septicemia were both common synonyms for puerperal fever. Another means of hiding maternal deaths due to puerperal fever was to list "peritonitis" or "septicaemia" alone, with no mention of childbirth. The cause of death was, strictly speaking, correct but incomplete. The deaths so recorded promptly disappeared into one of two "dustbin" categories in the vital statistics: "peritonitis of unstated origin" or unspecified "septicemia." This same camouflage was practiced in the United States and was referred to by those two eminent analysts of maternal mortality, Grace Meigs (1917) and Robert Morse Woodbury (1926).10 There were also other ways to hide maternal 10. Grace Meigs, Maternal Mortality from all Conditions Connected with Childbirth in the United States and Certain Other Countries, Children's Bureau Publications, no. 19., U.S. Dept. of Labor (Washington, D.C.: Government Printing Office, 1917); Woodbury, (n. 3) Maternal Mortality, pp. 7-16.

326 Journal of the History of Medicine :

Vol. 54, April iggg

deaths. In the 1890s a Midwestern American doctor noted wryly that it was amazing how many women appeared to have died from malaria just after having a baby.11 That such strategies accounted for the large majority of hidden maternal deaths was shown by William Farr and his successors at the General Register Office in London.12 Like"detectives, they investigated a large sample of women of childbearing age who had died but whose death certificates did not indicate childbearing. They uncovered some hidden maternal deaths, mostly deaths recorded as peritonitis. But shortage of staff made it impossible to investigate all the deaths in which such a cover-up had occurred. For this reason, I decided to try to quantify these hidden deaths, starting with peritonitis. The favorite hiding place for hidden maternal deaths, "Peritonitis of unstated origin," was a large but rapidly diminishing category of death. It contained (in round numbers) 23,000 deaths in the 1890s, 3,800 in the 1920s, and only 1,600 in the 1940s. At first sight it looks as if peritonitis was a deadly nineteenth-century disease which gradually disappeared, but this is not so, and the explanation is simple. Although appendicitis and duodenal ulcer were almost certainly common diseases in the nineteenth century, they only became officially recognized as registerable causes of death in 1902 and 1911, respectively. Both diseases could cause peritonitis. Having no other place to go, they were placed in "peritonitis of unstated origin" before 1902 and 1911, swelling that category to a great size. After those dates they were provided with a classification of their own. But for several years deaths from these surgical diseases continued to gravitate into "peritonitis" because doctors are slow to change their ways. Thus the number of deaths in the category "peritonitis of uncertain origin" gradually withered away, while those in the categories "appendicitis" and "duodenal ulcer" increased. The apparent changes in incidence were, in fact, changes in the categorization of causes of death. For those with an interest in the mortalityfromsurgical diseases, it would be easy to fall into the trap of believing that appendicitis and duodenal ulcer suddenly increased in the early years of this century. 11. C.S. Bacon, "The mortality from puerperal infection in Chicago," Am. Gynecol. Obstet.J., 1896, 8, 426-46, pp. 430-31. 12. Supplement to the Annual Report of the Registrar General (London: Her Majesty's Stationery Office, 1895). pp. xxii-xxviii.

London : Measurement of Maternal Mortality

327

What does all this have to do with hidden maternal deaths? As it happens, the causes of death from peritonitis are virtually all more common in males than females. Therefore, if one expresses female deaths from peritonitis of unstated origin as a percentage of male deaths, the result should always be less than 100. If, however, maternal deaths were hidden in this category, female deaths should increase proportionately in the ages of childbearing, even to the point where they exceed male deaths. This is just what I found (Table 4). (The features shown in Table 4, incidentally, were found in every decade from the 1880s to the 1940s.) Further, as Table 4 shows, the highest percentage of female deaths occurred in the maximum age of childbearing, 25—35 years, as one would expect if they were maternal deaths in disguise. Above 45-50 years, after childbearing, male deaths outnumbered female deaths. As the accuracy of death registration improved steadily through the twentieth century, one would expect the excess of female deaths in the "peritonitis" category to decrease in number, as it did. Female deaths due to peritonitis virtually disappeared after 1950. I carried out a similar exercise with the category "septicemia" (the common causes of which were much more common in males than females), but I found little evidence of a substantial number of hidden deaths. Returning to peritonitis, if one assumes that' the excess of female deaths over male in ages 15-44 were hidden deaths due to puerperal fever, it is easy to quantify them. The number of hidden maternal deaths was greatest around 1880-1900, and then decreased. Although I have not uncovered all the hidden maternal deaths, the work I have done suggests that although some maternal deaths were hidden, they were too few to cause serious distortion. Further, such distortion as there was diminished throughout the first half of this century. At most, if all hidden maternal deaths had been detected and added to the published figures, the MMR would have been raised by about 10 percent in the late nineteenth century and considerably less by the 1920s. There is some indirect evidence that careless and deceitful miscertification caused relatively little distortion of the published rates of maternal mortality. In Switzerland, death certification was carried out by specially trained public health officers who had no clinical responsibility for mothers who died in childbirth and therefore had no reason to distort the data. If miscertification had seriously understated

oo

79 86 88 98 96 86 74 65 177

87

196

5O-S4 55-59 60-64

65+

in

56 78

72

78

90

165

68 68 203

43 56

7i

67 74

48

47

no

50

39

46

82 81

68 63

106

100

27 28 104

33

75

17 26

12

72

36 63 93

57 H3 23

300

8 14 16

18

6

200

450

100

90 62

100

277

10

400 219

36

9 5 4 9

10

25

290 29

% 86 58

5

2

120

414

12

29

171

39

21

204 136

9

7

10

10

8 4

20

10

336 50 9

85

17

23 20

70 28 304 140

59 144

64

54

108

100

797

795

226

141

152

237 92

189 82 300 630

M

The third columns in each period show the excess of female deaths as a percentage of male deaths, with the excess being most prominent in the age of childbearing, especially the age groups 25—34. The excess of female deaths in children under the age of 15 was due to a condition which has now virtually disappeared, "primary pneumococcal peritonitis" which occurred in female but not in male children before the onset of puberty. It was most common between the ages of 3 and 13.

45

3O-34 35-39 40-44

38 43

61

133

56

10-14 15-19 20-24 25-29

5-9

40

387 542

129

59

2137 125

%

F

M

F

%

F

1128 152

M

86 56 64

Total Under i 1-4

Ages

1940-1950

1941-1950

1931-mo

Deaths Recorded in the Reports of the Registrar General as "Peritonitis of Unstated Origin"; Total Deaths for Both Sexes in Certain Age-Groups, and Female Deaths Expressed as a Percentage of Male Deaths, England and Wales

TABLE 4

London : Measurement of Maternal Mortality

3 29

maternal deaths in other countries, one would have found the Swiss figures to be unexpectedly high. They were not. They were of the same order as those in France, Germany, England, and Wales. What do we learn from all this? You may remember Sam Weller's story about the charity school boy who, having learned the alphabet, expressed serious doubts whether it was worth going to so much trouble to learn so little.13 The same might be felt about this catalogue of statistical complications and confounding factors. Is their analysis worth the trouble? The answer must be that of course it is. The only available measure of the effectiveness of maternal care in the past is maternal mortality. Comparisons of maternal care between countries and regions, between doctors and midwives, or home and hospital deliveries, and between different obstetrical techniques demand an appreciation of the reliability and the potential sources of error in the published statistics of maternal mortality. Without such information one is left with unsubstantiated assertions based on anecdotal evidence. For demographers, however, I am preaching to the converted.

13. Charles Dickens, The Pickwick Papers (first published London, 1837).