Vous consultezEstimating the Number of Abortions in France, 1976-2002

AuteursClémentine Rossier du même auteur

Claudine Pirus[*] [*] Institut national d’études démographiques, France
Translated...
suitedu même auteur

I - Three sources of data on abortions in France: notification forms, SAE, PMSI

9 The statistical system based on notification forms, initiated in 1975, was fleshed out in the 1990s by two further sources of data on abortions: the annual statistics of healthcare facilities (SAE)[1] [1] The full names of all acronyms are given on the last page. ...
suite available in digital form since 1995, and the medical statistics database (PMSI) set up in 1996. The data content varies between sources, as do the types of bias encountered.

1 - Notification forms

10 The notification forms are by far the richest source of information about the women seeking abortions and the associated circumstances. They contain medical information on the pregnancy termination (date and place of operation, type of establishment, abortion technique, anaesthetic, complications, length of hospital stay, duration of pregnancy, elective or therapeutic abortion) and socio-demographic data about the patient (place of residence, age, nationality, place of birth, marital and cohabitation status, employment, partner’s employment, number, date and outcome of any previous pregnancies). The notification forms are distributed, collected and processed by the Ministry of Health. The health and welfare directorates (DDASS) in each département distribute the notification forms to the hospitals, the regional health and welfare directorates (DRASS) collect them, and the research, study, evaluation and statistics directorate (DREES) processes the data. Analysis and publication of the statistics is handled by INED, in liaison with INSERM. Publication of the notification form data was suspended in 1998, after an unsuccessful experiment with optical scanning (the 1998 notification forms were merely counted manually). The DREES has recently processed the 2002 and 2005 notification forms. A simplified version of the notification form, compatible with optical scanning, was introduced by DREES at the end of 2004[2] [2] DREES released a highly simplified version of the notification...
suite.

11 From the 1980s, INED questioned the exhaustiveness of the statistics, since the notification forms appeared to under-estimate the number of abortions in France (Blayo, 1985, 1995, 1997). The causes of this under-recording in the notification forms were perceived as being of two types. First, some abortions were apparently not being recorded in the notification forms for various reasons (omission, disorganization, no stock of blank notification forms, lack of information, etc.). These abortions were properly recorded, it would appear, in the other information systems (hospital and social security statistics) but not on notification forms. Chantal Blayo estimated that this cause of under-recording amounted to 11% to 14%, depending on the year (1995, p. 781). She measured the extent of these omissions by comparing the notification form figures with those from hospital surveys (the forerunners of the current annual statistics of healthcare facilities) supplied by the statistical studies and information systems service (SESI, which became the DREES in 1998). In 1992, for example, some 27,000 abortions were apparently recorded in hospital surveys but not on the notification forms.

12 A second reason for the under-recording of abortions affected all the information systems (notification forms, hospital statistics and social security records). On the basis of a number of field surveys, C. Blayo and her colleagues were able to show that certain practitioners in the private sector recorded induced abortions as spontaneous abortions, so as to bill them at a higher rate or to perform more abortions than allowed under the legal quota. C. Blayo estimated that this infringement of the abortion regulations concealed some 32,000 abortions in the various data sources in 1992.

2 - Annual statistics of healthcare facilities (SAE)

13 SESI estimated the number of abortions on the basis of annual hospital surveys (estimates published in Blayo, 1995, for years 1986-1992). For each hospital, the number of notification forms collected was compared with the number of abortions indicated by the hospital surveys, and the higher figure taken. The purpose of these annual hospital surveys was to provide information about the structure of hospital care available in each geographical area. The surveys collected data on various types of hospital stay (short, medium, long), certain activities (pregnancy terminations, emergencies, surgical procedures), staff, equipment, etc. For abortions, only the total number of elective and medical or therapeutic abortions was requested, and the total number of procedures per technique (medical or surgical), the total number of terminations under anaesthetic, and the total number of abortions undergone by minors. The SAE was completed once a year by hospital management (director or secretary), usually on the basis of the hospital’s admissions records.

14 DREES considers that the list of hospitals supplying annual statistics is complete. However, some hospitals (less than 2% of active units) do not answer the survey, and others (a larger proportion) do not complete all parts of the questionnaire. DREES and DRASS issue reminders to the hospitals that have not replied, but not to those whose information is incomplete. The hospital annual statistics have been used for many years, with two major changes: computerization in 1995 and a new questionnaire in 2000. Statistics for these two years are therefore considered to be less exhaustive than the others. DREES also considers that the hospital annual statistics did not become more exhaustive in the 1990s, unlike the PMSI data.

3 - The medical statistics database (PMSI)

15 The aim of the PMSI is to inventory all medical acts in all hospitals and their costs, in order to determine the hospitals’ expenditure profiles, decide on their budgets (for public and semi-private hospitals) and monitor the prices charged. The PMSI database contains information recorded on each patient at the end of their stay by the doctor or department secretary on a form, or entered directly via special software. Medical acts are described and coded using the classification of diseases; the coding is verified by another designated doctor. The data are then forwarded to the hospital accounts department which bills all medical acts. Data on elective abortions contained in the PMSI are therefore individual, like the notification form data, but unlike the hospital annual statistics data. They include selected information about the women who have had an abortion (place of residence and age) and the act itself (length of hospital stay, technique, anaesthetic).

16 DREES (Buisson et al., 2003, p. 3) estimates that the PMSI measures 96% of medical, surgical and obstetrical activity. This high figure dates from 1998, since private hospitals (with more than 100 beds) only joined the PMSI in February 1997 (Buisson, 2005). Furthermore, the PMSI did not distinguish between therapeutic and elective abortions until 2001; a finer coding system was introduced in that year, but has not yet been adopted by all hospitals[3] [3] Before 2001, the PMSI had the following codes for recording...
suite. However, it is easy to distinguish spontaneous abortions (hydatidiform moles, other abnormal products, spontaneous abortions) from induced abortions (therapeutic, elective) in the PMSI.

17 Although the PMSI coverage is less exhaustive than the SAE, the PMSI data are more precise than the those of the SAE, despite coding problems, because of fee-for-service billing and the double-check by another physician in each hospital. DREES does not systematically compare the number of abortions recorded for each hospital in the two sources; but it does calculate an overall correction (by hospital size) between the two sources on the basis of geographical area.

4 - Substantial differences between sources

18 How has the number of elective abortions in metropolitan France varied since 1975? Although the question is simple, there is no direct answer. Comparison of the various abortion data sources reveals a number of discrepancies (Table 1 and Figure 1). The number of abortions recorded in the notification forms increased during the early years of this statistical source to nearly 183,000 induced abortions in 1983. The figure fell slightly afterwards (162,000 notification forms in 1987), then stayed fairly stationary until 1997, when 164,000 abortions were recorded[4] [4] The number of notifications recorded for 1995 and 2002 was...
suite. These figures fall short of reality, as we have seen. According to INED estimates, there were 250,000 abortions in 1976 (this is also the INED estimate for 1966), 262,000 in 1980, 230,000 in 1987 and 225,000 in 1993. An updated estimate by France Prioux (2002) provides a figure of 220,000 abortions in France in 1997. However, according to the SAE figures, there were approximately 180,500 abortions in 1994 and 206,600 in 2002: the number apparently increased slightly in the late 1990s (Le Corre and Thomson, 2000; Vilain and Mouquet, 2003; Vilain, 2004; Vilain, 2005).

Abortions in metropolitan France estimated from various sources, 1976-2002

Abortions in metropolitan France estimated from various sources, 1976-2002

Table 1 - Number of abortions in metropolitan France by source of information, 1976-2002

19 Comparisons between these different data sources raise two questions. First, the method used by INED to estimate the number of abortions in France from 1976 to 1997 has never been published, and its underlying hypotheses have not been discussed. The presence of new abortion data sources provides an opportunity to reassess these earlier estimates, because the INED figure for 1997 (220,000) does not coincide with the SAE figure for that year (189,000). Are the INED figures overestimates? And, whatever the precise estimated volume of abortions, are the trends indicated by the estimates reliable? In other words, did the number of abortions really fall in the early 1980s (Blayo, 1995 and 1997; Prioux, 2002)? Or has the number of abortions been stable for thirty years (Bajos et al., 2004)?

20 The second question raised by the comparison of different abortion data sources is the exhaustiveness of hospital data, a factor which has never been evaluated. Do the difficulties in collecting abortion notification forms also affect the collection of hospital data? Why did the number of abortions recorded in the hospital statistics rise in the second half of the 1990s? Was this a real rise or a gradual disappearance of the reasons for under-recording of abortions?

21 In addition to the sources already cited (notification forms, SAE, PMSI) and the publications using those sources, our work is based on interviews conducted from January to April 2005 in 27 institutions dealing with abortions (9 hospitals, 8 clinics, 4 family planning centres, 3 DRASSs and 3 DDASSs) in three French regions[5] [5] We chose to survey two of the three regions considered by...
suite (Table 2). The interviews in hospitals and clinics concerned the ways in which the various abortion statistics are recorded (notification forms, SAE, PMSI, social security forms). In each establishment visited, we interviewed, individually or in groups, all staff involved in recording the statistics: physicians, midwives, nurses, accounts clerks, receptionists, secretaries, hospital director, etc. In the family planning centres, most of the information gathered concerned the organization of centres performing abortions in the region. In addition to these field interviews, we also questioned relevant staff in the DDASSs and DRASSs of the regions concerned.

Table 2 - Abortion professionals surveyed

II - Abortion estimates published by INED from 1976 to 1997

22 Let us re-examine the abortion estimates produced by INED for 1976-1997. Since the method used for estimating abortion statistics was not published, a number of attempts were needed to reproduce the results. An initial attempt by spatial smoothing (increasing the ratio of abortions to births in départements where the ratio was lower than in adjacent départements) was unsuccessful: the figures obtained were noticeably lower than C. Blayo’s published estimates. This method uses the hypothesis of a relatively uniform abortion frequency throughout France. However, it is difficult to defend this hypothesis: the 2002 SAE data show a ratio of 23 induced abortions (therapeutic and elective) per 100 births in Mayenne département, and 101 abortions per 100 births in Corsica. Adjacent départements may have sharply differing abortion rates.

23 The second attempt was more successful. We selected for each year those départements with the highest abortion frequency (ratio of abortions recorded in notification forms to the number of births), calculated the average abortion frequency for that subgroup of départements, and applied that average to France as a whole. But what should that subgroup include? First we selected for each year those départements with an abortion ratio above 27 per 100 births, since in her research, C. Blayo often divided abortion frequencies into four classes, and 27 per 100 births was the lower limit of her highest class. To illustrate this method, let us take 1986, when 10 départements had a ratio above 27 abortions per 100 births; their average ratio was 31.1 per 100 births. Applying that figure to the total number of births in France that year (776,714), we obtain 241,464 induced abortions, close to C. Blayo’s estimate (239,000). We did the same calculation for 1976, 1981, 1986, 1991 and 1996 (Figure 2, Hypothesis 1) and each time came very close to Blayo’s estimates.

24 The number of départements above the 27 abortions per 100 births threshold varies from one year to another: there were 6 départements in this subgroup in 1976, 14 in 1981, 10 in 1986, 12 in 1991 and 14 in 1996. To simplify this method of estimation we took for each year the top 5, 7, 10, 13 and 15 départements and applied the subgroup average ratio to the number of births in metropolitan France. Although the trend in the estimated number of abortions is always the same, whatever the threshold (decline in the early 1980s, relative stability after 1986), the level of the estimates varies considerably according to the threshold. We approximate the level of C. Blayo’s estimates by applying to France as a whole the abortion ratios of the 13 départements with the highest abortion frequencies recorded in notification forms (Figure 2, Hypothesis 2).

Total abortions in metropolitan France estimated from notification forms, 1976-1996

Total abortions in metropolitan France estimated from notification forms, 1976-1996

25 It is clearly hazardous to use this principle to estimate the actual level of induced abortions in France from 1976 to 1996, since the figures vary according to the size of subgroup and any choice is arbitrary. But is the principle valid for estimating the abortion trend in France? In other words, is it legitimate to suppose that the abortion trends in the 5, 10 or 15 top départements with the highest abortion frequencies recorded in notification forms can be extrapolated to France as a whole? To settle the matter, we applied this method to the number of births in France, i.e. we estimated the total number of births in metropolitan France by using the crude birth rate of the 13 départements with the highest birth rates, for 1976, 1981, 1986, 1991 and 1996. As Figure 3 shows, although our method does not provide the actual level of births in France, it does give an accurate estimate of the trend over that period.

26 The trend suggested by these estimates (decline in abortions in the early 1980s, then stabilization) appears plausible when compared with the data on unplanned births. The proportion of unplanned births declined until the mid-1980s and then levelled out (Régnier-Lollier, 2005). The overlap of these two trends in time appears to indicate a fall in the number of unplanned pregnancies until the latter half of the 1980s (since an unplanned pregnancy leads either to a birth or an induced abortion). Successive surveys on contraception in France do indeed reveal that the transition from “natural” contraceptive methods to “medical” ones, which occurred progressively from 1970 to 1985, was complete by the late 1980s (Toulemon and Leridon, 1991; Rossier and Leridon, 2004): these improvements in prevention caused a decline in the total number of unplanned pregnancies over this period, and consequently fewer unplanned births and abortions.

Births in metropolitan France, 1976-1996

Births in metropolitan France, 1976-1996

27 For these reasons we conclude that although INED’s published estimates of the number of abortions in France from 1976 to 1997 do not provide very accurate numbers, they do indicate a genuine trend: a decline in the number of abortions in the early 1980s, and a stabilization from the mid-1980s to the mid-1990s.

III - Missing notification forms deduced from hospital statistics

28 We now examine the number of abortions in France from the mid-1990s. To address this question, we have various sources of information: hospital data (SAE and PMSI), which show a slight increase in abortions from 1995 to 2002, and the notification forms, which record significantly fewer abortions than the hospital statistics, and cannot be used to track a trend over time, since we only have data for 1996, 1997 and 2002, and the latter year was affected by collection problems. We examine first the difference between the notification forms and the hospital statistics, and then consider possible causes of under-recording that might affect all abortion figures, including hospital statistics.

29 In 1995, there were 23,500 fewer abortions recorded in the notification forms than in the hospital statistics, and in 2002, the shortfall was 69,000. How can this be explained? Evidently, there is no intention of concealing abortions, since they are recorded in the hospital statistics: this under-recording in the notification forms must be simply attributable to negligence. Private-sector practitioners are the main culprits, as shown first by C. Blayo’s field observations (1995), and then by a statistical comparison of notification forms and hospital sources (Vilain, 2004). Table 3 confirms this analysis: from 1994 to 1998, 3% to 5% of abortions performed in public establishments were not recorded in the notification forms, compared with 25% to 34% of abortions performed in private establishments. In 2002, the shortfall was larger in both sectors: 21% of abortions in the public sector recorded in hospital statistics were not recorded in notification forms, compared with 67% in the private sector.

Table 3 - Relative deviation between number of abortions recorded in SAE and notification forms [(SAE – forms)/SAE] from 1994 to 2002 in metropolitan France by type of establishment (%)

30 Interviews with medical staff in 2005 confirmed these results. In private establishments where the notification forms are completed, the form is usually filled out by the doctor alone, sometimes with the help of the patient; secretariats do not check the number of notification forms completed and simply forward them to the DRASS. Some private-sector practitioners we contacted have never heard of the notification form or have deliberately stopped filling in forms they consider to be “too much trouble” – note that they have sole responsibility for this task. In public establishments, however, the forms are completed either by nurses or by paramedical staff (counsellors, secretaries), with occasional help from the physician. In most cases, the secretariat also checks the number of notification forms against the number of admissions recorded before forwarding them to the DRASS. In some cases, organizational problems in public establishments (e.g. absence of a particular member of staff) temporarily disrupt the completion of notification forms.

31 Medical staff are currently divided as to the value of statistics introduced thirty years ago when the context was quite different. Some practitioners (in both private and public sectors) believe that it is still important to have information on the profiles of women who abort; most of them also consider that the new notification form introduced by the DREES in 2005 records insufficient information, and argue that the hospital statistics only give the crude numbers of abortions and a few medical details. Some supporters of notification forms nevertheless suggest that these exhaustive but cumbersome statistics should be replaced by periodical surveys. Other practitioners consider that the notification form statistics are no longer necessary. Some, especially in the private sector, find completion, even of the simplified form, too time-consuming; others say they do not want to complete forms for statistics that have not been published for nearly ten years. Some argue against the notification forms on ideological grounds: would it not be sufficient to record the basic data contained in the hospital statistics, now that the profile of women seeking abortion is known and does not greatly vary? Things that were important when abortion was liberalized are perhaps no longer so: is it not time to treat termination as a medical practice like any other?

32 Practitioners’ willingness to complete the notification forms also depends on the collection system used by the DRASS in their region, which itself depends on the number of abortions and their breakdown between public and private sectors. For example, the Île-de-France DRASS (Paris region) checks the exhaustiveness and quality of notification forms received, sends statistics based on the notification forms to the DDASSs, but does not issue reminders to establishments to improve the submission rate. This region faces a high demand for abortions and relies heavily on the private sector, so pressurizing practitioners to return notification forms would be ill-advised in such a context. The Midi-Pyrénées DRASS (Toulouse) adopts a different policy. It claims to have reduced the rate of non-return of notification forms in recent years by deliberately sending reminders to establishments. Note that demand for abortions is lower in Midi-Pyrénées than in Île-de-France, making it possible to monitor notification form returns more closely. But the regional directorate has no powers to compel poorly responding establishments (many of them private in this region). The Nord-Pas de Calais DRASS (Lille) is even more enterprising: it regularly publishes a statistical study of regional abortion figures and sends it to establishments. This encourages them to return their notification forms, to a degree the directorate considers exhaustive. Its task is simplified by the fact that the vast majority of abortions are performed in the public sector (96% in 2002, according to the SAE).

IV - Elective abortions not recorded in any information system

33 C. Blayo (1995) estimates that a number of abortions are not recorded either in notification forms or hospital statistics. They are probably recorded as spontaneous abortions. Until 2001, a legal limitation[6] [6] Article L178 1 of Act 75-17 of 17 January 1975 relating...
suite on the number of abortions performed in private establishments provided an incentive for some doctors not to record abortions above that number. The Nisand report (1999, pp. 9-10) mentioned this “quota” as one reason for the under-recording of abortions. The private practitioners we interviewed were unanimous on this point, however: although the legal quota may have been a problem for professionals immediately after legalization, when there were not yet many abortion services, this limit has not apparently held back their activity in recent years.

34

35 Doctors had another reason for recording abortions as other medical procedures, as pointed out in the Nisand (1999) and IGAS (2002) reports: the official tariff was too low. Unlike other medical procedures, for which private doctors have a certain latitude in their choice of procedure and fees, the fees for abortion procedures, covered by the social security system since 1982, are fixed in advance by a series of regulations. Our respondents all considered that the existing set fee did not enable private sector practitioners to achieve a return on investment; they even claimed to make a loss. The latest tariffs (set in 1991) were revised in 2004[7] [7] Detailed tariffs were published in the decree of 23 July...
suite, but the increase was considered insufficient.

36 Given the low tariff for abortions, private sector practitioners have reacted in a variety of ways. The simplest response is to refuse to perform the procedure. A gynaecologist working in Île-de-France explained that some clinics refused to engage in this unprofitable business:

37

38 Another alternative is to perform abortions without making a profit. This is the case for doctors who perform the procedure as a matter of principle, and those for whom abortions are only a marginal activity.

39

40 A third possibility is to attempt to make a profit within the legal tariff. Some doctors opt for the highest set fees, which include a longer hospital stay or general (rather than local) anaesthetic, and may prescribe a scan at their establishment. Since there is no set fee for the first consultation, some doctors have no qualms about charging a high price. Longer hospital stays and general anaesthetic are justifiable for surgical abortions, but are not necessary for most medical ones. Medical techniques may therefore be less attractive for some private practitioners. Using an analysis of notification forms, Kaminski and Crost (1997) showed that surgical abortions are performed more frequently under general anaesthetic in the private sector than in the public sector. Similarly, Jones and Henshaw (2002), comparing the development of medical abortion in France, Great Britain and Sweden, report that France is characterized by a slower development of this technique in the private sector. These two findings may be attributed, at least in part, to the attempts made by private sector doctors to balance their budgets or achieve a profit. In the following excerpt from an interview, a family planning counsellor in Île-de-France describes the wide variety of abortion tariffs.

41

42 The last reaction is to record the abortion as some other procedure that can be billed at a higher tariff. Most of our respondents mentioned the “K30” ploy, already reported by C. Blayo in the early 1990s[8] [8] C. Blayo estimates that this ploy accounted for some 10%...
suite, referring to the social security code for uterine evacuation, the procedure for incomplete spontaneous abortions[9] [9] Until the contents of the uterus are entirely evacuated,...
suite. Today, the advantage of recording an induced abortion as a spontaneous abortion is not so much to be eligible for a higher rate of social security reimbursement, but rather to be able to charge extra fees.

43

44 However, none of our respondents stated that they themselves used this ploy to charge more than the legal set tariffs.

1 - Spontaneous abortion statistics

45 Examination of the statistics for miscarriages and other spontaneous abortions may provide some enlightenment here. The term “spontaneous abortion” covers miscarriage, ectopic pregnancy and other abnormal products of conception. Spontaneous abortions requiring hospital admission are recorded in the PMSI and we have data for 1998 to 2002[10] [10] The PMSI is reliable from 1998, since 1996 and 1997 were...
suite. The trend over France as a whole was a slight reduction in spontaneous abortions treated in hospitals (–8.4% from 1998 to 2000), but in the private sector only (Table 4).

Table 4 - Spontaneous and induced abortions in metropolitan France, 1998-2002

46 An initial explanation for this trend might be the development of medical abortion techniques, also used for the outpatient treatment of spontaneous abortions[11] [11] Dr Danielle Hassoun, Saint-Denis hospital, and Dr Véronique...
suite, which reduces the number of hospital admissions. However, this would not explain why the reduction was restricted to the private sector; moreover, the PMSI also records short hospital stays.

47 Another explanation is possible: the lower number of spontaneous abortions treated in the private sector in recent years might be attributable to a lower number of induced abortions “wrongly” recorded as spontaneous. The rise in the number of induced abortions recorded in the SAE from 1998 to 2002 (+6,759) is broadly the same as the fall in the number of spontaneous abortions treated in hospital (–8,065). In addition, the rise in the number of induced abortions recorded in the SAE since the mid-1990s has been concentrated in the public sector, whereas the reduction in spontaneous abortions treated in hospital, as stated above, concerns the private sector only (Table 4). These divergent trends may conceal stability in the two phenomena. One may suppose that practitioners recording induced abortions as spontaneous abortions gradually stopped performing abortions, and that the demand they had previously met was absorbed by the public sector and became visible in the statistics.

2 - An estimate of induced abortions in 1998 and 2002 based on spontaneous abortions

48 To calculate the magnitude of possible false recording of induced abortions in 2002, we estimated the number of induced abortions (elective and therapeutic) in metropolitan France in 2002 on the basis of statistics for spontaneous abortions. The principle of estimation was as follows: we assumed that the recording of induced abortions as spontaneous abortions to charge more than the set tariffs was negligible in those départements where the vast majority of induced abortions were performed in the public sector. In other words, we assumed that the number of spontaneous abortions recorded in the PMSI was accurate for those départements, since it was not inflated by false recording of induced abortions. For 2002, the SAE indicate that in 35 départements, over 90% of induced abortions are performed in the public sector (Appendix Table).

49 The frequency of spontaneous abortion depends on biological factors (genetic anomalies, mother’s age, health and birth history, sex of foetus, multiple pregnancy, intra-uterine growth retardation), social factors (marital status of parents, cultural group, educational level, social class, occupation, exposure to toxic substances and work environment) and behavioural factors (smoking, alcohol) (Gourbin, 2002). Although biological factors appear to be predominant in first trimester foetal death, the weights of the various factors have not been evaluated for spontaneous abortions from the second trimester onwards. The frequency of spontaneous abortions is therefore likely to vary from one département to another. This is especially true since access to healthcare in the event of spontaneous abortion varies according to the structure of health services in the woman’s region of residence and to her social characteristics. For example, in the 35 départements whose statistics we deem reliable (i.e. where the public sector performs at least 90% of induced abortions), the ratio of spontaneous abortions to births varies from 9.0% in the Calvados to 23.9% in the Haute-Sâone, with an average of 11.8%[12] [12] For the 25 départements where 100% of induced abortions...
suite (see Appendix Table). We assume that this average is applicable to the whole of metropolitan France, i.e. that these 35 départements (with reliable spontaneous abortion statistics) are representative of the whole country in terms of frequency of spontaneous abortions.

50 We applied this ratio to the total number of births in the other 60 départements (Table 5) and obtained an estimate of 65,000 spontaneous abortions for these départements. In all, we estimate that there were 90,000 spontaneous abortions in France in 2002, which is close to the 88,000 actually recorded in the PMSI. It may be inferred that the doctors performing abortions were no longer under-recording abortions in 2002. The spontaneous abortions they record are most likely real spontaneous abortions; the hospital statistics therefore probably also provide an exhaustive record of induced abortions in 2002. This is not the case for 1998, when the estimated number of spontaneous abortions (88,000) was lower than the number recorded (96,000). One may suppose that the extra 8,000 spontaneous abortions recorded are actually disguised induced abortions. The real number of induced abortions would therefore be approximately 207,000 in 1998 instead of the 199,000 indicated by the SAE.

51 Although the hypotheses appear to be robust, this estimation method is nevertheless highly sensitive to the number of spontaneous abortions recorded in the départements we have taken as a reference. For 1998, for example, if 1% fewer spontaneous abortions had been recorded in the 24 départements where more than 90% of induced abortions were performed in public hospitals (in other words 170 fewer spontaneous abortions than the 16,966 observed), then the estimated number of induced abortions would rise to 207,700 (583 more); if 3% fewer spontaneous abortions had been recorded in these 24 départements (509 fewer), the estimated number of induced abortions would be 209,100 (over 2,000 more).

52 However, these estimates allow us to conclude that the number of spontaneous abortions treated in hospital most probably remained stable from 1998 to 2002, as did the number of induced abortions. Consequently, the number of induced abortions appears to have been stable at the end of the 1990s, rather than rising, as indicated in the hospital statistics. Our estimates tend therefore to show that the increase in the number of induced abortions in the hospital statistics at the end of the 1990s is a statistical artefact, because fewer induced abortions in the private sector were being recorded as spontaneous abortions.

53 This change may be explained by a number of factors: first, the hospital reform of 1996 led to the closure of many private clinics, and gynaecological services were concentrated in the largest hospitals. The strategy of declaring induced abortions as spontaneous abortions was possible in small structures with a small staff, who knew each other personally, but it became quite difficult in larger hospitals with many more staff. In addition, the creation of the PMSI in 1996 most probably contributed to the change, since this information system makes it possible (indeed is intended) to check the social security reimbursement applications from each hospital, and thus identify fraud.

Table 5 - Estimated number of induced abortions based on spontaneous abortion statistics in 1998 and 2002, metropolitan France

V - New estimates of induced abortion figures since 1975

54 We have seen that C. Blayo’s estimates appear to correctly reflect the trend in the number of abortions from 1976 to 1997. In addition, estimates based on spontaneous abortion statistics provide an approximate volume for induced abortions in 1998 and 2002. Consequently, the information required to make new estimates of the number of induced abortions in metropolitan France from 1976 to 2002 are available. To this end, we take the trends indicated by C. Blayo’s estimates and assume that the number of abortions was the same in 1996 as in 1998 (given the stability observed between 1998 and 2002). We set the number of abortions, therefore, at 205,000 in 1996, a reduction of some 20,000 in the number proposed for that year by C. Blayo’s method of estimation. We then subtract this same number (20,000) from the (recalculated) estimates made by C. Blayo for previous years, to obtain new estimates of the number of abortions in France in 1976, 1981, 1986 and 1991. Given the relative uncertainty in calculating the number of abortions from spontaneous abortion statistics, the number of abortions for 2000 could have been set at 200,000, 210,000 or even 215,000, requiring us to raise or lower the estimates for previous years.

New estimates of induced abortions in metropolitan France, 1976-2002

New estimates of induced abortions in metropolitan France, 1976-2002

Table 6 - Number of induced abortions and mean number of abortions per woman from 1976 to 2002, new INED estimates, metropolitan France

55 The results of these calculations are given in Table 6. To estimate the number of abortions in the intervening years, we simply interpolated from 1976, 1981, 1986, 1991 and 1996. We then calculated the age-specific abortion rate and, by addition, the total abortion rates (TAR), namely the average number of abortions a woman would have during her lifetime if she experienced the age-specific abortion rates for that year, a synthetic indicator of abortion intensity that does not depend on the population age structure.

56 It can be seen from these revised estimates that the total abortion rate fell in the early 1980s. Whereas in 1976 more than 6 out of 10 women would have had an abortion in their lives if they had experienced the abortion rate for that year, by 1988 they would have been 5 out of 10, a figure that remained steady until 2002.

Conclusion

57 Estimating the number of abortions where statistics are incomplete is a difficult exercise (Rossier, 2003). Rather than providing exact figures, the task is to examine various methods of estimation, reconsider their underlying hypotheses and assess their credibility with respect to the given context. We have attempted to gather enough clues to draw a number of lessons. Two basic conclusions emerge from this work: 1) the increase in the number of abortions measured by the statistics from 1995 to 2002 does not correspond to reality, but reflects the disappearance of a strategy for charging more than the legal tariffs for abortions by recording induced abortions as spontaneous abortions; 2) the frequency of induced abortions was higher in the early 1980s: the reduction recorded during this decade corresponds to the replacement of natural contraceptive methods by medical ones, which led to a decline in unintended pregnancies. Use of abortion has been stable since the end of the 1980s. It is therefore inaccurate to say that the frequency of abortion has not fallen in France in the last thirty years; rather it fell during the first fifteen years and has remained stable during the last fifteen.

58 We have seen that each of the current sources of induced abortion data is imperfect: the hospital statistics contain little information (total number, breakdown by technique and age) but at least the data are (now) exhaustive; the notification forms contain a large amount of information but only for a proportion of abortions, which nevertheless appears to be representative of the whole (Rossier, Confesson and Bringé, 2006).

59 Abortion remains a socially stigmatized practice (Boltanski, 2004), because it threatens the maternal component of women’s identity. Women’s empowerment involves access to unstigmatized abortion. But the collection of abortion data by the health authorities may well contribute to stigmatizing the women who have abortions. However, other medical practices are also recorded or notified in statistics, and abortion data are important for public health policy: they may be used to 1) measure the performance and the possible failings of sex education and birth control policies in different parts of the country and among more vulnerable sub-populations (young, foreign-born, deprived); 2) monitor equality of access to high-quality abortion services (technique, anaesthetic, waiting time) in different parts of the country and among various sub-populations. In all, we consider that the advantages of collecting information on induced abortion outweigh the disadvantages: the statistics help ensure access for all women to quality sex education, contraception and abortion services, so we believe it is useful to collect such data, despite the continued stigmatization such statistics may entail.

60 The next question is whether this information should be collected exhaustively or not. The variable success of collecting notification forms in the French regions shows that abortion services, overloaded with paperwork like rest of the French hospital system, can only continue to produce these statistics if the Ministry of Health (or its regional offices) makes a greater effort to process the forms rapidly and regularly communicate the results to practitioners. Exhaustive collection is time-consuming for practitioners and requires close involvement by the entire ministry to be operational. Wouldn’t periodical, shorter surveys be a more practical alternative? It does appear essential to make an exhaustive count of abortions, in order to compare France with other countries and thus assess the effectiveness of our birth control policy. However, exhaustive collection of data on the social characteristics of the women who have abortions (age, nationality, social background) and on abortion conditions (technique, anaesthetic, hospital stay, waiting time) is perhaps unnecessary: a survey of a sample of establishments would suffice. This sample would need to be representative of the various regions, because regional discrepancies in birth control policy and access to abortion are one of the essential dimensions of a national monitoring policy, and would thus have to be quite large. Replacing the notification forms by regular large-scale sample surveys would certainly lighten the task of practitioners, but it is highly likely that periodical surveys would be much more expensive and time-consuming for the Ministry of Health than the routine return of notification forms. The political will to support access to birth control and abortion may also waver according to government priorities, and expensive repeat surveys are more likely to be abandoned or postponed than the routine compulsory collection of data.

Annexe

Appendix

Induced abortions (elective/therapeutic) recorded in SAE, proportion of abortions performed in private sector according to SAE, spontaneous abortions recorded in the PMSI and births bydépartementin 2002

Appendix table - 

Ain (01) 0 782 866 3,768 Alpes-de-Haute-Provence (04) 0 216 430 1,130 Hautes-Alpes 05) 0 174 568 1,469 Ardèche (07) 0 341 542 3,483 Ardennes (08) 0 428 633 3,131 Aveyron (12) 0 255 612 2,583 Cher (18) 0 388 750 2,873 Côte-d’Or (21) 0 638 1,315 6,668 Creuse (23) 0 117 211 525 Indre-et-Loire (37) 0 813 1,343 6,938 Jura (39) 0 290 620 2,792 Landes (40) 0 397 594 2,522 Loir-et-Cher (41) 0 393 761 3,547 Haute-Loire (43) 0 233 306 1,126 Loiret (45) 0 776 1,950 8,005 Lozère (48) 0 105 127 1,492 Maine-et-Loire (49) 0 997 1,771 10,951 Haute-Marne (52) 0 259 475 2,109 Mayenne (53) 0 411 467 3,518 Haut-Rhin (68) 0 1,129 1,958 8,953 Haute-Saône (70) 0 365 432 1,525 Savoie (73) 0 586 1,323 4,768 Somme (80) 0 850 1,712 7,481 Vendée (85) 0 757 922 5,464 Territoire de Belfort (90) 0 266 407 2,148 Oise (60) 2 1,133 2,243 9,231 Aisne (02) 3 792 1,334 6,550 Nord (59) 3 3,893 8,504 39,342 Eure (27) 4 1,139 1,149 5,089 Pas-de-Calais (62) 6 2,300 3,292 17,550 Doubs (25) 7 756 1,579 7,506 Orne (61) 7 391 680 3,626 Puy-de-Dôme (63) 7 708 1,705 6,692 Saône-et-Loire (71) 7 670 1,281 5,688 Calvados (14) 8 806 1,852 9,004 Ariège (09) 10 173 455 1,142 Loire-Atlantique (44) 11 1,560 3,270 16,700 Morbihan (56) 12 821 1,666 8,017 Haute-Savoie (74) 12 853 2,504 9,386 Ille-et-Vilaine (35) 13 1,175 2,509 12,593 Manche (50) 13 593 839 5,419 Côtes-d’Armor (22) 14 601 1,271 6,144 Eure-et-Loir (28) 14 593 1,329 5,028 Haute-Vienne (87) 14 297 1,045 4,190 Indre (36) 15 299 505 2,237 Corsica (2A and 2B) 17 402 773 3,245 Charente-Maritime (17) 17 639 1,509 5,636 Corrèze (19) 17 283 567 2,561 Allier (03) 18 407 741 3,864 Gers (32) 18 196 238 1,012 Cantal (15) 19 148 419 1,588 Isère (38) 19 1,681 3,109 13,460 Meurthe-et-Moselle (54) 19 928 2,353 9,046 Aude (11) 20 445 1,225 3,278 Sarthe (72) 21 746 1,302 6,565 Drôme (26) 22 701 1,391 4,848 Loire (42) 22 956 1,989 9,873 Pyrénées-Atlantiques (64) 22 711 1,832 7,290 Seine-Maritime (76) 22 1,666 3,520 16,672 Aube (10) 23 510 1,462 2,719 Rhône (69) 24 2,725 6,179 26,780 Yvelines (78) 25 1,910 3,089 17,452 Tarn (81) 26 421 1,004 3,559 Lot (46) 27 182 394 1,296 Meuse (55) 27 280 365 1,740 Nièvre (58) 27 317 499 2,180 Deux-Sèvres (79) 27 406 689 3,316 Hautes-Pyrénées (65) 28 275 646 2,278 Yonne (89) 29 474 829 3,412 Finistère (29) 30 1,077 2,131 9,952 Hérault (34) 30 1,266 4,000 12,161 Vaucluse (84) 30 863 2,637 7,781 Lot-et-Garonne (47) 32 453 960 3,484 Bas-Rhin (67) 32 1,412 3,157 13,374 Pyrénées-Orientales (66) 33 528 1,750 4,519 Vienne (86) 33 554 1,035 4,789 Moselle (57) 35 1,579 2,806 11,820 Gironde (33) 36 1,965 5,190 16,501 Vosges (88) 37 503 1,009 4,350 Alpes-Maritimes (06) 42 2,136 4,143 11,776 Bouches-du-Rhône (13) 42 2,805 9,991 24,673 Seine-et-Marne (77) 42 1,923 4,237 13,992 Dordogne (24) 43 437 907 3,018 Paris (75) 43 3,950 13,654 42,147 Marne (51) 46 852 1,592 7,872 Charente (16) 47 458 1,029 3,257 Var (83) 47 1,494 3,591 9,333 Hauts-de-Seine (92) 55 2,826 8,787 25,670 Gard (30) 60 932 2,065 6,859 Tarn-et-Garonne (82) 61 302 657 2,096 Essonne (91) 61 1,789 4,996 15,691 Seine-Saint-Denis (93) 65 3,231 13,149 24,476 Val-de-Marne (94) 67 2,132 6,776 17,357 Val-d’Oise (95) 73 1,840 4,608 17,337 Haute-Garonne (31) 84 1,784 4,539 14,572 Total 88,019 205,627 761,630 Sources: DREES and INSEE

Bibliographie

References

Bajos N., Moreau C., Leridon H., Ferrand M., 2004, “Why has the number of abortions not declined in France over the past 30 years?”, Population and Societies, no. 407, 4 p.

Blayo C., 1985, “L’avortement légal en France”, Population et Sociétés, 187, 4 p.

Blayo C., 1995, “L’évolution du recours à l’avortement en France depuis 1976”, Population, 50(3), pp. 779-810.

Blayo C., 1997, “Le point sur l’avortement en France”, Populations et Sociétés, no. 325, 4p.

Boltanski L., 2004, La condition fœtale : une sociologie de l’engendrement et de l’avortement, Paris, Gallimard, 420 p.

Buisson G., Carrasco V., Coldefy M., Thomson E., Trigano L., 2003, “L’activité des établissements de santé en 2001 en hospitalisation complète et partielle”, Études et Résultats, no. 238.

Buisson G., 2005, “Redressements du Programme de médicalisation des systèmes d’informations (PMSI)”, Document de travail de la Drees, Série Statistiques, no. 80.

Gourbin C., 2002, “La mortalité fœtale” in Caselli G., Vallin J., Wunsch G. (eds.), Démographie : analyse et synthèse, Les déterminants de la fécondité, vol. II, Paris, INED, pp. 211-250.

Igas (Inspection générale des affaires sociales), 2002, Rapport d’activité du groupe national d’appui à la mise en œuvre de la loi du 4 juillet 2001 relative à l’interruption volontaire de grossesse et à la contraception, Paris, La Documentation française.

Ined, 1966, “Rapport de l’Ined à Monsieur le Ministre des Affaires sociales sur la régulation des naissances en France”, Population, 21(4), pp. 645-690.

Ined, 1976-1996, “Statistiques de l’avortement en France : annuaire 1976-1996”, Paris, INED.

Jones R. K., Henshaw S. K., 2002, “Mifepristone for early medical abortion: experiences in France, Great Britain and Sweden”, Family Planning Perspectives, no. 3, pp. 154-161.

Kaminski M., Crost M., 1997, “Les interruptions volontaires de grossesse : des caractéristiques stables”, Solidarité Santé, no. 2, pp. 85-95.

Le corre M., Thomson E., 2000, “Les IVG en 1998”, Études et Résultats, no. 69.

Le naour j.-Y., Valenti c., 2003, Histoire de l’avortement : xixe-xxe siècle, Paris, Ed. du Seuil, 387 p.

Memmi D., 2003, Faire vivre et laisser mourir : le gouvernement contemporain de la naissance et de la mort, Paris, La Découverte.

Nisand I., 1999, L’IVG en France : propositions pour diminuer les difficultés que rencontrent les femmes, Paris, ministère de l’Emploi et de la Solidarité, 1999/02, http://www.ping.be/planning-familial/nisaubry.html.

Prioux F., 2002, “Recent demographic developments in France”, Population-E, 57(4-5), p. 689-728.

Régnier-loilier A., 2005, “Évolution des naissances et des grossesses non désirées en France. Réflexions méthodologiques et éléments d’interprétation”, paper presented at the 25th International Population Conference, IUSSP, Tours, France, 18-23 July 2005.

Rossier C., 2003, “Estimating induced abortion rates : a review”, Studies in Family Planning, 34(2), pp. 87-102.

Rossier C., Leridon H., Groupe Cocon, 2004, “The pill and the condom, substitution or association? An analysis of the contraceptive histories of young women in France 1978-2000”, Population-E, 59 (3-4), pp. 387-414.

Rossier C., Michelot F., Bajos N., Groupe Cocon, 2007, “Modeling abortion as a process, with an application to French survey data”, Studies in Family Planning (forthcoming).

Rossier C., Confesson A., Bringé A., 2006, Analyse de la qualité des données des bulletins d’interruptions de grossesse de 2002, http://www.ined.fr/statistiques_ivg/2002/qualite_donnees_2002.pdf

Rossier C., Michelot F., groupe Cocon, forthcoming, “Le choix de la méthode d’IVG en France”, in de La Rochebrochard É. (ed.), L’individu face à la médicalisation de la vie reproductive : Quels choix ? Quelles difficultés ? , Cahiers de l’INED, Paris, INED.

Toulemon L., Leridon. H., 1991, “Vingt années de contraception en France : 1968-1988”, Population, 45 (4), pp. 777-812.

Vilain A., Mouquet M.-C., 2003, “Les interruptions volontaires de grossesse en 2001”, Études et Résultats, no. 279.

Vilain A., 2004, “Les interruptions de grossesse en 2002”, Études et Résultats, no. 348.

Vilain A., 2005, “Les interruptions de grossesse en 2003”, Études et Résultats, no. 431.

Notes

[ *] Institut national d’études démographiques, France
Translated by Roger Depledge

[ 1] The full names of all acronyms are given on the last page.

[ 2] DREES released a highly simplified version of the notification form in early 2004, which was replaced some months later by a fuller version. Since 2005, the notification form has not contained data on the woman’s conjugal status or her partner’s employment, and detailed information on previous pregnancies has been replaced by the number of children and number of previous terminations. One further piece of information is recorded: place where procedure is performed (doctor’s office or hospital).

[ 3] Before 2001, the PMSI had the following codes for recording induced abortions: medicalized abortion (O04), other types of abortion (O05), abortion without further indication (O06), failed attempted abortion (O07) and problems related to unwanted pregnancy (Z64.0). It was hard to distinguish therapeutic abortions from elective ones. Since 2001, each act can be given two codes: a primary diagnosis (mandatory) and an associated diagnosis (optional). Medicalized abortion (O04) is a primary code; the associated code Z64.0 (problems related to unwanted pregnancy) can be used to record elective abortions. For therapeutic abortions the associated code is “abnormal findings on antenatal screening” (O28).

[ 4] The number of notifications recorded for 1995 and 2002 was particularly low, as a result of collection problems in certain regions (for 2002, see Rossier, Confesson and Bringé, 2006).

[ 5] We chose to survey two of the three regions considered by the IGAS report (IGAS, 2002) to pose “problems” of access to abortion: the Paris region (29% of all abortions in metropolitan France are carried out in Île-de-France, SAE 2002) and Midi-Pyrénées (the third problem region is Provence-Alpes-Côte d’Azur). These regions suffer from a shortage of public-sector abortion centres, and a relatively high proportion of abortions are carried out in the private sector (55% in Île-de-France, 57% in Midi-Pyrénées, according to SAE 2002). In contrast, in Nord-Pas-de-Calais, women are overwhelmingly treated in the public sector (4% in the private sector, SAE 2002). In all, 39% of all abortions in France are carried out in these three regions (SAE 2002).

[ 6] Article L178 1 of Act 75-17 of 17 January 1975 relating to elective abortion stated, “In the establishments cited in Article L176 [private], the number of pregnancy terminations performed each year shall not exceed one-quarter of the total number of surgical and obstetrical procedures. Any infringement shall result in the closure of the establishment for one year. Any repeated infringement shall lead to permanent closure”.

[ 7] Detailed tariffs were published in the decree of 23 July 2003 relating to set fees for elective termination of pregnancy, Journal Officiel de la République Française of 28 July 2004, available in French only at http://www.admi.net/jo/20040728/SANP0422519A.html

[ 8] C. Blayo estimates that this ploy accounted for some 10% of the abortions recorded in hospital statistics in 1992 (Blayo, 1995).

[ 9] Until the contents of the uterus are entirely evacuated, the spontaneous abortion is incomplete. After diagnosis, women admitted for incomplete spontaneous abortion are either placed under observation for the spontaneous abortion to complete naturally, or dilatation and curettage (D&C) is performed if the evacuation is slow to occur.

[ 10] The PMSI is reliable from 1998, since 1996 and 1997 were trial years.

[ 11] Dr Danielle Hassoun, Saint-Denis hospital, and Dr Véronique Lejeune, Saint-Antoine hospital, personal communications.

[ 12] For the 25 départements where 100% of induced abortions were performed in the public sector in 2002, the same exercise produces an average of 12.2% for the ratio of spontaneous abortions treated in hospital to births.

Résumé

Quelle a été l’évolution du nombre d’interruptions volontaires de grossesse en France métropolitaine depuis 1975 ? Les sources de données sur l’avortement (la statistique des bulletins établie depuis 1975 et les statistiques hospitalières depuis le milieu des années 1990) ne concordent pas sur ce point. Compte tenu de l’incomplétude des bulletins, l’Ined avait produit par le passé des estimations du nombre total d’avortements : il y aurait eu 250 000 avortements en 1976, 262 000 en 1980, 230 000 en 1987 et 225 000 en 1993. D’après les statistiques hospitalières, le nombre d’interruptions volontaires de grossesse s’établirait à 180 500 en 1994 et 206 500 en 2003. Dans cet article, nous montrons que la tendance indiquée par les estimations de l’Ined (baisse dans les années 1980 et stabilité au début des années 1990) semble correcte ; nous montrons aussi que la hausse du nombre d’avortements indiquée par les statistiques hospitalières de la fin des années 1990 semble être un artefact lié à l’amélioration de ces données.

PLAN DE L'ARTICLE

• I - Three sources of data on abortions in France: notification forms, SAE, PMSI
  • 1 - Notification forms
  • 2 - Annual statistics of healthcare facilities (SAE)
  • 3 - The medical statistics database (PMSI)
  • 4 - Substantial differences between sources
• II - Abortion estimates published by INED from 1976 to 1997
• III - Missing notification forms deduced from hospital statistics
• IV - Elective abortions not recorded in any information system
  • 1 - Spontaneous abortion statistics
  • 2 - An estimate of induced abortions in 1998 and 2002 based on spontaneous abortions
• V - New estimates of induced abortion figures since 1975
• Conclusion
• Annexe

Clémentine Rossier et Claudine Pirus « Estimating the Number of Abortions in France, 1976-2002 », Population (english edition) 1/2007 (Vol. 62), p. 57-88.
URL : www.cairn.info/revue-population-english-2007-1-page-57.htm.
DOI : 10.3917/pope.701.0057.