Community-based evaluation survey of immunizations in Burkina Faso.

Research/Recherche Community-based evaluation survey of immunizations in Burkina Faso V. Schwoebel,1 A.-V. Dauvisis,1 B. Helynck,2 E. Gomes,2 G.F. Drejer,3 M. Schlumberger,1 L. Bibane,4 & H. Rumke5 A cluster sample survey was conducted in January 1989 in 3 provinces of Burkina Faso to evaluate an immunization programme (based on two contacts, providing inactivated poliomyelitis vaccine plus DPT) that had been launched in 1982-84. The objectives were to estimate neonatal tetanus (NNT) mortality and poliomyelitis prevalence in the study area. The target population (using the same sample of house- holds comprised 2107 live infants born during the preceding year for the NNT survey, and 17 154 children aged 0-9 years for the poliomyelitis survey. The NNT mortality rate was 3.3 per 1000 live births, and the poliomyelitis prevalence rate was 2.8 per 1000 children aged 5-9 years. Dates of onset of poliomyelitis cases among children aged 0-9 years and the numbers of children at risk during the 10-year recall period, reconstituted with demo- graphic indicators taken from standardized life-tables, were used to calculate the incidence rates of poliomyelitis. These rates could be compared in the 5-year period preceding the survey, and showed a decreasing trend consistent with routine surveillance data. Introduction Evaluations of immunization programmes tend to focus mainly on operational issues and use vaccina- tion coverage as the principal indicator of the suc- cess of the programme. Surveys using the WHO/EPI cluster sampling method (I) are routinely used to provide estimates of coverage. Frequency and trends of target diseases, though essential for the evaluation of the programme, are often difficult to obtain through routine surveillance in countries where access to health services is difficult and the health information system is still weak. To meet the need for obtaining information on the outcomes of vacci- I Association pour la Promotion de la Medecine Preventive, Paris, France. Requests for reprints should be sent to Dr V. Schwoebel at her present address: European Centre for the Epidemiological Monitoring of AIDS, H6pital National de Saint- Maurice, 14 rue du Val d'Osne, 94410 Saint-Maurice, France. 2 Organisation de Coordination et de Cooperation pour la Lutte contre les Grandes Endemies, Bobo-Dioulasso, Burkina Faso. 3 Stichting Redt de Kinderen (Save The Children Fund), The Hague, Netherlands. 4 Ministere de la Sante et de l'Action Sociale du Burkina Faso. 5 Rijksinstituut voor Volksgezondheid en Milieuhygiene (National Institute of Public Health and Environmental Protection), Biltho- ven, Netherlands. Reprint No. 5318 nation programmes, guidelines on community-based surveys on poliomyelitis prevalence (2)a and neona- tal tetanus mortality' have been developed by WHO, which can be used to estimate the baseline incidence of these diseases. Since 1982 a Dutch Foundation, Stichting Redt de Kinderen (SRK), has provided funding for an immunization programme in four provinces of Bur- kina Faso, with technical support from the Associa- tion pour la Promotion de la Medecine Preventive (APMP). Coverage surveys conducted in 1987 in the three provinces where the programme first started indicated a 38% prevalence of children immunized according to the adopted schedule in one of the pro- vinces and 52% in the other two,c placing them in the top 25% of the 29 provinces of the country.d The a Bernier, R.H. Prevalence survey techniques for paralytic polio: an update. Unpublished document WHO/EPI/GEN/84/3, 1984. b Galazka, A. & Stroh, G. Neonatal tetanus. Guidelines on the community-based survey on neonatal tetanus mortality. Unpub- lished document WHO/EPI/GEN/86/8, 1986. c OCCGE-Unite de Vaccinologle. Rapport d'evaluation de la couverture vaccinale du Bam, Namentenga et Sanmatenga. Bobo-Dioulasso, 1987. d Ministere de la Sante et de l'Action Sociale du Burkina Faso. Enquete nationale de couverture vaccinale par province: Rapport. Bobo-Dioulasso, 1988. Bulletin of the World Health Organization, 70 (5): 583-590 (1992) © World Health Organization 1992 583 V. Schwoebel et al. impact of the programme was difficult to assess, since neonatal tetanus (NNT) cases had been repor- ted separately from tetanus cases in other age groups only after 1985, and since the number of reported cases of poliomyelitis and neonatal tetanus was believed to considerably underestimate the true inci- dence. As part of the evaluation of its immunization programme which included a survey of coverage in each province, SRK, together with APMP, the Orga- nisation de Coordination et de Cooperation pour la Lutte contre les Grandes Endemies (OCCGE), the Ministry of Health of Burkina Faso, and the Rijks- instituut voor Volksgezondheid en Milieuhygiene (RIVM) carried out a survey to estimate the inci- dence of neonatal tetanus and the prevalence of poliomyelitis in the three provinces where the im- munization programme had been launched between 1982 and 1984. The survey, which was conducted in January 1989, was designed to obtain cross-sectional estimates of the two diseases. However, since base- line incidence was not available for either of them, it was decided to test whether this survey could also indicate the trend of poliomyelitis incidence over a ten-year period. Materials and methods Survey area and population The three provinces (Bam, Sanmatenga and Namen- tenga), which are contiguous and situated in the north-central part of the country, are essentially rural. More than 95% of the population belongs to the Mossi ethnic group and uses the language called More. The immunization programme was based on a simplified immunization schedule using the en- hanced inactivated poliovirus vaccine (eIPV) com- bined with diphtheria-pertussis-tetanus (DPT) vaccine (3), produced by the RIVM (Bilthoven, Nether- lands). Health centres were responsible for providing immunizations to the population living within a radius of 10 km (48% of the population at the time of the survey), either in the centre or during outreach activities. BCG and DPTP1 (first dose of DPT and poliovirus vaccines), were scheduled at 3 months, and DPTP2 (second dose), measles and yellow fever (YF) vaccines at 9 months; a minimum interval of 2 months was required between DPTP1 and DPTP2. Mobile teams, which covered the more remote popu- lation (52%), visited each village every 6 months, and delivered BCG and DPTP1 vaccines at 3-8 months, and DPTP2, measles and YF vaccines at 9-14 months. Women aged 15-44 years received up to 5 doses of tetanus toxoid (TT) following the WHO-recommended schedule. The target population comprised all children bom between 1/1/79 and 31/12/88. The NNT mor- tality survey was limited to all babies bom from 1 to 13 months before the survey, i.e., between 1/12/87 and 30/11/88, thus allowing each infant to have been at risk of contracting NNT during its entire neonatal period. All children bom during the ten-year study interval were included in the poliomyelitis preva- lence survey. Sampling method The WHO/EPI cluster sampling method for esti- mating disease incidence was used (4). Sample size calculations were based on the number of children needed to get a reasonable estimate of the inci- dence of NNT. A sample of 2100 live births was chosen based on an expected NNT mortality rate of 10 per 1000 and a design effect of 1.7,e i.e., 70 live births in each cluster. It was estimated that the house- holds visited for completing this sample would also provide approximately 15 000 children aged 0-9 years. This sample size was considered adequate to estimate with sufficient precision a poliomyelitis prevalence of 5 per 1000. A single sample of 30 clusters representative of all three provinces was selected by systematic sam- pling using the list of village populations available from the last census (1985). In each cluster selected, interviewers verified on site the administrative limits of the village with the village leader. Starting from the geographical centre of the village, a direction was randomly chosen. Houses along this direction were numbered up to the limits of the village, and one was randomly selected. From there, interviewers proceeded to the nearest houses until 70 live births were included. Case ascertainment. Mothers were asked whether they had given birth to a live baby in the past ten years. For those children bom from 1 to 13 months before the survey (NNT sample), potential cases of NNT were screened by registering neonatal deaths with the aid of a calendar of traditional events. Sub- sequently, the diagnosis of NNT was ascertained by one of the 7 participating physicians during a second visit to the family, using a standardized question- naire.e Blood samples were taken from those mothers of potential NNT cases who had evidence of having received TT injections, and the tetanus anti- body titres were measured blindly at the RIVM using a toxin-binding inhibition assay. Results of this test correlate well with the standard mouse neutralization test (5, 6). Titres of 0.1 IU/ml and more were con- sidered as protective. e See footnote b, page 583. 584 WHO Bulletin OMS. Vol 70 1992 Evaluation survey of immunizations in Burkina Faso From among the children born during the 10- year recall period (poliomyelitis sample), potential cases of poliomyelitis were screened by asking for those with any "walking disability". These children were examined by one of the authors (G.F.D.), a paediatrician with tropical experience. Dates of onset of poliomyelitis were estimated by asking for the child's age at onset. Case definitions are summarized in Table 1. Rate calculations The NNT mortality rate was calculated using the total number of live births registered in the NNT sample as a denominator. The poliomyelitis preva- lence rate was calculated for living children aged 5-9 years and compared with those of other surveys that usually use this age group as a target population. Incidence rates for each of the ten years of the recall period were calculated using all cases enumer- ated in the survey. As denominators, we used the number of children at risk (aged 0-4), in person- years, reconstituted with life tables for each calendar year. Calculating these numbers directly would have been possible if the date of birth and age at death had been recorded for each child of the sample. How- ever, these data were not collected because they were considered as too unreliable. Instead, indicators taken from standardized life tables prepared by the United Nations (7) (West model, infant mortality rate of 140 per 1000, life expectancy of 47.2 years) were used. The annual number of births between 1979 and 1987 was first calculated using the number of live births in 1988 in the sample and the population growth rate. Next, for each of the birth cohorts, the probabilities of dying at ages 0 to 9 were applied, and the number of person-years-lived below the age of 5 years in each calendar year was calculated. Total number of person-years-lived was finally Table 1: Criteria for defining cases of neonatal tetanus and poliomyelitis, by screening and ascertainment, in the evaluation study of immunizations in Burkina Faso Disease Neonatal tetanus Screening (interviewer) Born between 1/12/87 and 30/11/88 Died within 30 days Poliomyelitis Born between 1/1/79 and 31/12/88 Walking disability or "any problem of the leg" Ascertainment (physician) Normal during .2 days Stopped suckling General stiffness or fits Flaccid paralysis one or two legs Normal sensation Acute onset obtained by summing up figures from each cohort for each calendar year. The annual incidence rates obtained were then multiplied by two correction fac- tors: 1.33 to account for lethal forms of poliomyelitis and complete recoveries, and 1.25 to account for forms involving only the upper limb (2). Confidence intervals around the observed rates were obtained from tables for binomial proportions, after adjusting denominators for the design effect introduced by the cluster sampling method (4). The x2-test for trend was used to assess trends of inciden- ce rates calculated under the life table model. Logistics and costs The three provinces were surveyed within 17 days by 30 local interviewers organized in teams of 2, with one vehicle for 2 teams. Among the 11 supervi- sors, 6 were physicians who interviewed the mothers of potential NNT cases. Additionally, an independent team (the paediatrician and a medical student) examined all potential cases of poliomyelitis. The teams progressed in the same direction, which made it easier for the paediatrician who had to visit each team every 2 or 3 days. Field work, including train- ing of interviewers, involved a total of 1292 person- days (714 for interviewers and 578 for supervisors and other personnel). The total costs were approximately US $77 000 (300 CFA = $1): $17 000 (22%) on supplies, vehicle maintenance and petrol during the survey; $20 000 (26%) on salaries and expenses for local staff; and $40 000 (52%) for survey preparation, data col- lection and analysis, report preparation, international travel, and salaries and expenses for expatriate super- visors, physicians, logistic officers and statisticians. Results Neonatal tetanus Among the 2107 live births enumerated in the sample, 63 neonatal deaths were reported (neonatal mortality rate = 30 per 1000; 95% CI = 22-43 per 1000), among which 7 (11%) had signs and symp- toms compatible with neonatal tetanus. NNT mortal- ity rate was 3.3 per 1000, with 95% CI of 1.3-7.0 per 1000. The design effect introduced by the cluster sampling method (4) was 1.03, and was thus negli- gible. The male-to-female sex ratio was 1.03 for all neonatal deaths, and 0.75 for NNT cases. Three of the 7 mothers of NNT cases had received TT injec- tions before the estimated date of birth of the child (respectively 2, 3 and 3 doses that met the criteria of a minimum interval between injections) and were WHO Bulletin OMS. Vol 70 1992 585 V. Schwoebel et al. tested for antibody titres. Two had protective anti- body titres (.0.1 IU/ml). Poliomyelitis Among the 17 154 children born between 1/1/79 and 31/12/88 enumerated in the sample, 109 had a walk- ing disability, of whom 24 (22%) were ascertained to be poliomyelitis cases (Table 1). Three cases had been contracted outside the study provinces and were eliminated from further analysis. Of the remaining 21 cases, 19 were observed in children aged 5-9 years. The prevalence rate of lameness due to polio- myelitis was 2.8 per 1000 (95% CI = 1.7-4.6) in this age group. The design effect due to cluster sam- pling was 1.35. Fourteen of the 21 children with poliomyelitis were male (M:F ratio = 2:1), 20 (95%) were under 3 years old at onset of the disease, 52% were under 2 years. Five children had received 2 doses of DPTP. Of these, three had received the 2 doses after the reported year of disease onset while the other two may have received the 2 doses before disease onset (in the same year as the first reported symptoms). Calculations of the annual number of children at risk, i.e., all children aged 0-4 years, with the life- table model, showed that this number increased be- tween 1979 and 1983, whereas from 1984 to 1988 both the size and the age structure of the population at risk remained fairly stable (Table 2). Using these numbers as denominators, annual incidence rates were calculated. During the first 5-year recall period (1979 to 1983), the variability of the age structure of the population at risk did not permit any comparison of rates from one year to another. However, in the second 5-year period (1984 to 1988), during which the age structure was Fig. 1. Poliomyelitis incidence rates calculated in the survey sample and reported in Burkina Faso, 1979-88. 1 co 0.9 --------------------------------------------------- x 0.6 Calculated 0.4 ------------------------- ---- --------- 0.3 - --------------------- - 0.2 Reported a: 0.1 -_ _ _ _ _ _ _ _- 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 Years relatively stable, incidence rates ranged from 0.7 in 1984 to 0 per 1000 person-years in 1987 and 1988 (Table 2 and Fig. 1), demonstrating a significant decreasing trend (P = 0.034, X2-test for trend). Discussion Neonatal tetanus The NNT mortality rate found in this study was 3.3 per 1000 with a 95% confidence interval of 1.3 to 7.0 per 1000. For the same year, the incidence of neonatal tetanus reported through the national sur- veillance system of Burkina Faso was only 0.5 per 1000, which confirms the frequent observation that routine surveillance data usually underestimate the extent of NNT incidence (8). Although the estimate from this survey is somewhat imprecise, it can none- theless be considered as relatively low compared with rates reported in similar surveys conducted in other countries (8) and is one-third of the estimated rate for Africa (9). Two major protective factors are known to be effective in preventing NNT: sterile delivery practice and immunization of the mother with tetanus toxoid. Although we lack precise infor- mation on the conditions of deliveries in this area, we know that in 1986 only 17%, 24% and 27% of these deliveries, respectively in the 3 provinces, were attended by qualified personnel (Ministere de la Sante et de l'Action Sociale du Burkina Faso, un- published data). By contrast, the immunization programme was effective in reaching a substantial part of the women of reproductive age, as was shown by the vaccination coverage survey conducted in Jan- uary 1989 in the same three provinces.f In this sur- vey, 34-61% of women aged 15-44 years had docu- mented evidence of at least two doses of tetanus toxoid. Therefore it is probable that the low rate of NNT can be attributed to the protection of a large proportion of newboms by the vaccination of their mothers. Nevertheless, the number of cases may represent an underestimation. We found a neonatal mortality rate (NMR) of 30 per 1000. If we assume that the infant mortality rate (IMR) in Burkina Faso was 134 per 1000, and that neonatal deaths accounted for 30% of infant deaths (10), the NMR should have been 40 per 1000. It is known that in such surveys, mothers may underreport their infant deaths. This is I Ministere de la Sante et de l'Action Sociale du Burkina Faso, SRK, APMP, OCCGE. Evaluation epid6miologique du Programme elargi de Vaccination des provinces du Bam, San- matenga et Namentenga: rapport final. Paris, 1989. 586 WHO Bulletin OMS. Vol 70 1992 Evaluation survey of immunizations in Burkina Faso Table 2: Study children at risk of poliomyelitis, in person-years (N), number of cases of poliomyelitis (n), and inci- dence rates by year, Burkina Faso, 1979-88 Birth cohorts 1979 N n 1980 N n 1981 N n 1982 N n 1983 N n 1984 N 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 767 1460 3 1 788 1363 3 1499 2 809 0 1330 0 1400 2 1540 832 0 1312 650 0 1 1367 1347 0 0 1438 1404 1 0 1582 1478 1 1 854 1626 0 1 877 0 1985 N 1986 N n 1987 N n 1988 N n TOTAL N n Crude incidence rate per 1000 Correcteda incidence rate per 1000 667 0 1384 685 0 0 1442 1422 705 0 0 0 1518 1482 1461 1 0 0 1670 1560 1522 0 1 0 902 1716 1602 0 1 0 926 1762 0 0 952 723 0 1501 0 1564 0 1646 0 1811 0 0 973 0 767 2248 3 2 3671 5 5102 3 6553 2 7382 3 7583 7791 1 2 8004 0 8218 0 3.9 0.9 1.4 0.6 0.3 0.4 0.1 0.3 0 0 6.5 1.5 2.3 1.0 0.5 0.7 0.2 0.4 0 0 a Multiplied by 1.33 and 1.25 to take into account lethal forms, complete recoveries and forms involving only the upper limb. less likely to be due to recall failures than to cultural conceptions about the point at which life is really starting: when a child dies very young, particularly in the neonatal period, both birth and death may be omitted (11). The low rate observed could also have been due to the failure of mothers to identify NNT as the cause of death of their infant. Indeed, standardized questions seemed difficult for some mothers to answer, but the sensitivity of the case definition would be very difficult to assess. With regard to the specificity, we observed that three out of 7 mothers of NNT cases had documented evidence of TT immu- nization before the estimated date of birth of the child, with two of them having protective levels of antibodies at the time of the survey. Because it is very unlikely, although still possible, that these anti- bodies were due to doses received after the birth of the child (but not recorded on the vaccination card), it is most probable that these two cases were false posi- tive cases. While the validity of a diagnosis of NNT based on interviewing the mothers is believed to be excellent in areas where the incidence of neonatal tetanus is high and the disease is well known in the community (12, 13), such a method is likely to be- come less reliable in regions like our survey area with a low NNT incidence. Nevertheless, the NNT mortality rate found in our study is unlikely to have been substantially underestimated, and this result con- firmed the impact of the immunization programme in the area. Poliomyelitis The prevalence rate of 2.8 per 1000 that we observed in children aged 5-9 years is relatively low com- pared with the rates reported in other lameness surveys conducted between 1979 and 1981 in several African countries. In these studies, the rate ranged from 2.4 to 13.2 per 1000 in the same age group (14). WHO Bulletin OMS. Vol 70 1992 587 V. Schwoebel et al. Important factors in the prevention of polio- myelitis are a safe water supply, adequate sanitation, and immunization. Undeniable improvements in terms of safe water delivery have been made in Burkina Faso within the last ten years, where now a sub- stantial number of villages are equipped with bore- holes, but improvements in sanitation have been slow. On the other hand, the moderate prevalence rate found in this study seems consistent with an im- pact due to the immunization programme, which had been launched in the three provinces between 1982 and 1984. In 1987, the coverage of children aged 12-23 months with two doses of eIPV ranged from 47% to 65%,9 although it fell to 38% in 1989.h We can estimate that after 1983 approximately half of each cohort of children below the age of 2 years were vaccinated. It has been shown (15) that two doses of eIPV have a vaccine efficacy close to 90%. The expected impact of the programme in terms of cases of poliomyelitis prevented is therefore substan- tial. The WHO lameness survey method is designed to give an estimate of the prevalence rate of polio- myelitis in children aged 5-9 years, which is a sum- mary figure representing the result of risks encoun- tered by this cohort over the preceding 10 years. It can give no indication of the trend of the disease inci- dence, and if a major change such as an effective vaccination programme had taken place in recent years, the effect on this measure would have been limited (2). However, the reconstitution of the popu- lation at risk (in person-years) made it possible to compare incidence rates of poliomyelitis within the 5 preceding years, assuming that children at risk were those aged 0-4 years. A decreasing trend was observed in poliomyelitis incidence, which parallel- led the decrease in cases reported in the whole coun- try (Fig. 1). Although the period (after 1983), during which very few cases were observed, could have been a period of low incidence independently of the immunization programme, our results seem consis- tent with an impact of vaccination in the area. There are however several limitations to the method used. When the target age group of the sur- vey is extended to children aged 0-9 years, new dif- ficulties appear in identifying cases in younger chil- dren. The screening method, i.e., asking for any "walking disability", may have missed cases in chil- dren below 12-18 months of age. Indeed the inter- viewers spontaneously modified the question into a question about "any problem of the leg", which made the screening method very sensitive. On the 9 See footnote c, page 583. h See footnote f, page 586. other hand, our method of case ascertainment could be considered as very specific because the same experienced paediatrician examined all suspected cases, which reduced the number of false positive diagnoses to a minimum. Estimation of annual incidence rates in the last 5 years is subject to other additional limitations. It was based on the correct identification by the mother of the child's age at onset of the disease, which was probably not very reliable. However, conceming cases within the last 5 years only, it could be expec- ted to be more accurate. As for the denominator used, the calculation of the annual number of chil- dren at risk in person-years was made with a model based on the assumptions that natality and mortality remained stable during the period. Assumptions regarding the stability of mortality may be chal- lenged. Indeed the infant mortality rate in Burkina Faso is believed to have decreased from 167 per 1000 in 1975 to 134 per 1000 in 1985 (Ministry of Health data). However, no events that could cause an acute increase in mortality or decrease in natality were observed in the survey area within the last ten years, leading us to believe that relatively stable mor- tality and natality could reasonably be assumed. The cost-effectiveness of such large surveys has been challenged on the basis of their sample sizes, the cost, and their inability to provide information on trends (16). However, widening the age range of the children surveyed slightly increased the time and cost of the survey; the major part of the time was devoted to moving from house to house and to iden- tifying all the members in each household. This addi- tional effort enabled us to collect valuable informa- tion on the trends of the disease in the recent 5 years which could not have been provided by the classic lameness survey method. Conclusion The neonatal tetanus mortality surveys and polio- myelitis lameness surveys provide information on the extent of these two diseases, which cannot be obtained through routine surveillance data. By com- bining the two in a single survey using the same sample of households, we greatly reduced the costs involved in conducting two separate surveys. In addition, by extending our poliomyelitis sample to children aged 0-9 years, we obtained information on the trends of this disease during the preceding 5 years. In areas where the incidence of the two diseases is still high or moderate, and where data cannot be obtained through other sources, the combined survey can provide useful information on their incidence 588 WHO Bulletin OMS. Vol 70 1992 Evaluation survey of immunizations in Burkina Faso and help evaluate the impact of immunization pro- grammes. Acknowledgements We are grateful to Nancy Binkin for critically reviewing the manuscript, and Dr Roger Bernier, Dr Ralph Henderson and Dr Susan Robertson for their useful comments and suggestions. Resume Enqudte au niveau communautaire pour l'evaluation de la vaccination au Burkina Faso Pour 6valuer les programmes de vaccination, il est essentiel de connaitre les niveaux d'incidence des maladies cibles et leurs tendances. Lorsque cette information est absente ou peu fiable, des enquetes au niveau communautaire sur la pr6va- lence de la poliomy6lite et sur la mortalit6 impu- table au t6tanos n6onatal peuvent fournir des donn6es de base pour estimer l'incidence de ces maladies. Toutefois, ce type d'enquete n'a pas e conqu pour fournir des informations sur les ten- dances. Un programme de vaccination (administration de deux doses de vaccin antipoliomy6litique inac- tiv6 associ6 au DTC) lanc6 entre 1982 et 1984 dans trois provinces du Burkina Faso (Bam, San- matenga et Namentenga) a fait l'objet d'une telle evaluation en janvier 1989. L'enquete, men6e selon la technique de l'6chantillonnage par grappes, avait pour principal objectif d'estimer la mortalit6 imputable au t6tanos n6onatal et la pr6- valence de la poliomy6lite dans la region. En outre, les donnees ont 6t6 analys6es pour voir s'il 6tait possible d'en extraire des informations sur l'6volution de l'incidence de la poliomy6lite au cours des annees precedentes. Les populations cibles etaient constitu6es des enfants n6s vivants au cours de l'ann6e precedente en ce qui concer- ne l'enquete sur le t6tanos n6onatal et des enfants ages de 0 a 9 ans pour la poliomy6lite. Le meme echantillon de menages a fourni le nombre de sujets n6cessaire a la r6alisation des deux enquetes (respectivement 2107 naissances vivantes et 17 154 enfants). Le taux de mortalite imputable au t6tanos neonatal a ete evalue a 3,3 pour 1000 naissances vivantes et le taux de prevalence de la poliomyeli- te a 2,8 pour 1000 enfants ag6s de 5 a 9 ans. Ces taux sont faibles si on les compare aux taux releves dans des enquetes similaires pratiqu6es dans la meme r6gion. Toutefois, compte tenu des titres d'anticorps trouv6s chez certaines meres, on peut s'interroger sur la specificit6 du diagnostic retrospectif de t6tanos n6onatal fond6 sur l'interro- gatoire des meres. Pour calculer les taux d'incidence de la polio- myelite, on a note la date d'apparition de la mala- die chez les enfants ag6s de 0 a 9 ans et on a estime le nombre d'enfants a risque au cours des dix ann6es precedentes en reconstituant r6tros- pectivement des cohortes grace aux indicateurs d6mographiques des tables standard de mortalit6. Ce n'est qu'au cours des cinq ann6es pr6c6dant l'enquete que la structure d'age de la population d'enfants a risque a ete suffisamment stable pour que l'on puisse comparer les taux d'incidence. Au cours de cette p6riode, l'incidence de la poliomy6- lite a pr6sent6 une tendance a la baisse significa- tive, passant de 0,7 a 0 pour 1000 personnes- ann6e (P = 0,034), ce qui est en accord avec les donn6es de surveillance nationale. Le fait d'utiliser le meme 6chantillon de popu- lation pour 6valuer la mortalit6 imputable au t6ta- nos n6onatal et la prevalence de la poliomy6lite s'est revele a la fois pratique et 6conomique. En outre, des informations utiles ont e recueillies sur l'6volution de l'incidence de la poliomyelite au cours des cinq ann6es pr6c6dentes en incluant les enfants ag6s de 0 a 9 ans dans la population cible de l'enquete et en notant la date approxima- tive de I'apparition de la maladie chez ces enfants. References 1. Henderson, R.H. & Sundaresan, T. Cluster sam- pling to assess immunization coverage: a review of experience with a simplified sampling method. Bul- letin of the World Health Organization, 60: 253-260 (1982). 2. Laforce, F.M. et al. Clinical survey techniques to estimate prevalence and annual incidence of polio- myelitis in developing countries. Bulletin of the World Health Organization, 58: 609-620 (1980). 3. Stoeckel, P. et al. Bases theoriques, applications pratiques et premiere evaluation d'un programme elargi de vaccination des enfants africains. Bulletin de la Soci6t6 de Pathologie exotique, 76: 846-852 (1983). 4. Rothenberg, R.B. et al. Observations on the appli- cation of EPI cluster survey method for estimating disease incidence. Bulletin of the World Health Organization, 63: 93-99 (1985). 5. Hendriksen, C.F.M. et al. The toxin-binding inhibi- tion test as a reliable in vitro alternative to the toxin neutralization test in mice for the estimation of teta- WHO Bulletin OMS. Vol 70 1992 589 V. Schwoebel et al. nus antitoxin in human sera. Journal of biological standardization, 16: 287-297 (1988). 6. Hendriksen, C.F.M. et al. Combined estimation of tetanus and diphtheria antitoxin in human sera by in vitro toxin-binding inhibition (ToBI) test. Journal of biological standardization, 17: 191-200 (1989). 7. United Nations. World demographic estimates and projections: 1950-2025. ST/ESA/SER.R/79, New York, 1988. 8. Stanfield, J.P. & Galazka, A. Neonatal tetanus in the world today. Bulletin of the World Health Organi- zation, 62: 647-669 (1984). 9. Expanded Programme on Immunization. Neonatal tetanus in Africa. Weekly epidemiological record, 62 (1): 305-308 (1987). 10. Fargues, P. Ages au deces et niveaux de mortalite: 6valuer le taux d'enregistrement des d6cbs a partir de leur structure par age. Application a la Tunisie. Population, 39: 47-76 (1984). 11. Brass, W. et al. The demography of tropical Africa. Princeton, Princeton University Press (1968). 12. Garenne, M. & Fontaine, 0. Enquete sur les causes probables de deces en milieu rural au S6n6gal. In: Mesure et analyse de la mortalit6: nouvelles approches. Paris, INED/UIESP/PUF, 1988, pp. 123-141. 13. Zimicki, S. L'enregistrement des causes de deces par des non m6decins: deux experiences au Ben- gladDesh. In: Mesure et analyse de la mortalite: nouvelles approches. Paris, INED/UIESP/PUF, 1988, pp. 101-122. 14. Bernier, R.H. Some observations on poliomyelitis lameness survey. Review of infectious diseases, 6: 371-375 (1984). 15. Robertson, S.E. et al. Clinical efficacy of a new, enhanced-potency, inactivated poliovirus vaccine. Lancet, 2: 897-899 (1988). 16. Cutts, F.T. The use of the WHO cluster survey method for evaluating the impact of the expanded programme on immunization on target disease inci- dence. Journal of tropical medicine and hygiene, 91: 231-239 (1988). 590 WHO Bulletin OMS. Vol 70 1992