SEARCH

SEARCH BY CITATION

Keywords:

  • schistosomiasis;
  • Schistosoma mansoni;
  • high endemicity communities identification;
  • LQAS;
  • Madagascar

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

Lot quality assurance sampling (LQAS) was evaluated for rapid low cost identification of communities where Schistosoma mansoni infection was hyperendemic in southern Madagascar. In the study area, S. mansoni infection shows very focused and heterogeneous distribution requiring multifariousness of local surveys. One sampling plan was tested in the field with schoolchildren and several others were simulated in the laboratory. Randomization and stool specimen collection were performed by voluntary teachers under direct supervision of the study staff and no significant problem occurred. As expected from Receiver Operating Characteristic (ROC) curves, all sampling plans allowed correct identification of hyperendemic communities and of most of the hypoendemic ones. Frequent misclassifications occurred for communities with intermediate prevalence and the cheapest plans had very low specificity. The study confirmed that LQAS would be a valuable tool for large scale screening in a country with scarce financial and staff resources. Involving teachers, appeared to be quite feasible and should not lower the reliability of surveys. We recommend that the national schistosomiasis control programme systematically uses LQAS for identification of communities, provided that sample sizes are adapted to the specific epidemiological patterns of S. mansoni infection in the main regions.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

Schistosomiasis is endemic throughout Madagascar and represents one of its most important public health problems. In response, public health authorities are implementing a national control programme (Roux et al. 1994).

Recent advances in chemotherapy, epidemiology and morbidity assessment have changed the concept of schistosomiasis control [World Health Organization (WHO) 1993]. As schistosomiasis-associated morbidity is correlated with the intensity of infection, which in turn is usually associated with a high level of prevalence, current recommendations are to control morbidity by repeated systematic treatment in hyperendemic areas where people are most likely to present complications. In doing so, one accepts a persistence of transmission, but avoids development of high intensity infections leading to severe complications (Butterworth et al. 1991; Ravaoalimalala et al. 1994; Boisier et al. 1994). Hence it is important to be able to identify sites with a high prevalence of infection. Although schistosomiasis transmission occurs in focuses within a given area, communities may show very different levels of endemicity (Jordan et al. 1993), necessitating the assessment of individual communities. Conventional sampling is too expensive and time-consuming in a country such as Madagascar, which has neither enough health staff nor laboratory capability in peripheral regions and limited financial resources. The lot quality assurance sampling method (LQAS) seemed a good option to achieve the goal.

The LQAS was originally designed for manufacturing inspection where it was necessary to keep sampling costs to a minimum. LQAS is used to verify whether a lot of goods issued from a factory can be accepted for dispatch or not. The decision about the quality of a given lot is based on the probability that the number of defective items in a sample selected from that lot is less than or equal to some critical value (Lemeshow & Taber 1991).

The LQAS has been field-tested in public health, most often in evaluating immunization coverage (Singh et al. 1996; Tawfik et al. 2001). We describe an application of LQAS to the identification of communities with high prevalence of Schistosoma mansoni infection in Madagascar.

Objectives

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

The objective of the study was to assess the ability of LQAS reliably and quickly identify communities with high endemicity of S. mansoni to be included in the national control programme. Two points had to be explored. First, we had to identify the best sampling plan, considering the current distribution of schistosomiasis, the cost of field surveys and the laboratory capability of testing numerous stool samples. The best sampling plan would associate lowest sample size and highest statistical power for a large range of prevalences. Secondly, we had to test purely operational aspects such as the possibility of assigning the management of field surveys to non-specialists like teachers.

Study area

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

The study was conducted in the region of Ihosy between April and June 1998. Ihosy is located about 440 km from the capital city Antananarivo in southern Madagascar. This region includes two distinct areas. The first is the large Horombe plateau (altitude 1000–1200 m), characterized by very low population density and villages built close to water bodies located in depressions. The second is the area of Ihosy (altitude 700–800 m), where rivers and water bodies are numerous and natural water meets inhabitants' water needs.

Study population

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

As children are usually the most infected age group and because schoolchildren are easier to reach than any other category of population, the study population was children aged 5–14 years attending 32 primary public schools and six private schools of six communities of the region: Zazafotsy, Ankily, Mahasoa, Sahambano, Analavoka and Ambatolahy. In those six communities no previous antischistosomal treatment had been delivered at a community level.

Statistical method

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

The LQAS aims to test if a proportion P (i.e. the true proportion of infected individuals) in a studied community is superior or equal to a given proportion P0 (threshold prevalence beyond which health planners will intervene). LQAS is not designed to provide an accurate estimation of the proportion P in the community but only classifies P as being above or under a given value. The procedure is set up as a one-sided test as follows:

  • the null hypothesis to be tested is H0: P ≥ P0, meaning that the studied community can be considered as being hyperendemic;
  • the alternative hypothesis Ha: P < P0, meaning that the studied community is not hyperendemic.

The choice of these hypotheses is explained by the fact that wrongly identifying a community as low-prevalence would be a serious error (α-error), while falsely identifying it as high-prevalence (β-error) would merely represent a financial problem, providing that treatment of non-infected individuals is not dangerous.

If less than or equal to a critical number of d subjects are found presenting the study parameter (d denotes the number of infected individuals in a sample of n subjects), the hypothesis H0 is rejected (and P < P0 is admitted). If the number of individuals who present the study parameter is higher than d, then the null hypothesis is not rejected (P ≥ P0 is admitted). As soon as d + 1 subjects have been found presenting the studied parameter, the study of the remaining individuals of the sample can be stopped, unless one wants to assess also the prevalence and its confidence interval.

Setting the sample size n and the number d of permissible infected individuals among the sample can be achieved in two ways. The first is the use of a series of tables, where different upper and lower performance levels are shown for given type I and type II errors. The second is the use of Receiver Operating Characteristic (ROC) curves plotting the probability of rejecting H0 according to the true prevalence of infected individuals in the lot (Lemeshow & Taber 1991). Each ROC curve incorporates information on a unique combination of n and d (referred to as a candidate sampling plan). The horizontal axis of a ROC curve corresponds to the proportion P in the population who are infected. The vertical axis correspondsto the probability of rejecting the null hypothesis (H0: P ≥ P0) and of concluding that the community is not hyperendemic.

The choice of a sampling plan consists of selecting that ROC curve containing a point corresponding to both the selected threshold prevalence and α levels of interest in a study, while also having a steep enough slope in the region on the left of P0 to indicate sufficient statistical power associated with a value of n as small as possible. Each combination of n and d generates a single curve. As the sample size n increases, or critical value d decreases, the ROC curve will shift downward indicating that, for a given α level, we will be able to test for a lower value of P0. These interrelationships should be considered carefully in order to gain a better intuitive sense of the relationship amongst these parameters (Moïse et al. 1986; Lemeshow & Taber 1991). In this study, where each school was considered a lot, a sample was randomized in which the observed number of infected children was compared with a previously identified critical value corresponding to the threshold defining hyperendemic infection.

Sampling method

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

Seven ROC curves were designed to test a threshold prevalence of 60%(Figure 1). The curves are labelled according to the combination of n and d: (16; 6) (14; 5) (12; 4), etc. The sampling plan (16, 6) which had the highest statistical power was chosen to be implemented on the field. We considered that the value n = 16 was the acceptable maximum according to the laboratory capabilities. After an exhaustive census of the schoolchildren, voluntary teachers were given brief operating instructions and then randomized 16 children in each school under direct supervision of the study staff.

image

Figure 1. Receiver operating characteristic (ROC) curves of seven sampling plans designed to test a threshold value of prevalence of 60%.

Download figure to PowerPoint

In addition to this sampling trial in the field to test teachers' capability to participate in the schistosomiaisis control programme various other samples ranging from n = 4 to n = 16 (two trials for each plan) were randomly constituted in the laboratory from the pupil list of each school, using Epi Info software.

Parasitological method

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

All schoolchildren (whether or not they had been randomized) from studied schools were asked to provide a fresh stool specimen. Specimens were transferred into glass tubes with merthiolate-iodine–formalin (MIF) allowing good conservation of stool from the field to the laboratory. Teachers had to collect and to transfer stools only for randomized pupils, while specimens from other children were managed by the study staff.

Data analysis

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

For each school, the stool examination of all the children was considered as the reference survey and its results were used to characterize S. mansoni infection in the community. An area was defined as hyperendemic if the prevalence was ≥60% among schoolchildren; as mesoedemic if the prevalence ranged between 30% and 60%; and as hypoendemic if the prevalence was <30%.

To assess the capability of LQAS to correctly identify areas hyperendemic for S. mansoni, in each school the result observed with LQAS among the randomized children was compared with the prevalence calculated from the reference survey. According to the definitions, the critical value of prevalence to be tested was 60%. The sensitivity of LQAS (number of schools classified as hyperendemic by LQAS/number of schools found hyperendemic from the reference survey) and its specificity (number of schools classified as non-hyperendemic by LQAS/number of non-hyperendemic schools from the reference survey) were calculated.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

A total of 2437 pupils (50% males) aged 5–14 were enrolled in the study. The number of children was more than 30 in all schools but one which had only 29. The reference survey classified the 38 areas where schools were located as follows: hypoendemic, 7 (18.4%); mesoendemic, 13 (34.2%); and hyperendemic, 18 (47.4%) (Table 1).

Table 1.  Results of exhaustive examination by school
SchoolsNoStool examination positive (%)CI
  1. No: number of examined schoolchildren; CI: 95% confidence interval.

Vatobe404 (10.0)2.3–23.7
Androtsy Sud315 (16.1)5.5–33.7
Tolohomiady11319 (16.8)10.7–25.3
Anosibe479 (19.1)9.1–33.3
Nanarena PC4410 (22.7)11.5–37.8
Androtsy Nord5613 (23.2)12.9–36.4
Ambinda338 (24.2)11.1–40.9
Mahasoa Nord16550 (30.3)23.4–37.9
Ambararata Nord8634 (39.5)29.2–50.7
Misonjy3313 (39.4)22.9–57.9
Analavoka3715 (40.5)24.7–57.9
Nanarena Bat2913 (44.8)26.4–64.3
Anadabo5928 (47.5)34.3–60.9
Isifotra10150 (49.5)39.4–59.6
Lambomena6433 (51.6)38.7–64.2
Ankily8042 (52.5)41.0–63.8
Tritriva7037 (52.8)40.5–64.9
Menalafika5831 (53.4)39.9–66.7
Bemandresy6034 (56.6)43.2–69.4
Ihivoka7845 (57.7)46.0–68.8
Mahavelo4428 (63.6)47.8–77.6
Ivaro Est5032 (64.0)49.2–77.1
Ambia8758 (66.6)55.7–76.4
Sonjorano5638 (67.8)54.0–79.7
Morafeno4028 (70.0)53.5–83.4
Irina EPP8259 (71.9)60.9–81.3
Ambatolahy8361 (73.5)62.7–82.1
Irina EPC5844 (75.9)62.8–86.1
Ivaro Ouest4837 (77.1)62.7–88.0
Ampandratoka6350 (79.4)67.3–88.5
Sahambano9885 (86.7)78.4–92.7
Mahasoa Sud122107 (87.7)80.5–93.0
Kelivondraka3733 (89.2)74.6–97.0
Bevaho6762 (92.5)83.4–97.5
Voatavo4441 (93.2)81.3–98.6
Ankazobetroka3331 (93.9)79.8–99.3
Zazafotsy9691 (94.8)88.3–98.3
Ambinanitelo4545 (100)92.1–100
Total243710141423

All LQAS plans, including the (16, 6) one tested in the field, correctly identified the 18 hyperendemic areas. In most cases examination of about half of the n specimens of the sample was sufficient to reach and exceed the critical value d of positive stool samples, and testing could be stopped. LQAS plans (16, 6) (14, 5) (12, 4) and (10, 3) also correctly identified the seven hypoendemic zones. However, in mesoendemic areas LQAS often led to misclassification. The best global results were obtained with plan (16, 6) which correctly identified 27 schools among 38 (Table 2). The sensitivity of LQAS was 100% with all sampling plans except for trial 1 of plan (8, 2). The specificity ranged from 45% to 55% except for the two trials of plan (4, 0), whose specificity was 15% and 30%, respectively (Table 3).

Table 2.  Results of exhaustive examination (reference) and results of different LQAS plans
SchoolsType(16, 6)(14, 5)(12, 4)(10, 3)(8, 2)(6, 1)(4, 0)
12121212121212
  1. Exhaustive examination Type: H, hyperendemic (prevalence > 60%); M, mesoendemic (prevalence between 30 and 60%); h, hypoendemice (prevalence < 30%); 1 and 2, coding respectively, for the first and the second trial of a given sampling plan.

  2. LQAS method: C, Correctly identified by LQAS; I: misclassified by LQAS.

AMBATOLAHYHCCCCCCCCCCCCCC
 AmbinaniteloHCCCCCCCCCCCCCC
 BevahoHCCCCCCCCCCCCCC
 Ivaro EstHCCCCCCCCICCCCC
 Ivaro OuestHCCCCCCCCCCCCCC
 KelivondrakaHCCCCCCCCCCCCCC
 MenalafikaMIIIICIIIIIIIII
 MorafenoHCCCCCCCCCCCCCC
 Nanarena BeMIIIIICIIIICIII
 SonjoranoHCCCCCCCCCCCCCC
 TritrivaMIIICIIIIIIIIII
ANALAVOKAMCCCCIICIIICCII
 AmbindahCCCCCCCCCCCCII
 AnosibehCCCICCCCCCCCIC
 IsifotraMIICIIIIICIIIII
 MahaveloHCCCCCCCCCCCCCC
 MisonjyMCICCCICIICICII
 NanarenahCCCCCCCCCCCCIC
ANKILYMIIIIICIIIIIIII
 Ambararata NordMICICIIICIIIIIC
 AmbiaHCCCCCCCCCCCCCC
 AmpandratokanaHCCCCCCCCCCCCCC
 BemandresyMIIIIIIIIIIIIII
 AnadaboMIIIICCICICICII
 Androtsy NordhCCCCCCCCICCIII
 Androtsy SudhCCCCCCCCCCCCCC
 IhivokaMIIIIIIIIIIIIII
 Irina PCHCCCCCCCCCCCCCC
 Irina PPHCCCCCCCCCCCCCC
 LambomenaMIIIIIIIIICCIII
 TolohomiadyhCCCCCCCCCICIII
 VatobehCCCCCCCCCCCCCC
MAHASOA NORDMCCCCCCIICCICCC
SAHAMBANOHCCCCCCCCCCCCCC
ZAZAFOTSYHCCCCCCCCCCCCCC
 AnkazobetrokaHCCCCCCCCCCCCCC
 Mahasoa SudHCCCCCCCCCCCCCC
 VoatavoHCCCCCCCCCCCCCC
Table 3.  Sensitivity and specificity of different LQAS plans
PlansPlan (16, 6)Plan (14, 5)Plan (12, 4)Plan (10, 3)Plan (8, 2)Plan (6, 1)Plan (4, 0)
Sensitivity (%)
 Trial 110010010010094,44100100
 Trial 2100100100100100100100
Specificity (%)
 Trial 150555545455015
 Trial 250555545504530

Concerning operational aspects, all teachers performed easily and correctly the randomization of pupils according to the procedure given by the study staff. They did not exibit any unwillingness to transfer stool specimens into MIF tubes.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

The reference parasitological survey confirmed that the prevalence of S. mansoni infection was highly heterogeneous within the study area. Such a context, showing considerable variation among lots, inclines to test all the communities and fully justifies the use of LQAS.

The LQAS identified correctly all hyperendemic communities, even when using sampling plans having lowest values for n and d. On the other hand, communities with a prevalence <15% have been correctly recognized by all sampling plans. These findings are in accordance with observations made by other authors. Valadez et al. (1996) used LQAS during routine household visits to assess the technical quality of Costa Rican community-based health workers and concluded that LQAS could precisely identify quality at the two ends of a continuum: adequate or very inadequate. It was less sensitive to identify community-based health workers within the middle category. On the other hand, Singh et al. (1996), when comparing evaluation of immunization coverage in a primary health care centre area in Rajasthan using both LQAS and the 30-cluster sampling method recommended by WHO's Expanded Programme on Immunization (EPI), were faced by a homogeneous and low level of immunization coverage. They concluded that it constituted an impractical situation in which to apply LQAS, and the results obtained were therefore not particularly favourable to LQAS.

In our study, LQAS often failed to correctly identify as non-hyperendemic schools with prevalences between 45% and 60%, even when using our largest sampling plan (16, 6). This lack of statistical power in contexts of intermediate prevalences was especially marked when using low sample sizes. Wrongly identifying a community as hyperendemic may be a comparatively minor problem, considering that systematic praziquantel therapy would be beneficial to the inhabitants even at prevalence rates around 40%. However, low specificity could lead to large expense on antischistosomal drugs in areas with a high proportion of communities with medium prevalence. There is no all-purpose sampling plan (n, d) and it is highly advisable to obtain some basic knowledge on the epidemiological patterns of the infection in a given area prior to planning an LQAS survey and selecting a sample size n.

Computers allow easy calculation of potential sampling plans (n, d) and drawing of ROC curves in order to study what performances can be expected from each pair (n, d) in a given epidemiological situation. Hence pre-established tables that usually propose only a limited number of figures can be avoided, and it is possible to have a critical view of the sampling scheme. Lanata and Black (1991), in a review of advantages and constraints of LQAS in health surveys in developing countries, proposed as an alternative methodology to use a two-stage sampling scheme.

If the sampling strategy is to be cost-effective and fast, it must consist of delegating most of the work, except study design and laboratory testing. The progress of the study in schools confirmed that it was realistic to involve people working outside the field of public health, provided that they were given clear instructions on how to randomize. Most teachers did not show any reluctance to handle stool specimens. Lanata et al. (1990) in Peru also observed that it was possible to rely on local technicians for randomization when using LQAS.

We obtained a clear confirmation that LQAS is a valuable tool when a large number of communities needs be tested and resources to manage the survey are scarce. The sampling plan (n, d) must be chosen after careful consideration and taking into account previous epidemiological findings on the region. Although all sampling plans show very high sensitivity, the specificity may be very low especially when faced to intermediate prevalence if using smallest sample sizes. Making a decision will constitute an arbitration between satisfactory statistical power and logistical considerations.

Finally, it is worth noting that using LQAS is in accordance with WHO's aim to promote quick and inexpensive assessment methods to identify communities where schistosomiasis prevalence and morbidity are high (Chitsulo et al. 1995; Lengeler et al. 2000; Utzinger et al. 2000). LQAS is now routinely used for screening hyperendemic communities in the national schistosomiaisis control programme of Madagascar.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References

The study received support from the World Bank and from the French Fond d'Aide et de Coopération (FAC). We are grateful to the Ministry of Health, to the schoolchildren, to their teachers and to the villagers for their patience and co-operation. We thank Dr R. Migliani and Dr J.P. Manshande for helpful scientific discussion and invaluable advice and Dr Pascaline Ravoniarimbinina and Mr Théophile Rafamantanantsoa for technical assistance.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Objectives
  6. Study area
  7. Study population
  8. Statistical method
  9. Sampling method
  10. Parasitological method
  11. Data analysis
  12. Results
  13. Discussion
  14. Acknowledgements
  15. References
  • Boisier P, Ravaoalimalala VE, Seriey J et al. (1994) Infection àSchistosoma mansoni dans un foyer d'hyperendémie du moyen ouest de Madagascar: épidémiologie, morbidité et effets secondaires liés au traitement par praziquantel. Archives de l'Institut Pasteur de Madagascar 61, 4348.
  • Butterworth AE, Sturrock RF, Ouma JH et al. (1991) Comparison of different chemotherapy strategies against Schistosoma mansoni in Machakos District, Kenya: effects on human infection and morbidity. Parasitology 103, 339355.
  • Chitsulo L, Lengeler C & Jenkins J (1995) L'identification Rapide de Communautés Prioritaires Pour le Contrôle de la Schistosomiase Urinaire. OMS: Genève.
  • Jordan P, Webbe G & Sturrock RF (1993) Human Schistosomiasis. Cambridge University Press: Cambridge.
  • Lanata CF & Black RE (1991) Lot quality assurance sampling techniques in health surveys in developing countries: advantages and current contraints. World Health Statistics Quarterly 44, 113119.
  • Lanata CF, Stroh JRG, Black RE et al. (1990) An evaluation of lot quality assurance sampling to monitor and improve immunization coverage. International Journal of Epidemiology 19, 10861090.
  • Lemeshow S & Taber S (1991) Lot quality assurance sampling: single and double-sampling plans. World Health Statistics Quarterly 44, 115132.
  • Lengeler C, Makwala J, Ngimbi D et al. (2000) Simple school questionnaires can map both Schistosoma mansoni and Schistosoma haematobium in the Democratic Republic of Congo. Acta Tropica 74, 7787.
  • Moïse A, Salamon R, Commenges D et al. (1986) L'utilisation des courbes ROC. Note pédagogique. Revue d'Epidémiologie et de Santé Publique 34, 209217.
  • Ravaoalimalala VE, Rein J, Boisier P et al. (1994) Evolution de la bilharziose àSchistosoma mansonià Ampefy (Lac Itasy) de 1986 à 1993. Archives de l'Institut Pasteur de Madagascar 61, 4952.
  • Roux JF, Ravaoalimalala VE, Rakoto L et al. (1994) Bases de réflexions et grandes lignes du Programme National de lutte contre les bilharzioses à Madagascar. Archives de l'Institut Pasteur de Madagascar 61, 5861.
  • Singh J, Jain DC, Sharma RS et al. (1996) Evaluation of immunization coverage by lot quality assurance sampling compared with 30-cluster sampling in a primary health centre in India. Bulletin of the World Health Organization 74, 269274.
  • Tawfik Y, Hoque S & Siddiqi M (2001) Using lot quality assurance sampling to improve immunization coverage in Bangladesh. Bulletin of the World Health Organization 79, 501505.
  • Utzinger J, N'Goran EK, Tanner M et al. (2000) Simple anamnestic questions and recalled water-contact patterns forself-diagnosis of Schistosoma mansoni infection among schoolchildren in western Côte d'Ivoire. American Journal of Tropical Medecine and Hygiene 62, 649655.
  • Valadez JJ, Brown LD, Vargas WV et al. (1996) Using lot quality assurance sampling to assess measurements for growth country's primary health care system. International Journal of Epidemiology 25, 381387.
  • WHO (1993) The Control of Schistosomiasis. WHO Technical Report Series 830.

Authors Dr Pascal Boisier, Chef Unité d'Epidémiologie, CERMES BP 10887, Niamey, Niger. E-mail: pascal.boisier@cermes.ne Dr I. Jeanne, SIG Télédétection et santé, CERMES, BP 10887, Niamey, Niger. E-mail: ijeanne@cermes.ne M. A. Jutand, Institut de Santé Publique, d'Epidémiologie et de Développement, Université Victor Segalen, 146 rue Léo Saignat, 33076 Bordeaux Cedex, France. E-mail: Marthe-Aline.Jutland@isped.u-bordeaux2.fr Dr Léon Paul Rabarijaona, Unité d'épidémiologie, Institut Pasteur de Madagascar, BP 1274 101, Antananarivo, Madagascar. E-mail: leon@pasteur.mg (corresponding author). Dr Voahangy Elizabeth Ravaoalimalala, Divison Bilharziose Cysticercose, Ministère de la Santé, BP 1274, Antananarivo, Madagascar. E-mail: andriv@pasteur.mg Professor Jean-Felix Roux, CERMES, BP 10887, Niamey, Niger. E-mail: jeanfelix.roux@laposte.net Professor Roger Salamon, Institut de Santé Publique, d'Epidémiologie et de Développement, Université Victor Segalen, 146 rue Léo Saignat, 33076 Bordeaux Cedex, France. E-mail: Roger.Salamon.isped.u-bordeaux2.fr