1 National Institute of Epidemiology, Mayor VR Ramanathan Road, Chetpet, Chennai600 031,Tamil Nadu, India. E-mail: nieicmr{at}vsnl.com
2 Directorate of Rural and Medical Health Services, Tamil Nadu, India
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods A representative sample of 64 000 people drawn from eight districts of Tamil Nadu state, India, with maximum allowable number of 25 cases was considered, using LQAS methodology to test whether leprosy prevalence was at or below 7 per 10 000 population. Expected number of cases for each district was obtained assuming Poisson distribution. Goodness of fit for the observed and expected cases (closeness of the expected number of cases to those observed) was tested through 2. Enhancing factor (design effect) for sample size was obtained by computing the intraclass correlation.
Results The survey actually covered a population of 62 157 individuals, of whom 56 469 (90.8%) were examined. Ninety-six cases were detected and this number far exceeded the critical value of 25. The number of cases for each district and the number of cases in the entire surveyed area both followed Poisson distribution. The intraclass correlation coefficients were close to zero and the design effect was observed to be close to one.
Conclusions Based on the LQAS exercises leprosy prevalence in the state of Tamil Nadu in India was above 7 per 10 000. LQAS method using clusters was validated for monitoring leprosy elimination in high endemic areas. Use of cluster sampling makes this method further useful as a rapid assessment procedure. This method needs to be tested for its applicability in moderate and low endemic areas, where the sample size may need increasing. It is further possible to consider LQAS as a monitoring tool for elimination programmes with respect to other disease conditions.
Accepted 28 August 2003
Smallpox, poliomyelitis, guinea worm disease, tuberculosis, and leprosy are some of the diseases that have been targeted for eradication or elimination. The concept of eradication is clear as it aims at removal of the disease-causing agent and hence total absence of infection or disease. Theoretically it is conceivable to plan for evaluation of an eradication programme. Elimination is just a higher level of control, with the expectation that the disease will disappear in due course when the force of infection is brought below a threshold level. However, there is hardly any scientific evidence to predict what this threshold level, in terms of prevalence, should be for diseases like leprosy or tuberculosis. Measurements for elimination programmes are therefore not easy. In the context of leprosy, rapid and standard methods are needed to monitor progress towards the goal of elimination, at the country level in smaller countries and at sub-national levels in large countries, where leprosy has been endemic.1 Once low levels of prevalence of leprosy are reached, the health administrators' concern will shift to targeting smaller areas such as a state or region with high leprosy prevalence for necessary intervention and programme strengthening. Programme-based case detection data may not identify these areas. Traditional sample surveys in each state or region are laborious, expensive, and time-consuming. Programme managers need rapid assessment techniques for monitoring the programmes through an independent mechanism.2
Using real life data, we have earlier demonstrated that lot quality assurance sampling (LQAS) could be of particular value in initially high endemic areas for leprosy elimination monitoring.3,4 There are certain methodological issues to be addressed before it could be recommended for implementation. We assumed that leprosy cases were more or less evenly distributed in high endemicity situations and the number of cases followed Poisson distribution (curve for rare events). Random sample selection of individuals for leprosy examination is operationally inconvenient in the field. We considered a two-stage survey with households as the sampling units for operational convenience. The above critical factors needed to be examined using real life data before considering the applicability of LQAS for monitoring progress towards leprosy elimination.
LQAS is a quality control tool in industry. Items or goods of the same type, size, or grade in a production process are grouped to form homogeneous lots. A simple random sample of n items from each lot is taken for inspection to identify the number of defectives. If the number of defectives is less than or equal to maximum acceptable number (critical value) then the lot is regarded as having defective items within acceptance limits. On the other hand if the number of defectives exceeds the critical value on or before the complete examination of all the items in the sample, further examination of the sample items is discontinued and the lot is rejected. A lot in the health field may be a sample of people. More details of LQAS are available elsewhere.5
The use of LQAS has gained importance in the health field over the last 15 years. However, its use for leprosy elimination monitoring has been suggested only recently. After the introduction of leprosy elimination programmes, based on fixed duration multidrug therapy (MDT), a rapid decline in leprosy prevalence has been documented in high endemic regions. For example, in Tamil Nadu state in south India leprosy reportedly dropped from 53 per 10 000 in 1990 to 4 per 10 000 in 2001.6 The objective of this paper is to examine the validity of LQAS by using real life data from an initially highly leprosy endemic state, for leprosy distribution and for the design effect (enhancing factor for the sample size) if a two-stage cluster sample survey is employed.
![]() |
Materials and Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
Sample size, definitions, and examination procedures
The sample size that we provided for was adequate to conclude that the prevalence levels could be between 4 and 7 per 10 000 for the state. Details of the design are available from our earlier work.3,4
The sample size, assuming Poisson approximation, with 5% level of significance and a power of 90% required to test the hypothesis, was determined as 53 000 people. Allowing 20% as the margin of non-response, a sample of 64 000 people in the state was considered sufficient. The maximum allowable number of leprosy cases (d, critical value) in this sample was 25. If the number of cases in the sample exceeded the critical value (d) of 25, prevalence in the state was to be regarded as above 7 per 10 000.
For the purpose of sampling, Tamil Nadu state was divided into North, South Central, and West Zones. To have a representative sample in the state, two districts in each of the four zones were selected using probability proportional to size (PPS) linear systematic sampling. In each district 8000 people were to be selected. This sample was further divided into rural and urban populations, proportionate to the rural/urban ratio of the population in the district. Villages and towns were selected using PPS linear systematic sampling. Within each selected village/town, a fixed number of 100 households was selected using linear systematic sampling. Thus, in order to cover the entire sample 18 villages/towns were selected. The first household in each village/town was randomly selected and subsequent households were obtained following systematic sampling.
Information regarding number of households for each village or enumeration block was obtained. Sampling interval (K) was computed by dividing the number of households in the village or enumeration block by 100. A random entrance/exit to the village or enumeration block was selected. A random number below K for the household was further selected. Using the random number identified the first household, from this household every kth household was selected till 100 households were attained. Reasons for choosing a household and not the individual were (1) sampling frames of individuals were not available and (2) selection and examination of only certain individuals in a household would be practically difficult.
The survey was conducted from 6 January 2001 to 30 March 2001. Information on age, sex, and residential status of every member in the household was collected from the head of the household or any senior member of the family through a structured questionnaire. It was decided to complete the survey for the entire sample to generate data on leprosy distribution. Leprosy inspectors screened all the resident members of the household that were present at the time of the first and repeated visits by clinical examination for leprosy. A case of leprosy is defined as an individual who has manifestations of leprosy and who needs treatment. Supervisory staff confirmed the diagnosis of the cases. All the cases were classified by type, i.e. single skin lesion (SSL), paucibacillary (PB), or multibacillary (MB). A case was defined as PB if he/she had 5 skin lesions, with not more than one nerve lesion, and all the slit skin smears were negative for acid fast bacilli (AFB). A person was considered as an MB case if he or she had >5 skin lesions or more than one nerve lesion or the individual's skin smear was positive for AFB. Leprosy patients with a single patch receiving single dose treatment were not considered for prevalence, but, if put on 6-months PB MDT, they were considered in prevalence. New cases were those who were diagnosed for the first time through the present survey. An old case is defined as one who is already diagnosed and is under treatment. The Directorate of Medical and Rural Health Services of Tamil Nadu and the National Institute of Epidemiology, Chennai monitored data quality.
Five to six experienced paramedical staff members were formed into teams to cover a given district. Each team consisted of one health educator, three to four leprosy inspectors and one laboratory technician from the state government. Laboratory technicians were involved in collecting slit-skin smears. Medical officers from the Directorate of Public Health, Tamil Nadu and the National Institute of Epidemiology, Chennai monitored the quality of clinical examination.
Statistical analysis
To investigate whether leprosy cases in the sample survey area followed Poisson distribution the following procedure was adopted. Frequency distribution for the number of households (clusters) with 0, 1, 2, 3, and 4+ cases in the survey was obtained. Expected distribution of households was calculated from the Poisson probability distribution. The null hypothesis, that the observed frequency distribution of households by the number of cases follows the Poisson distribution, was tested through 2 (goodness of fit). A similar procedure was adopted for each district included in the survey.
To examine whether implementation of the two-stage cluster sampling design, instead of simple random sampling as envisaged in LQAS, had any impact in terms of increase in the standard error of the estimated prevalence the following procedure was adopted. Using the spreadsheet containing the population and number of cases by cluster, the estimated prevalence and the standard error for each district were computed. The intraclass correlation coefficient (measure of variability between the households as compared with variation within the households) was computed. Using the intraclass correlation coefficient and the cluster size, the design effect (D), was computed. A similar procedure was adopted for the total sample of eight districts.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
From Table 2 it is clear that the coverage was about 98% in children. The coverage was more (96%) among adult females and for adult males it was 80%. Twenty-nine cases detected in the survey were re-examined by two medical officers and for 28 of them diagnosis was confirmed (Table 3).
|
|
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
LQAS aims to provide information with respect to prevalence levels of the disease in a state or region. It does not provide estimates of the prevalence. LQAS thus needs to be undertaken periodically, rather than as a one-time effort, to help programme managers to understand whether progress is being made over a period of time. WHO, in its review, indicated that there was increasing application of the LQAS technique for assessing a variety of health care parameters including disease incidence.7 Recently, LQAS was used for identifying high-risk districts of neonatal tetanus, to assess and validate whether neonatal tetanus has been eliminated in Rajasthan state in India.8
In this study we tried to examine all the individuals in the sample. However, some people were not available at the time of visits, or they had temporarily migrated, or they were seriously ill, or were not willing to undergo the examination. These non-respondents could be a crucial group. We were successful in keeping the proportion of non-response to less than 10%. We presume non-response is probably not related to any stigma attached to leprosy because generally in an endemic state like Tamil Nadu, it is very low. However, no efforts could be made to find the reasons for non-response.
Trained and experienced leprosy inspectors were deployed for the present survey. The population of Tamil Nadu is generally receptive to routine periodical screening for leprosy as part of leprosy control activities. However, omission of a certain proportion of early leprosy cases cannot be ruled out. This problem might lead to misclassification and underestimation of the actual level of prevalence.
There were 13 known cases of leprosy in the examined population, indicating that the known prevalence of leprosy was only 2.3 per 10 000 in the population examined. This is far below the expected prevalence of 7 per 10 000 in the state. On the other hand even if only PB or MB leprosy cases are considered, the prevalence in the examined population is 11.5 per 10 000.
The number of leprosy cases in the population of each district and the combined selected districts followed Poisson distribution. We made a similar observation earlier in an endemic district of Tamil Nadu.4 The Poisson distribution of leprosy in endemic areas appears to be true in view of the low prevalence of leprosy in the statistical sense.
The near-zero intraclass correlation coefficients (variability among the cases) for the households indicated that distribution of leprosy among households was similar. This observation is similar to our previous exercise in a hyper endemic district of Tamil Nadu state.4
While carrying out LQAS we adopted operationally convenient two-stage cluster-sampling design in place of simple random sampling. A recent application of LQAS for neonatal tetanus elimination also used a cluster sampling procedure instead of simple random sampling.7 Since the observed design effect in our study was nearly one, there is no need to increase the sample size in similar situations. Thus pragmatic modifications to the LQAS design using cluster sampling, are a feasible proposition.
KEY MESSAGES
|
![]() |
Acknowledgments |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Mittal BN. The National Leprosy Eradication Programme in India. World Health Stat Q 1991;44:2329.[Medline]
3 Gupte MD, Narasimhamurthy B. Lot quality assurance sampling (LQAS) for monitoring leprosy elimination programme. Int J Lepr 1999;67:14349.[ISI]
4 Murthy BN, Subbaiah M, Boopathi K, Ramakrishnan R, Gupte MD. Lot Quality Assurance Sampling (LQAS) for monitoring leprosy elimination in an endemic district in Tamil Nadu. Indian J Lepr 2001;73:1119.[Medline]
5 Lemeshow S, Taber S. Lot quality assurance sampling: single and double sampling plans. World Health Stat Q 1991;44:11532.[Medline]
6 Masood Ahmed TH. National Leprosy Eradication Programme, Tamil Nadu: trend in leprosy. In: Gupte MD (ed.). Report on the Workshop on Impact of MDT on Trend of Leprosy. Madras, India: Janatha Printing & Publishing Co. Pvt. Ltd, 1994, pp. 2123.
7 Robertson SE, Anker M, Roisin AJ, Macklai N, Engstrom K, LaForce M. The lot quality technique: a global review of applications in the assessment of health services and disease surveillance. World Health Stat Q 1997;50:199209.[Medline]
8 Evaluation of neonatal tetanus elimination in Rajasthan-India. Wkly Epidemiol Rec 2003;78(5):2532.[Medline]