|Year : 1996 | Volume
| Issue : 4 | Page : 207-211
Methods for estimating prevalence and incidence of senile cataract blindness in a district
J Bhattacharjee, Devadethan, RS Sharma, NK Saini, KK Datta
Epidemiology Division, National Institute of Communicable Diseases, Delhi, India
DHE, Epidemiology Division, National Institute of Communicable Diseases, 22 Shamnath Marg, Delhi: 110054
Source of Support: None, Conflict of Interest: None
The problem of senile cataract blindness (SCB) is very acute in India, contributing to 80 per cent of total blindness. The national objective of reducing the prevalence of blindness from 1.49 per cent, during 1986-89 to 0.3 per cent by AD 2000 has necessitated the surgical correction of SCB to be the major activity. With the introduction of District Blindness Control Society (DBCS), there has been a substantial increase in the number of operations of SCB in those districts. However, in the absence of standard, feasible, simple and cost effective methods to estimate the prevalence and incidence of SCB, the DBCS may find it difficult to plan and execute its major activity in a realistic way. The paper suggests two such methods for the use by DBCS. Only five seemingly rational assumptions have been adopted for the purpose. The authors feel that proper field testing is required to be sure about the reliability and validity of these methods.
Keywords: Senile cataract, blindness, prevalence, incidence.
|How to cite this article:|
Bhattacharjee J, Devadethan, Sharma R S, Saini N K, Datta K K. Methods for estimating prevalence and incidence of senile cataract blindness in a district. Indian J Ophthalmol 1996;44:207-11
|How to cite this URL:|
Bhattacharjee J, Devadethan, Sharma R S, Saini N K, Datta K K. Methods for estimating prevalence and incidence of senile cataract blindness in a district. Indian J Ophthalmol [serial online] 1996 [cited 2020 Jan 20];44:207-11. Available from: http://www.ijo.in/text.asp?1996/44/4/207/24562
About one-sixth of world's total blind, with visual acuity <3/60, live in India. A multi-centric surveyconducted jointly by the World Health Organisation (WHO) and National Programme for Control of Blindness (NPCB) in India during 1986-89, estimated the total blind in the country to be about 12 million, with visual acuity <6/60, giving a prevalence rate of 1.49%. About eighty per cent of this blindness was found to be due to senile cataract.
The main objective of NPCB is to reduce the prevalence of blindness from the current 1.49% to 0.3% by AD 2000. The concept of District Blindness Control Society (DBCS) has shown very good results in the project districts and is being implemented in phased manner throughout the country. These registered district level bodies with financial support directly from Federal Government, have been necessitated to organise and execute the blindness control activities in more coordinated and effective ways. As most of the blindness in the country is due to senile cataract, operative correction of this condition has become the major activity of DBCS. But, to be effective, each DBCS is expected to know about the prevalence and incidence of the disease in the district.
Prevalence of senile cataract blindness (SCB) can vary from district to district because of differential age-specific proportion of population, death rates, operation rates and incidence rates. Presently there is no uniform, reliable, feasible and cost-effective method of estimating the prevalence of the disease at district level.
Incidence rate of SCB, though a very useful indicator from programme point of view, cannot be estimated easily because of chronic and nonfatal course of the disease. In poliomyelitis, the lameness prevalence can be easily manipulated to calculate incidence rate of the disease. But, the situation here is quite different. On the one hand, one-fifth of prevalence can be taken as incidence, which is too simple to accept and on the other, mathematical models can be used, which are too complex to be understood by many health officials at the district level.
Keeping the above problems in mind, this paper suggests alternative and simple methods for estimating SCB prevalence and subsequently, its incidence rate by using the same prevalence data. It is felt that, the applicability of proposed methodologies would be wide and useful from public health point of view, particularly in the context of developing world.
| Materials and methods|| |
Estimation of prevalence
In conventional SCB survey representing a district, the sample size can be calculated by using the formula: n = (1.96)2pq/L2, where "p" = proportion of SCB in district population, "q" = 1-p and "L" = allowable error. Considering the national average of "p" = 1.2% (80 % of 1.49 %) and fixing "L" at 10% of "p", the population sample size becomes 31,629. The problem arises in selecting and examining 31,629 people representing the district. Though the task is not impossible, such exercise is neither operationally feasible and cost-effective, nor necessary from the view point of a field oriented method. Moreover, inclusion of people of all age groups in the sample will be unnecessary, as the disease mostly occurs in elderly and old people.
In this respect, three issues are to be sorted out. To restrict the age group, an assumption has been made that, the disease occurs only in ≥ 40 years. This restriction of age-group would lead to a substantial reduction in sample size. The sample size can be further reduced by taking an acceptably small error level in term of absolute percentage. To ease the problem of selecting a representative sample from the whole of the district, a modified cluster sample technique is proposed.
Considering the proportion of people ≥ 40 years in the population to be 22.3% in India and expected population proportion of SCB to be 1.2%, the expected proportion of SCB in people ≥ 40 years becomes 5.4%. The error level is proposed at 1% in absolute term (which is about 18.5 % of age specific prevalence). Thus the initial sample size, using the formula referred above, becomes 2000 persons of ≥ 40 years. Since cluster technique of sampling is proposed in this method, some design effect is expected due to the variability between cluster as units and individuals. Taking the expected design effect to be 2, the net sample size becomes 2000 X 2 = 4000. Regarding the number of clusters to be surveyed, 30 clusters being the minimum acceptable number, is proposed. Thus, 30 clusters covering a district are to be surveyed and 135 persons of ≥ 40 years age are to be examined for SCB in each cluster. The expected number of population to be covered in each cluster is 605.
For the purpose of survey, following case definitions are proposed. "Suspect case" is one in the target age group complaining of progressive and painless diminishing vision. History will be sufficient for this. "Probable case" is defined as any suspect case who is unable to read "E" card from ten feet distance (visual acuity corresponding to <6/60), as examined by a trained surveyor. A "Confirmed case" is any probable case showing central eye opacity similar to "cataract reference card" (a very good quality coloured photograph of senile cataract) on side by side comparison by trained surveyor. Both such cards have been developed by DANIDA Support Unit in India.
The total 30 clusters are to be distributed in urban and rural areas as per population proportion. The survey in urban area will be restricted to District Headquarters city, for making the process easy. All the municipal wards of the city will be listed and numbered. The required number of clusters will be selected randomly from this list. In each selected ward, a spot will be selected blindly from the city map. After reaching the spot, the first household will be selected by consulting numbers of a currency note and contiguous households surveyed till 135 persons of ≥ 40 years are examined for SCB. In rural area, all the subcentres will be listed and numbered. From this sample frame, the required number of cluster will be selected randomly. In the next step, all the villages under the selected subcentres will be listed and numbered separately and one selected randomly from each list. The first household in each selected village is then identified from a central spot randomly and contiguous households surveyed as mentioned above. If a selected village does not have 135 persons of ≥ 40 years, the nearest village within same subcentre will be taken in to account, and so on, to achieve the total of 135.
The most important aspect of this survey is the response rate in the selected households. Care has to be taken in terms of timing of survey and rapport of the surveyors with the community. Response rate of 90% has to be achieved. Only 10% substitution (14 out of 135) is allowed in each cluster. Ensuring correct and uniform technique of eye examination is another aspect of concern. Thorough and standarised training of the surveyors (paramedical/teachers) can solve this problem. All data can be collected in a structured proforma shown in [Table - 1].
Estimation of Incidence
Some convenient assumptions have been made in this method. The initial ones are that, the population under study is stable and maximum longevity of people is 89 years. The materials needed for the method include some primary and secondary data. The primary data on SCB prevalence can be obtained by conducting district based survey as proposed above. Secondary data pertain to age-group specific (≥ 40 years) population, proportion and death rates, as well as, same age-group specific number of SCB operated during the preceding year in the district. The population based data on age specific SCB operation in a district for a given year may not be readily available any where in India. However, such data from a DBCS, with good data base, can be the best approximation, where the proposed methodology can be applied initially.
From the above information age-group specific Decennial Cumulative Incidence (DCI) of SCB and age-group specific proportion of operations are calculated and applied, along with death rates on a hypothetical cohort of 1000 people aged 40 years and not having cataract. The additional assumptions include that, the annual SCB operation would continue as that of previous year and death rate among SCB is twice as that of general population.
Following this imaginary cohort for next fifty years through four sub-periods is considered sufficient to calculate the expected total number of new SCB cases that will develop during this total period. In this process, the cohort is also expected to be exhausted. Since the method is concerned to find out occurrence of incident SCB cases in the cohort, the persons who would die or develop the condition would be excluded from the cohort at the end of each sub-period.
Further corrections to account (i) for the number of annual SCB operated upon by using the proportions estimated earlier and (ii) for the SCB who died and not reflected in the age-group specific DCI concept, with a differential death rate will give the net incident cataract blind cases in the cohort.
| Results|| |
Estimation of Prevalence: Due to various constraints, the authors have not been able to field test the methodology. But, it has been tried to justify that, the methodology is workable and worthy of trial. Attempt is, therefore, made to share this concept with Epidemiologists and Programme Managers, some of whom may feel inclined to undertake the trial on these lines and arrive at the prevalence rate of SCB in their respective districts.
B. Estimation of Incidence: For the purpose of illustration, actual data have been taken from Rohtak district of Haryana state in India. Since age-specific proportions and death rates of district Rohtak were not available, state data [7,8] have been applied on the district in [Table - 2] along with the number of SCB found between April 1993 and March 1994 and number operated during 1993 (Source: District Health Office, Rohtak, Haryana, through Saini, N.K., personal communication) to find out the age-specific prevalence and DCI of SCB, as well as, population proportion operated for the condition. As per assumption, prevalence, as well as, DCI of senile cataract blindness in 40-49 years age group is 0.24%. In the next age group (50-59 years), 4.40% is the prevalence. But, 0.24 per cent of them had developed the condition when they were 40-49 years old. On removal of this quantum from 4.40%, only 4.16% of 50-59 years group actually developed the condition. Similar reallocations are done for other groups as well in [Table - 2], to fix the age group specific DCI of SCB cases. These yardsticks are then applied on the above mentioned cohort of 1000 persons just 40 years old, without any cataract and followed for next fifty years. The mechanics of the calculation that will help to estimate person years of observation and development of new SCB in the said cohort during the fifty years period of observation, are shown in [Table - 3]. Accordingly total person years of observation = (972 X 10) + (892 X 10) + (697 X 10) + (315 X 20) = 31910. [Table - 4] describes the mechanics of further corrections to estimate the net incident cases of SCB in the cohort.
Calculation for estimating incidence rate of senile cataract blindness in district Rohtak, Haryana State, India (1994):
| Discussion|| |
Estimation of Prevalence: The methodology proposed, like any other new one, has to address to different issues. The WHO-NPCB survey has shown that, the problem of senile cataract blindness is virtually absent before 40 years of age. Its prevalence starts rising sharply after 45 years of age. To net all such SCB in early age group, the cut off age under this assumption has been taken as 40 years. Increasing it to 50 years may lead to a requirement of smaller sample size for prevalence survey. But in the process SCB of the most priority age group, though small in number, would be missed.
As the prevalence of SCB is very low, it is difficult to propose, in this field oriented method, a very small error level while calculating the sample size, which may increase it by many folds. Thus, 18.5% error level of age specific prevalence (one per cent in absolute term) is considered reasonable. The issue of design effect of cluster sampling has been attempted to be solved by adopting an expected design effect of 2, which is usually an accepted one. Moreover, since in the present method, clusters will be randomly selected, the design effect is expected to be lower still.
Another important issue is the case definition of SCB. Before embarking on the survey of all 30 clusters, five clusters are to be selected randomly from already selected 30 clusters in such a way that, one should be from urban area and four from rural area. In these five selected clusters, an Ophthalmologist will accompany the survey team but, shall conduct independent eye examination for SCB, as per his expertise, of all 135 persons in each of the five selected clusters, who have been examined by the survey team. Expectedly, there will be 36 persons with SCB in the five clusters. The data of Ophthalmologist will be used as gold standard to calculate the sensitivity and specificity of the case definition. In a field oriented approach, a highly sensitive case definition is desirable. Therefore, if the sensitivity and specificity of case definition are found to be less than 90 and 75 per cent respectively, a reconsideration of case definition, critical review of training of surveyors, survey technique and supervision has to be done and necessary modifications made.
To find out the reliability of the method, its replications, in the same selected villages and city wards of the same district, have to be conducted by different independent survey teams, each team masked to the results obtained by the other teams. The first household in each cluster and the subsequent direction, as well as, households to be surveyed, have to be same, as far as possible, for all different set of survey teams. This will help to know about the extent to which the measurement made by proposed methodology is consistent. Subsequently, the 95 per cent "Confidence limits" of the results, obtained from surveys conducted in different districts, will tell about the precision of the measurement made, i.e. the ranges within which the population values would lie. The con sistency in narrow confidence limits is indicative of the validity of result.
Though all the issues mentioned above are one time problems, it is also felt important that, the survey should have a good service support to get co-operation from the community and take DBCS closer to the people.
Estimation of Incidence: Incidence rate of senile cataract blindness in Rohtak district, during 1994, has been found to be 460/100,000/year. Though in the present cohort approach assumptions of stable population and same surgery rate of SCB for next fifty years have been made, the result, however, pertains to the year of study, i.e. 1994 only. There is no way to extrapolate the same rate for future. A longitudinal study conducted in Central India, estimated the incidence rate of SCB to be 470/100,000/year. However, without field testing, similarity of such results does not guarantee about the validity and reliability of the proposed method. Measurement of incidence rate directly by longitudinal study can give precise estimate at the expense of time, cost and simplicity. The methodology suggested in this paper, though indirect, can give quick result without any additional cost and in a much simpler way.
The crux of the present senile cataract blindness problem in India lies in its backlog of operation. For planning an effective blindness control strategy, all DBCS should have clear idea about the load of disease, as well as, new cases that are expected each year in the district. Such district based information is more relevant in our country where, great amount of diversity is expected between the districts. Though SCB per se, does not cause increased death rate, disability caused by the disease and associated illnesses, increase the death rate in SCB. For this obvious reason, differential death rate (double) has been assumed for persons with SCB. This can be modified on the basis of expert opinion or actual data. As a practical approach, SCB prevalence survey for a district may be conducted at the beginning of alternate DBCS year. Calculating cataract blindness incidence rate from its prevalence data would be a very convenient tool for future planning of the major activity of DBCS. Once the methods are accepted by the Programme Managers, a simple computer software can be developed to calculate the incidence rate by feeding the required primary and secondary data.
| References|| |
Director General of Health Services. Govt. of India. Present status of National Programme for Control of Blindness (NPCB). 1992, pp 1, 81.
DANIDA Support Unit New-Delhi. India. District Blindness Control Society Manual, p 10
Park JE, Park K. Textbook of preventive and social medicine. (13th ed) Jabalpur, Banarasidas Bhanot. 1991, p 141.
Foster A. Worldwide blindness, increasing but avoidable. Seminars Ophthalmol. 8: 166, 1993.
Podgor MJ, Leske MC. Estimating incidence from age specific prevalence for irreversible diseases with differential mortality. Statistics in medicine 5:573, 1986.
Registrar General of India. Sample registration system. 1990, p 111.
Registrar General of India. Census of India. Series 6. Haryana. 1981, p 60.
Registrar General of India. Census of India 1991. Series 1. Part 1. 1992, p 38.
Minaassian DC, Mehra V. In: 3.8 million blinded by cataract each year. Reference materials. DANPCB, India, 1994.
Report of WHO consultation. Executive summary. The use of intraocular lenses in cataract surgery in developing countries. WHO/PBL/91.1. 1990, p 21.
[Figure - 1]
[Table - 1], [Table - 2], [Table - 3], [Table - 4]