|Year : 2012 | Volume
| Issue : 7 | Page : 14-22
Index based mapping of high risk behaviours for HIV among female sex workers in India
Vasna Joshua1, MD Gupte1, Rajatashurva Adhikary2, Ramesh S Paranjape3, Mandar K Manikar3, G.N.V. Brahmam4, J Mahanta5, BM Ramesh6
1 National Institute of Epidemiology (ICMR), Chennai, India
2 Family Health International, New Delhi, India
3 National AIDS Research Institute (ICMR), Pune, India
4 National Institute of Nutrition (ICMR), Hyderabad, India
5 Regional Medical Research Centre (ICMR), Dibrugarh, India
6 Karnataka Health Promotion Trust, Bangalore, India
|Date of Submission||02-Aug-2010|
|Date of Web Publication||1-Dec-2012|
National Institute of Epidemiology (ICMR), R137, 3rd Avenue, Tamil Nadu Housing Board, Ayapakkam, Chennai 600 077
Source of Support: None, Conflict of Interest: None
| Abstract|| |
Background & objectives: Integrated Behavioral and Biological Assessment (IBBA) study is the first cross-sectional survey to study large number of covariates of HIV/STI (sexually transmitted infection) in India. Generally, districts are identified as of HIV high or low based on its prevalence. Instead, it would be optimal to label the districts based on several high-risk related covariates in the concurrent set up. The objectives of the present study were to obtain an index for each district, to discover 'natural' clusters and a map with Kriged estimates.
Methods: The study population consisted of 10461 female sex workers (FSWs) from 29 sites spread over 24 districts from five HIV high prevalent States. Covariates based on demographic characteristics, sexual practices, knowledge of HIV/STI and biological variables were studied. The analyses were done on weighted estimates based on principal component analysis, cluster analysis and Kriging technique. Five factors were extracted and improved using varimax rotation and standardized factor scores obtained. Natural clusters in a multivariate setting were identified. Each district was expressed as geographic co-ordinates and using the standardized scores the Kriged estimates were obtained.
Results: The proxy determinants were 'never used a condom', 'wanted to use a condom but did not use', 'experience of condom breakage' and 'current STI that needs a doctor'. Dimapur district stood first rank demanding the greatest attention. The cluster analysis branded Dimapur, Warangal, Prakasam, and Chittoor districts as a cluster, which required greatest attention and kriged estimates showed the high-risk concentrated regions as Andhra Pradesh, Maharashtra and northeast region.
Interpretation & conclusions: The results of this study may help the programme officials and policy managers to concentrate on the key factors, and districts/regions, which need greater attention in the order of priority.
Keywords: Factor analysis - female sex workers - HIV - Kriging technique - mapping
|How to cite this article:|
Joshua V, Gupte M D, Adhikary R, Paranjape RS, Manikar MK, Brahmam G, Mahanta J, Ramesh B M. Index based mapping of high risk behaviours for HIV among female sex workers in India. Indian J Med Res 2012;136, Suppl S1:14-22
|How to cite this URL:|
Joshua V, Gupte M D, Adhikary R, Paranjape RS, Manikar MK, Brahmam G, Mahanta J, Ramesh B M. Index based mapping of high risk behaviours for HIV among female sex workers in India. Indian J Med Res [serial online] 2012 [cited 2020 Sep 29];136, Suppl S1:14-22. Available from: http://www.ijmr.org.in/text.asp?2012/136/7/14/104166
Worldwide, AIDS is an equal opportunity disease for women  . Women are generally more susceptible in contracting the HIV infection simply because of their receptive nature. According to WHO, in India there are a large number of impoverished and disabled women who are not economically sound and hence they are forced to find opportunities for their survival or betterment. Behavioural factors of such women getting indulged in unprotected sex (without the use of condoms) and with multiple sexual partners place them at the greatest risk of contracting HIV and other sexually transmitted infections (STIs). According to National AIDS Control Organization (NACO), there are 8.3 lakh female sex workers (FSWs) in India  . Initially NACO demarcated Indian States as high, medium and low prevalence States and later as Category A, B, C and D based on the HIV prevalence rate (based on antenatal care attendees) in the districts, for prioritization of programme implementation  . When several variables related to HIV/STI transmission are available, it would be ideal to identify the risk status of the States/districts by considering the multiple high risk related covariates. Hence an attempt was made to (i) find an index or a score for each district surveyed, based on multiple high-risk related covariates of HIV/STI concomitantly; (ii) get a map based on 'natural' clusters (districts) in a multivariate set up; and (iii) obtain a map overlay of India with kriged estimates.
| Material & Methods|| |
Study type: Integrated Behavioural and Biological Assessment (IBBA) study is the first cross-sectional survey to study both the behavioural and biological variables of HIV/STI in India. Numerous covariates of HIV/STI transmission for FSWs were studied.
Study population: Five highly endemic States for HIV/STI were identified namely Andhra Pradesh, Karnataka, Maharashtra, Tamil Nadu and Nagaland. The study population consisted of 10,461 respondents from 29 high risk groups from different sites (hereafter referred to as sites) spread over 24 districts surveyed among the above five States. Forty-four covariates of HIV/STI transmission for FSW based on demographic characteristics, sexual history, condom practices, knowledge and awareness of HIV/STI and biological variables were examined. The study period varied for each site from August 2005 to September 2006. Female sex workers were defined as any female 18 yr or older, who sold sex for money at least once in the preceding one month of the survey.
Sampling strategy: A probability sampling method was used in all sites. A conventional cluster sampling was used when numbers and persons do not change like brothels; a time-location cluster (TLC) sampling was used for street-based FSWs and respondent-driven sampling (RDS) was used for populations in which FSWs did not congregate in identifiable locations and the population is hidden like bar girls.
The target sample size for FSWs was 400 per group, per district. There were fewer than 400 members in case of less population in the district or when the refusal rate was high.
The methodology, data collection, ethical consent, weighting procedure, laboratory methods, etc. are discussed elsewhere , .
Statistical analysis: Principal component analysis (PCA) was used to find optimal ways of combining variables into a small number of subsets and factor analysis was used to identify the structure underlying such variables and to estimate scores. Our attempt was to make the original data set into relatively a smaller number of independent factors and to find the estimates of factor scores, which is a linear combination of standardized indicators ,, .
All the estimates used in the analysis were weighted based on the inverse probability of selection. The original data set contained 44 high-risk related covariates. These covariates were examined using the correlation matrix. Fifteen highly correlated covariates were identified and used for further analysis. Majority of the FSW populations were highly concentrated in the five selected States. The final data set used in the analysis was of size 29 (sites) x15 (covariates) [Table 1].
|Table 1: Percentage distribution of respondents according to surveyed districts of India, 2005-2006 by 15 covariates|
Click here to view
Using the method of principal components a smaller subset of five factors extracted (Eigen value greater than one). These factors were improved using varimax rotation and factor scores obtained. Then the initial score for each site was obtained using per cent variation as weights on factor scores. The initial scores were standardized for comparative purposes. The above analysis was done using the SPSS software  .
Agglomerative methods like average linkage method, complete linkage method and Ward's minimum variance method were used. The methods started with a matrix of distances between the dissimilarity of the covariates and the sites were grouped using one of the above methods. The square of the Euclidean distance measure was used.
Cluster analysis dendograms were used to portray the 'natural' clusters of districts (districts which are similar in characteristic of the variables). The optimum cluster was done using the methods, which gave similar results. Dendograms were obtained using SYSTAT software  . The clusters were ranked on the basis of average standardized scores. Mapping of the natural clusters for the 24 districts (average was used if more than one site) was done using ArcGIS software  .
Kriging technique is an earth science technique based on regionalized variable theory  . Using the method of least squares, it provides a means of interpolating values for points not physically sampled using the available information and the spatial arrangement of the data set. The standardized score (as proportion) for each surveyed district was expressed as Zobs (xi, yi) where (xi, yi) are the geographic co-ordinates (latitude and longitude). Zobs(xi, yi) were modeled by means of a variogram using geo-statistical (kriging) technique and the kriged estimates were obtained using the ArcGIS software package  .
| Results|| |
Bartlett's test of sphericity was found to be highly significant (P<0.0001). Thus the indicators selected for the analysis were well correlated as required for factor analysis to be valid. The five factors extracted together explained about 82 per cent of the total variation. This was considered to be a satisfying result for factor analysis. The five factors identified by factor analysis as by priority, factor loadings that are larger (>0.6) are listed in [Table 2].
Factor I. The condom practices (d) and currently suffering from STI that require doctor (g), Factor II. sexual partners (e), Factor III. biological variables (a), Factor IV. demographic variables (b), and Factor V. sexual debut (c) (alphabets as explained in [Table 1]).
Initial score for various States are listed in [Table 3] wherein the last column represents the corresponding standardized score in the descending order. Seventeen sites had an index of above 50 (more than the average), indicating need for greater interventional care to bring down the HIV/STI transmission in the surveyed sites.
|Table 3: The standardized scores for the districts surveyed in India 2005-2006|
Click here to view
All the three agglomerative methods gave four as the minimum optimal number of natural clusters required to describe the variability of the data in a multivariate set up [Table 4]. The average linkage method and Ward's minimum variance method gave similar results. In literature these two methods out performed compared to other methods , . It identified Dimapur, Warangal, Prakasam, and Chittoor districts as a cluster (average score 87.9) under one umbrella which required greatest attention, Yavatmal, Pune and Kolapur as a second priority cluster (average score 67.5), Belgaum, Guntur, East Godavari, Karimnagar, Hyderabad, Bellary and Visakhapatnam as a third cluster (average score 64.9) and the remaining districts as the fourth cluster (average score 34.4). The mapping revealed a portrait of natural clusters in a multivariate set up [Figure 1].
|Figure 1: Natural clusters based on several high risk related covariates of FSWs in selected districts of India, 2005-2006.|
Click here to view
|Table 4: Clusters obtained using square of the Euclidean distance and various agglomerative methods using covariates of HIV for Indian States|
Click here to view
The map obtained [Figure 2] gave an optimal unbiased representation of multiple covariates of HIV/STI transmission with Kriged estimates. It depicted the regional variation and the HIV/STI high risk concentrated regions (hot spots), and regions at the greater risk of developing the infection. The kriged estimates showed the high-risk concentrated regions as central part of India with a few hot spots in Andhra Pradesh, Maharashtra and most of northeast region.
|Figure 2: Kriged estimates based on the analysis of several high risk related covariates of FSW in India, 2005-2006.|
Click here to view
| Discussion|| |
The districts surveyed in the study were selected based on the size of the high-risk populations (FSW). All the districts except Chennai were prioritized for programme implementation as category-A by NACO. The present exercise revealed that the proxy determinants are the condom practices (never used a condom, wanted to use a condom but did not use and experience of condom breakage) and current STI symptoms that need a doctor. According to NACO, "women who do not have the power to discuss sex and HIV risk with their husbands or sexual partners may not be able to avoid HIV infection"  . A study among Vietnamese female sex workers examined predictors of HIV testing and revealed that successful negotiation of condom use and self-reported sexually transmitted infections (STIs) had a protective effect  .
Greater care should be focused towards safer sexual practices (supportive environment like condom usage policy in labelled brothel areas as recommended by NACO), effective use of condoms and early treatment of STIs (sexual health) would bring down the HIV/STI infection to a greater extent.
Dimapur was the only district studied in the north east region, the FSWs in this region were highly hidden, stood first in rank and needs the greatest attention or intervention. In terms of the prevalence rate, this district was identified as a medium care district (below median value), whereas, when several high risk related variables were considered, it needed greatest attention. Dimapur has already been labelled as a commercial hub of Nagaland and owing to its unique location in the foothills and population comprising people from different regions of the India, fuel the different manifestations of STI and HIV infection in the region  .
All the districts surveyed in Andhra Pradesh stood above the average score and in the second priority for intervention. According to Andhra Pradesh AIDS Control Society (APSACS), HIV/AIDS in these districts had already reached at alarming proportions due to trading sex centres and floating population  . Warangal had the second highest score of 88 and this district is well known for highest HIV positivity among antenatal clinic (ANC) attendees and STD clinic attendees  .
In Maharashtra, most of the districts surveyed needed greater care and among those Yavatmal and Mumbai (brothel based area) districts deserved greater priority. As such the district Yavatmal had the second highest HIV prevalence of 37.3 per cent and the high-risk related variables were on the higher side. The above fact about Mumbai was emphasized by NACO by stating that "Mumbai sits at the epicentre of the HIV/AIDS epidemic - the city is home for the largest brothel-based commercial sex work in India"  .
In Karnataka, the district Belgaum attained the top priority of rank 5 and needed greater attention. The State's first case of HIV was identified in this district and it has highly-concentrated areas of high risk groups  , and has been marked 'red' on the HIV/AIDS map of the country .
In Tamil Nadu, Dharmapuri and Salem districts needed more care compared to Madurai, Chennai and Coimbatore. These two districts are already identified as areas with high intensity of HIV/AIDS  . Similar hierarchy was found among the HIV high prevalent states even in the general population, which has been reflected by the National Family Health Survey (NFHS-3) data  .
The districts identified as a group using cluster analysis showed agreement with the index. The map showed three hot spots, which are found in the northeast region, Andhra Pradesh and Maharashtra. According to the mapping of Roberts  , Nagaland, Manipur, the coastal Andhra Pradesh, northern Karnataka, and southern Maharashtra are considered as hot spots for commercial sex work and HIV/AIDS.
Enlightening the FSWs about the importance of their well being, making right decisions, safe sexual practices and immediate attention in treating the current STI may bring down the HIV transmission to a greater extent. The score together with the mapping exercise will be helpful for the policy makers and programme officials to identify regions that require additional interventions and to devise more efficient strategies for the reduction of HIV/STI infection in India.
This analysis, besides identifying the proxy determinants, quantified the health care needs. The usage of binary indicators in the principal component analysis might include biases to the covariance structure, and hence the factor loadings, and smaller reported proportion of explained variance  . This could be one of the limitations of the study. In this study the newly emerging technique of earth science was attempted to a unique data set available. If finer grid points like more number of districts (spread all over India) surveyed, it would have been more precise to pin point the pockets in the country for remedial action.
| Acknowledgment|| |
The authors acknowledge Bill and Melinda Gates Foundation for their financial support. The authors thank the IBBA study team at National AIDS Research Institute (NARI), Pune; National Institute of Nutrition (NIN), Hyderabad; National Institute of Epidemiology (NIE), Chennai; National Institute of Medical Statistics (NIMS), New Delhi; Karnataka Health Promotion Trust (KHPT), Bangalore; Regional Medical Research Council (RMRC), Dibrugarh and Family Health International (FHI) for active participation in the study.
| References|| |
|1.||Cline RJW, McKenzie NJ. Women and AIDS: The lost population. In: Parrott RL, Condit CM, editors. Evaluating women's health messages: A resource book. Thousand Oaks, CA: Sage Publications; 1996. p. 382-401. |
|2.||National AIDS Control Organization, 2004. Available from: http://www.nacoonline.org/NACO, accessed on March 30, 2009. |
|3.||Prioritisation of districts for programme implementation, NACO. Available from: http://www.nacoonline.org/Quick_Links/HIV_Data/, accessed on March 30, 2009. |
|4.||Saidel T, Adhikary R, Mainkar M, Dale J, Loo V, Rahman M, et al. Baseline integrated behavioural and biological assessment among most at-risk populations in six high-prevalence states of India: design and implementation challenges. AIDS 2008; 22 (Suppl 5): S17-34. |
|5.||Ramesh BM, Moses S, Washington R, Isac S, Mohapatra B, Mahagaonkar SB, et al. Determinants of HIV prevalence among female sex workers in four south Indian states: analysis of cross-sectional surveys in twenty-three districts. AIDS 2008; 22 (Suppl 5): S35-44. |
|6.||Manly BFJ. Multivariate statistical methods - A primer. 3 rd ed. New York: Chapman & Hall; 1986. p. 76-89. |
|7.||Sekhar CC, Indrayan A, Gupta SM. Development of an index of need for health resources for Indian States using factor analysis. Int J Epidemiol 1991; 20 : 246-50. |
|8.||Bhagavandas M, Joshua V. Mapping co-variates of mortality upto age of five years for Indian states; Indian J Public Health 2003; 47 : 22-6. |
|9.||Statistical package for the social sciences (SPSS) for Windows (version 14), Chicago, Illinois, USA: SPSS Inc, 2006. |
|10.||SYSTAT 8.0 for windows: Unparalleled research quality statistics and graphics statistics (Copyright 1998 by SPSS Inc.). Bangalore: Cranes Software International; 1998. |
|11.||Environmental Systems Research Institute (ESRI), ArcGIS Desktop: Release 9.2. Redlands, CA, USA: Inc.: ESR; 2006. |
|12.||Carrat F, Valleron AJ. Epidemiologic mapping using the "kriging" method: application to an influenza-like illness epidemic in France. Am J Epidemiol 1992; 135 : 1293-300. |
|13.||Indrayan A, Kumar R. Statistical Choropleth Cartography in Epidemiology. Int J Epidemiol 1996; 25 : 181-9. |
|14.||Bansal AK, Indrayan A. A computer based statistical study of cartography in mortality upto age of one year. Indian Pediatr 1998; 30 : 1251-8. |
|15.||National AIDS Programme Management, implementation of HIV prevention care and treatment strategies. module 6, 2007. Available from: http://www.searo.who.int/LinkFiles/Publications_Preliminar__pages.pdf, accessed on March 30, 2009. |
|16.||Grayman JH, Nhan DT, Huong PT, Jenkins RA, Carey JW, West GR, et al. Factors associated with HIV testing, condom use, and sexually transmitted infections among female sex workers in Nha Trang, Vietnam. AIDS Behav 2005; 9 : 41-51. |
|17.||Ovung Mhabemo E. Research on different manifestations of STDs/HIV/AIDS in one of the high prevalent districts in India. Int Conf AIDS 2004; 15: abstract no. B10008. - Bangkok, Thailand. |
|18.||The Hindu. December 31, 2008. Available from: http://www.hindu.com/2008/12/31/stories/2008123154330600.htm, accessed on March 30, 2009. |
|19.||Frontline. Vol. 21 (5), February 28 - March 12, 2004. Available from: http://www.hinduonnet.com/fline/fl2105/stories/20040312003109600.htm, accessed on March 30, 2009. |
|20.||Saadhan HIV/AIDS Helpline - Mumbai, India, the communication Initiative Network. Available from: http://www.comminit.com/en/node/118860/347, accessed on March 30, 2009. |
|21.||The Hindu. March 1, 2009. Available from: http://www.hindu.com/2009/03/01/stories/2009030152540300.htm, accessed on March 30, 2009. |
|22.||The Hindu. December 2, 2007. Available from: http://www.thehindu.com/2007/12/02/stories/2007120252100300.htm, accessed on March 30, 2009. |
|23.||Azad N. The Independent Commission for People's Rights and Development) concept paper "Trafficked female labour: coping mechanisms for poor households". Available from: www.hawaii.edu/global/projects_activities/Trafficking/_Azad.doc, accessed on March 30, 2009. |
|24.||International Institute for Population Sciences (IIPS), National Family Health Survey (NFHS-3), 2005-2006. Mumbai: IIPS; 2006. |
|25.||Robert S. HIV in India - A complex epidemic. N Engl J Med 2007; 356 : 1089-93. |
|26.||Kolenikov S, Angeles G, Socioeconomic status measurement with discrete proxy variables: Is principal component analysis a reliable answer?: Available from: http://web.missouri.edu/~kolenikovs/papers/polychoric-technical-4.pdf, accessed on March 30, 2009. |
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4]