Out-of-pocket Expenditures and Poverty: Estimates From NSS 61 st Round. Indrani Gupta Institute of Economic Growth Delhi

Transcription

1 DRAFT Out-of-pocket Expendtures and Poverty: Estmates From NSS 61 st Round Indran Gupta Insttute of Economc Growth Delh Paper presented for consderaton of the Expert Group on Poverty, Plannng Commsson 12 May 2009

2 Introducton: Out-of-pocket payment and poverty Health care fnance n developng and low ncome countres s stll predomnantly based on out-of-pocket (OOP) payments, and the lack of prepayment mechansms lke nsurance. In the absence of nsurance, an llness not only reduces welfare drectly, t also ncreases the rsk of mpovershment due to hgh treatment expendtures. It s now wdely acknowledged that health care expendtures can drve ndvduals and households nto poverty. The lterature around out-of-pocket payments and ts mpact on the economc status of households has grown tremendously over the past. Catastrophc payments for health are defned n relaton to the household resources, most often proxed by aggregate consumpton. A smple rato of health expendture to consumpton expendture can be used to estmate how hgh the health spendng of households s n comparson to ther total consumpton. A threshold of 10 percent s commonly used wth the ratonale that above ths the household may be forced to sacrfce other basc needs, sell productve assets, ncur debt or become mpovershed (Pradhan and Prescott 2002, Ranson 2002, Wagstaff and van Doorslaer 2003, Russell 2004). A semnal work on catastrophc health expendtures n 59 countres publshed by Lancet (Xu et al (2003) ndcated that there was wde varaton n the proporton of households facng catastrophc payments from out-of-pocket health expenses. The authors dentfed three key precondtons for catastrophc payments as the avalablty of health servces requrng payment, low capacty to pay, and the lack of prepayment or health nsurance. The authors concluded that ndvdual, partcularly n poor households, can be protected from catastrophc health expendtures by reducng a health system's relance on out-ofpocket payments and provdng more fnancal rsk protecton. Another study from Burkna Faso (Su et al 2006) dentfed the key determnants of catastrophc health expendture as economc status, household health care utlzaton especally for modern medcal care, llness epsodes n an adult household member and presence of a member wth chronc llness.

3 A gven overall share of out-of-pocket fnancng may represent relatvely lttle fnancal rsk to households f t s low and s dstrbuted more or less proportonally to capacty to pay. However, that s not the case and a mult-country analyss (Musgrove and Zeramdn 2001) ndcates that at low ncomes, the out-of-pocket share s hgh on average, and extremely varable, from about 20 to 80% of all health spendng. Wth ncreasng ncome, not only does the average share fall sharply, but the range narrows. Several studes of Indan vllages to determne why households descent nto poverty (Krshna 2004, Krshna et al 2005, Krshna 2006) fnd that n a majorty of cases of declne nto poverty, three prncpal factors are at work: health expenses, hgh-nterest prvate debt, and socal and customary expenses. Irrespectve of dstance to health care faclty, health care expenses fgured promnently n more than half of all cases of declne nto poverty. The dscusson, debates and evdence around the effect of OOP payments on health and poverty outcomes was so ntense that n 2005 the Member States of WHO adopted a resoluton encouragng countres to develop health fnancng systems amed at provdng unversal coverage. Unversal coverage was seen as a system desgned to obtan access for all to approprate promotve, preventve, curatve and rehabltatve servces at an affordable cost. Insuffcent health coverage, poor health servces, low publc health spendng all go to determne the level and extent of OOP n countres. The poverty-nducng effect of OOP expendtures has also led to a sgnfcant lterature on methodologcal ssues around estmaton of poverty that takes nto account health expendture. Ths paper revews the exstng methodology, proposes a methodology for Inda, presents some prelmnary estmates usng data from the 61 st round of the NSS, and also dscusses the concerns that reman n measurement and estmaton of both health expendtures and poverty.

4 Poverty and out-of-pocket health expendture: standard methodology It s now commonly acknowledged that standard poverty measures do not adequately reflect the health needs of ndvduals. A study n 11 countres of Asa estmated that 78 mllon people n Asa are not currently counted as poor despte the fact ther per capta household expendture net of health expendture falls below the extreme poverty threshold of $1 per day (Van Doorslaer 2006). The two key questons that need to be addressed n ths context are the followng: How to measure poverty takng nto account out-of-pocket (OOP) health expendtures? How does the head count raton or other poverty measures change when OOP s taken nto account? Based on earler papers and research, an expert team led by the World Bank Insttute has come out wth operatonal gudelnes enttled Analyzng Health Equty Usng Household Survey Data brought by the World Bank (O Donnell et al 2008). It gves n detals the optons avalable to practtoners and researchers for adjustng measures of poverty to take nto account expendture on health care. Subsequently, many researchers have used the more standard methodology descrbed n detal below. The estmates n ths note are based on ths methodology as well. Smply put, poverty needs to be measured takng nto account OOP spendng, snce health spendng s also now vewed as an essental expendture that enhances welfare lke food and other necesstes. The problem s that health spendng can sometmes be a functon of ncome, and may not always be essental. Thus, excludng all health spendng from total expendture to assess poverty can result n overestmaton of poverty. On the other hand, not ncludng any health expendture wll underestmate poverty, especally f the non-dscretonary part s qute sgnfcant, as t often s n developng countres. As contended by O Donnell and others, there are two condtons under whch the dfference between poverty estmates derved from household resources gross and net of OOP payments may approxmate the effect of health expendture on poverty. These are when:

5 (a) OOP payments are completely non-dscretonary, and (b) total households resources are fxed. In real lfe, nether of these two condtons s met; health expendtures are often a functon of ncome, especally at the upper tal of the ncome dstrbuton. Also, households augment ther resources n a varety of ways to meet unexpected expendtures. The other queston s: should poverty lnes be adjusted to reflect ncluson of essental health expendtures? The adjustments that have to be made to poverty estmates n the presence of sgnfcant health expendture would depend to a certan extent on how the poverty lne s beng calculated. If the poverty lne s calculated based on subsstence needs only,.e. an absolute poverty lne, then there may be no justfcaton of adjustng ths lne takng nto account health expendtures. If on the other hand, a relatve poverty lne s beng calculated, based on the mean or medan household expendture, there s certanly a justfcaton n adjustng t based on health expendture. The other problem s that the health expendtures are hghly stochastc across ndvduals and over tme, and vary sgnfcantly dependng on soco-economc characterstcs. Thus, the concern around the representatve ness of mean or medan health expendture always remans. However, t has been suggested that the poverty lne may be adjusted downwards by the mean of health spendng of households whose total expendture s about the same as the poverty lne (Wagstaff and van Doorslaer 2003). Measurng health expendture: measurement ssues A methodology s as good as the estmates t uses: measurng health expendture s fraught wth measurement errors, mostly because of the many factors that determne who spends how much and on what. a. Acute vs chronc condtons: acute llnesses and chronc llnesses need to be separated and treated dfferently. Whether a poverty lne should take nto

6 consderaton expendtures for long-term care that are less unpredctable than acute expendtures, or should focus only on sudden expendtures due to unforeseen llnesses s a pont that needs to be settled before makng any calculatons. b. Reference perod: whle the reference perod s often algned wth the type of llness (acute or chronc), t need not always be so. A 30 days or 15 days recall perod versus a 365 days recall perod would gve very dfferent estmates, and one has to have an operatonal rule about how to annualze the short recall perod estmates and how to reduce the long recall perod estmates to a monthly fgure. c. Hosptalzaton vs out-patent care (OPD): Hosptalzaton s generally an unantcpated event, and cannot be attrbuted to everyone n the sample. Thus, treatment of expendture from hosptalzaton needs to be handled wth cauton. OPD expendture on the other hand s probably the most general of aggregates and safest to use n calculatons lke poverty estmates. However, hosptalzatons often have a more severe mpact on poverty than OPD, and therefore, need to be taken nto account under sutable assumptons. d. Items of health expendture: health expendture can be broadly dvded nto the followng tems: drugs & medcnes, consultatons, dagnostcs, hosptal stay & related tems, medcal applances and devces used, and other mscellaneous expendtures. The relatve weght of each of these tems n the total health expendture may be necessary to calculate before makng a decson on whether or not all these tems should be treated as essental. e. General vs health-specfc surveys: fnally, t must be noted that aggregates generated from general surveys of consumpton expendture are often qute dfferent from aggregates generated from detaled temzed questons of specfc surveys on health. To better understand the varablty on estmates for these varous categores, we take data from the 60 th and 61 st NSS rounds; whle the 61 st round s a standard Consumer Expendture Survey (CES), the 60 th round s for health. Also, the 60 th round was for the year 2004, whereas the 61 st round s for Snce the tme perods are qute

7 close, these two surveys are used to demonstrate why cauton and care need to be taken to calculate health aggregates that wll then be used to re-estmate poverty fgures. Table 1: Health aggregates obtaned from NSS 60 th and 61 st round Varable Reference 60 th NSS 61 st NSS perod January June July 2004-June % reportng hosptalzaton 365 days 2.4% 9% Rural 10% - Urban % reportng hosptalzaton 30 days NA 1.4% - Rural 1.5% -Urban Per capta hosptalzaton expenses, over all who reported hosptalzaton 365 days Rs Rural Rs Urban Rs Rural Rs. 958 Urban Per capta hosptalzaton expenses, over entre sample 365 days Rs. 151 Rural Rs 303 Urban Rs 120 Rural Rs 204 Urban Per capta hosptalzaton, over entre sample 30 days NA Rs. 11 Rural Rs. 14 Urban % reported OPD treatment 15 days 89% - Urban 82% - Urban NA % reportng OPD treatment 30 days NA 61% -Rural 63% - Urban OPD expenses over those who reported OPD treatment 15 days Rs 322 Rural Rs 385 Urban NA OPD expenses, over entre sample 15 days Rs. 21 Rural Rs. 31 Urban NA OPD expenses 30 days NA Rs. 27 Rural Rs Urban Share of drugs n OPD expenses 15 days 63% - Share of drugs n OPD 30 days NA 82% Share of drugs n hosptalzaton 365 days 25% 41% A quck glance at the table ndcates that utmost cauton that needs to be exercsed n calculatng health expendtures. For example, hosptalzaton dffers sgnfcantly dependng on the reference perod and dependng on the rate of hosptalzaton, the average amount spent on hosptal expenses wll also be very dfferent. Percent reportng OPD expenses were also qute dfferent n the two rounds, though the rounds are only

8 about a year apart. Smlarly, share of tems of expendtures also vary sgnfcantly across the two rounds. Poverty estmates are based on consumer expendture surveys. Therefore, the frst crteron of selectng a health expendture aggregate has to be that t must come from the same survey. Whle the 60 th health round also has consumpton expendture, t s more lke a rapd survey of consumpton expendture. In other words, 60 th round has detaled health and bref consumpton expendture, whle the 61 st round has bref health and detaled consumpton expendture. Snce poverty calculatons need detaled consumpton expendture, the health aggregates must perforce come from the same survey. In prncple t s possble to adjust the health expendture fgures by lookng at the patterns from the detaled health survey, whch can be done n subsequent analyses. Poverty calculatons are done based on Rupees per capta per month. Therefore, any health expendture should be as close to a monthly average expendture fgure per person. The dffculty les here: snce not the entre sample s gong to be sck, the average expendture over all those who reported sck s gong to be very hgh, and cannot be taken as the norm for the entre populaton. On the other hand, usng the entre survey populaton to arrve at a per person fgure may underestmate the total expendture an average household undertakes for the entre year on health. Ideally, total health expendtures reported over the entre year and not for the past 15 or 30 days should be used to make any calculatons regardng health expendture. However, snce the Plannng Commsson poverty fgures are based on the Unform Recall Perod (URP) of 30 days, the health expendtures wll have to be comparable and based on URP too. Table 2 presents the 3 man aggregates and ther values calculated based on the 61 st round of NSS. Interestngly, the medcal nsttutonal expendtures for both the recall perods gve almost the same estmate.

9 Table 2: Medcal expendture, NSS 61 st round Health aggregates NSS 61 st round Per capta estmates n Rupees all Inda Medcal nsttutonal expendture n the last 365 days 143 Medcal nsttutonal expendture n the last 30 days 12 Medcal non-nsttutonal expendture n the last 30 days 35 Methodology for health-adjusted poverty estmates The most common methodology for adjustng poverty lnes to take nto account health expendture s one offered by Wagstaff and Doorslaer (2003) and subsequently compled by the World Bank Insttute as mentoned above. Ths methodology s presented n detal below: Suppose that the poverty head count s calculated gross of out-of-pocket (OOP) payments. In other words, f household expendtures nclude health payments, the head count rato H gross can be wrtten n the followng manner: N gross gross = s = p 1 gross H N, where p s = 1 IF x < PL and s 0 otherwse, where s s the = 1 household sze. N s the number of households n the sample, and PL s the poverty lne. The net of health payments head count s obtaned by replacng net p wth p = 1 f gross ( x T ) PL (and 0 otherwse), where T s health expendture. < gross gross The poverty gap gross of health payments s g = p PL x ) and the mean of ths ( gap s gross G N gross s = g 1 = N. The net of health payment poverty gap can be obtaned by usng s = 1

10 g net net = p PL ( x T )) n the second equaton, where T s the out of pocket ( expendture. One can also normalze the poverty gap on the poverty lne such that gross G NG gross =, PL where the mean of ths G H gross gross MPG =, gves the ntensty of poverty. The net of gross payments normalzed gap can be obtaned smlarly. It can be argued that the poverty lne tself should also be adjusted downwards f poverty s to be estmated net of OOP payments. Ths can be done n cases where the poverty lne s nclusve of health needs. Absolute poverty lnes do not requre such adjustments. However, for poverty lnes that are relatve and hgher than the absolute poverty lnes, there may be some reason to adjust the lne downwards, especally f the assumpton s that a majorty of ndvduals get addtonal funds to cover ther health needs. One opton suggested by Wagstaff and van Doorslaer (2003) s to subtract from the poverty lne the average health spendng of households wth total expendture n the regon of the poverty lne. For Inda, the poverty lnes are estmated based on subsstence nutrtonal requrements, and also there very lttle addtonal resources for health for the majorty of the populaton (only 10% of the populaton have any form of health coverage). Thus, no adjustment s requred to the poverty lnes whle calculatng poverty net of OOP payments. Ths paper essentally uses the same methodology wthout adjustment to the poverty lne - to estmate the health-expendture adjusted poverty estmates usng the Consumer Expendture Survey data of the 61 st round of NSS survey. An earler paper (Garg and Karan 2008) also uses the same methodology to estmate the effect of out-of-pocket health expendtures on poverty, usng the Consumer Expendture Survey (CES) of NSS n ts 55 th round. These earler results are presented here as well for the sake of comparson.

11 Results The analyss uses the poverty lne calculatons arrved at by the Plannng Commsson to arrve at revsed poverty fgures. The results are presented n Table 3. Whle the methodology of net consumpton expendture s adopted, the poverty lne s not revsed. One reason for dong that comes from the 60 th health round of the NSS, whch ndcates that less than 1 percent of those who reported an alment n the past 15 days, and less than 0.5 percent of those who reported hosptalzaton n the last 365 days had any sort of rembursement for ther treatment. Table 3: Estmates of poverty wth and wthout health expendture adjustments Varable Rural Urban a. Head count rato 28.3% 25.6% b. Poverty gap (Rs) c. Health-expendture adjusted head count rato 31.9% 28.5% d. Poverty gap, health-expendture adjusted e. Percentage ncrease n poverty (c-a) f. Increase n poverty gap (d-b) As can be seen from the table, poverty ncreases by 3.6% and 2.9% for rural and urban areas respectvely when OOP spendng s adjusted for. The poverty gap show how much would have to be transferred to the poor to brng ther expendture up to the poverty lne, and the table ndcates that ths amounts ncreases for both rural and urban areas, and an addtonal Rs 3.5 and Rs. 4.9 per capta per month s the ncrease n the poverty gap because of OOP payments. Clearly, these fgures do not reflect how many poor ndvduals are made poorer by OOP, whch s also an mportant dmenson of poverty. Comparson wth estmates from NSS 55 th round (Garg & Karan 2008) Based on the 55 th round of CES data of the NSS, Garg and Karan use the same methodology to arrve at poverty estmates. These fgures are presented below n Table 4 along wth the current estmates from Table 3.

12 Table 4: Estmates from 55 th and 61 st rounds: a comparson (Garg & Karan) (from Table 3) Varable Rural Urban Rural Urban a. Head count rato b. Poverty gap (Rs) c. Health-expendture adjusted head count rato d. Poverty gap, healthexpendture adjusted e. Percentage ncrease n poverty (c-a) f. Increase n poverty gap (d-b) Snce both sets of estmates use the same methodology, these are comparable. The table shows that both rural and urban poverty have ncreased between the two rounds, as have the health expendture adjusted poverty. Interestngly, the gap between rural and urban poverty s more when health expendtures are taken nto account. Rural poverty ncreases slghtly more than urban poverty n both the perods, when health expendtures are adjusted for. Further, there was a 3.3% dfference between urban and rural poverty n the 55 th round, whch ncreased to 4.2%. The mpact on poverty gaps for both rural and urban areas n the latter perod s somewhat more pronounced than ts mpact on poverty head count rato. These results ndcate that OOP spendng s more poverty-nducng n the rural than the urban areas, and ts mpact on poverty has ncreased over the years. Inter-state varaton n health-expendture adjusted poverty Table 5 below presents the poverty head count rato for the states. Snce poverty rato of Assam s used for Skkm, Arunachal Pradesh,Meghalaya, Mzoram,Manpur,Nagaland and Trpura, only result for Assam s presented here. Smlarly, the poverty ratos of Taml Nadu s used for Pondcherry and A & N Island, that of Goa and Kerala are used

13 for Daman & Du and Lakshadweep respectvely, and that of Punjab s used for Chandgarh. Thus, the results are only presented for the prmary states here. Table 5: State-wse estmates of Head Count Rato wth & wthout adjustment for heath expendture Rural Urban No Adjusted % No Adjusted % State adjustment for health ncrease adjustment for health ncrease expendture expendture Andhra Pradesh Assam Bhar Chattsgarh Delh Goa Gujarat Haryana Hmachal Pradesh Jammu & Kashmr Jharkhand Karnataka Kerala Madhya Pradesh Maharashtra Orssa Punjab Rajasthan Taml Nadu Uttar Pradesh Uttarakhand West Bengal Dadra & N. Havel There are two ponts to note from ths table: the frst s that as expected, the ncrease n poverty when one takes nto account health expendture s almost always hgher n rural than urban areas wth the excepton of Andhra Pradesh and Rajasthan, where urban adjusted poverty ncreased more than rural adjusted poverty. Thus, rural health expendtures are more poverty-nducng than urban health expendtures. Secondly, some states have sgnfcant ncreases n poverty, whereas for some others health expendture does not make too much of a dfference to the poverty estmates.

14 For example, for Uttar Pradesh and Madhya Pradesh the ncrease n rural poverty s 5.8 and 5.4 percent respectvely. For Uttar Pradesh the ncrease s not as hgh for urban areas, but for Madhya Pradesh urban poverty also ncreases by 4.6 percent, one of the hghest among all the states. Interestngly, Kerala has hgh ncreases n both rural and urban poverty when health expendtures are adjusted for. Overall, rural poverty ncreases the most for the EAG states wth the excepton of Maharashtra, West Bengal and Kerala. The pcture s slghtly more vared for ncreases n urban poverty. These results are somewhat consstent wth earler fndngs from Garg and Karan: they also concluded that UP showed hgh ncrease n poverty, but whereas Bhar showed very hgh ncrease n ther calculatons, Bhar has a relatvely more modest mpact on poverty due to OOP spendng. Health expendture n total consumpton expendture Overall, 4.7% of total household expendture s spent on OOP health spendng. However, there s wde varaton across states, wth the poorer states showng much hgher health spendng than the other states. Table 6 presents the nter-state varatons n proporton spent on health, by rural-urban resdence to complete the pcture on povertynducng effect of OOP spendng. The table shows very hgh OOP spendng n states lke Uttar Pradesh, Chattsgarh, Kerala, Maharashtra and West Bengal. Most of these states also end up n the basket of states that have sgnfcant ncreases n poverty due to OOP spendng, except Rajasthan to some extent. Clearly, hgh proporton of OOP spendng across states does not necessarly ndcate hgh ncreases n poverty; however, snce the overlap s very hgh, t ndcates that n these states, the burden of hgh spendng s probably mostly n the lower quntles of the expendture dstrbuton, whch n turn ncreases poverty n these states.

15 Table 6: OOP as a percentage of total expendture States Rural Urban Total Andhra Pradesh Assam Bhar Chattsgarh Delh Goa Gujarat Haryana Hmachal Pradesh Jammu & Kashmr Jharkhand Karnataka Kerala Madhya Pradesh Maharashtra Orssa Punjab Rajasthan Taml Nadu Uttar Pradesh Uttarakhand West Bengal Dadra & N. Havel Summary and conclusons The standard methodology developed by the World Bank team for analyzng poverty nduced by out-of-pocket expenses was used n ths paper to estmate the lkely ncrease n poverty. Data from the Consumer Expendture Survey of the 61 st round of the NSS was used to arrve at OOP health expendtures, whch were then accounted for whle estmatng poverty. The analyss showed ncreases n poverty by as much as 3.6 and 2.9 percent for rural and urban Inda respectvely, f OOP health expendtures are accounted for. These estmates are hgher compared to the estmated mpact on poverty calculated from the 55 th round of the CES of the NSS.

16 The state-wse pcture also ndcates that most states wll experence sgnfcantly hgher poverty f OOP s taken nto account, wth the EAG states beng affected the most. However, states lke Kerala, Maharashtra and West Bengal are also among those states that are most affected; these are also states that have hgh proporton of health spendng. Inda currently has about 10 percent of ts populaton covered by some form of health nsurance. In the absence of health nsurance, the effect of hgh OOP expendture wll clearly mpact on poverty, pushng especally those who are slghtly above poverty lne nto poverty, and those already below poverty lne, nto further mpovershment. Whle poverty estmates need to take nto account OOP spendng to make the estmates meanngful, t s also equally mportant to push polcymakers to ntate programmes and polces to extend health coverage to a larger number of ndvduals. The challenge n health coverage s to be able to fnd a way to cover the nformal and unorganzed sector workers and ther dependents. Whle many schemes have been consdered and launched, the success rates have been very low, and Inda remans one of the countres wth least health coverage for those who need t the most. Tll the tme such a mass extenson of health coverage to take nto account catastrophc expenses occurs, health wll contnue to be an addtonal factor that nduces poverty, n addton to employment status and wages.

17 References Van Doorslaer, E., O. O Donnell, R. P. Rannan-Elya, A. Somanathan, S. R. Adhkar, D, Harbanto, C.G. Garg, A. N. Hern, M.N. Huq, S. Ibragmova, A. Karan, C.-w. Ng, B.R. Pande, R. Racels, S. Tao, K. Tn, L. Trsnantoro, C. Vasasvd, and Y. Zhao. Effect of Health Payments on Poverty Estmates n 11 Countres n Asa: An Analyss of Household Survey Data. The Lancet. 368(14): O Donnell, Owen, Eddy van Doorslaer, Adam Wagstaff, and Magnus Lndelow Analyzng Health Equty Usng Household Survey Data: A Gude to Technques and ther Implementaton. World Bank Insttute. World Bank Xu Ke, Davd B Evans, Ke Kawabata, Radh Zeramdn, Jan Klavus, Chrstopher J L Murray. Household catastrophc health expendture: a mult-country analyss. The Lancet Vol 362 July 12, 2003 Su Tn Tn, Bocar Kouyaté, & Steffen Flessa. Catastrophc household expendture for health care n a low-ncome socety: a study from Nouna Dstrct, Burkna Faso. Bulletn of the World Health Organsaton. Volume 84, Number 1, January WHO. Sustanable health fnancng, unversal coverage and socal health nsurance [A58/33]. Geneva: Musgrove Phlp and R. Zeramdn. A Summary Descrpton of Health Fnancng n WHO Member States. Commsson on Macroeconomcs and Health Workng Paper Seres Paper No. WG3 : 3. Date: June 2001 Krshna, Anrudh. Escapng Poverty and Becomng Poor: Who Gans, Who Loses, and Why? World Development Vol. 32, No. 1, pp , 2004 Krshna, Anrudh, M Kapla, M. Porwal and V. Sngh. Why Growth s not Enough: Household Poverty Dynamcs n Northeast Gujarat, Inda. Journal of Development Studes, Vol. 41, No. 7. October Krshna, Anrduh. Pathways Out of and Into Poverty n 36 Vllages of Andhra Pradesh, Inda. World Development Vol. 34, No. 2, pp , 2006 Pradhan, M and N. Prescott. Socal Rsk Management Optons for Medcal Care n Indonesa. Health Economcs 11: Ranson, M.K. Reducton of Catastrophc Health Care Expendtures by a Communtybased Health Insurance Scheme n Gujarat, Inda: Current Experences and Challenges. Bulletn of the World Health Organzaton. 80(8): Russell, S. The Economc Burden of Illness for Household n Developng Countres: A Revew of Studes Focusng on Malara, TB and HIV/AIDS. Amercan Journal of Tropcal Medcne and Hygene 71 (Supp 2)

18 Wagstaff, A and E. van Doorslaer. Catastrophe and Impovershment n Payng for Health Care: Wth Applcatons to Vetnam Health Economcs 12:

19