Out-of-Pocket and Catastrophic Expenditure on Health in Cambodia. Cambodian Socio-Economic Surveys 2004, 2007 & 2009 Analysis

Transcription

1 Out-of-Pocket and Catastrophic Expenditure on Health in Cambodia Cambodian Socio-Economic Surveys 2004, 2007 & 2009 Analysis

2 As a federally owned enterprise, we support the German Government in achieving its objectives in the field of international cooperation for sustainable development. Items from named contributors do not necessarily reflect the views of the publisher. Published by Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH Registered offices Bonn and Eschborn, Germany Friedrich-Ebert-Allee Bonn, Germany Phone: Fax: Dag-Hammarskjöld-Weg Eschborn, Germany Phone: Fax: info@giz.de Internet: Cambodian-German Social Health Protection Programme PO Box 1238, Phnom Penh, Cambodia Phone: Fax: Sector Project Providing for Health (P4H) p4h@giz.de Responsible Adélio Fernandes Antunes Photo credits Ursula Meissner Cambodia, April 2014

3 Out-of-Pocket and Catastrophic Expenditure on Health in Cambodia Cambodian Socio-Economic Surveys 2004, 2007 & 2009 Analysis

4 iv

5 Acknowledgements This report was produced under the overall direction of Dr Lo VeasnaKiry, Director of the Department of Health Planning and Information, Ministry of Health, Royal Government of Cambodia. The principal writers of the report were Dr Piya Hanvoravongchai, Dr Lo VeasnaKiry, Mr Ros Chhun Eang and Mr Adelio Fernandes Antunes. The secondary statistical analysis for this report was performed by Dr Piya Hanvoravongchai. The helpful comments and support from Mr Ros Chhun Eang, Chief of the Bureau of Health Economics and Financing of the Ministry of Health, and Ms Priyanka Saksena, are gratefully acknowledged. Primary data and documentation for the analysis underlying this report were provided by the National Institute of Statistics, Ministry of Planning, Royal Government of Cambodia, under the leadership of His Excellency Director General San Sy Than. The researchers are grateful for the facilitation of Mrs Birgitta Mannflet and Mrs Tiina Orusild. We are grateful to the participants of the Consultative Workshop on the Finding of the Secondary Analysis of CSES 2009, held at the Phnom Penh Hotel on 31 January, 2012, who provided helpful comments for the revision of the draft analysis report. Technical support for this report was provided by the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH on behalf of the German Federal Ministry for Economic Cooperation and Development, and by the World Health Organization. This study was part of GIZ and WHO s collaboration in the Providing for Health (P4H) partnership. P4H is a global network for Universal Health Coverage (UHC) and Social Health Protection (SHP). - Disclaimer: All reasonable precautions have been taken by the authors, contributors and their institutions to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either expressed or implied. The responsibility for the interpretation and use of the material lies with the reader. The present report was finalised in December The report was edited by Mr John Paul Nicewinter. Design, layout and illustration were done by Mr Justin Pearce- Neudorf, CamPOP Media ( com), under the coordination of Mr Itay Noy. The Forward and Executive Summary were translated from English to Khmer by Dr Ung Bunthoeun and reviewed by Dr Chhiay Song. Acknowledgements v

6 vi

7 List of Acronyms CBHI Community based health insurance CDHS Cambodian Demographic Health Survey CSES FFC GDP GIZ HEF HH Cambodian Socio-Economic Survey Fairness of financial contribution (index) Gross Domestic Product Deutsche Gesellschaft für Internationale Zusammenarbeit GmbH Health equity fund Household HSP2 Second Health Strategic Plan MOH OD OOP P4H RGC SFHF SHP USD WHO Ministry of Health Operational district Out-of-pocket expenditure(s) Providing for Health Royal Government of Cambodia Strategic Framework for Health Financing Social health protection US dollar World Health Organization List of Acronyms vii

8 Contents Acknowledgements List of Acronyms v vii Executive Summary 1 Introduction 1 Methods 1 Results 1 Policy Implications 3 Introduction 7 Methods 9 Illness and health care-seeking behaviour 9 Place of treatment or care 9 Household health expenditure 12 Out-of-pocket health expenditure 12 Health-related transportation expenses 13 Household capacity-to-pay and poverty level 13 Household consumption expenditure 13 Food expenditure 13 Household subsistence spending 13 The household s capacity-to-pay 13 Poverty level 13 Catastrophic health spending, impoverishment, and medical indebtedness 14 Out-of-pocket health expenditure as a share of household capacity-to-pay 14 Catastrophic health expenditure 14 Fairness in Financial Contribution Index 14 Impoverishment 14 Indebtedness through illness 14 Sub-group analysis 15 Economic quintile 15 Urban and rural groups 15 Fee exemption and health insurance status of households 15 Operational districts with health equity funds 15 Other sub-groups 15 Determinant analysis 15

9 Results Illness and health care-seeking behaviour Incidence of illness over the previous month 17 Reported type of illness 17 Incidence of illness by gender, age group, and economic quintile Incidence of health care-seeking behaviour 17 Health care-seeking behaviour among population groups 18 Health care fee exemption and health care-seeking behaviour 18 Determinants of care-seeking: control for illness, age, sex Choice of health care provider 24 Health care utilisation by facilities and services 24 Health care utilisation by different facilities and services for population groups 24 Multiple visits Health care spending Out-of-pocket expenditure 32 Out-of-pocket expenditure per capita among population sub-groups 32 Average out-of-pocket expenditure per household 32 Health insurance membership and health spending 34 Share of average out-of-pocket expenditure per household by age group across quintiles 34 Share of total cumulative out-of-pocket expenditure by population sub-group and year Determinants of health spending 36 Determinants of positive out-of-pocket health expenditure 36 Determinants of out-of-pocket expenditure level Household capacity-to-pay Catastrophic expenditure and poverty impact Catastrophic incidence Impoverishment Determinants of catastrophic health payments Debt from illness Incidence of indebtedness through illness Determinants of indebtedness as a result of illness 46 Discussion 49 Health care-seeking behaviour 49 Health care fee exemption, health equity funds 50 Health care spending and implications 51 References 53 Annex 55

10

11 Executive Summary Introduction The Royal Government of Cambodia (RGC) is aware of the challenges that direct and indirect health expenditures represent for the population, and has supported and begun a number of social health protection (SHP) initiatives targeting different segments of the population while working towards universal health coverage. The largest and moststudied of these initiatives are health equity funds (HEFs), which covered over one million poor beneficiaries in Other smaller initiatives include community based health insurance (CBHI). The RGC has also initiated the establishment of social health insurance funds for the formal sector, which are expected to launch in Data from studies such as the Cambodian Socio-Economic Survey (CSES) provide an opportunity to analyse the health care utilisation, out-of-pocket health payments (OOP), and catastrophic health expenditure incidence of Cambodian households. This study provides key findings and recommendations to support the RGC in its efforts to ensure equitable access to quality health care for all people by 2015, as outlined in the draft Master Plan for Social Health Protection and the Strategic Framework for Health Financing (SFHF ). It sheds new light on the level and distribution of OOP, its impact on catastrophic health expenditure and subsequent impoverishment in Cambodia. It complements the Cambodian Demographic Health Survey and Cambodian Socio-Economic Surveys Analysis Out-of-Pocket Expenditure on Health commissioned by the Ministry of Health (MOH) in Methods Data from the 2004, 2007 and 2009 CSES were analysed using methods developed by the World Health Organization. Descriptive statistical analysis of illness patterns, health care-seeking behaviour, and household health care spending, disaggregated by economic level and geographical area, were conducted. The level of catastrophic health expenditure and impoverishment resulting from health expenditures was also analysed, to evaluate changes over time. Figure 1: Illness rate, and percentage of ill people using medical services - Source: CSES 2004, 2007 and % 50% 25% 0 Results 52% 20% Year 2004 Illness 55% 17% 69% 14% Year 2007 Year 2009 Medical care Decreased incidence of illness. The incidence of reported illness in Cambodia declined from 2.25 episodes per year in 2004 to 1.87 episodes in 2009, with a similar decline in all sub-groups. In all three of the surveys, the elderly and children under five years old reported more illnesses than other groups. More ill people are seeking care. The percentage of ill people seeking care from medical providers increased substantially, from 52.2% in 2004 to 68.6% in This shows that overall health care access has improved since This increase was also pro-poor; poorer sub-groups had the largest increase in the percentage of ill people seeking medical care. Nevertheless, gaps between the poorer and richer sub-groups were still substantial. Executive Summary 1

12 Figure 2: Choice of health care provider by provider type [in %] - Source: CSES 2004, 2007 and % 80% 60% 40% 20% 0% Year % 1.6% 11.6% 23.3% Year % 1.5% 7.1% 39.2% Year 2009 (first visit) 9.2% 8.7% 2.1% 2.8% 0.2% 0.3% 40.8% Year 2009 (last visit) 41.1% 26.1% 27.5% 19.2% 18.0% 2.7% 3.1% 2.6% 3.4% 7.8% 6.1% 6.3% 7.7% 6.2% 8.5% 11.2% 9.4% Others Traditional healers Home care Pharmacies and stores (selling drugs) Private clinics Private hospitals Public hospitals Health centres Increased use of medical facilities. Pharmacies, drug stores and private clinics were still the most common choices for health care services. However, the percentage of people seeking care in public facilities (for the first visit) increased steadily, from 12% in 2004 to 20% in Similarly, the popularity of public health centres among people in the poorest economic quintile also increased since Rising income and rising out-of-pocket expenditures. Overall health spending in Cambodia surpassed the target of 5% of GDP in However, this was mainly because OOP accounted for around two-thirds of all health spending, (much higher than the 30% 40% target). OOP per capita and per year, excluding transportation and associated costs for seeking health care, rose from 59,640 Riels in 2004 (approximately USD 14.80) 2 to 117,852 Riels (approximately USD 28.30) 3 in This increase occurred at the same time as rapid income growth, as illustrated by the increase in the capacity-to-pay (see below), with household consumption also more than doubling over the same period. There were significant variations in the level of spending between age groups and by economic 1 Source: WHO s National Health Accounts (NHA); calculated as the share of total OOP from total GDP exchange rate: USD 1.00 = 4030 Riels exchange rate: USD 1.00 = 4150 Riels status. Richer sub-groups and older sub-groups spent more on health care, as they also used health services more. The average OOP continued to be lower at public health facilities than at private facilities, although much higher for hospitals (public and private) than for other provider types. Decline in catastrophic health expenditures. Household capacity-to-pay, as measured by total consumption minus subsistence income, more than doubled between 2004 and While OOP also increased substantially, it did not do so at the same rate. As a result, there was a significant decrease in catastrophic health expenditures, from 6.02% in 2004 to 4.27% in Catastrophic health expenditures occur when a household spends more than 40% of its capacity-to-pay on out-of-pocket health expenditures. A similar decrease was seen across all economic quintiles, despite strong increases in the utilisation of health care services over the same period. In addition, the incidence of household indebtedness due to illness decreased from 5.3% to 3.8% over the same period. Role of social health protection mechanisms. It is clear that the increase in household capacity-to-pay contributed substantially to the decrease in catastrophic health expenditures and household indebtedness. The CSES studies also indicate that social health protection mechanisms including HEFs, exemptions, and CBHI schemes have played a role in protect- 2 Executive Summary

13 Figure 3: Average annual* out-of-pocket expenditure per capita by sub-group [in Riels] - Source: CSES 2004, 2007 and ,000 30,000 25,000 20,000 15,000 10,000 5,000 0 M F I II III IV V Sex Age group Economic quintile Year 2004 Year 2007 Year 2009 * Calculated from 4 weeks figures multiplied by 365/28 ing vulnerable groups from catastrophic health expenditures. It should be noted that the formulation of survey questions in CSES 2004, 2007 and 2009 did not allow them to produce as much evidence regarding the impact of SHP mechanisms as we might hope for. Economic analysis of the 2009 survey, controlling for illness type and other factors, does indicate that residents of operational districts (ODs) with HEFs had lower OOP rates for health care. Similarly, residents of ODs with HEFs were less likely to suffer catastrophic expenditures or to become indebted due to illness. This study also found that after controlling for disease patterns and hospitalisation, households in rural areas were more prone to medical indebtedness and catastrophic expenditures, while larger households had lower catastrophic health expenditure risk. Interestingly, the same analysis indicates that when people who qualify for exemptions, or live in a HEF OD, spent money on health care, they were likely to spend more. Also, people exempt from paying health care fees were also more likely to be indebted due to illness. This may seem unexpected; however it should be remembered that HEFs and exemptions apply only to treatment at less expensive public facilities. So if a person who benefits from an exemption or HEF does spend money on health care, it will likely be in the more expensive private sector. Policy Implications The analysis of household survey data on health care utilisation and health spending confirms an improvement in health financing protection among Cambodians from 2004 to Several factors explain this improvement. Rapid economic growth over the previous decade resulted in increased consumption capacity, and a reduction in poverty among Cambodian households. The poverty headcount index, based on the national food poverty line, decreased substantially from 35% in 2004 to below 20% in However, evidence suggests social health protec- 4 Reference: Key indicators, Cambodian Strategic Development Plan Update Executive Summary 3

14 Figure 4: Percentage of households experiencing catastrophic health expenditure by highest and lowest economic sub-groups [in %] - Source: CSES 2004, 2007 and % 6% 3% 0 All Highest economic group Lowest economic group Year 2004 Year 2007 Year 2009 tion mechanisms, such as HEFs and fee exemptions, also had a positive influence on health care access and utilisation rate improvements over this period. Nevertheless, there are still a number of challenges that need additional policy interventions to further improve the country s health financing functions. As shown in the analysis findings, there is still significant variation in health care access across population groups. More than 20% of the population in the two lowest economic quintiles did not seek care when ill, compared to less than 10% among the highest economic quintile. The rate of health care utilisation was also lower among the lowest economic quintile, with less than half the hospitalisation rate of the highest economic quintile. The significantly higher incidence of medical debt and catastrophic health expenditure risk in rural areas deserves special attention and policy responses. Additionally, a potential rise in health care costs and subsequent financial burden could also pose a major threat to any further progress. Improving all aspects of service coverage is imperative to improve health financing protection among all Cambodians. One useful tool for monitoring health financing protection is the WHO s Asia-Pacific Health Financing Strategy targets. Applying these targets, the proportion of the overall population covered by prepayment schemes is still less than 50% (target >90%), and the coverage of vulnerable populations with social assistance and safety net programmes is currently only 68%, much lower than the 100% goal. SFHF was formulated to respond to such challenges by guiding the development of health care financing systems in the country. It aims to remove financial and other barriers to health care access for the poor, and to protect the population from the effects of catastrophic expenditures on health care. SFHF provides a framework to achieve universal coverage in the longer term (beyond 2015); the on-going development of social security schemes for formal private sector employees and civil servants will contribute to this development. The on-going expansion of HEFs to cover a higher proportion of the poor, and the planned start of social health insurance for the private sector under the National Social Security Fund, will be major milestones towards universal coverage. Harmonisation of the benefit packages and financing mechanisms could help ensure more efficient and equitable use of available funds. Building on these core social health protection mechanisms to cover the informal private sector will be a challenge, and will require registration with one of the mechanisms. At present, voluntary enrolment in HEFs is being implemented. This is in accordance with SFHF , but remains small-scale, and requires a policy decision on the actual level of contributions expected from voluntary members. Voluntary schemes may provide certain advantages in the transition towards universal coverage, such as the promotion of insurance and prepayment principles, establishing expertise and protection of people most at risk from financial burden. However, there is almost an international consensus, as outlined in the World Health Report 2010, that voluntary insurance alone cannot achieve universal coverage. The strategic path to achieve universal coverage will require transitional strategies and quick wins to gain public and political support. The presented results of the CSES analysis suggest that Cambodia has, however, made significant progress toward achieving its goal. 4 Executive Summary

15 The findings of this study emphasise again the importance of private providers in the provision of health care in Cambodia; they remain the first health care choice for most of the population, despite an increased reliance on subsidised public services. The current social health protection mechanisms only provide access to public services, and as the private sector grows, the challenge to make use of the comparative advantages of the sector to ensure public functions remains. This will require reflection on key issues such as contracting, dual practices and regulation of the sector. The RGC recognises the needs of Cambodia s rural population, and specific policy approaches will be necessary to ensure protection for the most vulnerable population groups in rural areas, as our findings suggest. Special attention is also needed to address the lower health care utilisation of this segment of the population. Further detailed studies may be necessary to identify key determinants of effective coverage of the existing social health protection schemes and inform policy-making. Executive Summary 5

16 6

17 Introduction The Cambodian Socio-Economic Survey (CSES) 2009 was a survey on household income, consumption, and asset ownership. The CSES 2009 had data on 11,971 households interviewed in 2009, with information covering 57,105 individuals. Using the CSES 2009, estimates of household consumption can be calculated using two distinct sources, according to methods used in the Cambodia Poverty Profile: monthly diaries of household consumption, and recall questions on household consumption. The only difference between these two methods is that an estimated rent value for nonproperty-owning households with no reported rent was not included in our analysis. The consumption data rely on different reference periods for various components, so the analysis adjusts for these differences accordingly. The CSES also contained questions about health insurance membership, illness in the past month, and health care-seeking behaviour. Health insurance membership was a new section in the CSES 2009 that was not available in previous CSES. The 2009 survey included information on how many visits to health care providers each ill individual made, and the type of first and last health care providers that the person visited. One major limitation in the CSES 2009 dataset is that there were no questions on inpatient admission or number of inpatient days in the survey. Household expenditure on health care was also reported as overall spending in the past month, with no breakdown by health provider type. However, the CSES 2009 contains a question on transportation costs related to health care, which is the first time this information was included in a CSES. Table 1: Basic database information used in the analysis, and source of data by type of survey - Source: CSES 2004, 2007 and 2009 Interview period Sample size CSES 2004 CSES 2007 CSES 2009 October 2003 to December 2004 January 2007 to December 2007 January 2009 to December 2009 Households 15,000 3,593 11,971 Individuals 74,719 17,439 57,105 Recall period Previous 4 weeks Previous 4 weeks Previous month Data availability Illness q14a06 q14ac06 q13bc02 Type of illness q14a07 q14ac07a to q14ac07e Chronic (q13bc03) Health care-seeking behaviours q14a08 q14ac09 q13bc06 Type of providers q14a09 q14ac10 q13bc09a,b Hospitalised q14a10 q14ac11 N/A Days hospitalised q14a11 q14ac12 N/A Health care expenditures q14a12 q14ac13 q13bc11 Health-related transport expenses N/A* N/A q13bc10 Overall consumption Yes Yes Yes Economic status quintiles From consumption data From consumption data From consumption data Insurance/ HEF membership * Not available N/A N/A Yes Introduction 7

18 8

19 Methods Descriptive statistical analysis was used to account for illness and health care-seeking behaviour. The sample weight was used to achieve the overall estimation for the country as well as for each population sub-group; individual sample weights were used for individual level analysis (e.g. careseeking, provider choices) and household sample weights were used for household-level analysis (catastrophic health expenditure incidence, etc.). Wherever possible, standard errors of the mean (in results tables) and 95% confidence intervals (in figures) were also presented. The 95% confidence intervals were calculated from the standard errors. The estimations of household capacity-to-pay, catastrophic health expenditure and impoverishment followed the methods described in the 2005 WHO document Distribution of health payments and catastrophic expenditures Methodology, by Ke Xu. Details of the methods and assumptions used in this analysis as well as any specific adjustments for the three datasets in this analysis are given in the sections below. Illness and health care-seeking behaviour The incidence of illness in the population can be estimated based on the number of individuals who reported having an illness, injury or health problem in the previous 30 days. Survey weights are applied in order to calculate national figures based on the average number of illness episodes per capita. Average illness episodes per capita = i (w i * ill i ) / i (w i ) when ill i = 0, 1 and w i = individual weight The percentage of care or treatment sought among all episodes of illness was calculated based on the number of people reporting an illness, injury or health problem, and the number who sought care or treatment for such illness: 5 Share of individual health care-seeking among all those reporting illness = i (w i * seekcare i ) / i (w i * ill i ) when seekcare i = 0, 1 Place of treatment or care CSES 2009 contains data on the number of health care visits each ill individual made in the last 30 days. It then asked specifically for the place of care for the first visit and, if more than one, the last place of care. The codes for different provider types in this report are the same as those used for the CSES, except for one minor change in the coding for public outreach activities (see code number 6 in the table below). The list of provider types from each CSES survey and their corresponding codes in the databases are provided below. In this analysis, the provider types from CSES were reclassified into eight groups for comparability across surveys. The groups are: (1) health centres, (2) public hospitals, (3) private hospitals, (4) private clinics, (5) Pharmacies and stores (selling drugs), (6) home care, (7) traditional healers, and (8) others 6. The data were further categorised into public, private and non-medical providers. The list of members of each health care provider group by survey year is provided in Table 3. The list of members of each health care provider category by survey year is shown in Table 4. Summary statistics on the proportion of different providers are calculated as a proportion of overall care or treatment sought: provider share = i (w i * provider mi ) / i (w i * seekcare ki ) where provider mi = 0, 1; m=1,..,6; and k=1,2,3 5 We also include individuals who had out-of-pocket health spending over the last month as care-seeking individuals even though they may not report that in the specific question. 6 The others category also includes an unknown place of visit from individuals who reported seeking care but did not specify the type of health care provider. Methods 9

20 Table 2: List of health care providers and corresponding codes in the survey databases - Source: CSES 2004, 2007 and 2009 CSES 2004 CSES 2007 CSES National hospital 1 National hospital (PP) 1 National hospital (PP) 2 03 Provincial hospital 2 Provincial hospital (RH) 2 Provincial hospital (RH) 3 02 Referral/district hospital 3 District hospital (RH) 3 District hospital (RH) 4 01 Health centre 4 Health centre 4 Health centre 5 5 Health post 5 Health post 6 6 Outreach 6 Provincial rehabilitation centre (PRC) or Community based rehabilitation (CBR) 7 7 Other public 7 Other public 8 05 Private hospital 8 Private hospital 8 Private hospital 9 06 Private clinic 9 Private clinic 9 Private clinic Dedicated drug store 10 Private pharmacy 10 Private pharmacy Doctor s/nurse s Home 11 Home/office of trained health worker/nurse 11 Home/office of trained health worker/nurse Patient s home/own home 12 Visit by trained health worker/nurse 12 Visit of trained health worker/nurse Other private medical 13 Other private medical Other shop selling drugs 14 Shop selling drugs/market 14 Shop selling drugs/market Healer/herbalist 15 Kru Khmer/shaman 15 Kru Khmer/shaman Monk 16 Monk/religious leader 16 Monk/religious leader Traditional midwife 17 Traditional birth attendant 17 Traditional birth attendant Other 18 Other 18 Other Table 3: Groups of health care providers and the corresponding survey codes - Source: CSES 2004, 2007 and Health centres 01 Health centre 2. Public hospitals CSES 2004 CSES 2007 CSES Referral/district Hospital 03Provincial hospital 04 National hospital 4 Health centre 5 Health post 1 National hospital (PP) 2 Provincial hospital (RH) 3 District hospital (RH) 4 Health centre 5 Health post 1 National hospital (PP) 2 Provincial hospital (RH) 3 District hospital (RH) 3. Private hospitals 05 Private hospital 8 Private hospital 8 Private hospital 4. Private clinics 5. Pharmacies and stores (selling drugs) 06 Private clinic 07 Doctor s/nurse s home 08 Dedicated drug store 09 Other shop selling drugs 6. Home care 10 Patient s home/own home 9 Private clinic 11 Home/office of trained health worker/nurse 10 Private pharmacy 14 Shop selling drugs/market 12 Visit of trained health worker/nurse 9 Private clinic 11 Home/office of trained health worker/nurse 10 Private pharmacy 14 Shop selling drugs/market 12 Visit of trained health worker/nurse 7. Traditional healers 11 Healer/herbalist 12 Traditional midwife 13 Monk 15 Kru Khmer/shaman 16 Monk/religious leader 17 Traditional birth attendant 15 Kru Khmer/shaman 16 Monk/religious leader 17 Traditional birth attendant 8. Others 14 Other (specify) 6 Outreach 7 Other public 13 Other private medical 18 Other 6 PRC or CBR 7 Other public 13 Other private medical 18 Other 10 Methods

21 Table 4: Categories of health care providers and the corresponding survey codes - Source: CSES 2004, 2007 and 2009 CSES 2004 CSES 2007 CSES Public providers 01 Health centre 02 Referral/district hospital 03 Provincial hospital 04 National hospital 1 National hospital (PP) 2 Provincial hospital (RH) 3 District hospital (RH) 4 Health centre 5 Health post 6 Outreach 7 Other public 1 National hospital (PP) 2 Provincial hospital (RH) 3 District hospital (RH) 4 Health centre 5 Health post 6 PRC or CBR 7 Other public 2. Private providers 05 Private hospital 06 Private clinic 07 Doctor s/nurse s home 08 Dedicated drug store 10 Patient s home 8 Private hospital 9 Private clinic 10 Private pharmacy 11 Home/Office of trained health worker/nurse 12 Visit of trained health worker/ nurse 13 Other private medical 8 Private hospital 9 Private clinic 10 Private pharmacy 11 Home/Office of trained health worker/nurse 12 Visit of trained health worker/ nurse 13 Other private medical 3. Non-medical providers 09 Other shop selling drugs 11 Healer/herbalist 12 Traditional midwife 13 Monk 14 Other 14 Shop selling drugs/market 15 Kru Khmer/shaman 16 Monk/religious leader 17 Traditional birth attendant 18 Other 14 Shop selling drugs/market 15 Kru Khmer/shaman 16 Monk/religious leader 17 Traditional birth attendant 18 Other Table 5: Summary data from the CSES surveys - Source: CSES 2004, 2007 and 2009 CSES 2004 CSES 2007 CSES 2009 Number of household observations 14,984 3,593 11,971 Total households (sum of household weights) 2,599,166 2,629,487 2,938,650 Number of individual observations 74,719 17,401 57,105 Total population (sum of individual weights) 13,400,000 13,200,000 13,966,718 Average household size Geographic characteristics (households) % of households in the capital 7.95% 6.30% 9.91% % of households in other urban areas (outside capital) 10.25% 10.07% 10.22% % of households in other rural areas (outside capital) 81.79% 83.62% 79.87% Demographic characteristics (individuals) % of males 48.60% 48.50% 48.77% % of aged % 9.60% 9.91% % of aged % 23.80% 22.20% % of aged % 60.10% 61.03% % of aged % 6.50% 6.87% Methods 11

22 Household health expenditure In addition to illness and health care-seeking behaviour, the level of out-of-pocket expenditure for care and treatment, and the level of spending on transport to receive such care, are of interest to health funding policy analysts. Out-of-pocket health expenditure Out-of-pocket health payments (OOP) are made by households when they receive health services. The CSES contains data on how much individuals spent on medical care in the previous 30 days. It is assumed that the spending reported in the survey was out-of-pocket, with no reimbursement from a health insurance provider or other source. The CSES 2009 questionnaire asked only about the total amount of health care spending by each household member in the previous 30 days, with no detailed information on the amount spent per specific episode of illness or for a specific health care provider. Therefore, in this analysis we did not try to estimate the amount of spending at each type of health care provider. The cumulative OOP for the country and average OOP per household were calculated at the household level using household sample weights: household OOP (oop h ) total OOP = i (oop i ) = h (w h * oop h ) average OOP per household = h (w h * oop h ) / h (w h ) In addition to average household-level health spending, the proportion and/or total OOP at various provider types or provider groups were calculated. Household OOP disaggregated by household member characteristics (age, sex, etc.) were also calculated. Two levels of estimation are included: the individual level and the household level. Average OOP per illness episode, by illness type, provider type and age group were calculated on an individual level using individual sample weights: OOP per episode of illness = i (w i * oop i ) / i (w i * ill i ) OOP per episode of illness by illness type = i (w i * oop i ) / i (w i * illtype i ) OOP per episode of illness by provider = i (w i * oop i * provider mi ) / i (w i * provider mi ) OOP per episode of illness by age group = i (w i * oop i * agegrp n ) / i (w i * agegrp n ) 12 Methods

23 Health-related transportation expenses The CSES 2009 includes the transportation costs for health care purposes (toop) over the past 30 days (not for each care-seeking episode). Expenditures on health carerelated transportation were then calculated as the average cost per capita using individual-level data and individual sample weights: average toop per visit = i (w i * toop i * seekcare ki ) / i (w i * seekcare ki ) average toop per visit by provider type = i (w i * toop i * provider mi ) / i (w i * provider mi ) The cumulative transportation expenses and share of transport expenses in total OOP were calculated using household-level data and household sample weights: household toop (toop h ) = i (toop i ) cumulative toop = h (w h * toop h ) Household capacity-to-pay and poverty level Estimation of household capacity-to-pay is necessary to categorise the catastrophic level of health payments. To calculate capacity-to-pay, data on household consumption expenditures and food expenditures were used to estimate household subsistence spending. All consumption variables (e.g., overall consumption, food, health) were converted into a monthly figure (1 month = 30.4 days). The analysis did not adjust for inflation, as it was assumed that the inflation rate over the survey period was minimal. Household consumption expenditure Household consumption expenditure (exp) comprises both monetary and in-kind payments on all goods and services, and the monetary value of the consumption of homemade products. Food expenditure Household food expenditure (food) is the amount spent on all foodstuffs by the household, plus the value of the family s own food production that is consumed within the household. Because food expenditure is used for subsistence estimation in the analysis, household consumption of alcoholic beverages, tobacco, food consumed out of the home (e.g., at hotels and restaurants) and prepared meals bought outside and eaten at home were not included in household food expenditure. However, those consumption items were included in the household consumption expenditure (exp). Household subsistence spending According to the WHO method, a food share-based reference line was used to estimate household subsistence level (se). This reference line is defined as the weighted average of equivalised food expenditure of the households whose food expenditure as a share of total household expenditure is between the 45th and 55th percentile of the population. The household equivalence scale is defined as: eqsize h = hhsize h β where hhsize h is the household size. The value of the parameter β is 0.56, according to the WHO s estimation. Equivalised food expenditure is therefore household food expenditure divided by the equivalent household size. Subsistence spending was calculated from the reference line by multiplying the equivalent household size. The household s capacity-to-pay The household capacity-to-pay (ctp) is defined as a household s non-subsistence spending. For households with food expenditure lower than subsistence spending (se h >food h ), non-food expenditure is used as non-subsistence spending: ctp h = exp h se h if se h <= food h ctp h = exp h food h if se h > food h Poverty level A household is regarded as poor (poor h ) when its per capita household expenditure is smaller than the nationally defined per capita poverty line (pl h ), which includes both Methods 13

24 food and non-food components. The poverty line varies according to the geographical location of the household whether it is in Phnom Penh, other urban areas, or rural areas. The poverty line per capita per day in 2009 for Phnom Penh, other urban and rural areas was 4,185 Riels, 3,443 Riels, and 3,231 Riels respectively. 7 The household poverty incidence was calculated using household poverty status and household weight while the poverty headcount (individual) was calculated using the household member s poverty status (same as household status) and individual weight: poor h = 1 if (exp h / hhsize h ) < pl h Fairness in Financial Contribution Index The distribution of household financial contribution across households was also summarised using the WHO s fairness of financial contribution index (FFC). The FFC is based on the mean of the cubed absolute difference between the oopctp of each household and the oopctp norm: where poor h = 0 if (exp h / hhsize h ) pl h Catastrophic health spending, impoverishment, and medical indebtedness Out-of-pocket health expenditure as a share of household capacity-to-pay The burden of health payments is defined as a household s out-of-pocket health expenditure as a percentage of the household s capacity-to-pay (oopctp): oopctp h = oop h / ctp h Catastrophic health expenditure Catastrophic health expenditure (cata) is when a household s total out-of-pocket health payments equal or exceed 40% of a household s capacity-to-pay or non-subsistence spending. The threshold of 40% may be adapted according to a country s specific situation, but enables cross-country comparison as per WHO methodology. The dummy variable on catastrophic health expenditure was constructed with 1 indicating a household with catastrophic health expenditure, and 0 indicating a household without catastrophic health expenditure: cata h = 1 if oop x / ctp x 0.4 The FFC ranges between 0 and 1. The more equitable the health financing system, the closer FFC will be to 1. Impoverishment A non-poor household is impoverished by health payments when it becomes poor after paying for health services. To assess the extent of impoverishment in the population, a dummy variable on the poverty impact of health payments (impoor h ) was created. It equals 1 when household expenditure per capita is equal to or higher than the national poverty line, but is lower than the national poverty line after deducting OOP, and 0 otherwise: impoor h = 1 if exp h / hhsize h pl h and (exp h oop h ) / hhsize h < pl h, otherwise impoor h = 0 Indebtedness through illness The incidence of indebtedness resulting from illness among Cambodian households was calculated. The average level of debt per household and the distribution of debt by economic quintile were estimated. Additionally, the characteristics of households with indebtedness through illness were explored, to identify the determinants of such indebtedness. cata h = 0 if oop x / ctp x < The poverty line is based on nutrition equivalent of 2,100 kilocalories and a proportion of non-food allowances. 14 Methods

25 Sub-group analysis Wherever feasible, analysis for population sub-groups was conducted to demonstrate the pattern of health care-seeking behaviour, health spending, and catastrophic impacts. The sub-group analysis includes the following. Economic quintile Households were grouped into economic quintiles (expenditure quintiles) using equivalised per capita household expenditure (eqexp h ). Equivalised per capita expenditure equals total household expenditure divided by equivalent household size: eqexp h = exp h / eqsize h Household weight was considered when grouping the population by quintile. Urban and rural groups Three geographical (region) categories were used in this analysis: capital, other urban and other rural. Capital refers to both urban and rural areas of Phnom Penh. Other urban includes all other urban areas outside Phnom Penh. Similarly, other rural covers all rural areas outside Phnom Penh. Summary statistics for each of these geographical groups were calculated, and household weight was used. Fee exemption and health insurance status of households that allows free or subsidised health care. 8 Additionally, a household is classified as having a fee exemption if any of their members received free or subsidised health care in the last 12 months because their names are on a list held by the local authorities. 9 Operational districts with health equity funds Additionally, we analysed data to explore the difference between households living in ODs with HEFs and those living in areas without. Households were mapped into ODs using the codes provided by Owen O Donnell and Gabriela Flores. Data on HEF availability by OD were provided by the MOH. Other sub-groups Analysis of individual-level statistics was also conducted whenever feasible, including sub-groups such as sex (male/ female) and age group (under five, five to 15, 15 to 60, and over 60 years old). Further disaggregation by smaller age groups (i.e , 30-45, and years old) was also conducted whenever feasible. Determinant analysis In addition to the descriptive analysis to explore the patterns of health care-seeking behaviour and health care spending, further analysis to identify key determinants of health care-seeking behaviour and health care spending were also conducted. Results on health care utilisation and health spending by health insurance membership were also presented. Because the CSES 2009 contains questions on health insurance membership, households were divided into those with fee exemptions and those without. A household is considered to have a fee exemption if they answer that any household member has a priority access card, health equity card, or any other document 8 This includes households that answered yes on question 4, section 13 of CSES 2009 that asked, Do you or any member of the household have a Priority Access Card, Equity Card, or any other document that allows free or subsidised health care? 9 Any member answering option 1 or 2 under question 2 of question 13, How did they obtain this free/ subsidised treatment? 1 = Household Priority Access Card, Equity Card, or other document that allows free or subsidised health care 2 = Name(s) are on a List of Poor Households held by the local authorities. Note that this does not include other answers, such as provider offered free care, private health insurance, CBHI, etc. Methods 15

26 Table 6: Incidence of self-reported illness in the previous month - Source: CSES 2004, 2007 and 2009 Table 7: Incidence of illness in the previous month by sub-group (individual weighted) - Source: CSES 2004, 2007 and 2009 Sex Male (s.e.) (0.0020) (0.0037) (0.0020) Female (s.e.) (0.0020) (0.0040) (0.0021) Age group Age (s.e.) (0.0052) (0.0115) (0.0056) Age (s.e.) (0.0024) (0.0049) (0.0026) Age (s.e.) (0.0018) (0.0032) (0.0018) Age (s.e.) (0.0072) (0.0141) (0.0075) Economic quintile Average illness episode per capita (standard error) (0.0014) (0.0029) (0.0016) 1st quintile (I) (s.e.) (0.0031) (0.0062) (0.0032) 2nd quintile (II) (s.e.) (0.0032) (0.0069) (0.0033) 3rd quintile (III) (s.e.) (0.0032) (0.0066) (0.0034) 4th quintile (IV) (s.e.) (0.0033) (0.0068) (0.0034) 5th quintile (V) (s.e.) (0.0030) (0.0048) (0.0032) Table 8: Health care-seeking in the previous month - Source: CSES 2004, 2007 and % of all reported ill who sought care 90.3% 91.5% 91.4% (s.e.) (0.25%) (0.54%) (0.31%) % of all reported ill who used medical providers 52.2% 55.2% 68.6% (s.e.) (0.42%) (0.97%) (0.51%) 16 Methods

27 Results 1. Illness and health care-seeking behaviour 1.1 Incidence of illness over the previous month but the incidence of illness among the lowest quintile is lower than the other economic groups, especially in Across all years, most population sub-groups in the lowest quintile have lower incidences of illness. 1.2 Incidence of health care-seeking behaviour On average, the incidence of illness in the previous month among the population significantly declined from 2004 to The average illness episode per capita per month (1 month = 30.4 days) was 0.187, 0.166, and for 2004, 2007, and 2009 respectively (Table 6). Reported type of illness CSES 2009 does not provide direct information on the type of illness experienced, unlike in 2004 or However, it does ask whether the reported illness was chronic over the past year, or was so severe that the person needed to stop work. Approximately 20.4 percent of all reported illnesses were chronic (s.e. 0.44%) and around one-eighth, or 12.5 percent, of all reported ill people had to stop work because of their illness (s.e. 0.36%). Incidence of illness by gender, age group, and economic quintile Table 8 shows the health care-seeking behaviour from 2009 compared with the results from CSES 2004 and The proportion of ill people seeking care was reportedly similar across all three surveys, at slightly more than 90 percent. In both 2004 and 2007, only a little more than half of all those who were ill visited health care providers, but this proportion significantly increased to over two-thirds in Almost 15 percent of the population reported having a health care consultation or treatment, or spending money on health care over the past month. Around seven percent (or half of those who reported seeking care) did so only once, and another four percent (or 28 percent of those who reported seeking care) sought care twice (Table 9). The average incidence of illness varied by sex and age group in 2009 (Table 7). Females, young children (under five), and the elderly (60 and over) demonstrated a higher incidence of illness than their counterparts. The difference between economic quintiles is less prominent, Table 9: Number of health care-seeking episodes in the previous month - Source: CSES 2009 % of population % of those who sought care at least once None 85.6% - Once 7.2% 49.8% Twice 4.0% 27.8% Three times 1.8% 12.7% Four times or more 1.4% 9.7% Total 100% 100% Results 17

28 Health care-seeking behaviour among population groups When comparing male and female health care-seeking behaviour, males tended to seek care less than females, though the difference was minimal (Table 10). Comparing the CSES 2009 with the 2004 and 2007 surveys, health care-seeking incidence decreased in 2009 among females more than it did among males. When looking at this behaviour across age groups, children under five and the elderly over 60 demonstrated higher health care use than other age groups (Table 37). Across economic quintiles, households in lower quintiles sought care less than households in higher quintiles, but the differences across economic quintiles became smaller in 2009 when compared to the differences in Some of the health care-seeking episodes are at medical care providers. In all sub-groups except children under five, the proportion of medical care use increased in Still, male individuals accessed medical care less than females and children, and the elderly accessed medical care more than other age groups. Individuals in higher economic quintiles also used medical care more than those in lower quintiles. When evaluating health care-seeking behaviour only among those that reported an illness, the proportion seeking medical care was roughly similar to the proportions in 2004 and 2007 for most population sub-groups (Table 12). Only children under five years old saw a significant decrease in health care-seeking behaviour when ill. Figure 1 summarises the annual reported incidence of illness and the percentage of ill individuals that did not seek care in 2004, 2007 and Females, young children (under five), and the elderly (aged 60 and over) had the highest incidences of annual illness. The differences between economic quintiles were less prominent in 2009 than in previous years. The proportion of ill individuals who did not seek care decreased marginally across almost all population sub-groups. The only sub-group that saw a significant increase in not seeking care was children under five. It was found that children under five in rural areas, and especially among higher economic quintiles, sought care less than other sub-groups. lower reported illnesses per capita and a lower proportion of health care-seeking behaviour than people in rural areas. However, among people reporting an illness, those in Phnom Penh have a significantly higher proportion of health care-seeking behaviour than ill people in other urban or rural areas. In other words, there was a higher proportion of reportedly ill people outside Phnom Penh who did not seek care. The proportion of ill people who used medical care was also much lower in rural areas than in Phnom Penh or other urban areas. Health care fee exemption and health careseeking behaviour The CSES 2009 survey asked each household whether they had any members receiving free or subsidised health care over the past 12 months. It also asked whether any members had a priority access card, HEF card, or any other document that would allow free or subsidised health care. The analysis of CSES 2009 data found that approximately 5.6 percent of the population have some form of HEF membership, or a document that allows them to receive health care for free or at a subsidised rate. This proportion varies by economic quintile; fee exemptions make up a larger proportion of households in the lowest economic quintile (11.3 percent) than in the highest quintile (1.8 percent; Table 15). Households with fee exemptions had higher average illness incidence per capita, and a higher proportion of care-seeking per capita than households without fee exemptions (Table 16). The average number of careseeking episodes was also higher. However, the percentage of individuals not seeking care when ill was higher among households with fee exemptions than households without fee exemptions, as shown at the bottom of Table 16. Table 14 presents the statistics on illness and health careseeking behaviour by geographical region. The population in Phnom Penh and other urban areas have significantly Results