Abstract
This paper compares data from two household surveys to assess the effect of questionnaire design on estimated use of health services and analyses this across geographic areas and different groups. Deficiencies in the design of Statistics South Africa's General Household Survey led to a substantial underestimation of utilisation (capturing less than a third of visits). The South Africa Consortium for Benefit Incidence Analysis survey, which was more comprehensive, indicated that three out of four outpatient visits are to public sector facilities. Medical scheme membership is the most important predictor of using a private provider, particularly for inpatient care. Socioeconomic status and rural versus urban residence also influence overall utilisation rates and use of public versus private providers. It is critical to improve the design of routine household surveys to monitor utilisation patterns during the implementation of the proposed health system reform.
1. Introduction
The routine availability of accurate data on the use of health services is important for health system planning and monitoring. It allows us to assess whether utilisation rates are appropriate relative to the burden of ill-health and whether health services are being used at the appropriate level of care, which has implications for allocative efficiency. It also allows us to evaluate the distribution of health service use across socioeconomic groups and between geographic areas, which provides insights into equity issues.
Access to accurate health service utilisation data is all the more important at this point, given that South Africa has decided to implement national health insurance (NHI), with the aim of achieving universal coverage. This is an ambitious health sector reform, and one of the main concerns raised by key stakeholders is whether a universal system, with access to needed health care for all, is affordable and sustainable. A prerequisite for assessing this issue is accurate health service utilisation data. This will make it possible to analyse the current patterns of what services are being used by which groups and also provide a critical baseline for projecting what resources are likely to be needed in the reformed health system, since projection models rely on demographic, service utilisation and service cost data (McIntyre & Borghi, Citation2012).
Utilisation data may be available from routine health information systems, but such systems usually only cover public sector health facilities and do not provide any information on the people who are using these services (e.g. demographic and socioeconomic characteristics). We therefore generally have to rely on household survey data. However, to be useful for modelling the affordability of health sector reforms, the survey must be able to produce utilisation rate data. Unfortunately, many household surveys that have information of relevance to the health sector are not appropriately structured to allow for the calculation of utilisation rates. Two serious deficiencies in questionnaire design make it difficult to calculate these rates. First, utilisation information in some household surveys is dependent on self-reported recent illness, i.e. this information is collected only if respondents indicate that they or other household members have been ill or injured within a specified recall period. Second, many surveys only ask about the use of one service, whereas many people use more than one service or make multiple visits to the same service provider during an illness episode (McIntyre & Ataguba, Citation2010).
This study compared data derived from two household surveys to assess the extent to which utilisation patterns differ according to how the questions are structured, and analysed the distribution of health service utilisation across geographic areas and different groups in South Africa. The paper presents the first comprehensive analysis of health service utilisation rates carried out in South Africa.
2. Methods
2.1 Data sources
Data from two household surveys were used for this analysis. The first is the General Household Survey (GHS), which has been conducted annually by Statistics South Africa since 2002. This is a nationally representative survey intended to elicit information on such matters as education, health, the labour market, dwellings, access to services and facilities, transport, and quality of life of the population. The 2008 GHS, which we used in this analysis to ensure comparability with the other survey, had a sample size of 24 293 households. The first two questions in the health section of the questionnaire asked whether the respondent had suffered an illness or injury in the previous month and if so what kind, and this was followed by the main question on use of services, which only elicited information about the most recent consultation (see ).
Table 1: Questions relating to health service utilisation in the GHS
The second household survey we looked at is what has been termed the South Africa Consortium for Benefit Incidence Analysis (SACBIA).Footnote1 This survey was specifically undertaken to obtain comprehensive data on health service utilisation in South Africa. It is a nationally representative household survey of over 21 000 individuals (reduced to 20 099 after data cleaning). Enumerator areas (EAs) were stratified by province, type of settlement (farm, informal settlement, tribal settlement, smallholding and urban settlement) and population group. In total, 960 EAs were selected across the nine provinces and five randomly selected households were interviewed in each EA, giving a total sample size of 4800 households. The EAs in each stratum were selected with a probability proportional to the size of the EA, defined as the number of households it contained. Fieldworkers were given extensive training to ensure the questions were well understood. Data were collected in May and June 2008. Twenty per cent of the questionnaires were subjected to telephonic ‘check-backs’ for verification, and double-entry data capture reduced errors. The data were weighted to national population levels. The questionnaire and study protocol were subject to ethical review by the University of Cape Town, and all respondents provided signed informed consent.
The SACBIA survey focused on health system issues exclusively, and particularly the respondents' use of services. The main differences in the design of the SACBIA questionnaire (see ) compared with the GHS were:
| • | Collection of utilisation data was not conditional on reporting illness or injury. While respondents were asked whether any member of the household had been ill or injured in the previous month, this question was asked after questions on health service utilisation. The question on outpatient health service utilisation asked whether any member of the household had used any health service in the previous month. | ||||
| • | Data were captured on all visits during the previous month (i.e. not just the most recent visit). Visits to multiple providers as well as multiple visits to the same provider were captured. | ||||
| • | A more detailed list of health care providers was used. In particular, the SACBIA survey asked for the name of the hospital if a hospital outpatient department had been used. This allowed us to categorise the hospital, according to level of care, into district, regional, provincial and central hospitals, all of which provide acute care. | ||||
Table 2: Questions relating to health service utilisation in the SACBIA survey
The same approach was used for admissions to hospitals, where data were captured on all admissions within the past year.
2.2 Data analysis
Disaggregated utilisation data are presented according to:
| • | Geographic areas (provinces and, where possible, rural, formal urban and informal urban areas), | ||||
| • | Race group, | ||||
| • | Socioeconomic quintiles (based on a composite index calculated using principal component analysis, which included a range of asset and related variables that were common to the GHS and SACBIA surveys), and | ||||
| • | Medical scheme membership status. | ||||
The main analysis of health service utilisation in this paper is based on the SACBIA survey, which does allow for the calculation of utilisation rates (i.e. average number of visits to an outpatient provider per person per year or average number of hospital admissions per 1000 population). Only the main ‘formal’ health care providers were included, i.e. traditional healers were not included, and nor were categories of providers such as clinic at workplace and psychologist because only small numbers reported using these providers). The data were adjusted for seasonality (only in the case of outpatient services, as hospital admissions had a one-year recall period) and were age–sex standardised.
We also undertook negative binomial regression analyses to examine the association between the utilisation rate and relevant explanatory variables. Utilisation rates are an example of count data, i.e. non-negative counts of events with no negative values but with skewed distributions (O'Donnell et al., Citation2008). Although it is possible to apply a Poisson model, the negative binomial model proved more effective in dealing with problems associated with overall dispersion. In addition, it relaxes the Poisson distributions' restriction of equal mean and variance (Land et al., Citation1996).
2.2.1 Seasonality adjustment
While some analysts simply multiply the number of outpatient visits reported in the previous month by 12, and then divide this by the total population to estimate annual utilisation rates (O'Donnell et al. Citation2008), seasonal variations should be taken into account as the rate calculated purely from the household survey will be under-reported when the month of reporting is less prone to specific diseases or health conditions, or over-reported when it is more prone (McIntyre & Ataguba, Citation2010).
In this study, seasonality indices were generated using data on the total visits to public sector facilities in each month from the District Health Information System. In the case of private sector services, the indices were generated from data on utilisation patterns in each month provided by the largest medical scheme administrators, given that the use of private services is heavily skewed towards medical scheme members. The index essentially compares utilisation of each type of service in the month(s) in which household survey data were collected, with the average monthly utilisation over a full year. The seasonality index is calculated as follows:
Although seasonal adjustment does not have a significant effect on the results (McIntyre & Ataguba, Citation2010), we have made these adjustments as this provides more precise utilisation estimates.
2.2.2 Age–sex standardisation
To make useful comparisons of utilisation across different geographic areas and socioeconomic groups, it is important to standardise for the age–sex distribution within each of these areas and groups. This is necessary because different age–sex groups have different levels of need for health care; very young children, the elderly and women of childbearing age in particular have relatively high health service needs. The effect of the age–sex distribution must be taken into account so that differences we find in utilisation patterns across geographic areas or specific groups are not due to differences in their demographic composition (Schokkaert & Van de Voorde, Citation2009).
The two basic methods of standardisation are the direct and the indirect methods. Although they both aim to adjust for correlation between health indicators such as utilisation rates and demographics (O'Donnell et al., Citation2008), the indirect method is the more frequently used to standardise utilisation rates and has the advantage of not requiring information from a reference population. In this study, the indirect standardisation method was applied using the following steps. First, we estimate Equation
Equation(2) below, using the following age–sex groups: infants (below five years), children (five to 14 years), adults (15 to 49 years), senior adults (50 to 59 years) and senior citizens (60 years and above) using ordinary least square (OLS) regression. The standardisation formula used is:
Second, the OLS parameter (
) estimates and individual values of the standardising variables (x
ji
), are used to generate the predicted values of the utilisation rate
. Finally, estimates of the indirect standardised health service utilisation rate,
, are obtained by estimating the difference between actual and predicted utilisation rate, plus the overall utilisation rate sample mean (
) as given in equation 3 below:
3. Results
3.1 Comparison of utilisation patterns across surveys
Similar levels of acute illness were reported across the two surveys. Nearly 14% reported an illness in the previous month in the GHS and nearly 18% in the SACBIA survey. The SACBIA survey found that 38% of those who reported using a health service in the previous month did not report being ill in that month, highlighting the importance of not restricting questions on health service use to those who report illness.
While the GHS 2008 survey indicated that 61% of those who had used an outpatient service due to illness in the previous month had used a public provider, the SACBIA survey indicated that 71% of health service visits in the past month had been to a public provider. To calculate these estimates, the denominator used for the GHS was the total number reporting service use (as only one visit was recorded per person). In the case of the SACBIA study, as multiple visits by the same person were recorded, the denominator is the total number of visits reported.
shows that there was a similar difference in reported use of public and private health care providers across the two surveys when the utilisation patterns are disaggregated by geographic area and socioeconomic and other groups. The differences are particularly striking for the middle three socioeconomic quintiles, for Africans and coloureds, and for respondents in some of the poorer provinces (the Eastern Cape, the Free State, Limpopo, Mpumalanga and the North West).
Table 3: Percentage distribution of utilisation of public and private providers
If we calculate utilisation rates on the basis of the GHS data, we find there was an average number of outpatient visits per person per year of 0.83 to public providers and 0.53 to private providers. In contrast, the SACBIA survey indicates an average of 3.1 to public providers and 1.14 to private providers.
3.2 Health service utilisation rates of public and private providers
All results presented in this and following sections are drawn from the SACBIA dataset.
and present an overview of the utilisation rates of public and private sector outpatient services for different geographic areas and groups. Private sector utilisation rates were higher in formal urban areas than in rural and informal urban areas, while public sector utilisation rates were higher in rural areas (see ). There were also considerable differences across provinces, with utilisation rates of public sector health services being highest in Limpopo and KwaZulu-Natal. Private sector utilisation rates were highest in the Western Cape, Gauteng and KwaZulu-Natal. The Northern Cape has the lowest overall utilisation rates and the lowest rates of public and private sector service utilisation.
Figure 1: Age–sex standardised utilisation of outpatient services by province and type of area (2008)
Figure 2: Age–sex standardised utilisation of outpatient services by medical scheme membership status, socioeconomic group and race group (2008)
There are striking differences in utilisation rates across race groups, with the highest rates of use of private providers being found in the Asian and white groups and of public providers in the African and coloured groups (see ). There was a similar trend across socioeconomic groups, with higher use of private providers by richer groups and of public providers by poorer groups (see ). The widest differences in public–private sector utilisation rates relate to medical scheme membership, with very low private sector utilisation rates for those who are not members of such schemes and high for those who are, and the reverse for public sector services (see ).
The overall inpatient utilisation rate was 78 admissions per 1000 population for public hospitals and 17 per 1000 population for private hospitals. and show very similar patterns of public–private utilisation rate differences across geographic areas and groups in terms of hospital admissions. Use of private providers is even more heavily restricted to medical scheme members and the highest income quintile for inpatient rather than outpatient services.
Figure 3: Age–sex standardised utilisation of inpatient services by province and type of area (2008)
Figure 4: Age–sex standardised utilisation of inpatient services by medical scheme membership status, socioeconomic group and ‘race’ group (2008)
The output of the multivariate analysis of outpatient utilisation (see ) shows that outpatient utilisation rates were lower for males than females. The analysis reinforces the finding that medical scheme members have higher utilisation rates for private services but lower rates for public facilities when compared to non-members. In terms of race groups, using whites as the basis for comparison, on aggregate all other race groups visit all services combined more frequently, but, when disaggregated into public and private services, the outpatient private sector utilisation rates were lower for Africans and coloureds than for whites. For all categories of outpatient services, residents of the Northern Cape have significantly lower health care utilisation rates than those from the Western Cape. People in the lower quintiles used public health care facilities more than those in the top quintile.
Table 4: Multivariate regression of outpatient utilisation (total, public and private services; average number of visits per person per year)
Multivariate analysis of inpatient service use (see ) indicates that being African, living in the Northern Cape, living in a rural area and falling within quintiles 1, 2 and 3 significantly predicts lower overall inpatient utilisation rates, and more importantly predicts utilisation of public sector hospitals. However, membership of a medical scheme and being in a higher socioeconomic quintile are positive predictors of overall inpatient utilisation rates. Medical scheme members have lower public hospital but higher private hospital inpatient utilisation rates than those who are not members of a medical scheme.
Table 5: Multivariate regression of in-patient utilisation (total, public and private services; average number of admissions per 1000 population per year)
3.3 Health service utilisation rates by type of provider
provides a more detailed breakdown of outpatient service utilisation rates according to different types of public and private providers. The highest utilisation rates of public sector clinics and district hospital outpatient departments were found for the African population, rural dwellers, the two poorest quintiles, and those who are not members of medical schemes. Relatively high use of regional, provincial and central hospital outpatient services was found for the Asian population, urban dwellers and the third and fourth socioeconomic quintiles.
Table 6: Age–sex standardised utilisation of outpatient services by geographic areas and various groups (2008; average number of visits per person per year)
There is a strong and consistent socioeconomic gradient in the use of private general practitioners. This is also the case for most other private providers, with the exception of the surprising finding as regards the poorest quintile's use of private hospital outpatient services, which is probably due to small numbers. There are also strong gradients in the use of the various private providers according to race, with Africans and coloureds having the lowest utilisation rates and whites and Asians the highest. The strongest predictor of use of all categories of private providers was membership of a medical scheme.
disaggregates utilisation rates of different categories of public hospitals according to geographic areas and various groups. The trends in the use of inpatient services and outpatient public hospital care are similar. For example, while there is far higher use of district hospital inpatient services in rural areas, the use of higher-level public hospitals is far higher in urban areas. However, the differences in admission rates to the various categories of public hospitals across race groups are less clear, with problems of small numbers likely to be affecting the reported rates for whites. The gradient in terms of socioeconomic status is clearer, with poorer groups using district hospitals more often and richer groups using provincial and central hospital inpatient services more often. The patterns of private hospital use are striking, with far higher utilisation rates in formal urban areas, in Gauteng and the Western Cape, among whites and in the highest socioeconomic quintile. The substantial difference in utilisation rates according to medical scheme status makes it clear that inpatient care in private hospitals is almost entirely restricted to medical scheme members.
Table 7: Age–sex standardised utilisation of public inpatient services by geographic areas and various groups (2008; average number of admissions per 1000 population per year)
4. Discussion
4.1 The importance of accurate utilisation data
The comparison of the GHS and the SACBIA surveys suggests that the more limited collection of data in the GHS provides an inaccurate depiction of public–private sector utilisation patterns. The SACBIA survey indicates consistently higher levels of utilisation of public sector providers than the GHS does. This makes intuitive sense, given that recording of service utilisation in the GHS is conditional on reporting of acute illness or injury. This would automatically exclude use of preventive services (such as antenatal services) and generally also visits to a provider for chronic care (as patients with chronic illnesses may not consider themselves to have been sick in the previous month). It is very likely that there is greater utilisation of public sector providers for these kinds of services. The reason for this is that private services tend to focus on curative care rather than preventive or promotive interventions and because the cost of routine check-ups and particularly medicines for chronic conditions in the private sector would be unaffordable for most South Africans. The extent to which the GHS questionnaire design misrepresents utilisation patterns is demonstrated by the substantially lower utilisation rates calculated from the GHS (1.36 outpatient visits per person per year) than from the SACBIA survey (4.24 visits per person per year).
It is critically important that accurate health service utilisation data are available during a period of major health system reform, not only for projecting likely future resource requirements in a reformed system, but also for monitoring the impact of policy change. A major concern of key stakeholders is the affordability and sustainability of the proposed NHI in South Africa. This is a legitimate concern, and the sustainability of a universal health system is strongly dependent on ensuring that overall health service utilisation rates do not increase too rapidly and that services are used at the appropriate level of care. Monitoring requires regular, accurate access to utilisation rate data. In Thailand, the National Statistics Office was persuaded to undertake certain household surveys (the Socioeconomic Survey and the Health and Welfare Survey) more regularly than was the norm over the period of implementing their universal coverage policy, to enable monitoring of changes in service utilisation (Tangcharoensathien et al., Citation2007). Although in South Africa the GHS is undertaken on an annual basis (while SACBIA was a once-off project specific survey), the design of the GHS questionnaire does not make it possible to calculate accurate utilisation rates, nor does it allow for adequate disaggregation across different types of health care providers.
The most important changes required to the GHS to provide sufficiently accurate data for monitoring purposes are as follows:
| • | Questions on health service utilisation should not be made dependent on reporting an illness or injury. Instead, we should ask directly about the use of outpatient services in the previous month (and inpatient services in the past year). If we want to ask about self-reported illness or self-assessed health status, this is best asked later in the questionnaire. Even though utilisation reporting may not be conditional on illness reporting, if we ask about illness just before asking about utilisation, there is likely to be a subconscious influence on utilisation reporting. | ||||
| • | The utilisation questions should ask about all services used and how many times each was used. | ||||
| • | More disaggregated provider categories should be used, particularly in relation to the level of care provided by public hospitals. However, we recognise that this will be very resource intensive. Most survey respondents are unlikely to know whether a particular hospital is classified as a district, regional, provincial or central hospital. It is therefore necessary to obtain the name of the hospital and to code it after data collection. It may only be feasible to collect such disaggregated data every few years rather than annually. | ||||
4.2 Key patterns of health service utilisation in South Africa
The SACBIA survey is the only national household survey dataset currently available in South Africa that allows for the calculation of health service utilisation rates. For this reason, it is used to explore the current pattern of health care utilisation in more detail. However, it should be recognised that the sample size of 4800 households does translate into small cell sizes for some disaggregations. This is particularly the case when utilisation rates are disaggregated by type of provider as well as by geographic area or types of population groups. Thus, while the analysis of public-private utilisation patterns presented in to can be regarded as stable, this is not necessarily the case with the detailed provider categories presented in and .
The analysis presented in this paper establishes that the vast majority of South Africans are heavily dependent on the public health sector for their health service needs. Despite policy efforts to improve accessibility and reduce inequalities in health services in South Africa in the past one and half decades, there is compelling evidence that medical scheme membership and socioeconomic status are the major drivers of the use of health services in the public as compared to the private sector. Those covered by medical schemes not only use private providers extensively, they also make considerable use of public health facilities, particularly for inpatient care in provincial and central hospitals. On the other hand, health facilities used by those who are not members of medical schemes are concentrated in the public sector, with very low use of private providers, especially for hospital outpatient and inpatient care and for dentists. A similar trend was observed in our analysis across socioeconomic groups, where the rich have a higher probability of belonging to a medical scheme and not only use private sector services much more than the poor, but also use public sector hospital services within higher-level facilities (provincial tertiary and national central hospitals) to a far greater extent than poorer groups. An interesting finding is that the use of higher-level public sector facilities was particularly high for the third and fourth quintiles rather than the richest (fifth) quintile. This is because 71% of medical scheme members fall within quintile five and hence primarily use private providers (McIntyre, Citation2010). In contrast, only 18% of scheme members are in quintile four and even fewer in quintile three, and they are therefore more reliant on public sector facilities, but seem to be better able to negotiate access to higher-level public hospitals than poorer groups.
There were clear disparities in the use of health services by type of residential area. Use of outpatient care in public clinics and district hospital outpatient departments was particularly high in rural areas, but not of higher-level public hospitals or private providers (the result for private hospitals is likely to be due to small cell sizes). However, there were low overall utilisation rates in rural areas for inpatient care, although again particularly for higher-level public hospitals. The variation in utilisation rates across provinces is closely aligned with differences in the distribution of public and private health sector facilities and human resources between provinces (Day & Gray, Citation2008). For example, the Eastern Cape has one of the lowest levels of utilisation for both outpatient and inpatient services and has below average levels of health care professionals and hospital beds relative to its population size. The Northern Cape, which has reasonable health personnel to population ratios, but which covers a large physical area and is very sparsely populated, has the lowest level of outpatient service utilisation, highlighting the impact of poor physical access to health facilities. Inter-provincial variations in utilisation rates are also likely to be linked to differences in the burden of disease across provinces. For example, the particularly high levels of overall utilisation in KwaZulu-Natal are likely to be related to the very large burden of HIV and AIDS in this province.
5. Conclusions
South Africans in general use more public than private sector services. On average, three out of every four outpatient visits in South Africa takes place in public sector facilities. There is even greater reliance on public sector facilities for inpatient care. Medical scheme membership is the single most important predictor of utilisation of private as opposed to public health care facilities.
From the perspective of proposed health system reforms, current utilisation patterns suggest that while the public health sector will need to continue to be the backbone of health service delivery, it will also be important to draw on the service delivery capacity of the private sector (as it currently provides about a quarter of all outpatient services). Another important finding is that current utilisation levels are already at over four visits per person per year for outpatient services and nearly 100 hospital admissions per 1000 population for inpatient services. The experience in Thailand (Tangcharoensathien et al., Citation2007) has shown that utilisation increases when there is improved financial protection and access to needed health care. It will thus be important to monitor utilisation rates closely as health system reforms are introduced in South Africa, as has been done successfully in Thailand.
The SACBIA survey was a once-off initiative and it is unlikely to be feasible to undertake such a survey on a regular basis. While it has provided important information previously unavailable in South Africa, the relatively small sample size (compared to surveys undertaken by Statistics South Africa, such as the GHS) limited the ability to undertake disaggregated analyses. With relatively few changes and little additional data collection effort, the GHS could allow for routine monitoring of progress with the NHI reforms from a service utilisation perspective. It is of critical importance that current deficiencies in the GHS and other regular household surveys are addressed in order to support successful health system reform in South Africa.
Acknowledgements
The household survey on which this paper is based was funded by the European Union through a grant to the Department of Health. Di McIntyre is supported by the South African Research Chairs Initiative of the Department of Science and Technology and the National Research Foundation. The usual disclaimers apply. The authors are grateful to all our colleagues who contributed to the SACBIA survey.
Notes
1The Survey was funded by the National Department of Health through a grant from the European Union and data were collected by the Community Agency for Social Enquiry. It was a collaborative initiative between the Health Economics Unit at the University of Cape Town, the Centre for Health Policy at the University of the Witwatersrand, the National Department of Health and the London School of Hygiene and Tropical Medicine.
Related Research Data
References
- Day , C and Gray , A . 2008 . “ Health and related indicators ” . In South African Health Review 2008 , Edited by: Barron , P and Roma-Reardon , J . 239 – 395 . Durban : Health Systems Trust .
- Econex . 2009 . What Does the Demand for Healthcare Look Like in SA? NHI (National Health Insurance) Note 3 , Stellenbosch : Econex .
- Land , K C , McCall , PL and Nagin , D S . 1996 . A comparison of poisson, negative binomial and semiparametric mixed poisson regression models . Sociological Methods and Research , 24 ( 4 ) : 387 – 442 . (doi:10.1177/0049124196024004001)
- McIntyre , D . 2010 . “ National Health Insurance: Providing a vocabulary for public engagement ” . In South African Health Review 2010 , Edited by: Fonn , S and Padarath , A . 145 – 56 . Durban : Health Systems Trust .
- McIntyre , D and Ataguba , J . 2010 . How to do (or not to do) … a benefit incidence analysis . Health Policy and Planning , 26 ( 2 ) : 174 – 82 . (doi:10.1093/heapol/czq031)
- McIntyre , D and Borghi , J . 2012 . Inside the black box: Modelling health care financing reform in data-poor contexts . Health Policy and Planning , 27 ( Suppl. i ) : i77 – 87 . (doi:10.1093/heapol/czs006)
- O'Donnell , O , Van Doorslaer , E , Wagstaff , A and Lindelow , M . 2008 . Analyzing Health Equity Using Household Survey Data: A Guide to Techniques and their Implementation , Washington , DC : World Bank .
- Schokkaert , E and Van de Voorde , C . 2009 . Direct versus indirect standardization in risk adjustment . Journal of Health Economics , 28 ( 2 ) : 361 – 74 . (doi:10.1016/j.jhealeco.2008.10.012)
- Tangcharoensathien , V , Limwattananon , S and Prakongsai , P . 2007 . “ Improving health-related information systems to monitor equity in health: Lessons from Thailand ” . In The Economics of Health Equity , Edited by: McIntyre , D and Mooney , G . 222 – 46 . Cambridge : Cambridge University Press .