1 of 36Secondary Analysis of CDHS 2000 by S. Fabricant

Introduction:

The Ministry of Health Cambodia, with assistance from the World Health Organization and other development partners including DfID, World Bank, the ADB, UNFPA, UNICEF, GTZ, CIDA and USAID has produced a sector-wide strategic plan for 2003-2007. The plan is based on a common strategy for the entire health sector and focuses on enabling health outcomes for all Cambodians.

In parallel, the Government of Cambodia is also drafting a national poverty reduction strategy. A need exists to analyse the current situation of health status, service coverage, utilization and expenditures among the poor. A wealth of primary data has been collected through the Cambodian Demographic and Health Survey (CDHS) 2000 and the Cambodian Socio-Economic Survey (CSES) 1999. A fair amount of secondary analysis of the CDHS has been accomplished with support from the World Bank, WFP and MACRO International [1]to assess some key indicators by wealth rankings based on ownership of assets at household level.

In this context, the current study is complementary to ongoing work by World Bank, World Food Program and MACRO in examining selected health and demographic indicators by socio-economic groupings based on household income, expenditure and asset ownership. Additional analyses based on education status and gender are also be performed to identify priority target groups and the appropriate service delivery strategies.

Methods used:

The purpose of this study was to determine the extent of wealth and urban-rural differentials for selected demographic and health variables in the population studied. The “household” and “women’s” recoded datasets from the Cambodia Demographic and Health Survey of 2000 were examined using the household wealth rankings calculated by the original World Bank consultants, ORC-Macro. These rankings were calculated from an index of indicators based mainly on the assets owned by a household, type of housing, and others, according to a uniform standardized methodology developed by World Bank[2]. Asset variables were weighted according to principal component analysis and summed to produce a Wealth Index. This in turn was used to divide all sampled households into wealth quintiles.

The original Final Report summarized urban/rural differences for most reported variables. In this report it was usually possible to analyze wealth differences separately among urban and rural groups (households and women). However, for some variables it was necessary to condense wealth groups into two groups in order to obtain reasonable numbers of cases, especially for the much smaller urban group. Unless otherwise noted, the “poorer” groups comprise quintiles I, II, and III, while the “richer” groups comprise quintiles IV and V.

The results here do not always coincide with those in the ORC-MACRO CHSS Final Report. In the course of the analysis it was observed that the number of cases used did not always agree with the number of cases used in the tabulations presented in the Final Report. Since the Final Report did not as a rule state how cases for tabulation were selected, the selection (i.e., filtering cases for analysis by another independent variable) may have been done differently for the analysis presented here. However, the differences are mostly small, and wealth-related trends are consistent.

Results:

The most significant socio-economic differences are described in this section. When a significant wealth difference is noted, the percentages for the method used are given only for the wealth quintiles with the most extreme rates, generally but not always the lowest (Q-I) and highest (Q-V). Unless otherwise noted, the percentages in the intermediate quintiles fall in between the lowest and highest in a fairly linear manner. Where the important results an be described in a sentence or two, no tables have been prepared. When it is likely to be useful to view the results in their entirety, a table showing the important parameters is given.

Section 1: Drinking Water and Sanitation

1.1 Source of drinking water (dry season):

Urban/rural differentials:

The most common sources for urban households are piped-in dwelling/yard/plot (23%), open public well (15%), tanker/vendor (13%) and river/stream/pond (12%). For rural households, the most common sources are river/stream/pond (29%), open public well (26%), and tubed/piped public well or borehole (12%).

Wealth differentials:

Urban: Only the wealthiest (Q-V) households have piped-in water (42% versus 0.1% for all the rest). Tanker truck/vendors are also used mainly by Q-IV and Q-V. Most Q-I urban households (58%) use an open public well, compared to only 2% of Q-V households. Use of river/stream/pond/lake water is most prevalent among Q-II (29%), Q-III (28%), and Q-IV (21%).

Rural: Poorer households make much greater use of open public wells (41%) and tubed /piped public wells/boreholes (16%) than do wealthier ones (6% and 7%). Q-I households make significantly less use (19%) of river/stream/pond/lake water than the average of 29%. Use of tanker/truck deliveries ranges from 0% for Q-I, to 17% for Q-V.

1.2Source of drinking water (rainy season):

Urban/rural differentials: The most common sources for urban households are rainwater (25%), piped-in dwelling/yard/plot (22%), and open public well (12%). For rural households, the most common sources are river/stream/pond (23%), open public well (22%), and rainwater (17%).

Wealth differentials:

Urban: As in the dry season, only the wealthiest (Q-V) households have piped-in water (39% versus 0.1% for all the rest). Tanker truck/vendors are also used mainly by Q-IV and Q-V. Most Q-I households (54%) use an open public well, compared to only 1% of Q-V households. Use of river/stream/pond/lake water is most prevalent among Q-II (26%) and Q-III (26%). Rainwater collection is most used by Q-IV (42%) and Q-V (28%) households.

Rural: Poorer households make much greater use of open public wells (41%) and tubed /piped public wells/boreholes (15%) than do wealthier ones (3% and 4% for Q-IV and Q-V). Q-I and Q-V households make somewhat less use of river/stream/pond/lake water than the intermediate groups. Use of tanker/truck deliveries declined in the rainy season for Q-V to 7%.

1.3Time to water source (dry season and rainy season):

Urban/rural differentials: In the dry season, 90% of urban households can access drinking water in less than 15 minutes, compared to 77% of rural households. In rainy season access improves slightly, with 94% of urban households and 81% of rural households having access in less than 15 minutes.

Wealth differentials:

Urban: Access to drinking water (the source notwithstanding), is nearly the same for all wealth groups, with 85% of urban Q-I households having access within 15 minutes, compared to 95% in the dry season. In rainy season these rates increase uniformly for all groups, to 89% for Q-I and 97% for Q-V.

Rural: While all wealth groups in rural areas suffer worse access than urban groups, the differentials are similar. Access within 15 minutes is 72% for Q-I households, and 78% for Q-V households. In the rainy season these improve to 76% for Q-I and 84% for Q-V.

Most Cambodian households have good access to drinking water in both dry and rainy seasons, but poorer households utilize sources such as streams and lakes and open wells which are more likely to be contaminated. This is especially true in rural areas.

1.4 Household sanitation facility

Urban/rural differentials: 47% of urban households use no facility/field, and another 10% use a traditional pit latrine. 23% use a flush system connected to a sewer and 17% an unconnected flush system In rural areas 86% of households use no sanitary facility, or a field.

Wealth differentials:

Urban: The use of flush systems, either connected or unconnected, is limited almost entirely to Q-V households. Use of pit latrines also increases with wealth, increasing from about 1% in the lower 3 quintiles, to 19% and 13% in Q-IV and Q-V respectively. Correponding to this usage is a large differential in having no sanitary facility at all, from 99% in the lower 3 quintiles to 11% for Q-V households.

Rural: Differentials are nearly as great as in urban areas, with about 95% of households in the 3 quintiles having no sanitary facility, compared to 79% for Q-IV and 23% for Q-V. As in urban areas, use of a flush system is nearly all by Q-V households. Use of pit latrines is also only about 2% in the 3 lowest quintiles, versus 18% in the two highest.

Sanitary waste disposal facilities are nearly non-existent in rural areas, and in urban areas are limited to wealthy households.

Section 2: Use of iodised salt

Urban/rural differentials: Iodine was found present in salt in 28% of urban households (71% not present, 1% not tested or not having salt). This compares with only 14% of rural households having iodised salt (85% not present, 1% not tested or having any salt.)

Wealth differentials:

Urban: 38% of Q-V and 20% of Q-IV households had iodised salt, compared with only about 15% in the three lowest quintiles.

Rural: Groups with the highest rates were Q-V with 25% and Q-I with 18%. The 3 intermediate groups had average rates of 11%.

Iodised salt is reaching urban households more effectively than rural households. Wealthy households use iodised salt at about twice the rate of poorer ones.

Section 3: Accidental death or injury

Note: The data is reported differently here from the CDHS Final Report because of the low incidence rates of accidents and injuries.

Urban/rural differentials: Injuries or accidents occurred to one or more household members in the past 12 months in 4% of both urban and rural households. 21% of those injured in urban areas died, and 20% died in rural areas. Accidents claimed the lives of 50% of accident victims in urban areas and 40% in rural areas.

Wealth differentials:

Urban: Accident and injury rates were lower in poorer (Q-I/II/III) rural households (3%) than in wealthier (Q-IV/V) rural households (5%). The number of cases was too low to permit valid comparisons of death rates.

Rural: Accident and injury rates among poorer and wealthier groups were nearly equal in rural areas. Resulting death rates were lower among poorer groups for accidents (17% vs. 41% for wealthier), but higher for injuries (24% vs. 13%).

Injuries are equally prevalent in urban and rural areas, but fatalities are more prevalent among the wealthy and urban households. Although the type of accidents were not recorded, it may be that many deaths are due to motor vehicle and motor farm equipment accidents, which are most likely to happen to wealthier rural and urban residents.

Section 4: Health care utilization and expenditures

4.1Illness or injury in previous 30 days:

Urban/rural differentials: Rural households experienced an illness/injury rate of 39%, compared with 31% for urban households.

Wealth differentials:

Urban: Wealth groups in urban areas experienced incidence rates ranging from 27% (Q-V) to 40% (Q-IV). There is no clear wealth-related pattern in the urban areas.

Rural: A similar distribution of incidence was seen in rural wealth groups, with Q-V the lowest at 32% and Q-IV the highest at 43%. The four lower income groups all had fairly high incidences (38% to 43%) compared with Q-V.

Residence / Incidence of illness/injury by wealth and residence groups
Wealth groups
Q-I / Q-II / Q-III / Q-IV / Q-V / All wealth groups
Urban / 31.7% / 36.9% / 29.6% / 39.8% / 27.4% / 30.9%
Rural / 38.0% / 40.1% / 41.1% / 43.0% / 32.3% / 39.4%
Both / 37.5% / 39.9% / 40.2% / 42.6% / 30.0% / 38.2%

Morbidity is fairly evenly distributed across income groups in both urban and rural areas with the exception of the wealthiest quintile, which enjoys about 25% less illness. Reported morbidity is almost identical to neighboring Vietnam.

4.2Proportion who sought advice or treatment (by age and sex)

(Note: this analysis and the following one used different age groupings than the original CDHS Final Report, with ages 0-1 yr, 1-5 yrs, 6-15 yrs, and >15 yrs corresponding more closely to standard mortality calculations. Tabulated cases include up to three treatments for each of the first three household members reported having an illness or injury.)

Urban/rural differentials: Individuals sought treatment in 93% of urban households that experienced at least one illness, compared with those in 89% of rural households. Males in urban households had nearly the same rate of treatment-seeking as males in rural households (91% vs. 90%), but the differential for females was higher, at 95% urban vs. 89% rural.

Wealth differentials:

Urban: The rate of treatment-seeking ranged from 83% for Q-I to 97% in Q-V, with the three intermediate income groups about halfway between these percentages.

Rural: All wealth groups in rural areas had nearly the same rate of seeking treatment, with a mean of 89% for all rural households.

Age and gender differentials: The same overall urban-rural and wealth differentials are seen in each age subgroup. Treatment-seeking rates decrease slightly with age for males in both urban (97% for 0-1 year to 92% for adults) and rural (91% for 0-1 year to 89% for adults) areas. Rates are nearly identical for all age groups of urban females, while for rural females, infants and adults both had slightly lower rates than children.

A very high percentage of Cambodians seek treatment when ill, but income appears to be more of a barrier in urban areas than in rural areas. Treatment-seeking is higher than in Vietnam, where only 72% of ill people sought treatment.

4.3Choice of Provider (by age, severity and gender)

(Note: this analysis used different age groupings than the original report, with ages 0-1 yr, 1-5 yrs, 6-15 yrs, and >15 yrs coinciding more closely with usual mortality rate calculations. Tabulated cases included up to three treatments for each of the first three household members reported having an illness or injury, for a total of 7,130 episodes. Condensed wealth groups (Q-I/II/III and Q-IV/V) were used to maintain a higher number of cases in urban area cells.)

Overall, 19% of cases were treated by a public provider, 38% by a private provider, and 43% received non-medical treatment. Severity of illness was a major factor in determining provider choice, as shown below. Very little serious illness is treated by non-medical providers.

Self-reported severity all illness, first treatment of first 3 hh members
Provider used / Serious / Moderate / Slight / Total % / n
Public / 40.8% / 20.5% / 10.5% / 19.5% / 1061
Private / 49.4% / 44.6% / 24.6% / 37.9% / 2064
Non-medical / 9.8% / 35.0% / 64.9% / 42.6% / 2320
All / 12.9% / 50.1% / 36.3% / 100.0% / 5445

Urban/rural differentials: Public providers are used in 14% of episodes in urban areas and in 19% of episodes in rural areas. The most frequently used providerwas a shop or market selling drugs, used in 31% of cases in urban areas and 35% in rural areas. Dedicated drugstores (16%) and private clinics (15%) were the next most prevalent in urban areas, while in rural areas privately paid visits of a trained health worker (19%) and private clinic visits (9%) were next most prevalent.

Wealth and gender differentials:

Urban: The urban rich make greater use of public providers (government hospitals, health centers, clinics, and health workers) than the urban poor. Urban males and females make nearly equal use of public providers, although the numbers are small. The urban poor make much greater use of market drugs than the wealthy, 50% for males and 45% for females versus 23% and 25% respectively. Corresponding to this difference is the greater use by the wealthy of central and provincial hospitals (combined 6% by poor males and 4% by poor females vs. 12% by wealthy males and 11% by wealthy females), of private clinics (8% vs. 19% for males and 11% vs. 17% for females), and of dedicated drugstores (4% poorer males vs. 20% wealthy males and 6% vs. 21% for females.)

Rural: The rural rich also make greater use of public health providers than do the poor. Males and females make nearly the same use of public providers. The rural poor make greater use of market drugs than the wealthy, 41% for males and 39% for females versus 27% and 24% respectively. Corresponding to this difference is the greater use by the wealthy of central and provincial hospitals (combined 4% by poor males and 5% by poor fmales vs. 10% by wealthy males and 12% by wealthy females), of private clinics (6% vs. 13% males and 7% vs. 14% females), and of dedicated drugstores (4% poorer males vs. 9% wealthy males and 3% vs. 9% for females.) Visits to or by trained health workers are used equally by rich and poor of both sexes, in about 18% of all episodes.

Age differentials: Age-related preferences were difficult to find in urban areas because of small numbers of cases, but provincial hospitals were used least by children 6-15 years, who instead made high use of other private providers, especially for the poor in this age group. Use of other providers was nearly equal among age groups, with the most prevalent for all being shops/market selling drugs, except for infants and adults from wealthy households, who used dedicated drugstores more than market drugs. In rural areas, adults made greater use of public providers than other age groups. Adults from wealthy households used public providers in 26% of episodes, compared with 17-21% in other age groups from wealthy households, and 15-17% of other age groups in poor households. Greatest use of market drugs was made for poor (49%) and wealthy (34%) children age 1-5. Dedicated drugstores and trained health worker visits were used equally by all age groups in a given wealth category.

There is great diversity of provider choice in Cambodia. Public sector providers are used in a minority of all illness episodes, primarily for serious cases. As is true in many low-income countries, self-treatment with market drugs is the most common choice for non-serious illness and even some of moderate severity, and wealthier people use registered pharmacies for this more than the poor. Gender differences in provider choice are fairly small, but rural females tend to use self-treatment somewhat less often than males. Formal providers of most types are used relatively more by adults than by children.

4.4 Mean treatment expenditures (by age and sex)

As seen in Section 4.1, 4,670 households (38.2% of the 12,236 households sampled) reported at least one household member was ill or injured in the past 30 days.

Notes: Sampled households reported expenditures both in riels and dollars. All riel expenditures were converted to dollars at the same rate as used in the CDHS Final Report ($1=R4000). Expenditures are reported here as the sum of transport costs and treatment costs in order to reduce the complexity of analysis. Transport costs, while not negligible, averaged only 10.5% of the total cost in urban areas and 7.3% of total costs in rural areas. For some of the following analyses mean expenditures are the total of the first three reported treatments for the first three household members reported ill. Where there were interesting differences between first and subsequent treatments, these were analyzed separately.