Although there are wide variations in mortality between developed and developing countries, socioeconomic inequalities in health exist in both the societies. The evidence shows not only that all countries have socioeconomic differentials in health, the magnitude and nature of these inequalities vary from place to place and time to time.

Lower mortality levels among children with higher parental socioeconomic status, often measured by occupation and education, have been observed in many developing countries 678. In Bangladesh, household economic status as measured by ownership of various household articles had found negative relationship with child mortality 9. Using data from Matlab area 10, a negative relationship was documented between mortality of children aged 1–4 and various socioeconomic factors.

Higher survival prospects for female children in industrial countries are nearly universal but exceptions to this pattern are observed for a number of developing countries 11. For rural Bangladesh, higher female mortality compared to male after the neonatal period was reported consistently 1213. An examination of cumulative life-table probabilities of mortality by sex of the children for the cohort born during 1973–75 in Matlab area revealed that in ages 1–4, mortality for females exceeded that for males by 59 per cent 14. A study conducted in Matlab area documented discrimination against female children in intra family food distribution and medical care; this was put forward as a likely mechanism for sex bias child health and survival 15.

Higher mortality rates in rural than urban areas have been documented in many developing countries 6. Based on Bangladesh Fertility Survey data, higher mortality among rural children was reported by some studies 912. After reviewing the relevant factors, it was identified that better medical, environmental, and socioeconomic conditions in the urban areas having a favorable effect 9.

Availability of better toilet facility was found to exert a positive influence on child mortality in many countries 16. A study in a rural area of Bangladesh reported a positive association of use of latrine and a negative association of household size with survival during the post-neonatal period 17. For childhood mortality an inverse relationship between dwelling area and use of fixed latrine was also observed in Matlab area 10.

Studies showing the effect of health intervention on reducing socioeconomic inequalities in mortality are few, however, results are also not conclusive.

In Costa Rica, it was documented that the socioeconomic differentials of infant mortality narrowed when targeted primary health care services improved 18 while a similar pattern was reported in a cross-national study that childhood mortality by mother's education narrowed with the increase in per capita health expenditure 19. In Colombia, it was found that increased access to health services disproportionately improved chances of child survival among less educated mother 20.

Based on Matlab data, although it was reported that the health intervention program had a greater effect on the risks of child death (1–4 years) of uneducated mothers than mothers with at some education 21 but re-analysis of the same data set did not confirm the result^a [Appendix A; 22]. Again, it was reported that the poor-rich difference of 1–4 years mortality had decreased over time in the intervention area 23 while it was reported that measles vaccination narrowed the poor-rich difference of mortality of aged 9–60 months 24. In another study, the impact of a woman-focused development programme on child survival was examined and found that there has been a 52% reduction of the pre-intervention level hazard of death of children during infancy of participant mother's compared to 31% reduction for the infants of on-participant mothers but during childhood (1–4 years) such reduction was similar in the two groups 25. It was reported earlier that the poor-rich ratio of under-five and infant mortality had increased overtime but poor-rich ratio of 1–4 years mortality declined slightly in the intervention area 5.

Based on Nigerian data, it was reported that presence of modern health facilities had widened the childhood mortality by mother's education 26. They argued that the educated mother has more knowledge of the modern world and enough self-confidence to secure urgent help for her sick children from medical professional. In Brazil, it was reported that the new interventions tend to increase inequity since they initially reach those who are already better off, at least until the wealthy reach a level beyond which little progress can be made 27.

Data for this study come from Matlab Upazila (sub-district) where the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) has been maintaining a field station since 1963. Matlab is a rural area located about 55 km South-east of Dhaka. The area is low-lying deltaic plain intersected by the tidal river Gumti and its numerous canals. The major modes of transport within the area are walking, country boat and in some cases small steamer or launch. Farming is the dominant occupation, except in a few villages where fishing is the means of livelihood 28. Most of the farmers are in marginal situations with less than two acres of land, and 40% of them are landless. For many families sharecropping and work on others' land on a daily wage basis have become the main sources of livelihood. Some people also work in mills and factories in different towns and cities but their families live in the study area. Rice constitutes the staple food and is harvested three times annually. Rates of illiteracy are high and increase with age.

In Matlab study area the ICDDR,B has been maintaining a Demographic Surveillance System (DSS) over 200,000 people since 1966. The DSS collects information on births, deaths, migration, marriages, divorce and household splits. The DSS events are collected by the Community Health Research Worker (CHRW) through monthly household visits and Field Research Supervisor (in the past, CHRW recorded events through fortnightly household visits and FRS accompanied by the CHRW visited the household every six weeks to complete the registration form). The DSS also maintains cross-sectional socioeconomic data and such data is available for 1974, 1982 and 1996.

The Matlab HDSS (DSS re-named as HDSS after integrating with the health data) area consists of ICDDR,B-service and govt-service areas. These two areas are almost similar in socioeconomic conditions 28 but differ in MCH-FP program 29. Most of the ICDDR,B-service area was exposed to a contraceptive distribution program during 1975–77 and has been exposed to the Maternal Child health and Family Planning services since October 1977 30. In the ICDDR,B-service area, MCH-FP services have been provided by the CHRWs through fortnightly home visits until December 1999 and then from fixed site clinics. In fact, the intervention began with the family planning services and all other health services have gradually been added over a period of time. The CHRWs provide contraceptives and basic medicines to mother and child and referred patients with complications to sub-centre clinics. In addition, there are four sub-centres that also provide MCH-FP services. Moreover, ICDDR,B has a free 60-bed diarrhoea treatment centre in Matlab town and the facility is used not only by Matlab people but also from the surrounding districts. These health services are: tetanus toxoid vaccination to mothers; measles, DPT and polio vaccination to all children (by mid-1986 all blocks were covered); diarrhoeal disease management by bari mothers; vitamin A and beta-carotene supplementation; maternity care and antenatal check-ups; Acute Lower Respiratory Infection (ALRI) detection and management with penicillin and oral cotrimoxazole; education intervention to reduce ALRI mortality; reproductive tract infections detection and management; community based safe motherhood; and health centre assisted deliveries.

Until mid-1970s, the government supported health service was available mainly in the urban areas. The services were more curative than preventive in nature. The government of Bangladesh accepted the primary health care concept as national health objective in 1978. Since then, the health care system was reoriented to provide essential health care to the general mass. In fact, the government significantly increased the funding of health sectors from early 1980s. The facilities include Maternal and Child Welfare Centre in urban and sub-urban areas, Upazila Health Complex at Upazila level and the Family Welfare Centre at union level. The government has also made primary health service facilities available at Rural Dispensaries and Satellite Clinics. In Matlab town, the government runs a 31-bed free general hospital along with few union-level health facilities. In addition, the government has been promoting oral rehydration therapy for diarrhoea management, and the immunization program against six major childhood diseases 31. In fact, immunization against childhood diseases is mainly delivered through satellite clinics (since mid-1980s) while family planning services are delivered at the doorstep until recently. These services, however, less intensive than those services in the ICDDR,B-service area.

As health services are added gradually in both the ICDDR,B-service and govt-service areas, four birth cohorts (1983–85, 1988–90, 1993–95 and 1998–00) under study benefited differently from these interventions. For example, measles vaccination to all children started in 1982 in two Blocks of the ICDDR,B-service area while such vaccination started in 1985 in other two Blocks. In the govt-service area, immunization program was initiated in late-1980s and slowly gained popularity and in recent years the coverage rate is about 70%. So, children in the earlier cohorts benefited less from interventions than (immunization coverage differ) those in the recent cohort.

The study used HDSS data from both the ICDDR,B-service and govt-service areas. Four population-based birth cohorts were selected for the study. The HDSS system registered 21,280 births for 1983–85, 20,370 births for 1988–90, 16,925 births for 1993–95 and 17,102 births for 1998–00. These births were subsequently followed for five years for death and out-migration to ascertain survival and migration status respectively.

The socio-economic status is defined here in term of assets, rather than income or consumption. The asset information was collected through the household questionnaire administered during the censuses of 1982 and 1996. These questions include ownership of a number of consumer items (radio, watch, etc), dwelling characteristics (wall and roof material), type of drinking water and toilet facilities, however, more consumer items were collected in 1996 than in 1982 census.

Household economic status is measured in this study by constructing a wealth index using asset ownership as it was validated 32. Each household asset for which all information collected was assigned a weight or factor score generated through principal components analysis. The resulting asset scores were standardized in relation to a standard normal distribution with a mean of zero and a standard deviation of one. Each household was assigned a standardized score for each asset, where the score differed depending on whether or not the household owned that asset. These scores were summed according to household and individuals were ranked according to the total score of the household in which they resided. The household were then divided into quintiles. Birth cohort-1983–85 and birth cohort-1988–90 were matched with the household asset score of 1982 and birth cohort-1993–95 and birth cohort-1998–00 were matched with household asset score of 1996^b [Appendix A].

Based on asset quintiles, two indicators of inequality are created: poor-rich ratio and concentration index 33. The ratio (poor-rich) does not provide any information about the middle three quintiles but provides a magnitude of differences between the poorest and the richest quintiles of the population. On the other hand, concentration index measures the extent to which a particular health status variable is distributed across all five asset quintiles- that is the concentration of inequality. The concentration indices take the values between -1 and 1. The closer is the index to zero for any one health indicator, the less concentrated is the wealth inequality for that indicator; conversely, the further away is the index from zero, the greater is the inequality. A negative value indicates disproportionate concentration of the variable among the poor while a positive value indicates that the poor are getting less than would be expected had the distribution been equitable.

Life table mortality rates were calculated for neonatal, infant and child mortality.