Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

9
Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh Golam Mostafa a, *, Jeroen K. van Ginneken b a Health and Demographic Surveillance Programme, Public Health Sciences Division, International Centre for Diarrhoeal Disease, Research, GPO Box 128, Dhaka 1000, Bangladesh b Netherlands Interdisciplinary Demographic Institute, P.O. Box 11650, 2502 AR The Hague, Netherlands Abstract Longitudinal data collected from the Demographic Surveillance System (DSS) in Matlab, a rural area in Bangladesh, are used for determining trends in and determinants of mortality of the elderly population (60 yr and over) in 1974–1996. The old-age mortality rate is high in Matlab, 1.2 times that of Sri Lankan and 1.5 times that of the Swedish elderly population in a comparable period. Mortality among the elderly population declined in 1974– 1982, but much less so in 1982–1996. Proportional hazards models were used for examining determinants of mortality in a sample of about 10,000 elderly persons. This multivariate analysis used information on several social and economic variables derived from the 1982 census and mortality data of this population which was followed prospectively in 1982–1992. Marital status was the single most important determinant: widows and widowers had 1.5 to 2 times higher risk of death compared to couples where both husbands and wives were alive. Social support in old age by children also plays a role, especially for women: women living with at least one son or daughter had 18% lower mortality than women living in a household without sons or daughters. Socioeconomic factors are also important. Those who had at least some education or were relatively auent had lower mortality than those with no education or who were less auent. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: Elderly; Mortality; Matlab; Bangladesh Introduction Bangladesh has made progress in passing through the second phase of the demographic transition although socioeconomic conditions continue to be un- favourable. The crude birth rate declined from 48 per 1000 in 1970 to 27 in 1995 and the crude death rate from 21 per 1000 in 1970 to 11 in 1995. In the same period, the total fertility rate (TFR) declined from 7.0 in 1970 to 3.3 in 1996–1997 and under-five mortality declined from about 250 in 1970 to 113 in 1995 (Mitra et al., 1997; UN, 1998). One of the consequences of this development is that the process of ageing in the population of Bangladesh has started. This can be seen by the following figures. In 1961 5.2% of the population of Bangladesh was 60 yr and older. This proportion increased to 6.1% in 1995 and is expected to increase to 9.1% in 2010. In absolute numbers this means an increase from 2.9 Social Science & Medicine 50 (2000) 763–771 0277-9536/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0277-9536(99)00295-6 www.elsevier.com/locate/socscimed * Corresponding author. Tel.: +880-2-881-0719; fax: +880- 2-886-050. E-mail address: [email protected] (G. Mostafa).

Transcript of Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

Page 1: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

Trends in and determinants of mortality in the elderlypopulation of Matlab, Bangladesh

Golam Mostafaa,*, Jeroen K. van Ginnekenb

aHealth and Demographic Surveillance Programme, Public Health Sciences Division, International Centre for Diarrhoeal Disease,

Research, GPO Box 128, Dhaka 1000, BangladeshbNetherlands Interdisciplinary Demographic Institute, P.O. Box 11650, 2502 AR The Hague, Netherlands

Abstract

Longitudinal data collected from the Demographic Surveillance System (DSS) in Matlab, a rural area inBangladesh, are used for determining trends in and determinants of mortality of the elderly population (60 yr andover) in 1974±1996. The old-age mortality rate is high in Matlab, 1.2 times that of Sri Lankan and 1.5 times that of

the Swedish elderly population in a comparable period. Mortality among the elderly population declined in 1974±1982, but much less so in 1982±1996.Proportional hazards models were used for examining determinants of mortality in a sample of about 10,000

elderly persons. This multivariate analysis used information on several social and economic variables derived from

the 1982 census and mortality data of this population which was followed prospectively in 1982±1992. Maritalstatus was the single most important determinant: widows and widowers had 1.5 to 2 times higher risk of deathcompared to couples where both husbands and wives were alive. Social support in old age by children also plays a

role, especially for women: women living with at least one son or daughter had 18% lower mortality than womenliving in a household without sons or daughters. Socioeconomic factors are also important. Those who had at leastsome education or were relatively a�uent had lower mortality than those with no education or who were less

a�uent. # 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Elderly; Mortality; Matlab; Bangladesh

Introduction

Bangladesh has made progress in passing throughthe second phase of the demographic transition

although socioeconomic conditions continue to be un-favourable. The crude birth rate declined from 48 per1000 in 1970 to 27 in 1995 and the crude death rate

from 21 per 1000 in 1970 to 11 in 1995. In the same

period, the total fertility rate (TFR) declined from 7.0in 1970 to 3.3 in 1996±1997 and under-®ve mortalitydeclined from about 250 in 1970 to 113 in 1995 (Mitra

et al., 1997; UN, 1998).One of the consequences of this development is that

the process of ageing in the population of Bangladesh

has started. This can be seen by the following ®gures.In 1961 5.2% of the population of Bangladesh was 60yr and older. This proportion increased to 6.1% in1995 and is expected to increase to 9.1% in 2010. In

absolute numbers this means an increase from 2.9

Social Science & Medicine 50 (2000) 763±771

0277-9536/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.

PII: S0277-9536(99 )00295-6

www.elsevier.com/locate/socscimed

* Corresponding author. Tel.: +880-2-881-0719; fax: +880-

2-886-050.

E-mail address: [email protected] (G. Mostafa).

Page 2: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

million in 1961 to 7.3 million in 1995 and to 13.2

million in 2010 (Hossain, 1997). Little is known aboutthe social, economic and health conditions of theelderly population (de®ned here as 60 yr and older) in

Bangladesh. In this article, we deal with the mortalityand its determinants among the elderly population. Weare particularly interested in the role of support net-

works in the households to which the elderly peoplebelong and in the role of socioeconomic factors.

Considerable research has been done on these topicsin developed countries (Berkman and Syme, 1979;Bowling, 1987; Feldman et al., 1989; Valkonen, 1989;

Hu and Goldman, 1990; Mackenbach et al., 1997) andsome in Bangladesh and other South Asian countries

(Cain, 1986; Myers, 1992; Rahman et al., 1992; Martinand Kinsella, 1994; Rahman, 1997). Such researchshows that relatives of elderly people, especially their

children, play an important role in providing socialand economic support. In a review of the available lit-erature, Martin and Kinsella concluded that much

more needs to be known about the in¯uence of livingarrangements of the elderly people on health and mor-

tality in Asian countries and on the role of social andeconomic factors (Martin and Kinsella, 1994).For many reasons, it is important to know more

about the trends in and determinants of old-age mor-tality. Bangladesh society is experiencing fairly rapidchanges in household size and composition, relation-

ships within the household and relationships amongthe household members of di�erent generations. Old

people depend on their adult children, particularlysons, for old-age support and security as there is nopublic system of social security. Increased migration of

the workforce, changes in family structure, a trendtoward smaller family size and other socioeconomicchanges may adversely a�ect the old-age support sys-

tem in the villages. Social, economic and demographicdevelopments have caused several changes at the indi-

vidual, family and societal levels, all of which in¯uencethe lives of elderly people.In Bangladesh, concern for the elderly population is

still low priority for the Government. This is partlydue to the relatively small size of the elderly popu-

lation at present. A more important factor may be thecultural perception that elderly people will be takencare of by the extended families to which they belong.

Therefore, there is no urgent need for governmentintervention. Moreover, there are more pressing socialproblems, such as malnutrition, housing problem, illit-

eracy and widespread poverty, which demand priorityfrom the Government. In spite of this, there is growing

concern about the welfare of the elderly populationand provision of appropriate care (Hossain, 1997).Moreover, it is recognized that, as larger number of

people become older, a larger proportion of the popu-lation will die from cancer, cardio-vascular diseases

and other diseases relating to old age (Smith, 1993).For the various reasons mentioned above, more needs

to be known on trends in and correlates of mortalityin old age. The Matlab project of ICDDR,B providesa unique opportunity to study this topic in detail.

Methods and procedures

Data used in this study were collected in Matlab, a

rural area of Bangladesh with a population of 212,000(at the end of 1996). In this area the InternationalCentre for Diarrhoeal Disease Research, Bangladesh(ICDDR,B) has maintained a Demographic

Surveillance System (DSS) since 1966. Matlab is about55 km Southeast of Dhaka and typical of many ruraland riverine deltaic areas of Bangladesh. Being ¯at and

low lying, the area is prone to annual ¯ooding throughmany canals and rivers which cris-cross the area.Communication with the capital city is di�cult and

within Ð Matlab travel is mostly done on foot or byrikshaw, or during the monsoon, by country boats. Atypical village consists of several baris or groups ofhouses around a central courtyard, which function as

economic and social units. A bari is usually made upof two or more patrilinearly related families. For thefemale members of a household, social contacts outside

the bari are restricted, which means that women typi-cally live within the con®nes of a bari. A substantialproportion of villagers is landless and major sources of

income are ®shing, agricultural labour and share crop-ping.The Matlab DSS area in 1977 was divided into two

sectors: The Maternal and Child Health and FamilyPlanning (MCF-FP) area where an intensive MCH-FPprogramme has been implemented and a comparisonarea where only government health and family plan-

ning services have been provided.DSS is a registration system of vital events (birth

and death), in- and out-migration and marriage and

divorce (Strong, 1992; Mostafa et al., 1998; vanGinneken et al., 1998). Data are collected by teams ofmale and female enumerators who visit all the house-

holds in the study area on a monthly basis. Due to thepresence of this continuous surveillance system it ispossible to determine mortality rates (per 1000 or100,000) for the total population as well as for the 60

yr and older people in the period 1974±1996. In ad-dition, censuses were carried out in the DSS area in1974, 1982 and 1996. They served to verify the infor-

mation collected by the enumerators during their regu-lar rounds and to collect socioeconomic information.By combining the information of regular surveillance

with the information of censuses, longitudinal data setscan be constructed in which information on social andeconomic conditions can be linked to subsequent mor-

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771764

Page 3: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

tality. These data provide, therefore, an opportunity toexamine the association between socioeconomic and

demographic risk factors and mortality of the elderlypopulation. For this study, a longitudinal data ®le wascreated with the records of over 10,000 persons aged

60 yr and over alive during the 1982 Census.Information on subsequent mortality and out-mi-gration was included in this ®le for each individual for

the period 1982±1992.Since Matlab is one of the few places in the develop-

ing world with data of good quality on mortality, it is

possible to make comparisons with other places andcountries. For this purpose death rates in Matlab in1990±1992 are compared with those of Sri Lanka andSweden in about the same period. Sri Lanka was cho-

sen because in South Asia it has the highest life expect-ancy at birth (71.2 yr in 1992) while Sweden wasselected because it has one of the highest life expectan-

cies (77.7 yr in 1992) in industrialized countries(UNDP, 1994).Analysis of the data on factors associated with mor-

tality was done by both bivariate and multivariatemethods. For the multivariate analysis a proportionalhazards life table model was chosen. It is an appropri-

ate statistical model in view of the fact that we havelongitudinal data which allow us to calculate life tablefunctions (mortality risks, hazard rates) and we haveinformation on a number of explanatory variables

which may in¯uence these life table functions. Detailsof assumptions equations and procedures are describedin, for instance, Selvin (1991). In our model, the

dependent variable is the mortality risk in the follow-up period determined with the life table method.Mortality risks are determined and compared for each

of the values of an explanatory variable with theremaining variables held constant. Each individual wasfollowed up for the duration of the study period oruntil death occurred. Information on a number of per-

sons was lost to follow-up at some points between1982 and 1992 as a result of out-migration (censoreddata). Information on social and economic character-

istics is used from the 1982 Census. Changes in charac-teristics, such as family composition on socioeconomicstatus, may have occurred in subsequent years, but

these changes could not be taken into account.

Results

In the DSS area of Matlab, the population (60 yrand older) at the end of 1996 was 14,968 or 7.1% of

the total population. These percentages were 5.6 and5.8 in 1974 and 1982 respectively. Both in 1974 and1982, there were more older men than women (for

example in 1974, 6.0% were men and 5.1% werewomen). In the most recent Census of 1996, the per-

centages of elderly men and women were very similar7.1 for men versus 7.0 for women (Razzaque et al.,

1998).Table 1 shows age-speci®c mortality rates of elderly

people by gender in 1974, 1982, and 1996 in Matlab.

These were chosen because censuses were conducted inthese years.Mortality rates were very high in 1974; this was

probably due to famine which deeply a�ected thecountry in that year. Mortality decreased considerablybetween 1974 and 1982 in all age groups. Between1982 and 1996, mortality has remained nearly the

same. Table 1 also shows important di�erences by gen-der. Mortality of elderly women in 1974 was muchhigher than for men; in 1982 there was hardly any

di�erence by gender; and in 1996, male mortality rateswere higher than female rates in nearly all age cat-egories.

Figure 1 shows details of trends in mortality of theelderly people between 1983 and 1996. This ®gure con-®rms that mortality of elderly population has changedlittle during the past 14 yr.

This is in sharp contrast with trends in other agegroups as can be seen in Fig. 2.

Table 1

Mortality rates of the elderly (60 yr and over) per 1000 popu-

lation by age and sex in di�erent years

Age groups 1974 1982 1996

male female male female male female

60±64 48.6 55.7 35.3 26.2 23.5 19.5

65±69 78.6 82.2 47.0 50.7 46.3 38.4

70±74 103.6 140.0 73.7 78.8 55.3 53.6

75±79 174.3 263.4 105.3 100.8 87.7 92.7

80+ 246.8 345.2 166.7 183.5 137.4 140.2

Total 95.8 116.9 65.6 59.5 51.7 44.1

Fig. 1. Mortality rate of elderly (60+) population by sex in

Matlab, Bangladesh, 1982±1996

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771 765

Page 4: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

Fig. 2. Probability of dying in di�erent age groups in Matlab, Bangladesh, 1983±1996

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771766

Page 5: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

There were sharp declines in mortality among thepopulation belonging to 0±4 and 5±14 yr age groups

between 1983 and 1996 while there were less strongdeclines in the 15±44 and 45±59 yr age groups.It should be noted here that the scales of the y-axes

of the four graphs in Fig. 2 are not the same.

Although there were declines in mortality in the 5±14and 15±44 yr age groups, the level of mortality wasmuch lower in these age groups than in the 0±4 and

45±59 yr age groups.Table 2 shows death rates in Matlab in 1990±1992

compared to Sri Lanka and Sweden in about the same

period while ratios of mortality in Matlab to those inSri Lanka and Sweden are shown in Table 3. Males inMatlab in the 60±74 yr age groups have on average a25 to 85% higher risk of dying than Sri Lankan and

Swedish males. After 80 yr of age, Matlab elderlymales have lower mortality than the Sri Lankan andSwedish males. This unexpected result may be due to

the very small number of deaths in Matlab in the high-est age groups. The corresponding ratios (Table 3) ofMatlab to Sri Lanka and those of Matlab to Sweden

for females are substantially larger than for males.Table 4 indicates that in Sweden mortality of maleswas nearly double that of females in 1991while in

Matlab, male mortality was only slightly higher. Sri

Lanka occupied a position in between Sweden andMatlab.

Table 5 provides, ®rst of all, information on thecomposition of the families of the elderly people. Forinstance, there were more elderly males (5538) thanfemales (4604) in 1982 when the Census took place.

This is a consequence of higher female than male mor-tality rates in all age categories in the decades prior tothe 1980s (Strong, 1992). It can also be seen that 88%

of the men are married (4895 out of 5538) while 77%of the women are widows (3550 out of 4604). Most ofboth men and women live in households with one or

more sons (81% for men and 67% for women), butthere are less daughters (42% of the men and only12% of the women have daughter at home).Table 5 also gives information on the relationship of

a number of demographic and socioeconomic variableswith mortality of elderly people.Persons with these characteristics have higher than

average mortality: being male, occupying another pos-ition in the household other than being the head, beinga widow or widower, being a Hindu, having no son or

one son in the household, having no daughters present,being of low socioeconomic status and being illiterate.Since the data in Table 5 refer to bivariate relation-

ships of the independent variables with mortality, there

Table 2

Mortality rates of elderly persons (60 yr and over) per 1000 in Matlab (Bangladesh), Sri Lanka and Sweden: age-speci®c mortality

rates in three countries

Age groups Males Females

Matlab (1990±92) Sri Lanka (1987) Sweden (1991) Matlab (1990±92) Sri Lanka (1987) Sweden (1991)

60±64 28.5 22.7 14.5 21.1 13.1 7.7

65±69 43.1 35.8 24.1 36.9 23.8 12.3

70±74 69.7 54.1 37.9 61.5 41.0 19.8

75±79 88.3 73.6 66.4 84.5 59.3 37.5

80+ 130.9 152.4 149.3 140.1 156.9 106.2

Total 54.9 48.9 46.6 47.1 37.6 26.1

Table 3

Mortality rates of elderly persons (60 yr and over) per 1,000 in Matlab (Bangladesh), Sri Lanka and Sweden: ratio of Matlab to

Sri Lanka and Swedish mortality

Age group Males Females

Matlab to Sri Lanka Matlab to Sweden Matlab to Sri Lanka Matlab to Sweden

60±64 1.26 1.96 1.61 2.74

65±69 1.20 1.79 1.55 3.00

70±74 1.29 1.84 1.50 3.11

75±79 1.20 1.33 1.42 2.25

80+ 0.86 0.88 0.89 1.32

Total 1.12 1.18 1.25 1.80

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771 767

Page 6: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

is a need to control for other variables. Results of adiscrete-time hazard model showing net associations

are given in Table 6.Widows had a 99% higher risk of dying in old age

than married women (the Reference Category) andwidowers had a 54% higher risk of death than married

men (RC). Women living with at least one son had18% lower mortality than women living in householdswithout a son (RC). There was no e�ect of having a

son (or sons) on mortality of men. Having a daughter(or daughters) in the household reduces the risk ofmortality by about 18% for women and 9% for men.

Female mortality was higher in large families (familysize from 6 to 8 persons) than in small families (RC).Being the head of a household is bene®cial for the sur-vival of males. Educated elderly people and those of

higher socioeconomic status experienced higher survi-val than the uneducated and those of low socioeco-nomic status (RC). Muslims had lower mortality than

Hindus (RC). Elderly persons living in the MCH-FParea enjoyed better health than those in the compari-son area (RC). Mortality increased with age, as was

already noticed in Tables 1, 2 and 3.

Conclusions and discussion

Data were analyzed on levels, trends in levels and

determinants of mortality of the elderly people (60 yrand older) derived from the demographic surveillancesystem maintained in Matlab, a rural area ofBangladesh. Mortality rates of elderly people were

higher in Matlab than in two countries Sri Lanka andSweden which were used as comparisons. Mortality inMatlab was 1.2 times higher than in Sri Lanka and 1.5

times higher than in Sweden.We also found that mortality of the 60 yr and older

population declined substantially between 1974 and

1982, but much less in subsequent years. The smallchange in mortality in 1982±1996 is in contrast to theconsiderable decreases in mortality of children and

adults during the same period. This is in accordance

with the ®ndings of an international study indicatingthat the introduction of medical technology and inno-vations in public health have bene®ted the young and

to a lesser extent, the adults, but not the elderly people(UN, 1982). It should be added here that there is evi-dence to suggest that in theory it should be possible to

introduce certain medical and health interventions inthe older population leading to improvement of health

and well-being in old age (Suzman, 1992; Thatcher,1992; Kannisto, 1994).Mortality of women in old age used to be higher

than of men but between 1980 and 1985, a reversal hasoccurred. It is likely, however, that the overall situ-

ation of women in old age continues to be disadvan-taged compared to men. It is also worth mentioningthat most women in old age are widows (Table 5).

That women in Bangladesh are disadvantaged, can bederived from studies in the 1980s showing that dis-crimination against women was taking place in

Bangladesh at that time (Chen et al., 1981; Bairagi,1986; ). These studies were limited to the younger age

categories, but there is no reason to assume thatunequal treatment of women is limited to childhoodand adolescence. The extent of the unequal treatment

of women may have decreased to some extent since theearly 1980s, but undoubtedly this continues.One of the main ®ndings of our analysis was that

social support in old age by marriage partners and/orby close relatives is an important determinant of mor-

tality. Widows or widowers had much higher mortalitythan elderly people who were married and lived with apartner. Likewise, the elderly people who were living

at home with a daughter or son experienced lowermortality than those who were living without theseclose relatives. Similar results have been found in stu-

dies carried out in developed countries (e.g. Hu andGoldman, 1990) and in a few studies in Asian

countries (Cain, 1986; Rahman et al., 1992; Mostafaand Rahman, 1995; Rahman, 1997).There is some indication that the presence of daugh-

ters in the households of the elderly people may bemore important for survival than the presence of sons.

It is also possible that the presence of sons or daugh-ters may be more bene®cial for elderly women than formen. The evidence we have presented on these points

is, however, inconclusive and more research is there-fore needed. It is also important to mention that theimpact of presence of sons and daughters on survival

of their parents is underestimated in our analysis. Weonly had data on the presence of sons and daughters

living in the households of the elderly people. In ruralBangladesh such households are part of an extendedfamily (a bari). It is likely that sons or daughters were

living in neighbouring households belonging to thesame extended family.

Table 4

Ratio of male to female mortality in Matlab (Bangladesh), Sri

Lanka and Sweden

Age groups Male/female (ratio)

Matlab Sri Lanka Sweden

60±64 1.35 1.73 1.88

65±69 1.17 1.50 1.96

70±74 1.13 1.32 1.91

75±79 1.04 1.24 1.77

80+ 0.93 0.97 1.40

Total 1.17 1.30 1.79

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771768

Page 7: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

Surprising was the ®nding that the impact of one ormore son(s) in the household was not associated withlower than average mortality of elderly men (Table 6).

This result is con®rmed when we look at the bivariaterelationships between presence of one or more son(s)and mortality of the elderly people (Table 5). The pre-sence of two or more sons was associated with lower

mortality, but not presence of one son. We do nothave an adequate explanation for this ®nding. Another®nding which we ®nd di�cult to interpret was that

mortality among the elderly people in the MCH-FParea was lower than in the comparison area (Table 6)in spite of the fact that this was not found with respect

to the bivariate relationship of area with mortality(Table 5). No interventions were implemented in theMCH-FP area which were speci®cally aimed at the

elderly people.We also found that elderly people who are educated

and are of high socioeconomic status are likely to livelonger than the uneducated and those of lower socioe-

conomic status. This was also found in developedcountries (Feldman et al., 1989; Valkonen, 1989;Olausson, 1991; Mackenbach et al., 1997) and to some

extent in developing countries (Rahman et al., 1992).There is much evidence to show that socioeconomicfactors, in particular education of the parents, are im-

Table 5

Death rates of elderly (60+) per 1000 population by selected variables included in the regression analysis

Variables Male Female

# Death # Popu Rate #Death # Popu Rate

Total (60+) 2915 5538 526.4 2272 4604 493.5

Position in household

Head 2211 4526 488.5 229 526 435.4

Others 704 1012 595.7 2043 4078 501.0

Marital status

Single 5 7 714.3 2 2 1000.0

Married 2494 4895 509.5 385 1009 381.6

Divorce 25 45 555.6 24 43 558.1

Widow 391 591 661.6 1861 3550 524.2

Religion

Muslim 2419 4673 517.7 1888 3902 483.9

Hindu 496 865 573.4 384 701 547.8

No. of living sons

No son 571 1039 549.6 751 1498 501.3

One son 1137 1904 597.2 1367 2549 536.3

2+ sons 1207 2595 465.1 154 557 276.5

No. of living daughters

No daughter 1863 3215 579.5 2049 4051 505.8

1 daughter 681 1415 481.3 208 502 414.3

2+ daughters 371 908 408.6 15 51 294.1

Family size

< 6 1111 2048 542.5 794 1813 437.9

6±8 1067 2063 517.2 974 1798 541.7

9+ 737 1427 516.5 504 992 508.1

Occupation

High status 1609 3336 482.3 ± ± ±

Low status 1306 2202 593.1 ± ± ±

Items owned

None 554 945 586.2 795 1007 789.5

1+ 2361 4593 514.0 1777 3597 494.0

Education

0 1628 2803 580.8 2174 4297 505.9

1±4 673 1300 517.7 72 203 354.7

5+ 611 1426 428.5 21 81 259.3

Area

MCH-FP (treatment) 1539 2886 533.3 1154 2315 498.5

Comparison 1376 2652 518.9 1118 2289 488.4

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771 769

Page 8: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

portant determinants with respect to survival in child-

hood (Hobcraft et al., 1984; Cleland and van

Ginneken, 1988). We have found that socioeconomic

factors are also important determinants of mortality in

old age. More research on the precise contribution of

di�erent socioeconomic factors to health and welfare

in old age is needed. In carrying out such studies,

more sophisticated mortality measures should be used

than we have done. Mortality rates used in Table 5

should, in the future, be based on person-years of ob-

servation as the denominator.

Before moving to the topic of policy implication of

our ®ndings, the question needs to be addressed if and

to what extent the results found in Matlab can be gen-

eralized for rural Bangladesh as a whole. An accurate

reply to this question is not possible, because of

absence of accurate ®gures on adult or old-age mor-

tality in rural Bangladesh. It is our impression that the

situation for the elderly population in Matlab is fairly

representative for those in rural Bangladesh in general.

An implication of our study is that the Government

of Bangladesh as well as those of other developing

countries should take more interest in the health and

welfare of elderly people. A national policy for the

elderly population should be formulated leading to

improvement in health care and social security.

Measures should be taken by the Government as well

as non-government organizations to provide health

care and social security, especially for the elderly

widowers and widows who do not receive proper care

and support from sons, daughters and other relatives.

Acknowledgements

This research was funded by the Department for

International Development of the United Kingdomand ICDDR,B: Centre for Health and PopulationResearch which is supported by countries and agencieswhich share its concern for the health problems of

developing countries. Current donors providing unrest-ricted support include: the aid agencies of theGovernments of Australia, Bangladesh, Belgium,

Canada, Saudi Arabia, Sweden, Switzerland, theUnited Kingdom and the United States of America; in-ternational organizations include United Nations

Children's Fund (UNICEF).We also thank Dr. Mizanur Rahman for his valu-

able suggestions on an earlier version of this paper.

References

Bairagi, R., 1986. Food crisis, child nutrition and female chil-

dren in rural Bangladesh. Population and Development

Review 12, 307±315.

Berkman, L.F., Syme, L.S., 1979. Social networks, host resist-

ance and mortality. American Journal of Epidemiology

109, 186±204.

Bowling, A., 1987. Mortality after bereavement: a review of

the literature on survival periods and factors a�ecting sur-

vival. Social Science and Medicine 24, 117±124.

Cain, M., 1986. The consequence of reproductive failure:

Dependence, mobility and mortality among the elderly of

rural South Asia. Population Studies 40, 375±388.

Table 6

Discrete time hazard regression coe�cients of the e�ect of demographic and socioeconomic variables on mortality of persons aged

60 and over in Matlab, Bangladesh. �P < 0.05; ��P< 0.01; ���P < 0.001 (two-tailed tests)

Independent variablea Coe�cient Relative risk

males

(n=5510)

females

(n=4562)

males females

Widow(er) 0.435��� 0.689��� 1.54 1.99

Son in household ÿ0.042 ÿ0.193�� 0.96 0.82

Daughter in household ÿ0.097� ÿ0.201� 0.91 0.82

Family size (6±8) b 0.196�� b 1.22

Head of household ÿ0.267��� ÿ0.102 0.77 0.90

Literate ÿ0.144��� ÿ0.301�� 0.87 0.74

High SES ÿ0.417��� b 0.66 b

Household items owned ÿ0.158��� ÿ0.211��� 0.85 0.81

Muslim ÿ0.233��� ÿ0.341��� 0.79 0.71

MCH-FP Area ÿ0.090� �0.122�� 0.91 0.88

Age 0.062��� 0.073��� 1.06 1.07

Constant ÿ5.040��� ÿ6.816��� ± ±

ÿ2(log likelihood) 15,212.9 12,037.1 ± ±

a For description of reference categories, see text.b Not included in the model.

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771770

Page 9: Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh

Chen, C.L., Huq, E., D'souza, S., 1981. Sex bias in the family

allocation of food and health care in rural Bangladesh.

Population and Development Review 7, 55±60.

Cleland, J.G., van Ginneken, J., 1988. Maternal education

and child survival in developing countries: The search for

pathways of In¯uence. Social Science & Medicine 27,

1357±1368.

Feldman, J.J., Makuc, D., Kleinman, J., Corononi-Huntley,

J., 1989. National trends in educational di�erences in mor-

tality. American Journal of Epidemiology 129 (5), 919±

933.

Hobcraft, J.N., McDonald, J.W., Rutstein, S.O., 1984. Socio-

economic factors in infant and child mortality, a cross

national comparison. Population Studies 38, 193±223.

Hossain, M.S., 1997. Population growth and structure. In:

Barkat, A., Howlader, S.R. (Eds.), Population and

Development Issues in Bangladesh. Ministry of Health

and Family Welfare, Government of Bangladesh, Dhaka,

pp. 85±114.

Hu, Y., Goldman, N., 1990. Mortality di�erentials by marital

status: An international comparison. Demography 27,

233±250.

Kannisto, V., 1994. Development of oldest-old mortality,

1950±1990: Evidence from 28 developed countries. In:

Monograph of Population Aging No. 1. Odense

University Press, Odense, Denmark.

Mackenbach, J.P., Kunst, A.E., et al., 1997. Socio-economic

inequalities in morbidity and mortality in western Europe:

a comparative study. Lancet 349, 1655±1659.

Martin, L.G., Kinsella, K., 1994. Research on the demogra-

phy of aging in developing countries. In: Martin, L.S.,

Preston, S.H. (Eds.), Demography of Aging. National

Academy Press, Washington, DC, pp. 356±403.

Mitra, S.N., Al-Sabir, A., Cross, A.R., Jamil, K., 1997.

Bangladesh Demographic and Health Survey, 1996±1997.

National Institute of Population Research and Training

(NIPORT), Mitra Associates and Macro International

Inc., Dhaka and Calverton, Maryland.

Mostafa, G., Shaikh, M.A.K., van Ginneken, J.K., Sarder,

A.M., 1998. Demographic Surveillance System Ð Matlab,

Registration of Demographic Events Ð 1996, Volume 28,

Scienti®c Report, No. 79. ICDDR,B, Dhaka.

Mostafa, G., Rahman, M., 1995. Do daughters also provide

old age security to parents? Paper presented at the

Population Association of America (PAA), April 1995,

San Francisco.

Myers, G., 1992. Demographic aging and family support for

older persons. In: Kendig, H., Hashimoto, A., Coppard,

L. (Eds.), Family Support for the Elderly: The

International Experience. Oxford University Press, New

York, pp. 31±69.

Olausson, P.O., 1991. Mortality among the elderly in Sweden

by social class. Social Science & Medicine 32, 437±440.

Rahman, O., Foster, A., Menken, J., 1992. Older widow mor-

tality in rural Bangladesh. Social Science & Medicine 34,

89±96.

Rahman, O., 1997. The e�ects of spouses on the mortality of

older people in rural Bangladesh. Health Transition

Review 7, 1±12.

Razzaque, A., Nahar, L. Sarder, A.M., van Ginneken, J.,

Shaikh, M.A., 1998. Demographic Surveillance System-

Matlab, 1996 Socio-Economic Census, Volume 29,

Scienti®c Report No. 83. ICDDR,B, Dhaka.

Selvin, S., 1991. Statistical Analysis of Epidemiologic Data.

Oxford University Press, New York.

Smith, D.W.E., 1993. Human Longevity. Oxford University

Press, New York.

Strong, M.A., 1992. The health of adults in the developing

world: the view from Bangladesh. Health Transition

Review 2 (2), 215±224.

Suzman, R.M., 1992. The robust oldest old: Optimistic per-

spectives for increasing healthy life expectancy. In:

Suzman, R.M., Wills, R.P., Manton, K.G. (Eds.), The

Oldest Old. Oxford University Press, New York, pp.

1151±1159.

Thatcher, A.R., 1992. Trends in numbers and mortality at

high ages in England and Wales. Population Studies 46,

411±426.

United Nations, 1982. Levels and Trends of Mortality since

1950: A Joint Study of the United Nations and World

Health Organization. United Nations, New York (ST/

ESA/SER.A/74, Sales No.E.81.XIII.3).

United Nations, 1998. World Population Prospects: The 1996

Revision. Department of Economic and Social A�airs,

United Nations, New York.

United Nations Development Programme, 1994. Human

Development Report 1994. UNDP, New York.

Valkonen, T., 1989. Adult mortality and level of education: a

comparison of six countries. In: Fox, J. (Ed.), Health

Inequalities in European Countries. Gower Publishing Co,

Aldershot, pp. 142±162.

van Ginneken J., Bairagi, R., Francisco A.D., Sarder A.M.,

Vaughan. P., 1998. Health and Demographic Surveillance

in Matlab: Past, Present and Future. Special publication

No. 72. ICDDR,B, Dhaka.

G. Mostafa, J.K. van Ginneken / Social Science & Medicine 50 (2000) 763±771 771