Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
7 Mortality and Morbidity: Is City Life Good for Your Health? What is it that is distinctive about urban health? Health is both an outcome and a determinant of economic development and in this way must be associated with urbanization. But when considering health in cities and rural villages, can one identify a distinctive urban health profile, or are the urban/rural differences less a matter of kind than of degree? In what ways are the concepts of urban diversity and social-spatial proximity reflected in health? The concept of an epidemiological or health transition provides a starting point for discussion of these questions. This transition is expressed in a shift from a situation in which communicable diseases are the primary causes of morbidity and mortality to one in which noncommunicable diseases predominate. As will be shown, the health transition is well under way in some cities, especially in the developing countries that have relatively high levels of income per capita. As the transition proceeds, urban populations will experience relatively more chronic disease, including cancers and heart disease; mental health will also be of growing concern. In many and perhaps most cities, however, the health transition is still in its early stages, and these cities will continue to grapple with communicable diseases for the foreseeable future. In addition to such long-standing challenges to health, some cities will face grave threats from new diseases (e.g., HIV/AIDS) and diseases that are reemerging with heightened virulence or resistance (e.g., tuberculosis). The spatial proximity of urban residents and their reliance on common public resources leave them more vulnerable to communicable health threats than are ru- ral residents, who enjoy a measure of protection owing to their spatial dispersion. This "urban penalty" was first observed in the Victorian era when city dwellers died at higher rates than their rural counterparts despite their greater average in- comes, but it has been in force throughout most of human history. It was only 259
260 CITIES TRANSFORMED when urban populations had begun to be protected by public health investments and when advances in the understanding of disease had progressed to the point that higher incomes could purchase effective treatments that urban populations could achieve higher levels of health on average (Preston and van de Walle, 1978; Ewbank and Preston, 1990; van Poppel and van der Heijden, 1997~. In a sense, then, the urban advantage that we now take for granted is a recent and possibly fragile development. Nothing locks this advantage into place. De- teriorating economic conditions, disinvestment by governments in urban public health infrastructure, and newly virulent communicable diseases could conceiv- ably cause the penalties last seen in the nineteenth century to reemerge. Of course, some urban groups may never have enjoyed much of an urban advantage; the poor and the politically overlooked may have been at least as exposed to disease as their rural cousins, and perhaps more so. In this chapter, we look carefully for evidence of a reemergent urban penalty. As will be seen, the evidence is mixed and contradictory, but that in itself may present a challenge to complacency. The uneven distribution of health is clearly apparent within cities; it can be seen, for example, in the spatial variation of crude death rates in Accra, depicted in Figure 7-1. Although age-adjusted mortality measures would be preferred, the variation in crude death rates is suggestive of stark intraurban inequities that are unlikely to be due to age composition alone. The spatial concentration of poor health has long been recognized by epidemiologists, and its economic and social aspects are increasingly being emphasized in the public health journals. One now _ . A::' _..::::::::: ~ ':::::::::::: . it: - :,:,:,:,:,:,:,:,:,:,:,:,:,:,:,: ::::::::::::: :::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,:,::::::::::: l .............................................................................................................................. _ .~. ~ 1 ~ ::::::::::::::::::::::::::::~:::::~ ~ . . ~ ~ L"""""""""""""~............................ ' ~ s . ~ :.:.:.:.:.:.:.:.:.:.:.:.:.:~:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.~ ~.~ c ~ ~ ~ i .1 / ~~ . ~ ~::: ::::::::::::: :::::::: ~ ~ ~ A1~'''''''''''''''''''''''''""""""""""""""""""""""""""""'~ 222222222222222222222'''':1::::::::::::::::::::::::::::: t . :-:-:-:-:-:-.~:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:-:= A.:.:.:.:.:.:.: :.:.:.:.:.:.:-:.: A--- i -- - I................ t:::::::::::::::::~::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::t 1"""""""""""""""""'' ~"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""~ ~ """"'I ~"""""""""""""""""""""""""""""""""""""""""""""'~ :.: ~ ~ :::::~.:.:::::::::::::E :.:.:.:.:.:.:.:.:.:.:.:.:.~ ~ :,: ~ . ~ """""""""""` :,: ~ ~ ~ ........................ ~ I _ _ ., ~........................... ............. .:.:.:.:.:,:,:ja C <3.0 · 7.0-9.0 C 3.0-5.0 · ,9.0 1 ~ 5.0-7.0 FIGURE 7-1 Crude death rates by neighborhood in Accra, 1991. SOURCE: Stephens, Timaeus, Akerman, Avlve, Maia, Campanario, Doe, Lush, Tetteh, and Harpham (1994b).
MORTALITY AND MORBIDITY 261 sees much discussion of the roles of local social capital and social networks in ur- ban health, and efforts are under way to bring empirical content to these concepts. Although few longitudinal studies of urban sites are fully functioning as of this writing (see Kahn and Tollman, 2002, for a list of new urban research sites), there is some prospect for linking sophisticated prospective social science research to the sophisticated programs of epidemiological research already under way. The socioeconomic diversity of urban populations makes possible the devel- opment of many specialized markets and functions. Some of these can directly affect the communicability of disease; consider, for example, the markets in which sex workers participate. Rural prostitution exists, of course, but its epidemiolog- ical role in the spread of sexually transmitted diseases (STDs) and HIV/AIDS is quite different from that of urban prostitution, which probably involves denser and more highly interconnected sexual networks (Yirrell, Pickering, Palmarini, Hamil- ton, Rutemberwa, Biryahwaho, Whitworth, and Brown, 1998; Pickering, Okongo, Ojwiya, Yirrell, and Whitworth, 1997~. At the same time, urban diversity supports the development of markets that are beneficial to health. As was seen in the pre- vious chapter in connection with reproductive health, the role of the private sector is far more prominent in urban than in rural areas. Urban health providers in the public sector operate alongside a great variety of private-sector providers, who range from traditional healers to highly specialized surgeons. Various fee-for- service arrangements are found in this variegated private sector, and fees are seen increasingly in the public sector as well. These arrangements raise issues of abil- ity and willingness to pay on the part of urban residents. The urban poor who are unable to pay fees may remain as dependent on subsidized public-sector services as their rural counterparts. This chapter begins by considering the distinctive aspects of urban health in more detail. Then, guided by the concept of the health transition, it reviews the spectrum of diseases that afflict urban residents and compares this spectrum with the rural burden of disease. Much of the discussion concerns adult health; spe- cial emphasis is placed on risks for injuries and mental ill-health that have interesting urban features and have too readily been overlooked. The chapter next turns to children's health, presenting results from analyses of the Demographic and Health Surveys (DHS) on children's nutrition (as measured by height for age and weight for height) and child survival. (The panel could not explore trends over time going back to the DHS predecessor surveys the World Fertility Surveys (WFS) but this would be a useful exercise for future research.) Although the DHS data are limited in their spatial resolution and usually do not allow neighborhood effects to be discerned, they do enable a general assessment of whether urban children suffer from a newly emergent urban penalty. To this end, results from the DHS surveys are compared with findings from spatially fo- cused studies of urban slums and squatter settlements. The chapter then turns to a discussion of treatment-seeking behavior in urban populations and the new con- figurations of urban health systems. The final section presents conclusions and recommendations.
262 CITIES TRANSFORMED DISTINCTIVE ASPECTS OF URBAN HEALTH To understand whether and how urban health is distinct from rural, two questions need to be addressed: (1) whether the determinants of urban health are the same as those of rural health, and if so, (2) whether the levels of these determinants differ. Among the subdisciplines of health research, one finds such questions addressed most directly in the field of urban nutrition. Ruel, Haddad, and Garrett (1999: 1887) argue that there are important, urban-specific aspects of food security and nutrition, which include the following: (a) the greater dependence of urban dwellers on cash income and less reliance on surrounding natural resources, which means that having a secure source of income is critical; (b) the likelihood that, even if formal safety net programs are more prevalent, informal social networks are weaker, possibly reducing the ability of the poor to deal with economic shocks; (c) the higher levels of women who par- ticipate in the formal labor force and work away from home, with potentially negative consequences for their ability to care for other household members and their children; (d) the changes in diet and exercise patterns that may increase risk of chronic disease and obe- sity even among low-income groups; (e) the significant obstacles that the poorest segments of the population may still face to gain access to public services, such as water, sanitation, and garbage disposal, even when these services are more available than in rural areas; (f) the increased exposure of urban-dwellers to environmental contamina- tion, which increases risk of illness and, especially, infectious disease; and (g) the legal obstacles that urban-dwellers face in attempting to improve their livelihoods in such areas as employment, land, and water use. Box 7.1, which draws on Harpham, Lusty, and Vaughan (1988) and Ruel and Gar- rett (1999), shows how the characteristics of urban life may be expressed in health outcomes and the distribution of illness. Poverty, environment, and psychosocial problems may lead to specific diseases or health conditions that are more com- mon in urban areas, including STDs, accidents, and depression. However, some conditions, including malnutrition and malaria, are prevalent in both urban and rural areas. THE DISEASE SPECTRUM In Latin America, the health transition has generally appeared first in urban lo- cations (Pan American Health Organization, PAHO; Tanner and Harpham, 1995) and has tended to proceed more rapidly in countries with higher levels of
MORTALITY AND MORBIDITY 263 BOX 7.1 Urban Problems and Their Health Implications Poverty-Related Problems Environmental Problems Psychosocial Problems Cash income and markets Hazardous informal Stress High intake of fats and sector Alienation refined sugars ("junk Inadequate water and Instability food") sanitation Out-of-home female labor force participation Prostitution Street children Within-house and community overcrowding Lack of land to grow food Lack of rubbish disposal Traffic Industrial pollution Air, water, and food contamination Implications Insecurity Smoking Drug abuse (including alcohol) Limited social support Insecurity of tenure Violence Lack of breastfeeding Accidents Depression Malnutrition Parasitic disease Anxiety Sexually transmitted Malaria Suicide diseases, including Dengue Cancer HIV/AIDS Respiratory infections Heart disease Tuberculosis Injuries Cancer Other infectious diseases (especially diarrhea!) urbanization. A link to urbanization is seen in studies of other developing regions as well. These studies document higher urban rates of cardiovascular disease, cancers, coronary heart disease, and accidents, and higher rural rates of malaria, malnutrition, maternal mortality, and respiratory disease.) Evidently, urban and rural health profiles do differ. ~ On the rural disease pattern, see Mbizvo, Fawcus, Lindmark, and Nystrom (1993), Mock, Sellers, Abdoh, and Franklin (1993), McCombie (1995), Fawcus, Mbizvo, Lindmark, and Nystrom (1995, 1996), and Root (1997). On the urban pattern, see reviews by Beevers and Prince (1991), Muna (1993), Walker (1995), Walker and Sareli (1997), and Walker and Segal (1997), and studies by McPake et al. (1999), Steyn, Fourie, Lombard, Katzenellenbogen, Bourne, and Jooste (1996), Ceesay, Morgan, Kamanda, Willoughby, and Lisk (1996), and Delpeuch and Marie (1997). Some of these studies present data on risk factors associated with disease, rather than rates of disease as such.
264 CITIES TRANSFORMED The reasons why they differ are of course complex, stemming from the many environmental, socioeconomic, and cultural changes that are associated with ur- banization. Lower rates of communicable disease and urban advantages in child survival can be linked to lower urban fertility, better immunization coverage, and easier access to and greater use of health services.2 As communicable diseases be- gin to decline in urban areas, noncommunicable diseases (including the so-called "diseases of affluence") and diseases associated with social instability can be ex- pected to rise in relative importance, particularly among adults (Feachem, Kjell- strom, and Murray, 1990~. The growing impact of mental ill-health, violence, accidents, and chronic disease is evident in disability-adjusted life year (DALY) predictions.3 It has been predicted that by 2030, depression, traffic accidents, and heart disease will be the leading burdens of disease in developing countries overall, replacing the 1990 leaders, which were respiratory disease, diarrhea, and perinatal conditions (World Health Organization, 1996~. Health transitions generally progress unevenly, producing more rapid change in some population groups than in others. This unevenness can be seen in com- parisons of adult and child mortality. In some cities, data registers allow the main causes of death to be identified and permit comparisons of adult and child causes. A 1996 analysis by Ministerio de Salud de Peru (1996) showed the main causes of adult death in Managua, Nicaragua, to be acute respiratory infection (which accounted for 11 percent of adult deaths), hypertension (8 percent), road traffic accidents (7 percent), stroke (6 percent), and pneumonia (5 percent). Adults in Managua die from a mix of communicable and noncommunicable diseases. The children of this city, however, still die mainly of communicable diseases, which account for the five leading causes of death. Acute respiratory infections and diar- rhea alone are responsible for 39 percent of infant and child deaths.4 Other cities in Latin America have reached a more advanced stage of the health transition. In San Pedro Sula, a rapidly growing city in Honduras with about half a million residents, the Honduras Ministry of Health (1999) found the main causes of adult death in 1999 to be violence (43 percent), cardiovascular disease (19 percent), cir- rhosis (15 percent), cancer (13 percent), and AIDS (8 percent). Even child deaths are dominated by noncommunicable causes (by rank, violence, cancer, cranial trauma, and anemia). 2See Bah (1993), Bahr and Wehrhahn (1993), Taylor (1993), Fawcus, Mbizvo, Lindmark, and Nystrom (1995), Brockerhoff (1994, 1995a), Brockerhoff and Brennan (1998), and Gould (1998). 3 The DALY measure combines the years of life lost as a result of premature death with years spent in an unhealthy state. A death is premature if it occurs before age 82.5 for a woman and age 80 for a man, these being the life expectancies achieved in Japan, which is the world's current leader in longevity. To be included in the DALY measure, disabilities must be classified by severity and duration such that one DALY is equivalent to a year of fully healthy life. These concepts and measures are explained in the executive summary of The Global Burden of Disease and Injury Series, available at http://www.hsph.harvard.edu/organizations/bdu/summary.html. 4These figures should be interpreted cautiously; the quality of cause-of-death reporting is not known.
MORTALITY AND MORBIDITY TABLE 7-1 Disability-Adjusted Years of Life Lost in Mexico, by Cause and Residence 265 Cause Diarrhea Pneumonia Homicide and violence Motor vehicle-related deaths Cirrhosis Anemia and malnutrition Road traffic accidents Ischemic heart disease Diseases of the digestive system Diabetes mellitus Brain vascular disease Alcoholic dependence Accidents (falls) Chronic lung disease Nephritis Rural Urban Rural Rank Urban Rank Rural/Urban 12.0 9.3 9.2 7.9 7.5 6.8 5.5 5.1 4.7 4.1 3.0 3.0 2.8 2.6 2.2 1 2 3 4 5 6 7 9 10 11 11 13 14 15 2.8 3.9 7.4 8.3 6.3 2.4 6.8 5.3 1.7 5.7 3.0 1.9 2.6 1.9 2.2 11 6 15 13 10 13 12 4.28 2.39 1.23 0.95 1.19 2.86 0.81 0.96 2.74 0.72 1.02 1.56 1.09 1.39 1.01 N(4E: 1991 estimates, expressed per 1000 population. SOURCE: Lozano, Murray, and Frenk (1999: 130~. Not only do cities differ as to their stage in the health transition, but some countries challenge the generalization that urban areas take the lead in the transi- tion. As Table 7-1 shows for Mexico, the 15 leading causes of DALYs lost in rural and urban areas are the same, although they appear in different rank order. Of the top five causes in urban areas, three (deaths related to motor vehicles, homicide and violence, and cirrhosis) are also in the top five in rural areas. The implications of the health transition are far-reaching, encompassing fac- tors as various as the range of drugs needed in urban primary health centers and the emphases required for effective health promotion programs. The discussion that follows focuses mainly on adults and addresses three types of disease injuries, mental health, and "lifestyle" diseases whose impact has not been sufficiently appreciated. I· ~ nJurles Drawing on a DALYs analysis, Zwi, Forjuoh, Murugusampillay, Odero, and Watts (1996: 593) call attention to the effects of injuries on health and well-being: ... world-wide, intentional injuries (suicide, homicide and war) ac- count for almost the same number of DALYs lost as either sexually transmitted diseases and human immunodeficiency virus (HIV) infec- tion combined, or tuberculosis. Unintentional injuries cause as many
266 CITIES TRANSFORMED DALYs lost as diarrhea, and more than those lost from cardiovas- cular disease, malignant neoplasm, or vaccine-preventable childhood infections. In developing regions of the world, in 1990, injuries in males aged 15-44 years led to 55 million DALYs lost, over one-third of those lost from all causes in this sex and age group. As this quotation makes clear, violence is one of the major causes of injuries. Much of the empirical work on violence has been carried out in Latin America and the Caribbean. This is for good reason: Latin America has the world's highest burden of homicides, which occur at a rate of 7.7 per 1,000 population, more than twice the world average of 3.5 per 1,000. Approximately 30 percent of all homicide victims in Latin America are adolescents, and young men are the most affected group (Frenk, Londono, Knaul, and Lozano, 1998; Pan American Health Organization, PAHO, cited in Grant, l999~. Violent crime is particularly prevalent in Latin America's large cities, and in these cities, it disproportionately affects men living in low-income neighbor- hoods (Barata, Ribeiro, Guedes, and Moraes, 1998; Grant, l999~. Data collected between 1991 and 1993 in Sao Paulo suggested that men aged 15-24 in low- income areas were more than 5 times likelier to fall victim to homicide than were men of the same age in higher-income areas (Soares et al., cited in Grant, 1999~. Gender roles and relations put men and women at risk of different types of violence. Higher rates of homicide are reported for men, but rape and domestic violence rates are higher for women. Heise, Raikes, Watts, and Zwi (1994) re- viewed community-based data for eight urban areas from different regions of the developing world and found that mental and physical abuse of women by their partners was common. with damaging consequences for women's physical and r ~ ~ ~ , ~ be, a, _ ~ En_~ psychological well-being. Traffic accidents are another major but often overlooked cause of urban death and injury (Mock, Abantanga, Cummings, and Koepsell, 1999; Kayombo, 1995; Byarugaba and Kielkowski, 1994~. Urban residents are often thought to be at greater risk of being involved in an accident than rural residents (Odero, Garner, and Zwi, 1997~. This supposition enjoys some empirical support (Mock, Aban- tanga, Cummings, and Koepsell, 1999), but higher rural accident rates at least for accidents causing injuries have been recorded (Odero, 1995~. Poor coun- tries invest less in their roads than do rich countries; they have fewer laws related to traffic and enforce them unevenly; and they probably have rates of alcohol con- sumption that are at least as high as those of rich countries. The combination puts men, adolescents, and young adults at particularly high risk of involvement in an accident. Traffic accidents account for 30 to 86 percent of all trauma-related hospital admissions, with a mean length of stay of 20 days (Odero, Garner, and Zwi, 1997, citing 15 and 11 studies, respectively). Since the majority of trauma facilities are located in cities, accidents not only are responsible for significant
MORTALITY AND MORBIDITY 267 mortality and morbidity among urban residents, but also place a heavy burden on urban health systems. Little is known about the urban incidence of other types of accidents, such as accidental falls, drownings, poisonings, and injuries from fire. Some unintentional injuries are likely to be more common in urban areas because of overcrowding and related factors (Bartlett, Hart, Satterthwaite, Barra, and Missair, 1999; Mock, Abantanga, Cummings, and Koepsell, 1999; Knobel, Yang, and Ho, 1994~. Urban communities are vulnerable to some forms of natural disaster, such as landslides, earthquakes, and floods. The urban poor in Rio de Janeiro, for example, are forced to live where landslides kill or leave homeless thousands every year.5 Mental Health According to the World Health Organization (1996), by 2020 unipolar depression is expected to account for the greatest burden of disease in developing countries. Indeed, community-based studies of mental health in developing countries already show that 12 to 51 percent of urban adults suffer from some form of depression (see 16 studies reviewed by Blue, l999~. Although these studies employ a range of samples, definitions, and instruments, their conclusions underscore the importance of mental ill-health in the urban spectrum of disease. A diverse set of risk factors is implicated, including lack of control over resources, changing marriage patterns and increased divorce rates, cultural ideology, long-term chronic stress, exposure to stressful life events, and lack of social support (Harpham, 1994~. There are differences across urban neighborhoods and even from person to person in abili- ties to devise successful coping strategies. Anxiety and depression are typically more prevalent among urban women than men and more prevalent in poor than in nonpoor urban neighborhoods. Those suffering from stress may be able to call upon various forms of emotional support, as well as material support in the form of goods, services, and information. These resources can help in coping with stress and mitigating its damage (Thoits, 1995~. Nonetheless, urban environments in general, and poor urban environments in particular, have a number of harsh physical and social aspects, including poor housing and services and limited prospects for good jobs and incomes (Ekblad, 1993; Fuller, Edwards, Sermsri, and Vorakitphoka- torn, 1993; Satterthwaite, 1993, 1995~. Day-to-day life in poor communities can subject individuals to sustained, chronic stress. As discussed in Chapter 5, poor urban residents often show great resilience and creativity in meeting such chal- lenges. Nevertheless, they can be beaten down by the chronic stresses of poverty, jolted by other stressful life events, and wearied by the constant need to improvise new coping strategies. 5Of the 568 major natural disasters that occurred in the world between 1990 and 1998, 94 percent took place in developing countries, and 97 percent of all natural disaster-related deaths occurred in those countries as well (World Bank, 2001).
268 CITIES TRANSFORMED In empirical analyses, differences in social support the degree to which basic social needs are gratified through interaction with others have been estimated to account for 5 to 10 percent of the variance in levels of mental ill-health (of all types) in different areas (Harpham, 1994; Committee to Study Female Morbidity and Mortality in Sub-Saharan Africa, 1996; Aidoo and Harpham, 2001; Harpham and Blue, 1995~. Urbanization can be associated with reductions in social support resulting from the breakdown or reorganization of family life, a potential increase in single-parent households and decrease in the support networks of extended fam- ilies, reduced fertility (and thus fewer children to care for parents in old age), and the need to work outside the home (Harpham, 1994; Parry, 1995; Harpham and Blue, 1995~. Although some of these aspects of stress appear to be worse in large developing-country cities than in small cities, researchers have not yet compared levels of social support and stress in large and small cities. Another knowledge gap has to do with the interrelationships between mental illness and social support. Most of the available research is cross-sectional, leav- ing open the possibility that mental illness itself reduces social support and that the positive relationship between social support and mental health may be over- stated. Also, the literature has yet to explore the contribution of community-level factors, such as levels of violence and social cohesion (Blue and Harpham,1998~. Regarding the latter, an ecological variable that may play a role is social capi- tal, or the density and nature of the network of contacts and connections among individuals in a given community. Strong social capital has been linked to re- duced mortality at the state level in the United States (National Research Council, 2000~. In low-income urban communities, social capital has been found to weaken as households' ability to cope decreases and community trust breaks down, and to be severely eroded by various forms of violence (Moser and McIlwaine, l999~. Chronic "Lifestyle" Diseases As noted above, urban areas have higher risk factors for and rates of diabetes, obesity, cardiovascular disease, cancers, and coronary heart disease. These are sometimes termed chronic "lifestyle" diseases, the idea being that they are at least partly attributable to behavior. Risk factors associated with this group of diseases include smoking; alcohol consumption; increased intake of fat and reduced in- take of fiber; lack of exercise; and inhalation of potentially toxic pollutants, such as carbon monoxide, sulfur dioxide, nitrogen oxides, and suspended particulate matter. In many developing countries, substantial proportions of the population are either underweight or overweight, with the increase in the overweight percentage being a recent development. One review (Delpeuch and Marie, 1997) suggests that over 30 percent of the national population is overweight in Latin America, the Caribbean, the Middle East, and Northern Africa. (The highest prevalence of obesity is found in Pacific and Indian Ocean island populations.) In Asia and
MORTALITY AND MORBIDITY 269 sub-Saharan Africa, the prevalence of obesity is low on average but is evidently higher in urban than in rural areas. Obesity tends to appear first among the affluent and then among low-income groups, including young children and teenagers. Its main causes include the adoption of lipid-rich diets and (more important) the re- duction in physical activity that often accompanies city life. Malnutrition during fetal development and early childhood is a predisposing factor for later obesity. Undernutrition, food insecurity, dietary excess, and obesity often coexist in urban populations. Popkin (1999: 1908) shows that more urbanized developing countries have a higher consumption of sweeteners and fats, noting that "a shift from 25 percent to 75 percent urban population in very low income countries is as- sociated with an increase of approximately four percentage points of total energy from fat and an additional 12 percentage points of energy from sweeteners." A1- though this pattern is often attributed to the urban rich, Monteiro, Benicio, Conde, and Popkin (2000) show that in urban Brazil, it is the city residents with more education who are less likely to be overweight. Much the same pattern has been observed in South Africa (South African Department of Health, 1998~. In Latin America, there is an upward trend in cancer mortality, which is especially marked for cancers of the lung, gallbladder, and breast (Timaeus and Lopez, 1996~. In Accra and Sao Paulo, circulatory disease has been found to be the second most important cause of death among those aged 15-44 and the most important cause for those aged 45-64 (Stephens, Timaeus, Akerman, Avlve, Maia, Campanario, Doe, Lush, Tetteh, and Harpham, 1994a). Community-based urban studies among the elderly likewise have documented high rates of mortality and morbidity due to chronic and lifestyle diseases (Belle, Baiyewu, Bamigboye, Adeyemi, Ikuesan, and Jegede, 1993; Allain, Wilson, Gomo, Mushangi, Senzanje, Adamchak, and Matenga, 1997~. The prevalence and increase of risk factors for these diseases among urban populations, coupled with gradual population aging, imply that they will become increasingly important causes of death. Yet they remain poorly described, particularly in African and Asian cities. Unfortunately, the DALYs data available from the Global Burden of Disease studies have not been systematically disaggregated by rural and urban place of residence. Diagnoses of chronic diseases may be better in urban than rural areas, and the studies cited above suggest that urban populations will continue to be at the forefront of health transitions. It remains uncertain just how urban disease patterns will be influenced by urban population growth, poverty, and emerging and reemerging communicable diseases. New and Reemergent Communicable Diseases Communicable diseases continue to be important causes of adult mortality in many urban areas. In Dar es Salaam, HIV/AIDS is the main cause of death among urban men and, along with maternal mortality, is the primary killer of urban women (Kitange, Machibya, Black, Mtasiwa, Masuki, Whiting, Unwin,
270 CITIES TRANSFORMED Moshiro, Klima, Lewanga, Alberti, and McLarty, 1996~. This study also found that when desperately ill, city residents sometimes return to their rural homes to die: 11 and 19 percent of the adult deaths recorded in two rural study sites oc- curred to formerly urban residents who had become ill. Where this pattern is com- mon, it may result in underestimates of adult mortality in urban-based community surveys. The importance of STDs is suggested by a study of Harare, where these dis- eases are the most common presenting complaint among adults at primary care clinics, accounting for one-quarter of the total case load (Wellington, Ndowa, and Mbengeranwa,1997~. As discussed in Chapter 6, many factors are involved: numbers and types of sexual contacts, gender roles and relations, and poor knowledge about or access to contraceptives (Mamdani, Garner, et al., 1993; Agyei, Mukiza-Gapere, and Epema, 1994; Pick and Obermeyer, 1996; Wellington, Ndowa, and Mbengeranwa, 1997~. The role of STDs in facilitating HIV transmission further underscores their importance. Tuberculosis is among the leading causes of death to adults in developing countries, killing an estimated 3 million people in 1995 (Dolin et al., 1994; cited in Boerma, Nunn, and Whitworth, 1999~. The interactions between HIV and tuberculosis and the spread of multi-drug-resistant strains of tuberculosis have increased concerns about a global resurgence of the disease. Urban crowding increases the risk of contracting tuberculosis (van Rie, Beyers, Gie, Kunneke, Zi- etsman, and Donald, 1999), and this fact, together with the higher prevalence of HIV in cities, implies that tuberculosis will probably become increasingly preva- lent. High-density low-income urban communities may be particularly at risk. The socioenvironmental conditions of urban areas are also implicated in vector- borne diseases, such as malaria, filariasis, dengue, chagas disease, plague, and typhus (Knudsen and Slooff,1992~. Studies highlighting the potential inadequa- cies of health systems in preventing and treating these conditions (Molbak, Aaby, Ingholt, Hojlyng, Gottschau, Anderson, Brink, Gansted, Permin, and Vollmer, 1992; Atkinson and Cheyne, 1994; Byass, Adedeji, Mongdem, Zwandor, Brew- Graves, and Clements, 1995; Sodermann, Jakobsen, Molbak, and Aaby, 1997) also point to the role of local and national governments in the provision of envi- ronmental sanitation and health services. The Urban Penalty The prospect of- communicable diseases proliferating in countries with weak gov- ernmental and public health capacities has raised concerns about new forms of urban penalties. The phrase "urban penalty" arose from analyses of English mor- tality data from the industrial revolution of the nineteenth century, which revealed that urban mortality rates (particularly from tuberculosis) were much higher than rural rates (Kearns, 1988, 1993~. Similar urban disadvantages were evident elsewhere in Europe. In 1875, for example, the urban infant mortality rate was
MORTALITY AND MORBIDITY 271 240 per 1,000 in Prussia, as against 190 per 1,000 in rural Prussia (Vogele,2000~. To the extent that such differences were due to communicable disease, the spatial concentration of urban populations rendered them more vulnerable to infection. Special risks faced the inhabitants of port cities, who were repeatedly exposed to new or newly mutated pathogens carried by sailors, passengers, and vermin. The urban disadvantage persisted until urban public health measures, such as invest- ments in the provision of clean water and sanitation, were introduced on a wide scale. These investments, when accompanied by increases in literacy and sup- porting socioeconomic change, began to drive down urban infant mortality rates in Prussia, with clear evidence of decline emerging as of 1893. By 1905, the ur- ban infant mortality rate had reached parity with the rural rate at a level of 170 per 1,000. The eradication of urban penalties was seen throughout much of Europe at the turn of the century. The reduction in urban mortality rates owed as much to social and political forces as to technical and scientific factors. In Great Britain, for ex- ample, constitutional arrangements and political organizations that recognized the importance of the sanitary movement are thought to have been critical (Szreter, 1997~. As discussed in Chapter 2, a similar configuration of public health, politi- cal, and social dynamics proved to be influential in the United States (Preston and Haines, l991~. Likewise, in Japan an early appreciation of the public health bene- fits achievable through concerted public information campaigns stressing hygiene and sanitation yielded infant mortality rates rivaling those of England at the turn of the century. This was a notable achievement given average levels of income in Japan, which at the time were far below those of England. But from 1908 until after World War II, an urbanizing Japan failed to marshal the infrastructure in- vestments needed to consolidate and sustain its early gains, and began to fall well behind England in progress against infant mortality (Johansson and Mosk,1987~. As this brief review of the historical experience makes clear, the emergence of an urban advantage in mortality is the product of several complementary de- velopments: sufficient public-sector resources to undertake infrastructure invest- ments, the engaged attention of nongovernmental groups, concentrated political will, and both scientific and public health expertise. Higher levels of national and household income are also important. Higher national incomes supply govern- ments with the means to extend public health investments and training, and lay the foundation for the development of private health care markets. Higher house- hold incomes enable effective health care to be purchased. Household incomes are also associated with literacy and education, which can heighten attention to health, foster a sense of personal agency, and encourage beneficial social and political interaction (Montgomery, 2000~. In view of the many factors involved in reducing mortality and the natural disadvantages faced by spatially concentrated urban populations, it would not be surprising to find evidence of slippage in the urban health advantage, nor would it be surprising to see instances in which rural populations have regained their earlier
272 CITIES TRANSFORMED advantage. Given the importance of income levels noted above, such reversals might well occur in regions that are enduring severe economic distress. RECENT EVIDENCE ON CHILDREN'S HEALTH AND SURVIVAL What empirical evidence is there of an emerging urban penalty for children in developing countries? For an overview, we turn first to data from the DHS on children's height and weight, and then consider child survival. Children's Height and Weight The DHS collects information on the height and weight of young children and provides age- and sex-standardized measures of height for age and weight for height. These measures are expressed in terms of standard deviations from the medians of international reference populations. The units are such that a value of100 represents 1 standard deviation below the reference median. Most of the values seen in these data for the developing world are negative, reflecting the many health disadvantages that face developing-country children. A child whose height for age falls 2 or more standard deviations below the reference median is often described as "stunted," and one who is of similarly low weight for height is said to be "wasted." Low height for age is often taken to represent the cumulative effects of long-term deprivation, whereas low weight for height is interpreted as reflecting more recent, shorter-term deprivation. A1- though conventional, these interpretations are understood to be a bit simplistic. In any case, cross-sectional data do not allow the long- and short-term influences on health to be easily distinguished. The children whose height and weight are measured by the DHS range in age from newborns to 5-year-olds, depending on the survey. To reduce errors that may arise from measuring the youngest children and to maintain a degree of compa- rability across surveys, we have included in the present analysis only children in the age range of 3-36 months. In poor countries that is, in most of the countries surveyed there are often substantial age differences in the extent to which chil- dren fall short of the reference medians, and to see clearly how children's health varies in other dimensions, a method of further standardizing the data by age is needed. We do this by estimating ordinary least-squares regressions in which the dependent variable is either the child's height for age or weight for height, and the explanatory variable is the child's age. The results of these regressions are summarized in predicted values for children at 24 months of age. Table 7-2 summarizes estimates of children's height and weight for the rural and urban sectors as a whole.6 As can be seen, only one country in Southeast Asia 6As elsewhere in this report, the table entries are mean values of survey-specific estimates, and countries with more than one DHS survey are downweighted in proportion to the number of surveys. The "total" row of each table provides averages across all countries, but these averages are dominated
MORTALITY AND MORBIDITY TABLE 7-2 Children's Height for Age and Weight for Height at 24 Months, Rural and Urban Areas 273 Rural Urban Rural Urban Height Height Weight Weight DHS Surveys in Region Na for Age for Age for Height for Height North African 2 - 155.0 - 94.9 10.8 32.2 SubSaharan Africa 24 - 1 84.9 - 1 35.9 - 62.5 - 43.3 East, Southeast AsiaC 1 - 139.0 - 64.5 - 78.4 - 51.6 South, Central, West Asia 10 - 176.8 - 13 1.6 - 52.5 - 42.8 Latin America 10 - 144.7 - 92.4 - 10.6 2.9 TOTAL 47 - 172.4 - 122.5 - 46.6 - 30.3 a Number of countries with DHS data on children's height and weight. b Egypt and Morocco only. c Thailand only. (Thailand) gathered height and weight data, and only two did so in North Africa (Egypt and Morocco). The table documents clearly what might well have been ex- pected: in each geographic region, children in urban areas are significantly more healthy, judging by their height for age, than children in rural areas. The height differences are on the order of half a standard deviation in the usual case (in the units shown here, this is about 50 points in the average survey; see the "total" row of the table). A country-by-country inspection of urban/rural differences shows that in all but one case (Uzbekistan), children's height for age is greater in urban than in rural areas, and this difference is statistically significant in all countries. The urban/rural differences in weight for height are smaller, being somewhat less than 17 points in the average survey, although slightly larger differences ap- pear in North Africa, Southeast Asia, and sub-Saharan Africa. In only six sur- veys (those of Comoros Islands in 1996, Mali in 1995, Kazakhstan in 1995, the Kyrgyz Republic in 1997, Turkey in 1993, and Yemen in 1991) is rural weight for height greater than urban; in all but one survey (that of Kazakhstan in 1995), the urban/rural difference is statistically significant. Among urban children, do these measures of health vary according to city population size? Table 7-3 shows that differences in children's height by city size range from substantial (on the order of 0.9 standard deviation for surveys in Latin America) to trivial. There is a suggestion in these estimates that in some re- gions (notably in South, Central, and West Asia), children in the largest cities (those of over 5 million population) are disadvantaged relative to those living in intermediate-sized urban areas (with populations over 100,000 but less than 5 million). This is not a pattern common to all regions, however, as is evident in by the estimates from sub-Saharan Africa and Latin America, the regions that have fielded the greatest number of DHS surveys. When we discuss statistical significance, we refer to standard errors that are calculated with an allowance for unmeasured effects at the level of sampling clusters.
274 . ·_. V: ·_. v Cal Cal so ¢ ·_. so ' - o. ¢ so ' - o. ·_. m ¢ Em ~ 0 0 0 0 0 ok 0 0 0 0 0 0 0 Cal ~ O V: · ~ O oo kD ~ O oo kD kD ~ ~ ~ ~ ~ kD O O O O O O O O O O O O 00 0 ~ ~ ~ ~ ~ ~ ~ O V: ~ ~ ._ 0 = O ~ ~ S.., 0 ~ · - 0 ~ ~ =~ 00 0 .e ~ 1 . . . . . 1 ~ 1 ~ 1 ' 1 ' 1 ' ¢ ~ ~ ') 00 O ~ ~ O kD 1 ~ ~ 1 ~ 1 1 .= . . 00 1 O . . O ~ 1 . . 00 O . . kD ~ kD O Ct ·_. C~ ¢ Ct ~ ¢ .~ ~ Ct ¢ ~ ·o V ¢ O O V: V: V: o E~ ~ kD ~ ~ ~ oo kD . . . . . . . kD ~ kD oo O ~ 1 1 1 1 C~ . C~ O Ct C~ .~ C~ C=d O ~ 00 0 ~ ~ 00 ~ ~ O. 1 1 1 o C~ C~ C~ C~ Ct C~ ·_. .= .= C~ . .~ Ct V: .. O kD O ~ ~ O ~ O ~ . . . . . . . . ~ ~ 00 ~ ~ 00 ~ 00 o oo ~ ~ ~ ~ 1 ~ 1 1 1 1 ·o ~ ~ O s~ o . . . . ~ kD ~ ~ 1 1 1 ~ O . . . . . kD kD w) 1 1 1 ' 1 Ct ¢ .~ .o ~ ¢ ¢ O ~ O Ct ·_. C~ ¢ C~ . - d =.~ ~ v: ~ E~
MORTALITY AND MORBIDITY 275 the case of Latin America, where height for age is greater in the largest cities than elsewhere. Indeed, each region exhibits some irregularities in the relationship be- tween city size and children's height. Much the same story emerges when one considers children's weight for height, as shown in the second panel of Table 7-3. The differences by city size are small and irregular, although again it appears that children in the largest cities of Latin America enjoy better health than children in smaller urban areas. Differences by city size in weight for height do not appear to be closely associated, in general, with differences in height for age. The rightmost column of Table 7-3 presents the results of statistical signif- icance tests applied to the survey-specific data. Shown in this column are the numbers of surveys in which the city size variables make a statistically signifi- cant contribution to explaining health in relation to the total number of surveys in which such a test could be applied. The figures shown provide a sense of the strength of the association across surveys in each region. Of 66 surveys in which the test could be applied, statistically significant differences emerged in 39 surveys for children's height but in only 26 surveys in the case of weight for height. Poverty and children's health Using the relative poverty measure described in previous chapters, one can assess whether children's height and weight vary significantly with poverty status in ur- ban areas. Table 7-4 summarizes the results of a multivariate analysis in which the explanatory variables for the urban subsample include the child's age, the city population size, and the household's poverty status. The remarkable aspect of the table is the near uniformity of the results. The children of poor urban households are shorter and weigh less than the children of nonpoor urban households, but poor urban children are taller and heavier than their rural counterparts. The intraurban differences are usually statistically significant with regard to height for age (being so in 46 of 67 surveys), but are less often significant where weight for height is concerned (meeting the criterion in only 23 of 67 surveys). When we compare the height of poor urban children with that of rural children, we find that in almost all surveys (60 of 67), the former are significantly taller for their age than the latter. In only five surveys those for the Comoros (1996), Madagascar (1997), Bangladesh (1996), India (1992), and Guatemala (1987) is there evidence of poor urban children being significantly shorter. Apart from the Comoros estimate, which puts poor urban children about 0.3 standard deviations further below the median than rural children, the differences in these surveys are quite small in absolute terms, being less than 0.1 standard deviation. With regard to height for age, almost all surveys suggest that poor urban children fare at least as well as rural children, and generally fare better. Comparisons of weight for height are less definitive. In 44 of 67 surveys, ur- ban poor children are estimated to be significantly heavier given their height than rural children, but in 16 surveys, urban poor children weigh significantly less.
276 o cat Cal Cal V: so o Cal Cal so ¢ so ' - o. ·_. Cal Cal Cal · _. v 1 m ¢ Em .= .= V: Cal V: o . Ad O ~ O O 0 O O ¢ ~ Cal .o O O v ~ .= 0 00 ~ ED ED ~ O ED oo ~ oo ~ ED ~ ED ¢ 00 so 1 t004 · ~ ED oo ~ ~ oo ~ ~ ED . . . . 00 0 0 00 1 ~ 1 1 1 ') ~ kD ~ kD ~ . . . ~D O O O ~ O kD O . . . ~ oo ~ 1 Ct .o ¢ .~ Ct t~ C~ .o ~ ¢ ¢ ~ Ct C~ s~ O ~ O ~ oo ~ . . . ~ O ~ 8 ~ 1 o Ct ~ s~ . ~ o ~ C~ .= .= C~ _4 C~ V: o ·~ o ~ .= C~ ~ s~ o o ~ ~, ~ .= s~ ~ o >\ ~o ,~ . . o. ~ ~ ~ o oo ~ ~D ~ ~ ~ ~ oo ~ oo ~ ~ ~ kD ~ o ~ ~ ~ ~ . . . . . . ~ O oo kD O ~ 1 1 1 1 ~ ~ ~ ~ ~D . . . . . ~ ~ kD ~ ~ ~ ~ kD kD 1 1 ' ~ ~ , 1 1 1 oo ~ ~ ~ o ~ . . . kD cd .~ ¢ ~ ~¢ E~ ~ oo ~ o ~ oo ~ ~ ~ ~ . . . . . o ~ oo ~ ~ ~ kD ~ ~ 1 1 1 1 cd .o ¢ .~ cd t~ c~ .o ~ ¢ ¢ c~ ~ s~ ~ ~ o ~ o ~D 1 oo 1 cd .~ ¢ ~ ~ ~¢ ·~ ~ ~ E~ s~ t~o. .o s~ o .o o · - ~ .o c~ ~ ~ ~o c~ ~ o ~ s~ ~ ~ s~ c~ ~ o ~ c~ ~ . c~ ~ s~ ~ o ~ o · - · - ~ c~ s~ ~ J _4 O O O ~ p.
MORTALITY AND MORBIDITY 277 Differences favoring rural children emerge in surveys of the Comoros (1996), Kazakhstan (1995), the Republic of Kyrgyzstan (1997), Mali (1995), and Yemen (1991), each of these being on the order of 0.2-0.4 standard deviations. Where this measure of children's health is concerned, there is substantial variation in the health advantage of poor urban children in relation to their rural counterparts. One finds an urban advantage more often than not, but in a substantial minority of surveys, rural children fare no worse than poor urban children. Changes over time in children's health In a subsample of 18 countries, one can examine changes over time in children's health from one survey to the next.7 Trends in children's height differ markedly across the regions. In North Africa (only Egypt and Morocco) and Latin America, the time trends in height are positive in both rural and urban settings, whereas in sub-Saharan Africa, either no change or a decline in children's height for age is typical. Contrary to some hypotheses, our analysis suggests that the urban advantage in health has persisted over the period covered by the DHS surveys, at least with regard to children's height for age. The changes over time in rural areas are of roughly the same magnitude and in the same direction as the urban changes. The rural change in height for age is about 70 percent of the urban change, on average. Summary Our comparisons of children's height and weight between urban and rural areas show that on average, urban children are indeed advantaged. The differences in these measures of health are both statistically significant and substantively im- portant. Examining countries with two or more DHS surveys, we could find no evidence of erosion in the urban advantage in these measures of child health. That an urban advantage is found is not especially surprising, although it would have been thought an anomaly a century ago. The more revealing comparisons involve urban children in poor households, those in nonpoor households, and rural children. There is strong evidence that the children of the urban poor are disadvantaged relative to other urban children. The evidence is equally strong in showing that, by the measure of height for age, the urban poor enjoy better health than do rural children, at least on average. Only a handful of surveys suggests that poor urban children fare worse than rural children in height for age, but a greater number of surveys indicate that in terms of weight for height, poor urban children may be less healthy than rural children. 7 The countries are Bolivia, Brazil, Cameroon, Colombia, the Dominican Republic, Egypt, Ghana, Guatemala, Kenya, Madagascar, Mali, Morocco, Niger, Peru, Senegal, Tanzania, Zambia, and Zimbabwe.
278 CITIES TRANSFORMED Chapter 5 documents sharp differences between urban poor and nonpoor households in access to public services (water supply, sanitation, electricity) that fall into a similar pattern, with the access of the urban poor to services being worse than that of the urban nonpoor but better than that of rural households. It is likely that poverty-related differences in children's health are due, at least in part, to such differences in access to services.8 If poor households have worse access to sanita- tion and clean water, for example, children in those households may be at greater risk of exposure to communicable diseases, in particular diarrhea! diseases. Infant and Child Survival In their interviews with mothers, DIES fieldworkers collect information on the dates of birth of all children who were born alive and ascertain the child's length of life in cases in which a child died. The panel estimated the probabilities of infant and child death using Weibull and Kaplan-Meier models of mortality risks, allowing the correlation in risk across children in the same survey cluster to influ- ence the standard errors of the estimates. For univariate, descriptive analyses, we used the Kaplan-Meier estimator; when covariates were included, we restricted the analyses to infant mortality and used a Weibull estimator. Some comment is in order about the way in which we handled migration and exposure to risk. In rural areas, all children born in the 5 years preceding the survey date were considered in the analysis. In urban areas, however, the question arose of how to deal with families that migrated to their current residence during the 5-year period. A child of a migrant family (1) could have been born in the family's previous location and died there, or (2) could have survived through the move and then faced the risk environment of the family's current residence, or (3) could have been born in the family's current place of residence and faced those risks from the beginning of its life. For migrant families, one clearly must differentiate these segments of a child's history so as not to confuse the risks of the current with those of the previous residence. A first difficulty is that there is very little information on the nature of the previous residence; generally, no more is known than that the residence was rural or urban. A second difficulty is that the date of the move is imprecisely defined. In the great majority of surveys, this date is known only in terms of years of duration in the current residence. For 18 countries, however, there is a monthly migration calendar in which the most recent move can be pinpointed to the month, and we used these calendars when possible. Faced with such complications, we decided to focus only on the migrant child's exposure to risk in the current urban location. Unless a monthly migration calendar happened to be available, the children of families who moved in the year leading ~Haddad, Ruel, and Garrett (1999) used DHS data from 11 countries (5 Latin American, 4 African, and 2 Asian) to demonstrate that the ratio of stunting prevalence between poorer and wealthier quintiles was greater within urban than within rural areas, and that intraurban differences (among socioeconomic groups) were greater than the urban/rural differentials.
MORTALITY AND MORBIDITY 279 up to the survey could not be included in our analysis since there was no way of knowing when in that year the move took place, and thus no way of determining precisely when the child was first exposed to the risks of the current urban envi- ronment. The conservative decision appeared to be to drop the case entirely. The results of the survival regressions are summarized in predicted values for infant mortality (denoted by 1q0) and, where possible, for child mortality (sqo) as well. Chapter 4 summarizes infant and child mortality in urban and rural areas (see Table 4-6~. The results given there provide evidence of a substantial ur- ban advantage, at least on the average. Here we explore the differentials by sub- group, focusing first on differences in urban children's mortality according to city population size. Table 7-5 shows the extent to which infant and child mortality rates vary by city size. At the aggregate level, there are sizable differences in both infant and child mortality by city size. Within particular regions, however, the results are not especially robust, and they are often not statistically significant, attaining significance in only 16 of 86 surveys in the case of infant mortality. The effects of poverty are more substantial. The estimates for infant mortal- ity shown in Table 7-6 suggest that the urban poor often, but not invariably, face mortality risks that are significantly greater than those faced by the urban nonpoor. The point estimates (converted to predicted values) shown in this table place urban poor children at a point midway on the risk spectrum, falling between the urban nonpoor and rural children. In comparing urban poor and rural children, we find that in some 57 of 87 surveys, the mortality risks facing the urban children are sig- nificantly lower. In 25 surveys, however, the urban poor face significantly higher risks.9 Changes in child mortality can be examined for 34 pairs of DHS surveys.~° On average, child mortality fell by roughly the same amount in both urban and rural areas. The average change in the estimated sqo between surveys is9.3 points per 1,000 in rural areas and8.1 points per 1,000 in urban areas. (The gap between surveys averages 5.2 years.) In 20 of the 34 cases examined, mortality fell in both rural and urban areas, with the declines being about equal in size. In 6 countries all in sub-Saharan Africa child mortality rose between the surveys, and in these cases the increases in mortality were slightly larger for urban than for rural children. But the sub-Saharan story is complicated: in the remaining 6 cases from this region, mortality fell in both rural and urban areas, with the urban declines being slightly larger on average. In summary, although the details of the situation need further clarification and there are a number of counterexamples to consider, especially in sub-Saharan Africa, the dominant pattern evident in the DHS data is that of decline in both urban and rural mortality, with the urban decline being greater in relative terms. Overall, this evidence does not support the hypothesis of an eroding urban 9These findings are derived from an application of the delta method to predicted values from the mortality models. lessee Chapter 6, footnote is, for the list of countries.
280 ·_. V: o o Cal Cal ¢ ·_. · _. sit o v m .~ V: o o S.., 0 0 ~ So, ~ 0 8 of So 0 ~ ~ 0 Cal _ 0 00 00 ~ ID ~ ~ ~ ~ 0 0 0 0 0 0 . . . . . . O 0 0 0 0 0 ~ ~ 00 ~ ~ ~ 0 0 0 0 0 0 . . . . . . O 0 0 0 0 0 00 ~ ~ 00 00 ~ ~ ~ ~ 0 ~ ~ ~ ID ~ ~ ~ ~ ~ ~ ~ ID ~ ID 0 0 0 0 0 0 0 0 0 0 0 0 . . . . . . . . . . . . O 0 0 0 0 0 0 0 0 0 0 0 o of ~0 o 00 ID ~ ~ . O. O. O 0. O. O. ~ 0 0 0 0 0 0 Sin o v kD ~ ~ kD ~ ~D o o o o o o . . . . . . o o o o o o O ~ oo kD kD ~ ~ o. ~ o o. o. o. o o o o o o ~ ~ ~ ~ oo ~ kD kD oo o ~ o o o o . . . . . . o o o o o o oo ~ kD ~ kD kD ~ ~ ~ kD ~ o o o o o o o ~ o o o o . . . . . . . . . . . . o o o o o o o o o o o o cd cd .~ · c~ c~ ¢ ¢ .o ¢ ,~ ~ cd ¢ '`a ~ cd ~¢ ~ ~ v ¢ ~ ~¢ ~ ~ v ¢ ~ =¢ ~ ~ ~ ~ ~ =¢ o ~ o o ~ ~ o ~ o o ~ ~
281 o ma o ~ o ~ sit o o Cal ~ o V: ·~ ~ .= ID ~ ~ ~ ~ Do o o o o o o o o ~ ~ Do ~ ~~ o ~ ~ ~ ~ m ~ o ·.5 o o o Ad o o Cal Ct .= ID m O ~ ~ it ~ hi.= ID ~ ~ ~ ~ O O O O O ~ 00 ID ~ ~ ~ ~ O O O O O oo o oo o .~ s~ t~o. c~ =~ - O kD oo ~ kD kD ~ o o o o o o o ~ =m ~ ~ ~ ~ ~ kD o o o o o o o o o o o o ~ ~ ~ ~ ~ kD oo O ~ ~ kD oo o ~ o o o o . . . . . . o o o o o o cd · c~ ¢ c~ ~ .o v ¢ ~ ·~ o ~ v: cd · s~ ¢ cd ~ ¢ ~ ~ v: o ~ .~ c~ ¢ c~ cd o v: E~ . ~ c~ Co ~o c~ .~ c~ ..
282 CITIES TRANSFORMED advantage. But there may well be urban subpopulations such as the urban poor whose advantage is eroding. Moreover, the DHS program has continued to field surveys in Eastern and Southern Africa, where the effects of the HIV/AIDS epidemic are being seen in infant and child mortality. As the epidemic proceeds, our conclusions about trends will need to be reassessed. In analyzing data from a number of DHS surveys, Brockerhoff and Brennan (1998) did find evidence of erosion in the urban advantage for large cities (those of over 1 million population). Relying on retrospective reports by mothers on the survival of their children, Brockerhoff and Brennan determined that mortality rates in the large cities were not declining as rapidly as those in smaller cities (of 50,000 to 1 million population) or rural areas. In light of these contradictory findings, as well as other case studies suggest- ing deteriorating health conditions among slum dwellers, we cannot draw strong conclusions about trends in the urban health advantage. It may be possible to in- corporate data from the late 1970s gathered in the WFS to bring the trends by city size into clearer focus. The city size categories used in the WFS resemble those employed in the DHS, but considerable effort would be required to put the two in a strictly comparable form. Although the panel lacked the resources to pursue this issue further, it should be given high priority in future research. Epidemics and Economic Crises As we have seen, data for sub-Saharan Africa clearly indicate deteriorating health conditions in a number of cases. As noted in Chapter 6 and earlier in this chapter, there is substantial evidence of higher HIV prevalence in large cities than in small cities and rural areas in Africa, although the epidemic is spreading outside the large cities (Boerma, Nunn, and Whitworth, 1999~. The implications for child mortality are spelled out by Timaeus (1998: S25~: The main determinant of the impact of HIV on child mortality is the scale of the epidemic on adults . . . child mortality could be expected to rise substantially in urban areas; towns with seroprevalence of 30 percent or more would experience a rise of one-third (in eastern and central Africa) and three-quarters (in Southern Africa). In these regions of Africa, where the epidemic is currently at its worst, it is difficult to imagine that an urban survival advantage can be maintained. Hanmer and White (1999) found a substantial rise in under-5 mortality in Zambia, one of the countries hardest hit by HIV/AIDS. The rise in mortality has been greater in cities than in rural areas and has led to a narrowing of the ur- ban/rural differential. Child-rearing practices, immunization, and the use of oral rehydration therapy have all been improving, and if other things were held equal, iiTo derive estimates specific to cities from the DHS retrospective birth histories, Brockerhoff and Brennan were forced to make strong assumptions about women's migration histories, which are not generally available in the same retrospective form. It is unclear whether the results thus derived are sensitive to assumptions about migration.
MORTALITY AND MORBIDITY 283 mortality should have declined. The HIV/AIDS epidemic may have produced a rise in infant mortality of about 5 points per 1,000 in rural areas and 13 per 1,000 in urban areas (it accounted for a quarter of urban under-5 mortality in 1996~. But precise estimates are difficult to make because, in addition to the HIV/AIDS epidemic, Zambia has experienced a marked economic decline, reduced health spending, some deterioration in the quality of health services, and an increase in child malnutrition. Changes in health associated with macroeconomic crises have been closely studied in the case of Indonesia, which suffered a sharp economic contraction in 1997-1998. Frankenberg, Beegle, Thomas, and Suriastini (1999) were able to trace the health impacts of this crisis. (Other aspects of the Indonesian crisis are examined in Chapter 8.) Use of health services declined overall during the crisis, and the composition of use shifted from public to private and traditional practi- tioners. There was a significant reduction in the proportion of children receiving vitamin A, which protects against various illnesses. There was also a decline in the quality of public health services relative to private providers, but both raised their fees. Interestingly, little change took place in contraceptive use; according to Frankenberg, Thomas, and Beegle (1999a: 1), this was because "contraception is a more appealing option than the risk of having an additional child in the current economic environment." The nutritional status of adults substantially worsened, but as far as short-term health measures are concerned, children appear to have been largely shielded from the worst effects of the crisis. Perhaps the most dramatic effects of economic crisis have been seen in the former republics of the Soviet Union, which have experienced sharp increases in adult mortality. Box 7.2 describes the effects of the crisis in Kazakhstan and the subsequent recovery in the country's major cities. BOX 7.2 Rising and Recovering Mortality During Transition in Cities of Kazakhstan The collapse of the socialist world generated a demographic as well as an economic crisis, the most prominent feature of which was sharply rising adult mortality. High-quality data clearly document these disruptions in Kazakhstan (Becker and Urzhumova, 2001~. From the late Soviet period through the initial postindependence crisis, life expectancy in Kazakhstan as a whole declined by 6.5 years among men and 5.6 years among women. As in most of the other former Soviet republics, rising mortality in Kazakhstan can be at- tributed to diseases of the circulatory system and to a pronounced increase in accidents, injuries, and traumas. As elsewhere, rising mortality was accompanied by sharply declin- ing incidence of fertility and marriage (Urzhumova and Becker, 1999~. Life expectancy began to recover in 1996. The recovery has been strongest among women and in Almaty (the commercial capital) and Astana (the political capital), where women have regained the survival rates characteristic of the late Soviet era. Indeed, for Almaty women under 40 and above 65 years of age, 1999 mortality rates were already below those recorded in 1991. In much of the country, however, recovery has not yet been firmly established, and whether it is imminent is unclear.
284 CITIES TRANSFORMED A PENALTY FOR THE URBAN POOR? Although we did not find compelling evidence of a declining urban advantage overall, there remains the possibility that a health penalty is in force for some urban subgroups. Do the spatially concentrated urban poor face such a penalty? The last major urban health review for developing countries (Bradley, Stephens, Harpham, and Cairncross, 1992) consolidated studies highlighting intraurban in- equities in morbidity and mortality. Taken as a group, these studies raise the pos- sibility that the urban poor may suffer a double burden of both communicable and noncommunicable disease. Bradley, Stephens, Harpham, and Cairncross (1992: viii) advance the view that "urban poor households sometimes have worse nutri- tional status than rural households, contributing to ill-health related to nutrition." The panel's review of DHS data uncovered no systematic tendency for poor urban children to fare worse than rural children, but identified a number of exceptions to the general rule. The urban poor may face a double burden not because of poverty alone, but rather due to its depth and spatial concentration. These possibilities cannot be pursued in detail with DHS data, but they can be explored through case studies of selected cities and low-income neighborhoods within these cities. Stephens, Timaeus, Akerman, Avlve, Maia, Campanario, Doe, Lush, Tetteh, and Harpham (1994a) and Stephens (1996) analyze data on socioeconomic status, indicators of environmental quality, and mortality from Accra and Sao Paulo, uncovering evidence of enormous disparities in the health status of urban populations living in the most and least deprived areas of these cities. This was the first empirical study to demonstrate that the most deprived neighborhoods suffer relatively high mortality rates due not only to diseases of the respiratory system and infectious and parasitic diseases, but also to diseases of the circulatory system. In Sao Paulo, deaths from external causes (homicides and traffic acci- dents) are three times higher in the most deprived area of the city. The figures for homicide are particularly striking: there is an 11-fold difference between the least and most disadvantaged subdistricts. The authors estimate that in Accra, up to two-thirds of adult deaths in the poorest three zones of the city could have been avoided had these areas faced the lower mortality risks of the most advantaged ar- eas. Similarly, if Sao Paulo had had a uniform socioeconomic and environmental profile rather than a profile of substantially unequal risks, more than half of the deaths in its most disadvantaged neighborhoods might have been avoided. As earlier writers on urban health have noted (Harpham, Lusty, and Vaughan, 1988; Harpham and Stephens, 1991), the effects of the clustering of residential poverty on health can be devastating. The negative externalities that accompany life in slum and squatter settlements can magnify the effects of individual poverty and inadequate public health services, such as a lack of clean water and sanitation. High mortality rates in urban slums have long been documented. For example, i2The most deprived areas in both of these cities account for a significant proportion of the popula- tion (44 percent in Sao Paulo and 67 percent in Accra).
MORTALITY AND MORBIDITY TABLE 7-7 Intraurban Differences in Infant Mortality Rates, Bangladesh, 1991 Gender National Rural Urban Urban Slum Total 90 93 68 134 Male 98 97 70 123 Female 91 89 65 146 NOTE: National rates are from Bangladesh Demographic Statistics, Bangladesh Bureau of Statistics, Government of Bangladesh; rural and urban rates are extrapolated from 1990 data in the Statistical Yearbook of Bangladesh. Urban slum rates are from the Urban Surveillance System, Urban Health Extension Project, International Centre for Diarrhoeal Dis- ease Research, Bangladesh. SOURCES: Harpham and Tanner (1995: 36~. 285 infant mortality rates in the slums of Dhaka, Bangladesh, were found in 1991 to be significantly worse compared not only with the rates among other urban residents, but also with rural rates (see Table 7-7~. Several recent empirical studies have highlighted urban neighborhood and contextual effects on health outcomes and mortality, indicating that the urban poor, especially those living in majority-poor communities, suffer from urban penalties based on their geographic location and its interrelationship with poverty, segregation, and other forms of deprivation. Infant mortality rates in the municipalities that make up Greater Sao Paulo varied in 1992 from 18 per 1,000 (in Sao Caetano do Sol) to 60 per 1,000 in Biritiba-Mirim. In 1980, the best-performing municipality had a rate of 29.3, while the worst had a rate of 152 (Industria de Embalagens Plasticas Ltda,1994~. All of these municipalities have seen substantial declines in their infant mortality rates since 1980. At that time, the average infant mortality rate for Greater Sao Paulo was 51.8 per 1,000, but by 1992 the rate had fallen by half, to 25.5 per 1,000. Moreover, the differences across municipalities have narrowed. Even so, if the subdistricts of Sao Paulo municipality are classified into four categories ac- cording to social, economic, and environmental criteria, the mortality rates from infectious diseases for children under age 4 living in the worst of the four cate- gories were more than four times higher than those for children living in the best of the four categories; there was also a nearly fourfold difference in deaths from res- piratory diseases (Stephens, Timaeus, Akerman, Avlve, Maia, Campanario, Doe, Lush, Tetteh, and Harpham, 1994b). Similarly, in Buenos Aires in 1990, infant mortality rates within local government areas (municipalities and the central fed- eral district) varied from a low of 15 in the federal district to a high of 30 in San Fernando and Tigre (Arrossi,1996~. Findings of a recent study in Rio de Janeiro indicate that intraurban variations in the postneonatal mortality rate are associated with the geographic pattern of poverty that is, neighborhood poverty clustering even after adjusting for the neighborhood poverty rate. These poverty clusters are associated not only with
286 CITIES TRANSFORMED economic poverty, but also with poor access to services, a poor environment, and deprivation and segregation, which affect health and mortality through cultural and social factors (Szwarcwald, de Andrade, and Bastos, 2002~. Finally, results of a recent comprehensive survey of the slums of Nairobi reveal that slum residents who lack basic services, adequate housing, and health services and who live amid others similarly disadvantaged, are worse off in almost ev- ery health dimension than those who reside elsewhere in Nairobi, rural Kenyans, and Kenyans overall (African Population and Health Research Center, 2002~. In terms of infant and child morbidity and mortality, children living in the slums have lower survival odds, significantly lower immunization levels (with the exception of tetanus), and a higher incidence of infectious disease. During the 1990s, in- fant mortality rates declined in urban Kenya and in Nairobi overall, increased in rural areas, but appear to have increased even more sharply in the Nairobi slums. Figure 7-2 compares the infant mortality rate (0-1 year) and the under-5 mortality rate for rural Kenya, urban Kenya, Nairobi, and the Nairobi slums. The mortality rates are highest in the slums, where they are more than twice the rate for Nairobi as a whole (African Population and Health Research Center, 2002~. The extent of poverty can be gauged by the risk factors that are present for different diseases. Table 7-8 lists some of the risk factors for child diarrhea in Accra. Among Accra households facing fewer than three of these risk factors, only 2 percent reported diarrhea; among those facing three or four risk factors, 150 - ~_ lo lo lo 100 . _ A) us .~ o 50 o I l Rural ~1 Urban _ All Nairobi _ Nairobi Slums Infant Child Under-five Mortality Mortality Mortality (1-to) (4q~) (5-to) FIGURE 7-2 Levels of infant and child mortality in Nairobi slums as compared with other parts of Kenya. SOURCE: African Population and Health Research Center (2002~.
MORTALITY AND MORBIDITY TABLE 7-8 Approximate Relative Risk of Diarrhea Among Children Under Age 6 in Accra Factor Odds Ratio 95% ConfidenceInterval Use pot for storing water 4.3 1.7-11.1 Water interruptions are common 3.1 1.4-6.6 Share toilet with more than 5 2.7 1.2-5.8 households Purchase vendor-prepared food 2.6 1.1-6.2 Open water storage container 2.2 1.1~.3 Outdoor defecation in locality 2.1 1.1-3.9 Many flies in food area 2.1 1.1-3.8 Do not always wash hands before 2.0 1.1-3.8 preparing food NOTE: Number of observations is 500. SOURCES: McGranahan, Songsore, andKjellen(1999), Songsore andMc- Granahan (1998). 287 the percentage rose to 14; and among those facing more than four risk factors, the percentage rose to 39. The deeper it is in poverty, the more risk factors an Accra household is likely to face. Children's health (in terms of diarrhea and acute respiratory infections) is associated with environmental correlates of poverty. With regard to migrant health, Brockerhoff (1994, 1995a) analyzed DHS data from 15 and 17 countries, respectively, to explore inter- and intraurban health differentials. Regional variations in patterns of mortality, health, and social indi- cators rendered generalization difficult. Nevertheless, socioeconomic status, short birth intervals, young maternal age, parental education, and (after controlling for these variables) in-migration of mothers from rural areas were found to be pow- erful predictors of infant survival in cities with populations of over 1 million. (Recall that most recent migrants to urban areas come not from rural areas but from other towns and cities; see Chapter 4.) The factors that might explain excess mortality among the children of migrants could not be assessed with the available data, but may include the threat of new infectious disease agents, temporary resi- dence on arrival in particularly poor housing environments, changes in care-giving practices, termination of breastfeeding, a decrease in income, and incomplete im- munization due to a lack of familiarity with services. In summary, the research on intraurban inequalities highlights the diverse ex- perience of different countries and communities, but in general provides evidence that the double burden of disease is not equally shared by all groups within cities. Rather, the urban poor die disproportionately of both infectious and chronic, degenerative diseases. Finally, it is understandable that much research on urban health is focused on absolute poverty rather than on measures of relative poverty. There are enor- mous intraurban differences in housing, income, sanitation, drainage, piped water,
288 CITIES TRANSFORMED environment risk, and access to services, whether expressed in absolute or relative terms (Mitlin, Satterthwaite, and Stephens, 1996; Mutatkar, 1995; Wang'ombe, 1995; Atkinson, Songsore, and Werna, 1996; Todd, 1996; Harpham,1997~. How might an emphasis on the relative aspects of poverty bring a fresh perspective to the issues? Because city residents are spatially proximate, the poor constantly confront visible reminders of their relative deprivation. Social comparisons are all but forced upon them. At times, presumably, the higher living standards enjoyed by other urban groups may exemplify the rewards to upward mobility. But for the discouraged and frustrated poor who have lost confidence in their prospects, these visible symbols of inequity are no doubt highly stressful. The very diversity of urban life may then increase anxiety and even threaten psychosocial well-being. Box 7.3 describes the work of Wilkinson (1996), who identifies the implications of urban inequalities for health and social cohesion. BOX 7.3 Wilkinson on Inequality In Wilkinson's view, social comparisons knowledge of "how the other half lives" can have a powerful effect on individual psychosocial well-being. With economic growth and the beginnings of the epidemiological transition, some groups secure access to higher living standards, while others continue to face severe deprivation. As inequities become more visible, the relative aspects of poverty affect subjective experience, and the consequent tensions and anxieties can have further damaging effects on health. The impact of relative poverty is powerfully described (Wilkinson, 1996: 215, cited in Blue, 1999:214: From the point of view of the experience of people involved, if health is being damaged as a result of psychosocial processes, this matters much more than it would if the damage resulted from the immediate physical ef- fects of damp housing and poor quality diets.... To feel depressed, cheated, bitter, desperate, vulnerable, frightened, angry, worried about debts or job and housing insecurity; to feel devalued, useless, helpless, uncared for, hope- less, isolated, anxious and a failure: these feelings can dominate people's whole experience of life, coloring their experience of everything else. It is the chronic stress arising from feelings like these which does the damage. It is the social feelings which matter, not exposure to a supposedly toxic ma- terial environment. The material environment is merely the indelible mark and constant reminder of the oppressive fact of one's failure, of the atrophy of any sense of having a place in a community, and of one's social exclusion and devaluation as a human being. As Wilkinson sees it, relative inequalities in income can lead to a breakdown in social cohesion, creating chronic psychosocial stress that adversely affects physical and mental health. The deterioration of community life and subsequent rise in violence and crime have a detrimental impact on all members of society, not only the poor. It is the proximity of the urban poor to what are frequently some of the richest people in the world that has been linked to urban tensions and social unrest (Massey, 1996~.
MORTALITY AND MORBIDITY 289 The terms in which poverty is described have political implications. As Mitlin, Satterthwaite, and Stephens (1996) note, when poverty is described in absolute rather than relative terms, this can suggest that the responsibility for alleviating poverty rests with poor individuals and households themselves. Somehow the societal mechanisms that bring about and maintain deprivation tend to be over- looked. A focus on relative poverty and inequality, by contrast, may force a recog- nition of inequities and thereby encourage governments and the poor to grapple with the institutional and societal factors that perpetuate poverty. HEALTH SERVICE PROVISION AND TREATMENT SEEKING A defining feature of urban health systems is the diversity of providers who offer a multiplicity of services. These providers include government services run by district councils, municipal councils, or state or central ministries; private (for- profit) hospitals, laboratories, and practitioners (offering modern or traditional services); and a variety of nongovernmental providers, including missions and charities (Lorenz and Garner, 1995~. Medications are also available in shops, pharmacies, markets, and various clandestine outlets. This diversity of providers in urban settings is an expression of the demand for private (for-profit) services; the very different ideas, needs, and purchasing power of diverse urban popula- tions; the availability of social or private health insurance coverage for some formal-sector employees; and a continuing urban bias in the provision of gov- ernment health services (Hanson and Berman, 1998~. Official statistics often fail to convey this diversity; as Hanson and Berman (1998) note, there are almost no data on pharmacies, nurses, traditional healers, and kiosks. The conceptual distinctions we have mentioned between the public and private sectors, for-profit and not-for-profit providers, the formal and informal sectors, and modern and traditional providers can be difficult to detect (see Giusti, Criel, and Bethune, 1997; Londono and Frenk, 1997; Ferrinho, van Lerberghe, and Gomes, l999~. The urban health sector presents many arrangements that defy easy categorization dual public and private practices by government-employed physicians, the introduction of private wards in public hospitals, the self-referral of patients between healers and Western doctors and between public and private facilities, and the introduction of user fees into government services. The urban system is less a system than a patchwork: there is often little formal interaction or cooperation among the different types of providers (Ogunbekun, Ogunbekun, and ~ ~7 Orobaton, l999~. Urban Treatment Seeking The pluralism of urban health care provision is both an outcome and a determinant of treatment-seeking patterns. In urban as in rural areas, household responses to illness are highly complex, influenced by perceptions of illness severity, views of
290 CITIES TRANSFORMED causation and appropriate therapy, and access to sources of treatment. Recogni- tion of symptoms, their definition in terms of recognized illnesses, and decisions about treatment are all influenced by a plethora of macro-, mesa-, and micro-level factors (Igun, 1979; Kleinman, 1980; Phillips, 1990; Berman, Kendall, and Bhat- tacharya, 1994; Andersen, 1995~. Responses can therefore very for differentill- nesses or syndromes, and in many cases more than one therapy source is contacted over the course of an illness. Against this general background of diversity, empir- ical studies have identified a number of key features of urban treatment-seeking patterns. Self-medication When they fall ill, many urban residents turn to the drugs available in private for-profit shops and pharmacies (see also Chapter 6~. For example, community studies show that at least 40 percent of those seeking treatment for fever pur- chase drugs from these sources, and for most this is their first response to illness (Glik, Ward, Gordon, and Haba, 1989; Carme, Koulengana, Nzambe, and Bodan, 1992; Mwenesi, 1993; Kilian, 1995; Chiguzo, 1999; Molyneux, Mung'Ala-Odera, Harpham, and Snow, 1999~. From the viewpoint of many urban residents, such self-treatment is a sensible first step because formal clinics and hospitals are less convenient and more expensive in time and money terms. Small shops can offer drugs on credit or sell them in small, affordable doses (Molyneux, Mung'Ala- Odera, Harpham, and Snow, l999~. Traditional healers Herbalists, diviners, midwives, fertility specialists, and spiritualists are an im- portant source of health care in Africa and in much of Asia and Latin America. These traditional healers are thought to be the main source of health care for up to 80 percent of rural residents in developing countries, and Good (1987) argues that they are retaining and even expanding their influence in many cities. Heal- ers have their specialties: empirical studies suggest that urban healers are more likely to be consulted for convulsions, nonspecific pains, and psychological prob- lems than for other illnesses and symptoms (Winston and Patel, 1995; Carpentier, Prazuck, Vincent-Ballereau, Ouedraogo, and Lafaix,1995~. Many urban residents are eclectic, using healers even while they are attending modern health facilities, and healers often assume an important role when modern services fail to effect a cure or are perceived to fail (Molyneux, Mung'Ala-Odera, Harpham, and Snow, 1999~. Private providers Private facilities are an important urban treatment source for STDs, malaria, tuber- culosis, and diarrhea (Brugha, Chandramohan, and Zwi, 1999; Brugha and Zwi,
MORTALITY AND MORBIDITY 291 1999; Hotchkiss and Gordillo, 1999; Molyneux, Mung'Ala-Odera, Harpham, and Snow, 1999; Ngalande-Banda and Walt, 1995~. The relatively high costs of pri- vate for-profit services might appear to put them out of the reach of the poor, but a growing body of research shows that many low-income urban residents are heav- ily dependent on such services (Ogunbekun, Ogunbekun, and Orobaton, 1999; Ngalande-Banda and Walt, 1995; Yesudian, 1994; Molyneux, Mung'Ala-Odera, Harpham, and Snow, 1999; Thaver, Harpham, McPake, and Garner, 1998; Deve- lay, Sauerborn, and Diesfeld, 1996; Hotchkiss, 1998; Hanson and Berman, 1998~. The preference for private services despite their higher costs as compared with public services is attributable to the availability of staff and drugs, easier physical access, shorter waiting times, extended or more flexible working hours, better in- terpersonal communication between staff and patients, and the promise of greater confidentiality. Nevertheless, government hearth services remain an important source of health care. Wyss, Whiting, Kilima, McLarty, Mtasiwa, Tanner, and Lorenz (1996), for example, interviewed residents of Dar es Salaam about their use of services in the previous 2 weeks. They found that although poor and rich alike use private fa- cilities, the poor rely more often on government health services. In some settings, richer households are as likely as poorer households to depend on public facili- ties (Hotchkiss, 1998; Makinen, Waters, Rauch, Almagambetova, Bitran, Gilson, McIntyre, Pannarunothai, Prieto, Ubilla, and Ram, 2000~. Selected Issues in Health Service Provision and Use Aspects of urban health service use that have attracted research attention include the Malfunctioning of the referral system, the impact of user fees on rates and patterns of use, quality of care as a key influence on treatment seeking, and urban/rural linkages. These are addressed in turn below. The referral system In many cities, those seeking treatment often bypass clinics and other facilities at the lower tiers of the public health system and present themselves directly to the outpatient clinics of city hospitals (World Health Organization, 1 993; Holdsworth, Garner, and Harpham, 1993; Sanders, Kravitz, Lewin, and McKee, 1998; Akin and Hutchinson, 1999~. This behavior may be sensible from the viewpoint of those seeking treatment, given their perceptions of the low quality of services available at the lower tiers. From the perspective of the health system overall, however, such behavior contributes to overcrowding and to poor quality of care in hospitals, which arguably should be specializing in the severe health problems that require more sophisticated treatments. Basic primary care and first-contact services have been introduced in many cities, although they often cannot keep pace with urban growth. Secondary and
292 CITIES TRANSFORMED tertiary services (offering major inpatient and specialist care) are also generally located only in cities. The principal weakness identified in urban health systems is at the second tier of services, these being the units that should manage referrals, supervise primary care and first-contact services, and provide basic care for pa- tients with obstetric difficulties and trauma (Lorenz and Garner, 1995~. A World Health Organization strategy to strengthen this tier is to promote the development of urban intermediate-level health services or "reference centers," either by up- grading health centers or by giving referral hospitals authorization to provide dif- ferent levels of care in the same institution (Sanders, Kravitz, Lewin, and McKee, 1998). As shown by a study of patient flow in a national referral hospital in Lesotho (Holdsworth, Garner, and Harpham, 1993), however, clinicians may not be quite as overburdened by patients as they appear to be. Patient load as such may be less of a factor than poor management of the patient flow resulting in the over- crowding and long delays experienced in outpatient waiting rooms. Nor is it clear that establishing reference centers would necessarily ease such crowding. In Zam- bia, Atkinson, Ngwengwe, Macwan'gi, Ngulube, Harpham, and O'Connell (1999) found that accessible, inexpensive reference centers could well attract even more patients to the public health system, including many who would otherwise have used self-medication. User fees Proponents of health-sector reforms have often given top priority to financing, proposing that user fees be introduced or raised and urging consideration of pre- payment and insurance schemes. The impact of such policies on the poor has been a subject of intense debate, with empirical studies yielding conflicting findings (see, for example, McPake, 1993; Gilson and Mills, 1995; Gilson, Russell, and Buse, 1995; Gilson, 1997; Stierle, Kaddar, Tchicaya, and Schmidt-Ehry, 1999~. In some urban areas, the poor spend a higher proportion of total household funds on health than do the nonpoor. In 1992, for example, the poorest tenth of Mexican urban families spent 5.2 percent of their income on health care, as compared with 2.8 percent for the richest tenth of families (Londono and Frenk,1997~. Although many studies have found that user fees reduce utilization of care, particularly for low-income households, some studies suggest that fees have little effect on de- mand and can even increase demand if higher fees are thought to be linked to better-quality cared Quality of care and the perception of services may be as i3Even where utilization has not decreased, it is important to distinguish between willingness to pay and ability to pay. Where willingness to pay among low-income groups exceeds ability to pay, the strategies adopted by the poor to pay for health services (such as claims on kin, loans, sales of assets, and shifting of resources from other critical needs) can have broader consequences for livelihoods and health (Russell, 1996).
MORTALITY AND MORBIDITY 293 important as price and income in the use of health services (McPake, 1993; Rus- sell, 1996; Okello, Lubanga, Guwatudde, and Sebina-Zziwa,1998~. The introduction of user fees can price some low-income groups out of the market unless more effective exemption or third-party payment mechanisms are devised (Stierle, Kaddar, Tchicaya, and Schmidt-Ehry, 1999~. It is not yet clear, however, how such mechanisms can be developed. Some researchers stress the need for better ways to identify the poor families that need exemptions; others stress the need to better monitor exemptions so as to prevent use of the subsidies by groups who can afford care; and still others stress the need to focus on the fundamental causes of poverty and inequity. Quality of care Sanders, Kravitz, Lewin, and McKee (1998: 366) conclude that "the inappropriate utilization of referral facilities will remain a problem until quality accessible (and affordable) primary and secondary level care is available." Quality of care is now mentioned frequently, and, as just noted, its importance is underscored by studies exploring how user fees affect utilization. The aspects of quality found to be asso- ciated with use include drug availability, prescribing and dispensing practices, the physical condition of health facilities, service availability, number of personnel, crowding and length of waiting time, attitudes displayed by health workers to- ward clients, and the degree of confidentiality (Hotchkiss, 1998; Brugha and Zwi, 1999; Bassett, Bijlmakers, and Sanders, 1997~. A study of rural and urban Zimbabwe highlights the interactions between nurses and women from the community (Bassett, Bijlmakers, and Sanders, 1997: 185~: All groups were agreed in much of their assessment of the state of health services. Clinic fees, drug shortages and long waiting times were all identified as sources of dissatisfaction and declining quality of care.... To community women, the expectation of abrupt or rude treatment was the main complaint about the health services. Commu- nity complaints were voiced most strongly in the urban areas, where accusations of patient neglect and even abuse suggested a heightened hostility between the clinic and community in the urban setting. Sev- eral explanations for nurse behavior were put forward, chief among them was elitism.... it is in urban areas that class differentiation is most advanced. [The perspective of nurses differed. For them] over- work and low pay promote the adoption of the attitude of an industrial worker to do what is required and no more. Most nurses work more than one job, not to get rich but to survive. As elsewhere, efforts to professionalize nursing have benefited only a few.
294 CITIES TRANSFORMED As noted above, the main reason for use of private facilities, despite what are often high costs relative to government services, has to do with quality of care. Private practitioners appear to be more responsive to patient needs, both in inter- personal relations and in the establishment of opening and closing times that suit community needs (Bennett, 1992; Thaver, Harpham, McPake, and Garner, 1998~. Nevertheless, the relatively few studies that have explored the quality of care in private facilities from the technical point of view (Nichter, 1996; Brugha and Zwi, 1999) have produced disturbing findings, such as inappropriate prescription of antibiotics as a prophylaxis for STDs. Urban/rural linkages At a time when thinking on urban health is broadening to include social aspects of the environment and recognizing the multiple factors that operate at different levels, attention must also be paid to the sociodemographic linkages between rural and urban areas. The importance of urban/rural links in health is most frequently illustrated in the transmission patterns of HIV (and other infectious diseases). As noted in Chapter 6, a number of studies in sub-Saharan Africa have reported large differences in HIV prevalence among urban areas, roadside settlements, and ru- ral areas (Boerma, Nunn, and Whitworth, 1999), with prevalence generally being higher in urban areas. In Africa, however, many urban/rural links spread HIV to rural areas. The spread of disease from rural to urban populations can also result when migrants lack immunity to an endemic urban disease and spread it upon their return to rural areas or when urban residents lack immunity to diseases prevalent in rural areas. In some contexts, then, the interaction between urban and rural populations contributes to a sharing of disease patterns and risk factors. Poor urban and ru- ral populations may also exhibit similar responses to ill-health. One study on the Kenyan coast demonstrated similar treatment-seeking patterns on the part of low-income urban and rural mothers in response to childhood fevers and convul- sions (Molyneux, Mung'Ala-Odera, Harpham, and Snow, 1999~. The similarity in responses was unexpected given the significant differences between the two groups in socioeconomic status and distance to health services. It may be that urban and rural households exchange information and ideas about illnesses and appropriate therapies through migration and mobility, as well as through com- munication among spatially dispersed family members. Molyneux et al. (1999) found urban/rural ties to be strong: · One of every three lifelong rural-resident mothers had a husband who lived elsewhere, most of them (80 percent) in urban areas. · One of three urban-resident mothers had spent at least 10 percent of her nights elsewhere during the previous year (or since migration into the current household of residence), mainly in rural areas.
MORTALITY AND MORBIDITY 295 · Some 10 percent of lifelong rural-resident mothers had spent at least one- fifth of their nights during the previous year with urban residents (either in visiting cities or in hosting city visitors in their households), and 14 per- cent of urban residents had spent at least one-fifth of their nights with rural residents. · Over 60 percent of urban-resident mothers reported regularly assisting one or more persons living elsewhere, and most of those who were assisted (90 percent) resided in rural areas. · Fully 74 percent of urban-resident mothers stated that they wished to "retire" in a rural area. The importance of moving beyond the urban/rural divide in urban health think- ing is also highlighted by studies exploring referral systems within districts and countries (see, for example Akin and Hutchinson, 1999; Okello, Lubanga, Guwatudde, and Sebina-Zziwa, 1998) and by studies documenting the return mi- gration to rural areas of ill urban family members (Kitange, Machibya, Black, Mtasiwa, Masuki, Whiting, Unwin, Moshiro, Klima, Lewanga, Alberti, and McLarty, 1996~. CONCLUSIONS AND RECOMMENDATIONS This chapter has reviewed the extant knowledge about urban health, including health services. It has explored urban/rural, interurban, and intraurban differences in health and examined access to and quality of health care services, emerging health threats, and treatment-seeking behaviors. The major findings presented relate to urban and rural differences in health and health services; the relatively disadvantaged health status of the urban poor; and the shift from communicable diseases to chronic diseases, injuries, and mental health problems in low-income cities. These findings, based on a comprehensive literature review and analysis of the DHS-United Nations urban database, serve as the basis for the conclusions and recommendations presented below. Conclusions Child survival and child health Infant and child mortality rates are higher on average in rural than in urban ar- eas. Rural infant and child mortality is higher on average for a variety of reasons, including better urban public infrastructure, higher levels of maternal education, and better access to health care. Infant and child mortality risks also differ by region, with mortality being predictably and significantly higher in sub-Saharan
296 CITIES TRANSFORMED Africa than in other regions. However, rural areas still have relatively high mor- tality rates in all regions. Within the urban hierarchy, mortality varies little by city size. There is no clear evidence of systematic erosion over time in the urban ad- vantage in infant and child survival, except in some areas of sub-Saharan Africa. However, urban poor children face mortality risks that are significantly greater than those faced by the urban nonpoor. On average, infant and child mortality has declined over time by about the same amount in both urban and rural areas in most countries for which data are available. In six sub-Saharan African countries, however, mortality has increased in both urban and rural areas, slightly more so in the former. In the other half of the African cases examined, however, infant and child mortality has declined somewhat, with slightly larger declines evident in urban areas. The rapid urban spread of HIV/AIDS in some African countries may help explain their urban mortality increases, although it is difficult to disen- tangle the effects of the epidemic from other contributing factors. Although urban poor children have lower mortality risks than rural children in a majority of coun- tries examined by the panel, rural children exhibit lower mortality risks in 25 of 87 surveys reviewed. If health is assessed in terms of heightfor age, children in urban areas are sig- nificantly healthier than those in rural areas. Considering variation by city size, the panelfound children living in the largest cities to be better Nathan those in the smallest cities in Latin America, but other regions showed no clear pattern. In terms of height for age, urban children are on average 0.5 standard deviation taller than rural children. Although urban and rural differences in weight for height are smaller than those in height for age, they display the same general pattern. In North Africa, Southeast Asia, and sub-Saharan Africa, urban children are signif- icantly heavier for their height than rural children, although the reverse is true in six countries. City size differences in weight for height show no interpretable pat- tern except in Latin America, where children in the largest (5 million plus) cities are better off than those in the smaller cities. With health measured by children's height and weight, we find no clear evi- dence of systematic erosion in the urban health advantage for children during the past two decades. However, poor urban children are much less healthy than non- poor urban children, and they are sometimes not as healthy as their rural coun- terparts. Although there are few countries for which changes over time can be examined, it appears that the urban health advantage for children has not changed during the past two decades. In North Africa and Latin America, children's height for age has increased in both urban and rural areas over time. In both urban and rural areas of sub-Saharan Africa, however, there has been either no change or a decline in children's height over time, perhaps because of severe famine, war and civil conflict, and economic crisis. Urban children overall may be healthier on average than rural children, but some urban children those in poor families are worse off in some countries. In
MORTALITY AND MORBIDITY 297 nearly all countries, poor urban children are shorter and weigh less than nonpoor urban children. And although urban poor children are usually taller for their age than rural children in most countries examined by the panel, they are shorter in a few countries. Considering weight for height, poor urban children are generally heavier than rural children, but in 16 countries the reverse obtains. The urban health penalty In summary, there is no clear and compelling evidence of an emerging urban health penalty that puts urban children at greater risk than rural children. How- ever, the urban poorare generally worse Nathan the urban nonpoorand in several cases may fare worse than rural children. The panel's review of the literature and analysis of the data yielded a somewhat conflicting picture of urban health. Mi- cro studies of selected urban neighborhoods often suggest an erosion of the urban advantage in health, yet the DHS data show little evidence of this erosion. Of course, DHS surveys do not allow the effects of spatially concentrated poverty to be examined in any depth, and it may well be the spatially concentrated dis- advantage that produces the urban penalties seen in the micro studies. This is a high-priority area for future research. The disease spectrum Urban populations face growing health threats in terms of injuries, mental health problems, and chronic lifestyle diseases. Although environmental risks and com- municable diseases continue to be important, urban populations are increasingly at risk of injuries due to violence and accidents, mental health problems, and chronic diseases (e.g., heart disease, diabetes). This transformation, commonly known as the epidemiological or health transition, means that communicable diseases such as malaria and cholera become less of a threat in urban areas as these areas con- tinue to grow. Some communicable diseases, however, including HIV/AIDS and tuberculosis, are still important factors in cities. Recommendations Although health policy makers clearly must not abandon rural areas, those work- ing within urban areas should focus on reaching the urban poor. The panel's analysis clearly indicates priority areas for policy analysis and action in urban health, and strongly suggests the need for more comparative inter- and intraurban data and research on health and health service delivery. The following recom- mendations generally dovetail with those given in Chapter 5 regarding social and economic differentiation and access to public services, and with those in Chapter 6 regarding reproductive health service delivery, fertility, and reproductive health research. It may be that integrating services and synthesizing research in all of
298 CITIES TRANSFORMED these areas would help identify problems and solutions in the three interrelated domains. Policy, data, and research needs are addressed below. Service delivery The deficient health status of poor urban children in comparison with their nonpoor urban counterparts and sometimes in comparison with rural children leaves no doubt as to the importance of a focus on urban poverty in nutrition and health programs. Urban policy makers in Africa in particular should focus on improving environmental conditions and access to health services, as well as curbing the spread of AIDS, to reduce infant and child mortality among the urban poor. Dealing with emerging health problems will require a shift in health services, including different drugs and medical equipment and training, as well as different types of health promotion campaigns. The panel has identified four substantive areas of focus for policy makers. The following recommendations are made in the hope that they will be clarified and augmented by future research: · Governments should consider both inter- and intraurban differences when designing health services and public health infrastructure in cities. · Both governments and international agencies should place high pri- ority on the urban poor in increasing access to health services and improving public health infrastructure. · Closer attention should be paid to health conditions in smaller cities, especially in sub-Saharan Africa. · The urban health sector should adapt its programs and communication strategies to address emerging health threats such as injuries, mental health issues, and chronic disease. Data collection Apart from a few small-scale individual city and neighborhood studies, inter- and intraurban data on mortality and health are generally un- available. Data on mortality, morbidity, health service delivery, and public health infrastructure should be collected for cities of different sizes and for various socioeconomic groups and neighborhoods within cities. National surveys will remain very important the panel's specific recommendations for the DHS are provided in Appendix F but we encourage the collection of data in small-scale studies that could complement and enrich the findings from such national surveys. Studies that permit comparisons among the ur- ban poor and rural populations are of particular interest. In addition, much could be learned about mortality through the judicious use of data from na- tional censuses, especially when a census contains information on the areas of origin of recent migrants. Research A lack of data has led to gaps in research in the areas of urban mortality, health, and health services and infrastructure. There is an urgent need for
MORTALITY AND MORBIDITY 299 comparative research on inter- and intraurban differentials in health and treatment seeking, the quality of urban health services, and perceptions of quality. Although the urban advantage would appear to have persisted over time, research on urban penalties is urgently needed in sub-Saharan Africa, where there are credible accounts of stagnation and even declines in child health in some countries. We were unable to fully explore health and mor- tality trends over time using data from the WFS in addition to those from the DHS, but this could be a fruitful avenue for future research.