Tobacco Documents Online

Objectives. This study exam- ined the socioeconomic precursors of disparities in maternal health by measuring the associations of nine neighborhood-level indicators of so- cial phenomena with low infant birthweight. Methods. Vital records and cen- sns data for the Chicago metropolitan area in 1990 were merged (n = 112 327); a logistic regression model predicting low birthweight was esti- mated by backward elimination. Results. v~r~th individual-level variables held constant, six neighbor- hood-level indicators predicted low birthweight, together contributing to a variation in rate of 5.5%. Commu- nity economic hardship and housing costs were positively associated with low birthweight, while community socioeconomic status, crowded hous- ing, and high percentages of young and African-American residents were negatively associated with low birth- weight. Conclusions. Maternal health inequalities should be explored in the context of historical segregation, so- cial stratification, the dynamics of social support, and resource sharing. among communities. Several commu- nity characteristics associated with povex~.T are negatively associated with low birthweight. The traditional focus on individual risk factors for low birthweight limits our under- standing. (Am J Public Health~ 1997; 87:597-603) Neighborhood ioclal t nvtronments and the Distribution of Low Birthweight in Chicago o oo o o Eric M. Roberts, PhD Introduction " Among the neighborhoods of the inner city are some that serve as settings for many of today's most acute and perplexing public health problems. Some authors posit the existence of a relatively new urban underclass, characterized by concentrated poverty and social disloca- tion, racial and economic segregation, and norms of behavior that are markedly different from those of the rest of society,t Whether or not one accepts the existence of the urban underclass, it is inarguable that maternal health indicators for certain inner-city populations are alarming. While the rates of low birthweight and infant mortality for the state of Illinois as a whole in 1990 were 7.6% and 10.7 per 1000, respectively, those for certain neigh- borhoods of Chicago were as high as 19.5% and 31.9 per 1000 (Illinois Depart- ment of Public Health, Division of Family Health, unpublished data). Our understanding of the causes of these inequalities, however, is simplistic at best. Discussions of racial disparities in health are often collapsed into uni- dimensional explanations of socioeco- nomic status,~ although the causes of these disparities are very likely multifacto- rial, involving economics, health behav- iors, social environments, community dynamics, and segregation. Furthermore, the story of social inequalities in maternal health is traditionally told in terms of individual risk factors for low birthweight or infant mortality, which directs attention toward individual determinants of health and away from economic and slrucmral ones. It is likely that social phenomena that affect people at the level of entire communities play a role in creating these inequalities. It is therefore important to understand how neighborhood characteris- tics that may reflect these phenomena, such as poverty rates, housing conditions, and demographic composition, could be associated with individual health status. This paper examines community- level indicators of social phenomena for associations with individual probabilities of low birthweight. Data are drawn from the metropolitan area of Chicago, which is well known not only for the extremes of social conditions encountered in different parts of the city but also for its diversity of well-defined and thoroughly studied neigh- borhoods. The goal is to direct discussion toward the social, economic, political, and historical contexts in which disparities in maternal health have arisen. Social Environments and Maternal Health The data suggesting that community social environments may be associated with individual maternal health come from a variety of sources. Several au- tbors4-6 have reported that social support-- whether from a mother's existing support network or provided as a clinical interven- tion-is associated with increased birth- weight and postuatal maternal health. Studies addressing the effects of neighbor- hood social environments on individu- als,7-9 however, generally use less "medi- cal" indicators of well-being as outcome measures. These investigations employ a variety of indicators of general neighbor- hood socioeconomic stares (SES), which is regarded as a rough gauge of neighbor- The author is with the Medical Scholars P~gram and the School of Social Work, University of Illinois at Urbana-Champaign, Urbana. Requests for reprints should be sent to Eric M. Roberts, PhD, Medical Scholars Program and School of Social Work, University of Illinois at Urbana-Champaign, 190 Medical Sciences Bldg, 506 S Marthews Ave, Urbana. IL 61801. This paper was accepted June 21, 1996. April 1997, Vol. 87, No. 4 American Journal of Public Health 597

Roberts TABLE 1--1ndividual- and Community-Level Descriptors, Based on Chicago 1990 Vital Records and Census Data and Used in Logistic Regression Analysis (n = 112 327) Variable Method of Computation/Unit Mean -+ SD Individual characteflstics Race/ethnicity= Hispanic Coded as 0 or I 0.19 - 0.39 White Coded as 0 or 1 0.52 +- 0.50 Age In years 26.68 -+ 5.92 Marital status Single, divomed, or wid- 0.64 _~ 0.48 owed = 0; married = 1 Education Years (0-18 ~-) 12.51 -+ 2.85 Prenatal care use Inadequate = O; intermedi- 1.57 -- 0.65 ate = 1; adequate = 2 Parity No. children born live to 2.10 +_. 1.26 mother, including present birth Cigarette smoking Cigarettes per day 1.52 -+ 4.90 Alcohol consumption Drinks per week 0.05 __ 0.71 Community characteristics Socioeconomic status Average of zscores of % of -0.40 -+ 0.94 white-collar workers, median family income, median adult education level Economic hardship Average of zscores of % of 0.58 -+ 1.22 unemployed adults, % of families in poverty % of young residents % residents under 18 years 27.49 _-+ 6.10 % of old residents % residents over 64 years 10.71 -+ 4.58 Stability % residents with same 45.85 +-. 9.81 address 5 years before % of African-American resi- ... 27.83 +_ 36.76 dents % of Hispanic residents ... 14.11 -+ 20.07 Median rent $ per month 505.59 -+ 125.62 Crowded housing rate % housing units with more 6.29 ± 5.84 than one person per room aAfdcan Amedcan is the reference category. hood "quality." They measure associa- tions with individual outcomes such as educational attainment, crime, teenage sexual behavior and childbearing, and adult success in the labor market. Ethnographic data suggest mecha- nisms by which social dynamics within a community could influence personal well- being. In particular, Anderson's analysis of urban, low-income communities de- scribes young people caught in a tug-of- war between two lifestyles, one emphasiz- ing the value of hard work, getting ahead, and having options for the future, and the other revolving around a number of problem behaviors, such as violence, drug use, and irresponsible sexual activity.I° Anderson argues that central to the pro- pagation of these former values in a community is the presence of close-knit. financially stable extended families; this effect is undermined, however, by wide- spread poverty and unemployment. Addi- tional work supports the assertion that kin networks play a large role in the well- being of single parents, particularly among African-American families. I I,~ 2 An examination of the effects of neighborhood characteristics on individ- ual well-bein.g naturally raises the ques- tion of how the neighborhood characteris- tics came to be in the first place. Much of the social differentiation of th~ city that we see today is the result of powerful individuals and groups' actively working to divert resources and maintain segrega- tion in the ongoing oppression that has occurred over most of the 20th cen- tl/Ty.13A4 The effects of social differentia- tion on personal well-being that we observe today, both negative and positive, must be seen in the context of this political and economic history. Methods Data Sources The data set for this project was constructed from the 1990 vital records for lllinois, which provided infant birth- . weights and maternal characteristics for every Illinois birth, and the 1990 United States Census of Population and Housing, which provided neighborhood descrip- tors. Only the portions of these data sets describing the six-county Illinois segment of the Chicago metropolitan area were used. Each birth in this area during 1990 represented one case in the data set; each case contained information about the infant's birthweight, the individual maternal characteristics, and the charac- teristics of the community in which the mother resided. Within the city of Chicago, the desirable unit of analysis was the commu- nity area (neighborhood). Census tract information was aggregated to this level, using the definitions of the 77 Chicago community areas currently used by the Illinois Department of Public Health. Outside the city, data were aggregated to the level of individual towns. Selection of Communio'-Level Indicators No comprehensive review of commu- nity-level indicators of social processes and their associations with individual well-being exists. In keeping with the exploratory nature of the present study, indicators were chosen that were sug- gested by the diversity of social science literature discussed above. Convention suggests the use of neighborhood SES as an indicator of neighborhood "quality." in spite of a relative paucity of theoretical framework justifying this construct. Rates of poverty and unemployment have been clearly implicated as being connected with individual well-being because of their effects on both personal resources and social dynamics. Ethnographic data suggest that community age distribution and stability can influence personal sup- port networks, and that norms of behavior within these networks can vary between communities of differing racial and ethnic compositions. Intemeighborhood varia- tions in housing costs and crowding are among the clearest legacies of segregation within the city and are therefore also indicated as potential covariates with individual health status. Coding procedures for these vari- ables are summarized in Table 1. Infant 598 American Journal of Public Health April 1997. Vol. 87. No. 4

birthweight was coded as 1 for less than 2500 g and 0 for ~00 g or more. Prenatal care use was coded according to the classification procedure of Kessner et al.ts as modified by Showstack et a1.16 Analysis A logistic regression model was estimated by backward elimination. This is to say that all available variables were loaded into the model initially; those without significant coefficients (a = .05, using the Wald statistic) in the resulting equation were then removed and the model reevaluated in a stepwise fashion. To assess model stability, coefficients were reestimated using three random 50% subsamples from the data set. Both univariate and specific combinations of the multivariate coefficients were re- viewed; race-specific subsets of the data were also analyzed. The goodness of fit of the model for areas of particularly high rates of poverty or low birthweight was assessed by the inclusion of dummy variables indicating births from these communities, the Score statistic of such a variable providing an assessment of the fit of the model for that community,t7 Results The entire sample described 131 457 births to residents of the six-county area in 1990. Of these, 112327 (85.4%) re- mained after deletion of cases with missing data. Rates of low birthweight within the city are shown by community area/neighborhood in Figure 1. In all, three community characteristics (stability, percentage of Hispanic residents, and percentage of old residents) and one individual characteristic (parity) were re- moved from the model. The results of the final logistic regression analysis are shown in Table 2. For comparison, these coefficients are shown alongside regressions of individual and community variables alone as well as with the univariate coefficients. Among the individual-level risk factors, the sign of the multivariate age coefficient is somewhat counterintuitive, since young mothers are commonly assumed to be at higher, not lower, risk for low birthweight. This elevated risk has been shown to disappear, however, when social factors such as poverty and stress are controlled, making these findings consistent with those of previous studies,ts Six community characteristics were found to be significant predictors of low Envtrmunent and Birthweight Percent LBW [] O.Ito 4 [] 4.~to 8 [] 8.1to 12 [] 12.1 to 16 [] 16.1 to 20 Source. Illinois Department of Public Health vital records. FIGURE 1--Incidence of low birthweight (LBW) in Chicago, by community area, 1990. birthweight. The index of economic hard- ship was positively associated with risk for low birthweight, as was median rent. Negatively associated with risk for low birthweight were the index of neighbor- hood SES, the percentage of African- American residents, the percentage of young residents, and the crowded housing rate. The robustness of the model was assessed by random 50% subsample analysis. Three of the 14 coefficients were observed to change in magnitude by more than 20% in different subsamples---these were the indices of economic hardship and SES and the crowded housing rate. Of these, the coefficient for crowded housing actually switched from a negative to a marginally positive value, so its interpreta- tion, out of the entire set, is the most questionable. Changes in the economic hardship coefficient ranged from 7.0% to 28.7%, while those in the SES coefficient ranged from 26.8% to 41.8%. The sizes of these associations were examined by entering different values for each independent variable into the model equation and observing the changes in the predicted probability of low birthweight; all other variables were held constant at their means, so the estimated change was due to that variable alone (Table 3). For most of the variables that are not nominal or ordinal, the comparison in Table 3 is between the 10th and 90th percentile values of each independent variable; this way, the probability of low birthweight is compared for mothers across the spec- trum of the sample. The one exception is a/cobol consumption; fewer than 10% Agril 1997, VoL 87, No. 4 American Journal of Public Health 599

Roberts of the mothers in the sample are recorded as using .any alcohol during pregnancy, so the comparison values are arbitrarily selected as zero and three drinks per week. This procedure suggests that, with all other variables controlled, maternal race and ethnicity is still the risk factor that explains the largest amount of social inequalities in risk for low birthweight; Hispanic mothers are 6.25% less likely than African-American mothers to have a low-birthweight infant. Other factors with large effects include the index of poverty and unemployment, marital status, prena- tal care use, and age. All of the other associations, although significant, repre- sented changes in the probability of low birthweight of less than 2% by them- selves, although the effect of changing several variables at once under this model could be quite pronounced. For example, if one holds all of the individual-level variables constant at their means, the model predicts a minimum rate of low birthweight of 4.4% in Hawthorn Woods and a maximum of 9.9% in Hainesville, a range of 5.5%. A goodness-of-fit analysis was con- dueled for every community with a poverty rate above 40% or a low-birthweight rate above 15%--a total of 18 communities. Of these, two (Roseland and Washington Heights) had low-birthweight rates that differed significantly from those predicted by the model (ct = .05, using the Score statistic). These tom- munities had rates 2.4% and 4.1% higher than predicted, respectively. These differences are as- sumed to be due either to non- linearities in the model or to social processes not measured in the equation elevating neigh- borhood low-birthweight rates. Discussion Social Stratification: Povert); Unemployment, and SES After maternal race and ethnicity, the most substantial risk factor in the model appears to be the index of economic hardship. Focusing on the variable as an aggregation of individual characteris- tics, one can posit that women in high- poverty, high-unemployment communi- ties have fewer material resources and therefore run higher risks for malnutrition, lower quality health services, and stress. This interpretation is the most intuitive and involves factors influencing maternal health only at the individual level. The economic hardship variable lends itself to broader interpretation, however, if O 600 ,a, mcHcan lourna~ofPub~ic Hcixl~ April 1997, Vol. 87, No. 4

Environment and Birthweight one chooses to f~us on the social en- vironments experienced by the women living in high-poverty neighborhoods. As mentioned previously, poverty and unem- ployment can serve to undermine the cultural standards in a community. These factors can serve to destabilize families, eroding the support network available to an individual mother. Furthermore, eco- nomic hardship can interfere with the stabilizing influence of intact families on the community as a whole. The influence of this factor can be seen as beating upon entire communities in. muldple ways, in- fluencing the local subculture and hinder- ing mutual support. The association of low birthweight with the more traditional measure of neighborhood "quality," the index of SES, can also be thought of as resulting from the social environments experienced by the residents or simply from an aggregation of individual characteristics to the neighborhood level. An important issue is that neither the differentiation of social environments nor the grouping together of people of similar social and economic standing is accidental. Rather, these phenomena represent the expression of the social structure through the geo- graphic differentiation of the city. This expression has been reinforced in Chicago through economics, immigration pres- 1314 sures, violence, and public policy. • Allocation of Resources: Rent and Birthweight The inclusion in the model of the median monthly rent in the community focuses more specifically on the issue of maternal resources. Since the general in- come level of the mothers in each community is largely held constant by the economic hardship and SES variables, women in conu'nunities with lower rents may have more money available for food, clothing, transportation, child care, and so forth. This is the most direct explanation for the measured positive association of community median rent with maternal probability of low birthweight. At one level, therefore, this coeffi- cient serves to further reinforce the connection of maternal resources to preg- nancy outcome. The association with rent, however, has additional meaning because of its historical connection with racial segregation. Furthermore, although the direct association of rent with low birth- weight is relatively small, segregation can be linked to low birthweight by other mechanisms. Since African Americans in general are less able to translate gains in TABLE 3--Predicted Probabilities of Having a Low-Blrthweight Infant for Selected Valubs of Risk Factors Predicted Predicted Value= Probability, % Difference, % Race/ethnicity Afdcan American 11.31 6.25 Hispanic 5.06 White 5.83 Age 19 y 5.71 2.31 34y 8.02 Marital status Single 8.69 - 2.52 Wed 6.02 Education 9 y 7.38 - 1.04 16 y 6.34 Prenatal cam use Inadequate 8.88 -2.52 Adequate 6.36 Cigarette smoking None 6.54 1.12 6/d 7.66 Alcohol consumption None 6.81 1.00 3 ddnks/wk 7.81 Economic hardship Index = -0.60 5.60 3.63 index = +2.44 9.23 Socioeconomic status Index = - 1.72 7.67 - 1.56 Index = +0.82 6.11 Median rent $370/mo 6.25 1.39 $679/mo 7.64 Crowded housing rate, % 0.93 7.13 -0.63 12.39 6.50 % of African.American 0.25 7.23 - 1.32 residents 98.09 5.91 % of young residents 20.57 7.32 - 1.06 35.97 6.26 Note. For each calculation, all other variables are held constant at their means. =See text for explanation of comparison values. education and income into improvements in residential location, African-American families are more likely not only to reside in higher-rent neighborhoods but to re- main in higher-poverty, lower-SES neigh- borhoods in spite of personal gains in income.~9-22 This is the consequence of the oppression, both by other residents and by commercial and institutional inter- ests, that has restricted the movements of African-American families over the 20th century.t3A4 These community characteris- tics are together associated with a large range in rates of low birthweight; with all variables except the economic hardship and SES indices and median rent held constant at their means, the model pre- dicts a minimum low-birthweight rate of 4.2% (Hawthron Woods, Long Grove) and a maximum of 12.7% (Oakland), for a range of 8.5%. Thus, forced residential segregation can be seen as a health issue, contributing to disparities in maternal health between races. Community Social Environments and Family Dynamics Percentage of African-American resi- dents. Other variables appear to have more to do with behavioral dynamics within communities and less to do with the distribution of power and resources in society. One finding that might be consid- ered unexpected in light of conventional discourse on race and maternal health is that the percentage of African-American residents in a community is negatively associated with low birthweight when individual maternal characteristics (includ- ing race) and neighborhood-level vari- ables are controlled. It is well known that being African American is a risk factor for low birthweight in our society, and among traditional risk factors, it is the most striking and consistent. However, this fact does not preclude the idea that characteris- tics of African-American communities could promote maternal health. AgHI 1997, VoL 87, No, 4 American Journal of Public Health 601

Roberts As shown in Table 3, this variable • contributes to a variation in low birth- weight of about 1.3% across the sample, an amount that is small compared with that attributed to individual race or eth- nicity. In no neighborhood does the model predict a lower rate of low birthweight for African Americans than for Whites or Hispanics. Furthermore, when the moth- ers in each group arc analyzed separately, the coefficient for the percentage of African-American residents is nonsignifi- cant (P = .2244) for births to White mothers (data not shown). The implication is that, although African Americans have a higher incidence of low bit&weight than other racial or ethnic groups, African- American (not to mention Hispanic) mothers who llve in predominantly Afri- can-Arnedcan neighborhoods have fewer low-birthweight infants than their peers in other neighborhoods when other commu- nity-level indicators are held constant. Furthermore, race/etlmicity is the only variable in the present investigation for which wc can definitively distinguish between individual and neighborhood associations; that is, wc know that the association between the percentage of African Americans in a community and low birthweight is duc to a community characteristic and not simply an aggrega- tion of individual characteristics. Documentation of positive coping behaviors in African-American communi- ties can bc found in sociologicalt~ and anthropologicaP2 studies. In particular, Stack~2 investigated a low-income African- American community and observed strik- ing patterns of interdependence. Stack noted that "the material and cultural support needed to absorb, sustain, and socialize community members in [the neighborhood] is provided by networks of cooperating kinsmcn."i2~3) Although nuclear families are lacking by White middle-class definitions, multigcnera- tional networks defined by neighbors assuming dc facto caretaking roles are plentiful and play a central role in the well-being of residents. Most important, "responsibility for providing food, care, clothing, and shelter and for socializing children within domestic networks may be spread over several households" that tend to be geographically close to one another. ~ ~p,~o~ The key is that Stack describes char- acteristics of the African-American com- muni~.; the support networks operate within geographic areas in which mem- bers can easily interact and work with one another. Accordingly, an African-Ameri- can mother living in isolation from any supportive network would not experience the advantages of such a Community. This support primarily benefits African-Ameri- can mothers, since the coefficient be- comes nonsignificant for White mothers when they are analyzed separately from the rest of the sample. Other indicators of support net- works. Two other neighborhood character- istics--the percentage of young'residents and the crowded housing rate--appear to be negatively associated with low birthweight, although the coefficient for crowded housing rate appears unstable in subsample analysis. Again, the signs of these coefficients are more easily under- stood when one considers that other indicators of wealth and SES are largely held constant in the model. High concen- trations of children and crowded housing units may be indicators of neighborhoods where mothers live in close proximity to their support networks, where childbear- ing is normalized, and where the responsi- bility for support and resource provision are spread out over a large number of people. Again, these are probably not indica- tors of maternal resources: rather, they are indicators that resources in the maternal network are shared. Conversely, these results imply that communities where mothers are relatively isolated from their support networks--either by distance or by conventions of self-sufficiencymare likely to have higher incidences of low birthweight. All together, these commu- nity characteristics are associated with a substantial range in rates of low birth- weight; with all variables except the percentages of African Americans, people under 18, and crowded housing units held constant at their means, the model pre- dicts a minimum low-birthweight rate of 4.6% (Riverdale) and a maximum ofg.1% (the Loop), for a range of 4.5%. Summary As indicators of social processes affecting individual mothers, the commu- nity-level variables under study here are quite crude, but associations with mater- hal health are still detectable. The pic- ture of social inequalities in rates of low birthweight that emerges from these coefficients is one driven by broad soci- etal influences extending well beyond the health-seeking behavior of individual mothers. At the societal level, social stratification determines which mothers possess the resources necessary to care for themselves and which do not. Beyond this stratification, certain neighborhoods ap- pear to be home to community and family dynamics that mobilize and distribute resources to promote maternal health. Individual prenatal care use and marital status are still important determinants of pregnancy outcome, but in light of the present findings, the focus on traditional risk factors to explain social inequalities in maternal health appears both narrow and superficial. The present analysis is exploratory in nature. Therefore, variable selection was limited to previously measured indicators suggested for use by a broad reading of the social science literature. As an explor- atory effort, this paper uncovers patterns in low-birthwcight distribution that sug- gest intriguing mechanisms for social inequalities in health that have previously gone uninvestigated. A thorough under- standing of these mechanisms will come with more focused data gathering at the individual level, particularly with the development of instruments to assess the effectiveness of social support and kin networks and personal resources. In this way, the truc connections between commu- nity social dynamics and health status can be further illuminated. [] References 1. Wilson WJ. The Truly Disadvantaged: The Inner Cit~; the Underclass, and Public Policy. Chicago, Ilk University of Chicago Press; 1987. 2. Kricgcr N. Analyzing socioeconomic and racial/ethnic patterns in health and health care. Am J Public Health. 1993;83:1086-- 1087. 3. Pappas G. Elucidating the relationships between race, socioeconomic shams, and health. Am J Public Health. 1994;84:892- 893. 4. Oakley A. Social support in pregnancy: the "soR" way to increase birthweight? Soc Sci Med. 1985;21:1259-1268. 5. Newton RW, Hunt LE Psychosocial stress in pregnancy and its relation to low birth wcight. BMJ. 1984;288:1191-1194. 6. Boyoe WT, Cbestcrman EA, Winkleby MA. Psychosocial predictors of maternal and infant health among adolescent moth- crs.AmJ Dis Child. 1991;145:267-273. 7. Jcncks C, Mayer SE. The social conse- quences of growing up in a poor neighbor- hood. In: Lynn LE Jr, McGcary MGH, eds. Inner-City Poverty. in the United States. Washington, DC: National Academy Press; 1990:111-186. 8. Crane J. Effects of neighborhoods on dropping out of school and teenage child- bearing. In: Jcncks C, Peterson PE, cds. The Urban Underclass. Washington, DC: The Brookings Institution; 1991:299-320. 9. May~r SE. How much does a high school's racial and socioeconomic mix affect gradu- ation and teenage fertility rates? In: Jcncks 602 Ameri,~'an Journal of Public Health April 1997. Vol. g7. No. 4

Environment and Birthweight C, Peterson PE, ~ls. The Urt~n Under- class. Washington. DC: The Brookings Institution; lqg1:321-341. 10. Anderson E. Neighborhood effects on teenage pregnancy. In: Jencks C. Peterson PE. eds. The Urban Uwderclass. Washing- ton, DC: The Bro~kings Institution; 1991: 375-398. 11. Hogan DP, Hao/-, Parish WL. Race, kin networks, and assistance to mother-headed families. Soc Forces. 1990;68:797-812. 12. Shack CB. All Our Kin: Strategies for Survival in a Black Communit): New York. NY: Harper & Row: t974. 13. Spear AH. Black Chicago: The Making of a Negro Ghetto. 1890-1920. Chicago, I!1: University of Chicago Press; 1967. 14. Hirsch AR. Making the Second Ghetto: Race and Ho~ing in Chicago. 1940-1960. Cambridge, England: Cambridge Univer- sity Press; 1983. 15. Kessner DM, Singer J. Kalk CE, Schlc- singer ER. Cantrasts in Health Status. Volume 1. Infant Death: An Analysis by Maternal Risk and Health Care. Washing- ton, DC: Institute of Medicine, National Academy of Sciences: 1973. 16. Showstack JA. Budetti. PP, Minkler D. Factors associated with birthweight: an exploration of the roles of prenatal care and length of gestation. Am J Public Health. 1984;74:1003-1008. 17. Hosmer DW, Lemeshow S. Applied Logis- tic Regression. New York, NY: Wiley Interscience; 1989. 18. Geronimus A. The effects of race, resi- dence, and prenatal care on the relationship of maternal age to neonatal mortality. Am J Public Health. 1986;76:1416-1420. 19. Krivo LJ. Housing price inequalities: a comparison of Anglos, Blacks. and Spanish- origin populations. Urban Affairs Q. 1982; 17:445-462. 20. Goodman JL, Streitwieser ML. Explaining racial differences: a study of city-to-suburb residential mobility. Urban Affairs Q. 1983; 18:301-325. 21. Villemez WJ. Race, class, and neighbor- hood: differences in the residential return on individual resources. Soc Forces. 1980; 59:414-430. 22. Massey DS. American apartheid: segrega- tion and the making of the underclass. Am J SocioL 1990;96:329-357. APHA Publications Board Invites Proposals for Book Projects APHA's Publications Board invites APHA members to submit proposals for publication as books. The Board is looking for manuscripts that speak to public health topics, especially to those not previously or not adequately addressed. We need your most innovative work, your dedication, and your enthusiasm to create the best possible public health book program that APHA can offer. If you are interested in making a submission or if you have a topic in mind, feel free to discuss it with the Chair of the Publications Board, Dr Berttina Wentworth, or with the APHA Director of Publications, Ellen T. Meyer. To reach either or to receive guidelines on making a formal submission, call the Association Office at (202) 789-5693; fax (202) 789-5661. Please send preliminary inquiries or formal proposals to Ellen T. Meyer, Director of Pub- lications, American Public Health Association, 1015 15th St, NW, Washington, DC 20005-2605. Please note that all inquiries about publication in the American Journal of Public Health must be sent to the Editor of the Journal, Dr Mervyn Susser, at the APHA Washington, DC, address given above. 0 O~ O~ 0 0 A~,ril 1997, Vol. ,97, No. ~ American Journal of Public Health 603