|Home | About | Journals | Submit | Contact Us | Français|
To measure racial/ethnic inequalities in child dental health and quantify the contribution of several household, neighborhood, and geographic effects to these disparities using a decomposition analysis.
Using data from the 2007 National Survey of Children’s Health, we measured and decomposed racial/ethnic disparities in selected child dental health and dental preventive care outcomes. We employed a decomposition model to quantify the extent to which demographic, socioeconomic, maternal health, health insurance, neighborhood, and geographic effects explain these disparities.
Significant racial/ethnic disparities in dental health were observed. Hispanic children had the poorest dental health and lowest preventive dental care use, followed by African-American children, compared to Whites. The model explanatory variables accounted for a large proportion of the disparities in dental health (58–77%) and for most of the disparities in preventive dental care (89–100%). Socioeconomic status (maternal education and household poverty level) was the single most relevant factor for explaining these disparities and accounted for 71% of the gap between African-American and white children in preventive dental care, and 55% of that gap between Hispanic and white children. Other relevant factors for explaining disparities included maternal health, age, and marital status, neighborhood safety and social capital, and state of residence.
Racial/ethnic disparities in child dental health in the US are mostly socioeconomically driven involving household and neighborhood contributors. Reducing these disparities requires policies that recognize the multilevel pathways underlying these disparities.
Race and ethnicity are significant determinants of dental health in different countries and racial/ethnic minority status is a well-known risk factor for poor dental health 1, 2. Racial/ethnic differences in children’s dental health are common in the US. Among children aged 2 to 11 years, African-American and Hispanic children are more likely to have decayed teeth and untreated dental problems than whites 3, 4. The rate of primary dentition caries in 1999–2004 was 55% for Mexican-American children, 43% for African-American, and 39% for White children 3. About 6.5% of non-Hispanic white children have fair/poor oral health, compared to 12.0% of African-Americans and 23.4% of Hispanics, with large racial/ethnic differences remaining after adjusting for age, sex, education, poverty level, dental insurance and parental preventive care attitude 5.
Racial/ethnic disparities also exist in children’s access to dental care in the US. Larger unmet dental care needs are observed in nonwhite children. Moreover, Hispanic children have the highest likelihood of never having seen a dentist 6, 7. In addition, among children who are publicly insured through Medicaid and CHIP, Hispanic and African-Americans have longer intervals between dental visits and higher tooth decay rates 8.
While documenting racial/ethnic disparities in child dental health is important, of greater relevance is identifying the pathways that explain these inequalities, which is needed for informing policies that can effectively reduce them. Although racial/ethnic disparities in child dental health have been well documented, few studies have explored their underlying pathways and none has formally quantified the contributions of socioeconomic, demographic and neighborhood characteristics to these disparities. Previous studies highlight some factors as relevant for racial/ethnic disparities including socioeconomic condition, health literacy, educational attainment, dental insurance, language barriers, and cultural characteristics 2, 9–11. However, these studies did not adequately characterize the individual contributions of these and other theoretically relevant factors to disparities. The purpose of this study was to fill this research gap by measuring racial/ethnic inequalities in child dental health in a nationally representative sample, and quantifying the extent to which the observed disparities are attributable to conceptually-relevant factors, using a decomposition analysis. In doing so, the study highlights important pathways contributing to child dental health disparities.
We employed data from the 2007 National Survey of Children’s Health (NSCH), a cross-sectional national survey conducted by the National Center for Health Statistics at the Centers for Disease Control and Prevention (CDC) to assess different aspects of children’s health, and access and use of health care services. Ours is the first study to use the 2007 NSCH data for studying child dental health disparities. The NSCH’s sampling design provides a nationally representative sample of children and adolescents (up to 18 years) in the US. After identifying a household, the NSCH randomly selected one index child and collected data from parents/caregivers on several demographic, economic, health/healthcare and neighborhood characteristics 12.
Our sample initially included 57,027 children aged 3 years or older for whom the NSCH interview was completed by their mothers in order to avoid errors in maternal and child data from other respondents. We excluded children younger than 3 years since they had a very low frequency of dental problems (less than 0.5%), and since the differences in this outcome between the study racial/ethnic groups were not significant (at p=0.05).
The final sample for our analyses ranged from 43,972 to 45,237 children who had complete information on one or more of the dental health outcomes and all explanatory variables.
We focused on studying non-Hispanic white, non-Hispanic black (African-American), and Hispanic children. We measured and decomposed racial/ethnic disparities in three child dental health and care outcomes. The first dental health outcome measured if the child had any of the following dental problems over the past 6 months: toothache, decayed teeth or cavities, broken teeth, and bleeding gums. The second dental health outcome was based on maternal rating of the child’s dental health status as excellent, very good, good, fair, or poor. We created a binary indicator combining excellent/very good/good versus fair/poor. The third outcome was the number of preventive dental care visits during the last 12 months. Since the American Academy of Pediatric Dentistry recommends two dental preventive per year13, we used a dichotomous indicator for two or more preventive dental visits.
We evaluated the contribution of several conceptually-relevant categories of household and neighborhood/area characteristics to the racial/ethnic disparities in the study dental outcomes. It is important to recognize the complexity and multiplicity of the underlying mechanisms influencing health and health behaviors when studying health disparities. Therefore, our choice of the explanatory variables was motivated by general microeconomic theory for health/healthcare demand and for health production supplemented with psychosocial and neighborhood effects given their importance for health. We briefly describe below the theoretical associations between the explanatory variables and the outcomes. Given that our model specification is motivated by theory, we retained explanatory variables even if they had statistically insignificant effects on the outcomes, as their omission may result in bias in effects of other explanatory variables. In addition, we only included explanatory variables showing significant differences between African-American or Hispanic children compared with whites because of their potential to explain the observed disparities in study outcomes; variables with a similar distribution by race cannot explain these disparities.
Demographics included maternal age, marital status, and whether the child was born in the US. Maternal age reflects parental skills, knowledge and experience in child health management. Marital status may influence the availability of childcare and economic resources. Birth in the US may associate with socioeconomic and cultural differences between American and immigrant children that are relevant for dental health. We did not include child gender and age because these were not significantly different between the study racial/ethnic groups and therefore could not explain the observed disparities.
Maternal health was measured from self-reported general health status (excellent/very good/good vs. fair/poor). Maternal health could reflect the presence of common risk factors for child general/dental health. It can also affect maternal ability to care for the child, as poor maternal health can reduce the availability of time and economic resources needed for enhancing child health.
The third category, socioeconomics, included household income level, household employment status, and highest maternal educational attainment. Socioeconomic status can affect child dental health in several ways, such as by affecting parental knowledge and ability to enforce optimal dental health-related behaviors and to access dental care. Socioeconomic status also affects parental and child psychosocial status (such as stress/anxiety) which may affect dental health behaviors. Furthermore, education can directly affect maternal efficiency/ability to identify child dental problems and access needed dental care.
Child insurance status, which can direct affect access to dental care, was represented by indicators for public or private insurance relative to being uninsured at the time of the interview. Household demographics included the total numbers of children and adults in the household, which may affect dental health through allocation of resources in the household and child care.
Neighborhood conditions may affect dental health and care in various ways including physical safety, social networking and information about healthcare/dental services, and supply of dental care providers. Given that neighborhood characteristics may vary by race/ethnicity, and contribute to racial/ethnic disparities in dental health, we evaluated nine neighborhood characteristics grouped into three categories. Neighborhood condition included the presence of litter or garbage, poorly kept or rundown housing, and availability of library/bookmobile. Neighborhood safety was measured by presence of vandalism and perception of child’s safety. Neighborhood social capital included indicators for people in the neighborhood helping each other, watching out for each other’s children, counting on each other, and adults helping children in case they get hurt or scared. Finally, given the geographic variation in race/ethnicity distributions and in dental health and care (such as due to the distributions of dental professionals), we also evaluated the extent to which the state of residence explained the racial/ethnic disparities in dental health. Table 1 presents the study variables and their distributions by race/ethnicity.
The use of Oaxaca-Blinder type decomposition models14, 15 has significantly contributed to understanding some of the underlying pathways for inequalities in health status and health care16–19. The basic premise for these models is to quantify the extent to which differences in the distributions of explanatory variables between two groups (such as minority versus majority) account for their differences on a certain outcome. This decomposition approach essentially estimates a multivariate model for the outcome that includes all the explanatory variables of interest, substitutes the means of the explanatory variables for one group, one at a time, by the means of the explanatory variables for the other group, and recalculates the difference in the conditional outcome means between the two groups after each explanatory variable mean substitution. The change in the outcome mean difference between the two groups with the mean substitution for a certain explanatory variable represents the contribution of that variable to the total outcome gap between these groups.
The Oaxaca-Blinder decomposition model is designed for linear outcomes. Given that we studied binary outcomes, we employed the Fairlie decomposition model for non-linear binary outcome models 20 which has been previously used to decompose health and healthcare disparities 17, 21, 22. Our goal was to quantify the contribution of each category of explanatory variables described above to racial/ethnic disparities in child dental health outcomes. We decomposed the disparities separately for African-American and Hispanic children compared to non-Hispanic whites. For each outcome comparison (such as any dental health problem between African-American and white children), a logistic regression was estimated including race/ethnicity (such as African-American versus white) and all explanatory variables. Next, conditional probabilities of the outcome (such as probability of any dental health problem) were predicted for each observation. Then, a subsample of an equal number of observations to the minority group (African-Americans) was randomly selected from the majority group (whites). Within this selected majority subsample and the minority group, each observation was rank-ordered by the predicted outcome probability describe above. Next, each observation in the minority group was matched with the observation from the majority subsample with an equal rank. Then, one variable at a time, the values of each explanatory variable in the minority group were replaced by the values of the matched observations for the same variable from the majority subsample, and the difference in the dental health outcome probability between the minority and majority groups was estimated. This difference represents the disparity explained by a particular variable. The total outcome difference explained by all variables is the sum of the differences explained by the individual variables (which may also be obtained by switching all variables at the same time from the minority to the majority values). Similar to the individual variable analysis, categories of variables may be evaluated for their combined contribution to the racial/ethnic disparity. Appendix 1 includes further illustration of the statistical model.
Given that the results depend on the particular randomly selected majority subsample, we obtained 2,000 randomly selected majority subsamples and averaged the decomposition results across these selected subsamples 20. This repeated majority subsample selection approximates the estimates that would be obtained if the total majority sample was matched to the minority sample. In addition, because the decomposition results for each variable category can be affected by the order when its variable values are switched from the minority to majority relative to the other categories, we randomly selected the category order at each replication for randomly selecting the majority subsamples. The 2000 replications are expected to approximate the average result from all possible variable-category orderings. All analyses were estimated using the NSCH sampling probability weights to obtain population-based results.
There were large differences in the prevalence of dental problems by race/ethnicity, particularly between Hispanic and white children (Table 1). About 40% of Hispanic children had at least one dental problem during the last year compared to 34% of African-Americans and 24% of Whites. More than one fifth of Hispanic children had their teeth reported in fair/poor condition compared with 10.6% of African-Americans and about 5% of whites. About 47% of Hispanic children received the recommended preventive dental care in the past year, compared to 51% of African-Americans and 61% of whites.
There were also important differences in the study explanatory variables by race/ethnicity. The rate of married mothers was highest among white children (82%) and lowest among African-Americans (41%). In addition, fair/poor maternal health was more than twice as common among Hispanics and African-Americans as whites (21% and 19%, respectively, versus 8%). There were considerable differences in all socioeconomic and human capital indicators by race/ethnicity, with Hispanics having the highest poverty level and lowest employment rate and maternal educational attainment. African-American children had the highest rate of public insurance, and Hispanic children were significantly more likely to be uninsured compared to African-American and white children. Compared to white and Hispanic children, African-American children were more likely to live in poor, rundown and less safe neighborhoods. White and African-American children generally lived in neighborhoods with the highest and lowest social capital, respectively.
Table 2 presents the percentages of the racial/ethnic disparities in the study dental health and care outcomes explained by the decomposition model. The model explanatory variables accounted for a substantial part of the racial/ethnic gaps in dental health and preventive dental visits. The model explained the entire gap in preventive dental visits between Hispanic and white children and 89% of that gap between African-American and white children. Also, the model explained 77% of the disparities in dental health problems and self-rated dental health between African-American and white children, and 65% and 58% of these disparities, respectively, between Hispanic and white children.
Table 3 presents the racial/ethnic differences in the study dental outcomes accounted for by each explanatory variable category. Graphs 1 and and22 depict the proportions of the total gaps explained by these variable categories for African-American and Hispanic children, respectively. Table A1 in Appendix 2 reports the contribution of each variable towards explaining these disparities. In general, household socioeconomic characteristics were the single most relevant category for explaining these gaps.
Among all model categories, demographic differences (mainly higher rates of unmarried mothers for African-American children) explained the largest percentage (19.5%) of the gap in dental health problems between African-American and White children. Lower socioeconomic status (lower education and higher poverty level) and neighborhood safety among African-American children were the next most relevant factors accounting for 16.4% and 14.2% of this gap, respectively. Similarly, lower socioeconomic status accounted for 30.9% of the gap in prevalence of fair/poor rated dental health between White and African-American children, followed by maternal health, child’s insurance status, and neighborhood social capital which explained 21.2%, 14.3%, and 13.8% of this gap, respectively. Lower socioeconomic status (lower education and higher poverty level) was also the most relevant for explaining the lower use of preventive dental care among African-American children (71% of this difference), followed by demographic differences (mostly younger maternal age), which accounted for 26% of the gap in preventive dental care use.
Almost 30% of the higher prevalence of dental problems among Hispanic children was explained by lower socioeconomic status (lower education and higher poverty level), followed by lower neighborhood safety which explained 11.3% of this gap. Similarly, lower socioeconomic status (also lower education and higher poverty level) was most relevant for explaining the higher rates of fair/poor dental health rating (24.3% of the gap), followed by the higher rates of poor maternal health which explained 9.3% of that gap and differences in insurance status which explained 9.1% of this gap. Finally, lower socioeconomic status (mostly higher poverty level) explained 55% of the lower use of preventive dental care among Hispanic children, followed by the state of residence, which explained 22% of this gap.
We found significant differences in children’s dental health by race/ethnicity. Compared to whites, Hispanic children had the poorest dental health and lowest preventive dental care use, followed by African-American children. More importantly, we were able to explain most of these disparities especially for preventive dental care use, with lower household socioeconomic status, mainly lower maternal education and higher household poverty level, among Hispanic and African-American children, generally being the single most important factor for explaining these disparities. Other relevant factors for explaining disparities in dental health include maternal health, age and marital status, although the effects of these variables are less consistent compared to socioeconomic status.
To our knowledge, this is the first study to formally decompose and quantify the extent to which several conceptually relevant social, economic, demographic and neighborhood characteristics explain racial/ethnic gaps in children’s dental health and preventive care use, especially with a nationally representative sample. The study findings are important as they reveal that most of these disparities are socioeconomically driven. They also suggest that reducing racial/ethnic gaps in child dental health requires broad and comprehensive population-based interventions beginning with improving household socioeconomic status, which may be the most effective pathway to reduce these disparities, and also enhancing neighborhood quality. Our results are consistent with previous studies highlighting the importance of socioeconomic factors such as income and education for racial/ethnic disparities in dental health 5, 11, 23, 24. The important role of socioeconomic status, mainly household poverty level and maternal education, is strongly supported theoretically which reinforces the validity of the results. Enabling characteristics such as income can substantially enhance access to preventive dental care both through increasing the ability to pay for dental care and to have better insurance, which is also relevant for explaining racial/ethnic disparities on its own. In addition, socioeconomic status and maternal education can strongly influence maternal/household knowledge and enforcement of optimal dental hygiene practices and dietary patterns. Higher unemployment, which explained part of the gap in fair/poor dental health rating between African-American and White children, may affect dental health beyond its effects on income and insurance, such as by affecting maternal psychosocial status and information gathering ability.
Demographics, maternal health, neighborhood characteristics, and geographic location also contribute on their own to racial/ethnic disparities in children’s dental health. This highlights the complexity of the pathways leading to disparities and the importance of recognizing these when considering policies and interventions to reduce health disparities. Of particular importance are the effects of maternal health, marital status and age, which vary by the outcome. Maternal health and marital status are relevant for explaining disparities in dental health, while maternal age is relevant for explaining disparities in preventive dental visits. These effects are consistent for both African-American and Hispanic children. A positive association between maternal age and child’s dental health through knowledge about child health and parenting skills that are relevant to dental health is supported in previous studies 25.
The observed effect of state of residence in explaining part of the disparity in preventive dental visits between Hispanic and white children may reflect differences between states in policies and the distribution of dental care providers and their participation in public insurance programs (Medicaid and CHIP). Further, reduced neighborhood safety and social capital, which are the most relevant neighborhood characteristics for the observed disparities, may affect dental health through restricting visits to dental providers or reducing the availability of nearby dental providers who are more likely to locate in safer neighborhoods. Previous research supports the role of neighborhood characteristics in health and health behaviors, in part through sharing relevant information for health 26. In addition, previous studies have found an association of neighborhood characteristics with dental health through neighborhood socioeconomic conditions, social capital, and availability of and access to, healthy foods 27–31.
Understanding racial/ethnic disparities in child dental health is highly relevant since poor dental health affects children’s physical and social functions and lifetime outcomes related to general health, human capital, and socioeconomic status. Dental health problems during childhood have been found to affect behavioral and cognitive functioning and to have potential long term effects on language, nutrition, systemic health, and quality of life 4, 32–35. The study highlights important pathways leading to racial/ethnic disparities and is informative for public health and population-wide interventions to improve child dental health and reduce disparities. Furthermore, the study provides a framework for future studies to further characterize such disparities, as further work is needed to fully characterize the underlying pathways and develop specific interventions. For example, the study had no information about household dental health-related behaviors, which may in part explain the observed effects of household socioeconomic status, or the unexplained gaps. Similarly, we had no data on maternal attitudes/behaviors that may explain the observed effects of maternal age and marital status. Future studies incorporating household dental health behaviors and knowledge are needed to explain the role of these factors for dental health disparities. Finally, we had no direct measures of preferences for dental health and prevention practices that may be related to cultural factors. While we were able to explain most of the disparities, cultural factors may still be relevant for the unexplained gaps and deserve further research.
In conclusion, the study finds that racial/ethnic disparities in child dental health and preventive care are largely explained by economic and social factors, but that they are complex by involving household and neighborhood contributors. Therefore, there is no single intervention or policy that can substantially reduce these disparities on its own. However, most of the pathways underlying these disparities are amenable to policy interventions. Policies aimed at reducing racial/ethnic disparities in child dental health should recognize the need for household- and neighborhood- level interventions.
Data analysis was in part supported by NIH/NIDCR grants R01 DE020895 and R03 DE022094.
Formally, the difference (C) in the dental health outcome probability (for example, probability of any dental health problem) between the majority and minority groups explained by the kth variable from K variables was estimated by averaging across the minority sample the change in the outcome probabilities when switching the value of variable k for each observation from the minority group to that of the matched majority group observation as follows:
where NA is the number of individuals in minority race/ethnicity group A (African-American or Hispanic), X includes the explanatory variables, O and A indicate values for X from the majority (white) and minority (African-American or Hispanic) race/ethnicity group, respectively, and F is the logit cumulative density function. The values for variables 1 through k−1 were assigned to the minority race/ethnicity group values, while those for variables k+1 through K were assigned to the majority group values. The total outcome difference explained by all variables is the sum of the differences explained by the individual variables (which may also be obtained by switching all variables at the same time from the minority to the majority values). Similar to the individual variable analysis, categories of variables may be evaluated for their combined contribution to the racial/ethnic disparity.
|Any dental problem||Poorly self-rated dental health||Preventive dental visits|
|White vs. African-American||White vs. Hispanic||White vs. African-American||White vs. Hispanic||White vs. African-American||White vs. Hispanic|
|Coefficient (Standard Error)|
|Demographics||Maternal age||0.0012 (0.0019)||0.0007 (0.0021)||0.0006 (0.0008)||−0.0004 (0.0007)||0.0207*** (0.0024)||0.0246*** (0.0029)|
|Mother’s married status (married)||−0.0206*** (0.0060)||−0.0055** (0.0023)||−0.0038 (0.0027)||−0.0005 (0.0007)||0.0032 (0.0064)||0.0007 (0.0026)|
|Child born in the US||−0.0007 (0.0007)||−0.0035 (0.0035)||0.0000 (0.0001)||−0.0041 (0.0026)||0.0000 (0.0007)||0.0000 (0.0035)|
|Maternal health (fair or poor)||−0.0068*** (0.0020)||−0.0085*** (0.0027)||−0.0121*** (0.0019)||−0.0147*** (0.0024)||0.0022 (0.0020)||0.0011 (0.0026)|
|Socioeconomics||Poverty level||−0.0125* (0.0065)||−0.0198** (0.0083)||−0.0098** (0.0039)||−0.0105* (0.0060)||0.0593*** (0.0066)||0.0682*** (0.0089)|
|Employment||0.0018 (0.0021)||0.0006 (0.0032)||−0.0038** (0.0018)||0.0002 (0.0021)||−0.0000 (0.0021)||−0.0051 (0.0034)|
|Maternal education||−0.0062*** (0.0020)||−0.0246*** (0.0075)||−0.0040*** (0.0014)||−0.0279*** (0.0057)||0.0057*** (0.0020)||0.0089 (0.0078)|
|Child insurance status||−0.0081 (0.0057)||−0.008 (0.0064)||−0.0082** (0.0039)||−0.0143*** (0.0046)||−0.0115* (0.0059)||0.0122* (0.0067)|
|Household demographics||Number of kids||−0.0008*** (0.0003)||−0.0045*** (0.0014)||−0.0002 (0.0002)||−0.0028* (0.0015)||−0.0008*** (0.0003)||−0.0043*** (0.0015)|
|Number of adults||−1.95e-06 (0.0023)||−0.0001 (0.0009)||0.0042** (0.0020)||−0.0010 (0.0013)||−0.0060 (0.0024)||0.0013 (0.0009)|
|Neighborhood amenities and condition||Library or bookmobile||0.0008* (0.0005)||−0.0008 (0.0009)||−0.0003 (0.0003)||−0.0014** (0.0006)||0.0006 (0.0005)||0.0008 (0.0009)|
|Litter or garbage on the street or sidewalk||0.0025 (0.0022)||0.0001 (0.0004)||0.0014 (0.0016)||0.0000 (0.0002)||0.0050** (0.0024)||0.0013*** (0.0005)|
|Poorly kept or rundown housing||−0.0019 (0.0012)||−0.0000 (0.0001)||0.0001 (0.0008)||0.0002 (0.0003)||−0.0006 (0.0013)||0.0000 (0.0002)|
|Neighborhood Perceived safety||Vandalism||−0.0064*** (0.0016)||−0.0072*** (0.0025)||−0.0017* (0.0010)||0.0005 (0.0019)||−0.0013 (0.0016)||−0.0041 (0.0027)|
|Feeling safe in the neighborhood||−0.0082** (0.0041)||−0.0095*** (0.0036)||−0.0024 (0.0028)||−0.0034 (0.0028)||−0.0009 (0.0038)||−0.0016 (0.0035)|
|Neighborhood Social Capital||People helping each other out||−0.0031 (0.0038)||−0.0045 (0.0038)||−0.0010 (0.0023)||0.0012 (0.0027)||0.0005 (0.0037)||−0.0014 (0.0040)|
|Watching out for each other’s children||0.0058** (0.0026)||0.0053** (0.0024)||−0.0016 (0.0019)||−0.0003 (0.0019)||0.0041 (0.0026)||0.0047* (0.0025)|
|Counting on people||−0.0125*** (0.0043)||−0.0082*** (0.0025)||−0.0057* (0.0029)||−0.0011 (0.0014)||0.0000 (0.0043)||0.0009 (0.0024)|
|Adults help if child got hurt or scared||0.0009 (0.0018)||0.0019 (0.0018)||0.0004 (0.0016)||−0.0014 (0.0010)||−0.0003 (0.0015)||−0.0014*** (0.0016)|
|Area fixed effects||−0.0017 (0.0028)||−0.0006 (0.0086)||0.0036 (0.0026)||−0.0097 (0.0072)||0.0013 (0.0030)||0.0288*** (0.0087)|
Contributor Statement:Dr. Wehby conceived the project and co-designed the study with Dr. Guarnizo-Herreño. Both coauthors contributed equally to statistical analysis, interpreting results, and writing the paper.
The authors adhere to the American Journal of Public Health Policy on Ethical Principles, and declare that they have no conflicting interests. The study uses publicly available data online which can be downloaded from the following CDC website: http://www.cdc.gov/nchs/slaits/nsch.htm#2007nsch This research is considered exempted by the Institutional Review Board, University of Iowa.
Carol Cristina Guarnizo-Herreño, Department of Health Management and Policy, College of Public Health, University of Iowa, Iowa City, IA, 52242, USA, Departamento de Salud Colectiva, Universidad Nacional de Colombia, Bogotá, Colombia.
George L. Wehby, Dept. of Health Management and Policy, College of Public Health, University of Iowa, 105 River Street, N248 CPHB, Iowa City, IA 52242, Phone: 319-384-3814.