[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
[Skip to Content Landing]
December 1, 2008

Determinants of Health Insurance Status for Children of Latino Immigrant and Other US Farm Workers: Findings From the National Agricultural Workers Survey

Author Affiliations

Author Affiliations: Department of Pediatrics, University of Texas Medical Branch–Austin and Dell Children's Medical Center of Central Texas, Austin (Dr Rodríguez); RAND Corporation, Santa Monica, California (Drs Elliott and Schuster and Ms Suttorp); Department of Pediatrics, David Geffen School of Medicine, University of California, Los Angeles (Ms Vestal); and Department of Medicine, Children's Hospital Boston and Harvard Medical School, Boston, Massachusetts (Dr Schuster).

Arch Pediatr Adolesc Med. 2008;162(12):1175-1180. doi:10.1001/archpedi.162.12.1175

Objective  To characterize the health insurance status of farm workers' children, an understudied topic.

Design  A population-based multistage survey.

Setting  Employer-based study conducted in the continental United States from 2000 to 2002.

Participants  A total of 3136 parents with children younger than 18 years and no children residing outside of the United States who participated in the US Department of Labor's National Agricultural Workers Survey, which is administered to a national probability sample of US farm workers.

Outcome Measures  Children’s parent-reported health insurance status.

Results  Of the farm workers who were parents, 87% were Latino, 81% were foreign born, 15% were migrant workers, 55% had less than a sixth-grade education, and 68% reported little to no English language proficiency. Thirty-two percent of all farm-worker parents, including 45% of migrant-worker parents, reported that their children were uninsured. In multivariate analyses, older parental age (odds ratio [OR] for parents aged 30-39 years, 1.71; 95% confidence interval [CI], 1.16-2.50; OR for parents aged ≥40 years, 3.07; 95% CI, 1.99-4.74), less parental education (OR, 1.52; 95% CI, 1.09-2.10), less time in the United States (OR, 0.68; 95% CI, 0.56-0.91 per 10 years in the United States), being a migrant worker (OR, 1.96; 95% CI, 1.31-2.93), and living in the Southeast (OR, 3.17; 95% CI, 2.00-5.02) or Southwest (OR, 3.91; 95% CI, 2.32-6.57) were significantly associated with having uninsured children.

Conclusions  Farm workers' children were uninsured at roughly 3 times the rate of all other children and almost twice the rate of those at or near the federal poverty level. Programs aimed at extending insurance coverage for children should consider the unique social barriers that characterize this vulnerable population of US children. Moreover, there is significant regional variation that may reflect varying levels of insurance resources and eligibility from state to state.

Access to and use of pediatric health services are suboptimal for Latino children. Children of farm workers, most of whom are Latino,1 are an especially vulnerable subset of the Latino pediatric population, with a poor overall health profile that includes high rates of physical, mental, and oral health conditions.2-6 Migrant children (those who move throughout the United States with their farm-worker parents) of Mexican American background are 2 to 3 times more likely to be rated in poor or fair health (compared with good or excellent health) by primary caregivers as nonmigrant Mexican American children.7 Many environmental hazards also affect farm workers' children, including high levels of pesticide exposure.8,9 Older children of farm workers often engage in agricultural work themselves,10 and agricultural work is among the most dangerous of occupations.11 Furthermore, a disproportionately high number of occupational injuries and fatalities occur among young farm workers.12 These health issues make access to primary health services an important priority for this pediatric population.

Health insurance improves children's access to and use of health care services,13-17 making children's health insurance an important proxy for children's health care access. Many barriers to health care services exist for children of low-income parents,13 yet there is little information available regarding farm workers' children's access to these services. This issue merits attention, especially in light of a report of the Latino Consortium, an expert panel sponsored by the American Academy of Pediatrics that asserted “an urgent need to more adequately address the health care . . . of Latino children.”18

In this study, we explored determinants of health insurance status of children of farm workers to further our understanding of health care access in this population. We applied the Behavioral Model of Health Services Use, a well-established conceptual model for health care access,19 to the farm-worker population to identify potential determinants of access in this population. We considered the effects of demographics as well as other social and enabling characteristics on their children's health insurance status.


We analyzed data from the National Agricultural Workers' Survey (NAWS), a randomly sampled, employer-based, multistage survey of field workers active in crop agriculture in the continental United States. Commissioned by the US Department of Labor, it remains the only nationally representative information source on the demographics and working and living conditions of US farm workers. Workers were sampled within employers, who were sampled within Farm Labor Areas in 12 geographic strata. The NAWS incorporates several items that measure occupational safety and health as well as general health status and access to health care services for farm workers and their families. The NAWS collected information from 10 058 farm workers from 2000 to 2002, randomly sampling farm workers in 3 cycles each year in consideration of the seasonality of farm labor because many farm workers migrate throughout the year to follow the crop harvest. The NAWS located and surveyed workers at their work sites, thus helping to avoid the problem of undercounting a population that has historically been difficult to reach.20 Response rates were 75% at the employer level and 90% at the worker level for a cumulative response rate of 68%; additional details on the sampling and weighting approach are available elsewhere.21

For this analysis, we examined farm workers' reports of their children's health insurance status using NAWS data from 2000 to 2002. Accordingly, the sample was limited to those respondents who reported being parents. In this 3-year period, 3253 of 10 058 respondents (32%) were identified as parents of at least 1 child younger than 18 years. Of these, 46 (1%) were excluded because they reported having at least 1 child living outside the United States (the outcome variable could not account for children living in different countries). Of the remaining sample of 3207 respondents, we omitted an additional 69 (2%) because they did not provide a yes or no answer to our primary outcome measure, including 12 who did not know their children's health insurance status. We also omitted 2 cases (0.6%) because they were missing sample weights, yielding a sample size of 3136 farm-worker–parent respondents.

Our study protocol received a claim of exemption from the UCLA institutional review board.

Theoretical model

The conceptual framework for this analysis is based on the Behavioral Model of Health Services Use.19 The model emphasizes the role of contextual and individual determinants as well as the influence of health behaviors and outcomes on health services access and use. Contextual determinants include factors such as health insurance eligibility criteria, population health indices, and environmental factors affecting a community. Individual determinants include factors such as demographics, personal beliefs, individual financial considerations, and perceived need. The model has been further expanded for vulnerable populations because of the belief that factors that contribute to the vulnerability of a given population also affect its corresponding health care access and use.22

Dependent variable

Our dependent variable is whether the respondents' children have health insurance (yes or no), as reported by the farm-worker parent who participated in the survey in response to the question, “Do your children have health insurance?”.

Independent variables

Independent variables were 6 demographic and social characteristics: sex, age (continuous, divided into groups younger than 30 years, 30-39 years, and 40 years or older), number of children (highest category is 3 or more children, scored linearly), educational attainment (highest grade, divided into 2 groups who attended sixth grade or lower or seventh grade or higher), marital status (unmarried vs married or living with partner), and Latino ethnicity (Mexican American, Mexican, Chicano, Puerto Rican, other Hispanic, or not Hispanic or Latino). Our conceptual model suggested 8 additional independent variables: born in the United States or Puerto Rico (yes or no), migrant status (yes or no), poverty (yes or no, as defined by federal guidelines for family poverty status), ability to speak English (not at all, a little, somewhat, or well, and operationalized as not at all, a little, somewhat, or well, scored linearly), self-reported literacy level (ability to read English, using same categories as previous variable and operationalized as not at all or a little, somewhat, or well, scored linearly), current immigration status (authorized or otherwise), region of the United States residing in at the time of the interview (East, Southeast, Midwest, Southwest, Northwest, or California), and years in the United States (continuous, highest category is 30 or more years). Because there was no difference in the children's insurance status when we compared US-born respondents with foreign-born respondents who had been in the United States for 30 or more years, we capped the variable at 30 or more years and included US-born respondents in the 30 or more years category.

Missing values

Five predictors had some missing values, but at rates of less than 1% (marital status, education, foreign birth, English proficiency, and migrant status); these were imputed as the modal category. Four dichotomous predictors were missing for 1% or more of cases, including family poverty status (6%), Latino ethnicity (1%), and authorized status (1%); missing categorical indicators were created for each of these variables. One ordinal variable, literacy, had 449 missing values (14%); these were singly imputed via regression from all other predictor variables, including a draw from the residual variance.

Statistical analysis

All logistic regression analyses employed NAWS person-level weights to account for the sampling design. However, in the public-use data set, a single composite person-level weight is calculated for each observation and was set as the probability weight for all logistic regression analyses. We used Stata version 9 (Statacorp, College Station, Texas) for all analyses.

We conducted bivariate logistic regressions of children's health insurance for each candidate predictor variable to assess its effect. We then performed a multivariate logistic regression on child insurance status. All bivariate predictors were included in the multivariate model. Both the bivariate and multivariate analyses used the adult farm-worker respondent as the unit of analysis. To illustrate the multivariate logistic regression, we also present the change in probability of insurance associated with a unit change in each independent variable.23 This approach is sometimes also referred to as covariate-adjusted proportions or recycled predictions. Recycled predictions are based on a specific logistic regression and illustrate the effect of each given variable in a logistic regression model by calculating the differences in probabilities of the outcome (here uninsured status) implied by the model at several levels of that independent variable, while holding all other covariates at their naturally occurring values, as illustrated below. Unlike the odds ratio form of the model, covariate-adjusted proportions illustrate the population-level policy effect associated with a given change in an independent variable.


Of the analysis sample, 68% were men, 87% were Latino, 81% were foreign-born, and 15% were migrant workers. Fifty-five percent had a sixth-grade education or less and 68% reported “a little” or “none at all” regarding their ability to speak English (Table 1). Sixty-six percent of respondents were located in California, the Southwest, or the Northwest, and respondents with unauthorized legal status accounted for 31% of the sample. Of all parents, 32% reported that their children were uninsured. Nearly half of migrant parents (45%) reported that their children were uninsured compared with 30% of nonmigrant parents (data not shown).

Table 1. 
Characteristics of Farm-Worker–Parent Respondents, National Agricultural Workers Survey, 2000 to 2002
Characteristics of Farm-Worker–Parent Respondents, National Agricultural Workers Survey, 2000 to 2002

Bivariate analysis for each candidate predictor and children's insurance status revealed statistically significant associations for 9 of the 14 independent variables (Table 2). Older parental age, Latino ethnicity, less than a sixth-grade education, meeting federal poverty level criteria, foreign birth, being a migrant worker, less time spent living in the United States, less ability to speak English, and residence in the Southeast or Southwest were significantly associated with children being uninsured. Insignificant bivariate predictors included parental sex, marital status, number of household children, immigration status, and literacy level.

Table 2. 
Predictors of Uninsured Status for Children of US Farm Workers
Predictors of Uninsured Status for Children of US Farm Workers

Multivariate analyses appear in Table 2. Parents aged 30 years or older were more likely to have uninsured children than parents younger than 30 years, as were parents with less than a sixth-grade education (compared with parents with an education level of seventh grade or higher) and parents who were migrant workers. Parents who had spent 5 years or less in the United States were more likely than those who had been in the United States for 6 years or more to have uninsured children. Residence in the Southeast or Southwest regions retained a significant association with children being uninsured in the multivariate model. The association between Latino ethnicity and uninsured children approached significance in the multivariate model (odds ratio, 1.96; 95% confidence interval, 1.00-3.84).

The multivariate regression is also displayed in the form of covariate-adjusted probabilities of children being uninsured associated with each independent variable (Table 2). In each instance, the final column of this table illustrates the change in prevalence of lack of insurance associated with a 1-unit increase in the corresponding independent variable or for the given category of the independent variable relative to the reference category, holding all other variables constant. For instance, parents aged 30 to 39 years or 40 years or older were, respectively, 9% or 21% more likely to have uninsured children compared with parents aged 30 years or younger. Parents with a sixth-grade education or less were 8% more likely to have uninsured children compared with those with at least a seventh-grade education, and migrant workers were 14% more likely than nonmigrant workers to have uninsured children. Finally, parents in the Southeast or Southwest were, respectively, 24% or 29% more likely than parents in California to have uninsured children.


Our findings highlight the particular vulnerability of US farm workers' children regarding health insurance coverage. Roughly one-third of farm workers' children lacked health insurance, which is 3 times the rate of all US children and approximately 1½ times the rate of both children in poverty and Latino children nationally.24

The sociodemographic factors significantly associated with children's health insurance status in this analysis were older parental age, lower educational attainment, migrant status, less time in the United States, and residence in particular geographic regions of the United States. Older parental age, which may serve as a proxy for having older children, was associated with greater probability of children being uninsured. This corresponds to national data on health insurance; a recent study found that the proportion of uninsured children tends to increase from the younger childhood years to adolescence and the proportion of children covered by Medicaid or public insurance sources steadily decreases as children get older.25 Lack of insurance for adolescents in particular negatively affects various health-access measures such as having a usual source of care, unmet medical needs, and a medical visit within the past year.26 These access problems may have particular relevance for farm workers' adolescent children, who are often engaged in agricultural labor and therefore vulnerable to both environmental and occupational health risks.10,27

Regarding parent education, the agricultural worker population had low educational attainment—more than half received a sixth-grade education or less. Low educational attainment was also associated with greater probability of uninsured children, consistent with previous work.28

Increased length of time in the United States was also associated with greater probability of children having health insurance, with each additional decade of US residency through 30 years corresponding to a 48% increase in the odds of children being insured. Because length of time in a different country is an important part of acculturation, this finding may in part reflect the influence of acculturation on the ability to obtain health insurance for one's family. However, other elements of acculturation such as English language proficiency and legal status were not significantly associated with children's health insurance status in the multivariate model. It is also notable that the association between Latino ethnicity and uninsured children approached statistical significance (odds ratio, 1.96; confidence interval, 1.00-3.84). Thus, ethnicity alone may affect health insurance status for farm workers' children, which is consistent with national racial/ethnic disparities in children's health insurance coverage.24 Alternatively, Latino ethnicity may correspond to unmeasured factors that influence children's health insurance status.

Parents' legal status did not independently influence children's health insurance status. Though parental illegal status might be considered a significant barrier to obtaining health insurance for one's children, our findings suggest that any associated fears need not necessarily prevent children from being enrolled in health insurance programs if other barriers to enrollment can be overcome.

Status as a migrant farm-worker parent was also independently associated with decreased odds of children being insured. This finding may reflect the unique obstacles for migrant farm workers in obtaining health insurance for their children. Though most migrant farm workers would presumably meet income guidelines for public health insurance programs such as Medicaid or the State Children's Health Insurance Program (assuming children are legal residents), many migrants move across state lines from season to season, thus making it more difficult for their eligible children to benefit from state-based programs. For this reason, farm-worker advocates have encouraged policies of health insurance portability and reciprocity of Medicaid benefits across states, though only a few initiatives have been attempted.29 Our findings also reveal significant geographic disparities in health insurance coverage for the children of farm workers, even after controlling for a variety of sociodemographic factors.

Our study has several limitations. First, given that our data are observational, we are not able to draw causal inferences between our predictor variables and children's health insurance status. Furthermore, we may not have a precise estimate of children's insurance rates in this population. This may especially be true for parents who have children with different health insurance statuses, because they are not able to provide different responses for multiple children. Parents are instead required to indicate that their children either have or lack insurance. Finally, a significant limitation is the data set's lack of information on individual children's health. Health status is an important determinant of insurance coverage in our theoretical model, especially for children with chronic or special health care needs who require higher-than-average health services and may qualify for special insurance programs funded by the state and federal government. Thus, our analysis is unable to account for the effect of health status (a proxy for the need for health services) on children's health insurance coverage.

These findings have important policy implications. They suggest that the low parental education among many farm workers as well as more recent immigration—which may in part reflect acculturation—negatively affect their children's health insurance status. These social disadvantages may warrant increased efforts to enroll and retain eligible children in health insurance programs. Outreach efforts would need to consider other barriers that impede insurance enrollment and retention, such as the complexity of applications, language barriers, the inaccessibility of enrollment sites in rural areas, and parents' fear of using services or misunderstanding of eligibility guidelines.13,30,31 Furthermore, our findings underscore a particular risk of lack of insurance among children of migrant farm workers. Thus, policies advocating for insurance portability and reciprocity of benefits across states may enhance children's insurance coverage for those families who move around the country. Finally, our findings not only indicate suboptimal insurance rates for farm workers' children nationwide, but also highlight significant geographic disparities in children's insurance, even after controlling for the variety of sociodemographic factors tested in our model. These findings mirror recent national statistics on children's health insurance that show that all but 1 of the states exceeding the national rate of uninsurance among children were in the Southern and Western regions; these geographic disparities likely reflect regional variation in employment, family income, and eligibility guidelines for Medicaid and the State Children's Health Insurance Program.25 Efforts to address disparities in health insurance for farm workers' children may lead to better access to health care and better health.

Correspondence: Roberto L. Rodríguez, MD, MPH, University of Texas Medical Branch, Austin-Pediatrics & Dell Children's Medical Center of Central Texas, 4900 Mueller Blvd, Austin, TX 78723 (rlrodriguez1@seton.org).

Accepted for Publication: March 22, 2008.

Author Contributions: Dr Rodríguez and Ms Suttorp had direct and full access to data used in this study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Rodríguez, Elliott, and Schuster. Acquisition of data: Suttorp. Analysis and interpretation of data: Rodríguez, Elliott, Vestal, Suttorp, and Schuster. Drafting of the manuscript: Rodríguez. Critical revision of the manuscript for important intellectual content: Elliott, Vestal, Suttorp, and Schuster. Statistical analysis: Rodríguez, Elliott, and Suttorp. Obtained funding: Rodríguez and Schuster. Administrative, technical, and material support: Vestal. Study supervision: Schuster.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the Robert Wood Johnson Foundation through the Robert Wood Johnson Clinical Scholars Program (Dr Rodríguez) and by grant U48/DP000056 from the Centers for Disease Control and Prevention.

Role of the Sponsor: The funding bodies had no role in the design and conduct of the study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.

Acknowledgment: We thank Kate Sommers-Dawes, BA, for assistance with the manuscript.

Mines  RGabbard  SSteirman  A A profile of US farm workers: demographics, household composition, income, and use of services: Office of Program Economics Research report #6  US Department of Labor Web site. http://www.dol.gov/asp/programs/agworker/report/main.htm. Published April 1997. Accessed January 4, 2008Google Scholar
Eshleman  MJDavidhizar  R Life in migrant camps for children: a hazard to health.  J Cult Divers 1997;4 (1) 13- 17PubMedGoogle Scholar
Slesinger  DP Health status and needs of migrant farm workers in the United States: a literature review.  J Rural Health 1992;8 (3) 227- 234PubMedGoogle ScholarCrossref
Nurko  CAponte-Merced  LBradley  ELFox  L Dental caries prevalence and dental health care of Mexican-American workers' children.  ASDC J Dent Child 1998;65 (1) 65- 72PubMedGoogle Scholar
Martin  SLKupersmidt  JBHarter  KS Children of farm laborers: utilization of services for mental health problems.  Community Ment Health J 1996;32 (4) 327- 340PubMedGoogle ScholarCrossref
Kupersmidt  JBMartin  SL Mental health problems of children of migrant and seasonal farm workers: a pilot study.  J Am Acad Child Adolesc Psychiatry 1997;36 (2) 224- 232PubMedGoogle ScholarCrossref
Dever  G Migrant Health Status: Profile of a Population With Complex Health Problems.  Austin, TX National Migrant Resource Program1991;
Curl  CLFenske  RAKissel  JC  et al.  Evaluation of take-home organophosphorus pesticide exposure among agricultural workers and their children.  Environ Health Perspect 2002;110 (12) A787- A792PubMedGoogle ScholarCrossref
Koch  DLu  CFisker-Andersen  JJolley  LFenske  RA Temporal association of children's pesticide exposure and agricultural spraying: report of a longitudinal biological monitoring study.  Environ Health Perspect 2002;110 (8) 829- 833PubMedGoogle ScholarCrossref
McCurdy  SASamuels  SJCarroll  DJBeaumont  JJMorrin  LA Injury risks in children of California migrant Hispanic farm worker families.  Am J Ind Med 2002;42 (2) 124- 133PubMedGoogle ScholarCrossref
National Safety Council, Injury Facts: 2000 Edition.  Itasca, IL National Safety Council2000;
Castillo  DNLanden  DDLayne  LA Occupational injury deaths of 16- and 17-year-olds in the United States.  Am J Public Health 1994;84 (4) 646- 649PubMedGoogle ScholarCrossref
Guendelman  SAngulo  VOman  D Access to health care for children and adolescents in working poor families: recent findings from California.  Med Care 2005;43 (1) 68- 78PubMedGoogle Scholar
Kempe  ABeaty  BLCrane  LA  et al.  Changes in access, utilization, and quality of care after enrollment into a state child health insurance plan.  Pediatrics 2005;115 (2) 364- 371PubMedGoogle ScholarCrossref
Zlotnick  CSoman  LA The impact of insurance lapse among low-income children.  J Urban Health 2004;81 (4) 568- 583PubMedGoogle ScholarCrossref
Committee on Children, America's Children: Health Insurance and Access to Care.  Washington, DC Institute of Medicine and National Research Council1998;
Bermudez  DBaker  LC The relationship between SCHIP enrollment and hospitalizations for ambulatory care sensitive conditions in California.  J Health Care Poor Underserved 2005;16 (1) 96- 110PubMedGoogle ScholarCrossref
Flores  GFuentes-Afflick  EBarbot  O  et al.  The health of Latino children: urgent priorities, unanswered questions, and a research agenda [published correction appears in JAMA. 2003;290(6):756].  JAMA 2002;288 (1) 82- 90PubMedGoogle ScholarCrossref
Andersen  RM Revisiting the behavioral model and access to medical care: does it matter?  J Health Soc Behav 1995;36 (1) 1- 10PubMedGoogle ScholarCrossref
 National agricultural workers survey.  US Department of Labor Web site. http://www.dol.gov/asp/programs/agworker/naws.htm. Accessed January 4, 2008Google Scholar
 Statistical methods of the National Agricultural Workers Survey.  US Department of Labor Web site. http://www.dol.gov/asp/programs/agworker/statmethods.htm. Accessed January 4, 2008Google Scholar
Gelberg  LAndersen  RMLeake  BD The Behavioral Model for Vulnerable Populations: application to medical care use and outcomes for homeless people.  Health Serv Res 2000;34 (6) 1273- 1302PubMedGoogle Scholar
Graubard  BIKorn  EL Predictive margins with survey data.  Biometrics 1999;55 (2) 652- 659PubMedGoogle ScholarCrossref
DeNavas-Witt  CProctor  BDLee  CH Income, poverty, and health insurance coverage in the United States: 2005.  Washington, DC US Government Printing Office2006;US Census Bureau Current Population Reports P60-231
Rosenbaum  SShin  P Migrant and seasonal farmworkers: health insurance coverage and access to care. http://www.kff.org/uninsured/upload/Migrant-and-Seasonal-Farmworkers-Health-Insurance-Coverage-and-Access-to-Care-Report.pdf. Accessed January 4, 2008
Newacheck  PWBrindis  CDCart  CUMarchi  KIrwin  CE Adolescent health insurance coverage: recent changes and access to care.  Pediatrics 1999;104 (2 pt 1) 195- 202PubMedGoogle ScholarCrossref
Carroll  DSamardick  RMBernard  SGabbard  SHernandez  T Findings From the National Agricultural Workers Survey (NAWS) 2001-2002: a Demographic and Employment Profile of United States Farm Workers: Research Report No. 9.  Washington, DC US Dept of Labor2005;
Granados  GPuvvula  JBerman  NDowling  PT Health care for Latino children: impact of child and parental birthplace on insurance status and access to health services.  Am J Public Health 2001;91 (11) 1806- 1807PubMedGoogle ScholarCrossref
Arendale  E Medicaid and the State Children's Health Insurance Program.  Buda, TX National Center for Farmworker Health Inc2001;
Rivers  PAPatino  FG Barriers to health care access for Latino immigrants in the USA.  Int J Soc Econ 2006;33 (3) 207- 220Google ScholarCrossref
Eberhardt  MIngram  DMakuc  D  et al.  Urban and Rural Health Chartbook: Health, United States, 2001.  Hyattsville, MD National Center for Health Statistics2001;