Public-use Data Research Articles

Health-Related Research Using Confidential U.S. Census Bureau Data

Economic studies on health-related issues have the potential to benefit all Americans. The approaches for dealing with the growth of health care costs and health insurance coverage are ever changing and information is needed on their efficacy. Research on health-related topics has been conducted for about a decade at the Census Bureau’s Center for Economic Studies and the Research Data Centers. This paper begins by describing the confidential business and demographic Census Bureau data products used in this research. The discussion continues with summaries of nearly 30 papers, including how this work has benefited the Census Bureau and its research findings. Some focus on data linkages and assessing data quality, while others address important questions in the employer, public, and individual insurance markets. This research could not have been accomplished with public-use data. The newly available data from the Agency for Healthcare Research and Quality and National Center for Health Statistics, as well as additional Census Bureau data now available in the Research Data Centers are also discussed.

Opinion on the safety of ‘Alfalfa protein concentrate’ as food

Opinion on the safety of ‘Alfalfa protein concentrate’ as food

School Recess and Group Classroom Behavior

This study examines the amount of recess that children 8 to 9 years of age receive in the United States and compares the group classroom behavior of children receiving daily recess with that of children not receiving daily recess. This is a secondary analysis of a public-use data set, the Early Childhood Longitudinal Study, Kindergarten Class of 1998-1999, third-grade data set. Children were categorized into 2 levels of recess exposure, that is, none/minimal break (<1 break of 15 minutes/day) or some recess. Some recess was further categorized into 5 levels on the basis of frequency and duration of recess. Child, parent, school, and classroom characteristics of those with and without recess were compared. The group classroom behavior was assessed by using the teacher's rating of class behavior. Complete data were available for 10301 to 11624 children 8 to 9 years of age. There were equal numbers of boys and girls (boys: 50.3%). Children exposed to none/minimal break (30%) were much more likely to be black, to be from families with lower incomes and lower levels of education, to live in large cities, to be from the Northeast or South, and to attend public school, compared with those with recess. Teacher's rating of classroom behavior scores were better for children with some recess than for those with none/minimal break. This finding was maintained in multivariate regression analysis. However, among children receiving daily recess, the teacher's rating of class behavior scores did not differ significantly according to the level of exposure. These results indicated that, among 8- to 9-year-old children, having > or =1 daily recess period of >15 minutes in length was associated with better teacher's rating of class behavior scores. This study suggests that schoolchildren in this age group should be provided with daily recess.

Recent Trends in Top Income Shares in the USA: Reconciling Estimates from March CPS and IRS Tax Return Data

Although the vast majority of US research on trends in the inequality of family income is based on public-use March Current Population Survey (CPS) data, a new wave of research based on Internal Revenue Service (IRS) tax return data reports substantially higher levels of inequality and faster growing trends. We show that these apparently inconsistent estimates can largely be reconciled once one uses internal CPS data (which better captures the top of the income distribution than public-use CPS data) and defines the income distribution in the same way. Using internal CPS data for 1967–2006, we closely match the IRS data-based estimates of top income shares reported by Piketty and Saez (2003), with the exception of the share of the top 1 percent of the distribution during 1993–2000. Our results imply that, if inequality has increased substantially since 1993, the increase is confined to income changes for those in the top 1 percent of the distribution.

A Hybrid Bayesian Network Model for Predicting Breast Cancer Prognosis

Objective: Breast cancer is one of the most common cancers affecting women. Both physicians and patients have concerned about breast cancer survivability. Many researchers have studied the breast cancer survivability applying artificial nerural network model (ANN). Usually ANN model outperformed in classification of breast cancer survivability than other models such as logistic regression, Bayesian network (BN), or decision tree models. However, physicians in the fields hesitate to use ANN model, because ANN is a black-box model, and hard to explain the classification result to patients. In this study, we proposed a hybrid model with a degree of the accuracy and interpretation by combining the ANN for accuracy and BN for interpretation. Methods: We developed an artificial neural network, a Bayesian network, and a hybrid Bayesian network model to predict breast cancer prognosis. The hybrid model combined the artificial neural network and the Bayesian network to obtain a good estimation of prognosis as well as a good explanation of the results. The National Cancer Institute's SEER program public-use data (1973-2003) were used to construct and evaluate the proposed models. Nine variables, which are clinically acceptable, were selected for input to the proposed models' nodes. A confidence value of the neural network served as an additional input node to the hybrid Bayesian network model. Ten iterations of random subsampling were performed to evaluate performance of the models. Results: The hybrid BN model achieved the highest area under the curve value of 0.935, whereas the corresponding values of the neural network and Bayesian network were 0.930 and 0.813, respectively. The neural network model achieved the highest prediction accuracy of 88.8% with a sensitivity of 93.7% and a specificity of 85.4%. The hybrid Bayesian network model achieved a prediction accuracy of 87.2% with a sensitivity of 93.3% and a specificity of 83.1%. The results of the hybrid Bayesian network model were very similar to the neural network model. Conclusion: In the experiments, the hybrid model and the ANN model outperformed the Bayesian network model. The proposed hybrid BN model for breast cancer prognosis predictin may be useful for clinicians in the medical fields, as the model provides both high degree of performance inherited from ANN and good explanation power from BN. (Journal of Korean Society of Medical Informatics 15-1, 49-57, 2009)

Transition to Adulthood: Delays and Unmet Needs Among Adolescents and Young Adults With Asthma

Scal P, Davern M, Ireland M, Park K. J Pediatr. 2008;152(4):471–475 PURPOSE OF THE STUDY. To examine the effect of the transition to adulthood on financial and nonfinancial barriers to care in youths with asthma. STUDY POPULATION. Studied were adolescents and young adults with asthma. Public-use data from the National Health Interview Survey conducted by the National Center for Health Statistics were analyzed. Data from the years 2000–2005 were pooled to provide a sample of 26 597 adolescents (12–17 years) and 19 998 young adults (18–124 years). METHODS. Subjects were classified as having delayed care because of financial barriers when during the previous 12 months they had delayed seeking medical care because of concerns about affordability. Similarly, an unmet need because of a financial barrier was identified when during the previous 12 months the respondents indicated that they had failed to receive needed medical care or prescription medication because they could not afford it. RESULTS. More young adults than adolescents encountered financial barriers that resulted in delays (18.6% vs 8%; P &amp;lt; .05) and unmet needs (26.6% vs 11.4%; P &amp;lt; .05). Delays caused by nonfinancial barriers were similar (17.3% vs 14.9%; P was not significant). CONCLUSIONS. Delays and unmet needs caused by financial reasons were significantly higher for young adults with asthma compared with adolescents with asthma. REVIEWER COMMENTS. It is crucial for everyone who treats children with asthma to recognize the potential vulnerability of these patients as they transition to adulthood. Appropriate counseling and written materials regarding health insurance might be helpful, as might providing lists of resources for free or reduced-cost care that are available in the local community.

Predictors of American elders' home stay: A secondary data analysis study

The purpose of this study was to identify the predictors of home stay for American elders. This study used a cross-sectional, descriptive, secondary data analysis design. The National Home and Hospice Care Survey 2000 (NHHCS 2000) public-use data files were used for this study. The sample included 9879 elders who were listed as either current or discharged patients from the NHHCS 2000. Based on multiple logistic regression analysis, the most predictive variable for an elder's home stay was whether the elder was currently living with family members. The overall model of 16 predictors was statistically significant in distinguishing between "home stay" and "not home stay" elders. The model correctly predicted 92.0% of the elderly participants regarding the home stay outcome. Health-care professionals could target these predictors in an attempt to develop interventions that assist elders to reside in their own home.

Pharmaceutical Innovation and U.S. Cancer Survival, 1992-2003: Evidence from Linked SEER-MEDSTAT Data

This study examines the impact of pharmaceutical innovation and other factors on the survival of U.S. cancer patients during the period 1992-2003. In particular, it investigates whether cancer survival rates increased more for those cancer sites that had the largest increases in the proportion of chemotherapy treatments that were "new" treatments. We control for other types of medical innovation, i.e. other pharmaceutical innovation, and innovation in surgical procedures, diagnostic radiology procedures, and radiation oncology procedures.Data on observed survival rates, the number of people diagnosed, mean age at diagnosis, and stage distribution are obtained from the National Cancer Institute’s SEER public-use data. Estimates of rates of innovation in chemotherapy and other treatment and diagnostic procedures are constructed from the MEDSTAT MarketScan database and other data sources. Treatment innovation indicators based on MEDSTAT data are likely to be useful, albeit noisy, indicators of the treatment innovation experienced by patients in SEER registries. This sampling error is likely to bias the coefficients on the treatment innovation measures towards zero.We compute weighted least-squares estimates of 6 versions of a survival model, based on different survival intervals and functional forms. The chemotherapy vintage coefficient is positive and significant in every model. This indicates that the cancer sites whose chemotherapy vintage (measured by the share of post-1990 treatments) increased the most during the period 1992-2003 tended to have larger increases in observed survival rates, ceteris paribus.We estimate how much lower the survival rate from all cancer sites combined would have been during 1993-2001 in the absence of post-1992 chemotherapy innovation. The estimates indicate that chemotherapy innovation accounted for 74% of the increase in the 1-year observed survival rate for all cancer sites combined during the period 1992-2001. Chemotherapy innovation accounted for smaller fractions of the increases in the 2-year and 3-year observed survival rates for all cancer sites combined during the same period: 51% and 29%, respectively.The coefficients on measures of non-pharmaceutical medical innovation (in radiation oncology, diagnostic radiology, and surgery innovation) are generally not significant. However these measures may be less reliable than the drug innovation measure. They were based upon the year in which the AMA established a new procedure code, which may be a far less meaningful indicator of innovation than the year in which the FDA first approved a drug. This topic warrants further research.

Secular Trends in the Rates of Preeclampsia, Eclampsia, and Gestational Hypertension, United States, 1987-2004

Few studies have reported on population-level incidence of or trends in the hypertensive disorders of pregnancy, and none report on data through 2004. We describe population trends in the incidence rates of preeclampsia, eclampsia, and gestational hypertension in the United States for 1987-2004. We analyzed public-use data from the National Hospital Discharge Survey (NHDS), which has been conducted by the Centers for Disease Control and Prevention, National Center for Health Statistics since 1965. We calculated crude and age-adjusted incidence rates and estimated the risk associated with available demographic variables using Cox regression modeling. Rates of preeclampsia and gestational hypertension increased significantly (by 25 and 184%, respectively) over the study period; in contrast, the rate of eclampsia decreased by 22% (nonsignificant). Women under the age of 20 were at significantly greater risk for all three outcomes. Women in the south of the country were at significantly greater risk for preeclampsia and gestational hypertension when compared to those in the Northeast. The increase in gestational hypertension may be exaggerated because of the revised clinical guidelines published in the 1990s; these same revisions would likely have reduced diagnoses of preeclampsia. Therefore, our observation of a small but consistent increase in preeclampsia is a conservative indication of a true population-level change.

The Occupational Mix Adjustment to the Medicare Hospital Wage Index: Why the Rural Impact Is Less Than Expected

Rural hospitals are heavily dependent on Medicare for their long-term financial solvency. A recent change to Medicare prospective payment system reimbursement--the occupational mix adjustment (OMA) to the wage index--has attracted a great deal of attention in rural policy circles. This paper explores variation in the OMA across and within urban and rural markets. Reasons why the effect of the OMA has been less than some rural advocates anticipated are discussed. Data were obtained from the fiscal year 2007 Final Occupational Mix Survey Data Public Use File and the fiscal year 2007 Final Rule Wage Data Public Use File. Descriptive statistics were generated to determine the need for the OMA and the potential impact of its application on hospitals located in rural markets. The average OMA for nonmetropolitan markets is greater than 1, indicating that hospitals in these markets use a less-skilled mix of labor than the national average. However, almost one third of nonmetropolitan markets had an OMA that was less than 1 and experienced a net decrease in Medicare reimbursement due to the OMA. There are several reasons why the impact of the OMA is smaller than many rural hospital administrators expected. The most important is that the adjustment happens at the market-level rather than for individual hospitals, so a small hospital's staffing mix may have almost no effect on the final payment adjustment. In rural markets, it appears that hospitals in micropolitan areas exert a large influence on the OMA.

Time Trends in Oligodendroglial and Astrocytic Tumor Incidence

Background: We hypothesized that the incidences of oligodendrogliomas, anaplastic oligodendrogliomas, and mixed gliomas have significantly increased from the early 1990s forward, while the incidences of anaplastic and grade II astrocytic tumors have significantly decreased. Methods: Data for the years 1973–2004 from the Surveillance, Epidemiology and End Results (SEER) public-use data and for 1985–2004 from six collaborating registries of the Central Brain Tumor Registry of the US (CBTRUS) were obtained. SEER*Stat was used to estimate age-adjusted incidence trends and annual percent change (APC) for selected histologies. Joinpoint regression was used to identify sharp changes in incidences occurring over time. Results: Using CBTRUS data, the incidences (per 100,000 person-years) of oligodendrogliomas (APC = 4.7), mixed gliomas (APC = 3.9) and anaplastic oligodendrogliomas (APC = 12.5) have all increased over time, while the incidences of astrocytoma not otherwise specified (APC = –8.1) and fibrillary astrocytoma (APC = –2.1) have decreased. Restricting the analyses to later years (1992–2004) using SEER data shows the incidence of oligodendrogliomas leveling off (APC = 0.5), while joinpoint analyses demonstrate a decreasing trend after 1998. Conclusions: This study has demonstrated that increases in oligodendroglial tumor incidence correspond to decreases in astrocytic tumor incidence over the same time period. Minimizing misclassification of glial tumors will be essential for accurately assessing incidence, survival, and mortality rates, as well as for identifying homogeneous subgroups for epidemiologic and treatment studies.

Measuring Labor Earnings Inequality Using Public-Use March Current Population Survey Data: The Value of Including Variances and Cell Means When Imputing Topcoded Values

Using the Census Bureau's internal March Current Population Surveys (CPS) file, we construct and make available variances and cell means for all topcoded income values in the publicuse version of these data. We then provide a procedure that allows researchers with access only to the public-use March CPS data to take advantage of this added information when imputing its topcoded income values. As an example of its value we show how our new procedure improves on existing imputation methods in the labor earnings inequality literature.

Incidence trends in primary malignant penile cancer

Objective To examine trends in the incidence of primary, malignant penile cancer in the United States. Methods and materials A total of 1,817 men with primary, malignant penile cancer diagnosed between 1973 and 2002 from the Surveillance, Epidemiology and End Results Program Public-use data were used for analysis. Incidence rates were calculated by clinical and demographic variables of interest and decade of diagnosis (1973–1982, 1983–1992, and 1993–2002) using Surveillance, Epidemiology and End Results-Stat 6.1, and trends were examined using the annual percent change statistic. Additional incidence calculations were performed to examine further racial/ethnic differences. Results The overall incidence of primary, malignant penile cancer from 1973 to 2002 was 0.69 per 100,000. Incidence decreased significantly over time: 0.84 per 100,000 in 1973–1982 to 0.69 per 100,000 in 1982–1992 to 0.58 per 100,000 in 1993–2002. Incidence increased with increasing age at diagnosis. The majority of cases had squamous cell carcinomas, graded as I or II, and originated at the glans penis. Incidence of unknown grade primary, malignant penile cancer decreased significantly over the last 30 years, as did incidence of primary site penis, not otherwise specified primary, malignant penile cancer. The incidence of regional stage disease also increased over time. From 1993 to 2002, White Hispanics had the highest incidence rates (1.01 per 100,000) followed by Alaska Native/American Indians (0.77 per 100,000) and Blacks (0.62 per 100,000). Conclusions The overall incidence of primary, malignant penile cancer in the United States has decreased, and these rates varied by race/ethnicity. Incidence rates increased with increasing age at diagnosis, and the incidence of regional stage disease increased over time, while incidence of unknown grade primary, malignant penile cancer decreased over the last 30 years.

Effects of different data‐editing methods on trends in race‐specific preterm delivery rates, United States, 1990–2002

In recent years, national vital statistics data indicate that the US preterm delivery rate has decreased among African Americans and increased among whites. These trends in preterm delivery rates may have been affected, in part, by improvements in the accuracy of gestational age reporting on birth certificates. Several data-editing methods have been developed with the intention of reducing the percentage of records with misclassified gestational age. We explored whether three data-editing methods yielded different trends in preterm delivery for years 1990-2002 among US non-Hispanic white and non-Hispanic African American singleton livebirths. Using National Center for Health Statistics (NCHS) public-use data, we assessed two published methods for editing gestational age, one by Alexander et al. and the other by Zhang and Bowes (Zhang/Bowes). We also assessed a third method that substitutes the clinical estimate (CE) of gestational age when the NCHS last menstrual period (LMP)-based estimate differs from the CE by more than 2 weeks (the LMP/CE method). The percentage of records excluded and/or reclassified by each method was calculated for years 1990, 1996 and 2002. Gestational age-specific birthweight distributions were plotted by race for each method. Preterm delivery rates were calculated and compared by method of editing over time. The percentage of records excluded or reclassified declined from 1990 to 2002 regardless of the method applied. For infants at 28-33 weeks' gestation using the NCHS edited data, birthweight distributions were bimodal, with the second (right-sided) mode showing a mean birthweight consistent with that of term infants. The second mode was less pronounced using the Alexander et al. and the Zhang/Bowes methods, and was not present when the LMP/CE method was used. From 1990 to 2002, preterm delivery rates increased for non-Hispanic whites regardless of method (range 21.3-31.3%). Preterm delivery rates increased slightly for non-Hispanic African Americans with the LMP/CE method (1.8%), and decreased with the other methods (range 6.7-10.8%). Different approaches to editing gestational age from vital records can result in variation in preterm delivery rates and trends. Uncertainty persists around the true trends in preterm delivery, especially among African Americans. Additional research is needed to identify the approach that results in the most accurate classification of gestational age.

Multiple imputation for national public‐use datasets and its possible application for gestational age in United States Natality files

Multiple imputation (MI) is a technique that can be used for handling missing data in a public-use dataset. With MI, two or more completed versions of the dataset are created, containing possibly different but reasonable replacements for the missing data. Users analyse the completed datasets separately with standard techniques and then combine the results using simple formulae in a way that allows the extra uncertainty due to missing data to be assessed. An advantage of this approach is that the resulting public-use data can be analysed by a variety of users for a variety of purposes, without each user needing to devise a method to deal with the missing data. A recent example for a large public-use dataset is the MI of the family income and personal earnings variables in the National Health Interview Survey. We propose an approach to utilise MI to handle the problems of missing gestational ages and implausible birthweight-gestational age combinations in national vital statistics datasets. This paper describes MI and gives examples of MI for public-use datasets, summarises methods that have been used for identifying implausible gestational age values on birth records, and combines these ideas by setting forth scenarios for identifying and then imputing missing and implausible gestational age values multiple times. Because missing and implausible gestational age values are not missing completely at random, using multiple imputations and, thus, incorporating both the existing relationships among the variables and the uncertainty added from the imputation, may lead to more valid inferences in some analytical studies than simply excluding birth records with inadequate data.

Comparison of Stage at Diagnosis of Cancer in Patients Who Are on Dialysis versus the General Population

Frequent medical encounters in patients with ESRD on dialysis may allow early detection of malignancies despite low rates of cancer screening in this population. It is therefore unclear whether dialysis patients are disadvantaged in terms of cancer diagnosis. This study compared stage at diagnosis of cancer in a population-based sample of patients with ESRD versus the general population. The Surveillance, Epidemiology, and End Results Medicare database was used to identify patients with ESRD and incident cancers from 1992 through 1999. Modified Poisson regression models were used to predict nonlocalized stage of cancer at diagnosis in patients with ESRD versus the general population, adjusting for demographics, cancer site, region, year of diagnosis, and comorbidity. Two general population comparisons were used: Standardized Surveillance, Epidemiology, and End Results public-use data and Medicare control subjects without ESRD matched 3:1 to patients with ESRD. A total of 1629 patients with ESRD and incident cancer were identified. Overall, the likelihood of nonlocalized stage at diagnosis was not significantly different for patients with ESRD versus the standardized Surveillance, Epidemiology, and End Results general population or matched Medicare control subjects. Stratifying by cancer site, colorectal cancers were significantly more likely to be diagnosed earlier in the ESRD group, whereas prostate cancers were significantly more likely to be diagnosed at a later stage. With the exception of prostate cancer, patients with ESRD are not more likely to present with later stage malignancies compared with the general population.

Alcohol and Hepatitis C Mortality Among Males and Females in the United States: A Life Table Analysis

Evidence from previous studies suggests that heavy alcohol use (HAU) exacerbates the rate of fibrosis progression in the liver and results in increased probability for premature death among patients with hepatitis C virus (HCV) infection. The current study uses population-based mortality data to investigate whether heavy drinking affects the age of death among individuals with HCV and, if so, whether this effect differs between men and women. A total of 7,263,163 death records in the United States between 2000 and 2002 were drawn from the Multiple Cause of Death (MCD) public-use data files compiled by the National Center for Health Statistics (NCHS). International Classification of Diseases, Tenth Revision (ICD-10) codes were used to identify the presence of HCV (B17.1 and B18.2) and HAU (as indicated by alcohol-induced medical conditions, F10 and K70) either as the underlying cause or as one of the contributing causes of death. The deaths were divided into 4 distinctive cause-of-death categories: HCV without HAU, HAU without HCV, HCV plus HAU, and all others. The mean ages of death and the cumulative probabilities of death derived from multiple-cause life table were compared across these categories. Hepatitis C virus deaths showed an excessive prevalence of HAU when compared with non-HCV deaths. Compared with deaths of HCV without HAU, the mean age of death was shortened for deaths of HCV plus HAU (from 55.1 to 50.0 years among males, and from 61.0 to 49.1 years among females). The cumulative probability of death before age 65 was much higher for the latter than the former group (0.91 vs 0.68 among males, and 0.88 vs 0.47 among females). While HCV alone showed a disproportionate effect on premature death in males, HAU presented a stronger effect in females, resulting in a "catching-up" effect that diminished the gender difference in age of HCV death. This study provides mortality-based evidence to further establish heavy alcohol consumption as one of the key risk factors contributing to premature deaths from HCV in the United States. More importantly, this study, for the first time, presents empirical evidence that alcohol consumption affects men and women differently in HCV mortality.

Religiousness, Health, and Health Behavior in Public-Use Data of the National Center for Health Statistics

Studies of religiousness and health-related variables in large, population-based cross-sectional or, preferably, longitudinal studies, which are often prohibitively expensive, are important to complement findings from the more commonly performed studies. Inadequately known among social science researchers, the national health surveys of the Centers for Disease Control and Prevention’s National Center for Health Statistics (NCHS) offer large, high-quality data sets to the public at no or nominal cost and hence offer important opportunities for research in the area of religion and health, religion and reproductive behavior, sociology of religion and psychology of religion. This report provides an overview of the data sets and a bibliography of prior research using these data, which is intended to suggest how the data of NCHS may be further exploited by researchers of religiousness and health.

Solving Problems of Disclosure Risk While Retaining Key Analytic Uses of Publicly Released Microdata

MEASURES USED TO PROTECT SUBJECTS in publicly distributed microdata files often have a significant negative impact on key analytic uses of the data. For example, it may be important to analyze subpopulations within a data file such as racial minorities, yet these subjects may present the greatest disclosure risk because their records tend to stand out or be unique. Files or records that are linkable create another type of disclosure risk-common elements between two files can be used to link files with sensitive data to externally available files that disclose identity. Examples of disclosure limitation methods used to address these types of issues include blanking out data, coarsening response categories, or withholding data altogether. However, the very detail that creates the greatest risk also provides insight into differences that are of greatest interest to analysts. Restricted-use agreements that provide unaltered versions of the data may not be available, or only selectively so. The public-use version of the data is very important because it is likely to be the only one to which most researchers, policy analysts, teaching faculty, and students will ever have access. Hence, it is the version from which much of the utility of the data is extracted and often it effectively becomes the historical record of the data collection. This underscores the importance that the disclosure review c ommittee s trikes a g ood b alance b etween protection and u tility. In this paper we d escrib e our disclosure review committee's (DRC) analysis and resulting data protection plans for two national studies and one administrative data system. Three distinct disclosure limitation methods were employed, taking key uses of the data into consideration, to protect respondents while still providing statistically accurate and highly useful public-use data. The techniques include data swapping, microaggregation, and suppression of detailed geographic data. We describe the characteristics of the data sets that led to the selection of these methods, provide measures of the statistical impact, and give details of their implementations so that others may also utilize them. We briefly discuss the composition of our DRC, highlighting what we believe to be the important disciplines and experience represented by the group.

Evidence for a hallucinogen dependence syndrome developing soon after onset of hallucinogen use during adolescence

This study uses latent class methods and multiple regression to shed light on hypothesized hallucinogen dependence syndromes experienced by young people who have recently initiated hallucinogen use. It explores possible variation in risk. The study sample, identified within public-use data files of the 1999 National Household Survey on Drug Abuse (NHSDA), consists of 1186 recent-onset hallucinogen users, defined as having initiated hallucinogen use within 24 months of assessment (median elapsed time since onset of use -12 to 13 months). The recent-onset users in this sample were age 12 to 21 at the time of assessment and were between the ages of 10 and 21 at the time of their first hallucinogen use. The NHSDA included items to assess seven clinical features often associated with hallucinogen dependence, which were used in latent class modelling. Latent class analysis, in conjunction with prior theory, supports a three-class solution, with 2% of recent-onset users in a class that resembles a hallucinogen dependence syndrome, whereas 88% expressed few or no clinical features of dependence. The remaining 10% may reflect users who are at risk for dependence or in an early stage of dependence. Results from latent class regressions indicate that susceptibility to rapid transition from first hallucinogen use to onset of this hallucinogen dependence syndrome might be influenced by hallucinogenic compounds taken (for example, estimated relative risk, RR = 2.4, 95% CI = 1.6, 7.6 for users of MDMA versus users of LSD). Excess risk of rapid transition did not appear to depend upon age, sex, or race/ethnicity.