National Survey on Drug Use and Health, 2011 (ICPSR 34481)

Time Period(s)

2011

Date of Collection

2011

Data Collection Notes

In the income section, which was interviewer-administered, a split-sample study had been embedded within the 2006 and 2007 surveys to compare a shorter version of the income questions with a longer set of questions that had been used in previous surveys. This shorter version was adopted for the 2008 NSDUH and will be used for future NSDUHs.

For selected variables, statistical imputation was performed following logical inference to replace missing responses. These variables are identified in the codebook as "...LOGICALLY ASSIGNED" for the logical procedure, or by the designation "IMPUTATION-REVISED" in the variable label when the statistical procedure was also performed. The names of statistically imputed variables begin with the letters "IR". For each imputation-revised variable, a corresponding imputation indicator variable indicates whether a case's value on the variable resulted from an interview response or was imputed. Missing values for some demographic variables were imputed by the unweighted hot-deck technique used in previous surveys. Beginning in 1999, imputation of missing values for most variables was accomplished using predictive mean neighborhoods (PMN), a new procedure developed specifically for this survey. Both the hot-deck and PMN imputation procedures are described in the codebook.

Although the design of the 2011 survey is similar to the design of the 1999 through 2001 surveys, there are important methodological differences since 2002 that affect the 2011 estimates. Each NSDUH respondent since 2002 has been given an incentive payment of $30. This change resulted in an improvement in the survey response rate. In addition, in 2002 and 2011 new population data from the 2000 and 2010 decennial Censuses, respectively, became available for use in NSDUH sample weighting procedures. Therefore the data from 2002 and later should not be compared with data collected in 2001 or earlier to assess changes over time.

Since 1999, the survey sample has employed a 50-state design with an independent, multistage area probability sample for each of the 50 states and the District of Columbia.

The setup and dictionary files for Stata are designed to be compatible with StataSE, Version 8. This is a large data file requiring that approximately 400 megabytes of Random Access Memory be allocated to Stata. Operations within Stata, including conversion of the ASCII data to Stata format, are likely to be slow. Analysts may wish to download subsets of data from the SAMHDA Survey Documentation and Analysis (SDA) system for use with Stata.

Prior to the 2002 survey, this series was titled National Household Surveys on Drug Abuse.

Data were collected and prepared for release by Research Triangle Institute, Research Triangle Park, North Carolina.

To protect the privacy of respondents, all variables that could be used to identify individuals have been encrypted or collapsed in the public use file. To further ensure respondent confidentiality, the data producer used data substitution and deletion of state identifiers and a subsample of records in the creation of the public use file.

Previously published estimates may not be exactly reproducible from the variables in the public use file due to the disclosure protection procedures that were implemented.

Sample

A multistage area probability sample for each of the 50 states and the District of Columbia has been used since 1999. A coordinated sample design was developed for the 2005 through 2009 NSDUHs. The 2011 NSDUH is an extension of the 5-year sample design. Although there is no overlap with the 1999-2004 samples, the coordinated design for 2005 through 2009 facilitated a 50 percent overlap in second-stage units (area segments [see below]) between each two successive years from 2005 through 2009. The 2011 NSDUH continues the 50 percent overlap by retaining half of the second-stage units from 2010. This design was intended to increase precision of estimates in year-to-year trend analyses because of the expected positive correlation resulting from the overlapping sample between successive survey years. The 2011 design allows for computation of estimates by state in all 50 states plus the District of Columbia. States may therefore be viewed as the first level of stratification as well as a reporting variable. Eight states, referred to as the large sample states, had a sample designed to yield 3,600 respondents per state for the 2011 survey. This sample size was considered adequate to support direct state estimates. The remaining 43 states (which include the District of Columbia) had a sample designed to yield 900 respondents per state in the 2011 survey. In these 43 states, adequate data were available to support reliable state estimates based on SAE methodology. Within each state, sampling strata called state sampling (SS) regions were formed. Based on a composite size measure, states were partitioned geographically into roughly equal-sized regions. In other words, regions were formed such that each area yielded, in expectation, roughly the same number of interviews during each data collection period. The eight large sample states were divided into 48 SS regions each. The remaining states were divided into 12 SS regions each. Therefore, the partitioning of the United States resulted in the formation of a total of 900 SS regions. Unlike the 1999 through 2004 surveys, the first stage of selection for the 2005 through 2011 NSDUHs was Census tracts. The first stage of selection began with the construction of an area sample frame that contained one record for each Census tract in the United States. If necessary, Census tracts were aggregated within SS regions until each tract had, at a minimum, 150 dwelling units in urban areas and 100 dwelling units in rural areas. These Census tracts served as the primary sampling units (PSUs) for the coordinated five-year sample. One area segment (one or more Census blocks) was selected within each sampled Census tract. In advance of the survey period, specially trained listers had visited each area segment and listed all addresses for housing units and eligible group quarters units in a prescribed order. Systematic sampling was used to select the allocated sample of addresses from each segment. Beginning in 2002, each respondent who completed a full interview was given a $30 cash payment as a token of appreciation for his or her time. To improve the precision of the estimates, the sample allocation process targeted five age groups: 12 to 17 years, 18 to 25 years, 26 to 34 years, 35 to 49 years, and 50 years or older. The size measures used in selecting the area segments were coordinated with the dwelling unit and person selection process so that a nearly self-weighting sample could be achieved in each of the five age groups. In 2011, an oversample was included to help in measuring and reporting on the impact that the April 2010 Deepwater Horizon oil spill had on substance use and mental health along the gulf coast. The target sample was expanded by 2,000 cases in four Gulf Coast States (Alabama, Florida, Louisiana, and Mississippi). The achieved sample size for the 2011 survey was 70,109 persons. The public use file contains 58,397 records due to a subsampling step used in the disclosure protection procedures. A key step in the data processing procedures established the minimum item response requirements in order for cases to be retained for weighting and further analysis (i.e., "usable" cases). These requirements, as well as full sampling methodology, are detailed in the codebook.

Universe

The civilian, noninstitutionalized population of the United States aged 12 and older, including residents of noninstitutional group quarters such as college dormitories, group homes, shelters, rooming houses, and civilians dwelling on military installations.

Unit(s) of Observation

individual

survey data

Mode of Data Collection

audio computer-assisted self interview (ACASI)

computer-assisted personal interview (CAPI)

computer-assisted self interview (CASI)

Response Rates

Strategies for ensuring high rates of participation resulted in a weighted screening response rate of 86.98 percent and a weighted interview response rate for the CAI of 74.38 percent. (Note that these response rates reflect the original sample, not the subsampled data file referenced in this document.)

Original Release Date

2012-11-28

Version Date

2015-11-23

Version History

2013-03-28 Variable groupings added to ddi file.

2013-06-20 Released methodological resource documentation.

2012-11-28 ICPSR data undergo a confidentiality review and are altered when necessary to limit the risk of disclosure. ICPSR also routinely creates ready-to-go data files along with setups in the major statistical software formats as well as standard codebooks to accompany the data. In addition to these procedures, ICPSR performed the following processing steps for this data collection:

• Performed consistency checks.
• Standardized missing values.
• Created online analysis version with question text.
• Checked for undocumented or out-of-range codes.

2013-02-06 Questionnaires now being released.

2015-11-23 Covers for the PDF documentation were revised.

2014-05-19

Since the release of the previous version of the 2011 Public Use Data File and Codebook a number of variables have been added. Some of the additional variables are revised versions of previous variables and replace the ones removed from the data file.

A model to predict adult mental illness was revised in the 2012 NSDUH to produce more accurate estimates. The mental illness variables included in the 2011 NSDUH data files are based on the revised 2012 model.

One new geographic variable(PDEN00), has been added to the data file to replace the previous PDEN variable. In this case, the only difference is the variable name. The variable has been renamed to indicate which census data was used in its construction.

Weight

Due to unequal selection probabilities at multiple stages of sample selection and various adjustments, such as those for nonresponse and post-stratification, the 2011 NSDUH sample design is not self-weighting. Analysts are advised to use the final sample weight when attempting to use the 2011 NSDUH data to draw inferences about the target population or any subdomains of the target population. All estimates published in SAMHSA reports (such as the results from the 2011 NSDUH) are weighted using the final analysis weight for the full sample (ANALWT). For the public use file, the corresponding final sample weight is denoted as ANALWT_C, with the "C" denoting confidentiality protection. This sample weight represents the total number of target population persons each record on the file represents. Note that the sum of ANALWT_C, over all records on the data file, represents an estimate of the total number of people in the target population.

In the 2008 NSDUH's mental health module, the adult sample was split into a sample A who received the WHODAS questions LIREMEM through LIAD68 and a sample B who received the SDS questions MHAD66a through MHAD68. In the 2011 NSDUH, however, all of the adults in the sample received the WHODAS questions. Therefore, there is no need to have a separate adult mental health weight beginning in the 2009 NSDUH because the person-level analysis weight can be used to produce the adult mental health estimates.