Go to the Table Of Contents

Appendix D:
Other Sources of State-Level Data

A variety of surveys and data systems other than the National Survey on Drug Use and Health (NSDUH) collect data on substance use and mental health problems. It is useful to consider the results of these other studies when discussing NSDUH data. This appendix briefly describes one of these data systems that publish State estimates and presents selected comparisons with NSDUH results. The State-level survey that collects data on substance use and that is discussed in this appendix is the Behavioral Risk Factor Surveillance System (BRFSS), sponsored by the Centers for Disease Control and Prevention (CDC). Another CDC data system that provides State-level substance use estimates for most but not all States is the Youth Risk Behavior Survey (YRBS). Differences between the YRBS and NSDUH designs, and age groups used in NSDUH small area estimates, imply that comparisons of prevalence rates are not straightforward. However, ignoring these differences and examining estimates at a national level, the YRBS has generally shown to have higher prevalence rates but similar long-term trends compared with NSDUH (OAS, 2008a). For further details about the YRBS, see the CDC website at http://www.cdc.gov/HealthyYouth/yrbs/index.htm.

When considering the information presented here, it is important to understand the methodological differences between these surveys and the impact that these differences could have on estimates of substance use. Several studies have compared NSDUH estimates with estimates from other studies and have evaluated how differences may have been affected by differences in survey methodology (Brener et al., 2006; Gfroerer, Wright, & Kopstein, 1997; Grucza, Abbacchi, Przybeck, & Gfroerer, 2007; Hennessy & Ginsberg, 2001; Miller et al., 2004). These comparisons suggest that the goals and approaches of surveys are often different, making comparisons between them difficult. Some methodological differences that have been identified as affecting comparisons include populations covered, sampling methods, mode of data collection, survey setting, questionnaires, and estimation methods.

BRFSS is an annual, State-based telephone survey of the civilian, noninstitutionalized adult population aged 18 or older and is sponsored by the CDC. In 2007, BRFSS collected data from all 50 States, the District of Columbia, Puerto Rico, the U.S. Virgin Islands, and Guam using a computer-assisted telephone interviewing (CATI) design. BRFSS collects information on access to health care, health status indicators, health risk behaviors (including cigarette and alcohol use), and the use of clinical preventive services. More than 350,000 adults are interviewed each year. State estimates are presented on a yearly basis. BRFSS data are weighted based on the probability of selection of a telephone number, the number of adults in a household, and the number of telephones in a household. A final poststratification adjustment is made for nonresponse and noncoverage of households without telephones. The BRFSS State prevalence rates and confidence intervals presented in this report (in Table D.1 at the end of this appendix) are weighted design-based estimates (i.e., each respondent is weighted, and the survey design is accounted for in the estimates) from the 2007 survey. For more details about BRFSS, see the CDC website at http://www.cdc.gov/brfss.

There are three measures for which State estimates are produced for both BRFSS and NSDUH: past month alcohol use, past month binge alcohol use, and past month cigarette use. Here, the findings of the 2007 BRFSS State estimates and the combined 2006-2007 NSDUH State estimates for past month alcohol use are presented. In Table D.1, the 2007 BRFSS State estimates for past month alcohol use for adults aged 18 or older are shown alongside the pooled 2006-2007 NSDUH small area estimates for the same age group (by combining the 18 to 25 and 26 or older age groups). Table D.1 also includes p values that indicate whether the BRFSS and NSDUH estimates are significantly different from each other for a given State using an exact test described here. The method for comparing the two estimates is similar to what was described in Section A.12 of Appendix A. Here, the null hypothesis of no difference is tested, i.e., πb = πn (where πb is a State-specific BRFSS prevalence rate and πn is a State-specific NSDUH prevalence rate) or equivalently that the logs-odds ratio is zero, i.e., lor = 0, where lor is defined as The log-odds ratio, lor, is defined as the natural logarithm of the ratio of two quantities. The numerator of the ratio is Pi sub b, divided by 1 minus Pi sub b. The denominator of the ratio is Pi sub n, divided by 1 minus Pi sub n,
where ln denotes the natural logarithm. An estimate of lor is given by The estimate of the log-odds ratio, lor hat, is defined as the natural logarithm of the ratio of two quantities. The numerator of the ratio is p sub b, divided by 1 minus p sub b. The denominator of the ratio is p sub n, divided by 1 minus p sub n, where pb and pn are the

2007 BRFSS State estimates and 2006-2007 NSDUH State estimates, respectively (as given in Table D.1). To compute the variance of estimate of the log-odds ratio, lor hat, i.e., variance of the estimate of the log-odds ratio, lor hat, let Theta sub b hat is defined as the ratio of p sub b and 1 minus p sub b and Theta sub n hat is defined as the ratio of p sub n and 1 minus p sub n,

then Variance v of the estimate of the log-odds ratio, lor hat, is a function of three quantities: q1, q2, and q3. It is expressed as the sum of q1 and q2 minus q3. Quantity q1 is the variance of the natural logarithm of Theta sub b hat, quantity q2 is the variance of the natural logarithm of Theta sub n hat, and quantity q3 is 2 times the covariance between the natural logarithm of Theta sub b hat and the natural logarithm of Theta sub n hat.. The covariance term can be assumed to be zero because the BRFSS and NSDUH samples are independent.

The quantity variance of the natural logarithm of Theta sub n hat can be obtained by using the 95 percent PIs in Table D.1. For this purpose, let (lowern,uppern) denote the 95 percent PI for a given State-s:

Equation D-9     D

where U sub n is the natural logarithm of upper sub n, divided by 1 minus upper sub n, and L sub n is the natural logarithm of lower sub n, divided by 1 minus lower sub n.


The quantity variance of the natural logarithm of Theta sub b can be obtained by using the 95 percent confidence intervals (CIs) in Table D.1. For this purpose, let (lowerb,upperb) denote the 95 percent CI for a given States-s. Using the first-order Taylor series approximation, the variance can be calculated as follows:

Equation D-11     D


The p value for testing the null hypothesis of no difference (lor = 0) is given by p value = 2*P [Z ≥abs(z)], where Z is a standard normal random variate, Quantity z is the estimate of the log-odds ratio, lor hat, divided by the square root of the sum of the variance of the natural logarithm of Theta sub b hat and the variance of the natural logarithm of Theta sub n hat, and abs(z) denotes the absolute value of z.


As can be seen in Table D.1, for past month alcohol use, the NSDUH and the BRFSS estimates for most States are similar (i.e., at the 5 percent level of significance, only 13 of 51 States are significantly different). These estimates are also highly correlated. The correlation between the BRFSS and NSDUH estimates of past month alcohol use was 0.95.

Figures D.1 and D.2 were created by using the BRFSS State estimates and the NSDUH State estimates and categorizing the States into five quintiles similar to the process described in Section 1.2 of Chapter 1. As can be seen in Figures D.1 and D.2, the 10 States with the highest rates of alcohol use (States shown in red) were the same in the two surveys: Colorado, Connecticut, District of Columbia, Massachusetts, Minnesota, New Hampshire, North Dakota, Rhode Island, Vermont, and Wisconsin. The lowest rate of past month alcohol use for both surveys occurred in Utah (see Table D.1).

Note that although the BRFSS estimates are design based, the NSDUH small area estimates are model based. Also, the NSDUH small area estimates are based on the pooled 2006 and 2007 NSDUHs (2 years of data), whereas the BRFSS estimates presented here are based on the 2007 BRFSS survey (1 year of data). Although the BRFSS estimates are only based on 1 year of data, the BRFSS sample sizes for a given State are in general much larger than the sample sizes for the NSDUH over 2 years. The eight "large" States (California, Florida, Illinois, Michigan, New York, Ohio, Pennsylvania, and Texas) have a sample size of approximately 7,200 respondents for the 2006-2007 combined NSDUH data. For the 2007 BRFSS, three States (Illinois, California, and New York) had smaller sample sizes as compared with the NSDUH data, but the other five States had larger sample sizes. Overall, the BRFSS sample sizes for the eight "large" States varied from a low of 5,236 respondents in Illinois to a high of 39,549 respondents in Florida, with a median sample size of 9,367. For the remaining small sample States, the NSDUH sample size for the combined 2006-2007 data were approximately 1,800 respondents for each State. The BRFSS sample sizes for the small sample States were much larger. The BRFSS sample sizes for the small sample States varied from a low of 2,552 respondents in Alaska to a high of 25,881 respondents in Washington, with a median sample size of 6,603. Sample size differences of this magnitude explain why the NSDUH prediction intervals (PIs) are generally wider than the corresponding BRFSS confidence intervals (CIs).

Table D.1 Alcohol Use in Past Month among Persons Aged 18 or Older, by State: Percentages, Annual Averages Based on 2007 BRFSS and 2006-2007 NSDUHs
State 2007 BRFSS
(Estimate)
2007 BRFSS
(95% Confidence
Interval)
2006-2007
NSDUH
(Estimate)
2006-2007 NSDUH
(95% Prediction
Interval)
p value
NOTE: NSDUH estimates are based on a survey-weighted hierarchical Bayes estimation approach, and the 95 percent prediction (credible) intervals are generated by Markov Chain Monte Carlo techniques.
NOTE: p value: Probability of no change between the BRFSS and NSDUH estimates.
Source: SAMHSA, Office of Applied Studies, National Survey on Drug Use and Health, 2006-2007, and Behavioral Risk Factor Surveillance System, 2007.
Alabama 38.2 (36.4 - 40.0) 42.65 (39.36 - 46.00) 0.020
Alaska 54.8 (51.7 - 57.9) 59.22 (55.67 - 62.67) 0.065
Arizona 53.6 (50.7 - 56.5) 55.98 (52.51 - 59.39) 0.301
Arkansas 40.1 (38.3 - 41.9) 46.63 (43.07 - 50.23) 0.001
California 54.6 (52.8 - 56.4) 53.74 (51.88 - 55.60) 0.515
Colorado 62.6 (61.4 - 63.8) 64.42 (61.06 - 67.64) 0.313
Connecticut 65.8 (64.2 - 67.4) 66.98 (63.53 - 70.26) 0.538
Delaware 58.4 (56.2 - 60.6) 58.33 (54.66 - 61.90) 0.974
District of Columbia 61.5 (59.5 - 63.5) 63.07 (59.72 - 66.30) 0.426
Florida 54.4 (53.2 - 55.6) 56.29 (54.33 - 58.24) 0.107
Georgia 48.4 (46.8 - 50.0) 50.23 (46.82 - 53.64) 0.342
Hawaii 51.0 (49.2 - 52.8) 51.62 (47.84 - 55.38) 0.772
Idaho 46.6 (44.8 - 48.4) 45.95 (42.48 - 49.47) 0.746
Illinois 58.1 (56.3 - 59.9) 57.02 (55.09 - 58.93) 0.421
Indiana 50.4 (48.6 - 52.2) 54.12 (50.75 - 57.46) 0.056
Iowa 56.7 (54.9 - 58.5) 56.58 (53.26 - 59.84) 0.950
Kansas 49.7 (48.3 - 51.1) 56.46 (52.92 - 59.93) 0.001
Kentucky 31.4 (29.4 - 33.4) 44.15 (40.77 - 47.59) 0.000
Louisiana 46.2 (44.6 - 47.8) 53.95 (50.41 - 57.44) 0.000
Maine 57.3 (55.7 - 58.9) 56.88 (53.22 - 60.46) 0.835
Maryland 55.3 (53.7 - 56.9) 58.45 (55.06 - 61.77) 0.099
Massachusetts 65.3 (64.3 - 66.3) 64.78 (61.37 - 68.04) 0.769
Michigan 59.4 (57.8 - 61.0) 60.01 (58.19 - 61.81) 0.621
Minnesota 60.0 (58.0 - 62.0) 65.63 (62.04 - 69.05) 0.007
Mississippi 38.1 (36.5 - 39.7) 40.00 (36.73 - 43.36) 0.310
Missouri 51.9 (49.7 - 54.1) 54.24 (50.78 - 57.67) 0.263
Montana 56.4 (54.6 - 58.2) 59.42 (56.11 - 62.65) 0.115
Nebraska 56.0 (54.0 - 58.0) 57.17 (53.57 - 60.70) 0.576
Nevada 54.8 (52.4 - 57.2) 56.11 (52.13 - 60.02) 0.579
New Hampshire 63.7 (62.1 - 65.3) 66.78 (63.37 - 70.02) 0.107
New Jersey 55.7 (53.9 - 57.5) 59.28 (55.74 - 62.72) 0.076
New Mexico 50.7 (48.9 - 52.5) 47.79 (44.09 - 51.51) 0.168
New York 57.7 (56.1 - 59.3) 59.12 (57.12 - 61.09) 0.276
North Carolina 44.3 (43.1 - 45.5) 45.34 (41.78 - 48.96) 0.590
North Dakota 62.0 (60.2 - 63.8) 64.70 (61.13 - 68.12) 0.183
Ohio 53.8 (52.4 - 55.2) 55.00 (53.14 - 56.85) 0.312
Oklahoma 40.9 (39.3 - 42.5) 45.74 (42.06 - 49.47) 0.018
Oregon 59.8 (58.0 - 61.6) 59.75 (56.40 - 63.02) 0.979
Pennsylvania 54.6 (53.0 - 56.2) 56.62 (54.76 - 58.47) 0.106
Rhode Island 64.3 (62.3 - 66.3) 67.42 (64.08 - 70.58) 0.113
South Carolina 47.5 (46.1 - 48.9) 50.44 (47.03 - 53.85) 0.118
South Dakota 57.2 (55.4 - 59.0) 62.16 (58.77 - 65.43) 0.011
Tennessee 32.9 (30.7 - 35.1) 43.12 (39.60 - 46.71) 0.000
Texas 48.6 (47.4 - 49.8) 51.71 (49.84 - 53.58) 0.006
Utah 27.5 (25.7 - 29.3) 33.71 (30.47 - 37.12) 0.001
Vermont 64.3 (62.9 - 65.7) 64.14 (60.60 - 67.52) 0.933
Virginia 55.5 (53.5 - 57.5) 56.92 (53.43 - 60.35) 0.488
Washington 59.7 (58.9 - 60.5) 58.46 (54.97 - 61.88) 0.492
West Virginia 34.3 (32.5 - 36.1) 39.11 (35.76 - 42.56) 0.013
Wisconsin 68.3 (66.7 - 69.9) 66.25 (62.92 - 69.42) 0.263
Wyoming 55.4 (53.8 - 57.0) 60.21 (56.74 - 63.59) 0.014

Below is a map; click here for the text describing this map.

Figure D.1 Alcohol Use in Past Month among Persons Aged 18 or Older, by State: Percentages, Annual Averages Based on 2007 BRFSS

Figure D.1

Source: Centers for Disease Control and Prevention (CDC), Behavioral Risk Factor Surveillance System Survey, 2007.

Below is a map; click here for the text describing this map.

Figure D.2 Alcohol Use in Past Month among Persons Aged 18 or Older, by State: Percentages, Annual Averages Based on 2006-2007 NSDUHs

Figure D.2

Source: SAMHSA, Office of Applied Studies, National Survey on Drug Use and Health, 2006 and 2007.

Go to Top of PageGo to the Table of Contents

This page was last updated on .