Behind the Numbers: SAMHSA's Survey on Drug Use

By Rebecca A. Clay

"Watch out for the polar bear!" is not the kind of advice most people expect to hear while they're on the job. For Peter H. Law, however, the threat of a face-to-face encounter with a bear was just one of the hazards he confronted while collecting information about substance use and mental illness in isolated Alaskan villages.

As an interviewer for SAMHSA's National Survey on Drug Use and Health, Mr. Law traveled by helicopter through dense fog to get to one village and through a blizzard to reach another one. He tells the story of one remote town where he inched along narrow icy paths, with chained dogs lunging at him on one side and a steep drop-off on the other. At one point, he crawled on his hands and knees to get to a house raised on stilts high on a hillside.

"I felt like Indiana Jones," said Mr. Law. "What would probably be the trip of a lifetime for someone else is just another day at work for an interviewer in Alaska."

Each year, field interviewers like Mr. Law fan out across the Nation to knock on thousands of doors and collect information on residents' alcohol, tobacco, and illegal drug use. Formerly known as the National Household Survey on Drug Abuse, the National Survey on Drug Use and Health is the Federal Government's leading source of statistical data on the prevalence, patterns, and consequences of alcohol, tobacco, and illegal drug use among the civilian, non-institutionalized U.S. population age 12 and older.

The survey also collects information about mental illness and mental health treatment. SAMHSA's Office of Applied Studies (OAS) plans and manages the survey, which is conducted by a not-for-profit research organization, RTI International, in North Carolina.

Back to Top

A Massive Effort

Initiated in 1971 in response to growing concerns about substance abuse, the survey took place approximately every 2 years in its early days. When demand for up-to-date information increased in the late 1980s, the survey became an annual undertaking.

Today, the survey relies on more than 700 field interviewers to administer questionnaires during face-to-face interviews at the homes of a sample representing 98 percent of the U.S. population.

In 2002, field interviewers collected data from 68,126 people in houses, apartment buildings, homeless shelters, dormitories, rooming houses, migrant worker camps, halfway houses, and civilian quarters on military bases in all 50 states and Washington, DC. The survey excludes active-duty military personnel, homeless persons who do not use shelters, inmates at correctional facilities, and people in nursing homes, mental institutions, and long-term hospitals.

This massive effort produces a wealth of information, said Joseph C. Gfroerer, Director of the Division of Population Surveys within OAS.

"Policymakers, researchers, service providers, and others all turn to the survey results to inform their work," said Mr. Gfroerer. "The survey provides crucial information about the extent of substance use as well as the unmet need for prevention and treatment."

Because of shifting information needs and other key factors, the survey is constantly evolving. Some of the survey's changes reflect changes in types of drug use. The current survey asks about Ecstasy, a "club drug" often used by teenagers at late-night dances, for example. Two decades ago, most people had not heard of this drug, and it was not included in the survey. Other changes reflect new demographic realities. In 1999, for instance, the survey revised its questions about racial and ethnic background to better reflect the Nation's diversity.

Back to Top

Data Gathering

The sheer size of this nationwide effort makes the survey a challenge. And it can be difficult to persuade people to participate.

"Most people wouldn't respond positively to someone saying, ‘I'm here from the Government, and I want to ask you about your drug use,' " said Mr. Gfroerer. "But once people understand the purpose of the survey, and hear from our highly trained interviewers how data are collected and handled, they're usually willing to help." About 80 percent of those asked to participate actually complete the interview.

In the field, interviewers don't just decide at random which doorsteps to approach. Using statistical techniques, survey designers first compile a sample of more than 175,000 addresses and then narrow it down to eligible ones. Potential respondents in those households receive an introductory letter explaining the survey and all confidentiality safeguards.

Interviewers then head out to each location or home, using small, hand-held computers to collect basic demographic information about household members. The computer uses that preliminary information to select appropriate survey participants. Those who complete the interview receive a $30 payment as a thank-you.

To collect basic demographic information, interviewers use laptop computers. They read questions from their screens and then type responses into their keyboards. For more sensitive questions, interviewers offer their computers to respondents for privacy. Respondents can read the questions on screen or listen with headphones, and then they key their responses into the laptop. The entire process takes about an hour.

"These computer-assisted, self-interviewing techniques are designed to give respondents a very private way of answering questions," said Mr. Gfroerer. "That increases the chances that they'll be honest when they're detailing any illicit drug use and other sensitive behaviors." In fact, he pointed out, research shows that such techniques produce more honest answers than pencil-and-paper questionnaires.

Back to Top

A Wealth of Information

At the end of each day, interviewers transmit data collections electronically to RTI International by phone.

Once the raw data are received, computers perform complex editing procedures and check for consistency of reporting. For example, the computers may find a missing or ambiguous answer in response to a question about most recent drug use. Rather than "in the past 30 days" or "more than 12 months ago," the respondent answered with a vague "some point in lifetime." In that case, a process called imputation fills in the response that statistical models deem most likely to be correct.

The process assembles and taps into a set of respondents with complete data, similar to the respondent in question. The computers choose one respondent at random and then borrow his or her "good" answer to fill in the missing response.

The final step is the calculation of analysis weights. This step is designed to ensure that the data set is truly representative of the U.S. population as a whole. The survey intentionally samples a disproportionate number of young people, Mr. Gfroerer explained; to account for that oversampling, the survey assigns different weights to respondents. These adjustments help avoid skewed results.

The calculation of analysis weights also considers possible variations in participation rates in certain geographic areas and other key factors that could produce unrepresentative results.

Many publications are generated from this effort. The primary report, Results from the 2002 National Survey on Drug Use and Health: National Findings, more than 250 pages, provides a comprehensive summary of the latest findings.

A companion report, Overview of Findings from the 2002 National Survey on Drug Use and Health, offers a summary version with highlights of the most important findings and a brief discussion of methodological issues. OAS also produces specialized "short reports" on specific populations, age groups, and topics of interest.

For copies of the reports, contact SAMHSA's National Clearinghouse for Alcohol and Drug Information at P.O. Box 2345, Rockville, MD 20847-2345. Telephone: 1 (800) 729-6686 (English and Spanish) or 1 (800) 487-4889 (TDD). Publications can also be viewed on the SAMHSA Web site at www.oas.samhsa.gov.

Back to Top