N C H S - N H I S - Tobacco Information for Data Analysts, by Year

National Health Interview
Survey (NHIS)
Celebrating the First 50 years:
1957 - 2007

This section provides information in one place for data analysts who need to understand the basic changes to the NHIS over the years as they analyze tobacco data.

6.1. Tobacco Related Data Files
This section provides general and detailed information for data analysts who need to understand the changes to the tobacco related data files over the years.
To see a list of the NHIS files with tobacco data, See Section 6.1.2.

Tobacco data were obtained from questions on “Smoking Habits Pages” that were printed in the NHIS core questionnaire booklets for FY 1965 and FY 1966. The smoking information obtained was not released in the “core” files, but was released in small, separate smoking data files in FY 1965 and FY 1966. [1,2]

In 1966, the Bureau of the Census contracted with the NCHS Division of Health Interview Statistics to conduct a smoking habits survey as a supplement to the Census Bureau’s Current Population Survey (CPS) of June 1966. This study was designed to provide data comparable to an initial CPS smoking supplement of February 1955. The Division of Health Interview Statistics processed the data and has preserved copies of the file. As this survey was separate from the annual NHIS, it will not be discussed further here, except to note that it was the first DHIS-based survey to include a smoking status recode variable. The smoking status recode variable was not incorporated into the NHIS until 1970.

July 1967-December 1968 comprised an 18 month interim period during which NHIS data collection and processing switched from a fiscal year to the calendar year. Use of the calendar year began in January 1969 and allowed NHIS health data to be more compatible with other health-related statistics. For 1968, two sets of core files were released, one for Fiscal Year 1968 (July 1, 1967 – June 30, 1968), the other for Calendar Year 1968 (January 1, 1968- December 31, 1968). During the 12 month overlap period, July 1967-June 1968, a split sample design was used such that a half sample was administered using the old “condition approach” to disease conditions, and a half sample was administered using the new “person approach”. Information from subsequent tobacco use supplements could be combined with information on medical conditions obtained using the “person approach”. [2]

Core Files: From calendar years 1969 through 1996, data from the NHIS core questionnaire were released each calendar year with a minimum set of the following five files: household, person, condition, hospital, and doctor visit.

The 1970 smoking information was included in the core PERSON file. The smoking data from 1970, 1974, 1976, 1977, 1978, 1979, 1980 and 1983 were released in separate small files, containing mainly smoking data. Starting in 1985, tobacco related data were collected in selected years and generally released as part of more extensive, multi-topic prevention supplements. This approach continued through 1995.

For the most part, during these years, the tobacco use supplements were administered to one adult selected per household. For some years, questions on pregnancy and smoking were asked of all women in the household, including the sample adult, who met specified criteria. This information was released in a file separate from the core or other tobacco files.

In 1997, a redesigned NHIS was fielded that focused even more on the person by eliminating the files based on events (doctor visits and hospital visits) or conditions that were a part of the previous questionnaire design. A FAMILY level file was added to the NHIS data release in 1997.

For 1997 through 2003, the core information on cigarette smoking was released as part of the sample adult file, and any additional information on tobacco use was released in a separate cancer or prevention file. The exception was the “2000 Cancer Risk Factor” supplement information, which was included on the core sample adult file to facilitate analysis.

Through 2003, all of the basic demographic data and sampling information necessary to perform an analysis of the sample adult file was included on that file, although the correct analysis should start with the “person file”, and analyze only the sample adult subset, in order to preserve the complete sampling design structure. In 2004, sampling information appeared only in the HOUSEHOLD or FAMILY files. The demographic information appeared only in the PERSON files, with the core smoking information appearing in the SAMPLE ADULT file. The change required the user to combine files for most analyses.

In 2005, some of the demographic and sampling information was again included in the sample adult file, so that fewer files needed to be merged for proper analysis of tobacco use, other health behaviors, and condition data.

1. Ahmed, Paul I. and Gleeson, Geraldine A. Changes in Cigarette Smoking Habits Between 1955 and 1966. National Center for Health Statistics. Vital and Health Statistics 10(59), April 1970.

2. Wilder, Charles S. and Rivers, Carolyn W. Current Estimates from the Health Interview Survey, United States 1967, Series 10(52): 1-2. Washington DC: May 1969.