HEALTH PROMOTION AND DISEASE PREVENTION (HPDP) SAMPLE PERSON SUPPLEMENT NOTES INTRODUCTION 1. The 1990 NHIS of Health Promotion and Disease Prevention (NHIS-HPDP) Sample Person file is very similar to the 1985 NHIS of Health Promotion and Disease Prevention Sample Person file. Many items are identical. Some sections have minor changes from 1985. Two sections, mammography and radon, were new in 1990. Occupational health, which was included in the 1985 HPDP, was not part of the 1990 survey. 2. The NHIS-HPDP Sample person file consists of: a. The NHIS person record from the Core questionnaire (locations 1-200) b. Weight fields (locations 201-212) c. The special topic data (locations 336-773) 3. Item non-response was handled differently in 1990 than in 1985. In 1990, various types of item non-response were identified: (1) Respondent refused to answer the question (codes 7, 97, and 997); (2) The information was "not ascertained" (codes 8, 98, or 998) (including blanks when there should have been a response and impossible codes) and; (3) A response of "don't know" when the question was asked. For some items, it was determined that there was no meaningful difference between "not ascertained" and "don't know." In these cases, the two categories were both coded 9 (or 99 or 999, as appropriate), and labeled "unknown." In contrast, in the 1985 HPDP, for most questions, all three types of item non-response were coded 9 and labeled "DK or refused." For a few items, "refused" remained a separate category. The preservation of unambiguous "don't know" and "refused" categories in 1990 may be helpful, particularly when analyzing knowledge items or items of a sensitive nature. 4. The response rate for the NHIS-HPDP was 83.4 percent. This response rate was calculated as follows: Household response rate from Core (95.5 percent) multiplied by the HPDP sample person response rate (87.3 percent) = 83.4 percent - 1 - 5. Initially, dummy records were created for all HPDP non- respondents. Comparison of sociodemographic characteristics of respondents and non-respondents indicated that the two groups were not sufficiently different to warrant retaining dummy records in the file, especially since dummy records were not included on the 1985 HPDP file and their inclusion in the 1990 file would make trend comparisons more difficult. 6. Weights: Weights must be used to make accurate estimates based on data from the National Health Interview Survey. Two weights are included on the 1990 HPDP Sample Person file: a. The Final Basic Weight (location 207-212), calculated for each sample person, is the weight that will be used in most analyses of the HPDP data. This weight is the functional equivalent of the Annual Final Basic Weight found on the NHIS Person Record of the Basic Health and Demographic component of the survey (i.e., the Core questionnaire). The HPDP Final Basic Weight differs in two ways from that calculated for the Core data file: (1) the HPDP sample was re-weighted to account for non-response and selection of only one adult per family and, (2) the weight field on the Core data file is in binary format and the weight field on this file is in character format. b. The Interim Basic Weight (before age-sex-race adjustment), required by some software packages for variance estimation for surveys with complex sample designs, is also included on the file (loc. 201-206). 7. Estimating numbers of events or conditions a. To reduce respondent error, the recall period for questions about some events is limited to two weeks. These events are: bed days and other restricted activity days, work loss and school loss days, and doctor visits. The two-week variables are found in locations 98-107 and 120-121. Estimates of the total number of occurrences of these events in the population can be derived as follows: Number of events x 26 (number of two-week periods in a year) x Final Basic Weight = Total number of events occurring in the population during the data year, i.e., 1990. - 2 - Example: Number of bed days (loc. 100-101) x 26 x Final Basic Weight (loc. 207-212) = total number of bed days reported for the population in 1990. b. The recall period for acute incidence conditions is also two weeks and an annual estimate of the total number of acute incidence conditions is calculated using the same procedures as for two-week events: Number of acute incidence conditions x 26 x Final Basic Weight = Total number of acute incidence conditions occurring in the population during 1990. Note: An acute incidence condition is an acute condition with onset during the two weeks preceding the date of interview. c. The recall period for information on hospitalizations is 12 months. However, in calculating number of discharges and number of days in hospital (locations 132-141), only discharges occurring in the past 6 months are counted. Therefore, the weighted estimates for these events must be calculated as follows: (1) Number of discharges x 2 (number of 6-month periods in a year) x Final Basic Weight = Total number of discharges occurring in the population in one year. (2) Number of days in hospital associated with discharges occurring in the past 6 months x 2 (number of 6-month periods in a year) x Final Basic Weight = Total number of days of hospitalization occurring in the population in one year. 8. Calculation of rates for events and conditions: The number of events or conditions estimated for the population, as described in item 7, above, can be used as the basis for calculating rates of occurrence of these events (or conditions) per person and per 100 persons for the total U. S. population and for various population subgroups. Note: Only rates can be estimated from these data. The percent of the population experiencing a particular type of event during the data year cannot be estimated. [The percent of the population experiencing the event in the reporting period (i.e., two weeks or 6 months) can be estimated but is generally not meaningful.] - 3 - 9. Data on hospital episodes and days, based on a 12-month recall are locations 122-131. The Final Basic Weight is used for calculating estimates of these events in the same way it is used for all other person-based variables. These variables do permit estimating the percent of the population experiencing a hospital episode in the past year and the percent of the population having a specified number of hospital days. This description prepared by: Charlotte A. Schoenborn IDSB September, 1991 - 4 -