You are here: Home
Surveillance of Mortality and Morbidity in U.S. Workers
University of Miami Study Database Documentation
For the purposes of this Study, the Investigators have made a series of policy
decisions with regards to definitions, statistical analyses and other issues.
In addition, in the process of their explorations of the NHIS database,
the Investigators have encountered particular issues (e.g. “SOC” coding)
which are detailed below in hopes that this will aid other investigators.
Several of these explorations have resulted in detailed tabular data which
have also been made available to other investigators where indicated in the documentation below. Of note, this is an ongoing study and exploration, as such this Study Documentation will continue to be updated as necessary. The following issues have been addressed:
1986-1994 since these were the years with mortality follow up available through 1997 (as of 2003).
NHIS Working Population
"Currently employed" (paid and unpaid) within last 2 weeks prior to annual NHIS interview (Kaminski and Spirtas 1980)
18 yrs of age or older at time of annual NHIS interview
Use of Supplement Survey Data
Use only from the same year as their NHIS Survey
Do not use any supplements involving < 18 year olds
Weights for supplemental data
a) use supplemental survey weight if doing prevalence/incidence
If there is a year with > 2 or more supplements and overlap of participants, then
b) use NHIS annual survey weight for all other uses (for example if used as a covariate)
a) for Prevalence estimates, use only the larger Supplement and its weights
For years with > 2 or more supplements and no overlap and equal selection criteria, then use both supplements with appropriate supplement weight divided by the numberof supplements
b) for smoking enrichment, use both supplements but IF discordance, use the data from the larger Supplement
ex: 1991, use Health Promotions Supplement for Prevalence with appropriate
supplement weight; for smoking enrichment, use both Health Promotions and Smoking
Supplements with annual NHIS survey weight, if discordance of smoking information,
use Health Promotions smoking information
o ex: 1987 Cancer Control and Cancer Epidemiology Supplements were mutually exclusive and used similar selection criteria
Eliminate 1989 Diabetes and 1988 TAPS supplements due to very little data and/or too selective
Occupation Defined as current work in past 2 weeks prior to NHIS interview
"SOC" and "SIC" codes available
"SOC" and "SIC" codes mis-identified/labeled by NHIS documentation; really US Census Occupational and Injury codes
13 and 42 category recodes
used for initial evaluations
"SOC" codes used for specific occupational evaluations
For major analyses, selected occupations
with a 9 year average of >100,000 workers/occupation based on US Population of 1990
o Label/indicate if >30% SE (as per Brackbill R 1988)
"SOC" CodesIn 1992, the NHIS began using the 1990 version of the census occupational codes. Prior to 1992, the 1980 census codes were used. There were differences in some of the occupational categories between the 1980 version and the 1990 version.
The changed occupational codes were programmatically recoded in
the 1986-1991 data sets to categories compatible with the 1990 version. All of our reports are expressed in terms of the 1990 census occupational categories.
• Instruction Manual Part 19, Industry and Occupation Coding for Death
Certificates,Appendix D, 1993. Hyattsville, MD: 1992 http://www.cdc.gov/nchs/about/major/dvs/im.htm
Vital Statistics reference: Technical Appendix http://www.cdc.gov/nchs/data/natality/96linked/94mortad.pdf
Longest held job
• Only use as supplemental check
• 1986 Supplement on Longest Held Job (37,917)
• 1988 Supplement on Occupational Health (44,233)
o information on: current job, last 12 months job,
longest held job
concordance longest held job vs current job
• Burnett (1989), Cooper (1993) important references
Held Job in US Occupational Groups: The National Health Interview
Survey Journal Of Occupational and Environmental Medicine
Industry"SIC" codes available
but really US Census industry codes and are mislabeled as "SIC" by NHIS documentation
• Defined as current work in past 2 weeks prior to NHIS interview
Smoking/Tobacco UseVariable created for all NHIS participants:
• Create additional variable for all NHIS employed participants
• NHIS Questions: Have you smoked 100 > cigarettes in your life?.
What is your current smoking status?
o These two questions were asked of a subset of
adults in the following years: 1987, 1989, 1990, 1991, 1992,1993, 1994.
Keep Historic Smokers (ie known smoked > 100
cigs but not if current or former) as "Unknown or Missing"
These people can be used if using: Ever vs Smokers
but not if any detail on the smokers (then must set to "missing")
o Therefore Smoking Categories: Never, Ever (Former, Current), Unknown
No smoking questions from the 1994 second supplement on aging and
the 1994 followback survey
• See section on use of Supplement Survey Data for additional information
• Brackbill (1988), Nelson (1994) important references
Trends in US Smoking Rates in Occupational Groups: The National Health Interview Survey 1987-1994
Journal Of Occupational and Environmental Medicine
Only use as supplemental check
• 1988 Alcohol Supplement (43,809 subjects)
o Variable: Current Drinkers” = people who
answered “yes” to “in the past 12 months
the subject has had at least 12 drinks of any kind of alcoholic beverages.” (Question
• 1991 Alcohol and Drug Supplement (21,174 subjects)
o Variable Current Drinker =people who drank alcohol
in the past 12 months (ie. Answered
yes to tape location 337-338, item number 2, items and codes 01-12 (1-12 months drank).
Response to NHIS
• In the period 1986-1994, a nnual NHIS household survey response rates
ranged from 94% to 97% (1-9) ; annual response rates to the 1997-2002 adult core
ranged from 70% (in 1999) to 80% (in 1997) (10-15) .
1. Dawson DA, Adams PF. Current estimates from the National Health Interview
Survey, 1986. Vital Health Stat 10 1987:1-177.
2. Schoenborn CA, Marano M. Current estimates from the National Health
Interview Survey, 1987. Vital Health Stat 10 1988:1-233.
3. NCHS. Current estimates from the National Health Interview Survey,
1988. Vital Health Stat 10 1989:1-250.
4. Adams PF, Benson V. Current estimates from the National Health interview
survey, 1989. Vital Health Stat 10 1990:1-221.
5. Adams PF, Benson V. Current estimates from the National Health Interview
Survey, 1990. Vital Health Stat 10 1991:1-212.
6. Adams PF, Benson V. Current estimates from the National Health Interview
Survey, 1991. Vital Health Stat 10 1992:1-232.
7. Benson V, Marano MA. Current estimates from the National Health Interview
Survey, 1992. Vital Health Stat 10 1994:1-269.
8. NCHS. Current estimates from the National Health Interview Survey,
1993. Vital Health Stat 10 1994:1-221.
9. Adams PF, Marano MA. Current estimates from the National Health Interview
Survey, 1994. Vital Health Stat 10 1995:1-520.
10. Blackwell DL, Collins JG, Coles R. Summary health statistics for
U.S. adults: National Health Interview Survey, 1997. Vital Health Stat
11. Pleis JR, Coles R. Summary health statistics for U.S. adults: National
Health Interview Survey, 1998. Vital Health Stat 2002;10(209):1-121.
12. Pleis JR, Coles R. Summary health statistics for U.S. adults: National
Health Interview Survey, 1999. Vital Health Stat 2003;10(212):1-145.
13. Pleis JR, Benson V, Schiller JS. Summary health statistics for U.S.
adults: National Health Interview Survey, 2000. Vital Health Stat 2003;10(215):1-141.
14. Lucas JW, Schiller JS, Benson V. Summary health statistics for U.S.
adults: National Health Interview Survey, 2001. Vital Health Stat 2004;10(218):1-143.
15. Lethbridge-Cejku M, Schiller JS, Bernadel L. Summary health statistics
for U.S. adults: National Health Interview Survey, 2002. Vital Health Stat
Since only one household member used for the NHIS interviews (and
often for the supplements), proxy information is provided
for a large component
Always re-look at analyses using proxy as an interaction term
Vital Status Use NHIS algorithm criteria for matching with National Death Index
Keep proxy deaths
Make first day of each quarter the arbitrary date of death
Eliminate any deaths that occurred prior to date of NHIS interview
although keep these individuals in the dataset but will list as "missing" data
Assume if not dead then alive
• 2000 US Census
• 2000 US Mortality rates
• Variable categorical defined as <12 grade, 12 grade, > 12 grade
• Approximately 4000 missing data kept in analysis except for education
• Variable continuous
• Variable categorical defined as: 18-44, 45-64, 65+ years
Variable categorical defined as: White Non-Hispanic, White Hispanic, Black
Non-Hispanic, All Others
Sample Size Considerations
NHIS default of >30% of the Relative Standard Error (RSE) gets
and asterix to indicate that the data are unstable
• Do not use data with less than 45 individuals/yr
o Standardized Mortality Ratios (SMRs)
The Investigators have provided SAS program language to perform the SMR calculations.
In addition, Noble, Drever, and Shah give details on computing confidence
intervals, with formulas for events > 100 and table values for events between 0 and 100
o Logistic Regression
Obesity/Height & Weight Measurement
• In 1996 the National Center for Health Statistics (NCHS) truncated the range
of weight from 50-500 pounds to 98-289 pounds and height
from 36-98 inches to 59-76 inches in order to protect possible identification of survey participants.
• For the 1996 NHIS participants outside of the abovementioned ranges, Body Mass
Index (BMI) cannot be calculated.
• Beginning in 1997, the NCHS computed BMI values for all survey participants,
including those outside of the restricted range.
Acute & Chronic Disability Study Definitions
||Study Variable Definition
= 0 vs ≥ 1 Restricted Activity Days in Prior 2 Wks
AFD2 = 0 vs ≥ 1 Bed Days in Prior 2 Wks
AFD3 = 0 vs ≥ 1 Lost Work Days in Prior 2 Wks
CFD1 = 0-3 vs ≥ 4 Doctor Visits in Prior 12 months
CFD2 = 0 vs ≥ 1 Hospitalizations in Prior 12 months
HS1 = 0 if Health Self-rated Excellent or Good;
= 1 if Health Self-rated Fair
HS2 = 0 vs ≥ 1 Health Condition reported
1986 - 1994 NHIS Chronic Condition Prevalence