2014 NAPHSIS Meeting Date of Meetings – June 8 - 11, 2014 Presentation: Tuesday, June 10th NAPHSIS: VITAL RECORDS: A culture of Quality.

2014 NAPHSIS Meeting

• Date of Meetings – June 8 - 11, 2014• Presentation: Tuesday, June 10th

NAPHSIS:

VITAL RECORDS: A culture of Quality

A Use of Linked Data in the Study of US Mortality:

The National Longitudinal Mortality Study

(NLMS)

Norman J. JohnsonU.S. Census Bureau

National Longitudinal Mortality Study

Norman Johnson, Principle InvestigatorCensus Bureau

Sean Coady, Steering Committee Co-Chair,Principle Investigator National Heart, Lung, and Blood Institute of the NIH

Sean Altekruse, Steering Committee Co-Chair, Principle InvestigatorNational Cancer Institute of the NIH

•Overview of NLMS

•Study Design

•A Sampler of Study Findings

•Conclusion

•Overview of NLMS

•Study Design

•A Sampler of Study Findings

•Conclusion

NLMS in brief:

U.S. Census Bureau

- Current Population Surveys, - Annual Social and Economic Supplements & - A subset of1980 Census Data

linked to - Mortality Outcomes derived from the National Death Index

Purpose of NLMS:

To study the relationship between

demographic and socioeconomic factors

and

differentials in U.S. mortality rates.

NLMS Sponsors and Funding SupportersSponsors: The National Heart, Lung, and Blood Institute The National Cancer Institute The National Institute on Aging The National Center for Health Statistics The Food and Drug AdministrationFunders:The Office for Research on Women’s HealthThe Office of Behavioral and Social Science ResearchThe National Institute on Minority Health and Health DisparitiesThe NCI Center to Reduce Cancer Health Disparities

Overview of NLMS

Study Design

A Sampler of Study Findings

Conclusion

NLMS Study Design

Cohort Definition:39 Current Population Surveys (1973 – 2011) 1 1980 U.S. Census sub-sample(~3.8 million persons)

Mortality Definition:Deaths occurring between1979-2011 as identified in the National Death Index(~575,000 deaths identified)

Current Population Survey: National sample of householdsComplex sample designSamples of: Non-institutional U.S. populationTitle 13, U.S. Code, confidentiality restrictions

60,000 - 80,000 households / monthSelected sub-populations over-sampled(Hispanics, Blacks, Am Indians, small states)Data collected on all persons in the household(96% response rate)

NLMS – Cohort Selection Criteria

Those which had NDI linking variables:

Social Security Number,Name(first and last),Date of Birth (at least month & year)

Core CPS Variables for NLMS:

Age, Sex, Race, Marital Status, Education, Income Hispanic Origin, Ethnicity(Some information on tobacco use)

Extensive employment information:Occupation, Industry, Current Work Status, Insurance carriedNO DIRECT HEALTH INFORMATION

Other Variables

•Employment/Unemployment

•March Income Supplement data & Components of income

•Geographic variables

•Household composition

NLMS Mortality Update:

1. Step 1. Match NLMS records to the NDI(Intend to match every 2 years)

NDI matching algorithm involves: SSN, Name, Date of Birth and variations of these (NYSIS name, YOB +-2)

2. Step 2. NLMS scores the probability of a correct match (NLMS Bibliography

Ref # 6).Uses 12 personal identifiers producing

“true”, “questionable”, “false” matches.

NLMS Mortality Update(cont.):Step 3. Manual Review of Questionables

(~95% matches to the NDI are false positive matches)

MR conducted at the NPC using NLMS developed process called, EASY

(EASY allows direct comparison of Death Certificate to complete HH information)

Step 4. Update files with mortality information

Validation of Match Algorithm

Applied to American Cancer Society Study [Calle, Terrell, Am J of Epi, 1994]

15,000 persons with known vital status

With SSN - 97% of deaths identifiedW/O SSN - 87% of deaths identified

Mortality Information on NLMS:

All information from Death Certificate (age, sex, race, Hispanic, education, occupation)

Multiple Contributing CausesUnderlying Cause(Cause of Death according to the ICD, 9th & 10th Revisions)

Participation in CPS is VOLUNTARY

1. Permission is obtained to use the information in Administrative Records Studies before survey is administered2. Households which refuse are not included

in the NLMS.3. There is no follow-back to interviewees for

any reason. 4. No NLMS information is shared with other Census areas or other Government

agencies

Confidentiality Practices of the NLMS

1. Title 13 U.S. Code - Regulates Census Activities, including concerns for

privacy and confidentiality issues

2. NLMS Cohorts derived from CPS which is based on sampling from the 10-year

Census frame, so NLMS work must abide by Title 13 controls concerning

privacy and confidentiality of records.

Confidentiality Practices of the NLMS1. Take Title 13 Review Course every year

2. Sign a Title 13 Adherence Pledge twice a year3. Password restricted access to computers4. Restricted access to project files.5. Analytical files contain no PII data6. Output restricted to tabular and model parameter estimates, sparse cell restrictions.7. No individual data is ever released.

Penalties for Violations of ConfidentialityIf convicted:1. Loss of job2. Up to $500,000 fine

AND/OR3. Up to 5 years in jail as a Federal offense.

Overview of NLMS

Study Design

A Sampler of Study Findings

Conclusions

National Longitudinal Mortality Study Summary of Major Findings• Methodological• Demographic factors• Socioeconomic factors• Health care factors• Analytical cautions

•[See bibliography on Website]•[ Current bibliography list 86 articles

published in public health and scientific journals ]

METHODOLOGICAL PAPERS

•Algorithm to predict whether a match to theNational Death index is correct (Ref # 6) Journal of Chronic Diseases

•Studies comparing demographic data onthe death certificate with CPS data.

NCHS research staff has published four of the 86 journal articles to assess and document the quality of the information provided on the death certificate.

DEMOGRAPHIC PAPERS

•Mortality differentials by race/ethnicity

Black/White: Higher death rates in blacks (age <65)is decreased, but remains after adjusting for income.(Ref # 12). LANCET

Hispanic/Non-Hispanic: Hispanics show lower deathrates overall, and strongly lower for CVD. Rates even lower after adjustment for SES. (Ref # 14)

•Mortality differentials by marital status (Ref #48)Married persons have lowest death rates. Mortalityfollowing widowhood remains elevated for 5 years.

SOCIOECONOMIC PAPERS

•Backlund landmark paper on: Separation of income and education effects.(Ref#36)

Increased education and education are associated with reduced mortality at thresholds. For income, steepest decline for survival occur at lowest income.

SOCIOECONOMIC PAPERS

•Specific Occupations (Ref # 37)After adjustment for age, race, income, education, service workers have the highest mortality rates.

Specific occupations – food service workers have highest mortality, lawyers and math/sci professionalsthe lowest.

HEALTH CARE PAPERS

Differential mortality by health insurance status(Ref # 15) Archives of Internal Medicine

• Working uninsured have mortality rates 1.2 to 1.3 times higher than the working insured.

• Highest mortality for those <65 for those in Medicaid or Medicare.

• Selection effects and health care access likely to influence differentials in mortality.

Some Additional Research

1. The Dougy Foundation

2. Current Research on:

Amyotrophic Lateral Sclerosis

with Dr. Marc Weisskopf, Harvard School of Public Health

Overview of NLMS

Study Design

A Sampler of Study Findings

Conclusion:

HELP !!!A. We just completed a match to the NDI

B. WE ASK FOR: Dept. of Vital Records SUPPORT

C. Provide quick response to our request for copies of Death Certificates.

D. We need lowest possible cost

HELP !!! 1. Is there a low cost option for bulk orders or for applications to research?

We CAN provide HELP to reduce cost: 1. We can provide manpower 2. We can provide for pay for work

outside hours or extra hours staff pay.

NAPHSIS: VITAL RECORDS: A culture of Quality

The NLMS Website:

www.census.gov/nlms