Linkage of 1986–2009 National Health Interview Survey With ...

Serie

s2,

Nu

mb

er1

67Se

pte

mb

er2

014

Linkage of 1986–2009National Health InterviewSurvey With 1981–2010Florida Cancer Data System

Copyright information

All material appearing in this report is in the public domain and may bereproduced or copied without permission; citation as to source, however, isappreciated.

Suggested citation

Miller EA, Miller DM, Judson DH, et al. Linkage of 1986–2009 National HealthInterview Survey with 1981–2010 Florida Cancer Data System. National Centerfor Health Statistics. Vital Health Stat 2(167). 2014.

Library of Congress Cataloging-in-Publication Data

Linkage of 1986–2009 National Health Interview Survey with 1981–2010 FloridaCancer Data System.

p. ; cm.— (Vital and health statistics. Series 2, Data evaluation and methodsresearch ; number 167) (DHHS publication ; (PHS) 2014–1367)‘‘September 2014.’’Includes bibliographical references and index.ISBN 978–0-8406–0667-9 (alk. paper)—ISBN 0–8406-0667–2 (alk. paper)I. National Center for Health Statistics (U.S.), issuing body. II. Series: Vital andhealth statistics. Series 2, Data evaluation and methods research ; no. 167. III.Series: DHHS publication ; no. 2014–1367. 0276–4733[DNLM: 1. National Health Interview Survey (U.S.) 2. Florida Cancer DataSystem 3. Public Health—statistics & numerical data—United States. 4. HealthSurveys—United States. W2 A N148vb no.167 2014]RA407.3614.4’273—dc23 2014031050

For sale by the U.S. Government Printing OfficeSuperintendent of DocumentsMail Stop: SSOPWashington, DC 20402–9328Printed on acid-free paper.

Series 2, Number 167

Linkage of 1986–2009 NationalHealth Interview Survey With1981–2010 Florida CancerData System

Data Evaluation and Methods Research

Hyattsville, MarylandSeptember 2014DHHS Publication No. 2014–1367

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICESCenters for Disease Control and PreventionNational Center for Health Statistics

National Center for Health Statistics

Charles J. Rothwell, M.S., M.B.A., Director

Jennifer H. Madans, Ph.D., Associate Director for Science

Office of Analysis and Epidemiology

Irma E. Arispe, Ph.D., Director

Contents

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Data Files. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Linkage Approval and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Evaluation of Linked File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Linked NHIS–FCDS Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Figures

1. Number of female participants available for breast cancer risk analysis using survey tobacco questions: 1986–2009National Health Interview Survey linked with 1981–2010 Florida Cancer Data System . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2. Percentage of diagnosed late-stage cancers, by education level and sex: 1986–2009 Florida National Health InterviewSurvey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System after interview . . . . . . . . . . . . 7

3. Percentage of diagnosed late-stage cancers among those above and below federal poverty level, by sex: 1986–2009Florida National Health Interview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer DataSystem after interview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4. Percentage of diagnosed late-stage cancers, by marital status and sex: 1986–2009 Florida National Health InterviewSurvey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System after interview . . . . . . . . . . . . 8

5. Percentage of self-rated fair and poor health status, by cancer diagnosis and sex: 1986–2009 Florida National HealthInterview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System before interview(cancer survivor) and not linked (not diagnosed with cancer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

6. Percentage having any activity limitation, by sex: 1986–2009 Florida National Health Interview Survey participantsaged 18 and over linked with 1981–2010 Florida Cancer Data System before interview (cancer survivor) and not linked(not diagnosed with cancer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

7. Percentage of current smokers, by sex: 1997–2009 Florida National Health Interview Survey participants aged 18 andover linked with 1981–2010 Florida Cancer Data System before interview (cancer survivor) and not linked (notdiagnosed with cancer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Detailed Tables

1. Availability of selected questionnaire topics, by survey year: National Health Interview Survey, 1986–2009 . . . . . . . . . . 112. Number and percentage of linked participants, by demographic and cancer characteristics: 1986–2009 Florida National

Health Interview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System . . . . . . . . . 123. Number of linked participants and weighted percent distribution of survey and cancer time-period characteristics, by

sequence of survey and cancer diagnosis: 1986–2009 Florida National Health Interview Survey participants aged 18and over linked with 1981–2010 Florida Cancer Data System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4. Number and percentage of cancer types and demographic characteristics: 1986–2009 Florida National Health InterviewSurvey participants aged 18 and over linked with Florida Cancer Data System (cancers diagnosed during 2006–2010)compared with overall Florida Cancer Data System, 2006–2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

iii

5. Cancer stage at diagnosis, by education level, poverty level, marital status, and sex: 1986–2009 Florida NationalHealth Interview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System afterinterview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

6. Percentage of smoking status, self-rated health, and any activity limitation, by sex: 1986–2009 Florida National HealthInterview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System before interview(cancer survivor) and not linked (not diagnosed with cancer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

iv

BackgroundNational survey data linked with

state cancer registry data has thepotential to create a valuable tool forcancer prevention and control research.A pilot project—developed in acollaboration of the Centers for DiseaseControl and Prevention’s NationalCenter for Health Statistics (NCHS) andthe Florida Cancer Data System(FCDS) at the University of Miami—links the records of the 1986–2009National Health Interview Survey(NHIS) and the 1981–2010 FCDS. Theproject assesses the feasibility ofperforming a record linkage betweenNCHS survey data and a state-basedcancer registry, as well as the value ofthe data produced. The linkedNHIS–FCDS data allow researchers tofollow NHIS survey participantslongitudinally to examine factorsassociated with future cancer diagnosis,and to assess the characteristics andquality of life among cancer survivors.

MethodsThis report provides a preliminary

evaluation of the linked national andstate cancer data and examines bothanalytic issues and complicationspresented by the linkage.

ConclusionsResidential mobility and the number

of years of data linked in this projectcreate some analytic challenges andlimitations for the types of analyses thatcan be conducted. However, the linkeddata set offers the ability to conductanalyses not possible with either dataset alone.

Keywords: linkage • cancer registry• data evaluation

Linkage of 1986–2009 NationalHealth Interview Survey With1981–2010 Florida Cancer DataSystemby Eric A. Miller, Ph.D., Donna M. Miller, M.S., Dean H. Judson,Ph.D., Yulei He, Ph.D., Hannah R. Day, Ph.D., Keith Zevallos, andJennifer D. Parker, Ph.D., National Center for Health Statistics; JillA. MacKinnon, Ph.D., Monique N. Hernandez, Ph.D., Brad Wohler,M.S., and Recinda Sherman, Ph.D., Florida Cancer Data System,University of Miami Miller School of Medicine; Christina A.Fernandez, M.S.Ed., Laura A. McClure, M.S.P.H, William G. LeBlanc,Ph.D., Stacey L. Tannenbaum, Ph.D., Diane D. Zheng, M.S., andDavid J. Lee, Ph.D., University of Miami Miller School of Medicine;and Sharon L. Christ, Ph.D., Purdue University

Introduction

This report describes a pilot linkagebetween the 1986–2009 National HealthInterview Survey (NHIS) and the1981–2010 Florida Cancer Data System(FCDS), the cancer registry for Florida.The report includes a preliminaryevaluation of the linked cancer data anddescribes some of the analytic issuesand complications presented by thelinkage. This project is a collaborationbetween FCDS at the University ofMiami and the Centers for DiseaseControl and Prevention’s (CDC)National Center for Health Statistics(NCHS). Its goals are to demonstrateand evaluate the feasibility ofperforming a record linkage betweenNCHS national sample survey data anda state-based cancer registry, and toassess the value and utility of the data itproduces. In addition to anticipatedcancer studies that would use the linkeddata, information from the linkageprocess can help inform future linkagesbetween sample survey data and cancerregistry data, and may be able to be

extrapolated to inform other linkagesbetween national sample surveys andstate data.

Linkage of NHIS with cancerregistry data can potentially produce avaluable tool for cancer prevention andcontrol research. NHIS data containdetailed demographic and healthinformation that is not available fromcancer registries. For example, NHIScollects information on income,education level, occupation and industry,health insurance, and self-reportedhealth conditions. In addition, linkingmultiple years of cancer registry data tomultiple survey years adds alongitudinal component to thecross-sectional survey data. For surveyparticipants interviewed before theircancer diagnosis, researchers are able toexamine characteristics and risk factorsassociated with future cancer diagnoses,cancer stage at diagnosis, and survivaltime. For survey participants interviewedafter a cancer diagnosis, researchers areable to examine issues related to cancersurvival. Although NHIS asksparticipants if they have ever been

Page 1

[ Series 2, No. 167

diagnosed with cancer, NHIS–FCDSlinked data allow researchers to examinethe health characteristics and quality oflife of cancer survivors in much greaterdetail. For example, FCDS data provideinformation about cancer stage atdiagnosis, which has a substantialimpact on cancer survival and futurequality of life. As a result, linkagebetween these data sources providesresearchers with an opportunity toconduct a wide array of cancer studiesand examine relationships not possiblewith either data set alone.

Preliminary evaluation of the datadescribes the results of the linkage andcompares the cancer types anddemographic characteristics ofparticipants in the linked file to cancerpatients represented in FCDS data alone.In addition, some of the unique analyticissues presented by the linkage ofmultiple years of national health surveydata to multiple years of state-specificcancer registry data are discussed.Specifically, the creation of Florida-specific survey weights, the potentialimpact of residential mobility on thelinked file, and NHIS questionnaire andsample design changes over time aredetailed. Finally, preliminary tabulationsof cancer variables by factors availablethrough NHIS are shown to illustrateuse of the linked file.

NHIS–FCDS data are restricted-usedata that can be accessed only byapproved researchers through the NCHSResearch Data Center (RDC). RDCprotects the confidentiality of surveyrespondents by limiting access toapproved research proposals, providing asecure work environment, and reviewingresearcher output for potential disclosurerisk.

Methods

Data Files

National Health InterviewSurvey

NHIS is an annual, cross-sectionalhousehold survey of the U.S. civiliannoninstitutionalized population (CNI)

conducted by NCHS since 1957. NHISserves as a principal source ofinformation on the health of the nation.In addition to detailed demographicinformation, NHIS collects a wide rangeof health-related information includinghealth status and limitations, health careaccess and utilization, health insurance,and health behavior information(including specific supplements oncancer health behavior).

Prior to 1997, a core questionnaire,including demographic information andbasic health questions, was used foreveryone in the household, withsupplemental questionnaires used formore specific health topics. In 1997,NHIS underwent a sample redesign andquestionnaire revision. Whiledemographic and basic healthinformation were still collected oneveryone in the household, one adultand one child within each householdwere randomly selected to conduct amore detailed survey on more specifichealth topics (1). These changes areimportant for the linked data becausenot all variables have been collectedfrom the entire sample or collected overthe entire file time period. This isespecially true for many cancer-relatedvariables that are not available fromcancer registry data, such as tobacco useand cancer screening history. On theother hand, a number of variables havebeen collected across the entire surveytime period and are not typicallyavailable from cancer registry data.These include education level, income,health insurance coverage, occupationand industry, and body mass index(derived from self-reported height andweight). However, data collection inmany of these fields has changedslightly over time. Table 1 providesexamples of variable availability bysurvey year for selected topics in NHIS.

Florida Cancer Data System

Funded by Florida and CDC’sNational Program of Cancer Registries,FCDS is the incidence cancer registryfor Florida. Cancer registries areresponsible for collecting, managing,and analyzing data on incident cancercases and cancer deaths, and they areessential in monitoring progress in

cancer prevention and control. Withmore than 110,000 newly diagnosedcancer cases per year, FCDS is the thirdlargest registry in the country andrepresents approximately 6% of all U.S.cancer cases. FCDS is ‘‘gold certified’’by the North American Association ofCentral Cancer Registries based on thetimeliness and completeness of the datacollected. FCDS began collecting data in1981; however, the data system containscancer records dating to the 1950s,either from voluntary reporting to theFlorida Department of Health prior toFCDS or from previous diagnoses ofcancer among those with cancersdiagnosed from 1981 forward (2).

Cancer registries collect informationon the type of cancer, extent (i.e., stageat diagnosis), initial treatment, and basicdemographic characteristics. Cancerregistries collect data at the tumor level;therefore, persons diagnosed with morethan one incident cancer are included inthe registry for each tumor.Approximately one-fifth of persons inFCDS have been diagnosed withmultiple incident cancers. Tumors thatoccur as a result of cancer cellsspreading (metastasis) from the originallocation are not incident cancers. Cancerregistries are required to collect directidentifiers in order to eliminate duplicaterecords, consolidate multiple records,and conduct follow-up on vital statusthrough data linkage with the NCHSNational Death Index.

Linkage Approval andMethods

Designated Agent Agreementand project approval

NHIS data are protected by Section308(d) of the Public Health Service Actand by the Confidential InformationProtection and Statistical Efficiency Act(CIPSEA). Under CIPSEA legislation,FCDS was required to sign a DesignatedAgent Agreement, and all FCDSanalysts were required to completeNCHS confidentiality training andpaperwork before accessing confidentialNCHS data to conduct the linkage.Additionally, all NCHS record linkageactivities must be reviewed and

Series 2, No. 167 [ Page 3

approved by the NCHS Ethics ReviewBoard (ERB). ERB is a formallyappointed ethics review committeeestablished to ensure that researchinvolving human participants protectstheir rights and welfare and conforms tofederal regulations. ERB grantedapproval for the NHIS–FCDS datalinkage in May 2008. The project thenunderwent review and approval by theFlorida Department of HealthInstitutional Review Board, and thelinkage was initiated in spring 2009.

Linkage eligibility

To be eligible for FCDS linkage,NHIS participants had to providesufficient direct identifiers [e.g., name,date of birth, and Social Securitynumber (SSN)] and had not refused tohave their data linked. Linkageeligibility requirements varied acrossNHIS years due to changes in how ERBdetermined whether a survey participantwas deemed eligible or ineligible forlinkage. Before 2007, participants whorefused to provide direct identifiers wereimplicitly considered to have refusedrecord linkage. The refusal rateincreased between 1997 and 2006,reducing the number of NHISparticipants eligible for record linkages.Beginning in 2007, NHIS successfullyreduced linkage refusal rates by addinga short introduction prior to the requestfor SSN, requesting only the last fourdigits of SSN rather than the full ninedigits, and asking participants who didnot provide SSN (or Medicare number)for their explicit permission to link toadministrative records. Demographicdifferences have been found betweenthose who refuse to provide an SSN andthose who do not (3), which couldpotentially bias estimates calculatedfrom the linked data. Currently, thisissue is treated as a nonresponse bias,and analytic methods are used to adjustsample weights in an attempt to accountfor differences (4).

NCHS submission filepreparation and description

NCHS provided FCDS withelectronic data files containing directidentifiers and an NCHS-created control

number. The files used for matching didnot contain the NCHS survey public-useidentifiers available from the public-usefiles. For the NHIS–FCDS linkage, allNHIS survey participants meeting theeligibility criteria, including non-Floridaresidents, were submitted for linkage. Amatch was not requested for participantsdeemed ineligible for the linkage. Thedata were securely sent to FCDS on anencrypted CD.

Linkage process

FCDS used LinkPlus software (5) toconduct the linkage. LinkPlus, whichCDC developed for cancer registries,uses probabilistic linkage methods. Inaddition to finding exact matches,probabilistic methods can identifymatching records when not all matchingvariables agree or when variables onlypartially match. Records are givenscores for the number of matching orpartially matching fields, with certainfields given more weight than others.For example, matching on an SSN isgiven more weight than matching onplace of residence.

For this linkage, matching scoreswere based on SSN (nine and fourdigits), date of birth, last name, firstname, middle initial, sex, race (white,black, American Indian or AlaskaNative, Asian, other), five-character cityname, and five-digit zip code. Thecity-name field was truncated to fivecharacters to improve matchingefficiency and reduce errors due tomisspellings or nonstandardized citynames. To account for possibledifferences in first and last names due tomisspellings, alternative spellings, andclerical errors, LinkPlus includes theNew York State Identification andIntelligence System, or NYSIIS, whichis commonly used in data linkage toconvert names to phonetic codes.

Matches with scores of 30 andabove were considered true matches,and scores below 17 were consideredfalse matches. Matches with scores from17 through 29 were manually reviewedto determine if they were true matches.To assess the number of false or missedmatches, 100 consecutive records justabove and below the cutoffs werereviewed. Of those above the

upper-score threshold, one false-positivematch was identified. Of those belowthe lower threshold, matches appeared tobe missed in six records. The sixfalse-negative matches were all womenand fell just below the score cutoff,primarily due to inconsistencies betweenlast names on the two files.

Processing of return file atNCHS

The return file containing theextracted FCDS data for matched NHISrecords was remerged with thepublic-use survey identifiers to enablethe addition of survey data, and thenintermediate files were destroyed. Thereturn file was processed to create onerecord per person. Because so fewchildren were linked with FCDS, the filewas limited to adults aged 18 and overas of their NHIS interview. Variableswere added to provide linkage eligibilitystatus of survey participants and identifysurvey participants who moved after thesurvey interview. In addition, sampleweights appropriate to the linked datawere created and added to the file.

Creation of Florida-specificsample weights

NHIS public-use and restricted-useanalytic survey files include annualsample-weight variables, which areinversely proportional to eachparticipant’s selection probability andhave been adjusted for oversampling ofspecific subgroups and differentialnonresponse. NHIS sample weights areintended to be representative of the CNIpopulation of the United States in eachsurvey year.

Although other approaches arepossible, sample weights for the linkedNHIS–FCDS data were created torepresent the CNI population of Florida.The Florida-specific sample weightswere calculated for all linkage-eligibleNHIS participants, regardless of whetherthey linked with FCDS. Because thesample weights created for the linkedNHIS–FCDS data were representative ofthe Florida CNI population in eachsurvey year, survey participantsinterviewed in other states consequentlyreceived a sample weight of zero, even

[ Series 2, No. 167

if they linked with FCDS. As a result,using the Florida CNI weightingstrategy, survey respondents interviewedoutside of Florida who later moved toFlorida where they were diagnosed withcancer drop out of weighted analyses.

Two sets of Florida CNI sampleweights were created. The first werecreated for all survey participants aged18 and over in all NHIS years. Anadditional sample-weight variable wascreated to account for the sampleredesign and Sample Adult filebeginning in 1997. Because not allNHIS participants were eligible for thelinkage, NHIS sample weights forFlorida residents at the time of surveywere first adjusted for linkageineligibility using PROC WTADJUST inSUDAAN software (4). The adjustedweights were then poststratified to theestimated annual CNI population ofFlorida for the corresponding surveyyear. Two different sex- and age-specific(18–39, 40–64, and 65 and over)estimates of the annual Florida CNIpopulation were used to create thepoststratified weights. One method usedpopulation estimates directly from NHISsample weights for each survey yearamong Florida residents. Because ofsubstantial year-to-year variabilityamong population subgroups in earlieryears of NHIS, a second method usedtotal Florida population data from theFlorida Department of Health’s Officeof Health Statistics and Assessment(available from: http://www.floridacharts.com/charts/default.aspx). Thismethod estimated the CNI proportion ofthe total population using NHIS datafrom 1997–2009, when CNI populationestimates were more stable. The firststep compared the Florida CNIpopulation data from NHIS to the totalFlorida population for these 13 years tocalculate the average percent CNIpopulation over the time period. Thesecond step applied the average percentCNI to the total Florida population dataover the entire time period covered bythe linkage, 1986–2009, to estimate theCNI population for each survey year.The two sample weights were highlycorrelated (r = 0.99).

Evaluation of Linked FileTo conduct the evaluation,

Integrated Health Interview Series(IHIS) data, which harmonizes NHISdata across survey years (6), weremerged with the linked file to obtainseveral NHIS variables. For overallevaluations, the following NHISvariables collected at interview wereused: sex, age (18–39, 40–64, and 65and over), race and ethnicity (non-Hispanic white, non-Hispanic black,Hispanic, and all other races andethnicities), educational level [less thanhigh school degree, high schoolgraduate/General EducationalDevelopment (GED) certificate/somecollege, post-high school degree], andself-rated health status (excellent/verygood/good, fair/poor). IHISdocumentation contains furtherinformation on how the collection andsubsequent harmonized coding of raceand ethnicity and education informationhave changed over time (6). Except forthe presentation of unweighted numbers,(see Results and Detailed Tables),analysis of the file accounted for thecomplex survey design, and estimateswere weighted using the Florida-specificsample weights created for this file.Comparisons of unweighted percentageswere not statistically tested.

Evaluation of NCHS–FCDS datafirst consisted of an overall descriptionof NHIS participants who were linkedwith FCDS, including the unweightedand weighted distributions ofdemographic and cancer characteristicsin the linked population. Because linkedNHIS participants who were notresidents of Florida at the time ofsurvey receive a zero weight and dropout of weighted analyses, an assessmentwas made of how their characteristicsdiffered from the linked NHISparticipants who were Florida residentsat the time of survey. Additionalweighted examinations of cancer typesand characteristics related to the linkagetime period were made by sequence ofcancer diagnosis and surveyparticipation.

To assess the representativeness ofthe linked cancer cases in NHIS–FCDSdata relative to all Florida cancer cases,

the demographic characteristics andcancer types were examined for Floridalinked survey participants and theunlinked FCDS data for 2006–2010, themost current 5 years of data availablefor analysis. The evaluation was limitedto survey participants who had their firstcancers diagnosed during 2006–2010and those diagnosed after surveyparticipation, to examine incidentcancers only. Because the two groupsare not statistically independent,comparisons were not statistically tested.

Due to the potential loss of samplesize from linkage ineligibility, weightsthat were specific to Florida residents atthe time of survey, and theunavailability of certain variables duringthe linkage time period, an examinationwas made of how these factors couldaffect an analysis of the linkedNHIS–FCDS data. Specifically, thereduction in sample size afteraccounting for each of these factors wasdocumented, using a hypotheticalanalysis of female breast cancer andtobacco use as an example.

As an initial analysis of the linkeddata, basic statistics available from thelinked NHIS–FCDS data but notavailable from either data source alonewere calculated to examine preliminaryrelationships between demographic andhealth factors and cancer. For thisanalysis, an examination was made ofthe relationship between cancersdiagnosed at a late stage (i.e., withtumor having spread regionally or to adistant site in the body) anddemographic characteristics of thosewho were linked with FCDS. Inaddition, selected health characteristicswere compared between cancersurvivors (survey participants linkedwith FCDS prior to their surveyinterview) and survey participants whohad not been diagnosed with cancer(participants not linked with FCDS). Allestimates were calculated overall and bysex. Other information from IHIS usedfor this evaluation included maritalstatus (married, not married/widowed/separated), smoking status (never,current, former), and poverty statusbased on income above or below thefederal poverty level. Poverty

http://www.floridacharts.com/charts/default.aspx


status used the POORYN variable fromIHIS, which is based on income andfamily size and its relation to thepoverty threshold set by the U.S. CensusBureau for each survey year (6). Incomeis not imputed for this variable; as aresult, data were missing from 676Florida NHIS participants who werelinked with FCDS (11.4%). Thedemographic variables and activitylimitation variable were available for allyears of NHIS. Because smoking statuswas available only for select yearsbefore 1997 and was not alwayscollected on the full sample, smokingestimates reported here were based onsample adults in 1997–2009 NHIS data.

Results

Linked NHIS–FCDS DataFiles

Characteristics of linked surveyparticipants

NHIS participants aged 18 and overtotaled 1,708,723 during 1986–2009.Approximately one-third of theparticipants were ineligible for thelinkage (n = 575,327). After runningLinkPlus and performing manual review,NHIS participants who were linked withFCDS totaled 8,210. For this report,linked records with cancers missing adate of diagnosis (n = 59) or cancersreported prior to 1981 with noadditional cancers diagnosed after 1981were excluded (n = 41).

Of the remaining 8,110 linkedparticipants, 1,443 (17.8% unweighted)were linked with more than one tumorin FCDS (Table 2). Approximatelyone-half (unweighted) of the linkedsurvey participants were men (50.2%),approximately 90% were aged 40 andover at the time of survey (89.9%), andmore than 70% were non-Hispanicwhite (73.3%). Almost two-thirds(unweighted) of the linked participants(65.0%) were interviewed during1986–1996; the remaining 35.0% wereinterviewed during 1997–2009, after theNHIS sample and questionnaireredesign. By time period of cancer

diagnosis, 46.0% (unweighted) of linkedparticipants had cancers diagnosedduring 1996–2005.

When the Florida-specific weightswere applied, limiting the analysis toFlorida residents at the time of survey,the percentage of linked participantswho were Hispanic decreased and thepercentage of participants aged 65 andover increased (Table 2). This differencebetween the unweighted and weighteddistributions is likely a result of theweights accounting for the oversamplingof Hispanic participants in the NHISsurvey design. In addition, limiting thesample to Florida residents at the timeof survey resulted in a larger percentageof linked survey participants fromsurvey years 1997–2009 and with 0–5years between survey and cancerdiagnosis.

The total number of NHIS–FCDSlinked participants interviewed outsideof Florida was 1,829, 22.6%(unweighted) of all linked participants.Although not statistically tested, thesample of non-Florida linked surveyparticipants had higher percentages(unweighted) of men, non-Hispanicwhite persons, and post-high schooldegrees than the Florida linked surveyparticipants. Possibly due to beinghealthy enough to move, 85.3%(unweighted) of the sampled non-Floridalinked survey participants reportedexcellent/very good/good health status atthe time of survey; 76.7% (unweighted)of Florida linked survey participantsreported excellent/very good/good healthstatus. The majority of non-Floridalinked participants were interviewedduring 1986–1996 and close to 80%(unweighted) of their cancers (78.9%)were diagnosed during 1996–2010.

Table 3 presents the number anddistribution of cancers among linkedFlorida participants by sequence ofcancer diagnosis related to surveyparticipation. The cancer cases werelimited to the first cancer diagnosisamong those who were linked to morethan one cancer. The majority of linkedsurvey participants were diagnosed withcancer after they had participated in thesurvey, which allows researchers toexamine the characteristics and healthbehaviors of survey participants beforethey were diagnosed with cancer.

Among five commonly diagnosedcancers (prostate, female breast, lung,colorectal, and bladder), the number ofcases ranged from 200 to 724 when thecancer was diagnosed after surveyparticipation. More than 60% of linkedparticipants with cancer diagnosed afterthe survey (63.6%) were interviewedduring 1986–1996, and 84.3% of thecancers were diagnosed during1996–2010. When examining the lengthof time between survey and cancerdiagnosis, 44.3% of the linked surveyparticipants were diagnosed within 5years after the interview.

A total of 1,908 linked surveyparticipants had cancer diagnosed beforeparticipating in NHIS (Table 3). Thenumber of cases for the specific cancertypes ranged from 89 to 403. Consistentwith poor survival among patientsdiagnosed with lung cancer, theproportion of lung cancer survivors wassubstantially smaller for linked surveyparticipants diagnosed with cancerbefore rather than after the surveyinterview. Additionally, the slightlyhigher percentages of breast and prostatecancers among cancer survivors may bedue to the fact that many of thesecancers are diagnosed at an early stageand have high survival rates whendetected early (7). Approximatelytwo-thirds (weighted) of linked surveyparticipants with a cancer diagnosisprior to the survey (66.8%) were aged65 and over at the time of survey. Bytime period, 67.5% (weighted) of linkedsurvey participants with a previouscancer diagnosis were interviewedduring 1997–2009, and 79.9% of thecancers were diagnosed during1986–2005. More than one-half(weighted) were interviewed within 5years from their cancer diagnosis date inFCDS (53.5%).

Comparison with unlinkedFCDS

By inspection, the five commonlydiagnosed cancers among Florida linkedsurvey participants diagnosed during2006–2010 were consistent withunlinked FCDS data, with someexceptions. For example, female breastcancer was the most commonlydiagnosed cancer among Florida survey

[ Series 2, No. 167

participants, and lung cancer was themost commonly diagnosed cancer in theunlinked FCDS data (Table 4). By sex, asimilar percentage of men and womenare represented in the linked data, butthe percentage of men in the unlinkedFCDS data is 53.2%. The racial andethnic distribution of Florida linkedsurvey participants was generally similarto those with cancer diagnosis in theunlinked FCDS, except that theestimated percentage of non-Hispanicblack linked survey participants washigher than the correspondingpercentage in the unlinked FCDS.

Potential loss of analytic samplesize in linked NCHS–FCDSdata analysis

Although the large annual samplepopulations of NHIS make this linkagepossible, not all participants can beincluded in analyses of the linked datafor various reasons. This loss of analyticsample size can substantially reduce the

Number of female Florida N1986–200

49,550

Number of linkage-eligiblNHIS participants, 1

30,924

Not asked about tobacco usen = 16,725

No cancer(not linked with FCDS):

27,714

Answered tobacco questionsfrom available survey years1:

10,989

1See Table 1 of this report for details on the availability of tobacco qNOTE: NHIS is National Health Interview Survey; FCDS is Florida SOURCE: CDC/NCHS, 1986–2009 National Health Interview SuData System.

Figure 1. Number of female participants availabsurvey tobacco questions: 1986–2009 National1981–2010 Florida Cancer Data System

power of certain analyses or the abilityto examine differences betweensubgroups, and can introduce bias ifdifferences exist between those who areincluded in the sample and those whoare not. For example, consider tobaccouse and breast cancer risk (Figure 1): Ofthe 911,262 women aged 18 and over inNHIS during 1986–2009, 49,550 wereinterviewed in Florida. Of those, 18,626were ineligible for linkage. Afterexcluding women diagnosed with othercancers in FCDS, 27,714 womenwithout cancer and 724 womendiagnosed with breast cancer wereavailable for analysis. However, toconduct an analysis based on tobaccouse, which was not collected every yearor from the full sample, data areavailable only for 10,989 womenwithout cancer and 259 with breastcancer. Similarly, questions regardingcancer screening and family history ofcancer were included only in selectyears with NHIS cancer supplementalquestionnaires. Of the same 724 breast

HIS participants,9:

e female Florida986–2009:

Not linkage-eligiblen = 18,626

Excluding other cancersn = 2,486

Not asked about tobacco usen = 465

Answered tobacco questionsfrom available survey years1:

259

Breast cancer(linked with FCDS):

724

uestions across NHIS years.Cancer Data System.rvey linked with 1981–2010 Florida Cancer

le for breast cancer risk analysis usingHealth Interview Survey linked with

cancer cases, information on previousmammography use is available only for112 women.

Initial analysis of linkedNCHS–FCDS data

The weighted percentage ofparticipants diagnosed with late-stagecancer in the linked NHIS–FCDS datawas greater with decreasing educationlevel for both men and women(Figure 2, Table 5). Men (4.0%) andwomen (2.0%) with a post-high schooldegree were less likely to be diagnosedwith late-stage cancer compared withthose having less than a high schooldegree (8.0% men and 6.6% women).Similarly, men (5.1%) and women(3.8%) living above the federal povertylevel were less likely to be diagnosedwith a late-stage cancer compared withthose below the federal poverty level(12.6% men and 9.0% women), but thisdifference was statistically significantonly among men (Figure 3, Table 5). Bymarital status, only married women weresignificantly less likely to be diagnosedwith a late-stage cancer (2.9%)compared with those who were notmarried/divorced/separated (6.2%)(Figure 4, Table 5); however, marriedmen were significantly more likely to bediagnosed with late-stage cancer thanmarried women.

When comparing healthcharacteristics of cancer survivors(Florida NHIS participants who werelinked to FCDS prior to survey) withthose who had not been diagnosed withcancer (Florida participants not linkedwith FCDS before or after interview),the weighted percentage of cancersurvivors reporting they were infair/poor health (33.2%) was more thandouble that of participants who had notbeen diagnosed with cancer (13.0%)(Figure 5, Table 6). This association wastrue for both men and women. Cancersurvivors were also more likely to reporthaving some type of activity limitationcompared with those not diagnosed withcancer (Figure 6, Table 6). Both men(40.1%) and women (36.1%) who werecancer survivors were more than twiceas likely to report having an activitylimitation compared with those who had

* Figure does not meet standards of reliability or precision; relative standard error is greater than 30%.NOTES: Late-stage cancers include tumors that have spread regionally or to distant sites of the body at diagnosis.GED is General Educational Development high school equivalency diploma. Confidence interval is 95%. SOURCE: CDC/NCHS, 1986–2009 National Health Interview Survey linked with 1981–2010 Florida CancerData System.

Overall Men Women

Wei

ghte

d pe

rcen

t

0

5

10

15Less than high school High school graduate/

GED/some collegePost-high schooldegree

*

Figure 2. Percentage of diagnosed late-stage cancers, by education level and sex:1986–2009 Florida National Health Interview Survey participants aged 18 and over linkedwith 1981–2010 Florida Cancer Data System after interview

* Figure does not meet standards of reliability or precision; relative standard error is greater than 30%. NOTES: Late-stage cancers include tumors that have spread regionally or to distant sites of the body at diagnosis. Confidence interval is 95%. SOURCE: CDC/NCHS, 1986–2009 National Health Interview Survey linked with 1981–2010 Florida CancerData System.

0

5

10

15

20

Overall Men Women

Wei

ghte

d pe

rcen

t

Below poverty level Above poverty level

*

Figure 3. Percentage of diagnosed late-stage cancers among those above and belowfederal poverty level, by sex: 1986–2009 Florida National Health Interview Surveyparticipants aged 18 and over linked with 1981–2010 Florida Cancer Data System afterinterview


not been diagnosed with cancer (14.4%men and 16.4% women). Among thoseinterviewed during 1997–2009, thepercentage of cancer survivors whowere current smokers at the time ofsurvey was less than 11% for both menand women (Figure 7, Table 6),

approximately one-half the percentageamong women (20.2%) and less thanone-half the percentage among men(25.0%) who had not been diagnosedwith cancer.

Discussion

The linked NHIS–FCDS data allowresearchers to follow NHIS surveyparticipants longitudinally to examinefactors associated with future diagnosisof cancer and to assess thecharacteristics and quality of life amongcancer survivors. FCDS data providedetailed information on the type ofcancer and stage at diagnosis, whileNHIS provides detailedsociodemographic information andextensive self-reported data on healthcharacteristics, including healthconditions, access to health care, andhealth care utilization that are notavailable from cancer registry data. Forexample, using education and povertylevel, the linked data allow researchersto examine socioeconomiccharacteristics at the individual levelrather than relying on census tract-levelestimates, which is the commonapproach. Using self-rated health andthe presence of an activity limitation asanother example demonstrates the dataset’s ability to assess morbidity amongcancer survivors.

Overall, the linked NHIS–FCDSdata had a similar distribution of majorcancer types compared with unlinkedFCDS data, with a slight over-representation of breast cancer, whichmay be explained by a higherpercentage of women in the linked data.The linked NHIS–FCDS data also had aslightly higher percentage of non-Hispanic black cancer cases than theunlinked FCDS data. This could resultfrom oversampling in NHIS that wasnot fully adjusted for in the newweights.

Residential mobility proved to beone of the major challenges with thelinked NCHS–FCDS data. While thecurrent weighting strategy (creatingweights to the year of the survey) isvalid, it results in a nontrivial reductionin sample size by giving a zero weightto linked NHIS participants who wereinterviewed in other states but werediagnosed with cancer in Florida.Alternative weighting strategies arebeing investigated to reduce the loss ofsample size. However, noticeabledifferences were observed between the

Married Not married/widowed/separated

NOTES: Late-stage cancers include tumors that have spread regionally or to distant sites of the body at diagnosis. Confidence interval is 95%. SOURCE: CDC/NCHS, 1986–2009 National Health Interview Survey linked with 1981–2010 Florida CancerData System.

Wei

ghte

d pe

rcen

t

0

5

10

15

Overall Men Women

Figure 4. Percentage of diagnosed late-stage cancers, by marital status and sex:1986–2009 Florida National Health Interview Survey participants aged 18 and over linkedwith 1981–2010 Florida Cancer Data System after interview

NOTES: Cancer survivors are survey participants whose data were linked with the Florida Cancer Data Systembefore their survey interview; those not diagnosed with cancer were not linked with the data system.Confidenceinterval is 95%. SOURCE: CDC/NCHS, 1986–2009 National Health Interview Survey linked with 1981–2010 Florida CancerData System.

Overall Men Women

Wei

ghte

d pe

rcen

t

0

10

20

30

40

50Cancer survivor Not previously diagnosed with cancer

Figure 5. Percentage of self-rated fair and poor health status, by cancer diagnosis andsex: 1986–2009 Florida National Health Interview Survey participants aged 18 and overlinked with 1981–2010 Florida Cancer Data System before interview (cancer survivor) andnot linked (not diagnosed with cancer)

Page 8 [ Series 2, No. 167

demographic characteristics ofnon-Florida residents who were linkedwith FCDS and the demographiccharacteristics of linked Floridaparticipants, and caution is needed tonot bias inference by including

non-Florida residents withoutappropriate methods.

Out-migration is another potentialproblem in the data set. In this case, allsurvey participants from Florida aregiven a Florida sample weight for

analysis of the linked NHIS–FCDS data,but it is not possible to identify if someof these participants moved or werediagnosed with cancer in another state.The number of survey participants whowere Florida residents at the time ofinterview but moved out of Florida afterthe interview is unknown, but thenumber may not be trivial. Of the 3,526deaths among Florida surveyparticipants who were linked withFCDS, 7.8% (n = 275) died in anotherstate.

The loss of NHIS participants fromthe study population, due either tolinkage ineligibility or migration, mayadversely affect the generalizability ofstudy inferences or the ability tocalculate accurate prevalence estimates.Although the sample weights used forthe estimates in this report wereadjusted for linkage ineligibility andcalibrated to the Florida population, it isunclear how differences incharacteristics may influence resultsbetween those who are linkage eligibleand those ineligible, as well as betweenthose who migrated into or out ofFlorida compared with those who wereconstant Florida residents.

Using 23 years of NHIS dataallowed for a sufficient number ofFlorida NHIS participants to be linkedwith FCDS to conduct many types ofdata analyses. However, changes in thequestionnaire during this time periodlimit some types of analyses as well asthe ability to control for potentialconfounders. While NHIS data from1997 forward are more consistent, thenumber of cancers diagnosed from thosesurvey years limits the ability toexamine many individual types ofcancer; linkage eligibility criteria andthe provision of SSN also changedduring the time period, further reducingthe study population. In addition, surveyresponses may not be representative fortime-variant characteristics because ofthe potential for a substantial number ofyears to elapse between the survey andcancer diagnosis. Among participantsdiagnosed with cancer after their surveyparticipation, nearly 60% had more than5 years elapse between survey andcancer diagnosis (55.7%). Fortime-variant characteristics such as

NOTES: Cancer survivors are survey participants whose data were linked with the Florida Cancer Data Systembefore their survey interview; those not diagnosed with cancer were not linked with the data system. Confidenceinterval is 95%. SOURCE: CDC/NCHS, 1986–2009 National Health Interview Survey linked with 1981–2010 Florida CancerData System.

Overall Men Women

Wei

ghte

d pe

rcen

t

0

10

20

30

40

50Cancer survivor Not previously diagnosed with cancer

Figure 6. Percentage having any activity limitation, by sex: 1986–2009 Florida NationalHealth Interview Survey participants aged 18 and over linked with 1981–2010 FloridaCancer Data System before interview (cancer survivor) and not linked (not diagnosed withcancer)

NOTES: Current smokers have smoked at least 100 cigarettes in their lifetime and still currently smoke. Cancersurvivors are survey participants whose data were linked with the Florida Cancer Data System before their survey interview; those not diagnosed with cancer were not linked with the data system. Confidence interval is 95%.SOURCE: CDC/NCHS, 1997–2009 National Health Interview Survey linked with 1981–2010 Florida CancerData System.

Overall Men Women

Wei

ghte

d pe

rcen

t

Cancer survivor Not previously diagnosed with cancer

0

5

10

15

20

25

30

Figure 7. Percentage of current smokers, by sex: 1997–2009 Florida National HealthInterview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer DataSystem before interview (cancer survivor) and not linked (not diagnosed with cancer)


health insurance coverage or meetingcancer screening guidelines, the surveyresponse may not be representative ofthe full interval from or until cancerdiagnosis.

Conclusion

The pilot linkage of NHIS andFCDS demonstrates the feasibility oflinking national-level sample survey

data with state-based cancer registrydata. The linked data set provides theability to conduct analyses that are notpossible with either data set alone.NHIS adds participant-levelcharacteristics that are highly desiredamong researchers using cancer registrydata, and FCDS provides detailedoutcome information for those who arediagnosed with cancer after surveyparticipation and detailed baselineinformation for cancer survivors whoparticipate in the survey after theircancer diagnosis. The number of yearsof data linked in this project createssome analytic challenges, and limitationsexist in the types of analyses that can beconducted. For example, the loss ofsample due to linkage eligibility andmobility in and out of Florida make thedata more suited for examiningrelationships among factors than forother uses, such as prevalence estimates.The methodological issues encounteredwith these data may be relevant to othernational and state data linkages.

References

1. National Center for Health Statistics.1997 National Health Interview Survey(NHIS) public use data release. 2000.Available from: ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/1997/srvydesc.pdf.

2. Florida Cancer Data System (FCDS).2014. Available from: http://fcds.med.miami.edu/inc/welcome.shtml[Accessed February 18, 2014].

3. Miller DM, Gindi RM, Parker JD.Trends in record linkage refusal rates:Characteristics of National HealthInterview Survey participants whorefuse record linkage. Paper presentedat the 2011 meeting of the JointStatistical Meeting, July 30–August 4,2011, Miami Beach, FL. 2011.

4. Judson DH, Parker JD, Larsen MD.Adjusting sample weights forlinkage-eligibility using SUDAAN.Hyattsville, MD: National Center forHealth Statistics. 2013. Available from:http://www.cdc.gov/nchs/data/datalinkage/adjusting_sample_weights_for_linkage_eligibility_using_sudaan.pdf.

ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/1997/srvydesc.pdf

http://fcds.med.miami.edu/inc/welcome.shtml

http://www.cdc.gov/nchs/data/datalinkage/adjusting_sample_weights_for_linkage_eligibility_using_sudaan.pdf

[ Series 2, No. 167

5. Centers for Disease Control andPrevention. Registry Plus Link Plus.Available from: http://www.cdc.gov/cancer/npcr/tools/registryplus/lp.htm[Accessed July 15, 2014].

6. Minnesota Population Center and StateHealth Access Data Assistance Center.Integrated Health Interview Series(Version 5.0). Minneapolis, MN:University of Minnesota. 2012.Available from: https://www.ihis.us/ihis/[Accessed March 12, 2014].

7. American Cancer Society. Cancer Facts& Figures 2013. Atlanta, GA. 2013.

http://www.cdc.gov/cancer/npcr/tools/registryplus/lp.htm

https://www.ihis.us/ihis/

Table 1. Availability of selected questionnaire topics, by survey year: National Health Interview Survey, 1986–2009

Questionnaire topic

Survey year

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Tobacco use . . . . . . . . . . . . . . . . . 1X . . . . . . X X 2X 3X 1X 1X . . . 4X 4X 4X 4X 4X 4X 4X 4X 4X 4X 4X 4X 4XAlcohol use . . . . . . . . . . . . . . . . . X X . . . X X X . . . . . . . . . . . . 4X 4X 4X 4X 4X 4X 4X 4X 4X 4X 4X 4X 4XFamily history of cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4X . . . . . . . . . . . . 4X . . . . . . . . . . . .Time since last mammogram. . . . . . . X . . . . . . X . . . X . . . . . . . . . . . . . . . . . . 4X 4X . . . . . . 4X . . . 4X . . . . . . 4X . . .Time since last colorectal test . . . . . . 5X . . . . . . . . . . . . 5X . . . . . . . . . . . . . . . . . . . . . 4X . . . . . . 4X . . . 4X . . . . . . 4X . . .Body mass index . . . . . . . . . . . X X X X X X X X X X X X X X X X X X X X X X X XEducation level . . . . . . . . . . . . X X X X X X X X X X X X X X X X X X X X X X X XIncome . . . . . . . . . . . . . . . . . X X X X X X X X X X X X X X X X X X X X X X X XHealth insurance coverage . . . . . X X X X X X X X X X X X X X X X X X X X X X X XOccupation or industry. . . . . . . . X X X X X X X X X X X X X X X X X X X X X X X X

. . . Category not applicable.1One-half of sampled adults aged 18 and over.2All sampled adults, with partial-year data collection.3One-half of sampled adults, with partial-year data collection.4Sampled adults only.5One-half of sampled adults aged 40 and over, with partial-year data collection.

SOURCE: CDC/NCHS, National Health Interview Survey, 1986–2009.

Series

2,N

o.167

[P

age11

Table 2. Number and percentage of linked participants, by demographic and cancer characteristics: 1986–2009 Florida National HealthInterview Survey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System

Characteristic

All linked survey participants1 Florida residents at time of survey1 Non-Florida residents at time of survey1

Unweightednumber

Unweightedpercent

Unweightednumber

Unweightedpercent

Weightedpercent

Unweightednumber

Unweightedpercent

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8,110 100.0 6,281 100.0 100.0 1,829 100.0

Number of linked tumors

One. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6,667 82.2 5,147 82.0 81.9 1,520 83.1More than one . . . . . . . . . . . . . . . . . . . . . . 1,443 17.8 1,134 18.0 18.1 309 16.9

Sex

Male . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4,074 50.2 3,110 49.5 50.6 964 52.7Female. . . . . . . . . . . . . . . . . . . . . . . . . . . 4,036 49.8 3,171 50.5 49.4 865 47.3

Age (years)

18–39 . . . . . . . . . . . . . . . . . . . . . . . . . . . 821 10.1 607 9.7 9.2 214 11.740–64 . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,784 46.7 2,730 43.5 42.1 1,054 57.665 and over . . . . . . . . . . . . . . . . . . . . . . . . 3,505 43.2 2,944 46.9 48.7 561 30.7

Race and ethnicity

Hispanic . . . . . . . . . . . . . . . . . . . . . . . . . . 1,066 13.1 971 15.5 10.9 95 5.2Non-Hispanic white . . . . . . . . . . . . . . . . . . . 5,944 73.3 4,330 68.9 76.4 1,614 88.2Non-Hispanic black . . . . . . . . . . . . . . . . . . . 978 12.1 881 14.0 11.1 97 5.3Other . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 1.5 99 1.6 1.6 23 1.3

Education level

Less than high school. . . . . . . . . . . . . . . . . . 1,826 22.7 1,523 24.5 21.4 303 16.6High school graduate, GED, or some college . . . 4,562 56.8 3,498 56.3 57.7 1,064 58.4Post-high school degree . . . . . . . . . . . . . . . . 1,643 20.5 1,188 19.1 20.9 455 25.0Missing. . . . . . . . . . . . . . . . . . . . . . . . . . . 79 . . . 72 . . . . . . 7 . . .

Self-rated health

Excellent, very good, or good . . . . . . . . . . . . . 6,356 78.7 4,800 76.7 77.1 1,556 85.3Fair or poor . . . . . . . . . . . . . . . . . . . . . . . . 1,724 21.3 1,455 23.3 22.9 269 14.7Missing. . . . . . . . . . . . . . . . . . . . . . . . . . . 30 . . . 27 . . . . . . 4 . . .

Survey years

1986–1996 . . . . . . . . . . . . . . . . . . . . . . . . 5,271 65.0 3,782 60.2 52.6 1,489 81.41997–2009 . . . . . . . . . . . . . . . . . . . . . . . . 2,839 35.0 2,499 39.8 47.4 340 18.6

Cancer diagnosis2

Before 1986. . . . . . . . . . . . . . . . . . . . . . . . 443 5.5 354 5.6 5.7 89 4.91986–1995 . . . . . . . . . . . . . . . . . . . . . . . . 1,991 24.6 1,694 27.0 25.4 297 16.21996–2005 . . . . . . . . . . . . . . . . . . . . . . . . 3,705 45.7 2,812 44.8 45.0 893 48.82006–2010 . . . . . . . . . . . . . . . . . . . . . . . . 1,971 24.3 1,421 22.6 23.9 550 30.1

Years between survey and cancer diagnosis2

0–5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,307 40.8 2,861 45.5 47.6 446 24.46–10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,192 27.0 1,720 27.4 26.7 472 25.8More than 10 . . . . . . . . . . . . . . . . . . . . . . . 2,611 32.2 1,700 27.1 25.8 911 49.8

. . . Category not applicable.1Cancer identified through linkage with the Florida Cancer Data System could have occurred before or after survey participation.2Estimates limited to first cancer diagnosis among participants with more than one cancer diagnosis in the Florida Cancer Data System.

NOTE: GED is General Educational Development high school equivalency diploma.

SOURCE: CDC/NCHS, 1986–2009 National Health Interview Survey linked with 1981–2010 Florida Cancer Data System.


Table 3. Number of linked participants and weighted percent distribution of survey and cancer time-period characteristics, by sequence ofsurvey and cancer diagnosis: 1986–2009 Florida National Health Interview Survey participants aged 18 and over linked with 1981–2010Florida Cancer Data System

Characteristic

Florida residents at timeof survey with interview

before cancer diagnosis1

Florida residents at time ofsurvey with cancer diagnosis

before interview1

Unweightednumber

Weightedpercent

Unweightednumber

Weightedpercent

Total. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4,367 100.0 1,908 100.0

Cancer type

Female breast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724 16.4 401 20.9Prostate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666 14.8 403 21.2Lung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613 14.4 89 4.1Colorectal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 446 10.2 233 12.0Bladder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 4.4 120 6.4All others. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,716 39.8 662 35.4Missing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 . . . . . . . . .

Age at survey (years)

18–39. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 11.7 86 4.640–64. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,165 49.5 561 28.665 and over . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,681 38.9 1,261 66.8

Survey years

1986–1996. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,013 63.6 765 32.51997–2009. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,354 36.4 1,143 67.5

Cancer diagnosis

Before 1986 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 16.11986–1995. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753 15.7 941 43.21996–2005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,235 49.5 576 36.72006–2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,379 34.8 42 4.0

Years between survey and cancer diagnosis

0–5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,778 44.3 1,082 53.56–10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,239 27.2 481 25.8More than 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,350 28.5 345 20.7

. . . Category not applicable.1Estimates limited to first cancer diagnosis among participants with more than one cancer diagnosis in the Florida Cancer Data System.



Table 4. Number and percentage of cancer types and demographic characteristics: 1986–2009 Florida National Health Interview Surveyparticipants aged 18 and over linked with Florida Cancer Data System (cancers diagnosed during 2006–2010) compared with overall FloridaCancer Data System, 2006–2010

Characteristic

Florida residents at time of survey andcancer diagnosed during 2006–2010

Florida Cancer DataSystem, 2006–20101

Unweightednumber

Weightedpercent

Confidenceinterval2

Unweightednumber

Unweightedpercent

Cancer type

All cancers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1379 100.0 . . . 527,730 100.0Female breast . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 17.3 14.9–20.0 68,618 13.0Prostate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 13.0 11.2–15.0 73,746 14.0Lung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 13.0 11.3–15.0 81,023 15.4Colorectal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 9.7 7.6–12.4 49,592 9.4Bladder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2 3.3–5.2 24,978 4.7All others . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593 42.8 39.9–45.8 229,773 43.5

Sex

Male . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634 48.4 45.4–51.3 280,767 53.2Female . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745 51.6 48.7–54.6 246,963 46.8

Race and ethnicity

Hispanic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 12.4 10.1–15.2 68,008 12.9Non-Hispanic white . . . . . . . . . . . . . . . . . . . . . . . . . . 850 71.3 67.5–74.9 399,854 75.8Non-Hispanic black . . . . . . . . . . . . . . . . . . . . . . . . . . 234 14.2 11.9–16.8 49,712 9.4Other . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.1 1.4–3.1 10,156 1.9

. . . Category not applicable.1Available from: http://www.cancer-rates.info/naaccr/ [Accessed December 26, 2013].2Equals 95%.



Table 5. Cancer stage at diagnosis, by education level, poverty level, marital status, and sex: 1986–2009 Florida National Health InterviewSurvey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System after interview

Characteristic

Early stage1 Late stage2

NumberWeightedpercent

Confidenceinterval3 Number

Weightedpercent

Confidenceinterval3

Overall

Education level:Less than high school . . . . . . . . . . . . . . . . . . . . . . . 975 92.7 90.6–94.3 77 7.3 5.7–9.4High school graduate, GED, or some college . . . . . . . . . 2,318 95.2 94.1–96.1 108 4.8 3.9–5.9Post-high school degree . . . . . . . . . . . . . . . . . . . . . . 802 96.8 95.1–98.0 23 3.2 2.0–4.9

Poverty level:Below threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 89.5 85.1–92.7 29 10.5 7.3–14.9Above threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,302 95.5 94.6–96.4 150 4.5 3.7–5.4

Marital status:Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,800 95.8 94.9–96.6 117 4.2 3.5–5.1Not married, widowed, or separated. . . . . . . . . . . . . . . 1,311 93.4 91.9–94.6 92 6.6 5.4–8.1

MenEducation level:

Less than high school . . . . . . . . . . . . . . . . . . . . . . . 502 92.0 89.7–93.9 41 8.0 6.2–10.3High school graduate, GED, or some college . . . . . . . . . 1,061 94.3 92.8–95.5 61 5.7 4.5–7.2Post-high school degree . . . . . . . . . . . . . . . . . . . . . . 450 96.0 92.9–97.8 15 4.0 2.2–7.2

Poverty level:Below threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 87.4 81.7–91.5 15 12.6 8.5–18.3Above threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,667 94.9 93.5–96.1 83 5.1 3.9–6.5

Marital status:Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,565 94.8 93.2–96.0 79 5.2 4.0–6.8Not married, widowed, or separated. . . . . . . . . . . . . . . 458 92.6 89.9–94.7 37 7.4 5.4–10.1

WomenEducation level:

Less than high school . . . . . . . . . . . . . . . . . . . . . . . 473 93.4 90.4–95.5 36 6.6 4.5–9.6High school graduate, GED, or some college . . . . . . . . . 1,257 96.0 94.8–97.0 47 4.0 3.0–5.2Post-high school degree . . . . . . . . . . . . . . . . . . . . . . 352 98.0 95.5–99.1 8 *2.0 0.9–4.5

Poverty level:Below threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 173 91.0 82.7–95.6 14 *9.0 4.4–17.3Above threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,635 96.2 95.1–97.0 67 3.8 3.0–4.9

Marital status:Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,235 97.2 96.3–97.8 38 2.9 2.2–3.7Not married, widowed, or separated. . . . . . . . . . . . . . . 853 93.8 91.8–95.4 55 6.2 4.6–8.3

* Figure does not meet standards of reliability or precision; relative standard error greater than 30%.1Includes in situ and localized tumors at diagnosis.2Includes tumors that have spread regionally or to distant sites of the body at diagnosis.3Equals 95%.

NOTES: Education level was missing for 52 NHIS participants, income data were missing for 676, marital status was missing for 31, and cancer stage was missing for 19. GED is General EducationalDevelopment high school equivalency diploma.



Table 6. Percentage of smoking status, self-rated health, and any activity limitation, by sex: 1986–2009 Florida National Health InterviewSurvey participants aged 18 and over linked with 1981–2010 Florida Cancer Data System before interview (cancer survivor) and not linked(not diagnosed with cancer)

Characteristic

Cancer survivor Not previously diagnosed with cancer

NumberWeightedpercent

Confidenceinterval1 Number

Weightedpercent

Confidenceinterval1

Overall

Self-rated health:Fair or poor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669 33.2 30.1–36.6 6,756 13.0 12.4–13.7Excellent, very good, or good . . . . . . . . . . . . . . . . . . . 1,261 66.8 63.4–70.0 44,020 87.0 86.3–87.6

Any limitation:Yes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726 38.1 35.3–41.0 7,863 15.5 14.6–16.3No . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,213 61.9 59.0–64.7 43,074 84.5 83.7–85.4

Smoking status2:Never . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305 43.8 38.5–49.4 7,883 55.7 54.1–57.2Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 10.0 7.9–12.5 3,060 22.5 21.4–23.7Former. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312 46.2 41.0–51.4 2,922 21.8 20.8–22.9

MenSelf-rated health:

Fair or poor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 36.0 31.8–40.4 2,713 11.6 10.8–12.4Excellent, very good, or good . . . . . . . . . . . . . . . . . . . 596 64.0 59.6–68.2 20,445 88.4 87.6–89.2

Any limitation:Yes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377 40.1 36.4–43.9 3,383 14.4 13.5–15.5No . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577 59.9 56.1–63.6 19,840 85.6 84.5–86.5


WomenSelf-rated health:

Fair or poor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314 30.5 26.8–34.5 4,043 14.3 13.5–15.0Excellent, very good, or good . . . . . . . . . . . . . . . . . . . 665 69.5 65.6–73.2 23,575 85.8 85.0–86.5

Any limitation:Yes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 36.1 32.0–40.5 4,480 16.4 15.5–17.3No . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636 63.9 59.5–68.0 23,234 83.6 82.7–84.5


1Equals 95%.2Limited to survey years 1997–2009.



Vital and Health StatisticsSeries Descriptions

ACTIVE SERIES

Series 1. Programs and Collection Procedures—This type of reportdescribes the data collection programs of the National Centerfor Health Statistics. Series 1 includes descriptions of themethods used to collect and process the data, definitions, andother material necessary for understanding the data.

Series 2. Data Evaluation and Methods Research—This type ofreport concerns statistical methods and includes analyticaltechniques, objective evaluations of reliability of collecteddata, and contributions to statistical theory. Also included areexperimental tests of new survey methods, comparisons ofU.S. methodologies with those of other countries, and as of2009, studies of cognition and survey measurement, and finalreports of major committees concerning vital and healthstatistics measurement and methods.

Series 3. Analytical and Epidemiological Studies—This type ofreport presents analytical or interpretive studies based on vitaland health statistics. As of 2009, Series 3 also includesstudies based on surveys that are not part of continuing datasystems of the National Center for Health Statistics andinternational vital and health statistics reports.

Series 10. Data From the National Health Interview Survey—Thistype of report contains statistics on illness; unintentionalinjuries; disability; use of hospital, medical, and other healthservices; and a wide range of special current health topicscovering many aspects of health behaviors, health status, andhealth care utilization. Series 10 is based on data collected inthis continuing national household interview survey.

Series 11. Data From the National Health Examination Survey, theNational Health and Nutrition Examination Surveys, andthe Hispanic Health and Nutrition Examination Survey—In this type of report, data from direct examination, testing,and measurement on representative samples of the civiliannoninstitutionalized population provide the basis for (1)medically defined total prevalence of specific diseases orconditions in the United States and the distributions of thepopulation with respect to physical, physiological, andpsychological characteristics, and (2) analyses of trends andrelationships among various measurements and betweensurvey periods.

Series 13. Data From the National Health Care Survey—This type ofreport contains statistics on health resources and the public’suse of health care resources including ambulatory, hospital,and long-term care services based on data collected directlyfrom health care providers and provider records.

Series 20. Data on Mortality—This type of report contains statistics onmortality that are not included in regular, annual, or monthlyreports. Special analyses by cause of death, age, otherdemographic variables, and geographic and trend analysesare included.

Series 21. Data on Natality, Marriage, and Divorce—This type ofreport contains statistics on natality, marriage, and divorcethat are not included in regular, annual, or monthly reports.Special analyses by health and demographic variables andgeographic and trend analyses are included.

Series 23. Data From the National Survey of Family Growth—Thesereports contain statistics on factors that affect birth rates,including contraception and infertility; factors affecting theformation and dissolution of families, including cohabitation,marriage, divorce, and remarriage; and behavior related tothe risk of HIV and other sexually transmitted diseases.These statistics are based on national surveys of women andmen of childbearing age.

DISCONTINUED SERIES

Series 4. Documents and Committee Reports—These are finalreports of major committees concerned with vital and healthstatistics and documents. The last Series 4 report waspublished in 2002. As of 2009, this type of report is includedin Series 2 or another appropriate series, depending on thereport topic.

Series 5. International Vital and Health Statistics Reports—Thistype of report compares U.S. vital and health statistics withthose of other countries or presents other international data ofrelevance to the health statistics system of the United States.The last Series 5 report was published in 2003. As of 2009,this type of report is included in Series 3 or another series,depending on the report topic.

Series 6. Cognition and Survey Measurement—This type of reportuses methods of cognitive science to design, evaluate, andtest survey instruments. The last Series 6 report waspublished in 1999. As of 2009, this type of report is includedin Series 2.

Series 12. Data From the Institutionalized Population Surveys—The last Series 12 report was published in 1974. Reportsfrom these surveys are included in Series 13.

Series 14. Data on Health Resources: Manpower and Facilities—The last Series 14 report was published in 1989. Reports onhealth resources are included in Series 13.

Series 15. Data From Special Surveys—This type of report containsstatistics on health and health-related topics collected inspecial surveys that are not part of the continuing datasystems of the National Center for Health Statistics. The lastSeries 15 report was published in 2002. As of 2009, reportsbased on these surveys are included in Series 3.

Series 16. Compilations of Advance Data From Vital and HealthStatistics—The last Series 16 report was published in 1996.All reports are available online, and so compilations ofAdvance Data reports are no longer needed.

Series 22. Data From the National Mortality and Natality Surveys—The last Series 22 report was published in 1973. Reportsfrom these sample surveys, based on vital records, arepublished in Series 20 or 21.

Series 24. Compilations of Data on Natality, Mortality, Marriage, andDivorce—The last Series 24 report was published in 1996.All reports are available online, and so compilations of reportsare no longer needed.

For answers to questions about this report or for a list of reports publishedin these series, contact:

Information Dissemination StaffNational Center for Health StatisticsCenters for Disease Control and Prevention3311 Toledo Road, Room 5419Hyattsville, MD 20782

Tel: 1–800–CDC–INFO (1–800–232–4636)TTY: 1–888–232–6348Internet: http://www.cdc.gov/nchsOnline request form: http://www.cdc.gov/cdc-info/

requestform.htmlFor e-mail updates on NCHS publication releases, subscribe

online at: http://www.cdc.gov/nchs/govdelivery.htm.

http://www.cdc.gov/cdc-info/requestform.html

U.S. DEPARTMENT OFHEALTH & HUMAN SERVICES

Centers for Disease Control and PreventionNational Center for Health Statistics3311 Toledo Road, Room 5419Hyattsville, MD 20782

OFFICIAL BUSINESSPENALTY FOR PRIVATE USE, $300

DHHS Publication No. 2014–1367, Series 2, No. 167CS249357

FIRST CLASS MAILPOSTAGE & FEES PAID

CDC/NCHSPERMIT NO. G-284

For more NCHS Series reports, visit:http://www.cdc.gov/nchs/products/series.htm.

Linkage of 1986–2009 National Health Interview Survey With ...

Documents

Transcript of Linkage of 1986–2009 National Health Interview Survey With ...