NATIONAL IMMUNIZATION SURVEY SAMPLING ESTIMATION AND PUBLICUSE DATA

NATIONAL IMMUNIZATION SURVEY: SAMPLING, ESTIMATION AND PUBLICUSE DATA FILES _____________________ Michael P. Battaglia Abt Associates Inc. Meena Khare National Center for Health Statistics July 16, 2002

SAMPLE DESIGN n Quarterly list-assisted random-digit-dialing sample of telephone numbers (approximately 8, 500 interviews per quarter) n 78 geographic strata (IAP areas) n Goal is to have an approximately equal number of children with provider vaccination history data in each geographic area n Sample management of 78 simultaneous surveys n Combine 4 adjacent quarters to obtain annualized estimates

RDD SCREENING PROCESS n Dial sampled telephone numbers n Identify residential numbers n Screen households (HH) for presence of children aged 19 -35 months n Collect information on all age-eligible children in the HH n Stratified single-stage cluster sample design n Rare population (3. 5% eligibility rate) n Average of 437 children with HH interviews per IAP area in 2000

INTERVIEWING PROCEDURES n Use a variety of procedures to ensure high response rates and to reduce response rate differences between the 78 geographic areas (CASRO response rate of 79% in 2000) n Conduct interview with Most Knowledgeable Person in the HH n Vaccination history from shot card or recall n Demographic and socioeconomic characteristics n Permission to contact child’s vaccination providers (obtain provider name and address)

ESTIMATION n Children with completed household interviews Base sampling weight – reciprocal of selection probability § Adjustment for multiple voice-use phone lines in HH § Three level unit nonresponse adjustment Eligible HH but interview not completed Known HH but eligibility not determined Unknown status (not able to determine if sample telephone number is residential, nonresidential or nonworking)

ESTIMATION (CONT. ) n. Children with completed household interviews § Combine data from 4 adjacent quarters and divide weight by 4 § Poststratify to NCHS Natality File control totals within each geographic area by age category of child, race/ethnicity of mother and maternal education

ESTIMATION (CONT. ) n Children with completed household interviews § Use technique called modified poststratification to compensate for exclusion of nontelephone children (uses NHIS immunization Supplement data on vaccination up-to-date status of nontelephone versus telephone children, and Census and CPS telephone coverage rates

ESTIMATION (CONT. ) n In 2000, 36% of children did not have adequate provider data for use in estimation n Children with adequate provider data § Used response propensity (logistic regression) model with predictors from RDD survey § Within each geographic area divided children into (5) response propensity quintiles § Divided weights of children with adequate provider data by the weighted proportion of children with adequate provider data in their response propensity quintile

ESTIMATION (CONT. ) n. Children with adequate provider data § Used additional raking within each geographic area to ensure that weighted number of children with adequate provider data and their distribution on selected sociodemographic characteristics is the same as for all children with completed HH interviews

PURPOSE OF NIS PUBLICUSE FILES n. Provides opportunity for state and other public health researchers to analyze NIS data n. The NIS Public-Use File (PUF) can be used to form national, state and IAP-area estimates of vaccination coverage rates.

PURPOSE OF NIS PUFs (CONT. ) n. Analysis of age-appropriate vaccination coverage rates

STRUCTURE OF NIS PUBLICUSE FILES n. Each PUF consists of two primary levels of data § Data for all children with completed HH interviews (34, 087 in 2000) § Data for all children with adequate provider data (22, 958) – excludes DISPCODE 7 children

STRUCTURE OF NIS PUFs (CONT. ) n. The phrase "children with adequate provider data" refers to children for whom sufficient vaccination history information is obtained from their providers to determine whether they are up-to-date with respect to the recommended vaccination schedule.

STANDARD ERRORS n. Information in the data file can be used to calculate standard errors of the vaccination coverage rates that reflect the complex sample design used in the NIS.

AVAILABILITY OF NIS PUFs Year Number of Children with HH Interviews 1995 31, 997 Number of Children with Adequate Provider Data 16, 183 1996 33, 305 21, 099 1997 32, 742 22, 806 1998 32, 511 21, 827 1999 34, 442 22, 521 2000 34, 087 22, 958

GEOGRAPHIC IDENTIFIERS n. Each PUF includes IAP area and state identifiers n. Census Region n. MSA codes, county codes and ZIP codes are not included due to risk of disclosure 16

SUBGROUP ESTIMATES n. Demographic and socioeconomic variables in the file can be used to obtain national vaccination coverage rates for subgroups of the population.

SUBGROUP ESTIMATES WITHIN STATES AND IAP AREAS n. Data users should, however, be aware that estimates for such subgroups at the state or IAP-area level will have larger standard errors because of the smaller sample sizes n. Precision guidelines – sample size not less than 30, 95% CI half-width is not greater than 10 percentage points, half-width divided by estimate is not greater than 0. 50

GENERAL VARIABLE INFORMATION n. The variables in the NIS PUF fall into two major groupings: n. Variables that apply to all children with completed household interviews (n=34, 087 2000) and n. Variables that apply only to children with adequate provider data (n=22, 958 in 2000).

NIS PUF CODE BOOK The variables are organized into 9 sections: n. ID variables, weights and flag variables n. Household vaccination variables n. Demographic and socioeconomic variables n. Geographic identifier variables

NIS PUF CODE BOOK (CONT. ) n. Number of providers identified n. Number of providers responding n. Provider characteristics n. Provider-reported up-to-date vaccination variables n. Age in days and months at vaccination

COMPOSITE VARIABLES Household composite variables Include: n. Up-to-date status on individual vaccinations n. Up-to-date status on selected vaccine series n. Race of child and mother, household income n. Child ever have chicken pox?

COMPOSITE VARIABLES (CONT. ) PROVIDER DATA COMPOSITE VARIABLES INCLUDE: n. Vaccine-specific number of doses n. Up-to-date status indicators for individual vaccines and vaccine series n. Age at vaccination in months n. Age at vaccination in days

COMPOSITE VARIABLES (CONT. ) n. Provider characteristics in the NIS PUFs have included: nfacility type ntypes of care offered nparticipation in the Vaccines for Children Program

COMPOSITE VARIABLES (CONT. ) nparticipation in an immunization registry nclinical specialty of person who ordered child’s vaccinations nmedical home

ORDER OF RECORDS IN THE DATA FILE n. All records in the NIS public-use data file have been sorted: nfirst by the unique household identification number (SEQNUMHH) and nthen sorted by the unique child identification number (SEQNUMC) within each household

USE OF THE NIS SAMLING WEIGHTS n. The NIS PUF contains two child-level weights n. HY_WGT is the household interview weight variable for each child. It should be used to produce estimates for children with completed household interviews.

USE OF THE NIS SAMLING WEIGHTS (CONT. ) n. W 0 is the child-level weight for the children with adequate provider data (PDAT = 1) n. This weight should be used to produce estimates of vaccination coverage rates.

BASIC SUDAAN STATEMENTS NSEQNUMH = 1*SEQNUMHH; PROC SORT; BY ITRUEIAP NSEQNUMH; PROC CROSSTAB DATA=SUD_FILETYPE=SAS DESIGN=WR; WEIGHT W 0; NEST ITRUEIAP NSEQNUMH; 29

GUIDELINES FOR USING NIS PUFs n Use the documentation -- README, data user’s guide and the code book n The data file is available in ASCII format n A program is provided on the NIS web site and CD-ROMs for SAS users to convert the ASCII file to a SAS dataset n Software for analyzing complex sample designs includes SAS, SUDAAN and STATA (see Appendix F of Data User’s Guides for example of SUDAAN programs)

GUIDELINES FOR USING NIS PUFs (CONT. ) n Use the entire file for subgroup estimates n SUBPOPN statement in SUDAAN n For estimates by geographic area (e. g. , Maine, Texas) use the entire file and the SUBPOPN command in SUDAAN or just select the cases that belong the specific geographic area (standard errors will be the same) n If only interested in vaccination coverage estimates, just select children with PDAT = 1

GUIDELINES FOR USING NIS PUFs (CONT. ) n Missing values are imputed for key socio- demographic variables except family income n For the 2000 PUF age at the birth dose of Hepatitis B is imputed if the provider indicates that a birth dose date was given – see 2000 data user’s guide n Combining multiple years of NIS data to increase sample size of children within IAP areas and states – see 2000 data user’s guide for instructions and cautions

VISIT THE WEB SITES nhttp: //www. cdc. gov/nis nhttp: //www. cdc. gov/nip/coverage

NIS PUF WORKSHOPS Session 36: Wednesday, 8: 30 -10: 00 Session 41: Wednesday, 10: 30 -12: 00