Statistical Methodology of the National Immunization Survey: 1994-2002

1
Statistical Methodology of the National
Immunization Survey 1994-2002

• Philip J. Smith, Ph.D., Centers for Disease
  Control and Prevention
• David C. Hoaglin, Ph.D. Abt Associates Inc.
• Michael P. Battaglia, M.S., Abt Associates Inc.
• Meena Khare, M.S.., Centers for Disease Control
  and Prevention
• Lawrence E. Barker, Ph.D., Centers for Disease
  Control and Prevention

2
NIS Sampling DesignIndependent samples in
each of 78 Immunization Action Plan (IAP) areas.
These areas will begin to rotate soon
3
Target Pop 5.8M children 19-35 months of age
NIS Sampling Plan
RDD
PRC
3.4 million telephone numbers sampled 31,693 kids
19-35 m.o.a completing the RDD survey 21,410 kids
with sufficient provider-reported data to know
their UTD status
4
3 Important Advances NIS Survey Weights

• National
• Adjustment Percentage
• Children who Received No Vacc Doses 0.3
• Nontelephone Households 4.5
• Vaccination Provider Nonresponse 32.7

5
Rates per 100,000 Zero Dose Children 19-35 moa
6
Characteristics of Zero Dose Children

• Child Characteristics
• Non-Hispanic White 2-5x more likely than
  Hisp. or Non-Hisp Black
• Maternal Characteristics
• College education 2-3x more likely than 12
  years
• gt 30 years of age 2-3x more likely than moms
  20-29 years
• Households Characteristics
• Income gt75k/yr 2-6x more likely than
  75k/yr
• 4 children in HH 2-10x more likely than 1
  child

7
Variation in Telephone Coverage
8
How the Nontelephone Adjustment Works

• Research by Frankel et al. has shown that HHs
  that experienced an interruption in telephone
  service are very much like nontelephone
  households.
• Determine which HHs have had an interruption
• Increase the survey weights of children living in
  HHs with a recent interruption to represent both
  interruption and nontelephone HHs.

9
Characteristics of HHs with an Interruption

• Child Characteristics
• Hispanics 2x more likely than Non-Hisp.
  Whites
• Blacks 2x more likely than Non-Hisp.
  Whites
• Native Americans 3x more likely than
  Non-Hisp. Whites
• Foreign Born 2x more likely than US born
• Maternal Characteristics
• Divorced or Never Married gt 2x more likely than
  married
• lt 12 years of education gt 2x more likely than
  w/ some college
• Preferred language Spanish 2x more likely
  than English
• Maternal age 29 years gt 2x more likely
  than 30 years
• Households Characteristics
• Income below poverty 10x more likely than
  gt75k/yr
• 4 children in HH 2x more likely than 1 child

10
Provider Nonresponse Rates
11
How the Provider Nonresponse Adjustment Works

• Identify characteristics associated with provider
  nonresponse
• W/in each IAP area, group children into
  adjustment classes
• with similar propensities of having provider
  data
• Within each adjustment class
• Increase the survey weights of children with
  adequate provider reported vacc histories to
  represent all of the children in the class

12
Children w/out Adequate Provider Report

• Child Characteristics
• Non-Hipanaic Black and Hispanics
• 1.2x more likely than Non-Hisp. Whites
• Foreign Born 1.4x more likely than US born
• Maternal Characteristics
• No Shot Card 1.1 x more likely than shot card
  moms
• Households Characteristics
• Unknown HH income 1.4x more likely than married

13
Summary

• Progress
• Adjusting NIS vacc coverage estimates for
  nonsampling errors