Objective. To compare vaccination coverage among children 19 to 35 months of age from public housing developments where a free vaccine outreach program was in place with children residing elsewhere in the city.
Design. A household survey using a multistage cluster sampling method to compare community areas which accounted for 80% of measles cases during 1989 (high-risk stratum), areas which accounted for the remaining 20% of cases (low-risk stratum), and public housing developments (public housing stratum) having free, on-site vaccination services.
Setting. Inner-city Chicago households, April to May 1994.
Outcome Variables. Antigen-specific and series-specific coverage based on written records.
Results. Based on evaluation of 1244 children, citywide coverage for four doses of diphtheria-tetanus-pertussis vaccine, three doses of polio vaccine, and one dose of measles-containing vaccine (4:3:1) was 47% [95% confidence interval (CI), 40% to 55%]. Coverage was significantly lower among children residing in public housing (23%; 95% CI, 18% to 28%) compared with those residing in high-risk strata (45%; 95% CI, 38% to 52%) and low-risk strata (51%; 95% CI, 43% to 60%). Compared with white children (53%), coverage for the 4:3:1 series was lower among African-American children in public housing (29%) or outside public housing (36%). Moreover, 11% (95% CI, 8% to 14%) of children residing in public housing had never received any immunizations.
Conclusions. African-American children throughout Chicago, particularly in public housing, remain at increased risk for vaccine-preventable diseases and should be targeted further for vaccination services. Vaccination coverage remains low several years after a major outbreak of measles and implementation of a free vaccine outreach program. Cluster surveys may be useful for monitoring vaccination coverage in high-risk urban settings.
- CDOH =
- Chicago Department of Health •
- DTP =
- diphtheria-tetanus-pertussis •
- OPV =
- oral polio vaccine •
- MMR =
- mumps-measles-rubella vaccine •
- Hib =
- Haemophilus influenzaetype b •
- CI =
- confidence interval •
- OR =
- odds ratio
Despite the impact of vaccinations on the level of vaccine-preventable disease in the United States,1outbreaks of vaccine-preventable disease continue to occur and account for significant morbidity and mortality.2 During 1989, a total of 2232 cases of measles, 12% of the national total, were reported to the Chicago Department of Health (CDOH).3 4 Of these, 75% were in children under age 5 years (including 7 of the 8 reported measles deaths), 71% were African-American, and 23% were Hispanic children.4 Only 24% of the children afflicted with measles in Chicago had previously received measles vaccine.4 Approximately 10% of reported cases were in children residing in public housing (CDOH, unpublished data). During 1993 to 1994, Chicago also experienced one of the two large urban outbreaks of pertussis in the United States, predominantly among Hispanic children.5 6 The National Vaccine Advisory Committee concluded that failure to vaccinate vulnerable children was the principal cause of the 1989 to 1991 measles epidemic and recommended that vaccination coverage be assessed using feasible and economical methodologies in high-risk urban settings.7Beginning in 1991, the Chicago Department of Health and other local agencies began free outreach vaccination services using a mobile van at five public housing developments.
This report describes the results of a household survey conducted during April 1994 to compare vaccination coverage in subgroups of children in Chicago.
Children 19 to 35 months of age residing in the city of Chicago were the participants of the survey.8 A child was eligible if their birth date was between April 1, 1991 and September 1, 1992 and the parent/guardian or health-care provider furnished field workers with dates of vaccination from a record during a home or office visit, on the phone, or through the mail.
Stratification of the Population
Chicago is divided into 77 community areas, 876 census tracts, and more than 20 000 census blocks. To be able to compare immunization coverage in population subgroups with high measles incidence to those with low measles incidence, community areas were ranked by the number of measles cases reported during 1989. The highest-ranked community areas, which accounted for 80% of the measles cases, were arbitrarily defined as the high-risk stratum. The remaining community areas were defined as the low-risk stratum.
The population residing in the five public housing developments in which a weekly, mobile, on-site immunization program operated was classified as a separate stratum. The housing developments were Robert Taylor, Henry Horner, Cabrini Green, Rockwell Gardens, and Abla Homes; all five developments were located in high-risk communities. Of the 232 022 children under age 5 years in Chicago (1990 census), 10 503 (4.5%) resided in these five public housing developments, including 3746 children 19 to 35 months of age (Chicago Housing Authority, Residential Management System, April, 1994).
A total of 90 clusters, 30 each in the high-risk, low-risk, and public housing strata, were selected proportional to population size. The multistage, systematic sample selection procedure used in this survey was adapted from a method recommended by the World Health Organization Expanded Programme on Immunization to measure immunization coverage.9 10 The 1990 census and the family registers in public housing central offices provided the data for the sample frames for the high-risk, low-risk, and public housing strata, respectively.
At the first stage of sample selection for the high-risk and low-risk strata, 30 community areas in each stratum were systematically selected, proportional to the resident population of children under 5 years of age. At the second stage, census tracts in the selected community areas were randomly chosen. At the third stage, census blocks in the selected census tracts were systematically selected, proportional to the population of 2-year-old children. A cluster was defined as contiguous households, proceeding systematically from a starting point on the northeast corner of each selected census block, that had to be visited to identify a specified number of eligible children. Enumerators proceeded counterclockwise around the first block, and if necessary, around the next nearest and subsequent blocks, until a specified number of eligible children were recorded for each cluster.
For the public housing stratum, the first stage identified the number of clusters to be completed in each of the five public housing developments. The distribution of the 30 clusters was determined by systematic allocation, proportional to the population of 19- to 35-month-old resident children. The second stage was selection, in each of the developments, of a building to correspond with each cluster allocated within the development. The selection of buildings was proportional to the population of eligible children registered as residents of the buildings. A cluster was defined as contiguous apartments, proceeding systematically from a randomly selected floor in a selected building to other floors and buildings, if necessary, to identify a specified number of eligible children.
During a pilot test of the survey, it was found difficult to identify eligible children in the low-risk, compared with the high-risk and public housing strata. Based on that finding, a target of 7 children was set for low-risk clusters, and targets of 20 children per cluster were decided for high-risk and public housing clusters.
Eight 2-person teams of field enumerators made home visits to identify eligible children and record data. The enumerators included CDOH immunization program staff and communicable disease investigators. A written request to contact the CDOH by phone was left at each household when there was no answer at the initial visit. At least one additional home visit was made if the household did not contact the health department. Two additional staff members answered phones at CDOH, and each day the field enumerators checked with them for incoming information about their respective clusters. Data collected on the phone was reportedly read from the child's vaccination card.
Data collected included name, date of birth, race/ethnicity, name and address of provider, and dates of vaccination from written records. If vaccination records were not available at home, providers were contacted by mail, telephone, and personal visits to obtain vaccination dates. Providers were classified as public if they were on the CDOH list of recipients of federal government-purchased vaccine; otherwise, they were classified as private.
Antigen-specific coverage was measured as the percentage of 19- to 35-month-old children who had received three and four doses of diphtheria-tetanus-pertussis (DTP-3, DTP-4) vaccine, three doses of oral polio vaccine (OPV-3), one dose of mumps-measles-rubella vaccine (MMR), a valid dose of MMR (defined as administered at 12 months of age or greater), three and four doses of Haemophilus influenzaetype b conjugate vaccine (Hib-3, Hib-4), and three doses of hepatitis B virus vaccine. Series completion coverages were measured for the combinations of three doses of DTP, three doses of OPV, and one valid dose of MMR, or 3:3:1, and the combination of four doses of DTP, three doses of OPV, and one valid dose of MMR, or 4:3:1.
Data were analyzed using Epi Info Version 6.0.11 Cluster analysis was performed separately for each of the three strata. The 95% confidence intervals (CI) were calculated with adjustments for cluster survey design.12 Citywide coverage was weighted for the population of children under 5 years of age in each stratum. Comparisons and associations were evaluated using χ2 and odds ratios (OR).11
Between April 4, 1994 and May 13, 1994, enumerators completed 14 798 household visits, including 6534, 6272, and 1992 households in the high-risk, low-risk, and public housing strata, respectively. CDOH received 2751 telephone calls as a result of these visits, including 1102, 1394, and 219 from high-risk, low-risk, and public housing strata, respectively. A total of 1576 eligible children were identified, including 1486 (94.3%) from household visits and 90 (5.7%) from phone calls. Of these, 98 (6.2%) parents or guardians refused to participate or did not respond. An additional 234 (14.8%) were excluded because available provider information was insufficient, or the provider refused to participate. The refusal rate was similar in the three strata. An eligible child was evaluated every 4.1 households in the public housing stratum compared with every 12.3 households and every 27.3 households in the high-risk and low-risk strata, respectively. Of 1244 immunization records evaluated, 869 (69.9%) were records available in households, 285 (22.9%) from providers' office records, and 90 (7.2%) were obtained via telephone.
Resources required to conduct the survey included approximately 96 person-weeks of field work by CDOH staff. A private contractor was paid approximately $3000 to collect data from local health care providers. The clusters in public housing were the most efficient, yielding 484 evaluated children for approximately 18 persons-weeks of field work (27 children per person-week). In the high-risk stratum, 530 children were evaluated for approximately 48 person-weeks (11 children per person-week), and in the low-risk stratum, 230 children were evaluated for approximately 30 person-weeks (8 children per person-week).
Demographic, socioeconomic, and health characteristics of the high-risk and low-risk community areas are summarized in Table1 (CDOH Health Inventory, 1993). After stratification of the community areas, the incidence of measles during the 1989 epidemic period was 6 times greater in the area defined as high-risk compared with the area defined as low-risk.
The characteristics of children evaluated in the high-risk, low-risk, and public housing strata are summarized in Table2. A median age of 27 months was observed in all three strata and in all race/ethnicity groups. The proportion of children who were white was significantly higher in the low-risk compared with the high-risk and public housing strata (P < .001). Both the proportion of children who were African-American (P < .001) and who were Hispanic (P = .003) were also significantly higher in high-risk compared with low-risk and public housing strata. Households in the high-risk strata were more likely to have a vaccination record at home than either the low-risk (P = .02) or public housing strata (P < .001). The proportion of children who received services from providers of public vaccine was significantly higher among children in public housing than in the high-risk (P < .001) or low-risk strata (P < .001). The proportion of children receiving vaccine from providers purchasing vaccine privately was significantly higher in the low-risk stratum than in the high-risk (P < .001) or public housing strata (P < .001).
Antigen-specific and vaccination series coverage are summarized in Table 3. Children residing in the low-risk stratum had consistently higher point estimates of coverage than those in the high-risk stratum; however, the differences were not statistically significant. Coverage for all antigens and series were significantly lower in children residing in the public housing stratum compared with the low-risk and high-risk strata. The differences in the proportions of children who received any dose of MMR and a valid MMR dose ranged from 5% in high-risk and low-risk areas to 8% in public housing.
The citywide point estimates for 3:3:1 and 4:3:1 coverage were 57% and 47%, respectively. Children residing in public housing (11%) were more likely (OR, 3.4; 95% CI, 2 to 5.7; P < .0001) to have never been vaccinated than children outside of public housing (3%).
Vaccination status by race/ethnicity is shown in Table4. No significant difference was found in 4:3:1 coverage for Hispanic children compared with white children. However, coverage among African-American children was significantly lower than among white children. African-American children living outside of public housing were only one half as likely (OR, 0.49; 95% CI, 0.5 to 0.8) and African-American children living in public housing were only one third as likely (OR, 0.37; 95% CI, 0.4 to 0.7) as white children to have received the 4:3:1 series.
This survey of Chicago children, 19 to 35 months old, showed that the same minority subgroups that accounted for most measles cases during 1989 remained undervaccinated in 1994. African-American preschool children, particularly those residing in public housing developments, were one third to one half as likely to be vaccinated as white or Hispanic children—approximately 1 out of every 9 children age 19 to 35 months of age residing in public housing had not received any vaccinations. This suggests that free, mobile vaccination services provided on-site at public housing developments on a weekly basis has not redressed undervaccination of young children where their population density, and the potential for future outbreaks of vaccine preventable diseases, may be highest.
Children in the public housing developments reside amidst a population characterized by low-income, minority racial and ethnic groups, large family size, and young maternal age, all recognized risk factors for undervaccination.13-15 Demonstration projects in Chicago and other large cities have shown significant improvement in coverage among preschool children when Women Infant and Children benefits have been linked to age-appropriate vaccination status and services.16 In response to concerns about vaccination coverage among children in public housing, CDOH has implemented a computerized tracking system to monitor the vaccination status of young children and, along with other community organizations, has been providing door-to-door vaccination services for young children.
The survey was useful in identifying characteristics of subgroups of undervaccinated children. Those characteristics were consistent with the epidemiology of measles during the 1989 to 1991 Chicago epidemic. The National Immunization Survey, conducted by the Centers for Disease Control and Prevention, provides quarterly monitoring of current national and local vaccination coverage in the 50 states, the District of Columbia, and 27 urban areas.8 However, periodic local vaccination surveys may be important in some states and cities to monitor coverage in undervaccinated subgroups of the population that cannot be evaluated in such large surveys. Although retrospective surveys of children at school entry can provide a precise estimate of vaccination coverage, the children entering school should have received their basic series of immunizations approximately 3 years before entering school. Immunization programs need more timely information to target interventions. Before this 1994 survey, the most recent preschool vaccine coverage information available to CDOH was from retrospective data for 1987.17
The survey was more than twice as efficient in evaluating children in public housing than in high-risk, and more than three times as efficient than in low-risk, strata. These differences would be even greater if efficiency had been measured in terms of undervaccinated children. Vaccination coverage surveys based on similar cluster sampling methods may be most feasible in suburban and urban areas where the density of young children is high.
An advantage of this community-based survey was detection of children who had immigrated to Chicago and would have been missed if the sampling frames had been vital registration records such as birth certificates. Before this survey we pilot-tested random selection of birth certificates, and also found that 12 (48%) of 25 children were no longer residing at the addresses listed. Many children would also have been missed if a random digit dialing telephone survey had been done, as many households in the public housing and high-risk strata were observed to lack a functioning telephone. Assessment of records of vaccination coverage kept in health facilities is useful to monitor performance and to reduce missed opportunities for children vaccinated at the facilities. However, such assessments do not provide information on children not vaccinated in the facilities.
Although this cluster survey method was useful for assessing current coverage of undervaccinated subgroups of children in Chicago, its usefulness for monitoring coverage may be limited because of the level of precision on estimates of coverage obtained with the basic survey design. Precision and representativeness could be improved by increasing sample size and the numbers of clusters, respectively. However, in this survey approximately 3 days were required to complete an average cluster, and a substantial increase in the amount of field work would have strained available resources.
The interpretation of the results of this study are subject to other limitations. Histories of vaccinations without home or provider records were not considered valid and not recorded; thus, actual vaccination coverage may have been underestimated. Also, although census tracts in Chicago were originally delineated to represent uniform population units of approximately 4000 persons, population shifts in the city have resulted in unequal population distribution. Random selection of census tracts, without selection proportional to population size, or weighting for that stage during analysis may have biased the estimates of vaccination coverage.
Although not directly comparable with this survey because of different methodologies, vaccination coverage estimates from previous retrospective surveys in Chicago provide data which can be used to assess probable trends in coverage. The 3:3:1 series, for example, seems to have increased significantly from 47% in 1987 to 57% (95% CI, 51% to 64%) in 1994 in Chicago.17 The 4:3:1 series also seems to have increased significantly from 25% among 24-month-old children in 1990 to 47% among 19- to 35-month-old children in 1994.18 Coverage for the 4:3:1 series in Chicago was comparable to 1992 coverage estimated in Los Angeles using a similar cluster survey method.15 The largest increase in antigen-specific coverage since 1987 was with MMR, from 52% to 74%, suggesting that the Chicago measles epidemic may have led to improved vaccination coverage, particularly with MMR.18 A 1989 cluster survey identified a similar pattern after the measles epidemic in Los Angeles.19 Also, the recommendation during the Chicago epidemic to administer a dose of measles vaccine beginning at 6 months of age may have affected provider practices and contributed to the current 6% gap between receipt of an early dose and a valid dose of MMR. Citywide coverage for 3 doses of hepatitis B vaccine (14%) was the lowest of all the antigens measured. This can probably be attributed to recommended use of hepatitis B vaccine in the city's program since only 1994 and availability of this vaccine through public clinics since only 1995.
Possible mechanisms for the increase in vaccination coverage since the measles epidemic include greater parental motivation, increased provider compliance, changes in the health system to improve service delivery, and increased community participation and outreach efforts. Excluding infants under 1 year of age and assuming that MMR coverage in 19- to 35-month-old children remains stable until the age of school entry, it is estimated that at the time of the survey more than 45 000 children under 5 years of age in Chicago were vulnerable to measles and in need of MMR vaccination.
Undervaccinated subgroups of children need to be identified and monitored to effectively implement vaccination programs. With appropriate modifications, the World Health Organization cluster survey method designed for estimating vaccination coverage in developing countries may be useful in identifying and monitoring such subgroups in urban areas in the United States.
Essential contributions in carrying out the survey field work were made by Quincy Alexander, Chris Cartel, Helen Estevan, Herminia Guerrero, Kay Harper, Efraim Hernandez, Randy Holcombe, Ruby Lamon, Chang Lee, Gladys Martinez, Dorian Robinson, Janice Rohm, and Mamusa Tinson. Centers for Disease Control and Prevention public health advisors assigned to the Chicago Department of Health Immunization Program who provided extensive logistical support were Kristin Brusuelas, Edward Mihalek, and Stan Owens. Dr Marie Bausfield in the City of Chicago Department of Planning and Roslayn Shepherd in the Chicago Housing Authority Operations Division provided population profiles. We also thank Drs James McAuley, Edward Hoekstra, and Roger Bernier for their helpful suggestions.
- Received April 24, 1997.
- Accepted August 20, 1997.
Reprint requests to (T.K.) BOTUSA TB Project, Centers for Disease Control and Prevention, American Embassy Gaborone, State Department, Washington, DC, 20521-2170.
Dr Kenyon is currently with the Division of Tuberculosis Elimination, Centers for Disease Control and Prevention, BOTUSA TB Project, Gaborone, Botswana.
- Hinman AR
- Centers for Disease Control
- Centers for Disease Control
- ↵Training for mid level managers. The EPI coverage survey. Expanded programme on immunization. World Health Organization. WHO/EPI/MLM/91.10. 1991
- ↵Dean AG, Dean JA, Coulombier D, et al. Epi Info 6.0. A Word-processing, Database, and Statistics Program for Public Health on IBM-compatible Microcomputers. Atlanta, GA: Centers for Disease Control and Prevention; 1994
- ↵Cochran W. Sampling proportions and percentages. In: Sampling Techniques. 3rd edition. New York, NY: Wiley Publication in Applied Statistics; 1977:64–68
- Williams IT,
- Milton JD,
- Farrell JB,
- Graham NM
- Wood D,
- Donald-Sherbourne C,
- Halfon N,
- et al.
- Copyright © 1998 American Academy of Pediatrics