Objectives. Previous studies have indicated that provider characteristics are an important determinant of immunization coverage. The objectives of this study were to: 1) assess immunization coverage levels among 2-year-old children receiving care in private practices in 3 California counties; and 2) evaluate practice and patient risk factors for low immunization coverage.
Study Design. Cross-sectional chart review of immunization histories and provider survey of immunization policies.
Setting. Forty-five randomly selected, private medical practices in 3 counties in California.
Patients. Children 12 to 35 months old, followed by the participating practices.
Methods. Providers underwent a detailed assessment of their immunization coverage and completed a questionnaire describing their immunization policies and procedures. Immunization data were abstracted from randomly selected medical charts of children 12 to 35 months old. Only patients who met the criteria for active status (≥2 visits and ≥1 visit during the preceding 18 months) were included in analyses. Immunization coverage levels were calculated and logistic regression was used to estimate the risk of underimmunization associated with different practice and child characteristics.
Results. Of the 72 eligible practices that were contacted, 45 participated in the study, yielding a participation rate of 62%. The median immunization coverage of participating offices was 54% (range: 0%–91%). Multivariate analysis revealed 5 independent risk factors for underimmunization. The strongest predictors were having fewer than 50% active children in the practice and children having fewer than 8 visits to the provider. Other significant predictors were the percentage of patients in the practice on Medicaid, administering diphtheria-tetanus-pertussis 4 at a separate visit from the Haemophilus influenzae type b booster, and practice location.
Conclusions. These data provide new insights into immunization practices in an important clinical setting that has been poorly characterized previously. Immunization coverage levels were found to be low and significant risk factors for underimmunization were identified. Recommendations are made for immunization policy changes and targeting of immunization improvement interventions at practices that may be at risk for low immunization coverage. immunization, vaccination, immunization programs, primary prevention, private practice, child, preschool, pediatrics, family practice.
Despite the availability of highly effective vaccines, vaccine-preventable childhood infections continue to be a concern in the United States. The 1989–1991 measles epidemic, which resulted in >55 000 cases, >11 000 hospitalizations, and >130 deaths nationwide, demonstrates the impact of inadequate vaccine coverage.1 Many cases of vaccine-preventable diseases have occurred in unvaccinated, preschool-aged children.2Vaccination of preschool-aged children has been made a national priority, as evidenced by the 1990 Healthy People 2000report, which established the goal of completion of the basic immunization series by the second birthday in at least 90% of all children.3 There have been modest improvements in immunization coverage levels in recent years. The 1997 National Immunization Survey (NIS) conducted by the Centers for Disease Control and Prevention (CDC) found that 76% of children 19 to 35 months of age were up-to-date (UTD) on the basic series.4 Although these vaccination levels are the highest ever recorded among US toddlers, coverage levels are still far lower than the 90% goal established inHealthy People 2000.
The reasons for underimmunization among preschool-aged children can be broadly grouped into 3 categories: those relating to the health care system; those relating to the consumer (parent); and those relating to the health care provider. A number of studies have shown that provider-associated characteristics and behaviors are some of the most important predictors of underimmunization.5–8 Although several studies have examined the relationship between provider-related factors and immunization levels, most have been limited in their scope. Much of the research has concentrated on the impact of provider behaviors on immunization coverage in public clinics.9–12However, more than half of children in the United States receive their immunizations from private physicians.2 In California, it is estimated that two thirds of children are vaccinated in the private medical sector (E. Maes, personal communication, National Immunization Program, CDC, 1998). Of the immunization studies that have been conducted in private medical offices, nearly all have used small samples of 15 or fewer practices, and these have largely been nonrandomly selected.5,,7,13 Thus, further exploration of provider-related factors as determinants of immunization rates in the private practice setting is needed. This article presents data from a study of immunization coverage and provider characteristics conducted among a large, randomly selected sample of private medical practices in 3 diverse counties of California.
The objectives of this study were to: assess immunization coverage levels of 2-year-old children receiving care in private practices in 3 California counties; and 2) evaluate risk factors for low immunization coverage.
Study Setting and Practice Selection
Three California counties were selected for study based on 4 criteria: 1) demographic diversity within and between the counties; 2) the absence of existing immunization intervention activities similar to those implemented in the study; 3) representation of both northern and southern regions of California; and 4) relative accessibility. Contra Costa, San Joaquin, and San Bernardino counties differ substantially with respect to geographic location, population, income level, and racial diversity. Contra Costa is a small, densely populated, and heavily urbanized county located in the greater San Francisco Bay area of northern California. Two thirds of the residents of Contra Costa are white and 5% of families live below the poverty level. San Joaquin is small, less densely populated, and the least urbanized of the 3 counties. It is also located in northern California but is situated in the Central Valley region. Half of the residents of San Joaquin are white and 12% of families live below the poverty line. San Bernardino is a large, sparsely populated, and less urbanized county in southern California in which less than one half of residents are white and 10% of families live below the poverty level.14
The study included a random sample of 45 private, primary care practices located in Contra Costa, San Joaquin, and San Bernardino counties (shown in Table 2). To identify private providers who routinely administer childhood vaccines in each of the counties, 2 databases were used: 1) participants in the Vaccines for Children program; and 2) a listing of all board-certified family physicians and pediatricians provided by the American Board of Medical Specialists. After sorting by county, the 2 databases were merged and duplicates were eliminated. Because this study concentrated on private medical practices, providers whose practices were exclusively associated with public health facilities, university medical centers, and staff–model health maintenance organizations were also eliminated. From the remaining providers, 80 were selected from each county, using a random numbers table. Despite numerous attempts using a variety of resources, such as telephone directories and professional association listings, one third of the providers could not be located. Of the 240 randomly selected providers in 3 counties, 161 were located. All of these were screened over the telephone to determine whether they met preestablished inclusion criteria of: 1) working in a private medical practice; 2) providing immunizations to preschool-aged children; and 3) having a patient population of 25 or more children who were 1 to 3 years of age. Based on these criteria, 72 of the 161 located providers (45%) were deemed eligible. Recruitment into the study was initiated with an introductory letter that was then followed by 1 to several telephone calls. Of the 72 eligible providers, 45 (62%) agreed to participate and all were assessed by the deadline for the study. Practices that declined were asked to complete a questionnaire regarding immunization policies and practice characteristics for comparison with participating practices.
This study was approved by the Institutional Review Board of the University of California at Berkeley.
Children 12 to 35 months of age at the beginning of data collection (ie, born between September 19, 1993 and September 19, 1995) were selected. The following children were excluded from the study: 1) those who had documentation that they had moved or left the practice; 2) those whose parents refused vaccination; and 3) those who had had previous adverse reactions to vaccination. These conditions were determined by the presence of a note in the office chart. After excluding all visits made during the first month of life, the following criteria were used to define active patients within each practice: children with 2 or more visits to the practice, including at least 1 visit during the 18 months preceding the cutoff date for data collection.
Although our prescreening criteria were intended to eliminate practices with fewer than 25 age-eligible children, 3 of the assessed practices actually had fewer than 25 eligible children. Sample sizes of the number of children studied were selected to achieve 95% confidence intervals (CIs) of ±5%. Attributable to staff availability and access to practices, some CIs were wider. The uncertainty for immunization coverage levels ranges from ±.5% to ±10%, with a median of ±6%.
After determining that most practices could not provide complete and accurate patient lists, a 2-stage sampling method was developed. The size of each practice was estimated by counting the number of age-eligible children within a clustered, random sample of all medical charts. Based on this estimated population size, target sample sizes were determined for each practice. In practices where fewer than all eligible children were sampled, the survey population was selected using a second, clustered, random sample of all charts.
Data were collected between September, 1996 and September, 1997. To avoid a cohort effect, children born 12 to 35 months before the first date of data collection (September 19, 1996) were sampled. The first date of data collection was also used as the cutoff date for all subsequent data collection. Thus, immunizations and visits that occurred after September 19, 1996 were not included. Before initiation of the study, all data collection methods were pilot tested at 4 medical practices located outside of the study regions.
In each of the participating sites, trained interviewers administered a 15-minute questionnaire to both a physician and a nonphysician staff member (eg, nurse or medical assistant) who was directly involved in immunizing children in the office. The survey included questions about the office's immunization policies and estimates of the proportion of children in the practice who were insured by Medicaid. Additional information about practice characteristics was obtained through questioning office staff and onsite observation by research assistants.
Chart abstraction was used to determine immunization coverage levels for each practice. Chart abstractions were performed by a team of 7 trained research assistants who participated in regular meetings and received ongoing supervision for the duration of data collection.
The following information was collected from all charts: the child's date of birth, dates of all immunizations, and the dates and types of all visits to the practice. All data abstracted from charts were directly entered onsite into laptop computers using a customized version of the Clinic Assessment Software Application of the CDC. Interrater reliability was assessed by reabstraction of 57 charts by another research assistant. κ-indexes on the major dichotomous variables ranged from .87 to 1.00, indicating excellent agreement.
We defined UTD status on the basic immunization series at 12 and 24 months of age, based on the recommended immunization schedule in the 1994 edition of the American Academy of Pediatrics' Red Book15 that was in use when the children included in the study were between 12 and 24 months of age. Children considered UTD on the basic series at 12 months of age had received the following vaccinations in the first year of life: 3 doses of diphtheria-tetanus-pertussis (DTP), 2 doses of oral polio vaccine (OPV), and 2 doses of Haemophilus influenzae type b (Hib). Children considered UTD on the basic series at 24 months of age had received the following vaccinations in the first 2 years of life: 4 DTP, 3 OPV, 1 measles-mumps-rubella, and completion of the Hib series based on Red Book guidelines.16 The number of doses required for Hib series completion is variable and depends on the specific product used, as well as the age at initiation of the series. Because product information was not always available, we assumed the minimum number of required doses and determined completion of the series based on timing of doses. Although we collected data on hepatitis B immunizations, these data were analyzed separately and were not included in our definition of UTD on the basic series. We defined UTD status on hepatitis B immunizations as having received 3 vaccinations by 12 months or 24 months of age for the 12- and 24-month analyses, respectively.
Immunization assessment data from Clinic Assessment Software Application, as well as questionnaire data regarding the practices' characteristics and immunization policies, were analyzed using SAS 6.12 software (SAS Institute, Cary, NC).17 Only children who met the criteria for active status within a practice were included in estimates of UTD rates for the practice. Descriptive analyses were performed on all continuous outcomes, including the vaccination coverage levels per practice. Univariate and multivariate logistic regression analyses were performed with the child as the unit of analysis and a child's immunization status at 24 months of age as the outcome of interest. An odds ratio (OR) and 95% CI were estimated for each tested variable. Because the study children were clustered within practices, generalized estimating equations were used to account for the clustering effect by appropriately widening confidence intervals for all ORs.
Comparison of Participating and Declining Practices
Table 1 summarizes the characteristics of participating and nonparticipating practices. Twenty-two of the 27 sites that declined to participate (81%) provided information about their practice characteristics. There were no significant differences between the county (P = .82), practice type (P = .10), number of physicians (P = .99), and proportion Medicaid (P = .34) of participating and nonparticipating practices.
Practice Demographic Characteristics
Table 2 summarizes the types of practices included in the study. Although the study sites were randomly selected within each county, there were some important differences in the distribution of the demographic characteristics of the sites. San Bernardino had disproportionately fewer pediatric and larger sized practices, possibly reflecting the county's more rural setting. In addition, only 1 of the sites with an estimated Medicaid population of <10% was located in San Bernardino County. Conversely, none of the practices with an estimated Medicaid population of >50% were located in Contra Costa.
Practice Immunization Procedures
A complete discussion of the survey results regarding offices' immunization policies is the topic of a separate paper. Although the questionnaire covered a wide range of immunization procedures, only a subset of the variables was included in the analyses. Responses to several questions were too uniform to be useful variables and, therefore, were discarded. Additionally, the responses to the same questions of physicians and nonphysicians from the same practice were compared to check the validity of the survey data. We arbitrarily chose to use only the physicians' responses to those questions with 75% or more agreement between physician and nonphysician responders. The 5 questions that met this criterion were included as predictive variables in the analyses.
Sample Size and Proportion of Active Patients
We sampled a total of 4389 children from the 45 study sites. Of these, a total of 3414 children (78%) met the criteria for active status. Among all study sites, we sampled a total of 2325 children who were 24 months of age or older. Of these, a total of 1719 children (74%) met the criteria for active status. Because sample sizes were adjusted for each site based on the size of the target population and other factors, the total sample size varied widely among practices. The total number of children sampled per practice ranged from 16 to 149, with a mean of 98. Sample sizes for the number of children 24 months of age or older ranged from 6 to 80 per practice, with a mean of 52. The percentage of children within a practice who were classified as active also differed substantially among sites, ranging from 31% to 100% among all children and 33% to 100% among children who were 24 months of age and older. Three practices had fewer than 50% active children (Table 2).
Characteristics of Patients
The mean age of active patients in each practice ranged from 22 to 28 months of age among children 12 to 35 months of age and from 29 to 32 months of age among children 24 months of age and older. The mean number of visits per patient varied widely among sites, from a low of 4 to a high of 19.
Immunization Coverage Levels
Among active children, 4 vaccination coverage levels were estimated for each practice (Table 3): UTD at 12 and 24 months of age on the basic series, and UTD at 12 and 24 months of age on the basic series plus 3 hepatitis B vaccinations (HBV). The percentage of children UTD at 12 months of age ranged from 0% to 100%, with a median of 73%. At 24 months of age, the percentage of children who were fully immunized ranged from 0% to 91%, with a median of 54% and a mean of 58%. Practice coverage levels for the basic series plus 3 HBV were considerably lower for 12-month-old children, with a range of 0% to 89% and a median of 50%. At 24 months of age, coverage was only slightly lower for the basic series plus 3 HBV, with a range of 0% to 82% and a median of 52%.
Logistic Regression Analyses
Univariate logistic regression analyses were performed using a child's not-up-to-date (NUTD) status on the basic immunization series (4 DTP, 3 OPV, 1 measles-mumps-rubella, and completion of the Hib series) at 24 months of age as the outcome. All but 1 of the practice and child characteristics tested by univariate analysis were associated with underimmunization at 24 months of age (Table 4). The strongest predictor of being NUTD at 24 months of age was being NUTD at 12 months of age, which yielded an OR of 18.13 (95% CI: 12.21–26.90). This result is not unexpected, because the 2 variables are highly correlated, with a correlation coefficient of .57. For this reason, NUTD status at 12 months of age was not included in the stepwise procedure used to develop a final multivariate model for NUTD status at 24 months of age. The second strongest predictor in the univariate analysis was the percentage of active children in a practice. Patients in the 3 practices that had fewer than 50% active pediatric patients had 15 times the odds of being underimmunized, compared with patients in practices with a majority of actives (OR: 15.41; 95% CI: 8.08–29.40). The third strongest predictor was the mean number of visits per child in a practice. Children in practices that averaged fewer than 8 visits per child had 11 times the odds of being underimmunized, compared with other children (OR: 10.91; 95% CI: 2.80–37.10). Not surprisingly, the number of visits made by each child was also a significant predictor of that child's immunization status. Children with fewer than 8 visits had 4 times the odds of being NUTD (OR: 4.36; 95% CI: 3.14–6.05), and children with between 8 and 14 visits had twice the odds of being NUTD (OR: 1.90; 95% CI: 1.48–2.44), compared with children with more than 14 visits. Other significant characteristics associated with an increased odds of underimmunization in the univariate analysis were: location in San Bernardino and San Joaquin counties; higher percentage of Medicaid patients; family or mixed practice type (vs pediatric practice); administering DTP4 at a separate visit from the Hib booster; and having only 1 or 2 physicians in a practice. Of the 5 self-reported office immunization policies examined, only 1 yielded an OR that was significantly different from baseline (Table 5). Children in practices in which only the physician administered vaccines had twice the odds of being NUTD (OR: 2.31; 95% CI: 1.07–4.97), compared with children in practices in which nonphysician staff administered vaccines.
Multivariate logistic regression analysis was used to examine the independent effects of the various variables, again using a child's NUTD status at 24 months of age as the outcome. Backward, stepwise elimination was used to select a final model. In the most parsimonious model, only 5 factors remained as significant predictors of underimmunization (Table 6). The strongest predictor of NUTD status was attending a practice with fewer than 50% active children (OR: 5.77; 95% CI: 2.96–11.23). The second strongest predictor was having made fewer than 8 visits to the practice by 2 years of age (OR: 4.60; 95% CI: 3.14–6.75). Children with an intermediate number of visits, between 8 and 14, also had increased odds of being NUTD (OR: 1.76; 95% CI: 1.33–2.32) compared with children with more than 14 visits. Other practice characteristics significantly associated with increasing odds of being NUTD were: location in San Bernardino and San Joaquin counties; moderate percentage of Medicaid patients; and more frequent administration of DTP4 at a separate visit from the Hib booster. Family practice versus pediatric practice did not remain significant.
This assessment of immunization coverage among 45 randomly selected private medical practices in California found overall coverage levels to be strikingly low. The median coverage level for the basic series was 54% at 24 months of age, well below the Healthy People 2000 goal of 90% coverage among all 2-year-old children. Substantial evidence shows that children who are immunized in the private sector often have lower immunization coverage levels than their counterparts who are immunized in the public sector.6,,13,18,19 Despite these compelling results, there is a dearth of research on immunization coverage in the private setting, where two thirds of California's children and more than half of the nation's children are vaccinated. A great deal of research has suggested that provider-associated characteristics and behaviors are some of the most important risk factors associated with underimmunization. However, further description of this relationship among private medical practices has also been lacking. We must understand specific risk factors in this setting, so that we can efficiently focus interventions to improve immunization coverage.
This study was undertaken to further elucidate risk factors for low immunization coverage in private medical practices in California. We found that having <50% active pediatric patients in a practice was the most important independent predictor of underimmunization. After controlling for number of visits made to the provider and various other factors, children in practices with fewer than 50% active pediatric patients had nearly 6 times the odds of missing immunizations, compared with children in practices with greater than 50% active pediatric patients. It may be that practices with a minority of active patients function largely as urgent care providers and, thus, are not accessed by families for well-child care, including immunizations. It is unknown whether the children in these practices are medically homeless, meaning that they lack a primary care provider who offers immunizations and other preventive care. However, the percentage of active children in a practice may be a marker for an important risk factor for underimmunization that has not been previously reported. Although this theory is speculative, we believe that additional research in this area is warranted, because it may allow public health officials to target interventions toward medical offices in which children are at tremendous risk of missing immunizations.
In this study, the second most important predictor of vaccination status was the number of visits made by a child to the practice. Children who made fewer than 8 visits were at nearly 5 times the risk of being underimmunized, compared with children with 14 or more visits. This result is not surprising because increased contacts with providers allow for increased opportunities to immunize. This finding, combined with similar results of previously reported studies,6,,13suggests that children who make few office visits are a critically important target population for immunization interventions. Although the parents of these children may contribute to the problem by failing to make and keep appointments, many children actually make enough visits to receive all of their vaccinations (a minimum of 5) but still are not UTD because providers miss opportunities to immunize.1 Conventional wisdom holds that missed opportunities for vaccination are 1 of the single most important causes of low vaccination levels.1 Missed opportunities are caused by less than optimal provider immunization practices, such as deferring immunizations for invalid reasons and failing to administer simultaneous vaccinations. An intervention study that reduced missed opportunities found that the greatest improvement in immunization status was seen in children with the least office visits.20 Our findings support an emphasis on this subgroup of children who make few visits to their providers and would benefit most from efforts to eliminate missed opportunities. Providers may be able to encourage these high-risk children to return for well-care visits by instituting reminder and recall systems.21–23
Another significant risk factor for being incompletely immunized was the county in which the practice was located. Children in practices located in San Joaquin and San Bernardino counties had a greater chance of being underimmunized. Although the exact reasons for this are unknown, it is noteworthy that these counties have a higher percentage of persons living in poverty and a lower median family income than Contra Costa, the comparison county. Additional work should be performed to determine whether there are specific, modifiable factors associated with underimmunization in these areas.
Surprisingly, the percentage of Medicaid patients in a practice was not a clear predictor of vaccination status. Children in practices with a majority of Medicaid patients had no greater risk of underimmunization than did children in practices with <10% Medicaid patients. However, children in practices with an intermediate number of Medicaid patients (10%–50%) had twice the odds of underimmunization. Thus, the predicted result, increased odds of underimmunization associated with increased percentage of Medicaid population, was not found. These results may suggest that practices with a more homogeneous patient population (ie, largely Medicaid or largely private insurance) may do a better job of providing complete immunization coverage to their patients. Perhaps these practices are better able to adapt their immunization behaviors to suit their patient population. It is also possible that some of the physicians' estimates of percentage of Medicaid patients were inaccurate, resulting in misclassification of the practices.
A substantial decrease in UTD status for the basic series occurred between 12 and 24 months of age. Both study-wide mean and practice median coverage dropped from nearly 75% UTD at 12 months of age to <60% UTD at 24 months of age. In nearly all cases, the individual coverage levels of practices for the basic series were also substantially lower at 24 months of age versus 12 months of age. Similar drops in immunization coverage between 12 and 24 months of age have been previously reported by numerous researchers.6,,13,24,25
In contrast, coverage levels for the basic series plus 3 HBV were essentially the same for 12- and 24-month-old children, with study-wide means and practice medians at ∼50%. This finding suggests that many practices still had not begun routine administration of HBV in the first year of life, despite the 1991 recommendation of the CDC for universal immunization of infants against hepatitis B.26
The only office immunization policy that was significant in the multivariate model was administration of DTP4 at a separate visit from the Hib booster. This study found that children in practices that more frequently administer these vaccinations separately have twice the odds of missing immunizations, compared with children in practices that usually administer DTP4 and the Hib booster simultaneously. This finding is particularly important in view of the stagnant, low coverage levels for DTP4. In 1997, national coverage levels for DTP4 did not improve over 1996 levels of 81%.4 Some providers believe that giving DTP4 and Hib booster at separate visits will encourage parents to bring their children back for subsequent well-care appointments. However, a recent study27 found that 74% of parents were motivated to bring their children to well-child visits regardless of the need for immunizations and only 18% of parents needed a vaccine incentive to motivate them.
It is likely that other immunization policies are also important risk factors for underimmunization. This study was limited in its ability to show such effects for various reasons. First, the practice of splitting DTP4 and the Hib booster was the only immunization procedure that was determined from the medical record data. All other policies that were tested for significance were based on physician's questionnaire responses, the accuracy of which is unknown. Second, some reported policies were not used in the analysis because of discrepant answers from physician and nonphysician questionnaire respondents in the same practice. The third problem with detecting significant associations between office immunization policies and coverage levels resulted from the fact that only 1 physician from each practice completed the questionnaire. In offices with more than 1 physician, the surveyed physician's responses may not have been representative of all immunization providers in that office.
Another potential weaknesses of this study is that the sample of practices may be biased. Although sites were randomly selected from each county, the relatively low number of practices that could be located and the relatively low participation rates among located practices may have yielded a biased sample. For example, the large number of practices that could not be located may have been attributable in part to recent restructuring of the private health care system in response to the increase in managed care plans. Furthermore, providers who thought their immunization levels were especially low or high may have chosen not to participate in the study for reasons such as fear or lack of perceived need, respectively.
The percentages of fully immunized 2-year-old children found in the practices included in this study were lower, overall, than the coverage levels estimated by the NIS among children 19 to 35 months of age during the same period.28 This difference may in part reflect an older median age of 27 months of age among children included in the NIS survey. However, the primary reason for this discrepancy is most likely attributable to methodological differences between the 2 studies. The NIS is a population-based study that relies heavily on information from households that are reachable by phone. This study is practice-based and, therefore, captures only the data contained within the medical records from the study sites. Because patients are geographically mobile and/or have variable insurance status, children may leave a practice or receive immunizations elsewhere, without notifying the provider. Thus, as investigators of an office-based study, we faced some of the same problems that plague vaccine providers: determining which children are active patients of the practice and whether they have been vaccinated elsewhere. In the future, these problems may be solved by a well-functioning immunization registry. Until then, it is important to acknowledge that patients move in and out of practices. Although little is known about the number of providers visited by preschool-aged children in California, 1 study in Baltimore found that the vast majority (81%) of children saw the same provider for all primary care during the first 2 years of life.12 Nonetheless, in an effort to account for the nomadic behavior of some patients, this study included only active children in the calculation of a practice's vaccination rate. However, parents were not contacted to confirm the status of their children. It is likely that some patients who had left the practice were included in the study and some children who had received immunizations elsewhere were incorrectly classified as NUTD.
Misclassification, either of a child's active status or a child's immunization status, is one of the inherent difficulties of practice-based studies. Darden et al29 attempted to quantify the impact of these misclassifications on their office-based study. They found that when parent contact information was used to exclude inactive patients, vaccination coverage estimates increased by 4.5%. In addition, they found that supplemental information supplied by parents increased vaccination coverage estimates by 3.6% over those estimates based solely on medical records. The findings of Darden et al29 are reassuring because they suggest that misclassification may be a relatively small problem. Nonetheless, we would be remiss in failing to mention the possibility of confounding by such misclassifications. For example, this study found that smaller proportion of actives in a practice and fewer number of visits per child were the strongest predictors of underimmunization. However, practices with a large proportion of inactive pediatric patients may also have had more misclassifications of inactive children as active. Similarly, children with fewer visits may have had more misclassifications as NUTD, perhaps because they were immunized elsewhere. In both cases, these potential misclassifications may have confounded the relationship between the predictors and immunization coverage levels.
Despite the inherent difficulties, practice-based studies are a valuable research tool that allow a view of immunization coverage different from that provided by the NIS. Aside from the obvious benefit of clarifying the relationship among provider behaviors and immunization coverage, practice-based studies may also provide unique insights into vaccination barriers. For example, the practices in this study that had fewer than 50% active pediatric patients may have contained a subpopulation of patients that was highly nomadic or medically homeless. Similarly, the children that made few visits to their providers may also represent a subset of the preschool-aged population. Such patients may be more difficult to reach by phone and, therefore, may not be captured in population-based studies, such as the NIS.
In summary, we found surprisingly low immunization coverage levels in this study of 45 randomly selected private medical practices in California. Some important risk factors for underimmunization were identified. The strongest predictor of a child's incomplete immunization status was the practice characteristic of <50% active pediatric patients. This previously undetected association may indicate that children who receive care from offices that function primarily as urgent care providers are at great risk of underimmunization. This finding provides direction for further research and may help focus immunization intervention efforts. Other study findings associated with risk of underimmunization were the number of office visits made by the child and administering DTP4 at a separate visit from the Hib booster. This study reinforces that every visit must be viewed as an opportunity to immunize. Regardless of the nature of the visit, immunization records should be reviewed, all shots for which a child is due and eligible should be administered simultaneously, and a reminder or recall system should be used for low utilizers who are NUTD.
This work was supported by the California Department of Health Services, Immunization Branch and the Centers for Disease Control and Prevention, National Immunization Program.
This study was conducted within the context of a larger project, the Immunization Partnership, which was cosponsored by the following organizations: 1) California Department of Health Services, Immunization Branch; 2) University of California at Berkeley, School of Public Health; 3) American Academy of Pediatrics, District IX; and 4) California Academy of Family Physicians.
We thank the members of the cosponsoring organizations for their vision and guidance, as well as Kris Calvin, director of the Immunization Partnership, for her leadership. We are also grateful to the staff of the participating medical practices for assisting us with this research.
- Received June 22, 1999.
- Accepted January 29, 2000.
Reprint requests to (S.M.K.) Immunization Partnership, 853 Ramona St, Albany, CA 94707. E-mail:
- NIS =
- National Immunization Survey •
- CDC =
- Centers for Disease Control and Prevention •
- UTD =
- up-to-date •
- CI =
- confidence interval •
- DTP =
- diphtheria-tetanus-pertussis •
- OPV =
- oral polio vaccine •
- Hib =
- Haemophilus influenzae type b •
- OR =
- odds ratio •
- HBV =
- hepatitis B vaccinations •
- NUTD =
- ↵US Department of Health and Human Services. Epidemiology and Prevention of Vaccine Preventable Diseases. 3rd ed. Washington, DC: US Department of Health and Human Services; 1996
- ↵US Department of Health and Human Services. Healthy People 2000: National Health Promotion and Disease Prevention Objectives. Washington, DC: US Department of Health and Human Services 1990. DHS publication PHS 91-50212
- Taylor JA,
- Darden PM,
- Slora E,
- et al.
- Wood D,
- Donald-Sherbourne C,
- Halfon N,
- et al.
- McConnochie KM,
- Roghmann KJ
- Holt E,
- Guyer B,
- Hughart N,
- et al.
- Bordley WC,
- Margolis PA,
- Lannon CA
- ↵1990 Census Summary Tape, File 3. Available at: http://venus.census.gov/cdrom/lookup/846785298
- ↵American Academy of Pediatrics. Active immunization. In: Peter G, ed. 1994 Red Book: Report of the Committee on Infectious Diseases. 23rd ed. Elk Grove Village, IL: American Academy of Pediatrics; 1994:23
- ↵American Academy of Pediatrics. Haemophilus influenzae infections. In: Peter G, ed. 1994 Red Book. Report of the Committee on Infectious Diseases. 23rd ed. Elk Grove Village, IL: American Academy of Pediatrics; 1994:211–214
- ↵SAS Institute. Proprietary Software, Release 6.12. Cary, NC: SAS Institute; 1996
- Leduc D,
- Loeser H,
- Hercz L,
- Pless IB
- Szilagyi PG,
- Rodewald LE,
- Humiston SG,
- et al.
- Copyright © 2000 American Academy of Pediatrics