Examination of universal purchase programs as a driver of vaccine uptake among US States, 1995–2014

Abstract

Background:

Immunization against numerous potentially life-threatening illnesses has been a great public health achievement. In the United States, the Vaccines for Children (VFC) program has provided vaccines to uninsured and underinsured children since the early 1990s, increasing vaccination rates. In recent years, some states have adopted Universal Purchase (UP) programs with the stated aim of further increasing vaccination rates. Under UP programs, states also purchase vaccines for privately-insured children at federally-contracted VFC prices and bill private health insurers for the vaccines through assessments.

Methods:

In this study, we estimated the effect of UP adoption in a state on children’s vaccination rates using state-level and individual-level data from the 1995–2014 National Immunization Survey. For the state-level analysis, we performed ordinary least squares regression to estimate the state’s vaccination rate as a function of whether the state had UP in the given year, state demographic characteristics, other vaccination policies, state fixed effects, and a time trend. For the individual analysis, we performed logistic regression to estimate a child’s likelihood of being vaccinated as a function of whether the state had UP in the given year, the child’s demographic characteristics, state characteristics and vaccine policies, state fixed effects, and a time trend. We performed separate regressions for each of nine recommended vaccines, as well as composite measures on whether a child was up-to-date on all required vaccines.

Results:

In the both the state-level and individual-level analyses, we found UP had no significant (p < 0.10) effect on any of the vaccines or composite measures in our base case specifications. Results were similar in alternative specifications.

Conclusions:

We hypothesize that UP was ineffective in increasing vaccination rates. Policymakers seeking to increase vaccination rates would do well to consider other policies such as addressing provider practice issues and vaccine hesitancy.

1. Background

Childhood immunization against potentially life-threatening illnesses is widely viewed as a great public health achievement [1,2]. Indeed, a recent study finds that fully vaccinating a one-year birth cohort of US children results in 1.2 million additional quality-adjusted life-years, which translates into $184.1 billion in social value, or $45,000 per child [3]. Because of vaccines’ population-level impact in eradicating diseases, federal, state and local governments have played an important role in their purchase [4,5]. Section 317 of the federal Immunization Grants Program, which was expanded in 1991, assists jurisdictions with the purchases of essential vaccines such as polio and tetanus [6]. Section 317 has been associated with a significant increase in immunization rates [6].

Since the implementation of Section 317, several other federal reforms have impacted state-level purchasing power and childhood access to vaccines. Among these reforms is the Vaccines for Children (VFC) program, which was established by the federal government in 1993. VFC is intended to ensure vulnerable children have access to vaccines at no cost. Children under age 18 are VFC-eligible if they meet at least one of the following criteria: they are eligible for Medicaid, uninsured, American Indian or Alaska Native, or underinsured.¹ Through VFC, the Centers for Disease Control and Prevention (CDC) purchases vaccines directly from manufacturers at discounted prices, and distributes them to grantees such as state health departments and local public health agencies, who distribute the vaccines at no charge to private physicians’ offices and public health clinics registered as VFC providers. The implementation of VFC has coincided with higher immunization rates, the introduction of five new childhood vaccines, and a large reduction in vaccine-preventable diseases nationwide [7].

In the late 1990s, some states extended the VFC model to a “Universal Purchase” (UP) structure, with the intention of increasing vaccination uptake [8], and to reduce the burden on providers who may have had to finance the up-front vaccine costs or receive insufficient reimbursement for vaccines [9]. It should be noted, however, that there are many other factors associated with high immunization coverage other than vaccine cost [10].

States with UP programs buy all routinely recommended vaccines through the CDC purchasing contracts and provide them to all children, including those who are privately insured, through eligible providers within the state. In some states, UP programs were replaced by “UP Select” programs which either provide select vaccines or provide all routine vaccines only to some children. Sixteen states have had either a UP or UP Select program in place at some point between 1995 and 2014; as of 2014, seven states had UP programs, whereas three had UP Select programs for public providers and five had UP Select programs with select vaccines [11].

Initially, UP programs were funded by state appropriations, in conjunction with federal VFC funds [12]. State budget estimates for UP programs from 2005 ranged from $10 million in New Hampshire ($714 per eligible child) to $54 million in Massachusetts ($671 per eligible child) [13]. In recent years, state governments have assessed per-child fees on insurers rather than fund the programs through state appropriations [14]. The de facto result of this funding system is a reduction in the price of vaccines paid by private insurers, as their assessments are based on the CDC public sector contract price. This allows private insurers to benefit from the bargaining position of the public sector when acquiring vaccines.

Although UP programs were introduced over two decades ago, literature considering their effect on vaccination rates is sparse, with mixed findings. Freed et al. (1998) report an association between the North Carolina UP program and increased vaccination rates; however, their study only includes children who were born after UP program implementation [8]. Consequently, their results likely reflect both differences across the two birth cohorts as well as underlying time trends in vaccination rather than the effect of the UP program itself. Olshen et al. (2007) find no association between UP programs and adolescent vaccination rates for hepatitis B (HepB) and varicella zoster (varicella); however, the authors point out this may be due in part to low power in their study [15]. Finally, Stokley et al. (2006) find that children living in UP states are more likely to have three doses of pneumococcal conjugate vaccine (PCV) compared to children living in a non-UP state, but this difference is not significant after adjusting for child and maternal characteristics [16]. Our study contributes to the literature by estimating the association between UP programs, including UP Select, and vaccination rates.

2. Methods

2.1. Overview

We used state-level variation in timing of UP legislation from 1995 to 2014 to implement a difference-in-difference framework to estimate the association between UP programs and state-level vaccination rates for children aged 19–35 months. Regression analyses were conducted at the state- and individual-level. Fig. 1 provides the list of states which implemented a UP program during our study period, the type of program (UP or UP Select), and the years it was in effect.

Fig. 1.

UP and UP select Status by State, 1994–2014. Notes: Years that data for UP status were available are denoted with an ‘x’. source: CDC [15,16], Association of Immunization Managers [17], Institute of Medicine [18], and published lierature [9,11,19–27]. States without a UP or UP Select program during any of the years in our study period are not shown in this figure.

We focused on vaccination rates for vaccines recommended by the Advisory Committee for Immunization Practices (ACIP): polio, diphtheria and tetanus toxoids and acellular pertussis (DTaP/DT/DTP), measles or measles-mumps-rubella (MMR), hepatitis A (HepA), HepB, Haemophilus influenzae type b (Hib), varicella, PCV, and rotavirus. A subset of these vaccines is of specific interest because of their higher purchase price (e.g. PCV, varicella, and rotavirus) or their varying uptake rates (e.g. MMR, DTaP, and HepA); these may be more likely to be affected by UP programs. In addition, we also considered the proportion of children aged 19–35 months in a state who are up-to-date for a given set of recommended vaccines, specifically the 5, 6, and 7-series.²

2.2. State-level data

For the state-level analysis, the sample included state-year observations for all 50 states and the District of Columbia from 1995 to 2014. The dependent variable was the state-level vaccination rate for the selected vaccines (Polio, DTaP, MMR, HepA, HepB, Hib, varicella, PCV, and rotavirus) or the proportion of children who were up-to-date for the 5, 6, or 7-series. Vaccination rates and proportion up-to-date were collected from the CDC [17]. Trends in vaccination rates by UP status for PCV, rotavirus, and the 7-series are shown in Fig. 2; the others are in the Technical Appendix. PCV and rotavirus are two of the more costly vaccines to acquire, and therefore potentially more likely to benefit from UP programs. However, Fig. 2 shows that states without UP or UP Select programs had higher vaccination rates for PCV and rotavirus compared to states with UP programs.

Fig. 2.

Trends in Select Vaccination Rates by UP Status. Notes: PCV rates are for 3-doses, but the CDC only reported 4-dose vaccination rates in 2008.

Our independent variable of interest is an indicator for whether a state has either a UP or UP Select program. Information on UP program status was collected from the CDC, the Association of Immunization Managers, the Institute of Medicine, and published literature.[12,14,17–29] Information for UP programs was available for all 50 states in 1994, 2000, 2002, 2005–2009, 2011, and 2014; some states had other years of information available. For the years in which information was not available, we assumed that a state’s UP status remained unchanged between the two observed years. For example, Alaska reported having a UP program in 2002 and 2005. Therefore, we assumed Alaska had a UP program in 2003 and 2004. For states that did not have matching policies across years, we assumed the missing years took on the same policy as the earlier of the two values. For example, Hawaii reported having a UP program in 2002 and a UP Select program in 2005. We assumed the 2002 UP program remained in place for 2003 and 2004.

The state-level dataset also included state-level demographics, health measures, and vaccination-related legislation. Demographic variables were constructed using the 1995–2014 March Current Population Survey, and included average age, average family size, median household income, proportion of the state with a college degree, proportion of the state with a high school diploma, proportion of the state that lives in a rural area, and the proportion of the state that is white [30]. The state unemployment rate was collected from the Bureau of Labor Statistics [31]. Health measures included the proportion of the population with private health insurance (U.S. Census Bureau), and an indicator for whether the state has expanded Medicaid eligibility (National Governors Association, Kaiser Family Foundation) [30,32,33]. Vaccination-related legislation included an indicator for whether the state has either a daycare- or school-entry requirement for the vaccine of interest (CDC, IAC), and an indicator for whether the state allows personal, religious, or philosophical vaccination exemptions [34,35].

2.3. Individual-level data

For the individual-level analysis, we used the 1995–2014 National Immunization Survey (NIS), which contains the child’s vaccination information and characteristics for both the child and mother [17]. The sample included individuals with provider-reported vaccination records for at least one of the selected subset of vaccines. Children were excluded from the analysis if they were missing state information or reported living in a state different from where they were born.³ Finally, we augmented the individual-level NIS data with state-level data on UP program status, and vaccination-related legislation. Approximately 21% of individuals lived in states with a UP or UP Select program.

2.4. Statistical analysis

For the state-level analysis, we estimated a multivariate regression using ordinary least squares for each dependent variable of interest. We estimated four specifications with different sets of controls: (1) UP indicator only; (2) UP indicator and demographic controls; (3) UP indicator, demographic controls, and vaccination legislation controls; and (4) UP indicator, demographic controls, vaccination legislation controls, and health controls. All four specifications included state and year indicator variables.

Additionally, we performed the following sensitivity analyses as variations on our fourth model specification: (5) included separate indicators for full UP and UP Select; (6) reran the analysis omitting states that always had UP; and (7) reran omitting states that never had UP.

For the individual-level analysis, we estimated the probability that a child received the vaccine of interest according to a logistic model.⁴ We estimated three specifications with different sets of controls: (1) UP indicator only; (2) UP indicator and demographic controls; (3) UP indicator, demographic controls, and insurance status. All specifications included state and year indicator variables. The third specification included data for 2007–2014 because insurance variables were not available in all years of the NIS. We additionally performed sensitivity analyses as variations on our second model specification: (4) including separate indicators for full UP and UP Select; and (5) rerunning the analysis including only privately insured children in the sample.

3. Results

3.1. State-level results

State-level means for our model control variables are presented in Table 1 and predicted state vaccination rates given UP status based on our regression results are presented in Fig. 3. (Complete regression results are in the Appendix.) Because the results are consistent across specifications, we use Model 4, with the full set of covariates, for exposition purposes. Using DTaP as an example, our model predicts a state with average characteristics without UP would have a vaccination rate of 83.8%, while the same state with UP would have a vaccination rate of 83.3%, an insignificant difference (p > 0.10).⁵ UP programs do not have a statistically significant effect on vaccination rates for any of the considered vaccines (p > 0.10). Moreover, the estimated changes in vaccination rates from UP adoption are generally small, ranging from −1.84 (PCV) to 0.766 (HepA), which translates to a 2% decrease in PCV rates and a 1.5% increase in HepA rates. UP programs have no statistically significant effect on the proportion of children who are up-to-date for the 5-series, and are only significant for the 6-series and 7-series in certain specifications (p < 0.10). In general, point estimates are more precise for older vaccines, and less precise for newer vaccines (such as rotavirus).

Table 1.

State-level Data (1995–2014): means by State UP Status.

Variable	Full Sample	States with UP	States without UP
Demographics Age (years)	35.7	35.9	35.7
Family size	3.42	3.41	3.42
Median household income ($)	65,605	69,155	64,229
Proportion with college degree	0.184	0.190	0.182
Proportion with high school degree	0.459	0.460	0.458
Proportion rural	0.277	0.347	0.250
Proportion white	0.734	0.787	0.713
Unemployment rate (%)	5.64	5.30	5.76
Health controls
Proportion with private insurance	0.702	0.711	0.699
Expanded Medicaid (1 = yes, 0 = no)	0.325	0.463	0.272
Observations (N)	1020	285	735

Notes: All averages for state population characteristics represent the average within a state for the full population. VFC eligibility is only available 2010–2014. Descriptive statistics for the vaccination-related legislation are provided in the Technical Appendix.

Fig. 3.

Effect of UP Programs on State-level Vaccination Rates, 1994–2014. Note: Results are predicted values of state vaccination rates with and without UP programs at the mean values of the covariates. A state is defined as having a UP program if it has a UP or UP Select program. Predicted values were obtained from ordinary least squares regression of the given vaccination rate in the state-year, adjusting for demographic, vaccination and health covariates, and state and year indicators. Demographic controls included average age, average family size, proportion white, proportion with a college degree, proportion with a high school degree, proportion rural, median household income, and the unemployment rate. Vaccination controls included indicators for vaccination mandates and personal or philosophical exemptions. Health controls included the proportion of individuals with private insurance and whether a state has expanded Medicaid. Error bars give the 95% confidence intervals. Robust standard errors were clustered by state.

Results of the sensitivity analyses were similar to those of the main specification, and are presented in the Appendix.

3.2. Individual-level results

Summary statistics for the NIS sample are presented in Table 2, while Fig. 4 presents a child’s predicted likelihood of vaccination by the UP status of the child’s state, based on the individual regression results. Again using DTaP as an example, our model predicts a child in a non-UP state has a 96% chance of being vaccinated, while a child in a UP state has 97% chance, an insignificant difference (p > 0.10). Similar to the state-level results, UP programs have an insignificant effect on the likelihood that a child is vaccinated across all models and for all vaccines, with the exception of varicella in one specification. UP programs are significant in the varicella specification with full controls: children living in states with UP programs are 0.86 percentage points more likely to be vaccinated for varicella. This translates into a 1.2% increase in the proportion of children who are vaccinated for varicella. Results of the sensitivity analyses were similar to the main specification (see Appendix).

Table 2.

National Immunization Survey (1995–2014): means by State UP Status.

Variable	Full sample	States with UP	States without UP
Child characteristics Proportion aged < 24 months	0.300	0.299	0.300
Proportion white	0.591	0.697	0.563
Proportion female	0.488	0.488	0.488
Proportion firstborn	0.369	0.368	0.369
Mother/family characteristics
Proportion age > 30	0.590	0.624	0.582
Proportion married	0.726	0.768	0.715
Proportion with college degree	0.395	0.418	0.389
Proportion with income >$50 K	0.439	0.467	0.432
# of children in household	1.86	1.85	1.86
Observations (N)	525,683	108,119	417,564

Notes: Characteristics of the mother represent characteristics at the time of the survey.

Fig. 4.

Effect of UP Programs on Likelihood of Child Vaccination, 1994–2014. Note: Results are predicted values of the likelihood a child has received the given vaccine as a function of whether the child is living in a state with or without a UP program. Predicted values were obtained from logistic regression of a child’s vaccination status in a given year, adjusting for individual demographics, state vaccination policies, and state and year indicators. Individual demographic controls included child’s age, an indicator for whether the child is white, a female indicator, a firstborn indicator, an indicator for whether the mother has a college degree, and indicator for whether the family has income great than $50 K, an indicator for whether the mother is married, and an indicator for whether the mother is older than 30. State vaccination controls included indicators for vaccination mandates and personal or philosophical exemptions. State demographic controls included average age, average family size, proportion white, proportion with a college degree, proportion with a high school degree, proportion rural, median household income, and the unemployment rate. Predicted values were estimated at the mean values of the covariates. A state was defined as having a UP program if it has either a UP or UP Select program. Error bars give the 95% confidence intervals. Robust standard errors were clustered by state.

4. Discussion

The study findings suggest that UP programs do not significantly impact state-level vaccination rates. This result holds both in states with large shares of children eligible for VFC and in states with small VFC-eligible populations. The results are robust across vaccine types and several model specifications. Although states may implement UP programs with the goal of increasing vaccination rates [8], our finding that UP programs have an insignificant effect on vaccination rates makes sense conceptually. First, UP programs will not impact children who are already VFC-eligible when the programs are implemented, including about 50 percent of the pediatric population who receive vaccinations through VFC [36]. Second, among unvaccinated children in states that implement UP programs, only those who are not VFC-eligible, that is, privately-insured children, would potentially be impacted by UP programs. Therefore, UP programs may target a subset of the population that varies by state. Another possible explanation is that UP programs are aimed more at enhancing vaccine access for providers by eliminating out-of-pocket costs and reducing administrative processes associated with purchasing vaccines [9]. Nonetheless, it is informative that UP programs generally do not seem to increase vaccination rates.

Children may not get vaccinated for various reasons: they cannot afford it (i.e. they are cost-constrained), they have a contraindication, they are unaware of the need for a specific vaccine, providers are not conveniently available or parents hold philosophical or religious objections. Although UP programs could potentially cause individuals with personal or religious objections to vaccinate if the increased availability of vaccines makes them more salient to their children, we expect this effect to be small. Consequently, any potential effect of UP programs on vaccination rates should be driven primarily by privately-insured individuals who would not vaccinate without a UP program and who are cost-constrained. Therefore, UP programs would have a small effect if 1) a majority of privately-insured individuals would fully vaccinate in the absence of UP, or 2) a large majority of privately-insured individuals are not cost-constrained. Given that the 2010 Affordable Care Act required that private health plans cover all ACIP-recommended vaccines without cost-sharing, vaccine cost should not be a barrier for most privately-insured individuals [37].

A simple calculation supports our assertion that UP programs likely affect a small number of children. In 2014, approximately 8% of children living in states that did not have a UP program were privately insured and not up-to-date for at least one of the recommended vaccines, and therefore would be potentially impacted by a UP program. If we consider the 7-series, as many as 20% of children are either fully or partially unvaccinated and would be potentially impacted by UP programs. However, as we noted previously, vaccine cost should not be a barrier to vaccination for these children given they are covered by private insurance. Although UP programs would target up to 8–20% of children, in light of the findings from our paper, implementing UP programs is not likely to have significant impact on the vaccination rates for these children.

Consequently, states would do well to consider other policies that might be more effective in increasing vaccination rates. Examples of alternative policies include: limiting exemptions for school-entry vaccine requirements to medical exemptions [38], working with health insurance plans to implement programs for physicians that help manage the practice issues related to accessing and stocking vaccines, or mounting public relations campaigns to increase awareness of the importance of vaccination for both the individual and the community [39,40].

This study has several limitations. The effect of UP programs is identified by states that switch UP program status during our study period. Although 16 states had a UP or UP Select program during our study period, our results are identified by the six states that switched UP status (Hawaii, Nevada, North Carolina, North Dakota, South Dakota, and Wyoming). Although this may arguably result in low power to identify significant effects, the point estimates are relatively small, and translate to a range of −2% (PCV) to 1.5% (HepA) increase in the vaccination rate. Second, the NIS data is limited to children aged 19–35 months; therefore our results do not capture the effect of UP programs on older children. Future work could potentially utilize NIS-Teen to construct longitudinal vaccination records and determine the impact of UP programs for older children. Third, we were unable to find detailed information about which vaccinations were covered under UP Select programs historically; therefore, it is possible that certain UP Select states are misclassified in some regressions (most likely to affect PCV, varicella, and rotavirus). Finally, in our main analyses we did not utilize the variation in UP program classification, instead classifying all states that had either a UP or a UP Select program as having a UP program. We addressed this limitation in a sensitivity analysis in which we estimated separate effects of UP and UP Select; results were similar.

5. Conclusions

Various policies have been implemented with the goal of increasing vaccination rates, including vaccine finance programs such as VFC and UP, daycare- and school-entry vaccination requirements, and limiting vaccination exemptions to medical only. UP programs are unique in that they may end up targeting a small, relatively well-insured segment of the population who are just as likely to be vaccinated in the absence of UP. Moreover, the unvaccinated population affected by UP programs is more likely to be unvaccinated for nuanced reasons unrelated to vaccine cost. Hence, we find that the mechanism of UP programs are ineffective at increasing state-level vaccination rates. Although UP programs are financed through fees imposed on insurance providers and therefore may be relatively low-cost policies to implement from a state budget perspective, policymakers who want to increase vaccination rates will need to consider policy solutions other than UP programs.

Abbreviations:

VFC

Vaccines for Children

UP

Universal Purchase

CDC

Centers for Disease Control

HepB

hepatitis B

PCV

pneumococcal conjugate vaccine

ACIP

Advisory Committee for Immunization Practice

DTaP/DT/DTP

tetanus toxoids and acellular pertussis

MMR

measles-mumps-rubella

HepA

hepatitis A

Hib

Haemophilus influenza type b

NIS

National Immunization Survey