OBJECTIVE: We examined associations between influenza vaccination rates and Medicaid reimbursement rates for vaccine administration among poor children who were eligible for Medicaid (<100% of the federal poverty level in all states).
METHODS: We analyzed 3 consecutive National Immunization Surveys (NISs) to assess influenza vaccination rates among nationally representative children 6 to 23 months of age during the 2005–2006 (unweighted N = 12 885), 2006–2007 (unweighted N = 9238), and 2007–2008 (unweighted N = 11 785) influenza seasons (weighted N = 3.3–4.0 million per season). We categorized children into 3 income levels (poor, near-poor, or nonpoor). We performed analyses with full influenza vaccination as the dependent variable and state Medicaid reimbursement rates (continuous covariate ranging from $2 to $17.86 per vaccination) and terms with income levels as key covariates.
RESULTS: In total, 21.0%, 21.3%, and 28.9% of all US children and 11.7%, 11.6%, and 18.8% of poor children were fully vaccinated in the 2006, 2007, and 2008 NISs, respectively. Multivariate analyses of all 3 seasons found positive significant (all P < .05) associations between state-level Medicaid reimbursement and influenza vaccination rates among poor children. A $10 increase, from $8 per influenza vaccination (the US average) to $18 (the highest state reimbursement), in the Medicaid reimbursement rate was associated with 6.0-, 9.2-, and 6.4-percentage point increases in full vaccination rates among poor children in the 2006, 2007, and 2008 NIS analyses, respectively.
CONCLUSION: Medicaid reimbursement rates are strongly associated with influenza vaccination rates.
WHAT'S KNOWN ON THIS SUBJECT:
Although experts have suggested that provider reimbursement rates may affect childhood vaccination levels, this relationship is not well understood.
WHAT THIS STUDY ADDS:
This is the first study, to our knowledge, estimating the association between Medicaid provider reimbursement rates (for vaccine administration) and childhood influenza vaccination rates.
Children in the United States experience substantial morbidity from influenza disease.1,2 The Advisory Committee on Immunization Practices (ACIP) expanded its universal influenza vaccination recommendation to include all children 6 to 23 months of age in 20043 and all children 6 months to 18 years of age in 2008.4 However, childhood influenza vaccination coverage levels have remained low. In addition, wide disparities in influenza vaccination rates between poor and higher-income children and across states have been noted.5,6
Although experts have suggested that provider reimbursement rates may affect childhood vaccination levels,7,–,9 this relationship is not well understood. Presumably, higher reimbursement rates might lead providers to pursue more-aggressive vaccination efforts by implementing patient reminder/recall,10 reducing missed opportunities for influenza vaccination, or implementing other recommended strategies.11 Vaccine reimbursement is designed to cover both vaccine purchase costs and the other costs of vaccination (such as handling and administration). In the case of Medicaid, the federal Vaccines for Children program provides free vaccines, and Medicaid reimbursement should cover the rest of the costs of vaccination.12
A few studies have investigated the relationship between reimbursement and vaccination rates. A study in the 1990s that combined physician bonuses for performing vaccination with feedback noted some improvement in childhood immunization rates when physicians received financial bonuses on the basis of patients' up-to-date coverage (eg, $2500 for a 40% improvement from baseline levels).13 A recent study in New York, which has high Medicaid vaccine administration reimbursement rates, found that the Medicaid reimbursement rate was still less than the actual health care provider costs for influenza vaccination.14 Similarly, inadequate reimbursement rates, compared with provider costs for child vaccinations, were reported for other recent studies in Colorado15 and Georgia.16 Studies on adult influenza vaccination suggested that higher Medicare reimbursement rates for vaccination might lead to better vaccination levels.17 Therefore, some literature suggests that higher reimbursement rates may improve vaccination levels, although most of the evidence is either indirect or pertains to adult vaccinations.
Our primary analysis examined the individual-level association between receipt of influenza vaccine and state Medicaid administration reimbursement rates among Medicaid-eligible poor children. We hypothesized that higher Medicaid reimbursement rates would be associated with a greater likelihood of vaccine receipt, accounting for other factors. Our secondary analysis estimated the association between Medicaid administration reimbursement rates and statewide influenza vaccination rates, which were predicted with multivariable models in the primary analysis, among Medicaid-eligible poor children.
We conducted a cross-sectional analysis for each of 3 seasons, the 2005–2006, 2006–2007, and 2007–2008 influenza seasons, among a nationally representative sample of children 6 to 23 months of age in the 2006, 2007, and 2008 National Immunization Surveys (NISs)18 and compared vaccination rates across states with varying levels of statewide Medicaid reimbursement (Centers for Medicare and Medicaid Services, unpublished data, October 28, 2009). We restricted our analyses to this age group because data on influenza vaccination status were available in the NISs only for these ages.18
Our study period was determined by the availability of the NIS data; that is, the 2008 NIS was the most-recent data set as of July 2010.18 The 2005–2006 season was the first season in our analyses because the Centers for Medicare and Medicaid Services have collected data on statewide Medicaid reimbursement rates every other year since 2005, and thus data on the 2006 and 2008 Medicaid reimbursement rates were not available (Centers for Medicare and Medicaid Services, unpublished data, October 28, 2009). Therefore, 2005 reimbursement rates were used in the analyses for the 2005–2006 and 2006–2007 seasons and 2007 rates were used for the 2006–2007 and 2007–2008 seasons. The analyses using 2005 Medicaid reimbursement rates excluded children who resided in 2 states (Tennessee and Delaware) and Washington, DC (5.1% and 4.9% of the 2006 and 2007 NIS samples, respectively), and the analyses using 2007 reimbursement rates excluded children who resided in Washington, DC (2.1% and 1.5% of the 2007 and 2008 NIS samples, respectively), because data on the Medicaid reimbursement rates in those areas were not available (Centers for Medicare and Medicaid Services, unpublished data, October 28, 2009).
To make our analysis results nationally representative, the “svy” codes in Stata 10 (Stata Corp, College Station, TX) were used to account for the stratum (cluster) variable, the primary sampling unit identifier variable, and the weight variable.18 The unweighted and weighted study population sizes were 12 885 and 3.7 million, 9238 and 3.3 million, and 11 785 and 4.0 million for the 2005–2006, 2006–2007, and 2007–2008 seasons, respectively. We defined the dependent variable as being fully vaccinated according to ACIP guidelines; for example, in our 2006 NIS (2005–2006 season) analysis, children were considered fully vaccinated “if they had received either no doses of influenza vaccine before September 1, 2005, and 2 doses between September 1, 2005, and January 31, 2006, or the date of the interview (whichever was earlier)” or “≥1 dose of influenza vaccine before September 1, 2005, and 1 dose between September 1, 2005, and December 31, 2005.”18
To assess the potential influence of Medicaid reimbursement rates on vaccination rates, we needed to determine both reimbursement rates and children who were eligible for Medicaid. We defined the state Medicaid reimbursement rate for administration of influenza vaccine as a continuous variable ranging from $2.00 to $17.86 (average of $8.21 among 48 states) for the 2005 reimbursement rate and from $2.00 to $17.86 (average of $9.17 among 50 states) for the 2007 rate (Centers for Medicare and Medicaid Services, unpublished data, October 28, 2009) (Table 1). The 2006 NIS did not determine whether a child was enrolled in Medicaid during an influenza vaccination season. Therefore, our primary analysis (for all 3 years) used the NIS categorization of child income levels as a proxy for Medicaid eligibility, as follows: poor, below 100% of the federal poverty level; near-poor, between 100% of the federal poverty level and $75 000 per year; nonpoor, above $75 000 per year.18 Income below 100% of the federal poverty level is the common level for Medicaid eligibility across all states.19
Medicaid enrollment information was available in the 2007 and 2008 NISs but with some limitations (eg, missing data on exact enrollment periods and dual enrollment in Medicaid and private insurance for 11%–12% of children). Such measurement errors would bias the estimated unique association of Medicaid enrollment with vaccination toward the null in the analyses (ie, attenuation).20 We performed a supplemental analysis of the 2007–2009 data after creating 2 types of Medicaid enrollment variables (with and without dual enrollment in a private insurance plan); results were the same.
To evaluate the association between the Medicaid reimbursement rate and the vaccination rate among poor children, we estimated 2 types of measures, namely, the change in the likelihood of full vaccination among poor children relative to nonpoor children (ie, odds ratio [OR]) and the change in the absolute statewide vaccination rate (proportion vaccinated among poor children) associated with a $10 increase in the Medicaid reimbursement rate. The first type of measure was estimated in our primary analysis at the individual level, that is, a multivariable logistic regression model including an interaction term between the Medicaid reimbursement rate and the poor income category.
These logistic regression models also included another interaction term, for interaction between the Medicaid reimbursement rate and an aggregated income category (including both near-poor and unknown-income categories), to examine a possible association for children in this aggregated income category who generally were not enrolled in Medicaid and to assess potential confounding attributable to other unmeasured variables. If the interaction term for poor children was significant but the interaction term for children in this aggregated income category was not, then this would strengthen the association between Medical reimbursement and influenza vaccination rates among poor children.
The second type of measure, the change in statewide vaccination rates among poor children, was estimated through a linear (ordinary least-squares) bivariate analysis. In this analysis, the dependent variable was the state-level average of the probability of full vaccination among poor children (predicted through a primary analysis of the individual-level, multivariable, logistic regression model) and the single explanatory variable was the state-specific Medicaid reimbursement rate (N ≤ 50 states). This analysis was weighted by the population size of each state's poor children.
Our additional hypothesis was that providers who were mainly reimbursed through capitated payments would be less sensitive to the fee-for-service–based reimbursement rate for vaccine administration. To test this hypothesis, we created an interaction term including 3 variables, that is, poor status, Medicaid reimbursement, and lower penetration of Medicaid capitated payments. An available proxy measure of the third variable was the proportion of enrollees in comprehensive Medicaid managed care plans at the state level as of June 2008.21 A proxy dichotomous variable was created for the third variable, indicating whether the child resided in 1 of 24 states with a lower capitated-payment penetration rate (ie, lower than the median proportion [40%] of comprehensive Medicaid managed care plans among 50 states) or not. Presumably, a positive interaction term would support our additional hypothesis.
We used NIS individual-level independent variables, including demographic and socioeconomic factors, that have been suggested in the literature to be associated with vaccination rates5,22 (Table 2). We included state-level independent variables, including the unemployment rate, median household income (to adjust for macroeconomic effects on public assistance programs, such as the number of Medicaid enrollees and tax revenues, which in turn influence Medicaid reimbursement rates), and numbers of pediatricians and family physicians per child 0 to 19 years of age (as measures of access to care), as suggested in the literature,22 and obtained these measures from the Area Resource File.23 Among these 3 state-level variables, the unemployment rate and the median household income were analyzed as continuous variables, because these continuous variables have more variations than the corresponding categorical variables and thus are assumed to capture the macroeconomic effects more accurately.
Medicaid Reimbursement Rates
Table 1 displays state-level reimbursement rates and their rankings within the 50 states. Table 1 demonstrates the very wide variability across states and relatively small changes between 2005 and 2007.
Table 2 displays characteristics of the study population. In total, 23.1%, 25.4%, and 26.5% of children were in the poor income bracket in the 2005–2006, 2006–2007, and 2007–2008 influenza seasons, respectively.
Crude Influenza Vaccination Rates According to Demographic Characteristics
Of the children in our study, 21.0%, 21.3%, and 28.9% were fully vaccinated in the 2005–2006, 2006–2007, and 2007–2008 seasons, respectively, as shown in Table 3. The gaps across 3 levels of income categories were persistent in magnitude, with ∼10-percentage point gaps in the vaccination rates between poor and near-poor children, and between near-poor and nonpoor children in all 3 seasons. The exclusion of states without Medicaid reimbursement rate data had very little effect on the overall full-vaccination rates (eg, an increase of 0.1 percentage point at most in all seasons).
Medicaid Reimbursement Rates and Individual Vaccine Receipt
Table 4 shows the individual-level associations of the covariates with full-vaccination rates in multivariable logistic regression analyses. During the 2005–2006 season, poor children were 40% as likely to be fully vaccinated, compared with nonpoor children (OR: 0.40 [95% confidence interval [CI]: 0.25–0.65]; P < .001). As hypothesized, we found a positive association between state Medicaid reimbursement rates and influenza vaccination rates only among the population of poor children. The Medicaid reimbursement rate was not associated with coverage levels for all children (OR: 0.99 [95% CI: 0.97–1.02]; P = .71). However, the OR for the interaction between state Medicaid reimbursement rates and poor status was significant (OR: 1.05 [95% CI: 1.00–1.11]; P < .05). This indicates that a $10 increase (eg, from a US mean of $8 to a state maximum of $18) in the Medicaid reimbursement rate was associated with a 50% (ie, 5% × $10) relative increase in the rate of full vaccination among poor children, compared with the reference nonpoor children; a $1 increase was associated with a 5% relative increase in the full-vaccination rate.
Similar results were found for the 2006–2007 season. Compared with nonpoor children, poor children were 35% as likely (OR: 0.35 [95% CI: 0.21–0.68]) to be fully vaccinated (P < .01). The OR for the interaction between the state Medicaid reimbursement rate and poor status was significant (OR: 1.07 [95% CI: 1.01–1.13]; P < .05).
During the 2007–2008 season, a similar positive association with the Medicaid reimbursement rate (OR: 1.05 [95% CI: 1.01–1.09]; P < .05) was estimated only for poor children who resided in 24 states with lower capitated-payment penetration and not poor children in other states (OR: 0.99 [95% CI: 0.94–1.04]; P = .67). This result supported our additional hypothesis that providers who were reimbursed mainly through capitated payments would be less sensitive to fee-for-service–based Medicaid reimbursement rates.
Medicaid Reimbursement Rates and State-Level Vaccination Rates for Poor Children
Figure 1 shows the association between Medicaid reimbursement rates and statewide vaccination rates among poor children (predicted with multivariate logistic regression models in the primary analysis) (Table 4) on an absolute scale (ie, proportion with up-to-date vaccination status) with the size of circles representing the size of the poor-child population per state. Although substantial variability existed, a positive association (P < .01) was noted between Medicaid reimbursement and vaccination rates in 48 states during the 2005–2006 (Fig 1A) and 2006–2007 (Fig 1B) seasons and in 24 states with lower capitated-payment penetration during the 2007–2008 season (Fig 1C). The slope of 0.60 for the 2005–2006 season (Fig 1A) indicates that a $10 increase in the Medicaid reimbursement rate was associated with an increase in the state mean of full vaccination rates among poor children of 6.0 (ie, 0.60 × $10) percentage points. This association was 9.2 and 6.4 percentage points for the 2006–2007 and 2007–2008 seasons, respectively.
The use of the 2007 Medicaid reimbursement rate in the 2006–2007 season analysis yielded results that were similar to those obtained with the 2005 reimbursement rate, as described earlier. The supplemental analysis found that an interaction term between the Medicaid reimbursement rate and the Medicaid enrollment variable, instead of the poor variable, was statistically insignificant with 1 exception, that is, Medicaid enrollment for the 2006–2007 season (OR: 1.05; P < .05).
There was no statistically significant association between Medicaid reimbursement rates and rates of full influenza vaccination among children in the other 3 income categories (near-poor, nonpoor, and unknown income) in all 3 seasons. These findings strengthen the association between Medicaid reimbursement and vaccination rates among poor children. In addition, the Medicaid capitated-payment penetration rate (as of June 2008) did not have a statistically significant coefficient when included in analyses for the 2005–2007 seasons.
Our results suggest a positive, robust association between Medicaid reimbursement rates and influenza vaccine receipt among poor children. Although other studies noted variations in influenza vaccination rates according to insurance type,24 this is the first study, to our knowledge, that estimated the association between Medicaid provider reimbursement rates and childhood influenza vaccination rates. The magnitudes of the association were comparable (irrespective of season) when measured on the basis of either ORs in multivariable logistic regression models (OR: 1.05–1.07) or absolute scales in state-level bivariate linear models (0.60–0.92 percentage points per dollar increase). Because of the linear approximation in the latter state-level analysis, P values for the hypothesized association in a bivariate linear model tended to be smaller, compared with the more-accurate P values in individual-level, nonlinear, logistic regression models. The validity of the results presented was also strengthened by a supplemental analysis (results not shown) that noted that Medicaid reimbursement rates had similar associations with partial vaccination (ie, not fully vaccinated but received ≥1 influenza vaccine dose) as with full vaccination.
Three cautionary points should be noted in consideration of a potential policy option of increasing Medicaid reimbursement. First, it is important to note that many factors other than provider reimbursement rates affect vaccination rates.3 To optimize influenza vaccination coverage, multiple strategies are likely needed. Second, the estimated association might be attenuated if the reimbursement rate was much greater than $18; our estimates were based on the current statewide variation between $2 and $17.86. Third, our estimated association may be sensitive to other factors such as overall influenza vaccination rates (which may increase year by year), vaccine supply variations, and unusual situations such as that in the 2009–2010 season when additional novel influenza vaccines were needed.
The issue of reimbursement is complex, as implied by the mixed literature findings regarding the effectiveness of explicit performance incentives for both child and adult populations; approximately one-third of this literature shows modestly significant effects,13,25,–,27 another one-third indicates weak or null effects,28 and the remaining one-third demonstrates negative unintended consequences (reported only for adult populations).29 Studies documented that the Vaccines for Children program, which improved distribution and provider reimbursement for purchasing vaccines for Medicaid, was associated with reductions in physician referrals to health departments and also with increased overall vaccination rates,30,31 which suggests implicitly that vaccination delivery may be sensitive to provider reimbursement rates.
Our study contributes to the literature by providing a different type of implicit evidence at the national level, suggesting a reimbursement effect on providers. Because of such generalizability, our results could be used for a cost-effectiveness or cost/benefit simulation analysis of a future state/federal policy to increase reimbursement rates and also for justification of a future randomized controlled study to examine how an increase in Medicaid reimbursement rates for administration of vaccines may affect provider vaccination behavior (eg, more-aggressively using tracking systems, chart reminders, patient reminders, standing orders, and staff training).17
Our study has several potential limitations. The first limitation is that, although the associations were strong in the 2005–2006 and 2006–2007 seasons, they were strong in the 2007–2008 season only for poor children in 24 states with lower capitated-payment penetration and not for all poor children, as observed in the preceding 2 seasons. This finding for the 2007–2008 season may be attributable to less-accurate Medicaid enrollment data during this season or unknown factors.
The second limitation is that state dummy variables (or state-level immunization program variables) were not included in our final analyses because of perfect colinearity with our Medicaid reimbursement rate variable. The Medicaid reimbursement variable could have been confounded by other, unmeasured, state-specific characteristics that were only partly adjusted with our Census region variables. If such confounding had been serious, however, then we would have expected Medicaid reimbursement rates to be associated with influenza vaccination rates for nonpoor or near-poor children (and not just poor children), and we did not find this association in individual-level multivariable analyses. The third limitation is that, although we demonstrated strong associations between Medicaid reimbursement and influenza vaccination rates, we cannot prove a causal link.
Medicaid reimbursement rates for influenza vaccinations seem to be strongly associated with influenza vaccination rates among poor children. This positive association may be strongest during seasons shortly after new vaccine guidelines. An increase in Medicaid reimbursement rates is one policy option to improve influenza vaccine coverage levels for this nation's poor children.
This study was funded directly by the National Institutes of Health (grant 1K25AI073915).
We thank Chintan Pandya, MPH, MBBS, University of Rochester, for his research assistance, Jay Bhattacharya, MD, PhD, Stanford University, for his support in conception and interpretation of empirical analyses, and Stanley J. Schaffer, MD, MS, University of Rochester, for his comments on the manuscript. We also thank Jovonni R. Spinner (Centers for Medicare and Medicaid Services) for sharing the unpublished data on Medicaid reimbursement rates.
- Accepted August 10, 2010.
- Address correspondence to Byung-Kwang Yoo, MD, PhD, University of Rochester, School of Medicine and Dentistry, Department of Community and Preventive Medicine, 601 Elmwood Ave, Box 644, Rochester, NY 14642. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
Funded by the National Institutes of Health (NIH).
- ACIP =
- Advisory Committee on Immunization Practices •
- NIS =
- National Immunization Survey •
- OR =
- odds ratio •
- CI =
- confidence interval
- 3.↵Centers for Disease Control and Prevention, Advisory Committee on Immunization Practices. Prevention and control of influenza: recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2004;53(RR-6):1–40
- Santibanez TA,
- Santoli JM,
- Bridges CB,
- Euler GL
- Freed GL,
- Cowan AE,
- Gregory S,
- Clark SJ
- 12.↵Centers for Disease Control and Prevention. Vaccines for Children Program (VFC): for parents. Available at: www.cdc.gov/vaccines/programs/vfc/parents/default.htm. Accessed August 25, 2009
- Yoo B,
- Szilagyi PG,
- Schaffer SJ,
- et al
- Glazner JE,
- Beaty B,
- Berman S
- Coleman MS,
- Lindley MC,
- Ekong J,
- Rodewald L
- 18.↵Centers for Disease Control and Prevention. Datasets and related documentation for the National Immunization Survey. Available at: www.cdc.gov/nis/data_files.htm. Accessed February 14, 2010
- 19.↵Kaiser Commission on Medicaid and the Uninsured. Challenges of Providing Health Coverage for Children and Parents in a Recession: A 50 State Update on Eligibility Rules, Enrollment and Renewal Procedures, and Cost-Sharing Practices in Medicaid and SCHIP in 2009. Washington, DC: Henry J. Kaiser Family Foundation; 2009:22. Available at: www.kff.org/medicaid/upload/7855.pdf. Accessed July 1, 2010
- Greene WH
- Henry J
- Joyce T,
- Racine A
- 23.↵US Department of Health and Human Services, Health Resources and Services Administration, Bureau of Health Professions. Area Resource File (ARF). Rockville, MD: Health Resources and Services Administration; 2007
- Hillman AL,
- Ripley K,
- Goldfarb N,
- Weiner J,
- Nuamah I,
- Lusk E
- Copyright © 2010 by the American Academy of Pediatrics