Racial Disparity in Influenza Vaccination: Title and subTitle BreakDoes Managed Care Narrow the Gap Between African Americans and Whites? FREE

In the past decade, it has become clear that there are substantial racial disparities in the quality of health care in the United States, particularly for preventive services.^{1 - 8} Nationally, influenza vaccination rates have increased over the past decade; however, African Americans are persistently less likely than whites to receive influenza vaccinations.^{9 - 15} Prepaid health plans could play an important role in both increasing vaccination rates and reducing racial disparity. Plans have a financial incentive to increase delivery of potentially cost-saving preventive services, such as influenza vaccination to elderly persons and high-risk individuals, which may reduce health plan costs in the short-term.¹⁶ Plans may also monitor use of preventive services and encourage compliance.¹⁷ If limited access to health care or poor health education reduce African Americans' use of preventive services and if health plans address these issues, then racial disparity could be smaller for enrollees in managed care than for those with fee-for-service insurance.

Despite the importance of this issue, few studies have compared racial disparity in care for populations with managed care or fee-for-service insurance.^{18 - 21} Influenza vaccination offers an ideal opportunity to examine whether health plans increase use of preventive services compared with fee-for-service insurance and whether they reduce racial disparity in preventive service use.

In many areas of the United States, Medicare beneficiaries are allowed to choose to enroll in a managed care plan or retain traditional fee-for-service Medicare insurance.²² Those who opt for each type of insurance differ in their demographic characteristics, health status, and tendency to use services.^{23 - 24} Consequently, different vaccination rates among managed care enrollees and fee-for-service enrollees could be due to differences in populations rather than differing effectiveness of care under each type of insurance. We hypothesize that attitudes toward seeking or avoiding medical care ("care-seeking") might affect both the likelihood that an individual would enroll in managed care and the likelihood that the individual would actively seek or resist preventive services such as influenza vaccination. To reduce the possibility that this sort of selection bias would affect our results, we used detailed information about beneficiaries (including their attitudes toward seeking medical care) to create propensity scores that adjust for differences in the 2 populations.

The 1996 Medicare Current Beneficiary Survey (MCBS) was an in-person interview that collected data from a nationally representative sample of nearly 18 000 health care beneficiaries. The 1996 version of the MCBS oversampled beneficiaries who enrolled in managed care. In the autumn survey, beneficiaries were asked whether they had received a flu vaccination during the previous winter. They were asked about demographic characteristics (age, sex, race, Hispanic ethnicity, marital status, income, education, metropolitan residence, and region of residence), health status (which we dichotomized as "fair or poor" health vs "good," "very good," or "excellent" health), and the presence of comorbid diseases (we selected diabetes, heart disease, and chronic obstructive pulmonary disease).

To measure care-seeking attitudes, respondents were asked whether the following statements were true or false: "You worry more about your health than other people your age"; "You will do just about anything to avoid going to the doctor"; "When you are sick, you try to keep it to yourself"; and "Usually, you go to the doctor as soon as you start to feel bad." Another question was "During the past year, did you have any health problem or condition about which you think you should have seen a doctor or other medical person, but did not?" In addition, the survey asked whether the beneficiary had the same physician for more than 5 years. For a subset of respondents, the survey asked, "Did you visit a doctor since the last interview [1 year ago]?" If a respondent did not receive influenza vaccination, the interviewer asked for reasons and coded the responses using the following list: "didn't know it was needed," "shot could cause flu," "shot could have side effects or cause disease," "didn't think it would prevent the flu," "flu not serious," "doctor did not recommend the shot," "doctor recommended against getting shot," "don't like shots or needles," "inconvenient to get shot," "didn't think about it/forgot/missed it," "cost of shot," "had shot before/didn't need it again," or "other."

The response rate was 83.4%, yielding 17 794 responses. We used enrollment data from the MCBS to classify beneficiaries as enrolled in managed care or in fee-for-service insurance. We limited our analysis to beneficiaries who were older than 65 years because this group is clearly eligible (with very rare exceptions such as egg allergy or prior allergic reaction) to receive influenza vaccination. We excluded from the primary analysis respondents (n = 2804) who were younger than 65 years and residents (n = 231) of Puerto Rico where Medicare managed care is not available. We excluded individuals with race specified as unknown (n = 178), Asian (n = 71), Hispanic (n = 260), North American Indian (n = 10), or other (n = 355), and individuals who had been enrolled in managed care for less than 12 months at the time of survey (n = 534). Some of these categories overlapped so that, after exclusions, 13 674 respondents remained for analysis. Although we excluded those who claimed Hispanic race, we included African Americans and whites that claimed Hispanic ethnicity. Race was based on the respondent's self-report to the MCBS interviewer. We determined Hispanic ethnicity using the response to a separate question on the MCBS survey.

We compared managed care and fee-for-service populations overall and within each racial group on each of the sociodemographic characteristics, comorbid conditions, and care-seeking attitudes. We calculated the unadjusted percentage receiving influenza vaccination in all subgroups defined by each of the study variables, using χ² tests to evaluate statistical significance.

To adjust for differences between the managed care and fee-for-service populations, we calculated a propensity score for each respondent representing the probability that a person with those characteristics would select managed care. To calculate these propensity scores, we fitted a logistic regression model with the probability of managed care enrollment as the dependent variable and respondent characteristics—including race, Hispanic ethnicity, sex, age, health status, income, marital status, metropolitan residence, region of residence, diabetes, heart disease, respiratory disease, and the care-seeking variables—as independent predictors.

We also separately modeled the influence of these covariates among whites and African Americans by including interaction terms for race with each of the other covariates. For care-seeking attitudes, we used questions that ascertained information on an individual's underlying predisposition toward health care rather than soliciting attitudes directly about influenza vaccination. We expect that successful physicians or health care plans influence patients' attitudes about influenza vaccination to achieve compliance, so adjusting for these attitudes would be tantamount to controlling for the effect we are evaluating.

We used the propensity scores as weights to calculate the adjusted percentage of each group that reported influenza vaccination, weighting the fee-for-service group by the propensity for managed care enrollment and the managed care group by the propensity for fee-for-service enrollment. This rendered the weighted means for all covariates in the model equal for the 2 groups.²⁵ For analysis including the "doctor visit" variable, we analyzed the subset of 5522 beneficiaries for which data were available. Because the type of insurance might affect whether the patient had the same physician for more than 5 years and whether respondents made visits, we omitted these variables from our main model, but checked results when we included them.

We tested differences in the prevalence of vaccination between groups with the propensity score adjustment using a χ² test for weighted data. We used a t test to compare the difference in the disparities in African American and white vaccination rates between managed care and fee-for-service groups. We also selected the 5 most common reasons for failing to receive vaccination and compared the percentages reporting each reason across the 4 race and insurance groups using a χ² test. All analyses used sample weights provided by the MCBS for cross-sectional analysis so that the results are representative of the 1996 Medicare population aged 65 years and older (without the excluded groups).²⁶ We used SUDAAN software version 7.5 (Research Triangle Institute, Research Triangle Park, NC) to estimate SEs in the weighted analyses.

In 1996, 8.0% of the 13 674 MCBS respondents in our analysis were African American. Compared with whites, greater percentages of African Americans were female (61.7% vs 58.5%; P = .04), reported fair or poor health status (36.8% vs 22.1%; P<.001), had income of less than $10 000 (54.6% vs 21.8%; P<.001), were widowed, divorced, or separated (55.1% vs 38.2%; P<.001), had not graduated from high school (40.4% vs 18.7%; P<.001), and had diabetes (25.2% vs 13.8%; P<.001). African Americans were less likely to report Hispanic ethnicity (1.4% vs 3.3%; P = .008), to be older than 80 years (23.7% vs 26.7%; P = .03), or to report a respiratory condition (11.6% vs 13.9%; P = .06). The prevalence of heart disease did not differ significantly by race (20.4% vs 22.1%; P = .22).

Eleven percent of beneficiaries were enrolled in managed care with similar enrollment for African Americans (10.3%) and whites (11.1%) (Table 1). Compared with fee-for-service beneficiaries, managed care enrollees were significantly younger, less likely to report fair or poor health status, had higher income, were more likely to be high school graduates, more likely to live in metropolitan areas, more likely to live in the Pacific region, and less likely to have diabetes, heart disease, or respiratory conditions. The pattern of these insurance-related differences was qualitatively similar for African Americans and whites for all variables we examined. Among the subset of patients asked about physician visits, those in managed care were more likely than those with fee-for-service insurance to have a visit (87.3% vs 84.9%; P = .02); this difference was substantially larger among African Americans (90.4% vs 74.0%; P<.001) than among whites (87.0% vs 85.9%; P = .28).

The attitudes of managed care and fee-for-service beneficiaries toward seeking health care also differed significantly (Table 2). Managed care enrollees were less likely to report that they would avoid going to a physician or keep to themselves when sick. They were more likely to report that they would visit a physician as soon as they "feel bad." Substantially more fee-for-service beneficiaries than managed care beneficiaries reported having the same physician for longer than the past 5 years. Because of smaller sample sizes, most of the insurance-related differences in attitudes among African Americans were not statistically significant.

Many of the characteristics and attitudes of enrollees were significantly associated with the likelihood of receiving influenza vaccination (Table 3). Those reporting Hispanic ethnicity, widowed, divorced, or separated individuals, low-income individuals, and non–high school graduates were less likely than their counterparts to receive influenza vaccination, while those older than 80 years and with fair or poor health status, diabetes, heart disease, or a respiratory condition were more likely to receive influenza vaccination. Enrollees with attitudes indicating that they would avoid care were less likely to receive vaccination. Enrollees who reported a greater tendency to visit a physician or had the same physician for more than 5 years were more likely to receive influenza vaccination. In the subset analysis, enrollees who reported a visit were more likely to receive influenza vaccination (68.5% vs 42.1%; difference, 26.4%; 95% confidence interval [CI], 22.6%-30.2%).

Overall, 65.8% of Medicare beneficiaries reported that they received influenza vaccination. The percentage receiving influenza vaccination differed substantially by race (Table 4). Whites were more likely than African Americans to receive influenza vaccination (67.7% vs 46.1%; absolute disparity, 21.6%; 95% CI, 18.2%-25.0%). Managed care enrollees were more likely to receive influenza vaccination than fee-for-service beneficiaries (71.2% vs 65.4%; difference, 5.8%; 95% CI, 3.6%-8.3%). However, the unadjusted racial disparities were nearly identical for the managed care and fee-for-service populations (difference in racial disparity, 0%; 95% CI, −8.3% to 8.3%).

After propensity score adjustment, managed care enrollees were still more likely than fee-for-service beneficiaries to receive influenza vaccination (adjusted difference, 5.7%; 95% CI, 3.0%-8.3%). Adjustment changed the racial disparity somewhat. The disparity in fee-for-service, 24.9% (95% CI, 19.6%-30.1%), and the disparity in managed care, 18.6% (95% CI, 9.8%-27.4%), were both statistically significant but the absolute percentage point difference in racial disparity between the 2 insurance groups (6.3%; 95% CI, –4.6% to 17.2%) was not (Table 4). We found no significant statistical interaction effects between race and the other covariates in predicting the rate of influenza vaccination. The results were very similar when we included the variable "same physician for more than 5 years" and when we analyzed the subset that was asked about recent health care visits.

The 5 most common reasons beneficiaries offered for not receiving influenza vaccination included: (1) did not know it was needed (20.6%), (2) thought the shot could cause flu (18.4%), (3) thought the shot could have adverse effects (15.0%), (4) did not think it would prevent flu (14.5%), and (5) did not think about it or missed it (12.6%). The proportions of African Americans and whites citing each of these reasons did not differ significantly nor did the proportions offering these reasons differ between those with managed care and fee-for-service insurance.

In our study, beneficiaries enrolled in managed care insurance had significantly higher rates of influenza vaccination and this did not change when we controlled for other confounding factors including attitudes toward care. We did not find a statistically significant reduction in racial disparity by managed care, suggesting that even if there is a reduction, it is probably small.

Since 1993, when influenza vaccination became a Medicare-covered benefit, vaccination rates have increased. However, while whites met the goal for the Healthy People 2000 program (>60% vaccination among eligible individuals by the year 2000), African Americans have not.^{12 ,27} The reasons for racial disparity are incompletely understood but may include limited minority access to primary care, failure of clinicians to vaccinate minority patients during health care visits, limited awareness among minority patients of the need for vaccination, or misconceptions about the costs, adverse effects, risks, and benefits of vaccination.^{28 - 30} We had also hypothesized that managed care might reduce racial disparity, especially if disparities were related to barriers in access to health care or knowledge, both of which could be overcome by outreach programs and education.^{31 - 34}

Despite the importance of reducing racial disparity in health care, the literature on the impact of managed care on racial disparity is limited, few of the analyses have been based on national samples, and results have been conflicting. A national survey examining access to health care suggested that African Americans in managed care were more likely to report a usual source of health care than African Americans with other forms of insurance.¹⁹ In analyses of general access to health care in selected communities, health maintenance organizations (HMOs) were associated with a smaller racial disparity in self-reported "unmet health needs" but a larger racial disparity in ambulatory visit rates compared with non–HMO insurance.^{18 ,20} Among adults aged 18 years or older, many of whom would not be expected to receive influenza vaccination, rates of vaccination were higher in HMO compared with non–HMO settings but there was no significant racial disparity in any group.¹⁸ Two studies of prenatal care and low birth weight found similar racial disparities in both managed care and nonmanaged care settings in California.^{21 ,35}

It is sometimes assumed that rates of preventive service delivery depend more on individual attitudes and preferences than on the decisions or actions of clinicians.³⁶ Our results confirm that these attitudes are important, though not sole predictors of whether beneficiaries receive vaccination. Higher rates of vaccination in managed care, even among those with a tendency to avoid seeking care, suggest that individual reluctance can be overcome given the right interventions. For example, in a randomized trial, a health plan's case management program increased immunization of African American children.¹⁷ Our results also imply that differences in attitudes toward health care do not fully explain racial disparity, even though these attitudes differed significantly among African Americans and whites. We can only speculate about the other potential causes of disparity.

Our study has a few limitations. We could not independently verify self-reports of influenza vaccination using other data sources. Our results might be biased if the reliability of self-reports differs by race or insurance type. Prior studies suggest that reporting of preventive services is fairly reliable and consistent across socioeconomic groups.^{37 - 38} Our study design was nonrandomized and observational. The propensity score model allowed us to adjust our results for a broad range of potential confounders of the relationship between race, type of insurance, and probability of receiving influenza vaccination. We do not expect large biases related to factors other than those we used in the analysis and doubt that our conclusions would change significantly if we were able to include other factors. However, it remains possible that unobserved differences between the populations could bias our results.

In summary, for the eligible elderly population, the rate of influenza vaccination is higher among enrollees in Medicare managed care plans than among fee-for-service beneficiaries. Similar racial disparity in influenza vaccination exists for those enrolled in managed care and fee-for-service insurance. We remain hopeful that physicians and medical groups, whether working within managed care organizations or providing fee-for-service care, can build on the successful elements of health plan programs that increase vaccination of the population.³⁹ Health plans, physician groups, and public health agencies will need to develop new strategies to reduce racial disparity in influenza vaccination.