- Email: [email protected]
Ethnic Disparities in Vaccination Among the Elderly
Influenza Vaccine Supply and Racial/Ethnic Disparities in Vaccination Among the Elderly Byung-Kwang Yoo, MD, PhD, Megumi Kasajima, BS, Charles E. Phelps, PhD, Kevin Fiscella, MD, MPH, Nancy M. Bennett, MD, MS, Peter G. Szilagyi, MD, MPH This activity is available for CME credit. See page A3 for information
Background: The impact of vaccine shortages on disparities in influenza vaccination is uncertain. Purpose: The objective of this research was to examine the association between influenza vaccine supply and racial/ethnic disparities in vaccination rates among elderly Medicare benefıciaries. Methods: Cross-sectional multivariable logistic regression analyses were performed in 2010 to examine whether racial/ethnic disparities in vaccination rates changed across two consecutive seasons: from (Period 1) 2000 –2001 and 2001–2002 seasons through (Period 4) 2003–2004 and 2004 –2005 seasons. Self-reported receipt of influenza vaccine across consecutive years was examined among communitydwelling non-Hispanic African-American (AA); non-Hispanic white (W); English-speaking Hispanic (EH); and Spanish-speaking Hispanic (SH) elderly enrolled in the Medicare Current Benefıciary Survey (unweighted n⫽2306 –2504, weighted n⫽8.23– 8.99 million for Periods 1 through 4).
Results: During Periods 1 and 2, when vaccine supply increased nationally, adjusted racial/ethnic disparities in the influenza vaccination rate decreased by 1.8%–7.4% (W–AA disparity); 4.5%– 6.6% (W–EH disparity); and 6.6%–11% (W–SH disparity) (all p⬍0.001). During Period 4, when vaccine supply declined, adjusted disparities in vaccination rates increased by 2.3% (W–AA disparity) and 6.1% (W–EH disparity) but decreased by 6.6% (W–SH disparity) probably due to a “floor effect” (constant low rates among SH; all p⬍0.001). Conclusions: Improved vaccine supply was generally associated with reduced racial/ethnic disparities in influenza vaccination rates, whereas worse supply was associated with increased disparities. To avoid future widening of racial health disparities, policy options include stabilizing the vaccine supply and preferential delivery of vaccines to safety-net providers serving AA and Hispanic populations during a shortage. (Am J Prev Med 2011;40(1):1–10) © 2011 American Journal of Preventive Medicine
Introduction
E
liminating health disparities is an overarching goal of Healthy People 2010.1 Racial/ethnic disparities have been consistently documented for influenza vaccination among adults aged ⱖ65 years,2– 4 with Hispanic and African-American populations having lower influenza vaccination rates than white populations, despite evidence that influenza vaccination reduces influenza-related morbidity and is cost effective.5,6 In one nationally represen-
From the Department of Community and Preventive Medicine (Yoo, Kasajima), Provost’s Offıce (Phelps), Department of Family Medicine (Fiscella), Department of Medicine (Bennett), Department of Pediatrics (Szilagyi), University of Rochester, Rochester, New York Address correspondence to: Byung-Kwang Yoo, MD, PhD, 601 Elmwood Avenue, Box 644, Rochester NY 14642. E-mail: [email protected]. 0749-3797/$17.00 doi: 10.1016/j.amepre.2010.09.028
tative study,7 African Americans were 34% less likely to receive influenza vaccination than white Americans. Disparities in influenza vaccination may be exacerbated when vaccine supply declines or is delayed as vulnerable populations, including underserved racial and ethnic populations, may face relatively greater barriers to accessing vaccination during such times. There have been fıve seasons of delay/shortage since 2000 (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006),8 and the prospect of future supply problems remains.4 Roughly 15% of Medicare benefıciaries report diffıculties in receiving an influenza vaccination because of supply issues,9 and many high-priority groups reported that they did not attempt vaccination because of their perception of a shortage during the 2004–2005 season.10 One local population study noted that African-American adults were less likely to be vaccinated during a serious vaccine shortage season (2004 –2005), compared to a prior season with a moderate shortage relative to demand late in the season
© 2011 American Journal of Preventive Medicine • Published by Elsevier Inc.
Am J Prev Med 2011;40(1)1–10 1
2
Yoo et al / Am J Prev Med 2011;40(1):1–10
(2003–2004) (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8,11–13 Several challenges exist in assessing the impact of vaccine supply on racial/ethnic disparities. One challenge is that multiple factors such as overall coverage, supply, and health status vary over time. Thus analyses that compare populations need to adjust for underlying differences in these populations as well as time trends. A cross-sectional analysis cannot adequately control for potential time-varying confounders that may disproportionally affect people in each racial/ethnic group. A cohort analysis is more suitable because potentially confounding variables within the same population are likely to change relatively little between two consecutive seasons. The objective of the present study was to evaluate the association between influenza vaccine supply and racial/ ethnic disparities in vaccination rates across two consecutive vaccination seasons among nationally representative elderly Medicare benefıciaries. It utilized a cohort analysis of populations followed over two consecutive seasons with different vaccine supply. Thus, the current study provides a rigorous test of the hypothesis that racial/ethnic disparities in influenza vaccination rates are worse when vaccine supply is delayed or limited.
Methods Design Overview The 2000 –2005 Medicare Current Benefıciary Surveys (MCBSs)9 were analyzed in 2010 to examine levels and time trends in racial/ ethnic disparities across two consecutive vaccination seasons within each of four periods: Period 1, 2000 –2001 and 2001–2002 seasons; Period 2, 2001–2002 and 2002–2003 seasons; Period 3, 2002–2003 and 2003–2004 seasons; and Period 4, 2003–2004 and 2004 –2005 seasons. All analyses used Stata, version 10. MCBS is a multipurpose survey of a nationally representative sample of the Medicare population.9 Sample people are surveyed about their health experience three times a year over 4 years. Survey data are combined with participant Medicare claims. Survey response rates were 66%– 82% during the current study period. The 2000 –2001 season was characterized by a severe delay in supply, followed by improvement in2001–2002withmoderatedelay(CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8 The 2003–2004 season had a moderate shortage relative to demand late in the season because of unusually high demand and was followed by a severe shortage in the 2004 – 2005 season (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8,11,12
Setting and Participants The study population consisted of community-dwelling non-Hispanic African-American (African-American); non-Hispanic white (white); and Hispanic Medicare benefıciaries (aged ⱖ65 years). Analyses for each period examined the same cohort, that, is those who were enrolled in MCBS for two consecutive influenza seasons (i.e., unweighted n⫽2504, 2501, 2306, and 2411; and weighted n⫽8.73, 8.95,
8.23, and 8.99 million for Periods 1, 2, 3, and 4, respectively). Because prior research14,15 suggests that white–Hispanic disparities are largely explained by language differences, English-speaking Hispanic (EH) and Spanish-speaking Hispanic (SH) benefıciaries were distinguished based on whether Spanish was used in MCBS interviews.
Statistical Analysis Crude race/ethnicity-specifıc influenza vaccination rates were calculated fırst, and white–African-American (W–AA) and white– Hispanic (W–EH and W–SH separately) disparities were assessed for each season. Then, the changes in racial/ethnic disparities were measured for each period. Primary analysis (Tables 1–3 and Figures 1–4) used self-reported vaccine receipt data because (1) claims data were not complete for Medicare managed care plan (HMO) enrollees and (2) claims data do not capture vaccination given in settings that do not bill Medicare. Specifıcally, claims data underestimated the vaccination rate by 20%– 24% in the current study’s populations. Self-reported vaccine receipt data are collected in the fall of each year through the MCBS with the question Did you have a flu shot for last winter? The primary analysis, a cross-sectional multivariable logistic regression model, examined whether changes in racial/ethnic disparities remained after adjusting for demographics, socioeconomic factors, and other patient factors (Table 1) that have been found3,7,16 –21 to be related to influenza vaccine receipt among U.S. elderly adults aged ⬎65 years. Race/ethnicity-specifıc adjusted vaccination rates were predicted in this regression model, assuming all subjects in the model belong to a certain racial/ethnic group. Using the adjusted rates, this analysis calculated the racial/ethnic disparities in influenza vaccination rates per season and changes in disparities for each period. Next, the adjusted Wald test determined whether these disparities and differences in disparities between sequential seasons were statistically different from zero. A common set of cross-sectional weights was applied for each period.
Results Table 1 indicates that the patient factors changed little between the two consecutive seasons in both periods, as expected. Vaccine supply differed substantially at a monthly level and at a cumulative annual level (in December) between the two consecutive seasons in all periods except Period 3 (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8 Figures 1– 4 show crude and adjusted vaccination rates for the fırst and second season of each period. Tables 2 and 3 display the change in vaccination rates between the two seasons. These data show racial/ethnic disparities by season and corresponding changes in disparities between seasons where vaccine supply improved (Periods 1 and 2) and worsened (Periods 3 and 4). For all groups, the vaccination rate generally increased when vaccinesupplyimproved(Periods1and2)anddecreasedwhen vaccinesupplyworsened(Period4).Despitethesimilarvaccine supply between the two seasons in Period 3, vaccination rates increased in the 2003–2004 season during which an unusually early onset of a relatively severe epidemic led to higher vaccine demand late in the vaccination period and a consequent modwww.ajpm-online.net
January 2011
Table 1. Study population characteristics and influenza vaccine supply
Influenza season (vaccine supply) Cumulative influenza vaccine supply (million doses)a
Period 1
Period 2
Period 3
Period 4
(nⴝ2504, weighted nⴝ8.73 million)
(nⴝ2501, weighted nⴝ8.95 million)
(nⴝ2306, weighted nⴝ8.23 million)
(nⴝ2411, weighted nⴝ8.99 million)
2000–2001 (severe delay)
2001–2002 (moderate delay)
2001–2002 (moderate delay)
2002–2003 (no supply problem)
2002–2003 (no supply problem)
2003–2004 (moderate late-season shortage)
2003–2004 (moderate late-season shortage)
2004–2005 (severe shortage)
Nov⫽48.2 Dec⫽70.4
Nov⫽76.2 Dec⫽77.7
Nov⫽76.2 Dec⫽77.7
Nov⫽83.0 Dec⫽83.3
Nov⫽83.0 Dec⫽83.3
Nov⫽80.6 Dec⫽82.6
Nov⫽80.6 Dec⫽82.6
Nov⫽52.1 Dec⫽56.4
Outcome variable: influenza vaccine receipt (%) 68.4
72.6
70.9
73.9
71.2
74.8
73.8
66.8
84.9
84.9
83.7
83.7
84.2
84.2
84.4
84.4
Explanatory variablesb (%) Non-Hispanic white Hispanic (English-speaking) Hispanic (Spanish-speaking)
c
Non-Hispanic African-American
3.87
3.87
3.86
3.68
4.68
4.80
3.86
3.86
2.91
2.91
4.17
4.35
3.27
3.15
3.40
3.40
8.28
8.28
8.32
8.32
7.80
7.80
8.30
8.30
Female
58.4
58.4
57.7
57.7
58.1
58.1
57.4
57.4
Aged ⱖ75 years during first season
44.8
48.8
44.9
49.8
45.8
49.4
44.8
49.5
Medicaid enrollment
7.43
8.06
8.76
9.71
8.86
9.00
9.40
9.70
Subjective general health status (fair/poor)
18.8
20.1
19.3
20.3
17.1
17.8
18.3
18.7
High-risk chronic conditions for influenzad
35.9
39.7
39.7
41.9
39.4
40.5
43.5
44.6
Yoo et al / Am J Prev Med 2011;40(1):1–10
Influenza vaccine receipt (survey data)
a
Cumulative number of influenza vaccine doses distributed among the general U.S. population (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006)8 b Additional explanatory variables included education (high school diploma); income (ⱖ$30,000); number in household: two or more; metropolitan residence; private supplemental health insurance (Medigap); activity of daily living (ADL) limitation: one or more; instrumental activity of daily living (IADL) limitation: one or more; avoid medical care when sick; see physician soon when sick; presence of physician regularly consult with8; census region. All explanatory variables were dichotomous variables. c Used Spanish in at least one of MCBS interviews d Those with heart disease, lung disease, diabetes, renal disease, rheumatologic disease, dementia and/or stroke, identified by (1) Medicare claims data (inpatient and outpatient) associated with MCBS survey data among Medicare fee-for-service (FFS) enrollees and (2) both survey and claims data among Medicare managed care enrollees because of limited availability of claims21,22 MCBS, Medicare Current Beneficiary Survey
3
4
Table 2. Changes in racial/ethnic disparities in influenza vaccination rate in Period 1a–2b among Medicare elderly (% vaccinated) Period 1
Period 2
2000–2001 season (severe supply delay)
2001–2002 season (moderate supply delay)
Change from 2000–2001 to 2001–2002 season
2001–2002 season (moderate supply delay)
2002–2003 season (no supply problem)
Change from 2001–2002 to 2002–2003 season
White
71.3
74.8
⫹3.47***
73.9
76.9
⫹2.97***
African-American
53.7
61.9
⫹8.29***
52.8
58.1
⫹5.30***
60.6
67.6
⫹6.97***
64.8
68.6
⫹3.79***
35.6
45.4
⫹9.84***
52.7
49.5
⫺3.23***
White–African-American disparity
17.6*** (17.5, 17.8)
12.8*** (12.7, 12.9)
⫺4.82*** (⫺4.85, ⫺4.79)
21.2*** (21.1, 21.3)
18.8*** (18.7, 19.0)
⫺2.33*** (⫺2.37, ⫺2.29)
White–Hispanic (Englishspeaking) disparity
10.7*** (10.6, 10.9)
7.22*** (7.06, 7.38)
⫺3.51*** (⫺3.53, ⫺3.48)
9.10*** (8.96, 9.29)
8.32*** (8.14, 8.46)
⫺0.82*** (⫺0.85, ⫺0.79)
White–Hispanic (Spanishspeaking) disparityc
35.7*** (35.6, 35.9)
29.4*** (29.2, 29.56)
⫺6.38*** (⫺6.42, ⫺6.33)
21.2*** (21.1, 21.4)
27.4*** (27.3, 27.6)
⫹6.20*** (6.16, 6.24)
White
71.0
74.1
⫹3.05***
73.6
76.0
⫹2.45***
African-American
56.5
66.9
⫹10.4***
58.6
62.9
⫹4.29***
Hispanic (English-speaking)
59.5
69.2
⫹9.69***
68.1
75.1
⫹7.00***
Hispanic (Spanish-speaking)c
38.5
52.4
⫹13.9***
45.0
54.0
⫹9.03***
White–African-American disparity
14.5*** (14.4, 14.6)
7.15*** (7.08, 7.23)
⫺7.37*** (⫺7.44, ⫺7.31)
15.0*** (14.8, 15.2)
13.1*** (13.0, 13.3)
⫺1.83*** (⫺1.95, ⫺1.72)
White–Hispanic (English-speaking) disparity
11.5*** (11.5, 11.6)
4.91*** (4.85, 4.97)
⫺6.64*** (⫺6.69, ⫺6.584)
5.45*** (5.37, 5.52)
0.903*** (0.890, 0.917)
⫺4.54*** (⫺4.61, ⫺4.48)
White–Hispanic (Spanish-speaking) disparityc
32.6*** (32.5, 32.7)
21.7*** (21.5, 21.9)
⫺10.9*** (⫺11.0, ⫺10.7)
28.6*** (28.4, 28.8)
22.0*** (21.8, 22.2)
⫺6.58*** (⫺6.75, ⫺6.41)
Vaccination rate Crude
Hispanic (English-speaking) c
Hispanic (Spanish-speaking)
www.ajpm-online.net
Note: Values within parentheses are 95% CIs. a From the 2000 –2001 season to the 2001–2002 season b From the 2001–2002 season to the 2002–2003 season c Used Spanish in at least one of the MCBS interviews d Adjusted by all explanatory variables listed in Table 1 *p⬍0.05, ⴱⴱp⬍0.01,***p⬍0.001
Yoo et al / Am J Prev Med 2011;40(1):1–10
Adjustedd
January 2011
Table 3. Changes in racial/ethnic disparities in influenza vaccination rate in Period 3a– 4b among Medicare elderly (% vaccinated) Period 3 2002–2003 season (no supply problem)
2003–2004 season (moderate late-season shortage)
White
73.1
African-American
60.2
Vaccination rate
Period 4 Change from 2002–2003 to 2003–2004 season
2003–2004 season (moderate late-season shortage)
2004–2005 season (severe supply shortage)
Change from 2003–2004 to 2004–2005 season
77.0
⫹3.90***
77.6
71.1
⫺6.48***
63.5
⫹3.30***
53.5
43.0
⫺10.5***
Crude
74.6
⫹3.97***
71.1
57.5
⫺13.6***
43.7
⫺5.93***
33.1
30.7
⫺2.39***
12.8*** (12.7, 13.0)
13.4*** (13.3, 13.6)
⫹0.60*** (0.57, 0.63)
24.0*** (23.9, 24.2)
28.1*** (28.0, 28.2)
⫹4.04*** (4.00, 4.08)
2.4*** (2.23, 2.53)
2.3*** (2.18, 2.46)
⫺0.07*** (⫺0.08, ⫺0.05)
6.47*** (6.32, 6.63)
13.6*** (13.4, 13.7)
⫹7.08*** (7.06, 7.11)
23.4*** (23.2, 23.6)
33.2*** (33.0, 33.4)
⫹9.83*** (9.79, 9.87)
44.5*** (44.3, 44.7)
40.4*** (40.2, 40.6)
⫺4.09*** (⫺4.14, ⫺4.05)
White
72.4
76.9
⫹4.51***
77.0
70.1
⫺6.83***
African-American
62.7
66.3
⫹3.68***
57.2
48.1
⫺9.13***
Hispanic (English-speaking)
70.1
75.4
⫹5.32***
73.3
60.4
⫺12.9***
Hispanic (Spanish-speaking)c
64.2
36.4
⫺27.8***
40.9
40.7
⫺0.274
White–African-American disparity
9.73*** (9.58, 9.89)
10.6*** (10.4, 10.7)
⫹0.830*** (0.742, 0.925)
19.7*** (19.5, 19.9)
22.0*** (21.9, 22.2)
⫹2.30*** (2.13, 2.47)
White–Hispanic (English-speaking) disparity
2.30*** (2.26, 2.35)
1.50*** (1.47, 1.53)
⫺0.805*** (⫺0.829, ⫺0.781)
3.68*** (3.62, 3.74)
9.73*** (9.63, 9.83)
⫹6.05*** (5.97, 6.13)
White–Hispanic (Spanish-speaking) disparityc
8.16*** (8.03, 8.30)
40.5*** (40.3, 40.7)
36.0*** (35.8, 36.3)
29.5*** (29.3, 29.6)
⫺6.56*** (⫺6.74, ⫺6.38)
c
Hispanic (Spanish-speaking) White–African-American disparity White–Hispanic (English-speaking) disparity White–Hispanic (Spanish-speaking) disparityc Adjustedd
⫹32.3*** (32.2, 32.5)
Yoo et al / Am J Prev Med 2011;40(1):1–10
70.7 49.7
Hispanic (English-speaking)
Note: Values within parentheses are 95% CIs. From the 2002–2003 season to the 2003–2004 season b From the 2003–2004 season to the 2004 –2005 season c Used Spanish in at least one of the MCBS interviews d Adjusted by all explanatory variables listed in Table 1 *p⬍0.05, ⴱⴱp⬍0.01,***p⬍0.001 a
5
Yoo et al / Am J Prev Med 2011;40(1):1–10
6
% vaccinated
% vaccinated
Moderate supply delay Severe supply delay erate supply shortage 80 (CDC. Cumulative Adjusted monthly U.S. influenza vaccination rate vaccine distribution, 60 1999–2006. UnpubAdjusted lished data, 2006).8,11,12 disparity with In Period 1, with the white 40 largest vaccine supply increase, the adjusted Crude vaccination vaccination rate inrate 20 creased among W, AA, EH, and SH groups by 3.05%, 10.4%, 9.69%, 0 and 13.9%, respectively White African- Hispanic Hispanic African- Hispanic Hispanic White American (English- (SpanishAmerican (English- (Spanish(Table 2). Similarly in a speaking) speaking ) speaking) speakinga) Period 4, with the largest vaccine supply de2001–2002 season 2000–2001 season cline, these rates deFigure 1. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 1 creased among W, AA, (2000–2001 season to the 2001–2002 season) among Medicare elderly (detailed data in Table 2). EH, and SH by 6.83%, All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes 9.13%, 12.9%, and in racial/ethnic disparity across two consecutive seasons during the same period were 0.27%, respectively. significant (p⬍0.001). a Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews During Periods 1 and 2, when vaccine During Period 4, when vaccine supply substantially supply improved, racial/ethnic disparities narrowed. That is, worsened, adjusted W–AA and W–EH disparities increased the adjusted W–AA disparity decreased by 7.37%, W–EH by 2.30% (p⬍0.001) and 6.05% (p⬍0.001), respectively, as disparity decreased by 6.64%, and W–SH disparity decreased by 10.9% (all p⬍0.001) during Period 1 (Table 2). Similarly, hypothesized. However, W–SH disparity decreased by during Period 2, adjusted W–AA, W–EH, No supply delay Moderate supply delay and W–SH disparities 80 decreased by 1.83%, 4.54%, and 6.58% (all Adjusted p⬍0.001), respectively. vaccination 60 During Period 3, rate when vaccine supply Adjusted did not change but disparity with white 40 there was markedly high demand late in Crude vaccination the vaccination period rate 20 (CDC. Cumulative monthly U.S. influenza vaccine distri0 bution, 1999 –2006. White African- Hispanic Hispanic African- Hispanic Hispanic White Unpublished data, American (English- (SpanishAmerican (English- (Spanishspeaking) speaking ) speaking) speaking ) 2006),8,11,12 the adjusted W–AA and 2002–2003 season 2001–2002 season W–EH disparities changed little (–0.8% Figure 2. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 2 (2001–2002 season to the 2002–2003 season) among Medicare elderly (detailed data in Table 2). to0.8%)andtheW–SH All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes disparity increased in racial/ethnic disparity across two consecutive seasons during the same period were (32%, p⬍0.001) as significant (p⬍0.001). a hypothesized. Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews a
a
www.ajpm-online.net
Yoo et al / Am J Prev Med 2011;40(1):1–10
7
cases: the W–SH disparity changed direc80 tion (from crude Adjusted vaccination 6.20% to adjusted rate 6.58%) in Period 2, 60 and increased subAdjusted stantially in magnidisparity with white tude (from 9.83% to 40 32.3%) in Period 3. Crude A secondary analvaccination rate 20 ysis measured the proportion of individuals from each ra0 cial/ethnic group who White African- Hispanic Hispanic African- Hispanic Hispanic White stated that vaccine American (English- (SpanishAmerican (English- (Spanisha a unavailability/shortage speaking) speaking ) speaking) speaking ) was “one reason” for 2003–2004 season 2002–2003 season being unvaccinated. This analysis demonFigure 3. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 3 (2002–2003 season to the 2003–2004 season) among Medicare elderly (detailed data in Table 2). strates three points. All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes First, for all racial/ethin racial/ethnic disparity across two consecutive seasons during the same period were nic groups, this prosignificant (p⬍0.001). a portion was higher Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews during the two sea4.09% in the crude rate and by 6.56% in the adjusted rate sons with the most severe shortages or delays (2000 –2001 (i.e., from 36.0% to 29.5%, p⬍0.001), against the current and 2004 –2005). Second, disparities in those reporting they hypothesis. were unvaccinated because of shortages tended to be smaller Adjustment with a multivariable model generally yielded when vaccine supply increased in Periods 1 through 3. vaccination rates quite similar to crude rates except in two Third, in Period 4, a seemingly reverse Severe supply shortage Moderate late-season shortage time trend of W–SH 80 disparity was found Adjusted vaccination with regard to this rate proportion. This was 60 because the increase Adjusted in this proportion disparity with white was 2% among the 40 SH population, which Crude was lower than the vaccination 8%–9% increase amrate 20 ong the other three groups. Moderate late-season shortage
% vaccinated
% vaccinated
No supply delay
0 White
African- Hispanic Hispanic American (English- (Spanishspeaking) speakinga)
2003–2004 season
White
African- Hispanic Hispanic American (English- (Spanishspeaking) speakinga)
2004–2005 season
Figure 4. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 4 (2003–2004 season to the 2004 –2005 season) among Medicare elderly (detailed data in Table 2). All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes in racial/ethnic disparity across two consecutive seasons during the same period were significant (p⬍0.001), except White-Hispanic (Spanish speaking) disparity. a Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews January 2011
Discussion The results support the hypothesis that racial/ethnic disparities in influenza vaccination rates are reduced when vaccine supply is improved, and disparities are
8
Yoo et al / Am J Prev Med 2011;40(1):1–10
aggravated when vaccine supply is delayed or decreased. The one exception is the SH population in Period 4. The fındings from the primary analysis are supported by the parallel fındings of the secondary analysis which indicated that vaccine supply seems to affect AfricanAmerican and Hispanic populations more than nonHispanic white populations. Patient factors included in the analyses (Table 1) changed little during each period. The current supplemental analyses, which included additional patient factors among subpopulations (the number of medical encounters, vaccinationonly visits, and resistant attitudes and beliefs regarding vaccination), did not fınd a substantial change unique to a specifıc racial/ethnic group in each period. We speculate that the one exception to the general trend—the W-SH disparity trend—is due to a “floor effect,” as the adjusted vaccination rate among SH was as low as 40.9% (lower than that for W by 36%) in the fırst season in Period 4, and declined by only 0.27% among SH (compared to a 6.83% among W) when vaccine supply decreased in the subsequent season. Consequently, the adjusted W–SH disparity decreased by 6.56% during the second season in the period. Three possible causal pathways might conceivably explain these results: (1) Providers located in areas with a higher proportion of underserved racial/ethnic populations may have experienced a relatively larger drop in influenza vaccine provision during a season with a serious vaccine supply problem; (2) underserved racial/ ethnic populations may have been less likely to receive or respond to information on a vaccine supply shortage through a reminder or a communication campaign; and (3) fewer clinic visits (particularly vaccination-only visits) might have led to proportionately fewer opportunities for vaccination among underserved racial/ethnic populations. These possible causal pathways also were suggested indirectly by a quasi “dose–response” relationship. Disparity changes tended to be greater in Periods 1 and 4, with a greater difference in vaccine supply between seasons, compared to seasons in Periods 2 and 3 (with one exception—the W–SH disparity was larger in Period 3 than in other periods). The exception could be due to the Spanish-speaking Hispanic population’s vaccination pattern being potentially more sensitive to a late season vaccine shortage and also due to imprecision in the estimates for the Spanish-speaking Hispanic group because of their relatively small sample size. In general, the present study observed disparities comparable to those previously reported for African Americans 3,7,16 –19 and Hispanics.3 Consistent with the literature,3,7,16 –19 disparities persisted following adjustment for multiple confounders.
Our fındings differ from those of Link et al.,17 who observed no signifıcant change in disparities by race or ethnicity among adults aged ⱖ65 years during the 2004 –2005 season compared to prior seasons based on the Behavioral Risk Factor Surveillance System (BRFSS). Although they reported statistically signifıcant cross-sectional disparities by race and ethnicity, disparities (0.48 – 0.58 in OR) were generally smaller than those found in the present study (0.24 – 0.69 in OR). Differences in estimates between that study and the current one may reflect differences in design (cross-sectional vs cohort); sampling (random-digitdialing vs sampling from Medicare); survey administration (phone-based vs in-person); response rates (52.7% vs 73.6%); and control variables (e.g., language and smoking).9,17 Thus, caution is needed in comparing the current results in Tables 2 and 3 with those reported by Link and colleagues.17 For instance, the crude vaccination rate among African Americans in the 2001–2002 season was 61.9% and 52.8% in Period 1 and 2 (including a different cohorts), respectively (Table 2). If the current study had used a cross-sectional study design, this rate was almost equivalent to the average of these two cohorts, 57.4% [⫽ (61.9%⫹52.8%)/2], which is close to 57.6% reported in Link’s study.17 Accounting for this point, over the fıve seasons since the 2000 –2001 season, two data sources (MCBS and BRFSS) indicated that African-American and Hispanic elderly populations consistently experienced 10% and 15% lower vaccination rates, respectively.17 The MCBS analysis in the current study showed a greater gap (⬎25%) among the Spanish-speaking Hispanic elderly population. In other words, the vaccination change across seasons can be masked or underestimated by the changes in panel subjects in a cross-sectional study. This problem was avoided by the cohort analysis. However, a cohort analysis can include only those benefıciaries who lived two consecutive influenza seasons and hence may be healthier than the general community-dwelling elderly Medicare benefıciaries. This may somewhat diminish the generalizability of the results.
Limitations Although it is possible that this analysis omitted factors that affected disparities in vaccination rates, it is unlikely that the current fındings were confounded strongly for two reasons. First, some omitted individual factors (e.g., number of clinic visits),7,15,18 and area characteristics (e.g., primary care physicians per adults aged ⱖ65 years)7 have shown only limited effects on racial/ethnic disparities. Second, it is unlikely that the differential change in these omitted factors among racial/ethnic groups over the course of 2 years would be suffıcient to explain the current fındings. www.ajpm-online.net
Yoo et al / Am J Prev Med 2011;40(1):1–10
Additionally, data limitations did not allow the present study to account for three potentially related factors. These include (1) national or local vaccination campaigns; (2) effects of prior vaccine shortage on patient, provider, and health system expectations and unanticipated behaviors (an expected shortage may increase demand22 or may conversely discourage patients from seeking vaccine because of their perception of permanent shortage11); and (3) provider discriminatory behavior. Insofar as campaigns were distributed and targeted randomly, they would bias results toward the null. It is unclear how experiences with a prior shortage might affect different racial/ethnic groups or providers. However, it is not likely that provider discrimination would change over two seasons. The current study used self-reported data, not claims data, to determine the vaccination status despite the potential recall bias. Self-reported data in adults aged ⱖ65 years generally have good sensitivity (94%–100%) but relatively low specifıcity (38%–70%).23,24 However, any recall bias that differs by race/ethnicity is unlikely to be affected by vaccine supply.
Policy Implications Because reduced influenza vaccine supply could potentially increase racial/ethnic disparities, this analysis indicates that efforts to reduce vaccination disparities should focus on ensuring an adequate and consistent vaccine supply to improve overall vaccination rates. Specifıc policy implications to address disparities in immunization that might mitigate the impact of decreases in supply might include stabilizing the vaccine supply; the creation of an adult program similar to the Vaccines-for-Children (VFC) program to sustain delivery of vaccines to safety-net providers with limited vaccine investment resources such as federally qualifıed health centers and practices that serve large proportions of African-American and Hispanic patients; and active provider and patient reminder/recall systems and targeted communication campaigns.25
Conclusion Improved vaccine supply appears to be associated with reduced racial/ethnic disparities in influenza vaccination rates while constrained supply is associated with worsening disparities. In general, the elimination of racial and ethnic disparities in adult immunization will require resources for immediate implementation of evidence-based strategies for improving influenza immunization in settings that serve African-American and Hispanic patients. This study is supported by NIH/National Institute of Allergy and Infectious Disease (1K25AI073915). January 2011
9
No fınancial disclosures were reported by the authors of this paper.
References 1. Offıce of Disease Prevention and Health Promotion. USDHHS. Healthy People 2010. 2007; Available from: www.cdc.gov/nchs/ about/otheract/hpdata2010/abouthp.htm. 2. Fiscella K. Commentary—anatomy of racial disparity in influenza vaccination. Health Serv Res 2005;40(2):539 – 49. 3. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare benefıciaries. Health Serv Res 2005;40(2):517–37. 4. CDC. Prevention and control of influenza: Recommendations of the Advisory Committee on Immunization Practices (ACIP). Morb Mortal Wkly Rep 2007;56(RR06):1–54. 5. Nichol KL. Influenza vaccination in the elderly: impact on hospitalization and mortality. Drugs Aging 2005;22(6):495–515. 6. Maciosek MV, Coffıeld AB, Edwards NM, Flottemesch TJ, Goodman MJ, Solberg LI. Priorities among effective clinical preventive services: results of a systematic review and analysis. Am J Prev Med 2006;31(1):52– 61. 7. O’Malley AS, Forrest CB. Immunization disparities in older Americans: determinants and future research needs. Am J Prev Med 2006;31(2):150 – 8. 8. Wallace GS. Influenza vaccine distribution 2006 – 07: National Vaccine Advisory Committee, CDC; 2007, February 5. 9. Centers for Medicare & Medicaid Services. Medicare current benefıciary survey: CY 2000 –2006. Baltimore MD; 2006. 10. CDC. Experiences with obtaining influenza vaccination among persons in priority groups during a vaccine shortage—U.S., October–November, 2004. Morb Mortal Wkly Rep 2004;53(49):1153–5. 11. Charatan F. Widespread flu in U.S. exposes shortage of vaccine. BMJ 2004;328(7430):8. 12. Nelson R. Influenza vaccine shortage hits the U.S. Lancet 2003; 362(9401):2075. 13. Phillips-Caesar E, Coady MH, Blaney S, et al. Predictors of influenza vaccination in an urban community during a national shortage. J Health Care Poor Underserved 2008;19(2):611–24. 14. Fiscella K, Franks P, Doescher MP, Saver BG. Disparities in health care by race, ethnicity, and language among the insured: fındings from a national sample. Med Care 2002;40(1):52–9. 15. Rangel MC, Shoenbach VJ, Weigle KA, Hogan VK, Strauss RP, Bangdiwala SI. Racial and ethnic disparities in influenza vaccination among elderly adults. J Gen Intern Med 2005;20(5):426 –31. 16. Egede LE, Zheng D. Racial/ethnic differences in influenza vaccination coverage in high-risk adults. Am J Public Health 2003;93(12):2074 – 8. 17. Link MW, Ahluwalia IB, Euler GL, Bridges CB, Chu SY, Wortley PM. Racial and ethnic disparities in influenza vaccination coverage among adults during the 2004 –2005 season. Am J Epidemiol 2006; 163(6):571– 8. 18. Marin MG, Johanson WG Jr, Salas-Lopez D. Influenza vaccination among minority populations in the U.S. Prev Med 2002; 34(2):235– 41. 19. Schneider EC, Cleary PD, Zaslavsky AM, Epstein AM. Racial disparity in influenza vaccination: does managed care narrow the gap between African Americans and whites? JAMA 2001;286(12): 1455– 60. 20. Yoo BK, Frick K. Determinants of influenza vaccination timing. Health Econ 2005;14(8):777–91.
10
Yoo et al / Am J Prev Med 2011;40(1):1–10
21. Nichol KL, Wuorenma J, von Sternberg T. Benefıts of influenza vaccination for low-, intermediate-, and high-risk senior citizens. Arch Intern Med 1998;158(16):1769 –76. 22. Ompad DC, Galea S, Blaney S, et al. Access to influenza vaccine in East Harlem and the Bronx during a national vaccine shortage. J Community Health 2007;32(3):195–202. 23. Zimmerman RK, Raymund M, Janosky JE, Nowalk MP, Fine MJ. Sensitivity and specifıcity of patient self-report of influenza
and pneumococcal polysaccharide vaccinations among elderly outpatients in diverse patient care strata. Vaccine 2003;21: 1486 –91. 24. MacDonald R, Baken L, Nelson A, Nichol KL. Validation of self-report of influenza and pneumococcal vaccination status in elderly outpatients. Am J Prev Med 1999;16:173–7. 25. Ompad DC, Galea S, Vlahov D. Distribution of influenza vaccine to high-risk groups. Epidemiol Rev 2006;28:54 –70.
Did you know? Two AJPM articles per issue offer CME credits. Go to www.ajpm-online.net/cme/home for more information.
www.ajpm-online.net
Background: The impact of vaccine shortages on disparities in influenza vaccination is uncertain. Purpose: The objective of this research was to examine the association between influenza vaccine supply and racial/ethnic disparities in vaccination rates among elderly Medicare benefıciaries. Methods: Cross-sectional multivariable logistic regression analyses were performed in 2010 to examine whether racial/ethnic disparities in vaccination rates changed across two consecutive seasons: from (Period 1) 2000 –2001 and 2001–2002 seasons through (Period 4) 2003–2004 and 2004 –2005 seasons. Self-reported receipt of influenza vaccine across consecutive years was examined among communitydwelling non-Hispanic African-American (AA); non-Hispanic white (W); English-speaking Hispanic (EH); and Spanish-speaking Hispanic (SH) elderly enrolled in the Medicare Current Benefıciary Survey (unweighted n⫽2306 –2504, weighted n⫽8.23– 8.99 million for Periods 1 through 4).
Results: During Periods 1 and 2, when vaccine supply increased nationally, adjusted racial/ethnic disparities in the influenza vaccination rate decreased by 1.8%–7.4% (W–AA disparity); 4.5%– 6.6% (W–EH disparity); and 6.6%–11% (W–SH disparity) (all p⬍0.001). During Period 4, when vaccine supply declined, adjusted disparities in vaccination rates increased by 2.3% (W–AA disparity) and 6.1% (W–EH disparity) but decreased by 6.6% (W–SH disparity) probably due to a “floor effect” (constant low rates among SH; all p⬍0.001). Conclusions: Improved vaccine supply was generally associated with reduced racial/ethnic disparities in influenza vaccination rates, whereas worse supply was associated with increased disparities. To avoid future widening of racial health disparities, policy options include stabilizing the vaccine supply and preferential delivery of vaccines to safety-net providers serving AA and Hispanic populations during a shortage. (Am J Prev Med 2011;40(1):1–10) © 2011 American Journal of Preventive Medicine
Introduction
E
liminating health disparities is an overarching goal of Healthy People 2010.1 Racial/ethnic disparities have been consistently documented for influenza vaccination among adults aged ⱖ65 years,2– 4 with Hispanic and African-American populations having lower influenza vaccination rates than white populations, despite evidence that influenza vaccination reduces influenza-related morbidity and is cost effective.5,6 In one nationally represen-
From the Department of Community and Preventive Medicine (Yoo, Kasajima), Provost’s Offıce (Phelps), Department of Family Medicine (Fiscella), Department of Medicine (Bennett), Department of Pediatrics (Szilagyi), University of Rochester, Rochester, New York Address correspondence to: Byung-Kwang Yoo, MD, PhD, 601 Elmwood Avenue, Box 644, Rochester NY 14642. E-mail: [email protected]. 0749-3797/$17.00 doi: 10.1016/j.amepre.2010.09.028
tative study,7 African Americans were 34% less likely to receive influenza vaccination than white Americans. Disparities in influenza vaccination may be exacerbated when vaccine supply declines or is delayed as vulnerable populations, including underserved racial and ethnic populations, may face relatively greater barriers to accessing vaccination during such times. There have been fıve seasons of delay/shortage since 2000 (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006),8 and the prospect of future supply problems remains.4 Roughly 15% of Medicare benefıciaries report diffıculties in receiving an influenza vaccination because of supply issues,9 and many high-priority groups reported that they did not attempt vaccination because of their perception of a shortage during the 2004–2005 season.10 One local population study noted that African-American adults were less likely to be vaccinated during a serious vaccine shortage season (2004 –2005), compared to a prior season with a moderate shortage relative to demand late in the season
© 2011 American Journal of Preventive Medicine • Published by Elsevier Inc.
Am J Prev Med 2011;40(1)1–10 1
2
Yoo et al / Am J Prev Med 2011;40(1):1–10
(2003–2004) (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8,11–13 Several challenges exist in assessing the impact of vaccine supply on racial/ethnic disparities. One challenge is that multiple factors such as overall coverage, supply, and health status vary over time. Thus analyses that compare populations need to adjust for underlying differences in these populations as well as time trends. A cross-sectional analysis cannot adequately control for potential time-varying confounders that may disproportionally affect people in each racial/ethnic group. A cohort analysis is more suitable because potentially confounding variables within the same population are likely to change relatively little between two consecutive seasons. The objective of the present study was to evaluate the association between influenza vaccine supply and racial/ ethnic disparities in vaccination rates across two consecutive vaccination seasons among nationally representative elderly Medicare benefıciaries. It utilized a cohort analysis of populations followed over two consecutive seasons with different vaccine supply. Thus, the current study provides a rigorous test of the hypothesis that racial/ethnic disparities in influenza vaccination rates are worse when vaccine supply is delayed or limited.
Methods Design Overview The 2000 –2005 Medicare Current Benefıciary Surveys (MCBSs)9 were analyzed in 2010 to examine levels and time trends in racial/ ethnic disparities across two consecutive vaccination seasons within each of four periods: Period 1, 2000 –2001 and 2001–2002 seasons; Period 2, 2001–2002 and 2002–2003 seasons; Period 3, 2002–2003 and 2003–2004 seasons; and Period 4, 2003–2004 and 2004 –2005 seasons. All analyses used Stata, version 10. MCBS is a multipurpose survey of a nationally representative sample of the Medicare population.9 Sample people are surveyed about their health experience three times a year over 4 years. Survey data are combined with participant Medicare claims. Survey response rates were 66%– 82% during the current study period. The 2000 –2001 season was characterized by a severe delay in supply, followed by improvement in2001–2002withmoderatedelay(CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8 The 2003–2004 season had a moderate shortage relative to demand late in the season because of unusually high demand and was followed by a severe shortage in the 2004 – 2005 season (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8,11,12
Setting and Participants The study population consisted of community-dwelling non-Hispanic African-American (African-American); non-Hispanic white (white); and Hispanic Medicare benefıciaries (aged ⱖ65 years). Analyses for each period examined the same cohort, that, is those who were enrolled in MCBS for two consecutive influenza seasons (i.e., unweighted n⫽2504, 2501, 2306, and 2411; and weighted n⫽8.73, 8.95,
8.23, and 8.99 million for Periods 1, 2, 3, and 4, respectively). Because prior research14,15 suggests that white–Hispanic disparities are largely explained by language differences, English-speaking Hispanic (EH) and Spanish-speaking Hispanic (SH) benefıciaries were distinguished based on whether Spanish was used in MCBS interviews.
Statistical Analysis Crude race/ethnicity-specifıc influenza vaccination rates were calculated fırst, and white–African-American (W–AA) and white– Hispanic (W–EH and W–SH separately) disparities were assessed for each season. Then, the changes in racial/ethnic disparities were measured for each period. Primary analysis (Tables 1–3 and Figures 1–4) used self-reported vaccine receipt data because (1) claims data were not complete for Medicare managed care plan (HMO) enrollees and (2) claims data do not capture vaccination given in settings that do not bill Medicare. Specifıcally, claims data underestimated the vaccination rate by 20%– 24% in the current study’s populations. Self-reported vaccine receipt data are collected in the fall of each year through the MCBS with the question Did you have a flu shot for last winter? The primary analysis, a cross-sectional multivariable logistic regression model, examined whether changes in racial/ethnic disparities remained after adjusting for demographics, socioeconomic factors, and other patient factors (Table 1) that have been found3,7,16 –21 to be related to influenza vaccine receipt among U.S. elderly adults aged ⬎65 years. Race/ethnicity-specifıc adjusted vaccination rates were predicted in this regression model, assuming all subjects in the model belong to a certain racial/ethnic group. Using the adjusted rates, this analysis calculated the racial/ethnic disparities in influenza vaccination rates per season and changes in disparities for each period. Next, the adjusted Wald test determined whether these disparities and differences in disparities between sequential seasons were statistically different from zero. A common set of cross-sectional weights was applied for each period.
Results Table 1 indicates that the patient factors changed little between the two consecutive seasons in both periods, as expected. Vaccine supply differed substantially at a monthly level and at a cumulative annual level (in December) between the two consecutive seasons in all periods except Period 3 (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006).8 Figures 1– 4 show crude and adjusted vaccination rates for the fırst and second season of each period. Tables 2 and 3 display the change in vaccination rates between the two seasons. These data show racial/ethnic disparities by season and corresponding changes in disparities between seasons where vaccine supply improved (Periods 1 and 2) and worsened (Periods 3 and 4). For all groups, the vaccination rate generally increased when vaccinesupplyimproved(Periods1and2)anddecreasedwhen vaccinesupplyworsened(Period4).Despitethesimilarvaccine supply between the two seasons in Period 3, vaccination rates increased in the 2003–2004 season during which an unusually early onset of a relatively severe epidemic led to higher vaccine demand late in the vaccination period and a consequent modwww.ajpm-online.net
January 2011
Table 1. Study population characteristics and influenza vaccine supply
Influenza season (vaccine supply) Cumulative influenza vaccine supply (million doses)a
Period 1
Period 2
Period 3
Period 4
(nⴝ2504, weighted nⴝ8.73 million)
(nⴝ2501, weighted nⴝ8.95 million)
(nⴝ2306, weighted nⴝ8.23 million)
(nⴝ2411, weighted nⴝ8.99 million)
2000–2001 (severe delay)
2001–2002 (moderate delay)
2001–2002 (moderate delay)
2002–2003 (no supply problem)
2002–2003 (no supply problem)
2003–2004 (moderate late-season shortage)
2003–2004 (moderate late-season shortage)
2004–2005 (severe shortage)
Nov⫽48.2 Dec⫽70.4
Nov⫽76.2 Dec⫽77.7
Nov⫽76.2 Dec⫽77.7
Nov⫽83.0 Dec⫽83.3
Nov⫽83.0 Dec⫽83.3
Nov⫽80.6 Dec⫽82.6
Nov⫽80.6 Dec⫽82.6
Nov⫽52.1 Dec⫽56.4
Outcome variable: influenza vaccine receipt (%) 68.4
72.6
70.9
73.9
71.2
74.8
73.8
66.8
84.9
84.9
83.7
83.7
84.2
84.2
84.4
84.4
Explanatory variablesb (%) Non-Hispanic white Hispanic (English-speaking) Hispanic (Spanish-speaking)
c
Non-Hispanic African-American
3.87
3.87
3.86
3.68
4.68
4.80
3.86
3.86
2.91
2.91
4.17
4.35
3.27
3.15
3.40
3.40
8.28
8.28
8.32
8.32
7.80
7.80
8.30
8.30
Female
58.4
58.4
57.7
57.7
58.1
58.1
57.4
57.4
Aged ⱖ75 years during first season
44.8
48.8
44.9
49.8
45.8
49.4
44.8
49.5
Medicaid enrollment
7.43
8.06
8.76
9.71
8.86
9.00
9.40
9.70
Subjective general health status (fair/poor)
18.8
20.1
19.3
20.3
17.1
17.8
18.3
18.7
High-risk chronic conditions for influenzad
35.9
39.7
39.7
41.9
39.4
40.5
43.5
44.6
Yoo et al / Am J Prev Med 2011;40(1):1–10
Influenza vaccine receipt (survey data)
a
Cumulative number of influenza vaccine doses distributed among the general U.S. population (CDC. Cumulative monthly U.S. influenza vaccine distribution, 1999 –2006. Unpublished data, 2006)8 b Additional explanatory variables included education (high school diploma); income (ⱖ$30,000); number in household: two or more; metropolitan residence; private supplemental health insurance (Medigap); activity of daily living (ADL) limitation: one or more; instrumental activity of daily living (IADL) limitation: one or more; avoid medical care when sick; see physician soon when sick; presence of physician regularly consult with8; census region. All explanatory variables were dichotomous variables. c Used Spanish in at least one of MCBS interviews d Those with heart disease, lung disease, diabetes, renal disease, rheumatologic disease, dementia and/or stroke, identified by (1) Medicare claims data (inpatient and outpatient) associated with MCBS survey data among Medicare fee-for-service (FFS) enrollees and (2) both survey and claims data among Medicare managed care enrollees because of limited availability of claims21,22 MCBS, Medicare Current Beneficiary Survey
3
4
Table 2. Changes in racial/ethnic disparities in influenza vaccination rate in Period 1a–2b among Medicare elderly (% vaccinated) Period 1
Period 2
2000–2001 season (severe supply delay)
2001–2002 season (moderate supply delay)
Change from 2000–2001 to 2001–2002 season
2001–2002 season (moderate supply delay)
2002–2003 season (no supply problem)
Change from 2001–2002 to 2002–2003 season
White
71.3
74.8
⫹3.47***
73.9
76.9
⫹2.97***
African-American
53.7
61.9
⫹8.29***
52.8
58.1
⫹5.30***
60.6
67.6
⫹6.97***
64.8
68.6
⫹3.79***
35.6
45.4
⫹9.84***
52.7
49.5
⫺3.23***
White–African-American disparity
17.6*** (17.5, 17.8)
12.8*** (12.7, 12.9)
⫺4.82*** (⫺4.85, ⫺4.79)
21.2*** (21.1, 21.3)
18.8*** (18.7, 19.0)
⫺2.33*** (⫺2.37, ⫺2.29)
White–Hispanic (Englishspeaking) disparity
10.7*** (10.6, 10.9)
7.22*** (7.06, 7.38)
⫺3.51*** (⫺3.53, ⫺3.48)
9.10*** (8.96, 9.29)
8.32*** (8.14, 8.46)
⫺0.82*** (⫺0.85, ⫺0.79)
White–Hispanic (Spanishspeaking) disparityc
35.7*** (35.6, 35.9)
29.4*** (29.2, 29.56)
⫺6.38*** (⫺6.42, ⫺6.33)
21.2*** (21.1, 21.4)
27.4*** (27.3, 27.6)
⫹6.20*** (6.16, 6.24)
White
71.0
74.1
⫹3.05***
73.6
76.0
⫹2.45***
African-American
56.5
66.9
⫹10.4***
58.6
62.9
⫹4.29***
Hispanic (English-speaking)
59.5
69.2
⫹9.69***
68.1
75.1
⫹7.00***
Hispanic (Spanish-speaking)c
38.5
52.4
⫹13.9***
45.0
54.0
⫹9.03***
White–African-American disparity
14.5*** (14.4, 14.6)
7.15*** (7.08, 7.23)
⫺7.37*** (⫺7.44, ⫺7.31)
15.0*** (14.8, 15.2)
13.1*** (13.0, 13.3)
⫺1.83*** (⫺1.95, ⫺1.72)
White–Hispanic (English-speaking) disparity
11.5*** (11.5, 11.6)
4.91*** (4.85, 4.97)
⫺6.64*** (⫺6.69, ⫺6.584)
5.45*** (5.37, 5.52)
0.903*** (0.890, 0.917)
⫺4.54*** (⫺4.61, ⫺4.48)
White–Hispanic (Spanish-speaking) disparityc
32.6*** (32.5, 32.7)
21.7*** (21.5, 21.9)
⫺10.9*** (⫺11.0, ⫺10.7)
28.6*** (28.4, 28.8)
22.0*** (21.8, 22.2)
⫺6.58*** (⫺6.75, ⫺6.41)
Vaccination rate Crude
Hispanic (English-speaking) c
Hispanic (Spanish-speaking)
www.ajpm-online.net
Note: Values within parentheses are 95% CIs. a From the 2000 –2001 season to the 2001–2002 season b From the 2001–2002 season to the 2002–2003 season c Used Spanish in at least one of the MCBS interviews d Adjusted by all explanatory variables listed in Table 1 *p⬍0.05, ⴱⴱp⬍0.01,***p⬍0.001
Yoo et al / Am J Prev Med 2011;40(1):1–10
Adjustedd
January 2011
Table 3. Changes in racial/ethnic disparities in influenza vaccination rate in Period 3a– 4b among Medicare elderly (% vaccinated) Period 3 2002–2003 season (no supply problem)
2003–2004 season (moderate late-season shortage)
White
73.1
African-American
60.2
Vaccination rate
Period 4 Change from 2002–2003 to 2003–2004 season
2003–2004 season (moderate late-season shortage)
2004–2005 season (severe supply shortage)
Change from 2003–2004 to 2004–2005 season
77.0
⫹3.90***
77.6
71.1
⫺6.48***
63.5
⫹3.30***
53.5
43.0
⫺10.5***
Crude
74.6
⫹3.97***
71.1
57.5
⫺13.6***
43.7
⫺5.93***
33.1
30.7
⫺2.39***
12.8*** (12.7, 13.0)
13.4*** (13.3, 13.6)
⫹0.60*** (0.57, 0.63)
24.0*** (23.9, 24.2)
28.1*** (28.0, 28.2)
⫹4.04*** (4.00, 4.08)
2.4*** (2.23, 2.53)
2.3*** (2.18, 2.46)
⫺0.07*** (⫺0.08, ⫺0.05)
6.47*** (6.32, 6.63)
13.6*** (13.4, 13.7)
⫹7.08*** (7.06, 7.11)
23.4*** (23.2, 23.6)
33.2*** (33.0, 33.4)
⫹9.83*** (9.79, 9.87)
44.5*** (44.3, 44.7)
40.4*** (40.2, 40.6)
⫺4.09*** (⫺4.14, ⫺4.05)
White
72.4
76.9
⫹4.51***
77.0
70.1
⫺6.83***
African-American
62.7
66.3
⫹3.68***
57.2
48.1
⫺9.13***
Hispanic (English-speaking)
70.1
75.4
⫹5.32***
73.3
60.4
⫺12.9***
Hispanic (Spanish-speaking)c
64.2
36.4
⫺27.8***
40.9
40.7
⫺0.274
White–African-American disparity
9.73*** (9.58, 9.89)
10.6*** (10.4, 10.7)
⫹0.830*** (0.742, 0.925)
19.7*** (19.5, 19.9)
22.0*** (21.9, 22.2)
⫹2.30*** (2.13, 2.47)
White–Hispanic (English-speaking) disparity
2.30*** (2.26, 2.35)
1.50*** (1.47, 1.53)
⫺0.805*** (⫺0.829, ⫺0.781)
3.68*** (3.62, 3.74)
9.73*** (9.63, 9.83)
⫹6.05*** (5.97, 6.13)
White–Hispanic (Spanish-speaking) disparityc
8.16*** (8.03, 8.30)
40.5*** (40.3, 40.7)
36.0*** (35.8, 36.3)
29.5*** (29.3, 29.6)
⫺6.56*** (⫺6.74, ⫺6.38)
c
Hispanic (Spanish-speaking) White–African-American disparity White–Hispanic (English-speaking) disparity White–Hispanic (Spanish-speaking) disparityc Adjustedd
⫹32.3*** (32.2, 32.5)
Yoo et al / Am J Prev Med 2011;40(1):1–10
70.7 49.7
Hispanic (English-speaking)
Note: Values within parentheses are 95% CIs. From the 2002–2003 season to the 2003–2004 season b From the 2003–2004 season to the 2004 –2005 season c Used Spanish in at least one of the MCBS interviews d Adjusted by all explanatory variables listed in Table 1 *p⬍0.05, ⴱⴱp⬍0.01,***p⬍0.001 a
5
Yoo et al / Am J Prev Med 2011;40(1):1–10
6
% vaccinated
% vaccinated
Moderate supply delay Severe supply delay erate supply shortage 80 (CDC. Cumulative Adjusted monthly U.S. influenza vaccination rate vaccine distribution, 60 1999–2006. UnpubAdjusted lished data, 2006).8,11,12 disparity with In Period 1, with the white 40 largest vaccine supply increase, the adjusted Crude vaccination vaccination rate inrate 20 creased among W, AA, EH, and SH groups by 3.05%, 10.4%, 9.69%, 0 and 13.9%, respectively White African- Hispanic Hispanic African- Hispanic Hispanic White American (English- (SpanishAmerican (English- (Spanish(Table 2). Similarly in a speaking) speaking ) speaking) speakinga) Period 4, with the largest vaccine supply de2001–2002 season 2000–2001 season cline, these rates deFigure 1. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 1 creased among W, AA, (2000–2001 season to the 2001–2002 season) among Medicare elderly (detailed data in Table 2). EH, and SH by 6.83%, All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes 9.13%, 12.9%, and in racial/ethnic disparity across two consecutive seasons during the same period were 0.27%, respectively. significant (p⬍0.001). a Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews During Periods 1 and 2, when vaccine During Period 4, when vaccine supply substantially supply improved, racial/ethnic disparities narrowed. That is, worsened, adjusted W–AA and W–EH disparities increased the adjusted W–AA disparity decreased by 7.37%, W–EH by 2.30% (p⬍0.001) and 6.05% (p⬍0.001), respectively, as disparity decreased by 6.64%, and W–SH disparity decreased by 10.9% (all p⬍0.001) during Period 1 (Table 2). Similarly, hypothesized. However, W–SH disparity decreased by during Period 2, adjusted W–AA, W–EH, No supply delay Moderate supply delay and W–SH disparities 80 decreased by 1.83%, 4.54%, and 6.58% (all Adjusted p⬍0.001), respectively. vaccination 60 During Period 3, rate when vaccine supply Adjusted did not change but disparity with white 40 there was markedly high demand late in Crude vaccination the vaccination period rate 20 (CDC. Cumulative monthly U.S. influenza vaccine distri0 bution, 1999 –2006. White African- Hispanic Hispanic African- Hispanic Hispanic White Unpublished data, American (English- (SpanishAmerican (English- (Spanishspeaking) speaking ) speaking) speaking ) 2006),8,11,12 the adjusted W–AA and 2002–2003 season 2001–2002 season W–EH disparities changed little (–0.8% Figure 2. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 2 (2001–2002 season to the 2002–2003 season) among Medicare elderly (detailed data in Table 2). to0.8%)andtheW–SH All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes disparity increased in racial/ethnic disparity across two consecutive seasons during the same period were (32%, p⬍0.001) as significant (p⬍0.001). a hypothesized. Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews a
a
www.ajpm-online.net
Yoo et al / Am J Prev Med 2011;40(1):1–10
7
cases: the W–SH disparity changed direc80 tion (from crude Adjusted vaccination 6.20% to adjusted rate 6.58%) in Period 2, 60 and increased subAdjusted stantially in magnidisparity with white tude (from 9.83% to 40 32.3%) in Period 3. Crude A secondary analvaccination rate 20 ysis measured the proportion of individuals from each ra0 cial/ethnic group who White African- Hispanic Hispanic African- Hispanic Hispanic White stated that vaccine American (English- (SpanishAmerican (English- (Spanisha a unavailability/shortage speaking) speaking ) speaking) speaking ) was “one reason” for 2003–2004 season 2002–2003 season being unvaccinated. This analysis demonFigure 3. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 3 (2002–2003 season to the 2003–2004 season) among Medicare elderly (detailed data in Table 2). strates three points. All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes First, for all racial/ethin racial/ethnic disparity across two consecutive seasons during the same period were nic groups, this prosignificant (p⬍0.001). a portion was higher Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews during the two sea4.09% in the crude rate and by 6.56% in the adjusted rate sons with the most severe shortages or delays (2000 –2001 (i.e., from 36.0% to 29.5%, p⬍0.001), against the current and 2004 –2005). Second, disparities in those reporting they hypothesis. were unvaccinated because of shortages tended to be smaller Adjustment with a multivariable model generally yielded when vaccine supply increased in Periods 1 through 3. vaccination rates quite similar to crude rates except in two Third, in Period 4, a seemingly reverse Severe supply shortage Moderate late-season shortage time trend of W–SH 80 disparity was found Adjusted vaccination with regard to this rate proportion. This was 60 because the increase Adjusted in this proportion disparity with white was 2% among the 40 SH population, which Crude was lower than the vaccination 8%–9% increase amrate 20 ong the other three groups. Moderate late-season shortage
% vaccinated
% vaccinated
No supply delay
0 White
African- Hispanic Hispanic American (English- (Spanishspeaking) speakinga)
2003–2004 season
White
African- Hispanic Hispanic American (English- (Spanishspeaking) speakinga)
2004–2005 season
Figure 4. Crude and adjusted influenza vaccination rates and racial/ethnic disparities in Period 4 (2003–2004 season to the 2004 –2005 season) among Medicare elderly (detailed data in Table 2). All racial/ethnic disparities within the same season were significant (p⬍0.001). All changes in racial/ethnic disparity across two consecutive seasons during the same period were significant (p⬍0.001), except White-Hispanic (Spanish speaking) disparity. a Used Spanish in at least one of the Medicare Current Beneficiary Survey interviews January 2011
Discussion The results support the hypothesis that racial/ethnic disparities in influenza vaccination rates are reduced when vaccine supply is improved, and disparities are
8
Yoo et al / Am J Prev Med 2011;40(1):1–10
aggravated when vaccine supply is delayed or decreased. The one exception is the SH population in Period 4. The fındings from the primary analysis are supported by the parallel fındings of the secondary analysis which indicated that vaccine supply seems to affect AfricanAmerican and Hispanic populations more than nonHispanic white populations. Patient factors included in the analyses (Table 1) changed little during each period. The current supplemental analyses, which included additional patient factors among subpopulations (the number of medical encounters, vaccinationonly visits, and resistant attitudes and beliefs regarding vaccination), did not fınd a substantial change unique to a specifıc racial/ethnic group in each period. We speculate that the one exception to the general trend—the W-SH disparity trend—is due to a “floor effect,” as the adjusted vaccination rate among SH was as low as 40.9% (lower than that for W by 36%) in the fırst season in Period 4, and declined by only 0.27% among SH (compared to a 6.83% among W) when vaccine supply decreased in the subsequent season. Consequently, the adjusted W–SH disparity decreased by 6.56% during the second season in the period. Three possible causal pathways might conceivably explain these results: (1) Providers located in areas with a higher proportion of underserved racial/ethnic populations may have experienced a relatively larger drop in influenza vaccine provision during a season with a serious vaccine supply problem; (2) underserved racial/ ethnic populations may have been less likely to receive or respond to information on a vaccine supply shortage through a reminder or a communication campaign; and (3) fewer clinic visits (particularly vaccination-only visits) might have led to proportionately fewer opportunities for vaccination among underserved racial/ethnic populations. These possible causal pathways also were suggested indirectly by a quasi “dose–response” relationship. Disparity changes tended to be greater in Periods 1 and 4, with a greater difference in vaccine supply between seasons, compared to seasons in Periods 2 and 3 (with one exception—the W–SH disparity was larger in Period 3 than in other periods). The exception could be due to the Spanish-speaking Hispanic population’s vaccination pattern being potentially more sensitive to a late season vaccine shortage and also due to imprecision in the estimates for the Spanish-speaking Hispanic group because of their relatively small sample size. In general, the present study observed disparities comparable to those previously reported for African Americans 3,7,16 –19 and Hispanics.3 Consistent with the literature,3,7,16 –19 disparities persisted following adjustment for multiple confounders.
Our fındings differ from those of Link et al.,17 who observed no signifıcant change in disparities by race or ethnicity among adults aged ⱖ65 years during the 2004 –2005 season compared to prior seasons based on the Behavioral Risk Factor Surveillance System (BRFSS). Although they reported statistically signifıcant cross-sectional disparities by race and ethnicity, disparities (0.48 – 0.58 in OR) were generally smaller than those found in the present study (0.24 – 0.69 in OR). Differences in estimates between that study and the current one may reflect differences in design (cross-sectional vs cohort); sampling (random-digitdialing vs sampling from Medicare); survey administration (phone-based vs in-person); response rates (52.7% vs 73.6%); and control variables (e.g., language and smoking).9,17 Thus, caution is needed in comparing the current results in Tables 2 and 3 with those reported by Link and colleagues.17 For instance, the crude vaccination rate among African Americans in the 2001–2002 season was 61.9% and 52.8% in Period 1 and 2 (including a different cohorts), respectively (Table 2). If the current study had used a cross-sectional study design, this rate was almost equivalent to the average of these two cohorts, 57.4% [⫽ (61.9%⫹52.8%)/2], which is close to 57.6% reported in Link’s study.17 Accounting for this point, over the fıve seasons since the 2000 –2001 season, two data sources (MCBS and BRFSS) indicated that African-American and Hispanic elderly populations consistently experienced 10% and 15% lower vaccination rates, respectively.17 The MCBS analysis in the current study showed a greater gap (⬎25%) among the Spanish-speaking Hispanic elderly population. In other words, the vaccination change across seasons can be masked or underestimated by the changes in panel subjects in a cross-sectional study. This problem was avoided by the cohort analysis. However, a cohort analysis can include only those benefıciaries who lived two consecutive influenza seasons and hence may be healthier than the general community-dwelling elderly Medicare benefıciaries. This may somewhat diminish the generalizability of the results.
Limitations Although it is possible that this analysis omitted factors that affected disparities in vaccination rates, it is unlikely that the current fındings were confounded strongly for two reasons. First, some omitted individual factors (e.g., number of clinic visits),7,15,18 and area characteristics (e.g., primary care physicians per adults aged ⱖ65 years)7 have shown only limited effects on racial/ethnic disparities. Second, it is unlikely that the differential change in these omitted factors among racial/ethnic groups over the course of 2 years would be suffıcient to explain the current fındings. www.ajpm-online.net
Yoo et al / Am J Prev Med 2011;40(1):1–10
Additionally, data limitations did not allow the present study to account for three potentially related factors. These include (1) national or local vaccination campaigns; (2) effects of prior vaccine shortage on patient, provider, and health system expectations and unanticipated behaviors (an expected shortage may increase demand22 or may conversely discourage patients from seeking vaccine because of their perception of permanent shortage11); and (3) provider discriminatory behavior. Insofar as campaigns were distributed and targeted randomly, they would bias results toward the null. It is unclear how experiences with a prior shortage might affect different racial/ethnic groups or providers. However, it is not likely that provider discrimination would change over two seasons. The current study used self-reported data, not claims data, to determine the vaccination status despite the potential recall bias. Self-reported data in adults aged ⱖ65 years generally have good sensitivity (94%–100%) but relatively low specifıcity (38%–70%).23,24 However, any recall bias that differs by race/ethnicity is unlikely to be affected by vaccine supply.
Policy Implications Because reduced influenza vaccine supply could potentially increase racial/ethnic disparities, this analysis indicates that efforts to reduce vaccination disparities should focus on ensuring an adequate and consistent vaccine supply to improve overall vaccination rates. Specifıc policy implications to address disparities in immunization that might mitigate the impact of decreases in supply might include stabilizing the vaccine supply; the creation of an adult program similar to the Vaccines-for-Children (VFC) program to sustain delivery of vaccines to safety-net providers with limited vaccine investment resources such as federally qualifıed health centers and practices that serve large proportions of African-American and Hispanic patients; and active provider and patient reminder/recall systems and targeted communication campaigns.25
Conclusion Improved vaccine supply appears to be associated with reduced racial/ethnic disparities in influenza vaccination rates while constrained supply is associated with worsening disparities. In general, the elimination of racial and ethnic disparities in adult immunization will require resources for immediate implementation of evidence-based strategies for improving influenza immunization in settings that serve African-American and Hispanic patients. This study is supported by NIH/National Institute of Allergy and Infectious Disease (1K25AI073915). January 2011
9
No fınancial disclosures were reported by the authors of this paper.
References 1. Offıce of Disease Prevention and Health Promotion. USDHHS. Healthy People 2010. 2007; Available from: www.cdc.gov/nchs/ about/otheract/hpdata2010/abouthp.htm. 2. Fiscella K. Commentary—anatomy of racial disparity in influenza vaccination. Health Serv Res 2005;40(2):539 – 49. 3. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare benefıciaries. Health Serv Res 2005;40(2):517–37. 4. CDC. Prevention and control of influenza: Recommendations of the Advisory Committee on Immunization Practices (ACIP). Morb Mortal Wkly Rep 2007;56(RR06):1–54. 5. Nichol KL. Influenza vaccination in the elderly: impact on hospitalization and mortality. Drugs Aging 2005;22(6):495–515. 6. Maciosek MV, Coffıeld AB, Edwards NM, Flottemesch TJ, Goodman MJ, Solberg LI. Priorities among effective clinical preventive services: results of a systematic review and analysis. Am J Prev Med 2006;31(1):52– 61. 7. O’Malley AS, Forrest CB. Immunization disparities in older Americans: determinants and future research needs. Am J Prev Med 2006;31(2):150 – 8. 8. Wallace GS. Influenza vaccine distribution 2006 – 07: National Vaccine Advisory Committee, CDC; 2007, February 5. 9. Centers for Medicare & Medicaid Services. Medicare current benefıciary survey: CY 2000 –2006. Baltimore MD; 2006. 10. CDC. Experiences with obtaining influenza vaccination among persons in priority groups during a vaccine shortage—U.S., October–November, 2004. Morb Mortal Wkly Rep 2004;53(49):1153–5. 11. Charatan F. Widespread flu in U.S. exposes shortage of vaccine. BMJ 2004;328(7430):8. 12. Nelson R. Influenza vaccine shortage hits the U.S. Lancet 2003; 362(9401):2075. 13. Phillips-Caesar E, Coady MH, Blaney S, et al. Predictors of influenza vaccination in an urban community during a national shortage. J Health Care Poor Underserved 2008;19(2):611–24. 14. Fiscella K, Franks P, Doescher MP, Saver BG. Disparities in health care by race, ethnicity, and language among the insured: fındings from a national sample. Med Care 2002;40(1):52–9. 15. Rangel MC, Shoenbach VJ, Weigle KA, Hogan VK, Strauss RP, Bangdiwala SI. Racial and ethnic disparities in influenza vaccination among elderly adults. J Gen Intern Med 2005;20(5):426 –31. 16. Egede LE, Zheng D. Racial/ethnic differences in influenza vaccination coverage in high-risk adults. Am J Public Health 2003;93(12):2074 – 8. 17. Link MW, Ahluwalia IB, Euler GL, Bridges CB, Chu SY, Wortley PM. Racial and ethnic disparities in influenza vaccination coverage among adults during the 2004 –2005 season. Am J Epidemiol 2006; 163(6):571– 8. 18. Marin MG, Johanson WG Jr, Salas-Lopez D. Influenza vaccination among minority populations in the U.S. Prev Med 2002; 34(2):235– 41. 19. Schneider EC, Cleary PD, Zaslavsky AM, Epstein AM. Racial disparity in influenza vaccination: does managed care narrow the gap between African Americans and whites? JAMA 2001;286(12): 1455– 60. 20. Yoo BK, Frick K. Determinants of influenza vaccination timing. Health Econ 2005;14(8):777–91.
10
Yoo et al / Am J Prev Med 2011;40(1):1–10
21. Nichol KL, Wuorenma J, von Sternberg T. Benefıts of influenza vaccination for low-, intermediate-, and high-risk senior citizens. Arch Intern Med 1998;158(16):1769 –76. 22. Ompad DC, Galea S, Blaney S, et al. Access to influenza vaccine in East Harlem and the Bronx during a national vaccine shortage. J Community Health 2007;32(3):195–202. 23. Zimmerman RK, Raymund M, Janosky JE, Nowalk MP, Fine MJ. Sensitivity and specifıcity of patient self-report of influenza
and pneumococcal polysaccharide vaccinations among elderly outpatients in diverse patient care strata. Vaccine 2003;21: 1486 –91. 24. MacDonald R, Baken L, Nelson A, Nichol KL. Validation of self-report of influenza and pneumococcal vaccination status in elderly outpatients. Am J Prev Med 1999;16:173–7. 25. Ompad DC, Galea S, Vlahov D. Distribution of influenza vaccine to high-risk groups. Epidemiol Rev 2006;28:54 –70.
Did you know? Two AJPM articles per issue offer CME credits. Go to www.ajpm-online.net/cme/home for more information.
www.ajpm-online.net