Regional factors associated with pneumococcal vaccination coverage among U.S. adults with underlying chronic or immunocompromising conditions

ABSTRACT

The Centers for Disease Control recommends pneumococcal vaccination for U.S. adults aged 19–64 years with chronic or immunocompromising conditions, however, vaccination coverage is low and regional variations in coverage are rarely studied. This study examined pneumococcal vaccination coverage at the metropolitan statistical area (MSAs) level and identified regional factors associated with pneumococcal vaccination using the combined IBM® Watson Health MarketScan® Commercial and Medicare Supplemental databases. Pneumococcal vaccination coverage, clinical and socioeconomic factors were calculated for each MSA. Ordinary least square and spatial regression models were used to examine factors associated with vaccination. Results indicated that the national pneumococcal vaccination coverage was 13.4% with a large variation across MSAs (0–34%). The spatial error model, model with the best fit, showed that proportions of the population who were ≥50 years of age, received an influenza vaccine, or had health maintenance organization health plans were positively associated with pneumococcal vaccination coverage. In summary, we found that national pneumococcal vaccination coverage was low and there was substantial variation across MSAs. Regional factors identified may help inform interventions to improve pneumococcal vaccination coverage across geographies.

Introduction

Invasive pneumococcal disease (IPD) causes significant morbidity and mortality among U.S. adults, and is associated with substantial health care costs.^1–4 In 2019, the overall incidence of IPD in the U.S. was 8 cases per 100,000 adults 19–64 years of age and 24 cases per 100,000 adults ≥65 years of age.¹ The annual estimated cost of pneumococcal disease is between $1.9 and $5.0 billion, with a majority of these costs occurring among unvaccinated adults.^3,4 Adults with certain underlying chronic or immunocompromising conditions have a higher incidence of IPD as well as increased hospitalization and mortality rates once infected, and incur greater average health care costs associated with IPD treatment.^2,5,6

In 1997, the U.S. Centers for Disease Control and Prevention’s (CDC) Advisory Committee on Immunization Practices (ACIP) began recommending pneumococcal vaccines for adults aged 19–64 years with chronic or immunocompromising conditions.⁷ In 2019 the ACIP guidelines for adults 19–64 years of age recommended pneumococcal 23-valent polysaccharide vaccine (PPSV23 [Pneumovax®23, manufactured by Merck & Co., Inc., Rahway, NJ, USA]) for those with chronic medical conditions and sequential administration of pneumococcal 13-valent conjugate vaccine (PCV13 [Prevnar 13®, manufactured by Wyeth Pharmaceuticals LLC, a subsidiary of Pfizer Inc. New York, NY, USA]) and PPSV23 for those with immunocompromising conditions.⁸ In 2021 the ACIP guidelines for adults 19–64 years of age were further updated following the U.S. Food and Drug Administration’s (FDA) approval of new pneumococcal vaccines, and recommended a 20-valent conjugate vaccine (PCV20, [Prevnar 20™, manufactured by Wyeth Pharmaceuticals LLC, a subsidiary of Pfizer Inc. New York, NY, USA]) alone or pneumococcal 15-valent conjugate vaccine (PCV15, [Vaxneuvance™, manufactured by Merck & Co., Inc. Rahway, NJ, USA]) followed by PPSV23 for adults with certain underlying medical conditions.⁹ Despite the ACIP recommendations, pneumococcal vaccination coverage rates among U.S. adults aged 19–64 years with chronic or immunocompromising conditions remains lower than the CDC Healthy People 2020 target of 60% (2030 target is not yet established), with recent estimates ranging from 10.5–29.2%.^10–14

Individual-level factors associated with a higher likelihood of pneumococcal vaccination among U.S. adults 19–64 years of age with chronic or immunocompromising conditions have been identified.^12–16 There are also regional variations in pneumococcal vaccination coverage even after accounting for individual-level factors; however, there are limited published data on factors associated with this variation.¹³ The objective of this study was to examine regional variation in pneumococcal vaccination coverage among adults with underlying conditions, and to identify regional factors associated with pneumococcal vaccination.

Methods

Study population and data source

The main data sources for this study were the combined IBM® Watson Health MarketScan® Commercial and Medicare Supplemental databases.¹⁷ These databases represent the health care service use (e.g., enrollment records, inpatient, outpatient, ancillary, and drug claims) of about 200 million employees, dependents, and retirees in the U.S. with primary or Medicare supplemental coverage through privately insured fee-for-service, point-of-service, or capitated health plans. Both databases have been formally analyzed to ensure that data are de-identified. Since all study procedures were compliant with the U.S. Health Insurance Portability and Accountability Act, Institutional Review Board approval or specific informed consent were not required for this study.

We defined the study population as adults 19–64 years of age (on the date of initial diagnosis) who were newly diagnosed with a chronic or immunocompromising condition in 2013 (Table A1). Chronic conditions included all forms of the following: diabetes mellitus; chronic heart, liver, or lung disease; alcohol or tobacco dependence. Immunocompromising conditions included all types of the following: asplenia, cancer, chronic renal disease, HIV/AIDS, and organ transplantation. Additional inclusion criteria were having at least 2 administrative claims for a condition of interest at least 30 days apart, continuous health plan enrollment for at least 2 years before and 1 year after the initial diagnosis date (enrollment gaps of ≤45 days were permitted), prescription drug benefit, and record of primary residential address (metropolitan statistical area [MSA] code, recorded at initial diagnosis date). The U.S. Census Bureau defines an MSA as follows: “An MSA consists of one or more counties that contain a city of 50,000 or more inhabitants, or contain a Census Bureau-defined urbanized area (UA) and have a total population of at least 100,000 (75,000 in New England).”¹⁸ MSA was the most granular geographical level available in the databases and provides a large number of patients with conditions of interest for comparison across areas. Exclusion criteria were any record of pneumococcal vaccination or of a medical condition of interest in the 2 years prior to the initial diagnosis date. The follow-up period of this study ended on the earliest of the following dates for each adult: date of pneumococcal vaccination, date of death, or December 31, 2019.

Socioeconomic characteristics of the 369 MSAs represented in the study population were from publicly available data sources. We used the most recent available data on race/ethnicity (2020), having at least a bachelor’s degree (2017), and median personal income (2017) from the U.S. Census Bureau, the U.S. Bureau of Economic Analysis, and the State Science & Technology Institute, respectively.^19–21 The publicly available data were linked to the claims data at the MSA level, using a similar method to other studies on associations between regional factors and healthcare services.^22,23

Study measures

The dependent variable was MSA-level proportion of adults who received a pneumococcal vaccination. Pneumococcal vaccination was defined as receiving PPSV23 among adults initially diagnosed with chronic conditions and either PCV13 or PPSV23 among adults initially diagnosed with immunocompromising conditions; PCV15 and PCV20 received FDA approval after the study period and were not included. Pneumococcal vaccination was identified by Current Procedural Terminology version 4 (CPT4) codes in medical claims and National Drug Codes (NDC) in pharmacy claims (Table A1). Pneumococcal vaccination coverage was evaluated after the initial diagnosis date through the end of the follow-up period.

Using the existing literature as a guide, we selected regional factors describing the sociodemographic and clinical compositions of MSAs as independent variables, using MSA-level aggregate data.^12–14 The independent variables included the proportion of adults ≥50 years of age at initial diagnosis, of each race/ethnicity, having at least a bachelor’s degree, being covered by a health maintenance organization (HMO) health plan, being initially diagnosed with diabetes mellitus (a highly prevalent chronic condition) or cancer (the most common immunocompromising condition), being initially diagnosed with an underlying condition by a primary care provider, and receiving an influenza vaccination during the follow-up period. The independent variables also included MSA-level average annual number of office visits and median personal income.

Statistical analysis

Percentages for categorical variables and means/standard deviations (SDs) for continuous variables were calculated for each MSA. An ordinary least square (OLS) regression model was initially used to examine the association between independent variables and vaccination coverage without considering spatial dependency across MSAs. A global Moran’s I test was then conducted to examine whether pneumococcal vaccination coverage values were clustered or randomly distributed across MSAs.²⁴ The test returns a value of I between −1 and 1, with positive values suggesting spatial clustering of similar values, negative values suggesting the clustering of dissimilar values, and zero indicating a random spatial pattern. Since the Moran’s I value was statistically significant for pneumococcal vaccination coverage (see Results), we then assessed spatial dependency using spatial lag and spatial error regression models, similar methods to those recently described by others.^22,25–27 Spatial lag models assume that the observed value in any given spatial unit (an MSA, in the current study) is directly influenced by the values in neighboring units due to diffusion of people and behavior between adjacent MSAs. Spatial error models do not assume this spillover but rather assume that the residuals are correlated in the neighborhood of the spatial units and consider the effect of omitted spatially correlated variables that are not present in the model but that may affect the estimation.^25–28 Model fit statistics were obtained for each model and compared to identify the model with the best fit for this study.²⁹ A p-value of <0.01 was deemed significant in tests of differences. Data analyses were conducted using R version 4.0. Visx software was used to visually represent pneumococcal vaccination coverage by MSA on a map of the U.S.

Results

Study population and metropolitan statistical area characteristics

A total of 255,330 adults met all inclusion criteria and were included in the analysis (Attrition of study adults was described in a previous study).¹³ Table 1 summarizes the mean (SD) characteristics of the MSAs represented in the dataset. The proportion of newly diagnosed adults receiving a pneumococcal vaccination during the follow-up period was 13.4% across MSAs. The mean age was 48.5 years, and 55.7% of adults were ≥50 years of age. The average proportions of race/ethnicity across MSAs were 66.6% White, 14.7% Hispanic, 10.3% Black, 2.8% Asian, 0.8% American Indian and Alaska Native, and 0.2% Native Hawaiian and other Pacific Islander. Median personal income was $45,481 and 27.4% of adults had at least a bachelor’s degree. The proportion of adults covered by an HMO health care plan was 9.3%, 41.9% of initial diagnoses were made by a primary care physician, 18.9% of adults were initially diagnosed with diabetes mellitus and 17.3% with cancer. Influenza vaccination coverage rate during the follow-up period was 18.8%. There was a mean of 6.0 annual office visits per adult.

Table 1.

Population characteristics at the level of metropolitan statistical area.

Characteristic	Mean (SD)
Pneumococcal vaccination coverage	13.4% (5.2)
Age (years)	48.5 (1.9)
Population ≥50 years of age	55.7% (7.8)
Race/ethnicitya
White	66.6% (17.5)
Hispanic	14.7% (17.0)
Black	10.3% (10.5)
Asian	2.8% (3.0)
American Indian and Alaska Native	0.8% (3.0)
Native Hawaiian and other Pacific Islander	0.2% (0.4)
Median personal income	$45,481 (9,418)
Education level of at least a bachelor’s degree	27.4% (8.0)
Health Maintenance Organization health plan	9.3% (13.7)
Initial diagnosis by primary care provider	41.9% (8.0)
Diagnosis of diabetes mellitus	18.9% (6.9)
Diagnosis of cancer	17.3% (5.4)
Influenza vaccination during follow-up period	18.8% (6.7)
Average annual number of office visits	6.0 (0.7)

SD, standard deviation.

^aThe sum of the percentages is less than 100% due to missing information on race/ethnicity.

Regional variation in pneumococcal vaccination coverage

There was substantial variation in pneumococcal vaccination coverage rate at the MSA level. The MSAs with the 10 lowest and 10 highest pneumococcal vaccination coverage values are listed in Table 2. The lowest vaccination coverage rate (0.0%) was observed in Ames, IA and Cheyenne, WY and the highest (34.0%) in Ann Arbor, MI. The U.S. map in Figure 1 shows the locations of MSAs in which pneumococcal vaccination coverage rate among the study population was higher or lower than the national average (a different version of this figure was previously published).¹³ As shown in the map, a majority of the MSAs located in the U.S. Department of Health and Human Services (HHS) region of Seattle (15 out of 25) had vaccination coverage rates higher than the national average. In contrast, a majority of the MSAs in the San Francisco (24/34) and Atlanta (56/92) regions had lower pneumococcal vaccination coverage than the national average. Chicago (35/70) had a 50:50 split and Dallas had roughly a 50:50 split (25/48) of MSAs with a higher or lower vaccination coverage than the national average. The full list of MSAs by vaccination coverage quartile is provided in Table A2.

Table 2.

Metropolitan statistical areas with the lowest and highest pneumococcal vaccination coverage among adults 19–64 years of age newly diagnosed with a chronic or immunocompromising condition.

Lowest pneumococcal vaccination coverage			Highest pneumococcal vaccination coverage
HHS Region	MSA	Percent vaccinated	HHS Region	MSA	Percent vaccinated
Region 7	Ames, IA	0.0	Region 5	Ann Arbor, MI	34.0
Region 8	Cheyenne, WY	0.0	Region 7	Jefferson City, MO	32.2
Region 3	California-Lexington Park, MD	3.1	Region 8	Bismarck, ND	30.8
Region 4	Sebring, FL	3.3	Region 10	Walla Walla, WA	30.8
Region 10	Longview, WA	3.3	Region 6	Killeen-Temple, TX	29.0
Region 10	Yakima, WA	3.6	Region 5	Jackson, MI	26.9
Region 8 & 10	Logan, UT-ID	4.0	Region 6	Waco, TX	26.8
Region 6	Enid, OK	4.8	Regions 7 & 8	Sioux City, IA-NE-SD	26.0
Region 4	Ocala, FL	5.2	Region 7	Iowa City, IA	25.6
Region 6	Laredo, TX	5.2	Region 8	Billings, MT	25.3

HHS, U.S. Department of Health and Human Services; MSA, metropolitan statistical area.

Region 1- Boston. Includes: CT, ME, MA, NH, RI, and VT. Region 2- New York. Includes: NJ, NY, PR, and VI. Region 3- Philadelphia. Includes: DE, D.C., MD, PA, VA, and WV. Region 4- Atlanta. Includes: AL, FL, GA, KY, MS, NC, SC, and TN. Region 5- Chicago. Includes: IL, IN, MI, MN, OH, and WI. Region 6- Dallas. Includes: AR, LA, NM, OK, and TX. Region 7- Kansas City. Includes: IO, KA, MO, and NE. Region 8- Denver. Includes: CO, MT, ND, SD, UT, WY. Region 9- San Francisco. Includes: AZ, CA, HI, NV, AS, CNMI, FSM, GU, MH, and PLW. Region 10- Seattle. Includes: AK, ID, OR, and WA.

Figure 1.

Map of the United States representing metropolitan statistical areas with a higher (green) or lower (red) proportion of pneumococcal vaccination among adults 19–64 years of age newly diagnosed with a chronic condition or immunocompromising condition, compared with the national average of 13.4%.

Regional factors associated with pneumococcal vaccination coverage

The global Moran’s I value was 0.22 (p < .01), indicating a statistically significant spatial clustering of similar pneumococcal vaccination coverage in neighboring MSAs and the need for spatial models. Table 3 shows results from the OLS, spatial lag, and spatial error models. The spatial error model had the lowest Akaike information criterion (AIC) value and was selected as the model with the best fit for the study. The results of this model indicated that MSAs with higher proportions of adults who were ≥50 years of age, had an HMO health plan, or had received an influenza vaccine during the follow-up period had higher pneumococcal vaccination coverage. A 10% increase in the proportion of adults ≥50 years of age was associated with a 1.2% increase in pneumococcal vaccination coverage. Similarly, 10% increases in the proportion of adults receiving an influenza vaccination or being covered under an HMO plan were associated with a 1.1% or 0.8% increase in pneumococcal vaccination coverage, respectively.

Table 3.

Regional factors associated with pneumococcal vaccination among adults 19–64 years of age newly diagnosed with a chronic or immunocompromising condition.

Variable	Ordinary least square model			Spatial lag model		Spatial error model
	Beta (SE)	95% CI	p-value	Beta (SE)	p-value	Beta (SE)	p-value
Population ≥50 years of age	0.12 (0.04)	0.05, 0.19	<.01	0.12 (0.04)	<.01	0.12 (0.04)	<0.01
Race/ethnicitya
Hispanic	−0.30 (0.19)	−0.66, −0.06	<.01	−0.31 (0.18)	.09	−0.34 (0.19)	0.07
White	−0.24 (0.19)	−0.62, 0.10	.2	−0.26 (0.19)	.2	−0.29 (0.20)	0.13
Black	−0.26 (0.19)	−0.63, 0.10	.2	−0.28 (0.18)	.1	−0.30 (0.19)	0.11
Asian	−0.38 (0.24)	−0.86, 0.09	.11	−0.40 (0.23)	.08	−0.44 (0.24)	0.07
American Indian and Alaska Native	−0.18 (0.23)	−0.64, 0.27	.4	−0.18 (0.23)	.4	−0.20 (0.23)	0.4
Median personal income	0.001 (<0.001)	0.001, 0.001	.8	0.001 (<0.001)	.7	0.001 (<0.001)	0.6
Influenza vaccination during follow-up period	0.13 (0.04)	0.05, 0.21	<.01	0.12 (0.04)	<.01	0.11 (0.04)	<0.01
HMO health plan	0.09 (0.02)	0.05, 0.13	<.01	0.08 (0.02)	<.01	0.08 (0.02)	<0.01
Initial diagnosis by PCP	0.02 (0.04)	−0.05, 0.09	.5	0.03 (0.03)	.5	0.03 (0.03)	0.4
Education level of at least a bachelor’s degree	0.001 (<0.001)	0.001, 0.001	.04	0.001 (<0.001)	.03	0.001 (<0.001)	0.05
Diagnosis of diabetes mellitus	0.06 (0.05)	−0.04, 0.16	.3	0.06 (0.05)	.2	0.04 (0.05)	0.4
Diagnosis of cancer	−0.03 (0.06)	−0.14, 0.09	.6	−0.02 (0.06)	.7	0.001 (0.06)	0.99
Average annual number of office visits	−0.01 (0.004)	−0.02, −0.001	.03	−0.01 (0.004)	.04	−0.01 (0.004)	0.2
Model fit statistics
Rho				0.11
Lambda						0.24
Akaike information criterion	−1,147			−1,153.4		−1,164
Adjusted R2	0.17			0.19		0.22

CI, confidence interval; HMO, Health Maintenance Organization; PCP, primary care physician; SE, standard error.

^aThe reference group is Native Hawaiian and other Pacific Islander.

Discussion

This study assessed pneumococcal vaccination coverage at the MSA level among U.S. adults 19–64 years of age newly diagnosed with chronic or immunocompromising conditions, and identified factors associated with higher MSA-level vaccination coverage. Despite ongoing efforts to promote pneumococcal vaccination among this population, the mean vaccination rate at the MSA level was only 13.4%. While there was substantial variation in MSA-level pneumococcal vaccination coverage (0–34%), coverage was well below the CDC Healthy People target (60%) across all MSAs. Many adults do not know that they need vaccination and most healthcare providers do not recommend pneumococcal vaccination to adults with underlying conditions.³⁰ These results highlight the need to educate adults and providers about pneumococcal vaccination.

Vaccination coverage showed significant spatial dependence. For example, most MSAs in the Seattle HHS region had higher-than-average vaccination coverage, while most MSAs in the San Francisco and Atlanta HHS regions had lower-than-average vaccination coverage. In contrast, the Chicago and Dallas regions had a split of MSAs with higher or lower than average vaccination coverage. Our findings are generally consistent with a CDC report based on survey data on pneumococcal vaccination coverage at the county level for adults <65 years of age with select chronic or immunocompromising conditions.¹⁰ Specifically, MSAs with high or low vaccination coverage in our study were in the same areas as counties that the CDC reported to have high (e.g., Snohomish county, WA; Columbia county, OR; Washtenaw County, MI; Davie County, NC) or low (e.g., Tulare, CA; Highlands, FL; Webb, TX; Dorchester County, SC) pneumococcal vaccination coverage.

Research on regional factors associated with pneumococcal vaccination in the U.S. is limited. Previous studies have found substantial inter-state and inter-region variations in vaccination coverage for pneumococcal and other adult vaccines, after accounting for patient-level characteristics.^13,31 The current study went one step further to examine regional factors associated with vaccination. To our knowledge, this study has been the first to examine regional factors associated with pneumococcal vaccination coverage at the MSA level among adults with underlying conditions.

The spatial error model identified 3 MSA-level factors associated with higher pneumococcal vaccination coverage: proportion of adults ≥50 years of age, covered by an HMO health plan, or who had received an influenza vaccination. The association of pneumococcal vaccination with older age may reflect stronger vaccine promotion policies and better physician awareness in areas with higher proportions of adults ≥50 years of age.^32,33 Similarly, the association with influenza vaccination may reflect regional differences in how adult vaccines are funded, promoted, and delivered, for example by recommending administration of a pneumococcal vaccination during the same visit with influenza vaccination.³⁴ The association with HMO health plan coverage may be related to actual or perceived differences between plan types in the out-of-pocket costs of adult vaccinations.^35,36

Our study did not find an association between MSA-level pneumococcal vaccination coverage and 3 other regional socioeconomic factors (race/ethnicity, education, and income), some of which have been reported to associate with pneumococcal vaccination at the individual level.^12,15,37 Future research may examine associations with these variables at a more granular regional level, should healthcare data be available at this level. Future studies may also examine associations between pneumococcal vaccination and other factors such as local vaccine policies, programs, and funding, to the extent such data are available and linkable to healthcare data. Qualitative studies interviewing key stakeholders may also help to understand what common facilitators and barriers are among MSAs with higher or lower pneumococcal vaccination coverage than the national average.

Our findings may be used in multiple ways. First, with this data, MSAs with low pneumococcal vaccination coverage may recognize low vaccination coverage and identify appropriate interventions to increase pneumococcal vaccination. Second, the large variation we observed in pneumococcal vaccination requires an examination of patient, provider, and system barriers and enablers to vaccination in certain MSAs. Lastly, by examining the number of patients and costs associated with preventable pneumococcal disease in MSAs, healthcare professionals and policy makers may justify targeted reallocation of resources to increase pneumococcal vaccination coverage and lower the burden of pneumococcal disease in selected MSAs. These findings may also assist stakeholders in designing interventions. For instance, the finding that the percentage of adults covered under HMOs is associated with higher pneumococcal vaccination coverage may suggest the need to educate patients and/or providers that ACIP-recommended vaccines are covered without out-of-pocket costs by different types of commercial insurance plans.

Study limitations are noted. The observational design allowed us to identify associations, but not causality. The use of claims data is subject to inaccuracies or omissions of conditions.³⁸ We mitigated this risk by requiring adults to have at least 2 administrative claims for the same underlying condition. Further, the findings may not be generalizable to adults covered with noncommercial or Medicare plans. In addition, we did not examine the sequential administration of PCV13 and PPSV23 among adults with immunocompromising conditions, given the low rate of completed vaccination sequence (2.2%) shown in the literature and the focus of our study being factors associated with whether patients receive a pneumococcal vaccination.³⁹ Future studies may examine factors associated with completion of pneumococcal vaccination sequence among adults with immunocompromising conditions.

Conclusions

Pneumococcal vaccination coverage among U.S. adults with chronic or immunocompromising conditions varied substantially between MSAs and was well below the CDC target for this population in all MSAs. The MSAs with higher proportions of adults who were ≥50 years of age, had an HMO health plan, or received an influenza vaccine had higher pneumococcal vaccination coverage. The findings can inform interventions and resource allocation to improve pneumococcal vaccination coverage among all MSAs, especially those with low coverage.

Funding Statement

The study was funded by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.

Disclosure statement

Junqing Liu, Linda Shoener Dunham, and Kelly D. Johnson are employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA and may own stock in Merck & Co., Inc., Rahway, NJ, USA.

Author roles

Conception and design: Junqing Liu, Kelly D. Johnson, Linda Shoener Dunham

Analysis and interpretation of the data: Junqing Liu, Kelly D. Johnson, Linda Shoener Dunham

Drafting of the paper: Junqing Liu, Kelly D. Johnson, Linda Shoener Dunham

Revising the paper critically for intellectual content: Junqing Liu, Kelly D. Johnson, Linda Shoener Dunham

Final approval of the version to be published: Junqing Liu, Kelly D. Johnson, Linda Shoener Dunham

All authors agree to be accountable for all aspects of the work.