Association of Pneumococcal Conjugate Vaccine Use With Hospitalized Pneumonia in Medicare Beneficiaries 65 Years or Older With and Without Medical Conditions, 2014 to 2017

Key Points

Question  Is 13-valent pneumococcal conjugate vaccine (PCV13) use associated with reduced pneumonia hospitalization in US adults 65 years or older who have a high prevalence of underlying medical conditions?

Findings  In this cohort study of more than 24 million Medicare beneficiaries 65 years or older with and without underlying medical conditions across 50 US states and the District of Columbia, beneficiaries who received PCV13 had a 6.7% lower risk of pneumonia hospitalization overall, including 5.8% to 7.5% lower risk in adults with underlying medical conditions, compared with beneficiaries who did not receive any pneumococcal vaccines.

Meaning  The study results suggest that new PCV13 use may be associated with reduced risk of pneumonia hospitalizations among US adults 65 years or older, including among those with underlying medical conditions.

Importance  The association of 13-valent pneumococcal conjugate vaccine (PCV13) use with pneumonia hospitalization in older adults, especially those with underlying medical conditions, is not well described.

Objective  To evaluate the association of PCV13 use with pneumonia, non–health care–associated (non-HA) pneumonia, and lobar pneumonia (LP) hospitalization among US Medicare beneficiaries 65 years or older.

Design, Setting, and Participants  This cohort study with time-varying exposure assignment analyzed claims data from US Medicare beneficiaries 65 years or older enrolled in Parts A/B with a residence in the 50 US states or the District of Columbia by September 1, 2014. New Medicare Parts A/B beneficiaries within 6 months after their 65th birthday were continuously included in the cohort after September 1, 2014, and followed through December 31, 2017. Participants were censored if they died, changed enrollment status, or developed a study outcome. Most of the analyses were conducted from 2018 to 2019, and additional analyses were performed from 2021 to 2022.

Exposures  Use of PCV13 vaccination 14 days or more before pneumonia hospitalization.

Main Outcomes and Measures  Discrete-time survival models were used to estimate the incidence rate ratio (IRR) and number of pneumonia hospitalizations averted through PCV13 use. The adjusted IRR for the association of PCV13 vaccination with pneumonia hospitalization was used to estimate vaccine effectiveness (VE).

Results  At the end of follow-up (December 2017), 24 121 625 beneficiaries (13 593 975 women [56.4%]; 418 005 [1.7%] Asian, 1 750 807 [4.8%] Black, 338 044 [1.4%] Hispanic, 111 508 [0.5%] Native American, and 20 700 948 [85.8%] White individuals) were in the cohort; 4 936 185 (20.5%) had received PCV13 only, and 10 646 220 (79.5%) had not received any pneumococcal vaccines. More than half of the beneficiaries in the cohort were younger than 75 years, White, and had either immunocompromising or chronic medical conditions. Coverage with PCV13 increased from 0.8% (September 2014) to 41.5% (December 2017). The VE for PCV13 was estimated at 6.7% (95% CI, 5.9%-7.5%) for pneumonia, 4.7% (95% CI, 3.9%-5.6%) for non-HA pneumonia, and 5.8% (95% CI, 2.6%-8.9%) for LP. From September 2014 through December 2017, an estimated 35 127 pneumonia (95% CI, 33 011-37 270), 24 643 non-HA pneumonia (95% CI, 22 761-26 552), and 1294 LP (95% CI, 797–1819) hospitalizations were averted through PCV13 use.

Conclusions and Relevance  The study results suggest that PCV13 use was associated with reduced pneumonia hospitalization among Medicare beneficiaries 65 years or older, many of whom had underlying medical conditions. Increased PCV13 coverage and use of recently approved higher-valent pneumococcal conjugate vaccines may avert additional pneumonia hospitalizations in adults.

Introducing the pneumococcal conjugate vaccine for use among children has been associated with significantly reduced incidence rates of pneumococcal disease, not only among children who were directly targeted for vaccination, but also among older children and adults who were not targeted for vaccination, by preventing the transmission of vaccine-type pneumococci.¹ Three years after the 13-valent pneumococcal conjugate vaccine (PCV13) was introduced among US children in 2013, the incidence of invasive pneumococcal disease, defined as identification of pneumococcus from a normally sterile site, declined by 58% to 72% among US adults.² In 2014, the Advisory Committee on Immunization Practices (ACIP) recommended routine use of PCV13 in series with a dose of 23-valent pneumococcal polysaccharide vaccine (PPSV23) for all US adults 65 years or older.³ This recommendation was supported by findings from a large randomized clinical trial in adults 65 years or older in the Netherlands (CAPITA trial) that reported that PCV13 provides protection not only against invasive pneumococcal disease (75% vaccine effectiveness [VE]), but also against vaccine-type noninvasive pneumococcal pneumonia, defined as pneumonia with no identification of pneumococcus from a normally sterile site (45% VE).⁴ In adults, pneumococcal pneumonia contributes the largest proportion of pneumococcal disease burden. The reported efficacy against the first episode of all-cause community-acquired pneumonia from this trial was 5.1% (95% CI, −5.1 to 14.2). However, continued PCV13 use among children was expected to be associated with further reduced remaining burden of adult pneumococcal disease that was associated with serotypes contained in PCV13 and decrease the effectiveness of PCV13 in preventing pneumococcal disease in adults. Routine PPSV23 use for US adults 65 years or older had been recommended since 1984, but data on PPSV23 effectiveness against noninvasive pneumococcal pneumonia were considered to be inconsistent.³

The CAPITA trial did not enroll adults with immunocompromising conditions (ICs), and data on PCV13 VE in this group remain limited, even though the proportion of adults with ICs increases with age.^4,5 Additionally, determining the true burden of vaccine-type noninvasive pneumonia is challenging because of a lack of a standardized laboratory method to identify pneumococcal serotypes in noninvasive disease.⁶ In US adults 65 years or older, the estimated incidence of hospitalized community-acquired pneumonia (CAP) has ranged from 847 to 3500 per 100 000 persons.⁷ One study reported that approximately 5% of hospitalized all-cause CAP was due to PCV13-type pneumococci based on a urinary antigen detection assay used in research settings; however, this was still considered an underestimate of the true burden of PCV13-type disease.⁸ In October 2021, ACIP recommended the use of 15-valent pneumococcal conjugate vaccine (PCV15) followed by a dose of PPSV23, or the 20-valent pneumococcal conjugate vaccine (PCV20) alone for all adults 65 years or older and adults aged 19 to 64 years with certain underlying medical conditions and other risk factors.⁹ Use of PCV15 and PCV20 was licensed based on safety and immunogenicity data alone; understanding PCV13 VE against all-cause pneumonia could help inform the expected public health effect from use of PCV15 or PCV20 in adults. We evaluated the association between PCV13 and pneumonia hospitalization and estimated the number of pneumonia cases averted from PCV13 use among US Medicare Part A/B beneficiaries 65 years or older with and without underlying medical conditions during the first 4 years of routine PCV13 use in this population.

This cohort study with time-varying exposure assignment was reviewed by the US Centers for Disease Control and Prevention (CDC) and was conducted consistent with applicable federal law and CDC policy; therefore, institutional review board approval and informed consent were waived. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.¹⁰

Primary data sources were the Medicare claims and enrollment databases. Information on demographic characteristics and deaths were derived from the enrollment database. Information on pneumococcal and influenza vaccinations, health covariates, and outcomes was derived from Medicare Part A (inpatient), Part B (outpatient and community settings), and Part D (prescription drug) claims. Medicare beneficiaries who met the following 3 criteria on September 1, 2014, were included in the initial cohort: (1) 65 years or older; (2) residence in the 50 US states or the District of Columbia; and (3) enrollment in Medicare Parts A and B (without enrollment in Medicare Part C). After September 1, 2014, new beneficiaries who turned age 65 years and enrolled in Medicare Parts A and B were eligible to enter the cohort, as long as enrollment in Part B started within 6 months after their 65th birthday. Beneficiaries were followed until the end of study period (December 31, 2017) or until they were censored from the cohort due to (1) death; (2) moving out of the US, or when residence became unknown; (3) changing enrollment status; or (4) experiencing any of the study outcomes. Once censored, a beneficiary could not re-enter the cohort during a later month. Each beneficiary contributed 1 to a maximum of 40 continuous person-months to the data set (median, 17-21 person-months, depending on vaccination status and outcome). The exposure status and beneficiary characteristics (eg, age, sex, self-reported race and ethnicity, underlying health status, and state of residence) were updated monthly.

Outcomes of interest were identified based on inpatient claims. As the primary pneumonia outcome, we used the case definition created by Griffin et al,¹¹ in which hospitalized pneumonia was defined as a primary discharge diagnosis of pneumonia or a primary discharge diagnosis of meningitis, septicemia, empyema, or acute respiratory failure with a pneumonia diagnosis in any secondary position. We also evaluated 2 secondary outcomes that may be more specific for pneumococcal pneumonia. First, to exclude health care–associated (HA) pneumonia, we evaluated non-HA pneumonia, which was defined as a subset of hospitalized pneumonia cases in patients not admitted to a hospital or skilled nursing facility fewer than 30 days before pneumonia hospitalization. Second, we defined lobar pneumonia (LP) as an inpatient hospital claim with lobar or pneumococcal pneumonia in any discharge diagnosis position. International Classification of Diseases, Ninth Revision, Clinical Modification and ICD-10-CM codes used to identify these outcomes are included in eTable 1 in the Supplement.

We categorized beneficiaries into 4 mutually exclusive risk groups defined based on ACIP recommendations¹²: (1) those with ICs, (2) those with chronic medical conditions (CMCs), (3) those with ICs and CMCs (IC+CMC), and (4) low-risk individuals. Patients included in the IC group had asplenia, chronic kidney failure, generalized cancer, HIV infection, Hodgkin disease, iatrogenic immunosuppression, immunodeficiency, leukemia, lymphoma, multiple myeloma, nephrotic syndrome, sickle-cell anemia, or solid organ transplant, but no CMC. Patients included in the CMC group had alcoholism, chronic heart/liver/lung disease, cigarette smoking, or diabetes, but no IC. Patients included in the IC+CMC group had at least 1 IC and 1 CMC. Lastly, adults with no ICs or CMCs were classified as low risk. Influenza vaccination status was assigned using all available claims data starting August 1, 2014. Beneficiaries were considered to be vaccinated against influenza if (1) they received an influenza vaccine during August to April, when the vaccine was available for that flu season, and (2) the vaccine was administered 14 days or more before the first day of hospitalization for those who developed an outcome of interest or 14 days or more before the first day of the month for those without an outcome of interest. All covariates and their definitions are listed in eTable 2 in the Supplement.

Each month of the study follow-up, the cohort population was assigned 1 of the 4 mutually exclusive pneumococcal vaccination categories (PCV13 only, PPSV23 only, both vaccines, no vaccine) using claims data starting January 1, 2008, when vaccination status of the beneficiaries was captured consistently in the database. Vaccine doses were considered to be valid if they were administered 14 days or more before hospitalization for 1 of the pneumonia outcomes, or if they were administered 14 days or more before the first day of the month if the beneficiary did not develop an outcome that month (eMethods in the Supplement).

We performed a descriptive analysis comparing those who received PCV13 only and those who did not receive any pneumococcal vaccines. We estimated the association between pneumococcal vaccination status and each outcome of interest (hospitalized pneumonia, non-HA pneumonia, LP) using a discrete-time logistic regression model fit with generalized estimating equations (GEEs). The model assessed each outcome for each observation month and was adjusted for all covariates identified as confounders based on potential for association with the outcome and exposure of interest (including Charlson Comorbidity score, as previously described^13,14) (eTable 2 in the Supplement). We applied an exchangeable structure in the GEE model to account for within-participant correlation across months.

The VE was calculated as 1 minus the adjusted incidence risk ratio (IRR). The IRR was estimated as follows: (1) we calculated the marginal predictive probability (MPP) of pneumonia hospitalization using all eligible participants by plugging in the estimated parameters obtained in the GEE model described previously, assuming everyone had PCV13 only; (2) we recalculated the MPP but changed the vaccination status to no vaccine for all participants while keeping other covariate information intact; and (3) we calculated IRR as the ratio between MPPs calculated in steps 1 and 2. Steps 1 to 3 were repeated assuming everyone received PCV13 and PPSV23 in step 1, and in step 2, adults were assumed to have received PPSV23 only.

To estimate the number of hospitalizations averted from PCV13 vaccination, we first estimated the expected number of pneumonia hospitalizations in the absence of PCV13 (with or without PPSV23) using the following formulas:

• (Observed pneumonia hospitalizations in those who received PCV13 only) divided by IRR_{(PCV13 only)/(no vaccine)}

• (Observed pneumonia hospitalizations in those who received PCV13 and PPSV23) divided by IRR_{(PCV13 and PPSV23)/(PPSV23 only)}

• Subsequently, the number of hospitalizations averted was equal to the expected_{(number of hospitalizations)} minus the observed_{(number of hospitalizations)}

We applied this procedure within strata defined by age group and risk group and then summed the stratum-specific number of hospitalizations averted. We estimated IRRs as previously described, except that to calculate the number of hospitalizations averted, we used a subset of the entire cohort that had a specific vaccination status (PCV13 only, PCV13 and PPSV23) for the estimation steps previously described in steps 1 and 2.

We fit all models using Stata, version MP 15 (StataCorp). We used a customized Stata plug-in¹⁵ to estimate the counterfactual marginal probabilities and risk ratios. The variances of the marginal probabilities and risk ratios were calculated via the delta method.

In September 2014, 26.6 million beneficiaries were eligible for inclusion in the study cohort. At the end of the follow-up period in December 2017, 24.1 million beneficiaries remained in the cohort, representing approximately 49% of the US population 65 years or older. More than half were younger than 75 years, White, and had either ICs or CMCs (Table 1; eTable 3 in the Supplement). The proportion of those who received PCV13 with or without PPSV23 increased from 0.8% in September 2014 to 41.6% in December 2017. The median and mean months of follow-up by vaccination status are provided in eTable 4 in the Supplement. Among those who remained in the cohort in December 2017, compared with beneficiaries who did not receive any pneumococcal vaccination, those who received PCV13 only were more likely to be older (75 years or older, 51.3% vs 37.8%), have comorbidities (Charlson Comorbidity Index score ≥3, 29.4% vs 21.8%), have a higher frequency of outpatient visits (≥5 outpatient visits during the preceding year, 33.7% vs 22.8%) and hospital admissions (≥2 inpatient visits, 4.5% vs 3.5%), and have received an influenza vaccine (65.1% vs 23.2%) (Table 1). Among those who received PCV13 only, crude incidence rates per 100 000 person-months for hospitalized pneumonia, non-HA pneumonia, and LP were 131, 106, and 7, respectively; among those who did not receive any pneumococcal vaccination, the incidence rates were 137, 107, and 5, respectively (Figure). Hospitalized pneumonia, non-HA pneumonia, and LP incidence were higher among older age groups. Similarly, the incidence was higher among those with CMCs and ICs compared with those who did not have CMCs, ICs, or neither (Figure).

The VE for PCV13 only against hospitalized pneumonia was 6.7% (95% CI, 5.9%-7.5%) (Figure). The VE was lower among older age groups, although confidence intervals overlapped: 7.4% (95% CI, 6.2%-8.5%) in adults aged 65 to 74 years compared with 5.7% (95% CI, 4.6%-6.8%) in adults aged 85 years or older. The VE was higher for those at low risk compared with those with either ICs or CMCs: 15.1% (95% CI, 12.2%-18.1%) in adults at low risk compared with 5.8% (95% CI, 5.0%-6.7%) in adults with ICs and CMCs. Overall, PCV13 VE against non-HA pneumonia (4.7%; 95% CI, 3.9%-5.6%) and LP (5.8%; 95% CI, 2.6%-8.9%) were comparable with VE against hospitalized pneumonia; however, estimated PCV13 VE against LP in those with IC only (18.1%; 95% CI, 5.0%-31.2%) and in low-risk groups (34.6%; 95% CI, 24.6%-44.5%) was higher than estimated VE against hospitalized pneumonia and non-HA pneumonia in these subgroups. The VE comparing those who received PCV13 and PPSV23 vs those who received only PPSV23 against the outcomes of hospitalized pneumonia, non-HA pneumonia, and LP were generally smaller (3.8%, 1.8%, and 4.2%, respectively) (eTable 5 in the Supplement). Person-months of follow-up were smaller for this comparison compared with those who received PCV13 only and those who did not receive any pneumococcal vaccination, in which was associated with wider confidence intervals in some stratified analyses.

During the study period, 300 531 pneumonia, 241 279 non-HA pneumonia, and 16 810 LP hospitalizations were identified among beneficiaries who received any PCV13 vaccination (ie, with or without PPSV23), and PCV13 is estimated to have averted 35 127 pneumonia (95% CI, 33 011-37 270), 24 643 non-HA pneumonia (95% CI, 22 761-26 552), and 1294 LP (95% CI, 797–1819) hospitalizations overall (Table 2). The largest number of cases averted were among adults with ICs and CMCs, who had the highest pneumonia incidence (Table 2; Figure).

In this cohort study of Medicare beneficiaries 65 years or older, we estimated a PCV13 VE of 6.7% against pneumonia in hospitalized patients. Vaccine effectiveness was lower among older age groups compared with younger age groups, as well as in adults with underlying medical conditions compared with adults without underlying medical conditions. The overall estimate was similar to the PCV13 efficacy of 5.1% (95% CI, −5.1% to 14.2%) against a first episode of community-acquired pneumonia that was reported in the CAPITA trial.⁴ Adults with ICs were excluded from the CAPITA trial on enrollment, and to our knowledge, there are limited studies on PCV13 effectiveness in adults with ICs.¹⁶ The results of the present study suggested that PCV13 was also effective against hospitalized pneumonia in adults with ICs.

Previous observational studies that used administrative data to estimate PCV13 VE against all-cause pneumonia reported variable results; this may be due to differences in the populations that were included. A population-based cohort study of more than 2 million adults 50 years or older in Spain did not demonstrate a protective effect of PCV13 against pneumococcal pneumonia or all-cause pneumonia, although the characteristics of those who received PCV13 were significantly different compared with those who did not receive PCV13, including a significantly higher proportion of adults with ICs among the PCV13 recipients compared with nonrecipients (42.1% vs 8.6%).¹⁷ Effectiveness of PCV13 against all-cause pneumonia was demonstrated in other studies that were performed more recently. An insurance-based retrospective cohort study of adults 60 years or older in Germany showed lower 3-year cumulative incidence of all-cause pneumonia among those who received PCV13, with a VE of 12% (statistical significance was not reported).¹⁸ A cohort study of more than 40 000 members of Kaiser Permanente Southern California 65 years or older reported a PCV13 VE of 8.8% (95% CI, −0.2% to 17.0%) against a first episode of pneumonia.¹⁹ Another cohort study of more than 192 000 adults who were members of Kaiser Permanente Northern California reported a PCV13 VE of 10.0% (95% CI, 2.4%-17.0%) against hospitalized pneumonia.²⁰ These studies included a lower proportion of adults with ICs (2%-7%)^18,19 or comorbidities²⁰ compared with the present study population, in which approximately 40% had ICs. According to the National Health Interview Survey, more than 60% of adults 65 years or older are estimated to have multiple chronic conditions (arthritis, cancer, chronic obstructive pulmonary disease, coronary heart disease, current asthma, diabetes, hepatitis, hypertension, stroke, and weak or failing kidneys).²¹ Therefore, our study cohort, which captured approximately half of the US population 65 years or older from all 50 US states and DC, is more likely to be representative of US adults with underlying conditions compared with those in previously published reports.

The VE estimates against all-cause pneumonia allow for estimation of the proportion of vaccine-preventable disease burden among those with all-cause pneumonia.²² A prior study estimated that PCV13-type pneumococcal pneumonia comprised 5% of all-cause hospitalized CAP among adults 65 years or older, but the authors noted that this may still be an underestimate of the true burden of PCV13-type pneumococcal pneumonia.⁸ Applying the PCV13 VE against hospitalized vaccine-type pneumonia of 45%, as reported in the CAPITA trial,⁴ and assuming that PCV13-type pneumonia comprises 5% of hospitalized CAP cases, the estimated PCV13 VE against all-cause hospitalized CAP would be 2.3% (45% × 5%), which is lower than reported (5%-12%),^4,17-19 including estimates from the present study. These data suggest that there may be more pneumococcal pneumonia cases that are preventable by PCV13 than currently estimated.

This study is subject to several limitations. First, given the transition from International Classification of Diseases, Ninth Revision (ICD-9) to ICD-10 in October 2015, changes in coding practices might have been associated with the sensitivity or specificity of the outcome definitions during the study follow-up. A study using an ICD-10-CM algorithm derived from a validated ICD-9-CM algorithm showed that pneumonia cases identified in the ICD-9-CM era might be undercounted using ICD-10-CM, especially in adults with pneumonia associated with chronic obstructive pulmonary disease exacerbation.²³ However, the proportion of adults with CMC was similar among those who received PCV13 and those who did not receive any pneumococcal vaccine, so undercounting of pneumonia cases would have equally affected both groups. Additionally, we adjusted for year in the model and found similar estimates during the transition period. Second, we did not report the effectiveness of PPSV23 against pneumonia because of concerns of misclassification of PPSV23 vaccination status or potential for bias. An internal validation study comparing the vaccination status of beneficiaries who were captured by the CDC Active Bacterial Core surveillance and the Medicare claims data showed that PPSV23 status is more likely than PCV13 status to be missed in the Medicare claims data (eTable 6 in the Supplement). Coverage for PCV13 was very low among adults 65 years or older before September 2014,^24,25 whereas PPSV23 has been available longer, with stable vaccine coverage. Therefore, there might have been greater opportunities to miss PPSV23 doses vs PCV13 doses, including PPSV23 doses given before age 65 years.²⁵ Misclassification of PCV13 vaccination status, including missing doses given outside of Medicare Part A or B settings, remains a possibility, which could bias the VE estimates toward the null. Third, as with any observational study, and especially those using administrative data, residual confounding may be present due to unmeasured factors, despite adjustment made for several potential confounders.

The results of this cohort study of adults 65 years or older from across the US suggest that PCV13 was effective against hospitalized pneumonia during the period when indirect effects from PCV13 use in children have been associated with reduced PCV13-type disease burden in adults. Effectiveness was also observed among adults with ICs, for whom limited data on PCV13 effectiveness exist. Use of PCV15 and PCV20 in adults has the potential to further reduce pneumococcal disease incidence in US adults, given the broader serotype coverage and expanded eligibility for PCV receipt in adults younger than 65 years.⁹

Accepted for Publication: September 29, 2022.

Published Online: December 5, 2022. doi:10.1001/jamainternmed.2022.5472

Corresponding Author: Miwako Kobayashi, MD, MPH, 1600 Clifton Rd NE, MS H24-6, Atlanta, GA 30329 (mkobayashi@cdc.gov).

Correction: This article was corrected on January 23, 2023, to fix errors in the Key Points and Table 2.

Author Contributions: Drs Wu and Wang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Pilishvili and Lessa contributed equally to this article.

Concept and design: Spiller, Wu, Wang, Kelman, Whitney, Pilishvili, Lessa.

Acquisition, analysis, or interpretation of data: Kobayashi, Spiller, Wu, Wang, Chillarige, MaCurdy, Kelman, Deng, Shang, Whitney, Pilishvili, Lessa.

Drafting of the manuscript: Kobayashi, Wang, Pilishvili, Lessa.

Critical revision of the manuscript for important intellectual content: Kobayashi, Spiller, Wu, Chillarige, MaCurdy, Kelman, Deng, Shang, Whitney, Pilishvili, Lessa.

Statistical analysis: Spiller, Wu, Wang, Chillarige, MaCurdy, Deng, Shang.

Obtained funding: Kelman, Whitney, Pilishvili, Lessa.

Administrative, technical, or material support: Kelman.

Supervision: Chillarige, Kelman, Whitney, Pilishvili, Lessa.

Conflict of Interest Disclosures: Dr MaCurdy reported research funding from the US Centers for Medicare & Medicaid Services and the US Food and Drug Administration during the conduct of the study. No other disclosures were reported.

Funding/Support: Funding was provided by the US Centers for Disease Control and Prevention (CDC).

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC.

Additional Contributions: The following people provided advice on the study methods: Daniel Weinberger, PhD, Yale School of Public Health; Marc Lipsitch, DPhil, Harvard T.H. Chan School of Public Health; and Stephanie J. Schrag, DPhil, CDC. None received compensation for their contributions.