Association Between Social Vulnerability and Streptococcus pneumoniae Antimicrobial Resistance in US Adults Open Access

Abstract

Infections with drug-resistant Streptococcus pneumoniae (SP) are considered a serious threat by the US Centers for Disease Control and Prevention (CDC) [1] and account for approximately 830 000 deaths worldwide annually [2]. S. pneumoniae is also the fourth leading pathogen globally in terms of deaths attributable to or associated with antimicrobial resistance (AMR) [3].

Recent studies have demonstrated high rates of AMR in North American SP isolates in hospitalized and ambulatory adults [4–6]. In a 2022 study of 290 US hospitals, SP AMR rates were persistently high in adults. Among 34 039 SP isolates analyzed, almost half (46.6%) were resistant to ≥1 drug with high rates of resistance to macrolides (37.7%), penicillin (22.1%), and tetracyclines (16.1%) [5]. In 2018–2019, nearly 40% of pneumococcal bacteria collected from US ambulatory centers and hospitals were reported to be resistant to macrolides [4], whereas a 2017 study of 105 hospitals in the SENTRY US database reported that 16% of SP isolates were multidrug resistant (defined as intermediate or resistant to ≥3 drug classes) [6].

Use of pneumococcal vaccines has been associated with reduced AMR in SP [7–11]. However, recent data from the CDC Active Bacterial Core surveillance revealed an increase in AMR associated with nonvaccine-type serotypes in all age groups [12]. Resistance to frequently used antibiotics is an important concern for SP infections and associated patient outcomes, such as delayed disease resolution and increases in hospitalizations and deaths [5].

AMR may be exacerbated by social, cultural, and economic factors [13]. Evidence suggests antimicrobial-resistant infections disproportionately impact those living in socially vulnerable areas [14, 15] and factors such as socioeconomic status, minority status, and race are associated with a higher incidence of antimicrobial-resistant infections [15–18]. The CDC defines social vulnerability as the characteristics of a community that increase the potential for deleterious outcomes when faced with natural disasters or disease outbreaks [19]. Social vulnerability highlights social, economic, demographic, and geographic characteristics that determine risk exposure [20]. Important social factors that determine susceptibility to infections and subsequent need for antimicrobials include economic stability, education, gender, environment, livelihood, food, and nutrition [21]. Individuals who are underprivileged or vulnerable in one or more of these areas may experience a higher rate of infections and require more antimicrobials [22].

In the United States, racial disparities in the incidence of invasive pneumococcal disease (IPD) and pneumococcal vaccine coverage rates are evident [23–26], and clusters of IPD among persons experiencing homelessness have been reported [27, 28]. However, few studies describe the impact of social inequities on antimicrobial-resistant infections in the United States, particularly those attributed to SP [29, 30]. Accordingly, the CDC is prioritizing addressing health disparities and health equity issues related to antimicrobial resistance through the launch of several programs designed to increase equity across public health [31]. It has been suggested that a thorough investigation of all health inequities and their contributing factors is urgently needed to develop strategies addressing these issues [32].

Based on this need, we sought to evaluate the association between social vulnerability and SP AMR in US adults, utilizing the Social Vulnerability Index (SVI) developed by the Centers for Disease Control and Prevention/Agency for Toxic Substances and Disease Registry (CDC/ATSDR) [33].

METHODS

Design and Patients

In this retrospective study, antibiotic susceptibility of nonduplicate SP isolates (first noncontaminant SP isolate within 30 days) were collected from hospitalized and ambulatory patients between January 2011 and December 2022. The study included adult patients (≥18 years) who had zip codes of residence reported and matched with CDC SVI data.

Evaluation took place across 177 facilities in the BD Insights Research Database (Becton, Dickinson & Co., Franklin Lakes, New Jersey, USA). Isolates were defined as SP AMR if they were reported by the facilities as intermediate or resistant (non-susceptible [NS]) to ≥1 of the following antibiotics: macrolides, tetracyclines, extended-spectrum cephalosporins, or penicillin. Isolates from adult patients with zip codes of residence reported were matched with county-level CDC SVI data [34]. Isolates were categorized as IPD or non-IPD. Culture sources for IPD isolates were blood, cerebrospinal fluid, or a neurologic source, whereas sources for non-IPD isolates were respiratory or ear, nose, and throat [5].

Social Vulnerability Index

The SVI is an integrated metric that characterizes 4 themes: (1) socioeconomic status (SES), (2) household characteristics, (3) racial and ethnic minority status, and (4) housing type and transportation. Each theme is derived from a series of components, which are listed in Table 1. For clarification purposes, it is helpful to differentiate the household characteristics theme from housing type and transportation. The household characteristics theme describes individual family members, whereas housing type and transportation characterizes aspects of the physical home and transportation options.

For each of the 4 themes, percentiles for the 15 census tract level variables comprising each theme are summed, and the sum of these variables is ranked to determine an overall census tract percentile for each theme. To develop an overall theme ranking, theme percentiles are summed and ranked to determine a census tract percentile for the overall SVI [33]. Percentile ranking values range from 0 to 1, where higher values indicate higher vulnerability. Scores were presented as percentile rankings by county.

Statistical Analyses

Statistical analyses were performed in 5 parts; for each of the 4 SVI theme scores and the overall SVI score, a modeling analysis was conducted to evaluate the association between SVI score and SP AMR. In the exploratory phase of the analysis, descriptive statistics of AMR percent (number of non-susceptible isolated per 100 total tested isolates [% NS]) against each SVI ranking score were generated to explore the relationship of AMR and SVI. Bivariate associations of an SVI theme score or the overall SVI score with SP AMR was evaluated using the logistic regression method. These exploratory and bivariate analyses provide a preliminary look at the association between SVI and AMR without considering covariates. The final results reported in the study were based on the multivariable model-adjusted association between SVI score (overall and by each theme) and SP AMR using the generalized estimating equations (GEE) method with repeated measurements over time and nested within county to account for within-cluster correlations of data. The first-order autoregressive variance-covariance structure was used in the GEE models to account for temporal correlation of longitudinal data. A sensitivity analysis was also performed to evaluate potential differences between AMR for isolates with matched zip codes versus those that were unmatched.

The covariates considered in each GEE model included clinical factors, such as specimen collection setting (inpatient vs outpatient) and IPD versus non-IPD status based on source of culture collection. Additional covariates were patient demographics, sex, and facility characteristics (Table 2).

The models were tested using both SVI score as a continuous variable and as a categorical variable (categorizing SVI into 4 groups based on quartiles of SVI score). Similar results were achieved with the 2 parameterizations of SVI. The final model results were based on the GEE models using SVI scores as a continuous measure (Table 4). The effect of SVI on AMR was reported as the difference in risk of AMR per 10 percentage points of an SVI score. All analyses were conducted using SAS version 9.4 software (SAS Institute, Cary, NC).

RESULTS

The mean age of patients was 60.2 years (median 61 years); the majority were male (54.9%) and 50 years of age or older (78.2%), with 37.5% being between aged 50 and 74 years. Hospital characteristics are summarized in Table 2 and compared to a national sample of US hospitals; most hospitals (81%) were located in urban areas, over 40% had 100–300 beds, and 65% were non-teaching.

Across 40 151 SP isolates evaluated, a total of 8008 non-duplicate SP isolates (19.9%) with zip code of residence reported were matched with county-level CDC SVI data and analyzed (Table 3, Supplementary Table 1). Most isolates (5117; 63.9%) were derived from cultures collected within 3 days of hospitalization, whereas the remainder were collected in ambulatory settings (2022; 25.2%) or ≥3 days from hospitalization (869; 10.9%). Overall, 3333 (41.6%) SP isolates were categorized as IPD and 4675 (58.4%) as non-IPD. Respiratory samples were the most common across all culture source evaluated (52.8%). Of the 8008 SP isolates linked to 574 US counties, the overall rate of AMR was 49.9%. Among all isolates, AMR to macrolides, penicillin, extended-spectrum cephalosporins, and tetracyclines was 40.6%, 25.8%, 9.2%, and 16.9%, respectively; overall 13.4% were reported as multidrug resistance. AMR among non-IPD isolates was markedly greater compared with IPD isolates (58.4% vs 38%; P < .0001) and ≥3-drug resistance was also higher for non-IPD versus IPD (17.6% vs 7.5%).

The association between AMR and SVI score (overall and by each theme) is shown in Table 3, with higher SVI ranking scores indicating higher levels of social vulnerability. Specifically, for the SES theme, the observed percentage of AMR isolates ranged from 46.1% for the 25th–50th percentile to 52.1% for the 50th–75th percentile. Similarly, for the household characteristic theme, the observed percentage of AMR isolates ranged from 48.1% for the 25th–50th percentile to 51.7% for the 50th–75th percentile. In addition, for the 8008 SP isolates evaluated, 29.5% were below the overall median SVI score and 70.5% were above the overall median SVI score.

The multivariable model results demonstrate a significant association between the SES theme and AMR, with higher SVI scores (indicating greater social vulnerability) associated with increased risk of AMR. On average, a decile increase in SES vulnerability ranking score was associated with a 1.28% increased risk of AMR (95% confidence interval [CI], .61%, 1.95%; P = .0002; Table 4). Similarly, a single decile increase in household characteristic theme (ie, family members ≤17 or ≥65 years, individuals with a disability, single parents, English language proficiency) ranking score was associated with a 0.81% increased risk of SP AMR (95% CI, .13%, 1.49%; P = .0197).

No associations of SP AMR with the racial and ethnic minority status theme, the housing type and transportation theme, and overall SVI score were observed. Although not statistically significant, directionally, the results suggest that those from a higher race and ethnicity theme ranking score area (higher concentration of ethnic minority groups, Table 1) were at decreased risk of SP AMR (−0.39% (−1.01%, 0.22%) (P = .2116)). Similarly, those with a higher housing type and transportation theme (eg, multi-unit structures, mobile homes, crowding, no vehicle, group quarters) score were at decreased risk of SP AMR (−0.69% [−1.42%, 0.04%] [P = .0648]). Although for the overall SVI score, a slight increase in SP AMR was observed, the overall SVI score had no statistically significant association with ≥1 drug SP AMR (0.59% [−0.14%, 1.13%] [P = .1150]).

DISCUSSION

Social vulnerability assessments provide an effective means for better understanding of inequities in health outcomes by connecting social conditions and risk exposure [20]. In this exploratory study, the data showed that specific determinants of social vulnerability such as SES and household characteristics were associated with increased risk of SP AMR. These findings are not unexpected because lower SES has been associated with higher antibacterial use [22, 36], which can facilitate development of antibacterial resistance. Based on the current literature, this is the first published report to our knowledge to investigate the relationship between social vulnerability and SP AMR utilizing the SVI developed by the CDC/ATSDR.

At least one previous study utilizing the CDC/ATSDR SVI has produced similar results, where SES and household characteristics were found to be most closely associated with cardiovascular disease risk factors and coronary heart disease prevalence [19]. The SES theme is an important measure of social vulnerability, as it incorporates the percentage of individuals living below the poverty line, unemployment, housing cost burden, and lack of high school diploma. Moreover, the SES theme also considers health insurance status and those individuals with a higher SES theme score may be less likely to have insurance. A lack of health insurance coverage is increasingly considered a marker of lower SES and a barrier to healthcare access [37]. Furthermore, although age was included in the multivariable model, the household characteristics theme, which comprises persons ≤17 and ≥65 years, was associated with increased risk of SP AMR; this is consistent with expectations as these populations include individuals at greater risk for pneumococcal disease.

We did not find any association between county-level race/ethnicity and increased patient-level risk of SP AMR. However, it is important to note that we were not able to identify the race/ethnicity of the individuals from which the specific isolates were derived. Rather, data on the race/ethnicity for the study population were based on reported zip codes. Because the SVI is a county-level index and counties often contain multiple zip codes, there is potential for considerable variation in the racial/ethnic composition among the zip codes within a county. Still, in the context of race/ethnicity, the significant association between SES and increased risk of SP AMR is notable, because SES and race/ethnicity are strongly related [38].

The findings that higher socioeconomic vulnerability (ie, higher SES scores) are associated with increases in SP AMR suggest that addressing the disparities in SES may reduce AMR. Few studies have assessed the impact of social vulnerability on AMR attributed to SP, and the current results are consistent with prior studies suggesting that AMR disproportionately affects individuals who are socially vulnerable. For example, a 2017 study by See and colleagues found that socioeconomic factors explained in large part the racial disparities observed in invasive community-acquired methicillin-resistant Staphylococcus aureus (CA-MRSA) disease rates [15]. In a 2022 analysis of 43 studies in Canada, there was a positive association between higher income and protection from CA-MRSA [16].

An important tool against AMR is prevention, which can be accomplished with vaccination [39]. Importantly, several studies have also demonstrated racial disparities in pneumococcal vaccination coverage in the United States and lower incidence of NS-IPD after introduction of PCVs [12, 23–25]. The CDC assessed vaccination coverage among adults aged ≥19 years in the United States, based on an analysis of data from the National Health Interview Survey (NHIS) [23]. In that study, pneumococcal vaccination coverage for White adults aged 19–64 years at increased risk for pneumococcal disease was higher (26.3%) compared with Hispanic (16.7%) and Asian (13.8%) adults. Similarly, for individuals ≥65 years of age, pneumococcal vaccination coverage among White adults (72.4%) was higher compared with Black (50.8%), Hispanic (48.1%), and Asian (54.9%) adults [23]. More recently, similar health disparities were found for coronavirus disease 2019 (COVID-19) vaccination coverage, which was lower in individuals with greater social vulnerability [40, 41]. Based on the above evidence, careful consideration should be given to vaccine access when planning and developing vaccination strategies.

This study has several limitations. First, the 12-year assessment period could have introduced bias as the majority of data were collected after 2018 and results were not stratified by time period, although the temporal correlation was accounted for using the GEE modeling framework. As noted previously, this analysis did not evaluate the individual characteristics of patients, and the race and ethnicity of patients from which isolates were derived were unknown. Analyzing resistance based on local microbiology practices at each facility may have influenced reported resistance rates, as opposed to using a central laboratory using standardized susceptibility breakpoints. We could not verify that the laboratories used the updated breakpoints introduced by the Clinical and Laboratory Standards Institute (CLSI), which can take years to be fully adopted, but our data do reflect the information clinicians use to guide their daily management decisions [5, 42, 43]. The analysis was based on culture-positive isolates and the presence of symptomatic pneumococcal disease was not confirmed. There was also potential for selection bias due to a greater likelihood of testing patients who were more severely ill or lack of data on prior antibiotic use, which may have increased resistance rates. The overall prevalence rate of ≥1 drug SP AMR observed in this study (49.9%) was higher compared with several recent studies that have investigated SP AMR in hospitalized and ambulatory US adults [4, 5]. In this study, it was not possible to probe deeper to evaluate pneumococcal vaccination status of patients or SP serotypes. In addition, only approximately 20% of the SP isolates that were analyzed in this study were matched with zip codes. However, an evaluation of AMR for isolates with matched zip codes versus those that were unmatched did not reveal any differences (see Supplementary Table 1). Similarly, patient demographics and clinical characteristics were also comparable between the matched and unmatched cohorts. Finally, patients from eastern US census divisions (New England, Middle Atlantic, East North Central, East South Central, and South Atlantic) were overrepresented in the sample, accounting for 77% of the study population, compared with 23% for western regions. A region-specific imbalance such as this may obscure local level inequities [20].

These findings have important implications for pneumococcal vaccination, which is an effective strategy for reducing both pneumococcal disease and AMR infections [7–11]. Although pneumococcal vaccines have helped reduce AMR, resistance to antibiotics persists in US SP isolates, partly due to increases in pneumococcal disease caused by serotypes not covered by PCVs, which can be highly resistant to antibiotics. Important strategies to combat increasing SP AMR are increasing vaccination coverage, especially in those communities that face health inequities, as well as development of vaccines with broader coverage of serotypes observed in adults [5, 14].

In conclusion, higher levels of social vulnerability were observed to be associated with increased risk of SP AMR. These results contribute to evidence linking AMR to health disparities and inequities and suggest that AMR is impacted by SDOH.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The authors thank Nicole Cooper and the HealthCare Alliance Group, LLC, for providing manuscript support. They thank Centers for Disease Control and Prevention (CDC) staff members Emma Accorsi, Kristin Andrejko, James Baggs, Sophia Kazakova, and Sarah Yi for their review of and feedback on the manuscript.

Author contributions. Study concept and design: S. M., G. Y., C. S, N. C., M. W., K. Y., V. G. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: S. M., G. Y., C. S, N. C., M. W., K. P. K, V. G. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: G. Y., C. S., V. G. Obtained funding: S. M., V. G. Administrative, technical, or material support: S. M., G. Y., C. S., N. C., V. G. Supervision: S. M., G. Y., C. S., N. C., V. G. All authors read and approved the final manuscript.

Financial support. Funding for this research was provided by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA. S. M., N. C., M. W., and K. P. K. are employees of the funding source and played a role in the design and conduct of the study; interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References

Author notes