Epidemiology, vaccine effectiveness, and risk factors for mortality for pneumococcal disease among hospitalised adults in Singapore: a retrospective case-cohort study

Background: Streptococcus pneumoniae infections can lead to severe morbidity and mortality, especially in patients with invasive pneumococcal disease (IPD). This study evaluated factors associated with pneumococcal disease, pneumococcal vaccine effectiveness, and risk factors for mortality among hospitalised adults with pneumococcal disease in Singapore. Methods: Retrospective case-cohort study of patients tested for pneumococcal disease with streptococcal urinary antigen testing and at least one sterile site culture, during their admission to a tertiary hospital in Singapore from 2015-2017. Patients were dened as cases of IPD or non-IPD, or as controls, based on laboratory results and clinical diagnoses. Multivariable models were constructed to determine factors associated with IPD/non-IPD, and risk factors for mortality from pneumococcal disease. Vaccine effectiveness against IPD/non-IPD was estimated using a variation of the test-negative design. Results: We identied 496 pneumococcal disease cases, of whom 92 (18.5%) had IPD. The mean age of cases was 69.1±15.4yrs, and 65.5% were male. Compared with controls (N=9,181), IPD patients were younger (mean age 61.5±16.3yrs, vs 72.2±16.1yrs in controls; p<0.001) and with less co-morbidities [median Charlson’s score 1 (IQR 0-4), vs 3 (1-5) in controls; p<0.001]. IPD patients also had the highest proportions with intensive care unit (ICU) admission (20.7%), inpatient mortality (26.1%) and longest median length of stay [9 (IQR 8-17) days]. On multivariable analysis, IPD was negatively associated with prior pneumococcal vaccination (relative risk ratio=0.20, 95%CI 0.06–0.69; p=0.011). Risk factors for mortality among pneumococcal disease patients were ICU admission, diagnosis of IPD, age ≥ 85yrs


Background
Pneumococcal disease is caused by Streptococcus pneumoniae and can lead to severe clinical outcomes and death.S. pneumoniae is the most common cause of community-acquired pneumonia (CAP), estimated to cause 27.3% of CAPs worldwide 1 , and accounting for 29.2% of isolated pathogens from CAPs in Asia 2 . Other common sites of infection include otitis media and sinusitis. Invasive pneumococcal disease (IPD), de ned as isolation of S. pneumoniae from a normally sterile site (such as blood or cerebrospinal uid), carries an even higher risk of mortality and morbidity 3 .
Cases of pneumococcal disease are predominantly male, and patients 65yrs and above have the highest hospitalisation rates and case-fatality rates 4,5 . From 1995 to 2004, the overall mean annual hospitalisation rate for pneumococcal disease in Singapore was 10.9 per 100,000 population and the case-fatality rate was 3.2% 4 . Speci cally for IPD, locally reported case-fatality rates have ranged from 13.1% to 21.4% 6 . Pneumococcal disease may be transmitted via respiratory droplets from people with pneumococcal disease or asymptomatic individuals with carriage of the organism in their nasopharynx 7 . Risk factors for pneumococcal disease include presence of chronic diseases involving cardiovascular, pulmonary, hepatic, neurological or endocrine systems; immunosuppressed states; cigarette smoking history; alcohol abuse; recent in uenza infection; institutionalised status; male sex and extremes of age 8 . Successful empiric antibiotic therapy is vital to avoid treatment failure and subsequent costs 9 . However, increasing antibiotic resistance in S. pneumoniae in the community has been reported 10 . Hence, while updated treatment recommendations (especially for CAP) have reduced disease-associated complications and mortality, the risks from infection still remain signi cant 11 .
Previous studies in Asia have described the changing epidemiology of pneumococcal disease due to introduction of pneumococcal vaccines 19,20 . The aims of this study were to evaluate the factors associated with pneumococcal disease, the effectiveness of vaccination in preventing severe outcomes of pneumococcal disease, and the risk factors for pneumococcal disease-associated mortality, among hospitalised adults in Singapore.

Patient population and study design
We conducted a retrospective case-cohort study on patients suspected to have pneumococcal disease admitted to Tan Tock Seng Hospital (TTSH), the second largest adult acute tertiary care hospital in Singapore with 1,600 beds, from January 2015 through December 2017. Our cohort comprised patients admitted over the study period who were evaluated for pneumococcal disease, using streptococcal urinary antigen testing and at least one sterile site clinical culture during their admission. Cases of pneumococcal disease were de ned as either having i) a positive result on testing for streptococcal urinary antigen, or ii) growth of S. pneumoniae from sterile site cultures. For patients with multiple admissions for pneumococcal disease, only the rst admission was included for this study.
We further strati ed cases into those with IPD versus those with non-IPD. IPD was de ned as either having i) a sterile site culture (e.g. blood, joint uid, or pleural uid) positive for S. pneumoniae, or ii) a positive streptococcal urinary antigen test and a clinical diagnosis suggesting invasive disease (e.g. meningitis, septic arthritis, or pleural empyema), with cultures from the relevant site negative for any other causative pathogen. A non-IPD case was de ned as a patient with a positive streptococcal urinary antigen test, but without positive cultures or any clinical diagnosis suggesting invasive disease.
To identify factors associated with IPD and non-IPD, the cases were compared to a control group comprising the remainder of the cohort. For clarity, controls were de ned as patients admitted to TTSH over the same study period, with a negative test for streptococcal urinary antigen, and all laboratory cultures yielding negative results for S.pneumoniae. Again, for patients with multiple admissions, only the rst admission ful lling the de nition was considered. Patients who did not have both streptococcal urinary antigen tests and at least one sterile site culture (typically, at least one blood culture) performed during their admission were not considered part of the cohort, and were hence excluded from this study.

Data sources and classi cation
All data for the study was captured using the hospital's electronic medical record system, which comprehensively captures the demographics, clinical characteristics, laboratory results, and vaccination records of patients. Pneumococcal vaccination was de ned as ever having any record for PCV13 and/or PPSV23 given at least 14 days prior to the date of hospital admission. Inpatient mortality was de ned as death resulting from any cause during the pneumococcal disease-related admission. International Classi cation of Diseases, 10 th Revision (ICD-10) was used to establish patient diagnoses, comorbidities, risk factors, and infectious disease history. The co-morbidity status was assessed according to Charlson's co-morbidity score classi cation 21 .

Statistical analysis
Descriptive statistics were analysed for the overall cohort and patient subgroups, including the number and proportions for categorical characteristics, mean and standard deviation for normally distributed continuous variables, and median and interquartile range (IQR) for non-normally distributed variables.
We subsequently analysed differences in patient characteristics across our three patient subgroups (IPD, non-IPD and controls). Differences in proportions were assessed using Chi-square or Fisher's exact tests for categorical variables, while continuous variables were evaluated using the one-way ANOVA (for means) or Kruskal-Wallis test (for medians) as appropriate. Multinomial regression analysis was used to further evaluate patient factors associated with IPD and non-IPD. Factors to include in the model were determined based on a p value threshold of 0.05 on univariate analysis, using non-pneumococcal disease patients as the base comparison group, with calculation of relative risk ratios (RRRs) for independent variables. Within our multivariable model, we included pneumococcal vaccination status (receipt of PCV13 and/or PPSV23 vaccination at least 14 days prior to admission date) as a variable to determine its effectiveness in reducing risk of IPD / non-IPD. This was a variation of the test-negative design used for vaccine effectiveness studies.
Among the cases, we also evaluated factors associated with inpatient mortality. Univariate analysis of demographic and clinical variables was rst performed, and signi cant variables were included in a multiple logistic regression model with adjustment for demographic confounders. Unadjusted and adjusted odds ratios (ORs) and 95% con dence intervals (CIs) were calculated to determine the strength of association between variables and in-hospital all-cause mortality.
For all analyses, a p value of <0.05 was considered statistically signi cant. Furthermore, inclusion of variables into multivariable models also took into account con dence intervals and clinical signi cance of variables. All statistical analyses were performed using Stata version 13 (StataCorp 2013, College Station, TX).
In terms of clinical outcomes, IPD was associated with the highest proportion of ICU admissions (20.7%, versus 8.7% in non-IPD patients and 6.3% in controls; p<0.001) and in-hospital all-cause mortality (26.1% versus 11.4% and 9.1% respectively; p<0.001). Similarly, length of stay was longest for IPD patients (median duration 9 days, versus 8 days in non-IPD and control groups; p=0.003).
In our multivariable models (Table 2), IPD patients were much less likely than controls to have a history of any pneumococcal vaccination prior to admission (RRR=0.20, 95%CI 0.06-0.69; p=0.011). IPD was also positively associated with male gender (RRR=1.87, 95%CI 1.18-2.96; p=0.008), and negatively associated with increasing age groups, with very elderly patients (≥85yrs) 0.19 times less likely to have IPD compared to younger counterparts (<65yrs) (p<0.001). Although in uenza vaccination was negatively associated with IPD on univariate analysis, this factor was non-signi cant after multivariable adjustment. Non-IPD patients were also more likely than controls to be of male gender, although the effect size was smaller than in IPD patients (RRR=1.42, 95%CI 1.15-1.76; p=0.001). Patients ≥85yrs and those with Charlson's score of >3 were also less likely to have non-IPD, and no signi cant differences in vaccination status was observed.

Discussion
Among our cases comprising hospitalised patients with pneumococcal disease, majority were aged ≥65 years and of male gender, and about a quarter had a Charlson's score of >3. The ethnic distribution of the patients was re ective of the national census 22 . However, compared to the control group , IPD patients demonstrated a trend towards being younger, and with lower Charlson's scores. Non-IPD patients were more similar to the control group, but with only slightly lower mean age and Charlson's scores. In our study, patients in older age groups or with more co-morbidities might have had increased prior healthcare exposures, or had speci c risk factors (such as swallowing impairment) which resulted in them being more likely to develop infections 23,24 from other aetiologies other than S. pneumoniae 25,26 . However, in contrast to the age and co-morbidity trends, the ICU admission and in-hospital mortality rates were highest in IPD patients and lowest in the control group. This re ects the virulence of the disease and highlights the importance of early recognition, treatment, and prevention where possible.
Our ndings showed that prior pneumococcal vaccination was associated with a reduced risk of IPD by about 80%, after adjustments for age, gender, Charlson's score and in uenza vaccination status. This was consistent with previous studies evaluating the effectiveness of PCV13 and PPSV23 [13][14][15] . Pneumococcal serotypes covered by these vaccines have been shown to be well matched to those circulating in the Singapore population 6,10,27 . In particular, serotypes 3, 6B, 7F, 8, 19A and 23F are predominant in the 19-64yrs age group, while serotypes 3, 14 and 19A are predominant among those 65yrs and above 27 .
In addition, 50 (54.4%) of the IPD cases among our hospital patients were below 65yrs of age, of whom 24 (48.0%) had some form of chronic disease. A Japanese study of 10.4 million individuals demonstrated that younger adults with at least one medical condition were at greater risk of IPD compared to healthy older adults. For example, an adult aged 50-64yrs with an underlying medical condition had a higher risk compared to a healthy adult aged ≥65yrs 28 . Although the overall incidence in younger adults is lower than in older age groups 4,29 , the increased susceptibility of adults <65yrs with chronic diseases, demonstrated by our study in line with the above literature, suggests that targeted efforts to vaccinate this population might be bene cial in reducing IPD incidence.
Overall, our results support current recommendations to vaccinate at-risk individuals to prevent IPD. This may be provided by primary care practitioners as part of chronic disease management, or alongside other preventive services such as health screening. In addition, our hospital has ongoing vaccination programmes to identify and opportunistically vaccinate high-risk groups according to standardised protocols. In the inpatient setting, high-risk inpatients are counselled on and given vaccination prior to discharge, while in the outpatient setting, selected specialist clinics have a nurse-led programme for counselling and vaccination administration for high-risk patients, in conjunction with the medical consultation.
Our study, however, did not demonstrate any signi cant effect of PCV13 vaccination on reducing risk of pneumococcal disease, likely due to the low proportion of patients in our study with a clear documented history of PCV13 vaccination.
Apart from adult vaccination, widespread introduction of pneumococcal conjugate vaccine in children has been shown to reduce the incidence of IPD across all ages groups due to herd immunity, despite evidence of serotype replacement [30][31][32][33] . Singapore introduced PCV7 into the National Childhood Immunisation Programme (NCIP) in October 2009 34  Finally, our study also demonstrated key factors associated with in-hospital mortality among pneumococcal patients.
ICU admission was the strongest predictor, with those having admission being 23 times as likely as those who did not to die during their admission. Other factors included age ≥85yrs, higher Charlson's score and diagnosis of IPD. Medical teams should especially note the increased mortality risk in these patients and ensure prompt management to reduce the risk of death.
The strengths of our study include a systematic selection of cases and controls among hospitalised patients. As such, our study population is likely to represent the population at risk for pneumococcal disease and the risk of selection bias (if any) is reduced. Furthermore, we followed up the patients longitudinally to the point of discharge from the hospital.
Our ndings are useful to guide the identi cation of patients at high risk of in-hospital mortality from the disease.
Our study also has some limitations. Our de nition of cases included the use of results from streptococcal urinary antigen testing, which has an estimated sensitivity of 74.0% and speci city of 97.2% 36 . Sterile site cultures from multiple possible sources (e.g. blood, joint uid, or pleural uid) will also have varying microbial culture methods with different sensitivity and speci city. Some cases of pneumococcal disease may hence have been misclassi ed as controls (i.e. false negatives). However, such a misclassi cation could have reduced the magnitude of our ndings, but not negated them. As our data was obtained through hospital electronic medical records, clinical data from other sources (e.g. other hospitals or the primary care sector) were not available. However, majority of our patients tended to return back to our hospital if future admissions were required. Moreover, it is likely that this issue would result in nondifferential misclassi cation (if any) as data would not be captured differently across the study groups. We were unable to capture data for some known risk factors, such as smoking history and socioeconomic status. Our analysis of IPD cases was limited by the relatively small numbers, especially when divided into subgroups. Pneumococcal serotype data were also not available for this study, as this is not routinely performed for isolates. Nevertheless, we have reviewed data from serotyping studies conducted in the local setting 6, 10, 27 , which we believe to be applicable to our cohort. The generalisability of our study is limited to hospitalised adult patients.

Conclusion
In conclusion, while patients with pneumococcal disease tended to be younger and with less co-morbidities compared to those without pneumococcal disease, risk of ICU admission and death were nevertheless higher, especially in IPD patients. Mortality risk was highest in those with ICU admission, age ≥85yrs, higher Charlson's score and diagnosis of IPD. Pneumococcal vaccination was associated with reduced risk of IPD and should be encouraged among high-risk patients, including those from younger age groups.

Declarations
Ethics Approval and Consent to Participate: Ethics approval for this study was obtained from the Domain-Speci c Review Board of the National Healthcare Group (NHG DSRB No: 2017/00347).

Consent for Publication: Not applicable.
Availability of Data and Materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.
Funding: Not applicable.
Authors' Contributions: TC and MZT analysed and interpreted the data collected for this study, and contributed to the writing of the manuscript. WMK and AC contributed to study design, interpretation of data, and critically reviewed the manuscript. HJH conceived the study, analysed and interpreted the data, and was a major contributor in writing the manuscript. All authors approved the nal version of the manuscript for submission.  Table 3: Univariate and multiple logistic regression analysis of demographic and clinical factors associated with mortality among pneumococcal disease patients. OR = odds ratio; CI = confidence interval; AOR = adjusted OR; SD = standard deviation; IQR = inter-quartile range; ICU = intensive care unit.