Prognostic factors for mortality in invasive pneumococcal disease: a system review and meta-analysis

Risk factors associated with mortality in invasive pneumococcal disease remain unclear. The present work is a meta-analysis of studies that enrolled only patients with invasive pneumococcal disease and reported on mortality. Potentially eligible reports were identied from PubMed, CHAHL, and Web of Science, comprising 26 reports in total. Overall mortality for invasive pneumococcal disease was reported as 20.8% (95% condence interval (CI) 17.5–24%). Factors associated with mortality were age (odds ratio (OR) 3.04, 95%CI 2.5–3.68), nursing home (OR 1.62, 95%CI 1.13–2.32), nosocomial infection (OR 2.10, 95%CI 1.52–2.89), septic shock (OR 13.35, 95%CI 4.54–39.31), underlying chronic diseases (OR 2.34, 95%CI 1.78–3.09), solid organ tumor (OR 5.34, 95%CI 2.07–13.74), immunosuppressed status (OR 1.67, 95%CI 1.31–2.14), and alcohol abuse (OR 3.14, 95%CI 2.13–4.64). Mortality rates with invasive pneumococcal disease remained high, and these ndings may help clinicians provide appropriate initial treatment for this disease.


Background
Streptococcus pneumoniae commonly colonizes the upper airway asymptomatically, and is the cause of approximately 25-50% of cases of community-acquired pneumonia and meningitis. 1 The most severe form of pneumococcal infection, invasive pneumococcal disease (IPD), includes bacteremic pneumonia, meningitis, and septicemia with signi cant morbidity and mortality. In the United States of America, data from the Centers for Disease Control and Prevention showed 17 cases of IPD per 100,000 persons in all age groups. 2 In England and Wales, an IPD incidence of 7 per 100,000 persons in all age groups was reported, increasing to 21 per 100,000 among persons ≥65 years old. 3 The case fatality rate of IPD may reach 15-20% in adults and 30-40% in the elderly. 4 A marked reduction in the rate of IPD due to protein conjugate vaccine 7 and protein conjugate vaccine 13 serotypes has been reported for individuals ≥ 55 years old, with a much smaller decline in those 17-54 years old. 5 In the prospective international observational study, the overall mortality rate of pneumococcal bacteremia was 17%. 6 Although various articles have discussed risk factors for mortality from IPD, an overview has been lacking. This study was therefore designed to clarify prognostic factors contributing to death due to IPD.

Overview
The study protocol followed the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group statement and was registered to the University Hospital Medical Information Network (ID: UMIN000041377) 7,8 . The need for institutional review board approval was waived because of the systematic review nature of this research. H.C. and M.M. independently performed the searching process and data extraction, and subsequently built consensus results.
Search of the literature Three major databases (Medline, CHAHL, and Web of Science) were searched on July 30, 2020. The two reviewers independently extracted and recorded data for a prede ned checklist including the following items: study characteristics (i.e., country and year of study), characteristics of the cohort, and case de nitions. The following search formula was used: ("invasive pneumococcal disease" OR "pneumococcal infection" OR "Streptococcus pneumoniae infection" OR "pneumococcal bacteremia") AND ("mortality" OR "death") AND ("odds ratio" OR "relative risk" OR "hazard ratio").

Inclusion and exclusion criteria
No restrictions were placed on article types or publication language. To be included, a study had to include: 1) patients with IPD; 2) data outlining risk factors for 30-day mortality or in-hospital mortality after multivariable regression analysis; and 3) no restrictions on vaccination.
Exclusion criteria were as follows: 1) patients < 15 or > 65 years old only; 2) risk factors assessed by univariable analysis only; 3) risk factors assessed in categories of subgroup only; or 4) patients with human immunode ciency virus (HIV) infection only.
De nitions IPD was de ned as an illness occurring in association with the isolation of S. pneumoniae from a normally sterile body specimen including blood, cerebrospinal uid, peritoneal, pleural or synovial uid, or abscess aspirates or tissue specimens/swabs obtained intraoperatively, but excluding bronchoalveolar lavage. Immunosuppressed status was de ned as HIV infection, splenectomy, hematological malignancy, autoimmune disorder, presence of a transplant, or cancer chemotherapy within 4 weeks before the onset of bacteremia. Septic shock was de ned as sepsis-induced hypotension, persisting despite adequate uid resuscitation, along with the presence of hypoperfusion abnormalities or organ dysfunction. 9 Outcomes Potential prognostic factors were de ned in this study as any clinical information related to mortality such as age, sex, underlying diseases, focus of infection, and place of infection acquisition. Risk factors for 30-day and in-hospital mortalities were analyzed separately if > 3 articles were found in each subgroup. Quality assessment Two reviewers independently assessed the methodological quality of selected studies using the Newcastle-Ottawa Scale quality assessment to evaluate the quality of observational studies.
Disagreements among reviewers were discussed, with agreement reached by consensus. 10

Study search
Of these 26 articles identi ed, 2 demonstrated data from two independent populations (Fig. 1). Our analysis thus eventually extracted data from 16 countries for 27,742 patients with IPD, of whom 5810 died.
Characteristics of included studies The 26 articles were published between 2000 and 2020, reported from Spain and the United States (n = 3 each), Canada, Japan, Israel, South Korea, the Netherlands, the United Kingdom, and Taiwan (n = 2 each), and 6 other countries (n = 1 each). 6,12−36 Two articles from Sweden and Georgia were multi-country studies compromising patients from 5 and 10 countries (Table 1). Twenty-one studies assessed either only adults or only adults plus adolescents ≥15 years old. Six studies included patients with IPD at any age. Six studies included only patients with HIV infection. Only two studies discussed status of HIV infection separately and the remaining four studies classi ed HIV as one condition de ning an immunosuppressed status. Eight studies did not discuss the status of pneumococcal conjugate vaccination, seven studies revealed missing data for vaccination status, and nine studies discussed vaccination status (vaccination rates, 0-54.5%). No study identi ed a relationship between vaccination and mortality in multivariable analysis. Thirteen studies discussed in-hospital mortality as an outcome, twelve studies discussed 30-day mortality rates, and one study discussed 28-day mortality rate. The median in-hospital mortality was 23.0% (95% con dence interval (CI) 17.2-27.2%) and the median 30day mortality rate was 18.9% (95%CI 13.9-23.9%). The overall mortality rate from IPD included in this study was 20.8% (95%CI 17.5-24.0%). Median Newcastle-Ottawa Scale score was 6, suggesting that most studies were of acceptable quality. NOS: Newcastle-Ottawa Scale score; score ranges from 0 (worst) to 9 (best).
The I 2 statistic of 37.2% suggested mild heterogeneity of this model. No heterogeneity between inhospital mortality and 30-day mortality comparisons was revealed (P = 0.21).

Discussion
Despite the wide adoption of pneumococcal conjugate vaccination, the overall mortality rate from IPD has remained high, at 20.8%. This meta-analysis revealed older age (> 64 years old), septic shock, immunosuppressed status, underlying chronic diseases, solid organ tumor, alcohol abuse, nursing home, and nosocomial infection were prognostic factors for mortality from IPD. These results appear useful in understanding prognostic factors for mortality due to IPD because of the solid methodology in accordance with the MOOSE statement, with statistical power supported by more than 27,000 subjects.
Aging is a major risk factor for the development of virtually every lung disease, increasing both morbidity and mortality, while morbidities and mortalities from other prevalent diseases have declined or remained stable. 37 The conventional nuclear family model is becoming increasingly uncommon, and the majority of elder care is provided by relatives, albeit with varying patterns of involvement and responsibility across family structures. 38 Residence in a nursing home and nosocomial infection have been associated with the progression of aging. Comorbidity also plays an important role in affecting mortality. This study identi ed underlying chronic diseases, solid organ tumor, and immunosuppressed status as important factors to clinical progress.
Septic shock is a frequent complication of pneumococcal infection and causes high rates of morbidity and mortality. 39 The presence of septic shock on admission was the most strongly associated with risk of mortality among those factors. Unlike prognostic factors that cannot be changed, such as age, facility, and underlying diseases, outcomes might be improved with medical intervention for septic shock. Alcohol abuse is another preventable risk factor for pneumococcal disease, particularly among young adults. 40 This study also revealed alcohol abuse as a risk factor for mortality. Pneumococcal vaccination with PCV13 or PPV23 in adults is cost-effective and should be considered a priority for decision-makers, 41 and status of vaccination varied widely among studies (0-54.5%). Pneumonia vaccination was another effective means of preventing IPD, but much work remains to be done to increase the acceptance of pneumonia vaccination.
Limitations Several limitations to this study must be considered when interpreting the results. Given the nature of the disease in question, only a limited number of prospective studies were able to be enrolled. Second, the de nition of mortality in articles as in-hospital or 30-day mortality was merged together in some risk factors, due to limited numbers of articles. Third, prognostic factors for mortality differed markedly between articles. Fourth, the forms of vaccine status for studies included in this meta-analysis varied.
Fifth, the effects of serotype and antibiotic resistance were not discussed in the present investigation.

Conclusion
IPD still shows high mortality rates despite wide adoption of pneumococcal conjugate vaccines. The presence of septic shock represents one of the most important prognostic factors for mortality in IPD.

Declarations
Funding None.

Declaration of competing interest
The authors have declared that no competing interests exist.