Results
Patient Characteristics
Our study comprised 4,522 IPD case-patients, and 4,315 (95%) were matched with 8,837 controls; there were 2 controls/case-patient (Figure 1). Overall, 4,357 (96.4%) IPD case-patients had ≥1 hospitalized control; for some IPD case-patients (n = 165, 3.6%), nonhospitalized controls were required. The mean (±SD) age of case-patients was 55.8 (±17.7) years, and 56.7% of case-patients were male; this distribution remained consistent over time (Table 1; Appendix Table 1, https://wwwnc.cdc.gov/EID/article/28/8/21-2469-App1.pdf).
Figure 1.
Flowchart diagram of case inclusion for study of IPD long-term mortality rates in adults, Alberta, Canada. AHCIP, Alberta Health Care Insurance Plan; IPD, invasive pneumococcal disease.
Data on site of infection were available for patients only during 1999–2014. Of those patients, 67% had S. pneumoniae identified in >1 sterile body site. There were 2,008 (44.4%) cases of invasive pneumonia, of which 1,961 (97.7%) also had a positive blood culture; the remaining 2.3% had S. pneumoniae isolated from another sterile body site (pleural fluid, pericardial fluid, and peritoneal fluid). There were 116 (2.6%) cases of meningitis, 2,315 (51.2%) cases of bacteremia/sepsis, and 646 (14.3%) cases from an unspecified/other sterile source (not mutually exclusive) (Table 1).
All-cause Mortality Rate Within 30 Days
Within 30 days of the IPD diagnosis date (or pseudodate for controls), there were 614 deaths among IPD cases (1,915 deaths/1,000 person-years), compared with 348 deaths in the control group (510 deaths/1,000 person-years) (Figure 2, panel A). After adjustment, IPD cases were strongly associated with increased risk for 30-day mortality rate (adjusted odds ratio [aOR] 4.08, 95% CI 3.54–4.69; adjusted hazard ratio [aHR] 3.75, 95% CI 3.29–4.28) compared with controls (Table 2; Appendix Table 2).
Figure 2.
Invasive pneumococcal disease long-term mortality rates in adults, Alberta, Canada. Overall Kaplan-Meier survival estimates comparing case-patients with population controls. A) <30-day survival estimates; B) 30–90-day survival estimates. C) >90-day survival estimates.
We observed a major increase in risk for mortality rate in IPD case-patients in every age category for both male and female patients, as well as level of Elixhauser comorbidity category (Table 3). No interactions with age categories or Elixhauser scores were noted.
When stratified by year of IPD occurrence, we found that mortality rate differences decreased over time. These decreases were for case-patients >10 years ago (1999–2009) (aHR 4.66, 95% CI 3.75–5.78), for case-patients 5–10 years ago (2009–2014) (aHR 4.07, 95% CI 3.09–5.35), and for case-patients within the past 5 years (2014–2019) (aHR 2.88, 95% CI 2.33–3.56; p<0.001 for trend).
All-cause Mortality Rates at 30–90 Days
After removing IPD cases (and controls) who died or were censored within 30 days, we found that 3,888 case-patients (86.0%) and 8,459 (95.7%) controls remained. Compared with age- and sex-matched controls, for which 220 deaths (107 deaths/1,000 patient-years) occurred, we found that case-patients had 149 total deaths (158 deaths/1,000 patient-years) during 30–90 days after the diagnosis date (Figure 2, panel B). Both unadjusted and adjusted models demonstrated an increased risk for death for persons who had IPD (unadjusted HR [uHR] 1.49, 95% CI 1.21–1.83; unadjusted OR [uOR] 1.49, 95% CI 1.21–1.84; aHR 1.56, 95% CI 1.27–1.93; aOR 1.58, 95% CI 1.28–1.96) (Table 2; Appendix Table 2).
When we stratified patients by age group, we observed an increased risk for death only among those >75 years of age. We observed a major increased risk for death for male and female patients, as well as by level of Elixhauser comorbidity category (Table 3). However, few events occurred in patients who had only 0–1 comorbidities (n = 31, 1.2%). No major interactions with age or comorbidities were noted.
Case-patients entering the study >10 years ago had higher observed mortality rates during 30–90 days after diagnosis date than did controls (aHR 2.12, 95% CI 1.55–2.92). Case-patients identified during 5–10 years ago had an aHR of 1.40 (95% CI 0.84–2.33), and those identified <5 years ago had an aHR 1.19 (95% CI 0.85–1.66; p<0.001 for trend).
All-cause Mortality Rate After 90 Days
After removing IPD case-patients (and controls) who had an event within 90 days, we followed those who survived (and were not removed before 90 days) through March 31, 2019. At the end of this follow-up period, 1,174 case-patients (49 deaths/1,000 PYs) and 2,086 controls (37 deaths/1,000 PYs) died (Figure 2, panel C). Again, there was a major difference observed in mortality rates between case-patients and controls (uHR 1.32, 95% CI 1.23–1.42; uOR 1.35, 95% CI 1.24–1.47; aHR 1.43, 95% CI 1.33–1.54; aOR 1.49, 95% CI 1.36–1.64) (Table 2; Appendix Table 2).
Both groups that had 0–1 and ≥2 comorbidities showed higher mortality rates, as did every age group with the exception of patients >75 years of age. Again, female and male case-patients had similar increased risks when compared with controls. No interactions were noted. Cases identified >10 years ago had an aHR of 1.50 (95% CI 1.37–1.64), cases from 5–10 years ago had an aHR of 1.41 (95% CI 1.20–1.67), and cases from <5 years ago had an aHR of 1.26 (95% CI 1.06–1.51) (p<0.001 for trend).
All-cause Overall Mortality Rates
When observing the entire follow-up period of 20 years, we found that the median follow-up period was 3.9 (interquartile range 1.3–10.2) years. Overall, 1,937 case-patients died (81 deaths/1,000 PYs) compared with 2,654 controls (47 deaths/1,000 PYs) (Figure 2). By age, event rates in cases were as follows: <45 years, 1,324 case-patients (29.3%), 88 deaths (14.3%); 45–60 years, 1,388 case-patients (30.7%), 170 deaths (27.7%); 60–75 years, 1,054 case-patients (23.3%), 173 deaths (28.2%); and >75 years, 756 case-patients (16.7%), 183 deaths (29.8%). Unadjusted and adjusted models provided similar results: uHR 1.66 (95% CI 1.56–1.76), uOR 1.75 (95% CI 1.62–1.88), aHR 1.77 (95% CI 1.67–1.88), and aOR 1.97 (95% CI 1.81–2.14) (Table 2; Appendix Table 2).
Models stratified by age, sex, and comorbidity category showed that case-patients had increased mortality rates when compared with controls (p<0.01) (Table 3). Interaction models were tested, and none were noted. log-minus-log plots and Schoenfeld residuals were generated, and no violations were noted.
Case-patients identified >10 years ago (aHR 1.80, 95% CI 1.66–1.94]), 5–10 years ago (aHR 1.85, 95% CI 1.62–2.11]), and in the past 5 years (aHR 1.67, 95% CI 1.48–1.89) had similar estimates over time (p<0.01). However when compared with controls, we found that case-patients still had higher mortality rates (Figure 3; Appendix Table 3).
Figure 3.
Invasive pneumococcal disease (IPD) long-term mortality rates in adults, Alberta, Canada. aHRs describing illness risk comparing IPD cases versus controls after adjusting for age and Elixhauser comorbidity scores. Primary analysis: short (<30 days), intermediate (30–90 days), and long-term (>90 days) and overall (entire time period) follow-up. Secondary analysis: IPD cases and matched controls identified <5 years ago, 5–10 years ago, and >10 years ago. Error bars indicate 95% CIs. aHR, adjusted hazard ratio.
Sensitivity Analysis
In a sensitivity analysis, after excluding all IPD case-patients (and their controls) if no hospitalized controls were identified for the case-patient (n = 164, 3.6%), we found that our results were unchanged (<30 days aHR 3.71, 95% CI 3.24–4.24; 30–90 days aHR 1.43, 95% CI 1.16–1.78; >90 days aHR 1.37, 95% CI 1.27–1.47; overall mortality rate aHR 1.70, 95% CI 1.61–1.81). In addition, after we excluded IPD case-patients (and their matched controls) who had >1 IPD event (n = 142, 3.1%), we observed similar results (<30 days aHR 3.92, 95% CI 3.43–4.48; 30–90 days aHR 1.59, 95% CI 1.29–1.97; >90 days aHR 1.39, 95% CI 1.29–1.49; overall mortality rate aHR 1.75, 95% CI 1.65–1.86).
Emerging Infectious Diseases. 2022;28(8):1615-1623. © 2022 Centers for Disease Control and Prevention (CDC)
