Discussion
We estimated the comprehensive burden of CAP and AOM across all age groups by combining cases from both inpatient and outpatient care settings using ICD-9, ICD-10-CA and physician billing coding in health administrative data. Our findings suggest that the burden of CAP in older adults and AOM in young children remains substantially high in two of the most populous provinces in Canada. There remains a sizeable burden of CAP and AOM attributable to bacterial pathogens, including pneumococcus in the context of PCV programmes during the study period even with underestimation of outpatient burden in Ontario. No consistent direct and indirect benefits of PCV13 programme was observed in both provinces.
Applying the overall age-adjusted and sex-adjusted incidence rates in 2018, an estimated 70 397 bacterial pneumonia, 974 hospitalised and 4049 non-hospitalised pneumococcal pneumonia, and 222 572 ASOM cases occurred in Ontario and BC. As expected, the majority of the CAP cases in both provinces were identified from outpatient settings. Only a handful of studies have reported the incidence of CAP in outpatient settings. An annual incidence of 293–306 per 100 000 population was reported for all-cause CAP during 2005–2009 in individuals aged ≥15 years in a study conducted in Italian general practices.6 Another study reported an annual incidence of 470 per 100 000 population for all-cause outpatient CAP in French adults aged >18 years in 2011–2012.7 Our estimated annual rates of hospitalised and outpatient CAP for all age groups for the corresponding years are much higher than these reported rates. A previous systematic review estimated that approximately 27% of hospitalised adult CAP were attributable to pneumococcus.19 Pneumococcus was identified in approximately 25% of AOM cases in Japanese children.33 We observed a decline in the annual incidence rates of CAP, bacterial CAP and hospitalised pneumococcal pneumonia during pre-PCV13 years in both provinces, which likely resulted from the earlier use PCV7 or PCV10. Contrary to our expectation that PCV13 would have impacted the incidence rates of CAP and bacterial CAP, we observed relatively stable rates in Ontario and increasing rates of bacterial CAP in BC during our study PCV13 period. Hospitalised pneumococcal pneumonia seemed to decline slightly, particularly in older adults aged ≥65 years in both provinces, while non-hospitalised pneumococcal pneumonia increased specifically in BC. A substantial reduction in the incidence of hospitalised pneumococcal pneumonia, pneumonia of unspecified causative organism, and empyema was observed after PCV13 implementation among children aged <2 years in the UK; hospitalised pneumococcal pneumonia was also reduced in individuals aged 2–44 years, but increased in individuals aged ≥45 years.34 A reduction in hospitalised non-invasive pneumococcal CAP was observed only in infants during the post-PCV13 period compared with the pre-PCV13 period (6.2/100 000 vs 8.1/100 000) in Israel.35 The incidence of hospitalised all-cause pneumonia declined in Canadian children aged 0–17 years and older adults aged ≥65 years during early PCV13 years (2010–214)36; the incidence of pneumococcal pneumonia among hospitalised Canadian adults was also reported to have declined during early PCV13 years before increasing again in 2015.5
Our study identified several differences with respect to age-specific trends for CAP. We observed a very small decline in the cumulative incidence of bacterial CAP and hospitalised pneumococcal pneumonia from pre-PCV13 to PCV13 period in children aged <18 years in Ontario; furthermore, the incidence rate of bacterial CAP in BC increased substantially during PCV13 period compared with pre-PCV13 period, and the incidence rate of non-hospitalised pneumococcal pneumonia increased in children aged 0–4 years during PCV13 period. In contrast, the incidence of inpatient and outpatient non-invasive pneumococcal pneumonia using administrative database in US children aged <18 years decreased considerably during our study period from 2006 to 2009 (late PCV7) to 2014–2018 (late PCV13): from 102.0 to 32.2 per 100 000 patient-years in children <2 years, from 83.9 to 30.8 per patient-years in children aged 2–4 years, and from 34.1 to 12.5 per 100 000 patient years in children aged 5–17 years.37 In our study, the burden of bacterial CAP and pneumococcal pneumonia continued to increase with age in older adults aged ≥65 years even after introduction of PCV13 in children and publicly funded PPV23 in older adults. This phenomenon has also been observed in US-based studies evaluating hospitalised CAP in adults.38 We would expect influenza vaccination to reduce some of the burden of primary CAP caused by influenza virus and prevent secondary pneumonia caused by pneumococcus and other bacteria during vaccine-matched influenza seasons, particularly in older adults.39 For example, during 2016–2017 influenza season, 24% of children, 37% of adults aged 18–64 years with chronic medical conditions, and 69% of older adults aged ≥65 years were reported to have received influenza vaccine in Canada.40 The expected benefits of influenza vaccination on CAP and pneumococcal pneumonia may have been offset by suboptimal influenza vaccine coverage and the observed decline in vaccine effectiveness with influenza vaccination programme maturation.41 Nevertheless, the absence of aetiology data limits us from fully understanding the observed time trends of CAP, including bacterial CAP and pneumococcal pneumonia burden in our study.
We could not directly measure the impact of vaccination on AOM cases; however, we observed a decline in the incidence of AOM and ASOM from the pre-PCV13 period to the PCV13 period across all age groups in both provinces. This could have resulted from a reduction in the prevalence of pneumococcus as the causative organism of AOM in children with the introduction of PCV13 as reported previously,42 and reflects the direct benefit of PCV13 vaccine in children along with indirect and/or direct benefit in adults. However, a small reduction in AOM could also have resulted from influenza vaccination in children.43 The incidence of AOM in US children also declined from 1998 to 2018 using claims data from commercial plans (1111–727/1000 person-years in children aged <2 years, 517–400/1000 person-years in children aged 2–4 years, and 112–87/1000 person-years in children aged 5–17 years) and Medicaid (895–656/1000 person-years in children aged <2 years, 385–329/1000 person-years in children aged 2–4 years, and 98–87/1000 person-years in children aged 5–17 years).44 A lack of contemporaneous data on the burden of AOM in adults precluded comparison of our rates.
Substantial benefits of the childhood PCV have been observed in the USA in preventing healthcare visits for otitis media and hospitalisations for pneumonia in children.45 But more recently, partial replacement with non-vaccine serotype IPD across all age groups and an increase in non-PCV13 IPD hospitalisation in children aged <5 years have been reported in some high-income countries, including Canada.46–48 This could also apply to non-IPDs, such as CAP; the lack of considerable benefit of pneumococcal vaccines observed in our study may have resulted from pneumococcal serotype replacement with non-vaccine serotypes. Indeed, non-vaccine serotype CAPs have been reported in Canadian hospitalised adult patients with pneumococcal pneumonia, and these cases would not be prevented by PCV13 or PPV23; CAP cases caused by 22F and 33F serotypes were also reported.5 An increase in the incidence of hospitalised pneumococcal CAP, primarily due to non-vaccine serotypes and serotype 3 has also been reported in individuals aged ≥16 years in the UK.49 Similar increase in the incidence of hospitalised pneumococcal CAP with serotype 3 was observed in Canadians aged ≥16 years, which partly offset the indirect benefit of PCV13 conferred against 7F and 19A in adults.5 50 Pathogen-specific (eg, capsular polysaccharide and invasiveness) and host-related (eg, immunotolerance) factors together with a lower PCV13 vaccine effectiveness against serotype 3 with the current vaccine schedules and doses in children may explain the persistence of serotype 3 in adults.51 Nevertheless, limited evidence suggests 52.5% vaccine effectiveness of PCV13 against serotype 3 hospitalised CAP in older adults aged ≥65 years.52
Another reason for the increase we observed in CAP rates may be related to pneumococcal vaccine coverage and the differences in the vaccine schedule between Canada and the USA. Pneumococcal vaccination coverage remains suboptimal in both children and adults in Canada. Approximately 80% of Canadian children were reported to have received PCV by 2 years of age during 2013–2017,53 which is below the national coverage goal of 95%.54 Canada changed the PCV schedule to 2+1 back in 2010 while the USA had maintained the original 3+1 schedule. Studies have shown a considerable number of children in Canada are only partially vaccinated with the PCV vaccines and/or delayed their vaccination.55–57 It is possible that incomplete receipt of PCV vaccine is impacting the indirect benefits seen in adults in Canada compared with the USA. Additionally, PPV23 vaccination coverage in older adults also remained at 36%–58% during 2006–2019 in Canada, which is much lower than the national target coverage of 80%, and uptake of pneumococcal vaccine in younger adults (18–64 years) with chronic medical condition was estimated at 12%–25%.40 58 PPV23 has lower effectiveness against CAP hospitalisation (10%) and pneumococcal CAP hospitalisation (32%–51%) in older adults59 and minimum effect on all-cause pneumonia and pneumococcal pneumonia in adults.60 The US Advisory committee on Immunisation Practices (ACIP) recommended routine use of PCV13 in series with PPV23 for all adults aged ≥65 years in 2014 considering an estimated 20%–25% of IPD and 10% of CAP cases in US adults aged ≥65 years were caused by PCV13 serotypes that could be prevented in view of demonstrated efficacy of PCV13 against PCV13-type IPD and pneumonia in this population.61 This recommendation was made to achieve additional reductions in disease burden in adults aged ≥65 through the ongoing indirect effects of paediatric PCV13 programme and the direct effects of PCV13 in adults aged ≥65. However, the direct and indirect effects resulted in minimum changes in population-level incidence of pneumococcal disease in adults following implementation of this recommendation.62 Consequently, in 2018, the ACIP changed the recommendation to a routine single dose of PPV23 for all adults aged ≥65 years, and recommended shared clinical decision-making for use of PCV13 in adults aged ≥65 years without any immunocompromised condition, cerebrospinal fluid leak, or cochlear implant and who have not received PCV13 previously while recognising that some adults aged ≥65 years are at increased risk of exposure to PCV13 and would benefit from PCV13. In Canada, the NACI recommended the use of PCV13 in immunocompetent adults aged ≥65 years not previously immunised against pneumococcal disease who desire additional protection against PCV13 serotypes in addition to the routinely recommended PPV23 for all adults aged ≥65 years.63 NACI does not recommend inclusion of PCV13 in publicly funded immunisation programme for adults aged ≥65 years.64 Considering the low level of protection conferred by PPV23 vaccines and the observed persistent high burden of CAP and pneumococcal pneumonia in older adults during PCV13 period in our study, it may be worthwhile to consider the higher valent PCV vaccines for older adults and evaluating the effectiveness of PCV in this population. Newer generation, higher valent PCV vaccines, such as PCV15 or PCV20 that would provide extended protection against non-PCV13 or non-PPV23 serotypes causing CAP could maximise the potential direct and indirect benefits from polysaccharide vaccines and reduce CAP and pneumococcal pneumonia burden.
One key strength of our study is that we used data from multiple points of contact with the healthcare system (eg, hospitalisation, ambulatory and emergency care, and primary care) that enabled us to better approximate the overall healthcare burden of CAP and AOM by not focusing on serious episodes requiring hospitalisation. Our estimated CAP and AOM burden that could be potentially attributable to bacteria, including pneumococcus demonstrates the burden that could be preventable by the PCV programmes.
Our study had a number of limitations. One major limitation of our study is the inability to establish aetiology-specific pneumonia in primary care settings. This precluded identification and quantification of pneumococcal pneumonia cases among all-cause CAP that were treated in outpatient settings alone and consequently led to considerable underestimation of the burden of bacterial CAP and pneumococcal pneumonia. In the absence of routine microbiological testing to identify aetiological agents with subjectivity in clinical diagnosis and use of diagnostic codes in primary care settings, there remains uncertainty and some degree of imprecision in estimating the burden of CAP and AOM attributable to bacterial pathogen, including pneumococcus using diagnostic codes. There are limited data on the validity of pneumonia and AOM diagnostic codes used in primary care settings. However, previous studies in Canada have reported that in paediatric and adult populations, administrative billing codes for pneumonia and otitis media used in our study were reasonably accurate compared with medical records.65–67 Other studies have also validated codes in hospitalised, and ED visit paediatric pneumonia cases.68 69 Diagnostic codes similar to ours have been used to estimate the incidence of all-cause CAP hospitalisation in Canada previously.4 36 The incidence of paediatric AOM has also been estimated using administrative billing codes in the USA.44 As such, it is unlikely that our burden estimates are substantially overestimated because of false positive cases. Diagnostic and laboratory practices to identify a patient with CAP or AOM may vary among physicians in Ontario and BC. Our study used administrative databases and relied on the use of diagnostic codes. As a result, there is the potential for misdiagnosis, miscoding or misclassification of CAP and AOM, resulting in underestimation or overestimation of the burden.70 71 The differences in incidence between Ontario and BC (higher incidence of CAP and AOM in Ontario, while higher incidence of bacterial CAP, non-hospitalised pneumococcal pneumonia and ASOM in BC) most likely resulted from differences in case ascertainment/diagnosis and/or coding practices, although there may also have been local variability in viral and bacterial pathogen occurrence, including pneumococcal serotype replacement, and/or variable distribution of risk factors, particularly in adults. However, we would expect within province comparisons and trends over time to be less affected by these factors. Most of the bacterial CAP and pneumococcal pneumonia cases in BC were identified from the primary care settings and BC had much higher burden of bacterial CAP and non-hospitalised pneumococcal pneumonia than Ontario. A lack of general availability of billing codes for aetiology-specific pneumonia in primary care settings in Ontario likely resulted in the observed considerable underestimation of bacterial CAP and non-hospitalised pneumococcal pneumonia in Ontario. Furthermore, bacterial CAP and non-hospitalised pneumococcal pneumonia burden in BC also has the potential for overestimation in absence of microbiological confirmation in primary care settings. Although we included CAP and/or AOM cases from the primary care settings in both provinces, we also included cases treated at the ED from Ontario; the lack of ED visit data from BC may have led to an underestimation of CAP burden in BC.