High rates of bacterial coinfection in autopsy data from the 2009–2010 H1N1 influenza pandemic suggest synergies between influenza and pneumococcal disease in pandemic conditions. While influenza vaccination remains the primary tool to control the significant burden of influenza during both seasonal epidemics and pandemic seasons, pneumococcal vaccination also has the potential to mitigate the impact of pandemic influenza. During the 2009–2010 H1N1 pandemic, protection against the six additional serotypes in PCV13 compared to PCV7 likely would have prevented a portion of the pandemic-related pneumococcal cases, deaths, and costs.
In a typical influenza season, a PCV13 vaccination policy is expected to prevent an additional 13,400 IPD cases, 399,000 all-cause pneumonia cases, and 2,900 deaths, leading to cost savings of $472 million when compared to a PCV7 vaccination policy. We previously estimated that the PCV7 vaccination program in the US was cost saving for a normal influenza season, reducing pneumococcal-related costs by $1.6 billion per year; therefore the total cost savings of PCV13 versus a policy of no pneumococcal vaccination can be estimated at $2.4 billion per year. In a relatively mild pandemic similar to 2009–2010 H1N1, PCV13 is expected to prevent 22,800 IPD cases (1.7x normal season), 872,000 all-cause pneumonia cases (2.2x normal season), 3,700 deaths, and save $1.0 billion (2.2x normal season) compared to PCV7.
Our study is subject to a number of limitations inherent in the study design. First, the decision-analytic model is necessarily a highly simplified representation of the disease transmission and outcomes of pneumococcal disease. Although we accounted for some differences in treatment and outcomes using age stratification, we recognize that the U.S. population and healthcare delivery system are highly heterogeneous and may not be well represented by the relatively simple structure of this model. We also note that data used to estimate vaccine effectiveness and outcomes were derived and synthesized from a variety of sources, and this process of selection and interpretation is subject to bias. Although extensive sensitivity analyses to evaluate the effect of alternative parameter choices on our outcomes showed no change in the overall conclusions, we recognize that different assumptions may have yielded different results.
Costs used in this analysis were taken from published data and standard sources; the extent to which they reflect the true costs of administering medical care is unknown. Furthermore, this study was conducted from a third-party payer perspective rather than a societal perspective, and as such does not include costs of pneumococcal disease related to lost productivity, caregiver time, transportation, or other unreimbursed expenses. Inclusion of these costs presumably would have added substantially to the total cost burden of pneumococcal disease and the potential cost savings with PCV13. In addition, the model was estimated using U.S. data, and care should be used in generalizing our results to other settings and populations.
Because the data available to estimate pneumococcal coinfections were incomplete and based on autopsy data, a number of assumptions were necessary to convert these estimates of pneumococcal infection among influenza deaths to population estimates of coinfection. Although we attempted to be conservative in our assumptions, there remains a high level of uncertainty in these estimates. In particular, our calculations using obesity as a surrogate for all comorbidity is a simplification, and may not reflect the mechanism leading to excess deaths in groups such as those with asthma or immunosuppression, and pregnant women. Studies that examined obesity along with other comorbidities as a risk factor for poor outcomes in H1N1 have shown obesity to be the strongest and most consistent predictor. For example more than half of hospitalized H1N1 cases in California observed between April and August 2009 were obese and one-quarter morbidly obese , and obesity was identified as the strongest predictor of death in an analysis of hospitalized H1N1 cases in the United Kingdom (odds-ratio = 6.08; p = 0.01) . We further note, that Morgan and colleagues  estimated the odds-ratios of death for obese and morbidly obese persons versus normal weight persons among patients with chronic medical conditions that are risk factors for poor influenza outcomes (as identified by the Advisory Committee on Immunization Practices [ACIP]) and among patients with no chronic medical conditions. They found the odds of death to be significantly elevated only among obese and morbidly obese persons without chronic medical conditions; suggesting that the increased risk associated with comorbidity may be captured by the increase risk due to obesity in patient with both conditions.
It is also possible that our assumption of independence between obesity/comorbidity and bacterial coinfection as causes of death in persons with influenza is not valid. However, we note that there is considerable overlap in the comorbidities considered risk factors for both adverse pneumococcal disease outcomes and adverse influenza outcomes by ACIP [21, 47], suggesting that these conditions associated with poor influenza outcome also increase risk in the presence of pneumococcal infection. If there are synergies between comorbidity and coinfection, as suggested by a study in Malaysia , the rate of coinfection may be overestimated; however we found that even when we halved our estimate of the rate of coinfection, vaccination remained cost-saving in an assumed pandemic similar to 2009–2010 H1N1.