Invasive Pneumococcal Disease and Pandemic (H1N1) 2009, Denver, Colorado, USA

Pneumococcal prevention strategies should be emphasized during future influenza pandemics.


P neumonia caused by Streptococcus pneumoniae
(pneumococci) was a frequent complication of infl uenza during previous pandemics. In 1 autopsy series, ≈20% of deaths during the 1918 infl uenza pandemic were associated with pneumococci (1). Pandemic (H1N1) 2009 was the fi rst pandemic in which pneumococcal and infl uenza vaccines and antiviral drug treatment had the potential to change the interaction between pneumococcal infection and infl uenza.
Among early cases of pandemic (H1N1) 2009, pneumococcal complications were rarely reported (2)(3)(4). However, in October 2009, the Colorado Department of Public Health and Environment identifi ed a substantial increase in cases of invasive pneumococcal disease (IPD) in the Denver metropolitan area, concurrent with a peak in pandemic (H1N1) 2009-associated hospitalizations, raising the question of the role of this pandemic (H1N1) 2009 virus. We evaluated the IPD cases in October 2009 in terms of age, prevalence of concurrent conditions, severity of illness, evidence of co-infection with pandemic (H1N1) 2009 virus, use of antiviral drugs, and infl uenza and pneumococcal vaccination. We also assessed the possible contribution of changes in laboratory practices to the increase in reported IPD cases.

Epidemiologic Investigation
Classifi cation as IPD required isolation of S. pneumoniae from a normally sterile site, such as blood or cerebrospinal fl uid. IPD cases were identifi ed through the Active Bacterial Core surveillance (ABCs), a populationand laboratory-based system run continuously since 2000 in the 5-county Denver metropolitan area (population 2.4 million). To evaluate the magnitude of the apparent increase in IPD cases during October 2009, we compared the number of cases during that month to the mean number of cases occurring each October during 2004-2008 because the decrease in IPD rates that followed introduction of the 7-valent pneumococcal conjugate vaccine (PCV7) for US infants stabilized by 2004 (5).
To evaluate if IPD cases in October 2009 were epidemiologically different from IPD cases in previous, nonpandemic years, we compared the October 2009 cases to IPD cases in February 2009, the most recent local peak in seasonal infl uenza (H1N1). If the October 2009 cases were associated with pandemic (H1N1) 2009 and if this infl uenza affected IPD risk differently than seasonal infl uenza, we would expect the epidemiology of October 2009 IPD cases to differ from a month with predominant nonpandemic, seasonal infl uenza (H1N1) circulation. Data analyzed included epidemiologic information from the standard ABCs case report form (www.cdc.gov/abcs/ fi les/ABCs_case_report_form_2009.pdf). More detailed information on initial symptoms, diagnostic testing, clinical laboratory information, and clinical management was collected by chart abstraction.
Chart abstraction included review of physician notes, consultation reports, and laboratory results included in patients' medical records. For October 2009 cases, supplementary information was obtained from interviews with patients or their surrogates. For comparisons of underlying conditions and basic demographics, we ensured consistent methods across time by comparing data derived only from the ABCs case report form. Most October casepatients had pneumonia; for these, severity of illness was assessed by using the Pneumonia Severity Index, a wellestablished scoring system that incorporates concurrent conditions, laboratory fi ndings, and vital signs at clinical presentation (6). Additional data available for only October 2009 cases included vaccination, antiviral drug treatment, serologic test results, and intensive care unit (ICU) admission.
Infl uenza-associated hospitalizations were identifi ed by using the Colorado Emerging Infections Program. This program defi nes laboratory-confi rmed infl uenza infection as any positive rapid test or pandemic (H1N1) 2009 virusspecifi c real-time PCR result in a hospitalized resident of the surveillance area.
To identify the upper limit of potential pandemic (H1N1) 2009 cases among the October 2009 IPD cases, we used 2 approaches. First, because recent data (7,8) suggest that the sensitivity of PCR for pandemic (H1N1) 2009 virus decreases 5 days after symptom onset, patients with infl uenza-like illness (ILI) (fever plus cough or sore throat) and negative or unknown PCR results >5 days before the date of pneumococcal culture were considered to be potentially associated with pandemic (H1N1) 2009. Second, all IPD cases in excess of 2 SD above the mean number of IPD cases in October during 2004-2008 were defi ned as cases that may have been associated with pandemic (H1N1) 2009. Health care providers and the Colorado Immunization Information System were contacted to verify vaccination status for PCV7, pneumococcal polysaccharide vaccine (PPV23), and seasonal infl uenza and pandemic (H1N1) 2009 vaccines.
Infl uenza diagnostic methods investigated were realtime reverse transcriptase PCR, rapid infl uenza test, direct or indirect fl uorescent antibody, serologic analysis, and viral culture. Test results were separated by virus identifi cation (A, B, or both) and infl uenza A subtype (H1, H3, pandemic [H1N1] 2009, unsubtypeable, or other). We surveyed 13 clinical laboratories serving the 16 reporting hospitals to assess total numbers of IPD cases and total positive blood cultures in September-October 2009 and September-October 2008, and to identify if there had been any changes in laboratory procedures related to blood culturing.

Statistical Analysis
Data were analyzed with SAS version 9.2 (SAS Institute, Cary, NC, USA). The χ 2 test was used to compare proportions. Medians were compared by using the Wilcoxon ranked-sum test; p values <0.05 were considered signifi cant.

Descriptive Epidemiology
Fifty-eight cases of IPD were identifi ed in the Denver Metropolitan Area during October 2009, which was >3× the October average (mean ± SD 18.4 ± 4.7) during 2004-2008 ( Figure). Cases were reported from 16 of 20 Denver area acute care hospitals; these hospitals were distributed throughout the 5-county area. Forty-fi ve cases occurred in February 2009.
Medical records were abstracted for all 58 casespatients with IPD (Table 1). Compared with February case-patients, October case-patients were younger (median age 45 years vs. 54 years; p = 0.02), and the proportion of nonelderly case-patients (age <60 years) was higher (45 [78%] of 58 vs. 26 [58%] of 45; p = 0.03). After adjusting for different age distributions of the 2 groups, we found that October 2009 case-patients were more likely to have concurrent conditions, specifi cally, chronic lung disease ( Table 1).

Severity of Illness
Fifty-four (93%) of 58 IPD case-patients were hospitalized, and 47 (81%) had pneumonia. Seven casepatients died (12%), and 19 (33%) were admitted to the ICU. All patients who died were >40 years of age (Table  2). There were no major differences between the October 2009 and February 2009 case-patients in the proportion hospitalized or the case-fatality rate. A high (6/7, 86%) proportion of IPD case-patients 20-39 years of age who were hospitalized in October 2009 were admitted to the ICU. All 42 October pneumonia patients had Pneumonia Severity Index scores >2; the 7 case-patients who died all had pneumonia and scores of 4 or 5, which indicated severe illness and recommended hospitalization.
Of the 10 positive case-patients, 2 were given a diagnosis before admission (1 by PCR and 1 by PCR and rapid test) and 8 were given a diagnosis at admission (all patients had both a rapid test and PCR). For those patients given a diagnosis at admission, only 2 (25%) of the rapid test results were positive and all 8 PCR test results were positive. Of the 7 patients for whom infl uenza A virus subtype analysis was conducted, 6 had pandemic (H1N1) 2009 virus; the other virus was not subtypeable.
Fifty-three of 58 IPD case-patients had suffi cient information to evaluate for preceding ILI. Of these casepatients, 33 (62%) reported symptoms consistent with ILI >5 days before date of the culture that yielded pneumococci. All 10 infl uenza-positive case-patients reported ILI symptoms. Overall, 31 (94%) of the 33 case-patients with ILI were tested for infl uenza virus.
On the basis of IPD surveillance in Denver during 2004-2008, the maximum expected number of IPD cases during October of nonpandemic years was 28 (mean ± 2 SD 18 ± 5). If all remaining 30 IPD cases were considered to be excess cases associated with pandemic (H1N1) 2009, then 52% of all cases of IPD in October 2009 may have been associated with pandemic (H1N1) 2009.

Role of Vaccination
All 9 children had been vaccinated with PCV7 at appropriate ages. Of 38 adults 18-64 years of age, 29 (76%) had indications for PPV23 vaccination and 25 (86%) of them had available vaccination records; of these persons, 3 (12%) were vaccinated. The proportion of persons >65 years of age (universal PPV23 vaccination recommendation) with available vaccination records who were vaccinated was 67% (4/6).
Thirty-one (53%) of 58 IPD case-patients received antiviral medications during their illnesses; 20 (61%) of 33 patients who reported ILI symptoms received antiviral drugs and 9 (90%) of the 10 patients who had a positive infl uenza test result received antiviral drugs. According to medical records, the most common reason for not prescribing antiviral drugs was the duration of time since symptom onset. The 1 patient who had confi rmed pandemic (H1N1) 2009-associated IPD who died received antiviral drug treatment 72 hours after admission and 48 hours after confi rmatory infl uenza test results.

Serotypes of Pneumococci Causing IPD
Among the 47 (81%) cases with available isolates, 75% were caused by serotypes included in PPV23 and 4% by those included in PCV7. Serotype 7F was the predominant serotype identifi ed, and it accounted for 34% of the October cases. Serotype distribution was consistent with overall epidemiology of IPD in Colorado during nonpandemic periods. After adjusting for age, we found that the proportion of all case-patients in Denver with IPD caused by serotype 7F during 2004-2008 increased from 3% to 25%. This increase was observed when analyzing only October (0%-34%; p = 0.025) and all other months (2%-23%; p<0.0001). Among the 9 adult infl uenzapositive case-patients, 7 of the isolates were serotypes contained in PPV23. Five (71%) of 7 persons with an indication for PPV23 had a serotype covered by PPV23, and only 1 (20%) had received PPV23.

Laboratory Survey
Among 13 surveyed laboratories, 4 implemented changes in blood culture practices over the preceding year. Two hospitals increased the number of times that blood cultures were evaluated for growth, 1 hospital adopted a different automated culture system, and 1 hospital began using a different skin antiseptic. Eleven laboratories saw a higher number of blood cultures submitted during September-October 2009 than in September-October 2008 (overall increase of 16%). The proportion of cultures positive for pneumococci increased at 9 laboratories by an average of >4 additional isolates, which was a relative increase of 50% over the previous year (Table 3).

Discussion
An investigation of IPD cases in Denver during October 2009 showed 3× the average number of IPD cases identifi ed during October of the previous 5 years and a notable association with pandemic (H1N1) 2009 virus infection. Our fi ndings do not prove a causal relationship between pandemic (H1N1) 2009 and IPD. However, we confi rmed that 10 (17%) of 58 case-patients had infl uenza, and 2 estimates of the maximum proportion of IPD cases that may have been associated with pandemic (H1N1) 2009 showed that proportion was 52%-62%. Increases in testing for pneumococcal infection were modest and could not account for the magnitude of the increase in IPD incidence that we observed. However, the reported increase by area clinical laboratories in blood culture positivity for pneumococci supports a true increase in IPD incidence.
This investigation highlighted factors that may be distinct to IPD cases associated with pandemic (H1N1) 2009. Previous infl uenza pandemics have implicated secondary bacterial infection as a complication and cause of serious illness and death (1). These studies were based largely on autopsy series and histologic confi rmation, but were limited in their ability to evaluate clinical presentation, symptoms, and onset that may be distinct to IPD cases identifi ed during nonpandemic infl uenza seasons. During the 2009-10 infl uenza season, the predominant circulating infl uenza virus was pandemic (H1N1) 2009 virus (11). IPD cases mirrored the epidemiology of pandemic (H1N1) 2009, peaked at the same time, and affected younger persons. Because some cases of IPD occur every October and because some cases of IPD are likely attributable to seasonal infl uenza (12), we compared cases detected during October 2009 with IPD cases seen during a peak month of seasonal infl uenza activity. Attack rates for pandemic (H1N1) 2009 were likely higher than those for seasonal infl uenza (H1N1) (13 domestic and international sites (4,15,16) suggested that pandemic (H1N1) 2009-associated IPD was not unique to Denver. However, there were no widespread levels of IPD that were greater than expected. At the time of our investigation in Denver, whether increases in IPD in other ABCs sites were statistically or epidemiologically signifi cant was not clear. Since that time, increases have become apparent in other sites, although these increases were consistently more modest than those observed in Denver and were not as thoroughly investigated. In contrast to the previous 5 years, Denver experienced a peak in IPD in October 2009, which was likely attributable to pandemic (H1N1) 2009, and a second peak in December, which likely represented endemic disease. Prevention of IPD during future infl uenza pandemics should focus on vaccination and prompt diagnosis. This infl uenza pandemic was the fi rst in which pneumococcal vaccines and antiviral drug treatment were available. Among adults with IPD in Denver, vaccination rates for persons 18-64 years of age with indications for vaccination were less than the national rate, and vaccination rates for persons >65 years of age were similar to national estimates (17).
The increase in IPD cases during October, the peak month of hospitalizations of persons with pandemic (H1N1) 2009, might have been minimized if adults at highest risk for IPD had received the recommended polysaccharide vaccine. Increasing PPV23 coverage in populations with increased risk for IPD is a key prevention measure, especially in anticipation of infl uenza pandemics.
Introduction of PCV7 into routine childhood immunization programs in the United States resulted in dramatic reductions in rates of pneumococcal-related diseases and major changes in the epidemiology for all age groups (5,(18)(19)(20)(21)(22)(23). In Denver during October 2009, we identifi ed only 2 cases of IPD caused by serotypes included in PCV7, both in adults.
Whether vaccine against pandemic (H1N1 2009, which became available in Denver during late October, could have reduced the number of pandemic (H1N1) 2009-associated IPD cases is unknown. Antiviral drug administration was sometimes delayed or withheld despite national guidance for treatment even if >48 hours had elapsed from onset of illness (24), and such withholding may have changed the clinical course of some of the IPD cases. Outbreaks of S. pneumoniae have occurred in many settings (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39), and individual serotypes have been implicated in localized outbreaks (27)(28)(29)31,39). The variety of serotypes identifi ed in this outbreak indicates that the increase in IPD was not attributable to enhanced transmission of a single serotype. To address whether the increase in October 2009 refl ected a clonal outbreak of 7F, we analyzed the proportion of IPD cases in Denver that were serotype 7F during 2004-2010. During 2004-2009, the proportion of 7F increased (from 3% to 25%). When we evaluated all months except October during 2004-2009, the proportion of 7F still increased (from 2% to 23%). The proportion of IPD caused by 7F has been increasing in Denver over time and cannot be attributed to an increase in October 2009 alone or the pandemic. Furthermore, the distribution of serotypes was similar to serotype distributions in national (5) and Denver-specifi c IPD cases, which suggested that if pandemic (H1N1) 2009, was causally associated with this outbreak, it facilitated pneumococcal infection without a predilection for any particular serotype.
Our investigation had limitations. Low numbers of cases may have limited our ability to identify differences in the epidemiology of IPD during October 2009 and peak months of seasonal infl uenza activity. We were also unable to ascertain PPV23 vaccination histories for all cases, which may have underestimated PPV23 use. Of 47 infl uenza tests ordered, 9 (19%) were only rapid tests. The sensitivity of rapid tests for detecting pandemic (H1N1) 2009 ranged from 20% to 40% (40). Twenty-four (41%) of 58 IPD casepatients were not tested for pandemic (H1N1) 2009 by PCR (the standard for detection), which may have underestimated the number of confi rmed infl uenza-associated IPD cases. Some patients with negative test results may have been infected with infl uenza virus but were tested too late in the course of their illness. Finally, ILI does not capture all infl uenza cases and cases with infl uenza within 5 days of pneumococcal culture and not tested samples would not be included for a possible infl uenza-associated IPD case. In addition, ILI includes symptoms that occur frequently with signs and symptoms of pneumococcal pneumonia and may be a result of the symptom course of IPD rather than preceding infl uenza infection.
In conclusion, up to two thirds of IPD cases in Denver during October 2009 may have been associated with pandemic (H1N1) 2009. Pandemic infl uenza may have altered the epidemiology of IPD and shifted the age distribution to younger persons and to persons 18-64 years of age with an increased prevalence of underlying conditions. Missed opportunities for PPV23 vaccination were common. During future infl uenza pandemics, public health offi cials should increase awareness of the association between IPD and infl uenza among persons of greatest risk for infl uenza-associated IPD. Prevention efforts should include use of pneumococcal vaccines and vaccines for directly preventing infl uenza infection.