Increasing Incidence of Invasive Haemophilus influenzae Disease in Adults, Utah, USA

TOC Summary: The infection disproportionately affected patients >65 years of age.

B efore the introduction of the Haemophilus infl uenzae type b (Hib) conjugate vaccine, Hib was the most common cause of bacterial meningitis in children <5 years of age (1,2). Hib caused >80% of invasive H. infl uenzae disease among children (3). Since the introduction of the Hib conjugate vaccine, the incidence of invasive Hib disease in children has decreased by 99% (1,2). Several studies have reported a decrease in the incidence of invasive Hib in adults following widespread use of vaccine in children (4,5).
However, although the decrease in the overall incidence of invasive disease caused by Hib for children and adults has been a remarkable public health success, changes in the epidemiology of invasive disease caused by other H. infl uenzae have been observed. The incidence of H. infl uenzae invasive disease has shifted toward adults. In a study conducted in the Atlanta, Georgia, USA, metropolitan area before Hib vaccine introduction, adults comprised 24% of all invasive H. infl uenzae cases (6). A more recent population-based report from Illinois showed adults accounting for 77% of invasive cases (7). Other encapsulated strains, including H. infl uenzae type a (Hia), have emerged as notable causes of invasive disease among children in Utah, Alaska, and the Navajo nation (8)(9)(10). Given the changing epidemiology observed elsewhere, we sought to describe the epidemiology of invasive H. infl uenzae infections among Utah adults during 1998 through 2008.

Study Design and Population
We identifi ed all cases of laboratory-confi rmed invasive H. infl uenzae among Utah residents >18 years of age reported to the Utah Department of Health (UDoH) from 1998 through 2008. Invasive H. infl uenzae is an immediately reportable condition in Utah. We defi ned invasive H. infl uenzae according to the Council of State and Territorial Epidemiologists defi nition: a clinically compatible case that is confi rmed by isolation of H. infl uenzae from a normally sterile site such as blood, cerebrospinal fl uid, joint fl uid, pleural fl uid, or pericardial fl uid (11). The UDoH enhanced their passive surveillance program in 2005 by sending out reminders to major laboratories in Utah to report invasive H. infl uenzae cases. As part of these enhancement efforts, 1 large laboratory also began automated case reporting to the UDoH in 2005. Because these changes in surveillance could introduce ascertainment bias for the fi nal 4 years of this analysis, we conducted a parallel retrospective cohort analysis of laboratory-confi rmed invasive H. infl uenzae cases among Utah residents within the Intermountain Healthcare (IH) system during 1999-2008 (on the basis of availability of complete laboratory data). IH is a vertically integrated hospital system that provides care to roughly 50% of Utah adults (Scott Lloyd, pers. comm.). The estimated market share by age group is recalculated at regular intervals. Over this period, IH made no major changes to its microbiology practices that would alter the recovery of H. infl uenza isolates.

Serotyping
H. infl uenzae isolated from sterile sites was identifi ed by using standard culture techniques. Laboratories are requested to submit sterile site isolates of H. infl uenzae to the Utah Public Health Laboratory in Salt Lake City, Utah, for defi nitive serotyping by using monovalent antisera for H. infl uenzae types a-f. Isolates from reported H. infl uenzae cases not sent to the Utah Public Health Laboratory for serotyping were classifi ed as not typed. All other isolates were classifi ed as Hia-f or as nontypeable.

Demographic and Outcome Data
As part of its surveillance program, the UDoH collected demographic information including county of residence, race/ethnicity, age, gender, hospitalization, and death. Starting in 2005, the UDoH began to collect expanded clinical information including the presence of chronic medical conditions and clinical signs. These clinical data were collected at the local health department level. It should be noted that not all local health departments conducted such data collection, and collection methods varied across local health departments (e.g., chart review, patient interview, hospital information personnel interview).

Statistical Analysis
To calculate the incidence rate, we used the estimated annual Utah population among persons >18 years of age on the basis of the United States 2000 census (12). To derive an incidence for the IH cohort, we multiplied the estimated annual population by IH estimated market share of adult healthcare in Utah for a given year. Market share data were only available for 2000, 2002, and 2005-2008. For intervening years, we extrapolated using available data for the nearest years.
Categorical variables were evaluated by using standard χ 2 test for independence or Fisher exact test as appropriate. To evaluate associations between variables, we used ordinary logistic regression and exact logistic regression as appropriate. Poisson regression method with a log-linear model was performed to test for trend variation in the annual case rates for the IHC and UDoH datasets. We used SAS version 9.2 (SAS Institute Inc., Cary, NC, USA) for all statistical analyses.

Human Subjects Protection
We obtained approval for this study from the Institutional Review Boards of the University of Utah and IH; de-identifi ed case information was provided by the UDoH through a data sharing agreement. Informed consent was waived as all data were de-identifi ed.

Results
We identifi ed 121 cases of invasive H. infl uenzae in Utah adults reported to the UDoH during 1998-2008 (Table). The incidence of H. infl uenzae invasive disease was 0.66 cases per 100,000 person-years averaged over the 11 years. Incidence increased over the study period from 0.14 cases per 100,000 person-years in 1998 to a peak in 2008 of 1.61 cases per 100,000 person-years (p = 0.0023, by linear test for trend) ( Figure 1). We observed a marked seasonality with winter predominance: 41.3% of cases occurring in winter, 26.4% in spring, 21.4% in fall, and 10.7% in summer. We observed no signifi cant difference in distribution between sexes, with 64 cases in women and 57 cases in men (p = 0.65).
The incidence rates estimated by using the IH data showed a similar trend ( Figure 1). We identifi ed 94 cases of invasive H. infl uenzae, with an estimated average annual incidence of 1.05/100,000 person-years. The incidence increased from 0.54/100,000 person-years in 1999 to 1.27/100,000 person-years in 2008 (p = 0.002, by linear test for trend). Although the estimated rates differed, the trend in increasing incidence was similar for both data sources.
The incidence of invasive disease was markedly higher among persons >65 years of age; the average incidence in this age group was 2.74 cases per 100,000 person-years and increased to 6.14 cases per 100,000 in 2008. The average incidence for persons 18-34, 35-49, and 50-64 years of age were 0.25, 0.26, and 0.88 per 100,000 person-years, respectively. Compared with persons 18-34 years of age, patients >65 years of age had a 12-fold greater risk for invasive disease (risk ratio 12.5, 95% confi dence interval [CI] 7.29-21.32; p<0.0001). The incidence of invasive disease increased signifi cantly over the study period among those >65 years of age ( Figure 2).
The increases in invasive disease were largely because of nontypeable strains and Hif. The average incidence over 11 years of invasive disease caused by nontypeable strains was 0.23/ 100,000 person-years and rose from 0 in 1998 and 1999 to an average of 0.72/100,000 person-years during 2007 and 2008. During 1998-2008, the average incidence of disease caused by Hif was 0.14/100,000 person-years and rose to an average of 0.48/100,000 person-years during 2007 and 2008. The number of cases of Hia remained relatively steady, with an average annual incidence of 0.08/100,000 person-years.

Clinical Disease
Most (51%) invasive disease occurred in persons >65 years of age. Bacteremia accounted for 83% of all cases, meningitis for 12.4%, and other sterile sites for the remaining 4.6%. Nontypeable strains, Hib, and Hif accounted for 13 of the 15 cases of meningitis (Table). Infection with Hib was signifi cantly more likely to be associated with meningitis than nontypeable strains (odds ratio [OR] 6.1, 95% CI 1.3-29.0). While invasive disease was more likely to occur in the elderly, a greater proportion of disease among patients 18-49 years of age was due to meningitis compared with patients >50 years of age (25% vs. 9%; p = 0.05).
Data on underlying illness was available for 69 of the 121 case-patients. Forty-six (67%) of 69 patients had chronic medical conditions; 33% had >2. Twenty-four patients had diabetes mellitus; 15 were immunocompromised; 14 had chronic lung disease; 11 had a malignancy; 3 had asthma; 3 each had chronic obstructive pulmonary disease, congestive heart failure, chronic liver disease/cirrhosis, and chronic kidney disease; and 1 had a history of intravenous drug use. Forty-six patients had a diagnosis of pneumonia at the time of H. infl uenzae isolation. Patients with pneumonia were signifi cantly more likely to have >1 chronic illness (OR 4.48, 95% CI 1.48-13.58, p = 0.008) and signifi cantly less likely to have meningitis (OR 0.048, 95% CI 0-0.358, p = 0.0016).
The overall case-fatality rate was 22%. Two thirds of deaths were among those >65 years of age. Persons >65 years of age had a case-fatality rate of 29% compared to 15% for those <65 years of age (p = 0.08). The casefatality rate for patients with bacteremia (23%) was not signifi cantly higher than for patients with meningitis (13%). The case-fatality rate did not differ signifi cantly by serotype. Outcome did not differ signifi cantly by the  presence or absence of concurrent conditions (17% vs. 26% mortality rate, respectively), but data were only available for 69 patients.

Discussion
Despite the virtual elimination of Hib in areas with widespread use of Hib vaccine, invasive disease caused by H. infl uenzae remains a major source of illness and death. The incidence of invasive H. infl uenzae in Utah adults appears to be increasing; most of the increase in H. infl uenzae disease incidence was attributable to an increase in nontypeable and Hif strains. We found that invasive H. infl uenzae disease was most common in persons >65 years of age and was associated with a high mortality rate.
The overall incidence of invasive H. infl uenzae disease in adults in our population, 0.66 cases per 100,000 personyears, was similar to rates observed in Illinois (1.0/100,000 person-years in 1994) (7) and slightly lower than rates among Alaskans during 1991-1996 (1.4/100,000 personyears) (5). We found a signifi cant increase in the incidence of invasive H. infl uenzae over the study period by using 2 surveillance methods. Several studies in addition to ours suggest an increase in invasive disease caused by to nontype b H. infl uenzae among adults (5,7,13). The reasons for this change are unclear, and might refl ect changes in the organisms, changes in the number of persons at high risk, or perhaps waning of cross-immunity induced by exposure to Hib.
As the prevalence of Hib has decreased, other encapsulated serotypes seem to have emerged as major causes of invasive disease, including Hif in Illinois and Hia in Brazil, Manitoba, and Northwestern Ontario (7,10,14,15). Similar to fi ndings in Illinois, Hif was the second most common serotype in our analysis (25%), and cases of invasive Hif contributed substantially to the increase of invasive H. infl uenzae cases we observed in 2007-2008. Although Hia was the third most common isolate among adults in our study, it accounted for only 15% of all cases, a rate that is much lower than rates in Manitoba (29%) (15), Northwestern Ontario (42%) (10), and in children in Utah (8).
Nontypeable H. infl uenzae has been the most common isolate in virtually all published series among adults. In our study, it accounted for 43% of cases; similar rates were reported in other studies (3,7,(15)(16)(17). The emerging role of invasive disease because of nontypeable strains is intriguing because this organism has traditionally been considered a relatively noninvasive bacteria predominantly associated with community-acquired pneumonia, chronic obstructive pulmonary disease exacerbations, and otitis media (18). Whether invasive strains of nontypeable H. infl uenzae have distinct genotypic and phenotypic characteristics compared with noninvasive isolates remains largely unknown. Candidate virulence factors include the adhesin genes hmw and hia and IS1016, an insertion element that may confer Hib-like, encapsulated properties to nontypeable strains. However, to date no single set of virulence determinants has been conclusively associated with invasive nontypeable strains (19)(20)(21). Better understanding of the factors that confer invasive capabilities might lead to improved strategies for vaccine development.
Similar to the fi ndings in Spain, Illinois, and Alaska, rates of disease in our study were highest among older patients (3,5,7). The incidence increased in persons >65 years of age, and 29% of infections in this age group were fatal, emphasizing the disproportionate effect. Invasive H. infl uenzae was highly associated with concurrent illness. Among those for whom clinical data were available, 67% of our patients had an underlying condition, as has been observed in other studies (3,7). However, in those with and without concurrent illness, invasive H. infl uenzae showed a high overall case-fatality rate (22%). Other studies have reported that invasive H. infl uenzae is associated with high mortality rates, ranging from 13% to 29% (3,4,6,7 (24). Comparison of estimated rates from the IH laboratory database and UDoH surveillance data suggests that the sensitivity of reporting among adults was 65%. Changes in reporting behavior and changes to the passive surveillance program in 2005 are potential sources of bias. Several studies have demonstrated the inaccuracy of serotyping compared with molecular diagnostic techniques (25,26). Although serotyping is still an accepted standard, had we performed capsular gene analysis, it might have changed the serotype distribution. Additionally, the small number of identifi ed cases places limitations on analyses of serotype distribution and clinical characteristics. Finally, we had clinical information on concurrent conditions for only 57% of the cohort, and clinical information was not collected in a systematic manner across the state.
These limitations notwithstanding, we found that the incidence of invasive H. infl uenzae in Utah adults appears to be increasing, and this increase is caused mostly by the rising incidence of nontypeable H. infl uenzae and Hif. Invasive H. infl uenzae, including disease caused by nontypeable strains, has a high mortality rate. Persons >65 years of age are most affected by the disease and have the highest death rates. These data have implications for targeted adult H. infl uenzae vaccine development.