Background. Although hospitalizations due to invasive pneumococcal disease decreased after routine vaccination of young children with a 7-valent pneumococcal conjugate vaccine (PCV7) began in 2000, information on the trends in pneumococcal meningitis is limited.
Methods. We estimated national trends in rates of hospitalization for pneumococcal meningitis, using data from the Nationwide Inpatient Sample, 1994–2004. Pneumococcal meningitis cases and deaths were identified on the basis of the International Classification of Diseases, Ninth Edition, Clinical Modification coded primary discharge diagnosis, and rates were calculated using US Census data as denominators. The year 2000 was considered to be a transition year, and the average annualized rate after PCV7 introduction (2001–2004) was compared with that during the baseline years (1994–1999).
Results. During 1994–2004, there were 21,396 hospitalizations and 2684 deaths (12.5%) due to pneumococcal meningitis in the United States. In children aged <2 years, the average annualized rates of pneumococcal meningitis hospitalizations per 100,000 population decreased from 7.7 in 1994–1999 to 2.6 in 2001–2004 (change, −66.0%; 95% confidence interval [CI], −73.5% to −56.3%). Among children aged 2–4 years, the hospitalization rate decreased from 0.9 to 0.5 per 100,000 (change, −51.5%; 95% CI, −66.9% to −28.9%). Average rates also decreased by 33.0% (95% CI, −43.4% to −20.9%) among adults aged ⩾65 years. After PCV7 introduction (2001–2004), an estimated 1822 and 573 pneumococcal meningitis hospitalizations were prevented in persons aged <5 years and ⩾65 years, respectively. Overall, an estimated 3330 pneumococcal meningitis hospitalizations and 394 deaths were prevented in persons of all ages during 2001–2004 in the United States.
Conclusion. After implementation of routine childhood vaccination with PCV7, hospitalizations for pneumococcal meningitis decreased significantly for both children and adults. Most pneumococcal meningitis cases now occur among adults.
After the introduction of conjugate vaccines to prevent Haemophilus influenzae type b disease, Streptococcus pneumoniae became the leading cause of bacterial meningitis in the United States . The risk of developing pneumococcal meningitis is highest among young children, older adults, persons with chronic illnesses, and immunocompromised individuals [1–3]. Even with advances in medical care, the case-fatality rate has a range of 16%–37% in adults  and 1%–2.6% in children [1, 5]. Those who survive the disease have a 30%–52% risk of neurological sequelae [6, 7].
In 2000, a 7-valent pneumococcal conjugate vaccine (PCV7) was licensed and recommended for routine use in US children . In clinical trials, PCV7 prevented invasive pneumococcal diseases (e.g., bacteremia and meningitis) and reduced nasopharyngeal carriage of vaccine serotypes [8–10].
Studies have demonstrated decreases in invasive pneumococcal disease in all age groups after the introduction of PCV7 [11–14]. However, by 2002–2003, the rates of pneumococcal meningitis in adults aged ⩾50 years were reported not to have changed compared with the rates during the baseline years . The national impact of the PCV7 immunization program on pneumococcal meningitis in all age groups is currently unknown. We evaluated trends in the incidence and mortality of pneumococcal meningitis hospitalizations during 1994–2004, using the Nationwide Inpatient Sample (NIS), the largest source of inpatient data available in the United States.
The NIS. The NIS contains data on hospital inpatient stays from US states that participate in the Healthcare Cost and Utilization Project (sponsored by the Agency for Healthcare Research and Quality) . The NIS contains patient-level clinical and resource-utilization data and provides information on 5–8 million hospitalizations per year from ∼1000 hospitals. These hospitals constitute an ∼20% sample of community hospitals in the United States, including nonfederal short-term, general, and specialty hospitals. Participating hospitals are sampled by stratified probability sampling in 5 strata (ownership/control, bed size, teaching status, urban/rural, and US region), with sampling probabilities proportional to the number of US community hospitals in each stratum. The NIS collects data on all hospitalizations regardless of payment source, and weighting and sampling variables are provided for each year to calculate national estimates.
Study design. We evaluated the effect of the implementation of PCV7 vaccination on rates of pneumococcal meningitis, using an ecologic design. PCV7 was introduced in 2000, and vaccine coverage increased rapidly afterward. Thus, calendar years were considered a surrogate marker for vaccine uptake. NIS data from 1994–2004 were analyzed to allow the assessment of secular trends that preceded the implementation of the PCV7 immunization program.
Definition of bacterial meningitis hospitalizations. A hospitalization due to pneumococcal meningitis was defined as an NIS record with a principal discharge diagnosis (first-listed diagnosis) of pneumococcal meningitis (International Classification of Diseases, Ninth Edition, Clinical Modification [ICD-9-CM] code 320.1). In addition, we evaluated hospitalizations with a principal discharge diagnosis of streptococcal meningitis, to examine the possibility that some cases of pneumococcal meningitis could have been misclassified as streptococcal meningitis (ICD-9-CM code 320.2). We also evaluated rates of H. influenzae meningitis, which decreased rapidly in the early 1990s after routine childhood H. influenzae type b vaccination, and meningococcal meningitis, which has been reported to have gradually decreased since the late 1990s [1, 16, 17]. Rates of other bacterial meningitis (specified or unspecified) were also evaluated for comparison. In addition, these 5 mutually exclusive diagnostic groups were aggregated into an all bacterial meningitis group.
A secondary analysis identifying meningitis codes listed in any diagnosis field (rather than only the principal diagnosis) yielded results similar to those of the primary diagnosis. Patients with meningitis who died during their hospitalization were classified as meningitis deaths for the calculation of in-hospital mortality rates and case-fatality ratios.
Statistical analysis. National weighted frequencies of meningitis hospitalizations and their respective SEs were calculated using NIS inflation (DISCWT) and stratum (NIS_STRATUM and STRATUM) variables. Annual hospitalization rates were computed using NIS weighted frequencies as numerators and annual, mid-year population estimates from the US Census Bureau as denominators . Similarly, national estimates of meningitis mortality rates were calculated using weighted in-hospital meningitis deaths as numerators and US Census Bureau midyear population data (person-year estimates) as denominators. We also evaluated trends in annual meningitis hospitalizations and mortality rates from 1994 to 2004 by age group (age <2 years, 2–4 years, 5–17 years, 18–39 years, 40–64 years, and ⩾65 years).
In a separate analysis, we divided the study years into 3 time periods: 1994–1999 (baseline), 2000 (transition), and 2001–2004 (after PCV7 introduction). We considered 2000 to be a transition year and excluded it from this analysis because vaccine uptake started to increase substantially only after US governmental purchasing through the Vaccines For Children program began in June 2000 .
To estimate the impact of the immunization program, we calculated the average weighted pneumococcal meningitis hospitalization and mortality rates for the baseline years (the reference group) and for the years after PCV7 introduction. Rate differences and percentage changes were estimated by fitting outcome-specific Poisson regression models with adjustment for age group, calendar year, and their interaction, while accounting for the NIS sampling design. The estimated population for each age group during each calendar year was the offset term for the models. Comparisons of rates before and after the introduction of PCV7 were obtained through linear combinations of coefficients from the models [20, 21]. The number of pneumococcal meningitis hospitalizations and deaths prevented after PCV7 introduction was estimated by multiplying the estimated rate difference by the respective population estimates. Because changes in rates of streptococcal meningitis in infants might be associated with changes in early-onset neonatal group B streptococcal infection, we performed a secondary analysis for this group, excluding children aged <30 days.
Statistical analyses were performed using SAS 9.1 (SAS Institute) and Stata 8.2 (StataCorp). This study was considered to be exempt from review by the institutional review boards of Vanderbilt University and the Centers for Disease Control and Prevention.
Of the total 395,917,007 weighted hospitalizations in the United States during 1994–2004, 21,396 had a primary discharge diagnosis of pneumococcal meningitis. The median age of patients was 41 years (interquartile range [IQR], 3–59 years), and 53% were male. The median length of hospital stay was 10 days (IQR, 6–14 days), and 2684 (12.5%) of the patients with pneumococcal meningitis died during hospitalization.
Changes in pneumococcal meningitis hospitalizations. Annual rates of pneumococcal meningitis hospitalizations fluctuated before the introduction of PCV7 in 2000, decreased sharply in 2000 and 2001, and remained relatively stable during 2002–2004. The pneumococcal meningitis mortality rates showed a similar pattern, whereas the overall case-fatality ratios fluctuated (figure 1).
Overall, pneumococcal meningitis hospitalization rates decreased by 33.0% after PCV7 introduction. The average annualized rate in children aged <2 years decreased 66.0%, from 7.7 to 2.6 per 100,000 in this PCV7 target population. A 51.5% decrease in annual rates was also observed in children aged 2–4 years. During the same period, the rates decreased in older age groups as well. The average rates among persons aged ⩾65 years decreased from 1.2 to 0.8 per 100,000 population, representing a 33.0% decrease (table 1 and figure 2).
After implementation of routine vaccination with PCV7, the overall pneumococcal meningitis mortality rate decreased by 32.7% (table 2). Children aged <2 years had the largest decrease in mortality rate (from 0.37 to 0.18 per 100,000; 51.1% decrease), followed by persons aged ⩾65 years (from 0.34 to 0.19 per 100,000; 43.9% decrease). Because the decrease in hospitalization rates was larger than the decrease in mortality rates in children aged <2 years, the case-fatality ratio for pneumococcal meningitis (in-hospital deaths per meningitis hospitalizations) increased from 4.9 to 7.0 per 100 cases. In persons aged ⩾65 years, the case-fatality ratio decreased from 27.2 to 22.8 per 100 cases. There were no significant changes in mortality rates of pneumococcal meningitis in other age groups (table 2).
After introduction of PCV7, there was an average of 1572 pneumococcal meningitis hospitalizations annually, compared with 2199 during the baseline years. On the basis of the rate differences, we estimated that nationally, in the 4 years after PCV7 introduction (2001–2004), 1822, 360, and 573 pneumococcal meningitis hospitalizations were prevented in persons aged <5 years, 18–39 years, and ⩾65 years, respectively. In persons of all ages, there were 3330 fewer pneumococcal meningitis hospitalizations and 394 fewer deaths after PCV7 introduction, compared with during the baseline years (1994–1999). Because of these changes, the median age of patients increased from 37 years (IQR, 1–59 years) during the baseline years to 46 years (IQR, 18–60 years) after PCV7 introduction. During the baseline years, children aged <5 years accounted for 30% of pneumococcal meningitis hospitalizations, compared with 15% after PCV7 introduction; the percentage of hospitalized patients with meningitis aged ⩾65 years (∼20%) remained constant throughout the study period (figure 3).
Changes in nonpneumococcal bacterial meningitis hospitalizations. Similar to the trends observed in pneumococcal meningitis rates, there was an overall 17.5% decrease in the rates of streptococcal meningitis. However, the decrease was significant only in children aged <5 years. To assess the possible effect of changes in early-onset, group B streptococcal meningitis on the estimates, we excluded children aged <30 days. The average rate of streptococcal meningitis hospitalizations in children aged 1–23 months showed a 34.9% decrease after PCV7 introduction (table 1 and figure 4).
The overall rates of meningococcal and H. influenzae meningitis hospitalizations also decreased during the study period (table 1 and figure 4). Rates of meningococcal meningitis decreased in all age groups, whereas the decrease in H. influenzae meningitis rates occurred primarily in children aged <2 years. The decreasing trend in rates of both meningococcal and H. influenzae meningitis began in the 1990s.
The other bacterial meningitis group included diagnoses of unspecified bacterial meningitis (54.3%), staphylococcal meningitis (11.2%), tuberculosis meningitis (9.1%), and other specified and not elsewhere classified bacterial meningitis (25.2%). During the study years, rates of other bacterial meningitis hospitalizations also decreased. Table 1 shows that children aged <2 years had the largest decreases in other bacterial meningitis hospitalizations over time. However, most of these decreases occurred in the 1990s; after 2000, the rates increased modestly (figure 4).
After routine PCV7 vaccination in the United States, hospitalization rates for pneumococcal meningitis decreased significantly in young children, young adults, and elderly persons. We estimated that, during the first 4 years after PCV7 introduction (2001–2004), ∼1800 meningitis hospitalizations in children aged <5 years were prevented in the United States, resulting in a major change in the age distribution of pneumococcal meningitis cases. The majority of cases now occur in working-age and older adults.
The 66% reduction observed in national rates of pneumococcal meningitis hospitalizations in children aged <2 years, the target population for the PCV7 program, is consistent with reports from other population-based studies of pneumococcal meningitis in selected areas, including a 59% decrease in children aged <2 years by 2001 in 7 geographic areas  and a 69% decrease in children aged <5 years by 2003 in Massachusetts . Similarly, a 56% reduction in pneumococcal meningitis was observed between 1994 and 2004 in 8 children's hospitals in the United States . In addition, a 40% decrease in invasive pneumococcal disease cases was seen in infants aged 0–90 days after the introduction of PCV7 . Laboratory-based surveillance in these investigations confirmed that the decrease was limited to PCV7 serotypes, providing strong evidence of a causal association [12, 14, 22].
The NIS databases were large enough to detect significant decreases in meningitis in adults and children, as has been shown with all invasive pneumococcal disease [11, 12, 24]. Significant reductions in pneumococcal meningitis rates were seen in persons aged 18–39 years and aged ⩾65 years, for whom PCV7 is not recommended. This suggests an indirect vaccine effect, likely because of the reduced nasopharyngeal carriage of vaccine-type pneumococci in vaccinated children and decreased bacterial transmission to their close contacts . Although use of HAART has resulted in important decreases in rates of invasive pneumococcal disease among HIV-infected adults , the major effect of HAART took place before PCV7 introduction .
The influence of increasing uptake of the 23-valent pneumococcal polysaccharide vaccine (PPV23) in persons aged ⩾65 years during the study period for pneumococcal meningitis trends is unclear. The median national PPV23 coverage increased from 37% in 1995 to 60% in 2001 but has remained at similar levels since then. The decreases in invasive pneumococcal disease documented after PCV7 introduction have been specific to serotypes included in PCV7, whereas no reduction was seen in the 16 serotypes contained only in PPV23 .
These observed decreases have changed the epidemiology of pneumococcal meningitis in the United States, and, as the age distribution of pneumococcal meningitis continues to change after implementation of PCV7, it might be necessary to reevaluate prevalent pathogens and current age-based empirical treatment recommendations for bacterial meningitis . Moreover, the mortality rates for pneumococcal meningitis decreased after implementation of PCV7 vaccination, and an estimated 394 deaths were prevented during the first 4 years after PCV7 introduction. Nevertheless, despite the substantial vaccine effects and advances in treatment, the disease case-fatality ratio remains high, emphasizing the need for additional prevention strategies.
In addition to the decrease in pneumococcal meningitis hospitalizations, we observed a comparable reduction in hospitalizations for streptococcal meningitis in children aged <5 years. Recommendations for giving peripartum antibiotics to mothers colonized with group B Streptococcus were implemented before PCV7 and have resulted in a decrease in early-onset (at age 1–7 days) neonatal group B streptococcal invasive disease . The pattern of decrease in streptococcal meningitis remained similar after children aged <30 days were excluded to account for the decrease in early-onset neonatal group B streptococcal meningitis. Some observations suggest that this decrease in streptococcal meningitis might be partly due to misclassification of S. pneumoniae meningitis as “streptococcal meningitis.” First, the trend of streptococcal meningitis in children aged <2 years closely resembled that of pneumococcal meningitis in the same age group, showing a sharp decrease in hospitalization rates soon after routine PCV7 vaccination began (data not shown). Second, the reduction in streptococcal meningitis hospitalizations was significant only in children aged <5 years. If some of the pneumococcal meningitis cases were misclassified as streptococcal meningitis, our estimate of the impact of PCV7 vaccine would be conservative. For children aged <2 years, the decrease in streptococcal meningitis represented an additional 864 fewer cases of meningitis than were expected after PCV7 introduction, compared with during the baseline years.
The hospitalization rates for meningococcal and H. influenzae meningitis also decreased during the study period, beginning before the introduction of PCV7. The decrease in meningococcal meningitis rates is consistent with the cyclical nature of this disease and with surveillance reports that have shown a decrease in invasive Neisseria meningitidis disease since the 1990s . The quadrivalent A, C, Y, and W-135 meningococcal polysaccharide vaccine has been used in military recruits and certain high risk groups since 1980s and in college freshmen since 2000 [30, 31]. The relation of this targeted vaccine use with the decrease in disease rates, however, is unclear, particularly because use of this vaccine is restricted to persons aged >2 years and substantial decreases were seen in this age group. H. influenzae meningitis decreased after licensure and widespread use of H. influenzae type b conjugate vaccines for children at least 18 months old in 1987 and infants at least 2 months old in 1990 . Rates for other bacterial meningitis decreased from 1994 to 2000 but increased after 2000. The temporal pattern of these changes was different from that of pneumococcal meningitis, and there was no decrease in hospitalizations for other bacterial meningitis overall across the study period.
Findings from this study need to be interpreted in light of some potential limitations. For this ecologic study, we did not have individual vaccination records, and we considered calendar year as a surrogate marker for PCV7 coverage in the population. However, the observed decreases in pneumococcal meningitis were similar to the decreases in invasive pneumococcal disease reported in other surveillance studies in selected geographic areas [12, 14, 22] and were consistent with rapidly increasing vaccine uptake after PCV7 introduction . Vaccination coverage with ⩾3 doses increased to 68%, 73%, and 83% for children born in February 2000–June 2002, February 2001–May 2003, and February 2002–July 2004, respectively , despite the reported intermittent vaccine shortages during 2001–2004 . Moreover, in the absence of secular trends, the ecologic approach may be the preferred way to estimate the total program effect because it accounts for both direct and indirect vaccine effects.
We identified meningitis hospitalizations on the basis of the ICD-9-CM codes listed as primary discharge diagnoses, which are considered to be the main reasons for hospital admission  and are likely to be specific for a severe condition such as meningitis. Although rates of pneumococcal meningitis hospitalizations for children aged <2 years that were estimated from discharge data were similar in magnitude to those previously reported from active surveillance at selected US sites (10.3 cases per 100,000 population during 1998–1999 and 4.2 cases per 100,000 population in 2001) , NIS rates were somewhat lower, suggesting a lower sensitivity of this approach compared with that of active case finding.
Furthermore, although we explored information on rates preceding the implementation of PCV7 to assess secular trends, we could not rule out effects on our estimates from increasing PPV23 uptake by the elderly population, meningitis outbreaks, or changes in the coding of meningitis hospitalizations. Nevertheless, outbreaks of pneumococcal meningitis are uncommon in developed countries, and, although changes in coding practices for meningitis during the study period are unknown, these changes would be unlikely to affect selectively different age groups.
Information on pneumococcal serotype distribution or antimicrobial susceptibility of the bacterial isolates is not available in the NIS. Before PCV7 introduction, the vaccine serotypes (4, 6B, 9V, 14, 18C, 19F, and 23F) accounted for 73% of pneumococcal meningitis in US children aged <6 years . Although invasive disease caused by these serotypes has been reduced dramatically, the incidence of invasive disease caused by nonvaccine serotypes has been increasing in recent years [24, 37]. Although to date these increases have been small compared with overall disease reductions , continuous monitoring of changes in serotype distribution is necessary.
In conclusion, this study provides a comprehensive assessment of changes in pneumococcal meningitis hospitalization rates and age distributions of the disease after a routine PCV7 immunization program began in the United States. Results from this study contribute to the evidence supporting the overall nationwide beneficial effects of PCV7 on pneumococcal meningitis, the most common cause of community-acquired bacterial meningitis.
Financial support. Centers for Disease Control and Prevention and The Association for Prevention Teaching and Research (cooperative agreement TS-1392) and Agency for Healthcare Research and Quality (1R03HS016784).
Potential conflicts of interest. M.R.G. and C.G.G. received speaker honoraria from Wyeth, the manufacturer of the pneumococcal conjugate vaccine. C.J.T. and J.P.N.: no conflicts.