Bacterial meningitis: Aetiology, risk factors, disease trends and severe sequelae during 50 years in Sweden

Abstract Background Bacterial meningitis (BM) is a rare but severe infection. Few population‐based studies have characterised BM episodes and sequelae over long periods. Methods This was a population‐based observational cohort study with national coverage, using data on aetiological pathogens, sex, premorbid conditions, steroid pretreatment, severe sequelae and birth, death and diagnosis dates collected from 10,339 patients with BM reported to the National Board of Health and Welfare in Sweden between 1964 and 2014. Results During the 50‐year study period, the incidence of BM decreased in young children, but not in the elderly. The most common cause of BM was pneumococci (34%), followed by Haemophilus influenzae (26%), and meningococci (18%), mainly community acquired. Premorbid conditions were found in 20%. After the H. influenzae type b vaccine was introduced in 1993, the BM incidence decreased by 36%. Following pneumococcal conjugated vaccine introduction in 2009, the incidence and 30‐day mortality from pneumococcal meningitis decreased by 64% and 100%, respectively, in previously healthy children, and the 30‐day mortality decreased by 64% among comorbid adults. The BM incidence in immunosuppressed patients increased by 3% annually post vaccine introduction. The 30‐day mortality was 3% in children and 14% in adults, and the rate of severe sequelae was 44%. On average, patients lost 11 years of healthy life due to BM. Conclusion The introduction of conjugated vaccines into the childhood vaccination program has reduced the incidence of BM in young children, but not in adults. Post vaccine introduction, patients present with more premorbid conditions and other bacterial causes of BM, emphasising the need for a correct diagnosis when treating these infections.


Introduction
Bacterial meningitis (BM) is a rare infection that can be complicated by severe sequelae [1]. Infec-tion occurs among patients of all ages, who may have premorbid health conditions, in both community and healthcare settings [2]. Most patients have a single episode; rarely, the infection is recurrent [3]. Community-acquired BM (CBM) is distinct from healthcare-associated BM (HBM). HBM has recently been associated with healthcare facilities (HBMF), such as hospitals or nursing homes, and with clinical procedures resulting in post-neurosurgical infection, with or without an implanted, indwelling intracranial device [4][5][6]. Vaccines targeting BM-causing pathogens have been introduced in several countries. In Sweden, the Haemophilus influenzae type b vaccine (Hib) was introduced into the childhood vaccination program in 1993, and pneumococcal conjugate vaccines (PCVs) were launched nationally in 2009. The effects of vaccine introduction on BM and its severe sequelae remain to be determined, and patient groups with an increased risk of BM need to be identified for preventive measures. During the conjugate vaccine era, overall age-specific survival in Sweden has increased, in part due to new medical treatments for chronic diseases [7][8][9].
Here, we studied the incidence and aetiology of BM, and its associated premorbid conditions, 30day mortality and risk factors for severe sequelae among patients with BM over 50 years in Sweden. Additionally, the importance of specific predisposing conditions for recurrent BM (RBM) was investigated.

Data collection
We performed a 50-year population-based cohort study with national coverage of BM in Sweden. Patients diagnosed with pathogen-specified BM prospectively reported to the National Patient Register (NPR) using the Swedish version of the International Classification of Diseases (ICD) 7-10 were identified from the registers of the National Board of Health and Welfare. Data, including their aetiologic pathogens, date of birth, admission, discharge, death, migration, severe sequelae, clinical setting and premorbid conditions, were collected. The NPR contains Swedish inpatient data from 1964 onwards, with national coverage from 1987. National Swedish registers for open specialist care (2001 onwards) and pharmacology (July 2005 onwards) provided complementary information regarding severe sequelae and pharmacological pre-admission treatment, respectively, for patients diagnosed after these dates. This study was approved by the Uppsala local ethics committee.

Definitions
BM was defined as an episode of pathogenspecified BM identifiable using ICD 9/10 codes, as used in Swedish health care from 1987 onwards. The ICD codes were provided by the attending physicians and bacteria were mainly detected in the cerebrospinal fluid. Detailed definitions of age groups, Charlson comorbidities, risk factors for BM, pharmacological immunosuppression, HBM and pathogens, including ICD codes and classification as specific predisposing conditions, are available in the Supplementary Information (Materials and Methods, Tables S1 and S2). To minimise potential bias due to different ICD versions, coding practices and subnational coverage, no statistical analyses were performed on patient data from before 1987. To further test data validity, we collected 46 medical records of consecutive BM episodes from different Swedish hospitals (Table S3). RBM was defined as a subsequent separate BM episode with the same or different pathogens as in the index episode. Detailed definitions of the RBM are available in the Supplementary Information, Materials and Methods. Severe sequelae were defined as death within 30 days or new severe neurological sequelae in hospitals within either 90 days of admission, for acute cerebrovascular events, or 1 year of admission, for other severe neurological sequelae (hydrocephalus, epilepsy, paresis of one or more limbs, loss of vision and/or other cranial nerve dysfunctions excluding hearing impairment, sensorineural hearing impairment, depression with antidepressant pharmacological treatment, moderate and severe anxiety or attention deficit hyperactivity disorder) [1,10,11]. Pharmacological data were available from 2005 onwards only; therefore, data on severe sequelae were restricted to episodes reported thereafter. Detailed definitions of severe sequelae and disease burden are provided in the Supplementary Information.

Temporal trends of BM and the effects of conjugate vaccines
The BM incidence increased in children under 5 years from 1970 to 1980 and remained constant until the beginning of the 1990s, when it decreased dramatically (Fig. 1). Haemophilus was the dominant agent; however, after 1993, when the Hib vaccine was introduced, there were few Hib cases. The incidence of meningococci was also high between 1965 and the mid-1970s in young children, but decreased thereafter, despite the absence of a meningococcal vaccine in the vaccination program ( Fig. 1 and Figs S1 and S2). The incidence of episodes caused by gram-negative bacteria and streptococci was also most prominent in young children, and there were increases in neonatal BM (annual change 6%, 95% confidence interval [CI95%] 1%-11%, p = 0.01) and neonatal gram-negative BM (annual change 23%, CI95% 9%-39%, p = 0.001) during the conjugate-vaccine era ( Table 1, Fig. S3). Overall, E. coli was the most common gram-negative bacterium among neonates (28/34), in contrast to Pseudomonas, which was mainly found in adults (12/14) (Table S4), Listeria infections were predominantly found in the elderly (208/349, 60%) and were practically absent among neonates ( Table 1). The incidence of BM episodes caused by staphylococci, Listeria and gram-negative bacteria increased from 0.4 to 0.6 per 100,000 person-years (incidence rate ratio 1.8, CI95% 1.  Table 2).

Discussion
BM is a devastating disease carrying a risk of death and severe neurological sequelae among survivors.
To develop effective treatment and preventive strategies, we need knowledge on incidence trends, aetiology and risk factors over the long term. In this unique national population-based study, we examined BM over 50 years in Sweden using national registries. We found that the incidence of BM decreased dramatically in young children after the introduction of conjugated vaccines into the childhood vaccination program, leading to changes in the aetiology of BM during the study period, which varied with age group. The most dramatic decrease was observed following the introduction of the Hib vaccine in 1993, in accordance with previous studies [15]. The incidence of Hib has remained low since then, probably because nonvaccine-type strains have not expanded to any great extent in healthy carriers [16]. The incidence of pneumococcal BM also decreased after PCV introduction but not in nonvaccinated adults, which is in agreement with earlier studies showing a limited impact of PCVs on invasive pneumococcal disease in this age group [17,18]. This limited effect is probably influenced by the observed replacement in carriage during childhood of vaccine-type strains with nonvaccine type strains, which may spread to susceptible individuals, such as nonvaccinated adults [19]. Nonvaccine types have, in previous studies, appeared less capable of causing invasive diseases, including BM, in previously healthy individuals; they primarily affect the elderly and immunocompromised individuals, and generate a milder disease course [18][19][20]. We found that the incidence of meningococcal meningitis has decreased since the mid-1970s, particularly in young children, even though a general meningococcal vaccination was not introduced. The same trend has been observed in other European countries, including Finland, and several mechanisms have been proposed [21].  The incidences of neonatal BM and BM caused by pathogens associated with post-neurosurgical BM, Staphylococcus spp. and gram-negative bacteria were in accordance with those found in previous studies [22,23].
We also showed that during the conjugate-vaccine era, the incidence of BM in patients with immunosuppression surpassed that of BM in previously healthy individuals. At the same time, the proportion of pathogens associated with immunosuppression (Listeria monocytogenes, Staphylococcus spp., gram-negative bacteria) and RBM increased. These trends are probably the result of demographic changes, including increased age-specific survival among adults [7][8][9]. In line with previous findings, we found that several neoplastic, cerebrovascular and traumatic CNS lesions associated with blood-brain barrier dysfunction are risk factors for acquiring BM [3,[24][25][26][27][28][29][30].
Importantly, the overall mortality was 9%, and this decreased during the study period  in the elderly and unvaccinated adults, while it remained low in children (3%). At the beginning of the study period (1987)(1988)(1989)(1990)(1991), the all-cause 30-day mortality among adult patients with BM was 16% (126/780); towards the end of the study period (2010-2014), it was 9% (85/910) (OR 0.5, CI95% 0.4-0.7; p < 0.001). Mortality due to pneumococcal BM decreased during the study period, from 18% (94/524) at the beginning of the study period to 9% (42/466) by the end of the study period (OR 0.5, CI95% 0.3-0.7; p < 0.001). Post-PCV, a decrease was observed among children, comorbid adults and the elderly. Moreover, severe sequelae were present in 44% of the patients with BM, and the highest rates were found among patients with staphylococcal (62%) and pneumococcal (47%) infections. Thus, despite decreased incidence and improved prognosis, we found that BM patients, on average, lost about 11 years of healthy life, and pneumococcal BM takes the highest toll. These data show similar patterns to those seen in previous studies, but with slightly higher rates [21,31]. Furthermore, we showed that age, pathogen and immunocompetence, including prior neurosurgery, are important prognostic indicators for BM. However, in a non-neurosurgical setting, patient age, rather than premorbid conditions or recent association with a healthcare facility, is of prognostic value. When considering initial empiric treatment in a patient with suspected BM, fac-tors associated with opportunistic or resistant bacteria, such as immunosuppression, older age or recent hospital or nursing home care, are also important.
This study has several strengths but its limitations should also be considered. The study period was long, at 50 years; the source has national coverage and includes culture-negative episodes. Random erroneous reporting would result only in nondifferential misclassification, and the comorbidity data have been shown to be highly valid [32]. The Swedish setting, with a tax-financed healthcare system and unique individual numbers assigned to citizens at birth and to immigrants at contact with the healthcare system, allows for data tracking. NPR coverage has increased gradually from only a few hospitals in 1965 to near national coverage in 1975; thus, data from prior to 1975 should be interpreted with caution. We found no evidence of changes in data quality from 1987 onwards; thus, only data from this period were retained for the analysis. The medical records of 46 consecutive episodes of pneumococcal BM with 1 year of follow-up data from hospital admittance were retrieved from selected Swedish secondary and tertiary care hospitals from 2005 to 2014, and we found consistency between the reported registry data and original medical records (Supplementary Information, Materials and Methods, Fig. S2).
We conclude that this national study provides important information on BM and its consequences, with a decreasing incidence in younger age groups in the post-vaccination era, but with changes in aetiology and patient characteristics. Conjugated vaccines have reduced the number of BM cases caused by pneumococci and H. influenzae, but BM caused by meningococci has also decreased, even though no vaccine has been introduced against this pathogen. This information is relevant for future strategies to treat and prevent this severe infection.