Delays in diagnosis and treatment of vaccine preventable community acquired bacterial meningitis (CABM): a retrospective analysis at three tertiary care centers

Background Outcomes in community-acquired bacterial meningitis (CABM) are signicantly impacted by delays in diagnosis and treatment. This retrospective case series aims to describe the sociodemographic, epidemiological, and clinical variables including time to diagnosis and treatment of vaccine preventable CABM in three tertiary care settings in New York City (NYC). Methods A retrospective chart review was conducted of patients at Columbia University Irving Medical Center (CUIMC), Children’s Hospital of New York (CHONY), Mount Sinai Health System, and Weill Cornell Medical Center with CABM due to Haemophilus inuenzae type B, Streptococcus pneumoniae, and Neisseria meningitidis between January 1, 2012 and December 31, 2017. A descriptive statistical analysis was performed. Results Our case series consisted of 36 patients, 24 (66.7%) females, and 12 (33.33%) males with a median age of 42 years (IQR 55 years). Median time from presentation to lumbar puncture (LP) was eight hours (IQR 7). The median time from hospital presentation to diagnosis was 12 hours (IQR 9), and the median time from LP to diagnosis was three hours (IQR 5). Delay in diagnosis which is dened by more than 8 hours from hospital presentation, occurred in 13 patients (36.1%) due to initial misdiagnosis, most commonly systemic febrile and/or viral infections and otitis media. Conclusions Despite evidence of the importance of early diagnosis and treatment for CABM, this case series shows the ongoing challenges with early clinical diagnosis. Misdiagnoses were an underlying reason for delays from presentation to LP and to antibiotic treatment in the majority of our patients. This study in NYC identies ongoing major delays in diagnosis and antimicrobial treatment in CABM, and future studies are needed to identify mechanisms to improve time to antibiotic treatment and LP in CABM.

median age of 42 years (IQR 55 years). Median time from presentation to lumbar puncture (LP) was eight hours (IQR 7). The median time from hospital presentation to diagnosis was 12 hours (IQR 9), and the median time from LP to diagnosis was three hours (IQR 5). Delay in diagnosis which is de ned by more than 8 hours from hospital presentation, occurred in 13 patients (36.1%) due to initial misdiagnosis, most commonly systemic febrile and/or viral infections and otitis media.
Conclusions Despite evidence of the importance of early diagnosis and treatment for CABM, this case series shows the ongoing challenges with early clinical diagnosis. Misdiagnoses were an underlying reason for delays from presentation to LP and to antibiotic treatment in the majority of our patients. This study in NYC identi es ongoing major delays in diagnosis and antimicrobial treatment in CABM, and future studies are needed to identify mechanisms to improve time to antibiotic treatment and LP in CABM.

Background
Despite the effectiveness of vaccines in prevention of CABM, in those who acquire CABM mortality rates remain high at 10-30%, with untreated cases resulting in death within 24 to 48 hours (Glimaker et al 1 , Cohn et al 2 , World Health Organization 2017 3 ). Prompt and effective clinical assessment are essential as delays in diagnosis and treatment were noted in one study to lead to increased mortality by 12.6% per hour and result in permanent neurological sequelae including epilepsy, hearing loss, and neurocognitive de cits in at least 20% of survivors (Glimaker et al 1 , Oordt-Speets et al 4 ). A prospective study of individuals diagnosed with pneumococcal meningitis found that a delay in antibiotic administration of more than three hours after presentation was independently associated with three-month mortality rates (Auburtin et al 5 ). A comparable retrospective study found that the adjusted odds ratio for mortality was 8.4 times higher for those experiencing delays greater than six hours from presentation to appropriate antibiotic administration (Proulx et al 6   discharge codes, G00.8 and G00.9, 320.82, and 320.9 were reviewed in detail but few cases were related to the three pathogens of interest. The majority of the unspeci ed bacterial meningitis cases stemmed from shunt infections, were CSF culture negative, or caused by low-incidence pathogenic strains not addressed in this paper. As such, these case codes were excluded from this study. Additionally, patients under two months of age and those with any surgical hardware in the brain or skull were excluded. Sociodemographic data (age, sex, race, ethnicity, English pro ciency, employment status, and household size) and epidemiologic factors (recent sick contacts, recent travel within one month of admission, chronic health conditions, daycare enrollment, alcohol, tobacco, and drug use) were gathered from EMRs.
Clinical data including diagnostic evaluation, presenting symptoms, preceding illness, length of hospital and intensive care unit (ICU) stay, and need for intubation were also extracted. Time spent in the emergency department (ED) before inpatient admission, time to lumbar puncture (LP), time to etiological diagnosis de ned as rst positive CSF culture or CSF PCR, and time of initial central nervous system (CNS) antimicrobial coverage for BM were recorded. Glasgow Coma Scores (GCS) at presentation and Glasgow Outcome Scores (GOS) at discharge were gathered as well as the presence of hearing, behavioral and cognitive de cits, occurrence of seizures, and ability to complete tasks of daily living (ADLs) at discharge; three-to-six-month follow-up; and one-year follow-up where available. Delay from presentation to LP was de ned as more than 6 hours, delayed presentation to administration of antibiotics was de ned as more than 4 hours, and delay of diagnosis was de ned as more than 8 hours from time of hospital admission. Descriptive statistical analyses including mean, median, standard deviation, and interquartile range were calculated for each continuous variable. Analyses were conducted using RStudio version 3.4.3 (RStudio, Boston, MA 7 ) and SAS version 9.4 (SAS Institute Inc., Cary, NC 8 ).
The median number of days from initial neurological symptom to presentation at any clinical setting (OSH, clinic, or ED at one of our study sites) was one day (IQR 2 days). For individuals whose records included detailed time data (34, 94.4%), the median time from presentation to LP was 8 hours (IQR 7). The median time from presentation at hospital to diagnosis, de ned as the rst positive CSF culture or PCR result, was 12 hours (IQR 9), and the median time from LP to diagnosis was 3 hours (IQR 5). The median time from presentation to administration of CNS antimicrobial coverage was 4 hours (IQR 5).
Two (5.6%) patients, who were initially misdiagnosed, were non-English speaking. Five individuals (13.9%) had a delay of four hours or more from presentation to the administration of antibiotics with appropriate CNS coverage. All of these antibiotic administration delays were due to initial misdiagnosis ( Table 4).
The median length of hospital stay at the three study sites was 12 days (IQR 15), and 31 (86.1%) individuals were admitted to the ICU for a median stay of 6 days (IQR 6). Nineteen (52.7%) individuals were intubated, 11 (30.6%) had seizures, 3 (8.3%) of whom were in status epilepticus, and 8 (22.2%) had cerebral edema (   13 ). Overall survival rates, as well as the rates of most common sequelae of BM, were also comparable to those in previous studies.
This study reports a delay in time to LP and to antibiotic administration as shown by previous studies (Proulx et al 6 , Auburtin et al 5 ). Initial misdiagnosis accounted for all delays of three hours or more from presentation to antibiotic treatment and delays from presentation to LP. Some LPs were deferred due to concern for cerebral edema or patient hesitancy to consent to LP.
In this study, the median time from presentation to LP was eight hours, and the median time from presentation to appropriate antibiotic administration was four hours as compared to estimates of one to two hours in prior similar cohorts (Miner et al 14 , Bodilson et al 15 ). Importantly, while overall time from presentation to antibiotic administration was longer in this cohort than prior similar cohorts, individuals who were not initially misdiagnosed were administered antibiotics in two hours or less-similar to previously published treatment times. This disparity in the current cohort suggests that taking steps to prevent misdiagnoses can signi cantly reduce delays to LP and antibiotic administration, thereby likely improving clinical outcomes. Contributing factors to misdiagnosis in this study include initial presentation to an outpatient clinic and a preceding or coinciding illness veiling CABM.
This study has several important limitations. Firstly, the ICD discharge codes used to identify potential cases may not include all cases of CABM. This study also lacks detailed symptomology at presentation, which could possibly identify factors associated with misdiagnoses and diagnostic delays. Further, the retrospective design relies on data obtained through EMRs which is subject to the availability and completeness of user entries. Lastly, the study has a small sample size and focuses only on three sites in NYC. However, this study capitalized on the sociodemographic diversity of the NYC patient sample that uniquely contributes to the epidemiological characterization of CABM. This study also provides insight into how tertiary NYC EDs and hospitals manage CABM cases. Our dataset provides timestamped data for many clinical and demographic variables, enabling a more thorough exploration of the epidemiological associations underpinning diagnostic and treatment delays.

Conclusions
Future studies are needed to investigate contributing factors of diagnostic and treatment delays such as detailed clinical symptomology, language and literacy barriers, and geographic proximity/physical accessibility to care facilities that may contribute to poor prognosis. Increased awareness of the rami cations of delayed antibiotic treatment in mortality and permanent neurological sequelae can offer care providers better qualitative metrics by which to triage vulnerable cases. As the epidemiological pro le of BM continues to evolve by virtue of antibiotic resistant strains and serotype replacement, it is essential to optimize preventative, diagnostic and therapeutic measures to reduce mortality and increase quality of life for survivors. Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.
Funding: Jacqueline Gofshteyn, is supported by the NIH/NINDS NSADA-K12 Career Development Award (NS5250799523). Kiran Thakur is supported by the National Institute of Health, NINDS K23 NS105935-01 and NIH/NICHD 1R01HD074944-01A1 Authors' contributions: SDT collected and extracted and statistically analyzed data, and was a major contributor in writing the manuscript. MD collected and extracted and statistically analyzed data, and contributed to writing and editing of the manuscript. CYK contributed to the writing and revision of the manuscript. JVK statistically analyzed the data. NL collected and extracted the data. MH collected and extracted the data. EMS reviewed and revised the manuscript. BG collected the data, and reviewed and revised the manuscript. SSM conceptualized and designed the study, collected data, reviewed and revised the manuscript. DW reviewed and revised the manuscript. JSG reviewed and revised the manuscript. AKY reviewed and revised the manuscript. KTT conceptualized and designed the study, wrote the initial draft, reviewed and revised the manuscript. All authors read and approved the nal manuscript.