OUP user menu

Delays in the administration of antibiotics are associated with mortality from adult acute bacterial meningitis

N. Proulx , D. Fréchette , B. Toye , J. Chan , S. Kravcik
DOI: http://dx.doi.org/10.1093/qjmed/hci047 291-298 First published online: 10 March 2005

Abstract

Background: Bacterial meningitis continues to cause high mortality. Few studies address the possible association between this mortality and antibiotic administration delays.

Aim: To determine whether delays in antibiotic administration are associated with mortality from bacterial meningitis, and to identify inappropriate diagnostic–treatment sequences leading to such delays.

Design: Retrospective case record study.

Methods: We reviewed 123 cases of adult acute bacterial meningitis in 119 patients aged ⩾16 years admitted to hospital from January 1990 to March 2002, using multivariate regression analysis to assess the association between meningitis mortality and door-to-antibiotic time (the time elapsed between emergency room presentation and antibiotics administration).

Results: The case fatality rate was 13% (16/123). Adjusted odds ratios (OR) for mortality were: 8.4 (95%CI 1.7–40.9) for door-to-antibiotic time >6 h; 39.4 (95%CI 4.3–358.1) for afebrility at presentation; and 12.6 (95%CI 2.2–72.0) for severely impaired mental status at presentation. Factors associated with a door-to-antibiotic time of >6 h were: (i) failure to administer antibiotics prior to transfer from another institution (OR 21.8); (ii) the diagnostic–treatment sequence: head CT then lumbar puncture, then antibiotics (OR 5.6); and (iii) the absence of the classic meningitis triad (OR 4.9).

Discussion: There is an independent incremental association between delays in administrating antibiotics and mortality from adult acute bacterial meningitis. Inappropriate diagnostic–treatment sequences were significant predictors of such treatment delays.

Introduction

Adult acute bacterial meningitis is a potentially life-threatening condition requiring immediate recognition and treatment. Despite the development of more effective antibiotics, bacterial meningitis continues to cause high mortality.1 Textbooks of infectious diseases recommend appropriate antibiotic administration within 30 min of presentation. Surprisingly, recent studies consistently report substantial delays in the time to antibiotic administration, ranging from 2 to 4.9 h.2–5 We postulated that significant delays in antibiotic administration could be associated with adverse clinical outcome in bacterial meningitis.

Establishing an association between delays in antibiotic administration and adverse outcome would change management strategies in the emergency department. Previous studies exploring this possible association had significant flaws. Some assessed mixed populations of adults and children, or used broad, unclear definitions of ‘time to antibiotics’.6 Moreover, few studies have assessed time to treatment and its effect on mortality. A retrospective review of 171 cases of adult and child meningitis found significantly higher mortality rates in patients first given antibiotics as in-patients, as compared to those given antibiotics in the emergency department.6 In a second retrospective review of 93 children with bacterial meningitis, no correlation between morbidity and/or mortality with antibiotic time was established.5 Such inconsistent findings may be partly explained by failure to consider the many recognized predictors of mortality in meningitis as potential confounding factors. Based on these and other findings, leading authors state that this association remains inconclusive and speculative.4–7 A more careful analysis of an association between delays in antibiotic administration and adverse clinical outcome is needed, as it would have important diagnostic and therapeutic management implications in the emergency department setting.

Using a more precise, feasible, and objective approach of door to antibiotic time, we sought to determine whether delays in antibiotic administration were associated with mortality from adult acute bacterial meningitis, and to identify inappropriate diagnostic–treatment sequences that can lead to delays in antibiotic administration.

Methods

Study design and setting

We conducted a retrospective chart review of all patients with bacterial meningitis admitted to hospital over the 12-year period between January 1990 and March 2002. The hospital is a 1050-bed university-based centre on three campuses, providing primary care and tertiary referral services for a mixed urban and rural population in excess of one million. This study was approved by the institution's research ethics board.

Selection of patients

Patients were identified from the health record department using the International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes for bacterial meningitis. Those considered for inclusion had bacterial meningitis not occurring in the presence of an intraventricular neurosurgical prosthesis such as an external ventricular drain or a ventriculo-peritoneal shunt. Patients with proven viral, fungal or mycobacterial meningitis were excluded.

Specific inclusion criteria required subjects to be aged ⩾16 years with one of the following:7 (i) a positive cerebrospinal fluid culture; (ii) a negative cerebrospinal fluid culture with neutrophilic pleocytosis and at least one of: positive blood cultures, positive cerebrospinal fluid gram stain, cerebrospinal fluid high protein (>0.60 g/l) and low glucose (<2.7 mmol/l), positive cerebrospinal fluid bacterial antigen test; (iii) a negative cerebrospinal fluid culture without neutrophilic pleocytosis but with positive blood culture and at least one of: cerebrospinal fluid high protein (>0.60 g/l) and low glucose (<2.7 mmol/l), positive cerebrospinal fluid gram stain, positive bacterial antigen test; (iv) autopsy confirmation.

Neutrophilic pleocytosis was defined as an absolute neutrophil count ⩾100/mm3 in immunocompetent patients or ⩾50/mm3 in immunocompromised patients (HIV-infected, on anti-neoplastic chemotherapy or long-term corticosteroid therapy).

Data collection and processing

Each of the case records was reviewed by two independent reviewers. Data were recorded on a standardized data collection form. Discrepancies in the collected data were reconciled by a third reviewer (NP). Attempts were made to obtain missing data of patients transferred from peripheral hospitals from their respective health records departments via fax and mail. Records were reviewed for patient demographic data, historical and physical findings, biochemical variables, treatment parameters and outcomes.

The door-to-antibiotic time was defined as the time elapsed (in minutes) between the recorded time of presentation to the emergency room triage nurse to the time of the first administered dose of appropriate antibiotics. Appropriate empiric antibiotic treatment for meningitis was defined as a parenteral antibiotic regimen administered for the purpose of treating meningitis, whose choice and dose conformed with the recommended treatment guidelines at that time,8 with appropriate adjustments made in cases where a specific pathogen was isolated. Fever was defined as a temperature at presentation ⩾37.5°C. Patient mental status was categorized as: (1) alert and oriented; (2) lethargic or confused; (3) responsive to pain only; or (4) coma. Patients categorized as (2), (3) or (4) were considered to have ‘impaired mental status’ and those categorized as (3) or (4) were considered to have ‘severely impaired mental status’. A non-revealing lumbar puncture was defined as having a cerebrospinal fluid white blood cell count <10/mm3. Partial meningitis treatment was defined as either administration of an intravenous antibiotic >6 h or oral antibiotic >12 h prior to lumbar puncture, or the final dose of an antibiotic regimen administered within 5 days prior to lumbar puncture.

Sixteen biologically plausible potential predictors of mortality were selected for analysis after review of the medical literature but before study initiation.2,7,,9 The chosen covariates were age >60 years, male sex, severely impaired mental status at presentation, the occurrence of any seizure, afebrility at presentation, the causative micro-organisms Streptococcus pneumoniae or Staphylococcus aureus, prolonged fever after treatment initiation (>10 days), other focus of infection, corticosteroids given, and a cerebrospinal fluid white blood cell count <100/mm3. An intensive care unit admission ⩾3 days was chosen to represent a group of previously reported predictors of mortality including shock, mechanical ventilation and acid-base disturbance. Insufficient data were available for the use of the standard APACHE score. Finally, the following antibiotic timing parameters were chosen as potential covariates: a door-to-antibiotic time >6 h, a door-to-antibiotic time>the median value of the study, the diagnostic–treatment sequence CT of the head followed by lumbar puncture followed by antibiotics, and any combination of diagnostic-treatment sequences where antibiotics were not given first.

Eleven predetermined potential predictors of door to antibiotic time >6 h were selected, based on clinical experience. The potential covariates were age >60 years, the absence of headache, fever, or the meningitis triad (fever, neck stiffness, impaired mental status) at presentation, a non-revealing lumbar puncture, partial meningitis treatment, other focus of infection, failure to administer adequate antibiotics prior to a transfer from a peripheral hospital, and the following diagnostic–therapeutic sequences: (1) CT of the head followed by lumbar puncture followed by antibiotics, (2) lumbar puncture before antibiotics, and (3) CT of the head before antibiotics.

Outcome measure

The primary outcome was in-hospital death. Death was further defined as category I (meningitis was the underlying and immediate cause of death), category II (meningitis was the underlying but not the immediate cause of death) and category III (meningitis was neither the underlying nor the immediate cause of death). Criteria for these categories are outlined elsewhere.10

Statistical analysis

Data presentation was mainly descriptive, with frequencies reported as a percentage of meningitis cases. Univariate logistic regression assessed the association between single potential covariates and mortality. Variables with a univariate significance of p<0.10 were entered in a multivariate forward conditional logistic regression analysis, and those retaining a p<0.05 were included in the final model. A case was excluded from the individual analysis when relevant data was missing. Only death defined as category I or II was considered as a dependent variable.

Univariate logistic regression assessed the association between potential covariates and a door-to-antibiotic time >6 h. Results are reported as odds ratios with 95%CIs. To assess the incremental association between antibiotic administration delay and mortality, case fatality rates were calculated according to increasing door to antibiotic time intervals.

Other comparisons were made using the Mann-Whitney test and analysis of variance, as appropriate. All statistical analysis was performed using SPSS 11.5 (SPSS Inc).

Results

A total of 271 possible cases of bacterial meningitis were identified. Of these, 56 were excluded due to the presence of either an intraventricular neurosurgical prosthesis (45 cases), or a proven viral, fungal or mycobacterial aetiology (11 cases). Of the remaining 215 cases, 92 could not be included on the basis of age <16 years (38 cases), lumbar puncture not performed (26 cases) and not meeting at least one of the four cerebrospinal fluid result criteria (28 cases). Hence 123 cases of meningitis in 119 patients were retained. The final distributions of patients meeting each of the inclusion criteria were: (i) a positive cerebrospinal fluid culture (72 cases); (ii) a negative cerebrospinal fluid culture with neutrophilic pleocytosis and at least one of: positive blood cultures (19 cases), positive cerebrospinal fluid gram stain (8 cases), cerebrospinal fluid high protein and low glucose (19 cases), positive cerebrospinal fluid bacterial antigen test (3 cases); (iii) a negative cerebrospinal fluid culture without neutrophilic pleocytosis yet with positive blood culture and cerebrospinal fluid high protein and low glucose (one case); (iv) autopsy confirmation (one case).

Patient characteristics of the 123 cases are shown in Table 1. Forty-nine patients (40%) were transferred from a peripheral hospital, with the most common reason for transfer being unavailability of a CT of the head (22/49 cases, 45%). The reason for transfer was not indicated in 43% of cases. Transferred patients were more likely to be of male gender (p<0.0001), aged <60 (p = 0.001) and have subsequent cultures growing Neisseria meningitidis (p = 0.03). There was no significant difference in presentation mental status between these groups. The case fatality rate in the non-transferred patient group and transferred patient group was 16% vs. 14%, respectively (p = 0.84). The overall case fatality rate of the series was 13% (16/123 cases over 12 years). All deaths were attributed to meningitis as category I (11 cases) or category II (5 cases). The case fatality rate across the study period was consistent; no significant differences in rates between 3-year subgroups were identified.

View this table:
Table 1

Patient characteristics in 123 cases of adult acute bacterial meningitis

n (%)
Predisposing factor
Sinusitis25 (20)
    Recent*17 (14)
    Remote*8 (7)
Otitis media24 (20)
    Recent*16 (13)
    Remote*8 (7)
Neurosurgery19 (15)
    Recent*9 (7)
    Remote*10 (8)
Cerebrospinal fluid leak17 (14)
Immunosuppressant drug16 (13)
Cancer13 (11)
Head injury12 (10)
    Recent*3 (2)
    Remote*9 (7)
Diabetes mellitus11 (9)
Alcoholism10 (8)
Splenectomy7 (5)
Clinical features at presentation
Mean age in years (range)54 (19–86)
Male77 (63)
Peripheral hospital transfer49 (40)
Fever (⩾37.5°C)114 (93)
Neck stiffness95 (77)
Impaired mental status**104 (85)
Headache98 (80)
Nausea or vomiting77 (63)
Neck pain33 (27)
Ear findings***32 (26)
Photophobia31 (25)
Kernig's sign22 (18)
Brudzinski's sign16 (13)
  • Percentages may not equal their related sum because of rounding. *‘Recent’ denotes within 1 month of meningitis onset, ‘remote’ >1 month before meningitis onset. **Defined as lethargic or confused, responsive to pain only or coma. ***Includes ear pain, otorrhea or examination findings of otitis media.

The main predisposing factors for meningitis included sinusitis (20%), otitis media (20%), past neurosurgery (15%) and cerebrospinal fluid leak (14%). The most common presenting clinical features were fever (93%), impaired mental status (85%), headache (80%) and neck stiffness (77%). The meningitis triad (fever, neck stiffness and impaired mental status) was found in 63% of cases. Blood cultures were positive in 53% of cases (62/118 cultures drawn). The most commonly isolated pathogens were Streptococcus pneumoniae (56%), Staphylococcus aureus (9%), Neisseria meningitidis (7%), Haemophilus influenzae (7%) and Listeria monocytogenes (7%). There were only two cases of meningitis due to S. pneumoniae that was incompletely susceptible to penicillin (both with MICs of 1.0 mg/l) and there were no cases due to methicillin-resistant S. aureus. Predisposing factors for meningitis due to S. aureus were immunosuppressant drugs (3 cases), diabetes mellitus (2 cases), sinusitis (2 cases), mastoiditis (1 case), past neurosurgery (2 cases) and cerebrospinal fluid leak (1 case). One case of associated endocarditis was found.

The door-to-antibiotic time was available for 118/123 cases (96%). The median was 3.8 h (IQR 1.4–6.1). Door-to-antibiotic time was >6 h in 38 cases (32%) and >12 h in 15 cases (13%). Two patients died in the latter group. Antibiotics were administered 53% of the time prior to transfer from another hospital. The door-to-antibiotic time was significantly higher in those who did not receive antibiotics prior to transfer, compared to those who did (6.0 h vs. 2.4 h, p<0.001). The most commonly used antibiotic regimen was a third-generation cephalosporin plus penicillin or ampicillin (35% of cases), followed by one or more of a third-generation cephalosporin plus vancomycin (21% of cases) followed by a third-generation cephalosporin alone (16% of cases). Empiric antibiotic treatment was considered appropriate by type in 86% of cases and by dose in 83% of cases. Inappropriate antibiotic administration by type or by dose did not significantly predict mortality. The mean duration of intravenous antibiotic administration was 17.9 days (95%CI 15.0–20.8).

Table 2 illustrates the predictors of mortality. Univariate regression analysis identified afebrility at presentation, severely impaired mental status at presentation, a door-to-antibiotic time >6 h, age >60 years, meningitis due to S. aureus, treatment above the median door-to-antibiotic time (3.8 h) and the diagnostic–treatment sequence CT of the head followed by lumbar puncture followed by antibiotics as being significantly associated with mortality. Multivariate regression analysis identified afebrility at presentation, severely impaired mental status at presentation and a door-to-antibiotic time >6 h as being significant predictors of mortality. A door-to-antibiotic time >6 h independently conferred patients an 8.4× greater risk of mortality from meningitis (p<0.01).

View this table:
Table 2

Predictors of mortality in adult acute bacterial meningitis using logistic regression analysis

VariableNo. of casesCrude OR (95%CI)Adjusted OR (95%CI)
TotalAliveDead
Afebrility at presentation83515.6 (3.3–74.3)**39.4 (4.3–358.1)**
Severely impaired mental status at presentation181264.8 (1.5–15.4)***12.6 (2.2–72.0)***
Door-to-antibiotic time >6 hours3828106.8 (1.9–23.4)***8.4 (1.7–40.9)***
Age >60 years5340137.3 (1.9–27.0)***
Staphylococcus aureus8448.6 (1.9–38.8)***
Door-to-antibiotic time >3.8 h (study median)5948114.3 (1.1–16.2)*
CT of the head followed by lumbar puncture, followed by antibiotics241863.8 (1.1–12.7)*
Seizure13943.6 (0.9–13.6)
Male sex7765121.9 (0.6–6.4)
Intensive care unit admission ⩾3 days332761.8 (0.6–5.4)
Antibiotics not first685991.5 (0.4–5.2)
Streptococcus pneumoniae524571.1 (0.4–3.1)
Corticosteroids given302641.1 (0.3–3.9)
Other focus of infection565060.7 (0.2–2.0)
Cerebrospinal fluid white blood cell count <100/mm3111010.6 (0.1–5.4)
Prolonged fever (>10 days)6600.0
  • Statistical significance: *p<0.05; **p≤0.001; ***p<0.01. Responsive to pain only or coma. Any combination of diagnostic-treatment sequences where antibiotics were not given first.

Figure 1 illustrates the case fatality rate according to door to antibiotic time intervals. The case fatality rate increases sharply at the door to antibiotic time interval >6–8 h and continues to increase at the higher time interval.

Figure 1.

Case fatality rate according to door-to-antibiotic time interval in adult acute bacterial meningitis.

Figure 2 demonstrates the relative frequency of diagnostic–treatment sequences. The most common sequence was antibiotics followed by CT of the head followed by lumbar puncture (32%). The second most common sequence was CT of the head followed by lumbar puncture followed by antibiotics (22%). Unsurprisingly, door-to-antibiotic time was significantly higher when patients did not receive antibiotics first in the diagnostic-treatment sequence, compared to those who did (5.9 h vs. 2.3 h, p<0.001).

Figure 2.

Relative frequency (%) of diagnostic–treatment sequences in adult acute bacterial meningitis. The order in which the interventions are presented represents the order in which they were performed. Total percentages do not equal 100 because of rounding.

Table 3 illustrates the predictors of door-to-antibiotic time >6 h. Univariate regression analysis identified the variables: no antibiotic prior to transfer; the diagnostic–treatment sequence CT of the head followed by lumbar puncture followed by antibiotics; partial meningitis treatment; and absence of the meningitis triad, as significant predictors.

View this table:
Table 3

Predictors of door-to-antibiotic time >6 h: univariate logistic regression analysis

VariableNo. of casesCrude OR (95%CI)p
TotalDoor-to-antibiotic time ⩽6 hDoor-to-antibiotic time >6 h
No antibiotics given prior to peripheral hospital transfer21111021.8 (2.5–192.2)0.005
CT of the head followed by lumbar puncture followed by antibiotics249155.6 (2.1–14.7)<0.001
Partial meningitis treatment*6252100.2 (0.1–0.4)<0.001
Absence of meningitis triad** at presentation4219234.9 (2.1–11.3)<0.001
Absence of fever (≥37.5°C) at presentation7342.9 (0.6–14.1)0.17
Absence of headache at presentation2312112.3 (0.9–5.9)0.08
Age >60 years5130212.1 (0.9–4.5)0.07
CT of the head before antibiotics***191181.8 (0.6–4.8)0.28
Lumbar puncture before antibiotics2515101.7 (0.7–4.2)0.29
Other focus of infection5439150.7 (0.3–1.5)0.35
Non-revealing lumbar puncture151230.5 (0.1–1.8)0.26
  • *Administration of an intravenous antibiotic >6 hours or oral antibiotic >12 hours prior to lumbar puncture or the final dose of an antibiotic regimen administered within 5 days prior to lumbar puncture. **Triad includes fever, neck stiffness, impaired mental status. ***Any combination of diagnostic–treatment sequences whereby CT of the head was performed before antibiotics. Any combination of diagnostic–treatment sequences whereby lumbar puncture was performed before antibiotics. Cerebrospinal fluid white blood cell count <10/mm3.

Discussion

The management of suspected adult acute bacterial meningitis requires early administration of antibiotics. Although it has been suggested that delays in the administration of antibiotics may be associated with adverse clinical outcome, to our knowledge no study to date has clearly demonstrated this association.4–7

This study provides compelling evidence that delays in the administration of antibiotics are associated with death in adult acute bacterial meningitis. In the multivariate logistic regression analysis, a delay of >6 h in the administration of antibiotics after presentation independently conferred an 8.4-fold greater risk of death from meningitis. Furthermore, the effect of treatment delay on case fatality rate was incremental; greater delays resulted in higher case fatality rates.

Afebrility at presentation was the most significant predictor of mortality. Data collection did not include the distribution of temperatures in this subgroup; therefore the analysis could not identify those with hypothermia, a previously recognized predictor of mortality in sepsis.11 It is possible that the association between mortality and hypothermia may be even stronger than that with afebrility at presentation. In contrast to recent evidence,12 the use of corticosteroids in this study was not associated with a significant reduction in case fatality rates. This finding may be explained by the infrequent use of corticosteroids (25%) in our study, which was completed just prior to publication of a large randomized control trial.12 Other explanations relate to variability observed over the 12-year period in the corticosteroid type, route, dose, duration of treatment and timing relative to antibiotic administration. Alternatively, corticosteroids may have been used preferentially in patients with poor prognostic factors.

Significant delays in the administration of antibiotics has been assessed in other studies;4,,13 in this study, we demonstrated that such delays may be explained, at least in part, by inappropriate diagnostic–treatment sequences. We were alarmed to find that the second most common diagnostic–treatment sequence was CT of the head followed by lumbar puncture followed by antibiotics, a sequence found to confer a 5.6-fold increased likelihood of a door to antibiotic time >6 h. This observation suggests that some physicians prioritize accuracy of diagnosis over the immediate administration of antibiotics.

The perceived need for a CT of the head was a common reason for transfer to our institution, with only 53% of patients receiving antibiotics prior to such transfers. Other studies have reported similar low rates, all below 50%.5,14,,15 In this study, failure to administer antibiotics prior to transfer was a strong predictor of a door to antibiotic time >6 h (OR 21.8). This may be a point of intervention; since transfers between hospitals usually involve a conversation between transferring and accepting physicians, the administration of antibiotics prior to transfer can be ensured.

The prevalence of CT findings in bacterial meningitis that would contraindicate lumbar puncture is low; in one study, this was the case in only 2.7% of 113 patients.16 Moreover, in a recent prospective study enrolling 301 adults, the absence of certain baseline clinical features allowed clinicians to identify patients who were unlikely to have abnormal findings on a CT of the head, implying a low risk of herniation from lumbar puncture.17 This has led some to question the use of a CT of the head as a screening tool to identify patients at risk of herniation from lumbar puncture. Regardless of this point, the decision to perform a CT of the head before lumbar puncture should not prevent the immediate administration of antibiotics.

One perceived drawback with the immediate administration of antibiotics is that it may prevent isolation of the causative organism from the cerebrospinal fluid drawn on a subsequent lumbar puncture. However, despite antibiotic administration, investigations may still support a diagnosis of bacterial meningitis, often with an aetiological diagnosis. First, cerebrospinal fluid Gram stain or culture may be positive despite recent antibiotic administration.18,,19 Second, one or two doses of antibiotics will not significantly alter the presence of cerebrospinal fluid pleocytosis, elevated protein and reduced glucose sufficient to make the diagnosis of bacterial meningitis.18,,20 Finally, although the cerebrospinal fluid cultures may be negative, it is still possible to make an aetiological diagnosis from positive blood cultures drawn before initiation of antibiotics.18,,21 Clearly, the benefits of immediate administration of antibiotics in this setting greatly outweigh the potential drawbacks.

When considering the above findings, it is important to appreciate the limitations intrinsic to retrospective chart reviews and, as was the case in this study, a relatively small number of deaths, leading to wide confidence intervals. Although a broad series of cases was obtained using all related ICD-9 diagnostic codes for bacterial meningitis, the study remains subject to incomplete case identification. Despite efforts to obtain missing data from the health records departments of patients transferred from peripheral hospitals, completeness could not always be assured. Although door-to-antibiotic time was available in 96% of cases, accuracy of this documentation by nurses cannot be assessed, although there is no reason to suspect systematic bias. While taking into account the potential confounding effects of multiple independent covariates, the multivariate regression analysis identified a significant delay in the administration of antibiotics as an independent predictor of mortality. The possibility cannot be excluded, however, that this association may be weakened by another unrecognized covariate.

Although one could consider this study as hypothesis-generating due to its limitations, deploying efforts to further substantiate the findings would have to be balanced against the compelling results of this study. A door-to-antibiotic administration delay of >6 h was independently associated with mortality. An incremental time to treatment vs. mortality relationship was demonstrated. Moreover, the hypothesis is biologically plausible, and there are minimal drawbacks to administrating antibiotics early in suspected bacterial meningitis. Emergency room pathways assuring immediate antibiotic administration need to be developed and implemented for patients in whom bacterial meningitis is suspected.

Acknowledgments

Our thanks to Dr Dean Ferguson and Dr Amit Garg for their statistical advice. This study was presented as a slide presentation at the Interscience Conference on Antimicrobial Agents and Chemotherapy, Chicago IL, September, 2003.

References

View Abstract