Fever Interval before Diagnosis, Prior Antibiotic Treatment, and Clinical Outcome for Young Children with Bacterial Meningitis

MAJOR ARTICLE Fever Interval before Diagnosis, Prior Antibiotic Treatment, and Clinical Outcome for Young Children with Bacterial Meningitis Bema K. Bonsu 1 and Marvin B. Harper 2 1 Department of Medicine, Division of Emergency Medicine, and 2 Division of Infectious Diseases, Children s Hospital, Boston In young children, meningitis due to Streptococcus pneumoniae is preceded by a long interval from onset of fever to diagnosis of bacterial meningitis (hereafter known as fever interval ), during which time the patient frequently contacts a clinician. By means of retrospective chart review, we compared the fever interval that preceded diagnosis with the complication rate among 288 young children (age, 3 36 months) who had bacterial meningitis (1984 1996), as stratified by causative organism and prior antibiotic treatment. Pathogens included S. pneumoniae, Haemophilus influenzae type b, and Neisseria meningitidis. Pneumococcus species were associated with the longest fever interval prior to diagnosis of meningitis, the highest frequency of contact with a clinician before hospitalization, and the highest rate of documented morbidity or mortality. For S. pneumoniae, there was an association between antibiotic treatment received at prior meetings with a clinician and a reduced rate of meningitis-related complications (odds ratio, 0.14; P p.02). Antibiotic treatment during such meetings is associated with a substantial reduction in disease-related sequelae. Bacterial meningitis is a serious bacterial infection that may lead to lifelong debility or death [1]. In the pediatric age group, the highest incidence of bacterial meningitis is among children!3 years of age [2, 3]. The diagnosis of meningitis in such children is often preceded by an interval of fever that can be either brief or long. When this interval is long, there is a greater opportunity to visit with or contact a clinician prior to the time of diagnosis. If such visits are made, children may or may not receive antibiotics for unrelated minor Received 10 February 2000; revised 6 July 2000; electronically published 9 February 2001. Presented in part: Annual scientific meeting of the Pediatric Academic Societies (American Pediatric Society and the Society for Pediatric Research), New Orleans, on 4 May 1998 (abstract 371). This study was approved by the Institutional Review Board at The Children s Hospital, Boston. Reprints or correspondence: Dr. Bema K. Bonsu, Children s Hospital, Division of Emergency Medicine, 700 Children s Dr., Columbus, OH 43213 (Bonsub @pediatrics.ohio-state.edu). Clinical Infectious Diseases 2001; 32:566 72 2001 by the Infectious Diseases Society of America. All rights reserved. 1058-4838/2001/3204-0007$03.00 infections. For children who develop meningitis in spite of having received antibiotic treatment at an earlier visit, it is unclear whether such treatment has an impact on the rate of complications noted at the time of discharge from the hospital. Our study was conducted to estimate the length of the interval from onset of fever to diagnosis (hereafter known as fever interval ) in a sample of young children with bacterial meningitis, and to determine whether this interval varies by causative bacteria. In addition, we sought to evaluate the rate of contact with a clinician during this fever interval and to determine whether there was an association between the administration of antibiotics during such visits and the frequency of complications from meningitis caused by different bacteria. PATIENTS AND METHODS Patient selection. We conducted a retrospective chart review of all children 3 36 months of age with diagnosed acute bacterial meningitis (ABM) who were 566 CID 2001:32 (15 February) Bonsu and Harper

discharged from our institution, Children s Hospital, Boston, a large urban pediatric teaching hospital, from 1 January 1984 through 31 December 1996. Children were included in the review if they had ABM, which is defined by any of the following criteria: (1) CSF culture that tests positive for Haemophilus influenzae (type b or other), Streptococcus pneumoniae, Neisseria meningitidis, group B Streptococcus, gram-negative enteric rods, or Listeria monocytogenes; (2) a culture of blood samples that is positive for 1 of the aforementioned pathogens, along with a CSF WBC count of 10 cells/mm 3 and a CSF RBC count of 1000 cells/mm 3 ; or (3) a positive result of CSF latex agglutination assay or Gram stain, along with a CSF WBC count 500 cells/mm 3 and a CSF RBC count of 1000 cells/mm 3. Children were excluded if they had immunologic, complex cardiac, genetic, metabolic, neurological, or neurosurgical disease diagnosed prior to the diagnosis of ABM. Data abstraction. Data abstracted from charts included the organism that had caused meningitis, the age of the child, the highest recorded temperature before diagnosis (excluding the temperature noted at the time of diagnosis), the duration of fever from onset of illness to the time of diagnosis, occurrence or nonoccurrence of a clinician visit or contact before diagnosis, and antibiotic use during the illness that led to the diagnosis of meningitis. Fever was defined as any mention of fever or a temperature of 38 C that was documented in the chart during the illness that led to the diagnosis of ABM. Measured temperatures were standardized to rectal temperatures by the addition of 0.5 C to axillary or oral temperatures. Our primary independent variable was the type of pathogen that had caused bacterial meningitis. Our secondary independent variable was the use of an antibiotic prior to and within 7 days of the clinician visit at which meningitis was diagnosed. Dependent variables included the duration of fever from the onset of illness to the initial diagnosis of meningitis (the fever interval), the occurrence of a prior visit or contact with a clinician, and the complications noted at the time of the patient s discharge from the hospital. This information was abstracted as recorded in the charts. We read the discharge summary, the emergency record(s), the transfer notes, the consultation notes, the intake notes of the primary resident or medical student, and the nursing notes. The entry that we considered to have the most comprehensive and quantitatively specific data was utilized. Children were considered to be afebrile prior to the diagnosis of ABM if this was specifically stated in the chart or if no mention of either fever or a temperature of 38 C was noted during 2 separate reviews of the chart. Patients were excluded from review if the notes on the charts were unclear or if the most comprehensive entry contained inconsistencies with respect to prior antibiotic use, or status or duration of fever. Prior clinician contact was defined as a visit or other contact with a health care provider that occurred during the time of the illness (but prior to the diagnosis of ABM), and that resulted in an unrelated diagnosis (i.e., not meningitis) being made. For children who had prior clinician visit(s), the duration of fever at the time of the first visit, the time from the first visit to the diagnosis of ABM, and the number and site of these visits were recorded. If antibiotic therapy that a child was already undergoing was not discontinued, or if new antibiotic therapy was initiated by a clinician during the course of the illness that led to the diagnosis of meningitis, the type of antibiotic and the route of administration (oral vs. parenteral) were noted. The terms penicillins and cephalosporins generically were used to refer to any antibiotics within these classes of antimicrobial drugs. Any child who received an antibiotic for treatment of an unrelated infection within 7 days of the diagnosis of ABM was categorized as having pretreatment. A complication was defined as death or an adverse neurological or audiologic outcome that was noted at the time of discharge from the hospital and that was attributable to the episode of meningitis. Charts on which such information was incomplete or absent were flagged and were excluded from any analysis of complications. Statistical analysis. All analyses were performed by use of STATA software, release 5.0 (STATA). Continuous variables were summarized by computation of the mean and/or median (with interquartile range). CIs were calculated for proportions of categorical variables. We used the x 2 test or Fisher s exact test to compare proportions of categorical variables; the Mann- Whitney log-rank test, to compare medians; and Student s t- test, to compare continuous variables. We used analysis of variance (ANOVA) to compare variables when dealing with 3 or more groups. RESULTS Patients. A total of 288 children met the study criteria for diagnosis of meningitis. The median patient age was 10 months, with an interquartile range of 6 19 months (table 1). Ninetytwo children (32%) were 3 6 months old, 85 (30%) were 7 12 months old, 73 (25%) were 13 24 months old, and 38 (13%) were 25 36 months old. Causative organism. The diagnosis of ABM was established by a positive result on culture of CSF samples for 258 (90%) of the 288 children, by a positive result on blood culture and a CSF WBC count of 10 cells/mm 3 for 22 (8%), by positive result on Gram stain and a CSF WBC count of 500 cells/mm 3 for 4 (1%), and by a positive result on latex agglutination assay and a CSF WBC count of 500 cells/mm 3 for 4 (1%). H. influenzae type b caused 154 (53%) of the 288 cases of meningitis, S. pneumoniae caused 69 (24%), and N. meningitidis was responsible for 59 (20%). In addition, 5 cases (2%) were caused by group B streptococci, 1 case was caused by Salmonella, and 1 case was Bacterial Meningitis in Young Children CID 2001:32 (15 February) 567

Table 1. Comparison of the features and disease characteristics of and outcomes for 288 children with acute bacterial meningitis, according to the types of organism that caused meningitis. Feature Streptococcus pneumoniae (n p 68) Children with ABM caused by Haemophilus influenzae type b (n p 154) Neisseria meningitidis (n p 59) Other organisms (n p 7) Children with ABM due to all causes (n p 288) a Age, median mo (interquartile 7 (5 13) 11 (7 18) b 15 (9 26) b 3.5 (3.25 4) 10 (6 19) range) Median temperature, C (interquartile range) 40.0 (39.4 40.3) 40.0 (39.4 40.4) 39.9 (39.4 40.0) 40 (39.4 40.6) 40.0 (39.4 40.3) Fever interval 1 d c 56/64 (88) 105/149 (70) b 45/58 (78) 3/5 (60) 209/276 (76) Clinician contact 0 visits, no. (%) 0 (0) 4 (3) 1 (2) 0 (0) 5 (2) 1 visit, no. (%) 33 (48) 67 (44) 24 (41) 1 (14) 125 (43) 1 clinician visit and/or phone contact, no. (%) 33 (49) 71 (47) 25 (43) 1 (14) 130 (45) 2 visits, no. (%) 8 (11) 16 (10) 3 (5) 0 (0) 27 (9)!1-d interval between onset of fever and first visit d 23/33 (70) 37/67 (55) 20/24 (83) 0/1 (0) 80/125 (64) 1-d interval between first visit and diagnosis of ABM d 25/33 (76) 47/67 (70) 12/24 (50) b 1/1 (100) 85/125 (68) Pretreatment with antibiotics, no. (%) 21 (32) 48 (31) 8 (14) b 1 (14) 78 (27) Pretreatment with parenteral antibiotic 5/21 (24) 8/48 (17) 1/8 (13) 0/1 (0) 14/78 (18) Mortality and morbidity, e no. (%) 31/55 (56) 47/134 (35) b 5/44 (11) b 1/4 (25) 84/237 (35) Death, no. (%) 2 (3) 4 (3) 0 (0) 0/7 (0) 6/288 (2) NOTE. Data are n/n (%) of patients, unless otherwise indicated. ABM, acute bacterial meningitis; fever interval, period from onset of fever to diagnosis of ABM. a Includes 7 other bacterial pathogens: group B streptococci (5), Salmonella species (1), and nontypeable H. influenzae (1). b Statistically significant when compared with S. pneumoniae. c After exclusion of children whose records had no mention of fever. d After exclusion of the 5 children who made contact with but did not visit a clinician. e Children with full documentation of complications. caused by nontypeable H. influenzae. The age of the children varied according to the type of organism that caused meningitis ( P p.0001). Median ages, as related to the type of organism that caused meningitis, are presented in table 1. Height of fever prior to diagnosis. Twelve (4%) of 288 children either had no fever documented on 2 separate chart reviews or were specifically noted to have been afebrile. Sixtytwo children (21%) were described simply as having had fever (59 children) or high fever (3 children) prior to the diagnosis of ABM. A numeric temperature was documented on 214 charts (74%) during the illness that led to diagnosis. Of these temperatures, 212 (99%) were 38 C (range, 37.8 C 41.7 C). Thirty-eight (18%) of these 212 temperatures were between 38 C and 38.9 C, 59 (28%) were between 39 C and 39.9 C, and 115 (54%) were 40 C. Median temperatures (table 1) were not statistically significant when stratified by the type of organism ( P p.55). Duration of fever at the time of diagnosis. After the exclusion of children without fever, we determined that 67 (24%) of 276 children had fever for!1 day at the time of diagnosis. A total of 176 children (64%) had fever that lasted for 1 3 days (1 day for 84 patients, 2 days for 55; and 3 days for 37) and 33 (12%) had fever that lasted 13 days. Stratification done according to type of organism showed that S. pneumoniae (figure 1) was significantly more likely to have been associated with a fever interval of 1 day when compared with the rest of the sample (i.e., 56 of 64 patients vs. 153 of 212 patients [OR, 2.7; 95% CI, 1.2 5.9; P p.01]). This difference in the fever interval was statistically significant when S. pneumoniae was compared with H. influenzae type b (OR, 2.9; 95% CI, 1.3 6.5; P p.008), but not when it was compared with N. meningitidis (OR, 2; 95% CI, 0.7 5.2; P p.15; table 1). When adjustment was made for prior antibiotic treatment, S. pneumoniae was more likely to have been associated with a fever interval of 1 day at the time of diagnosis, in comparison with all other organisms (OR, 2.7; 95% CI, 1.1 6.1; P p.02). This difference was sta- 568 CID 2001:32 (15 February) Bonsu and Harper

Figure 1. Duration of fever prior to the diagnosis of meningitis in children with fever, stratified by the type of organism that caused meningitis. The asterisk indicates that P!.05 for comparison of the fever interval associated with meningitis caused by Streptococcus pneumoniae with that associated with other organism(s). tistically significant when S. pneumoniae was compared with H. influenzae type b (OR, 3.2; 95% CI, 1.3 7.7; P p.006), but not when it was compared with N. meningitidis (OR, 1.5; 95% CI, 0.6 4.2; P p.4). The recorded duration of illness among children with no documented fever ranged from 12 h to 14 days. The duration of illness was not recorded on 3 charts. Of the 9 remaining children, 6 (66%) had illness that lasted for 1 day. Rate and pattern of prior clinician visits. During the course of the illness that led to the diagnosis of meningitis, 130 (45%) of 288 children had contact with a clinician prior to the visit during which meningitis was diagnosed (table 1). Five (4%) of 130 such children did not visit a clinician, but it was documented that they had made contact with one. The remaining 125 children (96%) made 1 or more visits to a clinician that had resulted in an unrelated diagnosis being given. Of the 125 prior clinician visits, 89 (71%) were made at outpatient clinics or offices; 28 (22%), at emergency departments; 2, at home; and 6, at unspecified sites. Of the 125 children who made a prior clinician visit, 98 (78%) made 1 visit, 25 (20%) made 2 visits, 1 ( 1%) made 3 visits, and 1 ( 1%) made 5 visits. The longer the duration of fever preceding the diagnosis of ABM, the higher the probability that a child would have made contact with a clinician before the visit at which meningitis was diagnosed. Among children with no documented fever, 2 (17%) of 12 had contact with a clinician. For children with documented fever, 11 (16%) of 67 who had a fever interval of!1 day had contact with a clinician; conversely, when this interval was 1 day, 117 (56%) of 209 had contact. Therefore, among children whose charts documented fever, the OR of making prior contact with a clinician if the fever interval was 1 day versus!1 day was 6.5 (95% CI, 3.2 12.9; P!.0001). The rate of prior clinician contact in relation to the causative organism (not statistically significant) is shown in table 1. Antibiotic pretreatment and interval of fever before diagnosis. Antibiotics had been administered within the 7 days prior to the diagnosis of ABM in 78 (27%) of the 288 children (5 of 78 children did not visit a clinician but were prescribed an antibiotic over the phone). Fourteen (18%) of 78 children who were pretreated with an antibiotic received a parenteral antibiotic (6 received penicillins and 8 received cephalosporins). The remaining 64 children (82%) were administered oral antibiotics exclusively (penicillins were given to 38 patients; cephalosporins, to 13; trimethoprim-sulfamethoxazole, to 3; other, to 4; 11 antibiotic, to 4; and unspecified, to 2). All of the 4 children who were pretreated with 11 oral antibiotic (from 2 classes of antimicrobial agents) received as 1 of these drugs either a penicillin (3 children) or a cephalosporin (1 child). Of the children who were discharged after a prior clinician contact, 21 (64%) of 33 with meningitis caused by S. pneumoniae, 48 (68%) of 71 with meningitis caused by H. influenzae type b, 8 (32%) of 25 with meningitis caused by N. meningitidis, Bacterial Meningitis in Young Children CID 2001:32 (15 February) 569

and 1 of 1 with meningitis caused by group B streptococci (total, 78 [60%] of 130) were pretreated with an antibiotic. The proportion of children who received a penicillin or cephalosporin either alone or sequentially with another drug within 1 week of diagnosis of meningitis was 17 (81%) of 21 children with S. pneumoniae, 43 (90%) of 48 with H. influenzae type b, 8 (100%) of 8 with N. meningitidis, and 1 (100%) of 1 with group B streptococci. Eighty (64%) of the 125 children who visited a clinician had a fever interval of!1 day from onset of fever to the time of the first clinician visit. However, after their visit, children who were discharged to their homes after receiving antibiotics were more likely to have meningitis diagnosed at 1 day after their initial visit than were children who did not receive an antibiotic (59 of 73 children vs. 26 of 52 children; OR, 4.3; 95% CI, 1.9 9.3; P p.0003). Antibiotic pretreatment was associated with an increased likelihood of having an interval of 1 day from the first clinician contact to diagnosis; this was true for patients infected with all major organisms. However, although this increase was statistically significant for children with H. influenzae type b (35 of 44 pretreated children vs. 12 of 23 children who were not pretreated; OR, 3.6; P p.02) and N. meningitidis (6 of 7 pretreated children vs. 6 of 17 children who were not pretreated; OR, 11.0; P p.02), it was not for S. pneumoniae (17 of 21 pretreated children vs. 8 of 12 children who were not pretreated; OR, 2.1; P p.36). Antibiotic pretreatment and complications related to meningitis. A complete record of complications at the time of discharge from the hospital was available on 237 (82%) of 288 charts (55 [81%] of the 68 cases of S. pneumoniae, 134 [87%] of the 154 cases of H. influenzae type b, and 44 [75%] of the 59 cases of N. meningitidis had full documentation of complications). Thirty-one (11%) of the 288 charts were incomplete (auditory examination findings were pending), and 20 (7%) of 288 charts had no record of the clinical status at the time of discharge. Of the 130 children who had prior contact with a clinician, a complete record of complications noted at the time of discharge from the hospital was available for 110 (83%). The 20 remaining charts included 10 with no comment on clinical outcome (for 3 patients with S. pneumoniae, 4 with H. influenzae type b, 2 with N. meningitidis, and 1 with group B Streptococcus) and 10 that described neurological findings as normal; for these latter 10 patients, no audiologic evaluation had been done at the time of discharge from the hospital (for 3 patients with S. pneumoniae, 5 with H. influenzae, and 2 with N. meningitidis). Sixty-seven (61%) of 110 children with completely documented complications had received an antibiotic prior to being given a diagnosis of meningitis (table 2). When charts with incomplete or no documentation of outcome were excluded, there was no statistically significant difference in the rate of complication between children who received versus those who didn t receive antibiotic pretreatment (20 of 67 patients vs. 18 of 43 patients, respectively; OR, 0.59; 95% CI, 0.27 1.3; P p.19). However, after stratification was done according to the type of causative bacteria (table 2), we determined that children with S. pneumoniae meningitis who contacted a clinician and who received pretreatment with an antibiotic were less likely to develop complications (death, or neurological or audiologic sequelae) than were children who had clinician contact but who did not receive an antibiotic (6 of 17 patients vs. 8 of 10 patients, respectively; OR, 0.14; 95% CI, 0.02 0.79; P p.02). For children with meningitis caused by H. influenzae type b (OR, 0.69) and N. meningitidis (OR, 0), the probability of development of complications, if the child had previously contacted a clinician, was lower if the child had received pretreatment, but this was not statistically significant (table 2). When pretreated children were compared with the group that was comprised of all children who did not previously receive an antibiotic (children who did not have a prior clinician contact, and children who had a prior clinician contact but who did not receive antibiotic pretreatment), pretreatment was still associated with a statistically significant decrease in the rate of complications if the organism was S. pneumoniae (6 of 17 pretreated patients vs. 25 of 38 patients who did not receive pretreatment; OR, 0.28; 95% CI, 0.09 0.91). Similar comparisons for children with meningitis caused by H. influenzae type b and N. meningitidis were not statistically different. When all charts were included, there was no statistically significant difference in the mortality rate when stratification was done according to the type of organism that caused meningitis. Irrespective of the type of bacteria that caused meningitis, there was no difference in the mortality rate if meningitis was diagnosed at some interval of time after the first clinician visit (regardless of antibiotic pretreatment), rather than at the initial contact (table 2). DISCUSSION Our study shows that infection with S. pneumoniae and H. influenzae type b is associated with a long duration of fever and a high rate of clinician contact prior to the diagnosis of meningitis. The interval before diagnosis of meningitis is longer for children who were seen by a clinician and discharged to home after having received an antibiotic for an unrelated infection than it was for those who did not receive an antibiotic; however, this did not increase the rate of complications. If the causative organism was S. pneumoniae, such pretreatment was associated with fewer complications attributable to meningitis. Several studies [4 7] have reported the duration of illness and the rate of contact with a clinician prior to diagnosis of ABM in children. Previous investigators have described a gen- 570 CID 2001:32 (15 February) Bonsu and Harper

Table 2. Rate of complications associated with meningitis, as related to prior contact with a clinician and antibiotic use after such clinician contact, in 288 children with bacterial meningitis. Organism that caused meningitis, complication Proportion of patients with complications Meningitis diagnosed at first clinician contact and no prior antibiotic use Meningitis not diagnosed at first clinician contact (n p 110) Did not receive an antibiotic Received an antibiotic Total no. of patients evaluated for complication(s) Streptococcus pneumoniae Any 17/28 (61) 8/10 (80) a 6/17 (35) a 55 Death 1/35 (3) 1/12 (8) 0/21 (0) 68 Haemophilus influenzae type b Any 25/73 (34) 8/19 (42) 14/42 (33) 134 Death 1/83 (1) 2/23 (9) 1/48 (2) 154 Neisseria meningitidis Any 3/23 (13) 2/14 (14) 0/7 (0) 44 Death 0/34 (0) 0/17 (0) 0/8 (0) 59 Other organisms Any 1/3 (25) 0/0 (0) 0/1 (0) 4 Death 0/6 (0) 0/0 (0) 0/1 (0) 7 All organisms Any 46/127 (36) 18/43 (42) 20/67 (30) 237 Death 2/158 (1) 3/52 (6) 1/78 (1) 288 NOTE. Data are n/n (%) of patients, unless otherwise indicated. When all complications were analyzed, only patients with a complete record of complications at discharge were included. For analysis of mortality, all patients were included. a P!.05, for comparison of the rate of complications in children who received antibiotic treatment at a prior clinician contact versus those who did not. erally short duration of fever before the diagnosis of bacterial meningitis in children. Kilpi et al. [4, 7] reported that 141 (49%) of 286 children with meningitis (age range, 3 months to 15 years) had symptoms that lasted 24 h before the time of diagnosis. Similarly, Davis et al. [5] reported that 220 (47%) of 468 children (unspecified age range) with H. influenzae type b meningitis had symptoms that lasted 24 h before the time of diagnosis. According to the study by Kilpi et al. [4, 7], however, the group with the longest duration of symptoms (148 h) tended to be younger than the groups with symptoms of intermediate duration (124 48 h) or short duration ( 24 h; mean ages, 1.6, 3.5, and 3.2 years, respectively). This is in agreement with our finding of a generally long period of fever that precedes diagnosis of bacterial meningitis in children 3 years of age. In our study, S. pneumoniae was especially likely to be associated with a long fever interval; in fact, when children with no documented fever were excluded from the study, only 12% of cases of pneumococcal meningitis were rapid in onset (i.e., diagnosed!24 h after onset of fever). It has been reported that between one-quarter and one-half of all children who develop ABM have visited a clinician prior to having ABM diagnosed [8, 9]. In our study, 45% of young children sought care from a clinician prior to the diagnosis of meningitis, and 64% of the children who visited a clinician did so within 24 h of the onset of fever. The factors that determine if and when care is sought are complex and are not addressed by our study. Kallio et al. [8] reported that between 44% and 58% of patients had received prior antibiotic treatment, depending on the length of time between the last contact with a clinician and the visit at which a diagnosis of meningitis was made. Other studies have reported a total rate of antibiotic use (not restricted to patients who visit a clinician) of up to 40% of patients. The overall rate of patients who had prior antibiotic use in our study was 27%. However, among children who had a visit/ contact with a clinician, 63% of those with meningitis due to S. pneumoniae, 68% of those with meningitis due to H. influenzae type b, and 32% of those with meningitis due to N. meningitidis had received prior antibiotic treatment (overall rate of pretreatment, 60%). The results presented in previous reports [7, 9 12] conflicted with regard to the effect of outpatient antimicrobial use (preceding the diagnosis of meningitis) on the rate of long-term morbidity and mortality. Kilpi et al. [7] found no association between the use of antimicrobial pretreatment and the severity of meningitis. Likewise, Winkelstein [9] and Jarvis and Saxena [10] found no difference in the outcomes of patients who had Bacterial Meningitis in Young Children CID 2001:32 (15 February) 571

meningitis when they compared children who were previously treated with antibiotics to those who were not previously treated. Davis et al. [11] have found a trend toward lower mortality rates among children who were pretreated with antibiotics. None of these investigators performed a subanalysis of their results according to the type of organism that caused meningitis. Kaplan et al. [12] studied children with meningitis caused by H. influenzae type b and found that oral antibiotics were associated with a worse outcome and a longer duration of illness before definitive therapy was initiated. Our data showed no statistically significant difference in the outcome for pretreated versus nonpretreated children with meningitis caused by H. influenzae type b or N. meningitidis. However, when the causative organism was S. pneumoniae, antimicrobial pretreatment had a substantial benefit. The reasons for this substantial treatment effect are unclear. We are unaware of any previous studies that have specifically investigated the effect of prior treatment on the clinical outcome of children with meningitis due to S. pneumoniae. Our findings have special relevance for young children immunized against H. influenzae type b, since S. pneumoniae is presently the predominant cause of meningitis in this age group. S. pneumoniae is also the target of treatment-management strategies directed toward the prevention of serious bacterial infections, through the empirical use of antibiotics, in highly febrile children who appear well. Although debate regarding the usefulness of these strategies continues, it is probable, on the basis of our findings, that children benefit from empirical antibiotic therapy, even if such treatment fails to prevent meningitis or is only partially effective against early-onset meningitis. Furthermore, clinicians who prescribe antibiotics to young children for appropriate medical diagnoses need not be concerned that such treatment will worsen clinical outcome if pneumococcal meningitis should ensue. Our study is limited by the accuracy with which the duration of fever was recorded. Since data were abstracted retrospectively, it is difficult to evaluate the validity of the fever intervals as recorded in the charts. The frequency of clinician visits prior to the diagnosis of bacterial meningitis was probably underestimated because of our decision to equate the lack of a specific reference to a prior visit with the actual lack of a prior visit. Since our data reflect only short-term sequelae, we cannot extrapolate our conclusions to the long-term outlook for such children. However, since long-term recovery is uncommon, it is highly plausible that our findings would also apply to the long-term outcome for these children. Our study would have been enhanced if outcomes had been documented comprehensively in all of the charts. In conclusion, compared with children who have meningitis caused by other organisms, young children with S. pneumoniae meningitis have a longer duration of fever and greater likelihood of having had a prior clinician visit before bacterial meningitis was diagnosed. Prior antibiotic administration among such children is associated with a substantially improved outcome at the time of discharge from the hospital. Acknowledgments We acknowledge Drs. Elizabeth Bodner and Sue Torrey, for their assistance with data abstraction, as well as Dr. Gary Fleisher, for his helpful comments during the preparation of the manuscript. References 1. Baraff LJ, Lee SI, Schriger DL. Outcomes of bacterial meningitis in children: a meta-analysis. Pediatr Infect Dis J 1993; 12:389 94. 2. Schuchat A, Robinson K, Wenger JD, et al. Bacterial meningitis in the United States in 1995. Active Surveillance Team. N Engl J Med 1997; 337:970 6. 3. Baraff LJ, Bass JW, Fleisher GR, et al. Practice guideline for the management of infants and children 0 to 36 months of age with fever without source. Agency for Health Care Policy and Research [erratum appears in Ann Emerg Med 1993; 22:1490] [see comments]. Ann Emerg Med 1993; 22:1198 210. 4. Kilpi T, Anttila M, Kallio MJ, Peltola H. Severity of childhood bacterial meningitis and duration of illness before diagnosis. Lancet 1991; 338: 406 9. 5. Davis SD, Hill HR, Feigl P, Arnstein EJ. Partial antibiotic therapy in Haemophilus influenzae meningitis: its effect on cerebrospinal fluid abnormalities. Am J Dis Child 1975; 129:802 7. 6. Rothrock SG, Green SM, Wren J, Letai D, Daniel-Underwood L, Pillar E. Pediatric bacterial meningitis: is prior antibiotic therapy associated with an altered clinical presentation? Ann Emerg Med 1992; 21:146 52. 7. Kilpi T, Anttila M, Kallio MJ, Peltola H. Length of prediagnostic history related to the course and sequelae of childhood bacterial meningitis. Pediatr Infect Dis J 1993; 12:184 8. 8. Kallio MJ, Kilpi T, Anttila M, Peltola H. The effect of a recent previous visit to a physician on outcome after childhood bacterial meningitis [see comments]. JAMA 1994; 272:787 91. 9. Winkelstein JA. The influence of partial treatment with penicillin on the diagnosis of bacterial meningitis. J Pediatr 1970; 77:619 24. 10. Jarvis CW, Saxena KM. Does prior antibiotic treatment hamper the diagnosis of acute bacterial meningitis? An analysis of a series of 135 childhood cases. Clin Pediatr (Phila) 1972; 11:201 4. 11. Davis SD, Hill HR, Polly F, Ellis JA. Partial antibiotic therapy in Haemophilus influenzae meningitis: its effect on cerebrospinal fluid abnormalities. Am J Dis Child 1975; 129:802 7. 12. Kaplan SL, Smith EO, Wills C, Feigin RD. Association between preadmission oral antibiotic therapy and cerebrospinal fluid findings and sequelae caused by Haemophilus influenzae type b meningitis. Pediatr Infect Dis 1986; 5:626 32. 572 CID 2001:32 (15 February) Bonsu and Harper