Hearing Loss at School Age in Survivors of Bacterial Meningitis: Assessment, Incidence, and Prediction
,
* Department of Pediatrics, VU Medical Center, Amsterdam, the Netherlands
Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, the Netherlands
Center of Biostatistics, University Medical Center Utrecht, Utrecht, the Netherlands
|| Division of Neuropsychology, Wilhelmina Children’s Hospital, University Medical Center Utrecht, Utrecht, the Netherlands
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Objectives. To establish the incidence of sensorineural hearing loss in children who survived non–Haemophilus influenzae type B (Hib) bacterial meningitis, to highlight the actual percentage whose hearing was evaluated, and to develop a prediction rule to identify those who are at risk of hearing loss.
Methods. In 1999, we compiled a cohort of 628 school-aged children who were born between January 1986 and December 1994 and had survived non-Hib bacterial meningitis between January 1990 and December 1995. Presence of sensorineural hearing loss (>25 dB) was determined, based on information from questionnaires and medical records. Potential risk factors for hearing loss were obtained from medical records; independent predictors were identified using multivariate logistic regression analysis, leading to the formulation of a prediction rule.
Results. The incidence of hearing loss was 7%. The hearing of 68% of the children was evaluated as part of their routine follow-up after bacterial meningitis, resulting in the detection of 75% of the cases of hearing loss. The remaining 25% were detected after this follow-up had ended. Using a prediction rule based on 5 factors—duration of symptoms before admission >2 days, absence of petechiae, cerebrospinal fluid glucose level 
0.6 mmol/L, Streptococcus pneumoniae, and ataxia—62% of the postmeningitic children were selected as being at risk. All cases of hearing loss were in this at-risk group.
Conclusions. Hearing loss can be predicted satisfactorily. When the hearing of children who are predicted to be at risk is tested as part of their routine follow-up, no children with hearing loss need be missed.
Key Words: bacterial meningitis • cohort study • hearing assessment • hearing impairment • prediction
Abbreviations: Hib, Haemophilus influenzae type B • CSF, cerebrospinal fluid • AUC, area under the curve
Sensorineural hearing loss is the most common severe consequence of childhood bacterial meningitis, affecting approximately 9% of the children.1–5 Furthermore, bacterial meningitis is the leading cause of severe hearing loss acquired in childhood.4,6 As a result of vaccination, Haemophilus influenzae type B (Hib) meningitis has virtually disappeared from most Western countries,7,8 but it is not yet known to what extent this has influenced the incidence of hearing loss. Early identification and rehabilitation of hearing loss is essential for the acquisition of normal speech and language, as well as for the child’s educational and social development.9 Early identification of hearing loss is also important because ossification of the cochlea after meningitis may complicate cochlear implantation.10,11 Hearing evaluation, therefore, is recommended as part of the routine follow-up after bacterial meningitis.1,5,12,13 However, as up to 25% of the children do not undergo a formal hearing test after bacterial meningitis,12,13 identification of those who are at risk of hearing loss may help to ensure that they are evaluated. Previously proposed risk factors for hearing loss1,5,12,14,15 have not resulted in satisfactory prediction. The purposes of the present study were to 1) establish the incidence of sensorineural hearing loss in children who survived non-Hib bacterial meningitis, 2) report on the actual percentage whose hearing was evaluated after bacterial meningitis, and 3) develop a prediction rule to identify those who are at risk of hearing loss.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Postmeningitic Cohort
Files of the Netherlands Reference Laboratory for Bacterial Meningitis were searched in 1999 for data on eligible patients. This laboratory collects bacterial isolates and data (eg, pathogen, name, date of birth, hospital of admission) from approximately 80% of all bacterial meningitis cases in the Netherlands. The diagnosis of bacterial meningitis was based on the presence of bacteria being demonstrated in the cerebrospinal fluid (CSF). The inclusion criteria were date of birth between January 1986 and December 1994 and recovery from meningitis caused by Neisseria meningitidis, Streptococcus pneumoniae, Streptococcus agalactiae, Escherichia coli, or Listeria monocytogenes between January 1990 and December 1995. The exclusion criteria were meningitis caused by Hib or other less common pathogens, "complex onset" of meningitis (defined as meningitis secondary to immunodeficiency states, central nervous system surgery, cranial trauma, or CSF shunt infections, or relapsing meningitis), cognitive or behavioral problems before meningitis, and diseases developed after meningitis (eg, cancer), which could have caused cognitive or behavioral problems. These last 2 exclusion criteria were applied because this cohort was compiled as part of a study on academic and/or behavioral problems after bacterial meningitis.16
Inclusion criteria were met by 1605 patients, treated in 110 different hospitals. A pediatrician from every hospital involved sent the parents a standard letter requesting their participation. Participating parents were sent questionnaires on long-term outcome and on exclusion criteria.16 Subsequently, medical records of the children were studied to obtain detailed data on admission and follow-up. During this process, 860 (54%) children were lost for various reasons: pediatricians’ refusal (n = 205, 10 pediatricians), death (n = 40), addresses missing from medical records (n = 60), no response to letter requesting participation (n = 445), parents’ refusal to participate (n = 49), questionnaires not returned (n = 16), filling in of questionnaires incomplete (n = 11), no permission from parents to study medical records (n = 14), and data on bacterial meningitis incomplete in medical records (n = 20). Moreover, 117 (7%) children were excluded because of meningitis caused by Hib or other rare pathogens (n = 4), no meningitis (n = 1), meningitis "with complex onset" (n = 84), cognitive or behavioral problems before meningitis (n = 23), and diseases after meningitis that could have caused cognitive or behavioral problems (n = 5). Hence, this report concerns 628 children (57% boys). Mean age at infection was 2.4 years (range: 0–9.5), and mean age in June 2002 was 11.7 years (range: 7.5–16.3).
Hearing Loss and Hearing Evaluation
Hearing loss was defined as a perceptive loss of >25 dB4; the diagnosis was based on information from questionnaires and medical records. Also gathered from the medical records was information on timing and type of hearing evaluation. This information could not be found in 50 medical records, and so the report on this issue concerns a total of 578 children.
Risk Factors for Hearing Loss
By reviewing medical records, information was collected on all potential risk factors for hearing loss after bacterial meningitis. To reduce the number of factors to be evaluated, we selected only those reported in previous studies on permanent sensorineural hearing loss after (non-Hib) childhood bacterial meningitis1,5,12,14,15 (Appendix 1).17
The study was approved by the medical ethics committee of the University Medical Center Utrecht. Written informed consent was obtained from the parents or guardians of all children and from the children themselves if they were at least 12 years of age.
Data Analysis
A 2-sample t test was used for continuous data, and a 
2 test was used for nominal data. Statistical significance was considered when 2-tailed P values were .05. Associations between hearing loss and potential risk factors were examined using univariate logistic regression analysis. Predictors univariately associated with outcome (P < .10) were included in a multivariate logistic regression model. Continuous predictors were dichotomized. The model was reduced by excluding predictors from the model with P > .10, and the goodness of fit was estimated.18 The prognostic ability of the model, ie, to discriminate between patients with and without hearing loss, was estimated by measuring the area under the receiver operating characteristic curve (AUC).19 As multivariate regression analysis requires data to be complete for all patients,20 multiple imputation techniques were used to fill in missing variables. Per variable, no >20% of the data were missing.20 Imputation was repeated 5 times to take into account uncertainties in imputed data.21 From each of the imputed data sets, a prediction model was estimated. Averaging the regression coefficients and standard errors of these logistic models resulted in a single prediction model.21 Furthermore, random bootstrapping techniques were used to calculate a shrinkage factor to adjust for overfitting (ie, too optimistic estimates of the regression coefficients) and to validate the model.20 The final model was transformed into a clinical prediction rule. Regression coefficients were divided by the smallest coefficient, multiplied by 10, and rounded off to the nearest integer. By assigning points for each variable and adding the results, a score was obtained for each patient. Patients were classified according to their score. For simplifying the interpretation of the model, a nomogram was prepared and a cutoff point chosen. The data were analyzed using SPSS 9.0 and S-plus, with help of the libraries Hmisc and Design of Harrell22 and mice of Van Buuren and Oudshoorn.23
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Hearing Loss
Forty-three (7%) postmeningitic children had hearing loss. This was unilateral in 20 (3%) and bilateral in 23 (4%). In most children, hearing loss was severe (71–90 dB; n = 11) or profound (>90 dB; n = 21). Five children with profound loss received cochlear implants.
Hearing Evaluation
The hearing of 68% (395 of 578) of the children had been evaluated as part of the routine follow-up after meningitis. They either underwent a formal hearing assessment (46%)—auditory brainstem response in 132, audiogram in 97, test at an audiological center or by an otorhinolaryngologist in 38 children—or their hearing was screened using distraction methods (22%) at the child health clinic. Hearing loss was diagnosed in 32 of these children (75% of the children with hearing loss); with only 2 exceptions, hearing loss was detected within 6 months after bacterial meningitis. The hearing of 80 (78%) of the 103 children with pneumococcal meningitis and of 287 (58%) of the 495 children with meningococcal meningitis were tested as part of the routine follow-up after meningitis. The hearing of 5% (26 of 578) of the children was evaluated because of complaints that emerged after the outpatient clinic follow-up had ended. Hearing loss was diagnosed in 11 cases (25% of the children with hearing loss); 8 of these losses were >55 dB, and all but 1 were detected late (>6 months after bacterial meningitis). Of these 11 children, 3 had survived pneumococcal meningitis and 8 had survived meningococcal meningitis.
Risk Factors
Table 1 presents the associations between potential predictors and hearing loss. In the multivariate analysis, duration of symptoms before admission >2 days, absence of petechiae, CSF glucose level 0.6 mmol/L, S pneumoniae, and ataxia emerged as independent predictors (Table 2). A prediction rule was constructed by assigning points for each variable. A total score was computed for each patient, ranging from 0 to 100. The AUC of the score was 0.85 (95% confidence interval: 0.80–0.90). Figure 1 shows the nomogram of the rule, and Table 3 shows the numbers of subjects across categories of the score. With a cutoff of 0, all children with hearing loss were identified correctly. This cutoff point suggests that a patient is at risk when 1 of the predictors is present. With this cutoff, 62% of the children were selected as being at risk, ie, 60% of the children without hearing loss were falsely predicted as having hearing loss (false positives; Table 3). Alternatively, with a cutoff of 25 or more, 95% of the patients with hearing loss were identified correctly. Of the 2 (5%) children who were false negatives, 1 had mild and 1 severe loss. With this cutoff point, 49% of the children were selected as at risk, ie, 46% false positives (Table 3).
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The incidence of sensorineural hearing loss in children who survived non-Hib bacterial meningitis is 7%. Approximately one third of the Dutch postmeningitic children do undergo a hearing evaluation as part of their routine follow-up after bacterial meningitis, resulting in late detection of one quarter of the cases of hearing loss. Using a rule based on 5 easily obtainable predictors, 62% of the postmeningitic children were identified as being at risk of hearing loss. Had these children received audiological follow-up, none of the children with hearing loss would have been missed.
To appreciate the true value of these results and place them in a proper context, a few points should be borne in mind. Of the 1605 children selected from the reference laboratory for bacterial meningitis, 628 were included. Most children were excluded because parents or pediatricians refused to participate or because parents could not be contacted. Selection bias could have occurred. However, sex, age, and causative pathogens of the 628 participants and the 1605 originally selected children were very similar,16 suggesting that our cohort is representative of the pediatric non-Hib bacterial meningitis population. Second, children with previous cognitive or behavioral problems, children with diseases that could have led to cognitive or behavioral problems, and children with "complex onset" of meningitis were excluded. The prediction rule may not be applicable to these excluded groups. Third, the hearing of 27% of the children was not tested. It is possible that some of these children had small, unobserved losses and that we have underestimated the incidence of hearing loss. Fourth, we collected risk factors retrospectively from routinely documented medical information. As a consequence, not all data were available for each patient. We dealt with this problem by using multiple imputation techniques. Finally, using bootstrapping techniques, we demonstrated that the prediction rule is robust. Although the regression coefficients shrank after bootstrapping, the AUC did not change. The actual performance of this scoring rule, however, does need to be confirmed in a new group of postmeningitic children before it can be implemented in clinical practice.
All risk factors found in our study have previously been reported.1,5,12,14,15 However, until now, hearing loss had not been predicted satisfactorily and comprehensive prediction rules had not been provided. With our prediction rule, all patients with hearing loss could be predicted with a cutoff of 0. We chose this cutoff point because we find it unacceptable to miss any case of hearing loss. Whereas the number of children who require hearing assessment (62%) will be slightly fewer than in current practice (68%), no cases of hearing loss will be missed (compared with 25% currently). With regard to the factor S pneumoniae in the prediction rule, it is important to realize that vaccines against common serotypes of S pneumoniae are currently available. Vaccination campaigns using these vaccines may alter the incidence of hearing loss and thus the prediction rule.
The hearing of 22% of the children was screened at a child health clinic (in the Netherlands, the hearing of every child is screened using distraction methods at a child health clinic at the age of 9 months). Although these children were not formally tested, their hearing was screened and losses were detected within 6 months. Recently, however, a program for neonatal hearing screening shortly after birth has been implemented in the Netherlands. Consequently, hearing screening using distraction methods at the age of 9 months will disappear. Thus, more postmeningitic children will have to undergo a formal hearing assessment.
It is currently not clear which testing strategy is most effective; should children at risk be tested with an audiogram or auditory brainstem response, or should they be screened with oto-acoustic emissions and should only children who fail this screening be tested with an audiogram or auditory brainstem response? Another difficult issue concerns the timing of testing. Fluctuating hearing loss and deterioration of hearing in time have been reported.3,11,24 Hence, it may be insufficient to assess hearing just once (straight after discharge from hospital). Additional research into these issues is necessary.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Hearing loss is an important severe consequence of non-Hib bacterial meningitis, affecting 7% of the postmeningitic children. Using a prediction rule based on 5 easily obtainable patient characteristics, children who are at risk of hearing loss can be identified. The inclusion of hearing evaluation in the routine follow-up of these children after bacterial meningitis may prevent missed diagnoses of hearing loss. Additional research is required to validate the prediction rule further.
|
|
ACKNOWLEDGMENTS |
|---|
This research was supported by the Health Research and Development Council of the Netherlands (ZON, project 28-30470).
We thank the Netherlands Reference Laboratory for Bacterial Meningitis for help with patient inclusion. We are grateful to the pediatricians and the families who enthusiastically agreed to participate in this study. The help of H. van der Lei and D. Eijking with processing of the medical records, of M. Twijnstra with data management, and of B. Vollers-King with editing is gratefully acknowledged.
|
FOOTNOTES |
|---|
Received for publication Dec 9, 2002; Accepted Jul 7, 2003.
Reprint requests to (A.M.v.F.) VU Medical Center, Department of Pediatrics, Box 7057, 1007 MB Amsterdam, the Netherlands. E-mail: am.vfurth{at}vumc.nl
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
- Fortnum HM. Hearing impairment after bacterial meningitis: a review.
Arch Dis Child.1992; 67
:1128
–1133
[Free Full Text] - Baraff LJ, Lee SI, Schriger DL. Outcomes of bacterial meningitis in children: a meta-analysis. Pediatr Infect Dis J.1993; 12 :389 –394[Web of Science][Medline]
- Richardson MP, Reid A, Tarlow MJ, Rudd PT. Hearing loss during bacterial meningitis.
Arch Dis Child.1997; 76
:134
–138
[Abstract/Free Full Text] - Irving RM, Ruben RJ. The Acquired Hearing Losses of Childhood. Philadelphia, PA: Lippincott-Raven Publishers; 1998
- Woolley AL, Kirk KA, Neumann AM Jr, et al. Risk factors for hearing loss from meningitis in children: the Children’s Hospital experience.
Arch Otolaryngol Head Neck Surg.1999; 125
:509
–514
[Abstract/Free Full Text] - Fortnum H, Davis A. Epidemiology of permanent childhood hearing impairment in Trent Region, 1985–1993. Br J Audiol.1997; 31 :409 –446[Web of Science][Medline]
- Conyn-van Spaendonck MA, Veldhuijzen IK, Suijkerbuijk AW, Hirasing RA. Significant decline of the number of invasive Haemophilus influenzae infections in the first 4 years after introduction of vaccination against H. influenzae type B in children [in Dutch, English summary]. Ned Tijdschr Geneeskd.2000; 144 :1069 –1073[Medline]
- Peltola H. Haemophilus influenzae type b disease and vaccination in Europe: lessons learned. Pediatr Infect Dis J.1998; 17(suppl) :S126 –S132[CrossRef][Web of Science][Medline]
- Yoshinaga-Itano C, Sedey AL, Coulter DK, Mehl AL. Language of early- and later-identified children with hearing loss.
Pediatrics.1998; 102
:1161
–1171
[Abstract/Free Full Text] - Dodds A, Tyszkiewicz E, Ramsden R. Cochlear implantation after bacterial meningitis: the dangers of delay.
Arch Dis Child.1997; 76
:139
–140
[Abstract/Free Full Text] - Marx RD, Baer ST. Spontaneous recovery of profound post-meningitic hearing loss. J Laryngol Otol.2001; 115 :412 –414[Web of Science][Medline]
- Fortnum H, Davis A. Hearing impairment in children after bacterial meningitis: incidence and resource implications. Br J Audiol.1993; 27 :43 –52[Web of Science][Medline]
- Riordan A, Thomson A, Hodgson J, Hart A. Children who are seen but not referred: hearing assessment after bacterial meningitis. Br J Audiol.1993; 27 :375 –377[Web of Science][Medline]
- McIntyre PB, Berkey CS, King SM, et al. Dexamethasone as adjunctive therapy in bacterial meningitis. A meta-analysis of randomized clinical trials since 1988.
JAMA.1997; 278
:925
–931
[Abstract/Free Full Text] - Arditi M, Mason EO Jr, Bradley JS, et al. Three-year multicenter surveillance of pneumococcal meningitis in children: clinical characteristics, and outcome related to penicillin susceptibility and dexamethasone use.
Pediatrics.1998; 102
:1087
–1097
[Abstract/Free Full Text] - Koomen I, Grobbee DE, Jennekens-Schinkel A, Roord JJ, van Furth AM. Parental perception of educational, behavioural and general health problems in school-age survivors of bacterial meningitis. Acta Paediatr.2003; 92 :177 –185[Web of Science][Medline]
- Steyerberg EW, Eijkmans MJC, Habbema JDF. Stepwise selection in small data sets: a simulation study of bias in logistic regression analysis. J Clin Epidemiol.1999; 10 :935 –942
- Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons Inc; 1989
- Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve.
Radiology.1982; 143
:29
–36
[Abstract/Free Full Text] - Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med.1996; 15 :361 –387[CrossRef][Web of Science][Medline]
- Little RA. Regression with missing X’s: a review. J Am Stat Assoc.1992; 87 :1227 –1237[CrossRef][Web of Science]
- Harrell FE. Regression Model Strategies. New York, NY: Springer; 2001
- Van Buuren S, Oudshoorn CGM. Multivariate Imputation by Chained Equations: MICE V1.0. User’s Manual. Report PG/VGZ/00.038. Leiden, the Netherlands: TNO Prevention and Health; 2000
- Jayarajan V, Rangan S. Delayed deterioration of hearing following bacterial meningitis. J Laryngol Otol.1999; 113 :1011 –1014[Web of Science][Medline]
PEDIATRICS (ISSN 1098-4275). ©2003 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
This article has been cited by other articles:
|
|
 |
|
  C. M. P. Buysse, H. Raat, J. A. Hazelzet, J. M. Hulst, K. Cransberg, W. C. J. Hop, L. C. A. C. Vermunt, E. M. W. J. Utens, M. Maliepaard, and K. F. M. Joosten Long-term Health Status in Childhood Survivors of Meningococcal Septic Shock Arch Pediatr Adolesc Med, November 1, 2008; 162(11): 1036 - 1041. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Declau, A. Boudewyns, J. Van den Ende, A. Peeters, and P. van den Heyning Etiologic and Audiologic Evaluations After Universal Neonatal Hearing Screening: Analysis of 170 Referred Neonates Pediatrics, June 1, 2008; 121(6): 1119 - 1126. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. G. Millichap Risk of Hearing Loss After Bacterial Meningitis AAP Grand Rounds, February 1, 2004; 11(2): 17 - 18. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
