Accuracy of initial clinical diagnosis of acute bacterial meningitis in children from a malaria-endemic area of Papua New Guinea Jimmy Aipita, Moses Lamanb,c,*, Ilomo Hwaiwhanjea, Cathy Bonab, Naomi Pomata, Peter Sibab, Timothy M. E. Davisc and Laurens Manningc a
Modilon General Hospital, Paediatrics Unit, Madang, Papua New Guinea; bPapua New Guinea Institute of Medical Research, Madang, Papua New Guinea; cSchool of Medicine and Pharmacology, University of Western Australia, Fremantle Hospital, Fremantle, Western Australia, Australia *Corresponding author: Tel: +675 422 2909; E-mail: [email protected]
Received 17 December 2013; revised 1 April 2014; accepted 1 April 2014 Background: The diagnosis of acute bacterial meningitis (ABM) is challenging in resource-limited settings where cerebral malaria and viral encephalitis are also common. Methods: To assess the accuracy of an initial clinical diagnosis of ABM in a malaria-endemic area of Papua New Guinea (PNG), a retrospective chart review of hospitalized children aged 2 months to 10 years was conducted. Results: Of the 481 eligible children, 240 had an initial clinical diagnosis of ABM that was confirmed independently by trained research staff under standardized conditions, with laboratory support in only 84 (17.5%; 84/481). When compared with the final laboratory-confirmed diagnosis, an initial diagnosis of ABM had a sensitivity, specificity, positive predictive value and negative predictive value of 76% (95% CI 66–85%), 56% (95% CI 51–61%), 27% (95% CI 21–33) and 92% (95% CI 87–95%), respectively. There was discordance between initial and final diagnosis of ABM in 196 children; 176 initially considered to have ABM had an alternative diagnosis, while 20 without an initial diagnosis of ABM were confirmed to have ABM. Conclusion: These data show that initial misdiagnosis of ABM is common in a malaria-endemic area of PNG. A diagnostic algorithm using standardized assessment for meningeal irritation, coma and malaria parasitological testing needs further evaluation in this setting. Keywords: Acute bacterial meningitis, Malaria-endemic setting, Papua New Guinea
Introduction Acute bacterial meningitis (ABM) is a severe illness in children that frequently results in death or long-term neurological sequelae.1–3 Its presenting features include impaired consciousness, focal neurological deficits, signs of meningeal irritation and seizures. Because these signs and symptoms overlap with those of cerebral malaria and viral encephalitis,1,4 infections that are also relatively common in resource-poor tropical countries such as Papua New Guinea (PNG),3,5,6 diagnosis can be difficult on clinical grounds, especially when laboratory facilities are limited or unavailable. An accurate clinical diagnosis of ABM is, nevertheless, important in this setting because it allows initiation of appropriate antibiotic treatment and supportive measures, while avoiding overuse of antibiotics in children with alternative non-septic illness.7 Haemopilus influenzae Type B (Hib) and Streptococcus pneumoniae are the leading causes of ABM in PNG.8 This situation is unlikely to change for some time since the S. pneumoniae vaccine
has only recently been introduced into PNG in 2013 and although Hib vaccination was incorporated into national schedules in 2007, coverage in PNG remains below 70%, particularly in rural areas. Because of emerging resistance, ceftriaxone has replaced chloramphenicol as the recommended first-line treatment for ABM in PNG children.9 Although bacterial pathogens causing ABM remain sensitive to ceftriaxone, this may be temporary.10 In view of potential resistance and the cost of ceftriaxone, it is important to optimize clinical examination and bedside diagnostic tests in the assessment of children with suspected ABM. In a recent PNG study of hospitalized children, neck stiffness, deep coma and a bulging fontanel in children younger than 18 months at presentation, together with a negative parasitological test for malaria, were strong independent predictors of ABM as a final diagnosis.3 However, the diagnostic utility of clinical signs elicited by trained research staff under standardized conditions with laboratory support may not necessarily reflect usual care. Clinicians in countries such as PNG often work in relative isolation
# The Author 2014. Published by Oxford University Press on behalf of Royal Society of Tropical Medicine and Hygiene. All rights reserved. For permissions, please e-mail: [email protected].
444
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
ORIGINAL ARTICLE
Trans R Soc Trop Med Hyg 2014; 108: 444–448 doi:10.1093/trstmh/tru067 Advance Access publication 3 May 2014
Transactions of the Royal Society of Tropical Medicine and Hygiene
Materials and methods Study setting The present study was carried out at Modilon Hospital, the main referral hospital in Madang Province on the north coast of mainland PNG. A single consultant paediatrician and a registrar provide senior clinical support for Modilon Hospital Paediatric Division, with the assistance of resident medical officers, health extension officers and nurses. The annual entomological inoculation rate for Madang Province has recently been estimated at 37 for Plasmodium falciparum and 24 for P. vivax. 11 Co-incident sepsis, including ABM, in children with severe or cerebral malaria is infrequent.3,12 However, because ABM often results in high morbidity and mortality in PNG, local guidelines recommend lumbar puncture (LP) even in children without clinical signs of ABM. Due to limited diagnostic facilities, empirical treatment is common and clinicians assessing children at presentation with signs or symptoms of severe illness will often initiate management prior to receiving results from blood or cerebrospinal fluid (CSF) examination. Because of the large number of daily admissions, it is not feasible for all admitted children to be reviewed by the consultant paediatrician on the day of admission.
Study design The present study is a retrospective review of an initial clinical diagnosis of ABM. A previous prospective study was conducted at Modilon Hospital, from 2007 to 2010, which identified clinical predictors of ABM.3 The prospective study was conducted by a research team composing of six research nurses, a medical officer and an infectious diseases physician, who collected clinical and laboratory data independently from the hospital-based clinicians. During the 3-year study period, all children were admitted to the ward by the hospital clinicians independently and this enabled the assessment of the accuracy of an initial clinical diagnosis of ABM. An initial clinical diagnosis of ABM refers to the provisional diagnosis made by hospital-based clinicians, as documented in a child’s hospital record at the Outpatient Department or the Emergency Department, prior to ward admission and the availability of CSF results. This diagnosis is often based on signs and symptoms such as impaired consciousness, convulsions, neck stiffness, bulging fontanel, severe headache and other nonspecific signs present at the time of admission, as guided by the PNG paediatric national standard treatment guidelines.9 For many other children for who ABM was not the initial clinical diagnosis, ABM may be part of a differential diagnosis that could include malaria, pneumonia and other undifferentiated infections. In normal daily practice, hospital clinicians may not perform all the clinical examinations for meningism in a systematic manner,
or routinely record these in the patient record. Furthermore, outside the research setting there is likely to be wide variation in the ascertainment and interpretation of clinical signs and symptoms. Therefore clinical signs, features of meningeal irritation and level of consciousness reported were determined by the research team in the prospective component of the study.
Study patients Children aged 2 months to 10 years admitted to Modilon Hospital were screened for eligibility for enrolment into the prospective study.3 In that study, the decision to perform LP was made by hospital clinicians based on their initial clinical assessment, guided by the PNG national standard treatment guidelines.9 In the present study, children were considered eligible if they were recruited to the prospective study exploring predictors of ABM,3 and had an initial diagnosis recorded by the admitting hospital clinician documented in the hospital medical records. Written informed consent for participation was obtained from parent(s) or guardian(s) of the children. Approval was also obtained from the management of Modilon Hospital to extract information from hospital records. Proven ABM was defined as a CSF white blood cell count (WCC) of ≥20 cells/mm3 with a positive CSF/blood culture or bacterial antigen test.3,8 Probable ABM was defined as a CSF WCC count of ≥20 cells/mm3 with a negative CSF/blood culture or bacterial antigen test. Children without ABM were those with no laboratory evidence of ABM. Details of CSF analysis and microbiological cultures have been previously published.8 The initial diagnosis of malaria was based on microscopic examination of thick blood films and subsequently confirmed by PCR.12
Statistical analysis Comparison of categorical variables was by Fisher’s Exact or x2 tests as appropriate, while comparison of continuous variables was by Mann-Whitney U test. A two-tailed significance level of p,0.05 was used throughout. To assess the accuracy of an initial clinical diagnosis of ABM, children with proven and probable ABM were combined as it is standard practice to treat all children with a CSF leucocytosis as though they had proven ABM.
Results The hospital records of 481 children who fulfilled the inclusion criteria were retrieved. The median age of these children was 26 (inter-quartile range [IQR] 8–60) months and 56% were boys. The initial clinical and final laboratory-confirmed diagnoses are summarized in Table 1. In 240 children, ABM was the initial clinical diagnosis and not a differential diagnosis. A final diagnosis of ABM based on CSF examination was made in 84 children. Of these 84 children, ABM was proven and probable in 38 and 46 children, respectively. Children with proven ABM were significantly younger than those with probable ABM (8 [IQR 4–44] vs. 35.5 [IQR 12–60] months; p¼0.025) and had predominantly neutrophilic CSF WCC (763 [IQR 205–2200] vs. 43 [IQR 5–220] cells/mm3; p,0.001), while those with probable ABM had predominantly lymphocytic CSF WCC (15 [IQR 0–35] vs. 0 [IQR 0–10] cells/mm3; p¼0.04). The mortality rate was similar in both groups (proven ABM 24% vs. probable ABM 28%; p¼0.80).
445
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
and without the availability of relevant laboratory tests. Limited supervision and access to continuing medical education in this setting may impair the correct identification of physical signs that depend on technical competence and informed interpretation. In the present study, we aimed to determine retrospectively the accuracy of the initial ward-based diagnosis compared with the final diagnosis derived independently from detailed research data.
J. Aipit et al.
Table 1. Numbers of laboratory-confirmed proven and probable acute bacterial meningitis cases according to initial diagnosis Not ABM by initial diagnosis n¼241 (%)
Total n¼481 (%)
64 (26.7) 29 (12.1) 35 (14.6) 176 (73.3)
20 (8.3) 9 (3.7) 11 (4.6) 221 (91.7)
84 (17.5) 38 (7.9) 46 (9.6) 397 (82.5)
ABM: acute bacterial meningitis.
When compared with the final diagnosis, an initial diagnosis of ABM had a sensitivity, specificity, positive predictive value and negative predictive value of 76% (95% CI 66–85%), 56% (95% CI 51–61%), 27% (95% CI 21–33) and 92% (95% CI 87–95%), respectively. The positive and negative likelihood ratios for an initial diagnosis of ABM were 1.72 and 0.43, respectively. For children with ABM that was proven by culture or bacterial antigen testing, the sensitivity, specificity, positive predictive value and negative predictive value were 76% (95% CI 60–89%), 51% (95% CI 47– 56%), 12% (95% CI 8–16) and 96% (95% CI 93–98%), respectively. There was discordance between initial and final diagnosis of ABM in 196 children. Of children with an initial clinical diagnosis of ABM, 73.3% (176/240) did not have a final diagnosis of ABM. This included 28 children (15.9%; 28/176) with a normal level of consciousness (Blantyre coma score¼5), without signs of meningeal inflammation and no history of convulsions. Twenty children without an initial diagnosis of ABM had a final diagnosis of ABM. In this group of children, neck stiffness (60%; 12/20), Kernig’s sign (40%; 8/20) and deep coma (45%; 9/20) were commonly detected. A comparison of the clinical features and outcome of the two discordant groups are shown in Table 2. Neck stiffness, Kernig’s sign, deep coma and the presence of malaria parasites all had significantly different proportions between the discordant groups. Malaria or pneumonia accounted for 54.9% (218/397) of admissions in children without a final diagnosis of ABM (Table 3). Of the 481 children, a total of 40 (8.3%; 40/481) died and 56 (11.6%; 56/481) were discharged with at least some neurological disability. Of the 64 children in who an initial diagnosis of ABM was confirmed by the laboratory, 15 (23.4%; 15/64) died compared with 7 (35.0%; 7/20) of 20 with ABM where an initial diagnosis was not made (p¼0.38).
Discussion Early diagnosis and treatment of ABM is a critical issue that often determines prognosis in children with ABM. In countries such as PNG, children typically present at a late-stage of illness to hospitals and other healthcare facilities that have poor infrastructure and a lack of trained staff. The present study from a provincial hospital in PNG highlights the limitations of the initial clinical diagnosis of ABM and suggests that simple clinical algorithms may have
446
Initial ABM + final Initial ABM – final p-value diagnosis ABM – diagnosis ABM+ n¼176 (%) n¼20 (%) Age Neck stiffness Kernig’s sign Bulging fontanela Deep coma Convulsion(s) Malaria Outcome Died Survived with neurological sequelae Survived without neurological sequelae
29 (756) 39 (22.2) 11 (6.3) 1 (1.2) 33 (18.8) 118 (67.0) 39 (22.2)
29 (17–42) 12 (60) 8 (40) 0 (0) 9 (45) 12 (60) 2 (10)
12 (6.8) 18 (10.2)
7 (35) 3 (15)
146 (83.0)
10 (50)
0.43 ,0.001 ,0.001 1.00 0.02 0.62 0.05 0.002
ABM: acute bacterial meningitis. Bulging fontanel assessed in children ≤18 months (n¼84).
a
Table 3. Final diagnoses in children initially diagnosed with acute bacterial meningitis but proven not to have acute bacterial meningitis on cerebrospinal fluid investigations Final diagnosis
Frequency n¼397 (%)
Severe malaria Pneumonia Non-severe malaria Infections with other focus Febrile convulsions Encephalopathy Diarrhoeal illness Neurological disorders Tuberculosis Non-infective causes Rheumatic fever
92 (23.2) 67 (16.9) 59 (14.9) 59 (14.9) 38 (9.6) 27 (6.8) 25 (6.3) 14 (3.5) 7 (1.8) 7 (1.8) 2 (0.5)
potential value in this setting. Admitting local clinicians had a high index of suspicion for ABM but their initial diagnoses were often inaccurate, even when objective clinical signs of meningeal irritation were present. There are understandable influences on the clinical assessment of a severely unwell child in the rural tropics where malaria is common, and a detailed history of relevant
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
ABM Proven ABM Probable ABM Not ABM
ABM by initial diagnosis n¼240 (%)
Table 2. Clinical features and outcome for children with discordant initial and final laboratory-confirmed diagnoses of acute bacterial meningitis. Data are in median (interquartile-range) or numbers (percentage)
Transactions of the Royal Society of Tropical Medicine and Hygiene
assessing children with possible ABM. Although not formally validated, our data suggest that a standardized approach is likely to be more accurate than the present ward-based approach. The present data suggest that a standardized approach to clinical assessment and the availability of light microscopy and/or rapid diagnostic tests for malaria should guide initial clinical diagnosis. A simple algorithm incorporating key clinical features (neck stiffness, Kernig’s sign and deep coma) and the results of parasitological testing could be developed. Its uptake and accuracy in the field could be assessed through randomization of individual provincial hospitals with similar infrastructure and caseloads to usual care or educational intervention familiarizing clinicians with the ABM algorithm, with independent monitoring of diagnostic accuracy in each case. Such a detailed investigation was beyond the scope of the present study. Our study had limitations. Although every attempt was made to ensure the independence of the initial clinical diagnosis, the retrospective nature of the study cannot rule out instances in which the initial diagnosis was influenced by research methods and data. However, because the present comparative study was not pre-specified, it is unlikely that the practices and procedures of the admitting clinicians changed materially during the study. Second, because only children who had a LP were recruited, this may not be representative of all children presenting with severe illness. Nevertheless, as the final diagnosis was based on objective CSF findings, this bias was unavoidable.
Conclusions The present study shows that an initial ward-based diagnosis of ABM made by attending clinicians has limited diagnostic utility in PNG, resulting in substantial over-diagnosis and frequent missed diagnoses of ABM in children at high risk of death and neurological sequelae. This situation argues for the implementation and validation of improved clinical diagnostic algorithms complemented by better availability of relevant laboratory tests.
Authors’ contributions: JA and ML contributed equally to this work. JA, IH and NP were involved in the clinical management of patients; ML, CB, LM, PS and TMED were involved in the prospective component of this study; ML conceived the study and wrote the first draft of this manuscript; LM and TMED edited the manuscript. All authors read and approved the final manuscript. LM and TMED are guarantors of the study. Acknowledgements: We thank the children and their guardians for their participation. We also thank the staff of the Paediatric Ward at Modilon Hospital and research nurses and support staff of the PNG Institute of Medical Research for clinical and logistics support. Funding: This study was funded by the National Health and Medical Research Council (NHMRC) of Australia [grant #513782]. We also acknowledge infrastructure funding support from the Malaria Genetic Epidemiology Network Consortium. Competing interests: None declared. Ethical approval: The prospective component of this study was approved by the PNG Institute of Medical Research Institutional Review Board and the Medical Research Advisory Committee of the PNG Health Department (MRAC 07.37).
447
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
events such as pre-presentation empirical antibiotic administration and convulsions can be difficult to obtain. In addition, there are published data indicating minimal13 through to some clinical utility14 for signs of meningeal irritation when considered with other clinical signs or as part of prognostic scores that incorporate clinical and laboratory data.15 Nevertheless, neither the positive (1.72) nor negative (0.43) likelihood ratios indicate that an initial diagnosis usefully differentiated ABM from other severe illnesses. By contrast, our previous report from Modilon Hospital showed that clinical examination performed in a standardized manner by trained clinicians had reasonable diagnostic utility for ABM.3 The positive likelihood ratios for neck stiffness (4.3), Kernig’s sign (5.3) and bulging fontanel (7.6) were higher than an initial diagnosis of ABM and proved independent predictors of ABM.3 In addition, the presence of malaria parasites (positive likelihood ratio 0.04) also had good diagnostic utility in ruling out a diagnosis of ABM in this setting. The initial usual-care clinical assessment in the present study both missed and overdiagnosed ABM. A missed diagnosis (24% of ABM cases) is likely to have a major impact in a setting where diagnostic facilities and therapeutic options are limited, and the risk of death in an individual child is extremely high. Indeed, the mortality rate was 35.0% in children with ABM when the initial diagnosis of ABM was missed, compared with 23% when the initial diagnosis of ABM was correct. Although this difference was not statistically significant, it highlights that a missed diagnosis is associated with an outcome similar to that in a child with an initial diagnosis of ABM correctly made by the attending clinician. Indeed, it is possible that the consequences of a missed diagnosis of ABM in this study were attenuated by the empirical use of both artemether and antibiotic therapy, regardless of diagnosis. In addition, children with a missed diagnosis often had many of the clinical signs typical of meningeal irritation yet were not considered to have ABM. The discordant results do not appear to have occurred by chance alone. The detection of neck stiffness, Kernig’s sign and deep coma by study staff differed significantly from that of the local clinicians, perhaps reflecting a non-standardized approach to clinical examination outside a research setting. The PNG Department of Health recently changed the standard treatment guidelines for ABM. Chloramphenicol has been replaced by ceftriaxone and is now indicated only for severe pneumonia and some undifferentiated infections.9 One consequence of this change is that an attending clinician will now have to make an important practical decision for all severely unwell children, namely whether or not to prescribe ceftriaxone. The overdiagnosis of ABM in this study means that a high proportion of children without ABM receive ceftriaxone unnecessarily, resulting in increased costs to a health system that is already struggling with limited resources. Furthermore, widespread use of ceftriaxone will inevitably lead to an increased risk of antimicrobial resistance in hospital- and community-acquired bacterial pathogens. Our data contrast with similar studies performed in African children where up to 30% of proven ABM cases in Kenyan children would be missed if the diagnoses were based on clinical signs and CSF turbidity alone.16 However, these studies did not examine the use of an initial ABM diagnosis made independently as part of usual care. Large variation in clinical experience, local ABM and malaria epidemiology, and diagnostic facilities mean that our data may not be applicable to settings outside PNG. Nevertheless, the data from this study highlight difficulties for all clinicians when
J. Aipit et al.
References 1 Berkley JA, Versteeg AC, Mwangi I et al. Indicators of acute bacterial meningitis in children at a rural Kenyan district hospital. Pediatrics 2004;114:713–9.
3 Laman M, Manning L, Greenhill AR et al. Predictors of acute bacterial meningitis in children from a malaria-endemic area of Papua New Guinea. Am J Trop Med Hyg 2012;86:240–5. 4 Berkley JA, Mwangi I, Mellington F et al. Cerebral malaria versus bacterial meningitis in children with impaired consciousness. QJM 1999;92:151–7. 5 Muller I, Bockarie M, Alpers M, Smith T. The epidemiology of malaria in Papua New Guinea. Trends Parasitol 2003;19:253–9. 6 Lehmann D, Yeka W, Rongap T et al. Aetiology and clinical signs of bacterial meningitis in children admitted to Goroka Base Hospital, Papua New Guinea, 1989–1992. Ann Trop Paediatr 1999; 19:21–32.
10 Manning L, Laman M, Greenhill AR et al. Increasing chloramphenicol resistance in Streptococcus pneumoniae isolates from Papua New Guinean children with acute bacterial meningitis. Antimicrob Agents Chemother 2011;55:4454–6. 11 Michon P, Cole-Tobian JL, Dabod E et al. The risk of malarial infections and disease in Papua New Guinean children. Am J Trop Med Hyg 2007;76:997–1008. 12 Manning L, Laman M, Law I et al. Features and prognosis of severe malaria caused by Plasmodium falciparum, Plasmodium vivax and mixed Plasmodium species in Papua New Guinean children. PloS One 2011;6:e29203. 13 Waghdhare S, Kalantri A, Joshi R, Kalantri S. Accuracy of physical signs for detecting meningitis: a hospital-based diagnostic accuracy study. Clin Neurol Neurosurg 2010;112:752–7. 14 van de Beek D, de Gans J, Spanjaard L et al. Clinical features and prognostic factors in adults with bacterial meningitis. N Engl J Med 2004;351:1849–59.
7 Weber MW, Herman J, Jaffar S et al. Clinical predictors of bacterial meningitis in infants and young children in The Gambia. Trop Med Int Health 2002;7:722–31.
15 Oostenbrink R, Moons KG, Donders AR et al. Prediction of bacterial meningitis in children with meningeal signs: reduction of lumbar punctures. Acta Paediatr 2001;90:611–7.
8 Laman M, Manning L, Hwaiwhange I et al. Lumbar puncture in children from an area of malaria endemicity who present with a febrile seizure. Clin Infect Dis 2010;51:534–40.
16 Berkley JA, Mwangi I, Ngetsa CJ et al. Diagnosis of acute bacterial meningitis in children at a district hospital in sub-Saharan Africa. Lancet 2001;357:1753–7.
448
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
2 Peltola H. Burden of meningitis and other severe bacterial infections of children in Africa: implications for prevention. Clin Infect Dis 2001;32:64–75.
9 Paediatrics Society of Papua New Guinea. Standard Treatment for common illnesses of children in PNG. Port Moresby: Paediatrics Society of Papua New Guinea; 2011.
Modilon General Hospital, Paediatrics Unit, Madang, Papua New Guinea; bPapua New Guinea Institute of Medical Research, Madang, Papua New Guinea; cSchool of Medicine and Pharmacology, University of Western Australia, Fremantle Hospital, Fremantle, Western Australia, Australia *Corresponding author: Tel: +675 422 2909; E-mail: [email protected]
Received 17 December 2013; revised 1 April 2014; accepted 1 April 2014 Background: The diagnosis of acute bacterial meningitis (ABM) is challenging in resource-limited settings where cerebral malaria and viral encephalitis are also common. Methods: To assess the accuracy of an initial clinical diagnosis of ABM in a malaria-endemic area of Papua New Guinea (PNG), a retrospective chart review of hospitalized children aged 2 months to 10 years was conducted. Results: Of the 481 eligible children, 240 had an initial clinical diagnosis of ABM that was confirmed independently by trained research staff under standardized conditions, with laboratory support in only 84 (17.5%; 84/481). When compared with the final laboratory-confirmed diagnosis, an initial diagnosis of ABM had a sensitivity, specificity, positive predictive value and negative predictive value of 76% (95% CI 66–85%), 56% (95% CI 51–61%), 27% (95% CI 21–33) and 92% (95% CI 87–95%), respectively. There was discordance between initial and final diagnosis of ABM in 196 children; 176 initially considered to have ABM had an alternative diagnosis, while 20 without an initial diagnosis of ABM were confirmed to have ABM. Conclusion: These data show that initial misdiagnosis of ABM is common in a malaria-endemic area of PNG. A diagnostic algorithm using standardized assessment for meningeal irritation, coma and malaria parasitological testing needs further evaluation in this setting. Keywords: Acute bacterial meningitis, Malaria-endemic setting, Papua New Guinea
Introduction Acute bacterial meningitis (ABM) is a severe illness in children that frequently results in death or long-term neurological sequelae.1–3 Its presenting features include impaired consciousness, focal neurological deficits, signs of meningeal irritation and seizures. Because these signs and symptoms overlap with those of cerebral malaria and viral encephalitis,1,4 infections that are also relatively common in resource-poor tropical countries such as Papua New Guinea (PNG),3,5,6 diagnosis can be difficult on clinical grounds, especially when laboratory facilities are limited or unavailable. An accurate clinical diagnosis of ABM is, nevertheless, important in this setting because it allows initiation of appropriate antibiotic treatment and supportive measures, while avoiding overuse of antibiotics in children with alternative non-septic illness.7 Haemopilus influenzae Type B (Hib) and Streptococcus pneumoniae are the leading causes of ABM in PNG.8 This situation is unlikely to change for some time since the S. pneumoniae vaccine
has only recently been introduced into PNG in 2013 and although Hib vaccination was incorporated into national schedules in 2007, coverage in PNG remains below 70%, particularly in rural areas. Because of emerging resistance, ceftriaxone has replaced chloramphenicol as the recommended first-line treatment for ABM in PNG children.9 Although bacterial pathogens causing ABM remain sensitive to ceftriaxone, this may be temporary.10 In view of potential resistance and the cost of ceftriaxone, it is important to optimize clinical examination and bedside diagnostic tests in the assessment of children with suspected ABM. In a recent PNG study of hospitalized children, neck stiffness, deep coma and a bulging fontanel in children younger than 18 months at presentation, together with a negative parasitological test for malaria, were strong independent predictors of ABM as a final diagnosis.3 However, the diagnostic utility of clinical signs elicited by trained research staff under standardized conditions with laboratory support may not necessarily reflect usual care. Clinicians in countries such as PNG often work in relative isolation
# The Author 2014. Published by Oxford University Press on behalf of Royal Society of Tropical Medicine and Hygiene. All rights reserved. For permissions, please e-mail: [email protected].
444
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
ORIGINAL ARTICLE
Trans R Soc Trop Med Hyg 2014; 108: 444–448 doi:10.1093/trstmh/tru067 Advance Access publication 3 May 2014
Transactions of the Royal Society of Tropical Medicine and Hygiene
Materials and methods Study setting The present study was carried out at Modilon Hospital, the main referral hospital in Madang Province on the north coast of mainland PNG. A single consultant paediatrician and a registrar provide senior clinical support for Modilon Hospital Paediatric Division, with the assistance of resident medical officers, health extension officers and nurses. The annual entomological inoculation rate for Madang Province has recently been estimated at 37 for Plasmodium falciparum and 24 for P. vivax. 11 Co-incident sepsis, including ABM, in children with severe or cerebral malaria is infrequent.3,12 However, because ABM often results in high morbidity and mortality in PNG, local guidelines recommend lumbar puncture (LP) even in children without clinical signs of ABM. Due to limited diagnostic facilities, empirical treatment is common and clinicians assessing children at presentation with signs or symptoms of severe illness will often initiate management prior to receiving results from blood or cerebrospinal fluid (CSF) examination. Because of the large number of daily admissions, it is not feasible for all admitted children to be reviewed by the consultant paediatrician on the day of admission.
Study design The present study is a retrospective review of an initial clinical diagnosis of ABM. A previous prospective study was conducted at Modilon Hospital, from 2007 to 2010, which identified clinical predictors of ABM.3 The prospective study was conducted by a research team composing of six research nurses, a medical officer and an infectious diseases physician, who collected clinical and laboratory data independently from the hospital-based clinicians. During the 3-year study period, all children were admitted to the ward by the hospital clinicians independently and this enabled the assessment of the accuracy of an initial clinical diagnosis of ABM. An initial clinical diagnosis of ABM refers to the provisional diagnosis made by hospital-based clinicians, as documented in a child’s hospital record at the Outpatient Department or the Emergency Department, prior to ward admission and the availability of CSF results. This diagnosis is often based on signs and symptoms such as impaired consciousness, convulsions, neck stiffness, bulging fontanel, severe headache and other nonspecific signs present at the time of admission, as guided by the PNG paediatric national standard treatment guidelines.9 For many other children for who ABM was not the initial clinical diagnosis, ABM may be part of a differential diagnosis that could include malaria, pneumonia and other undifferentiated infections. In normal daily practice, hospital clinicians may not perform all the clinical examinations for meningism in a systematic manner,
or routinely record these in the patient record. Furthermore, outside the research setting there is likely to be wide variation in the ascertainment and interpretation of clinical signs and symptoms. Therefore clinical signs, features of meningeal irritation and level of consciousness reported were determined by the research team in the prospective component of the study.
Study patients Children aged 2 months to 10 years admitted to Modilon Hospital were screened for eligibility for enrolment into the prospective study.3 In that study, the decision to perform LP was made by hospital clinicians based on their initial clinical assessment, guided by the PNG national standard treatment guidelines.9 In the present study, children were considered eligible if they were recruited to the prospective study exploring predictors of ABM,3 and had an initial diagnosis recorded by the admitting hospital clinician documented in the hospital medical records. Written informed consent for participation was obtained from parent(s) or guardian(s) of the children. Approval was also obtained from the management of Modilon Hospital to extract information from hospital records. Proven ABM was defined as a CSF white blood cell count (WCC) of ≥20 cells/mm3 with a positive CSF/blood culture or bacterial antigen test.3,8 Probable ABM was defined as a CSF WCC count of ≥20 cells/mm3 with a negative CSF/blood culture or bacterial antigen test. Children without ABM were those with no laboratory evidence of ABM. Details of CSF analysis and microbiological cultures have been previously published.8 The initial diagnosis of malaria was based on microscopic examination of thick blood films and subsequently confirmed by PCR.12
Statistical analysis Comparison of categorical variables was by Fisher’s Exact or x2 tests as appropriate, while comparison of continuous variables was by Mann-Whitney U test. A two-tailed significance level of p,0.05 was used throughout. To assess the accuracy of an initial clinical diagnosis of ABM, children with proven and probable ABM were combined as it is standard practice to treat all children with a CSF leucocytosis as though they had proven ABM.
Results The hospital records of 481 children who fulfilled the inclusion criteria were retrieved. The median age of these children was 26 (inter-quartile range [IQR] 8–60) months and 56% were boys. The initial clinical and final laboratory-confirmed diagnoses are summarized in Table 1. In 240 children, ABM was the initial clinical diagnosis and not a differential diagnosis. A final diagnosis of ABM based on CSF examination was made in 84 children. Of these 84 children, ABM was proven and probable in 38 and 46 children, respectively. Children with proven ABM were significantly younger than those with probable ABM (8 [IQR 4–44] vs. 35.5 [IQR 12–60] months; p¼0.025) and had predominantly neutrophilic CSF WCC (763 [IQR 205–2200] vs. 43 [IQR 5–220] cells/mm3; p,0.001), while those with probable ABM had predominantly lymphocytic CSF WCC (15 [IQR 0–35] vs. 0 [IQR 0–10] cells/mm3; p¼0.04). The mortality rate was similar in both groups (proven ABM 24% vs. probable ABM 28%; p¼0.80).
445
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
and without the availability of relevant laboratory tests. Limited supervision and access to continuing medical education in this setting may impair the correct identification of physical signs that depend on technical competence and informed interpretation. In the present study, we aimed to determine retrospectively the accuracy of the initial ward-based diagnosis compared with the final diagnosis derived independently from detailed research data.
J. Aipit et al.
Table 1. Numbers of laboratory-confirmed proven and probable acute bacterial meningitis cases according to initial diagnosis Not ABM by initial diagnosis n¼241 (%)
Total n¼481 (%)
64 (26.7) 29 (12.1) 35 (14.6) 176 (73.3)
20 (8.3) 9 (3.7) 11 (4.6) 221 (91.7)
84 (17.5) 38 (7.9) 46 (9.6) 397 (82.5)
ABM: acute bacterial meningitis.
When compared with the final diagnosis, an initial diagnosis of ABM had a sensitivity, specificity, positive predictive value and negative predictive value of 76% (95% CI 66–85%), 56% (95% CI 51–61%), 27% (95% CI 21–33) and 92% (95% CI 87–95%), respectively. The positive and negative likelihood ratios for an initial diagnosis of ABM were 1.72 and 0.43, respectively. For children with ABM that was proven by culture or bacterial antigen testing, the sensitivity, specificity, positive predictive value and negative predictive value were 76% (95% CI 60–89%), 51% (95% CI 47– 56%), 12% (95% CI 8–16) and 96% (95% CI 93–98%), respectively. There was discordance between initial and final diagnosis of ABM in 196 children. Of children with an initial clinical diagnosis of ABM, 73.3% (176/240) did not have a final diagnosis of ABM. This included 28 children (15.9%; 28/176) with a normal level of consciousness (Blantyre coma score¼5), without signs of meningeal inflammation and no history of convulsions. Twenty children without an initial diagnosis of ABM had a final diagnosis of ABM. In this group of children, neck stiffness (60%; 12/20), Kernig’s sign (40%; 8/20) and deep coma (45%; 9/20) were commonly detected. A comparison of the clinical features and outcome of the two discordant groups are shown in Table 2. Neck stiffness, Kernig’s sign, deep coma and the presence of malaria parasites all had significantly different proportions between the discordant groups. Malaria or pneumonia accounted for 54.9% (218/397) of admissions in children without a final diagnosis of ABM (Table 3). Of the 481 children, a total of 40 (8.3%; 40/481) died and 56 (11.6%; 56/481) were discharged with at least some neurological disability. Of the 64 children in who an initial diagnosis of ABM was confirmed by the laboratory, 15 (23.4%; 15/64) died compared with 7 (35.0%; 7/20) of 20 with ABM where an initial diagnosis was not made (p¼0.38).
Discussion Early diagnosis and treatment of ABM is a critical issue that often determines prognosis in children with ABM. In countries such as PNG, children typically present at a late-stage of illness to hospitals and other healthcare facilities that have poor infrastructure and a lack of trained staff. The present study from a provincial hospital in PNG highlights the limitations of the initial clinical diagnosis of ABM and suggests that simple clinical algorithms may have
446
Initial ABM + final Initial ABM – final p-value diagnosis ABM – diagnosis ABM+ n¼176 (%) n¼20 (%) Age Neck stiffness Kernig’s sign Bulging fontanela Deep coma Convulsion(s) Malaria Outcome Died Survived with neurological sequelae Survived without neurological sequelae
29 (756) 39 (22.2) 11 (6.3) 1 (1.2) 33 (18.8) 118 (67.0) 39 (22.2)
29 (17–42) 12 (60) 8 (40) 0 (0) 9 (45) 12 (60) 2 (10)
12 (6.8) 18 (10.2)
7 (35) 3 (15)
146 (83.0)
10 (50)
0.43 ,0.001 ,0.001 1.00 0.02 0.62 0.05 0.002
ABM: acute bacterial meningitis. Bulging fontanel assessed in children ≤18 months (n¼84).
a
Table 3. Final diagnoses in children initially diagnosed with acute bacterial meningitis but proven not to have acute bacterial meningitis on cerebrospinal fluid investigations Final diagnosis
Frequency n¼397 (%)
Severe malaria Pneumonia Non-severe malaria Infections with other focus Febrile convulsions Encephalopathy Diarrhoeal illness Neurological disorders Tuberculosis Non-infective causes Rheumatic fever
92 (23.2) 67 (16.9) 59 (14.9) 59 (14.9) 38 (9.6) 27 (6.8) 25 (6.3) 14 (3.5) 7 (1.8) 7 (1.8) 2 (0.5)
potential value in this setting. Admitting local clinicians had a high index of suspicion for ABM but their initial diagnoses were often inaccurate, even when objective clinical signs of meningeal irritation were present. There are understandable influences on the clinical assessment of a severely unwell child in the rural tropics where malaria is common, and a detailed history of relevant
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
ABM Proven ABM Probable ABM Not ABM
ABM by initial diagnosis n¼240 (%)
Table 2. Clinical features and outcome for children with discordant initial and final laboratory-confirmed diagnoses of acute bacterial meningitis. Data are in median (interquartile-range) or numbers (percentage)
Transactions of the Royal Society of Tropical Medicine and Hygiene
assessing children with possible ABM. Although not formally validated, our data suggest that a standardized approach is likely to be more accurate than the present ward-based approach. The present data suggest that a standardized approach to clinical assessment and the availability of light microscopy and/or rapid diagnostic tests for malaria should guide initial clinical diagnosis. A simple algorithm incorporating key clinical features (neck stiffness, Kernig’s sign and deep coma) and the results of parasitological testing could be developed. Its uptake and accuracy in the field could be assessed through randomization of individual provincial hospitals with similar infrastructure and caseloads to usual care or educational intervention familiarizing clinicians with the ABM algorithm, with independent monitoring of diagnostic accuracy in each case. Such a detailed investigation was beyond the scope of the present study. Our study had limitations. Although every attempt was made to ensure the independence of the initial clinical diagnosis, the retrospective nature of the study cannot rule out instances in which the initial diagnosis was influenced by research methods and data. However, because the present comparative study was not pre-specified, it is unlikely that the practices and procedures of the admitting clinicians changed materially during the study. Second, because only children who had a LP were recruited, this may not be representative of all children presenting with severe illness. Nevertheless, as the final diagnosis was based on objective CSF findings, this bias was unavoidable.
Conclusions The present study shows that an initial ward-based diagnosis of ABM made by attending clinicians has limited diagnostic utility in PNG, resulting in substantial over-diagnosis and frequent missed diagnoses of ABM in children at high risk of death and neurological sequelae. This situation argues for the implementation and validation of improved clinical diagnostic algorithms complemented by better availability of relevant laboratory tests.
Authors’ contributions: JA and ML contributed equally to this work. JA, IH and NP were involved in the clinical management of patients; ML, CB, LM, PS and TMED were involved in the prospective component of this study; ML conceived the study and wrote the first draft of this manuscript; LM and TMED edited the manuscript. All authors read and approved the final manuscript. LM and TMED are guarantors of the study. Acknowledgements: We thank the children and their guardians for their participation. We also thank the staff of the Paediatric Ward at Modilon Hospital and research nurses and support staff of the PNG Institute of Medical Research for clinical and logistics support. Funding: This study was funded by the National Health and Medical Research Council (NHMRC) of Australia [grant #513782]. We also acknowledge infrastructure funding support from the Malaria Genetic Epidemiology Network Consortium. Competing interests: None declared. Ethical approval: The prospective component of this study was approved by the PNG Institute of Medical Research Institutional Review Board and the Medical Research Advisory Committee of the PNG Health Department (MRAC 07.37).
447
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
events such as pre-presentation empirical antibiotic administration and convulsions can be difficult to obtain. In addition, there are published data indicating minimal13 through to some clinical utility14 for signs of meningeal irritation when considered with other clinical signs or as part of prognostic scores that incorporate clinical and laboratory data.15 Nevertheless, neither the positive (1.72) nor negative (0.43) likelihood ratios indicate that an initial diagnosis usefully differentiated ABM from other severe illnesses. By contrast, our previous report from Modilon Hospital showed that clinical examination performed in a standardized manner by trained clinicians had reasonable diagnostic utility for ABM.3 The positive likelihood ratios for neck stiffness (4.3), Kernig’s sign (5.3) and bulging fontanel (7.6) were higher than an initial diagnosis of ABM and proved independent predictors of ABM.3 In addition, the presence of malaria parasites (positive likelihood ratio 0.04) also had good diagnostic utility in ruling out a diagnosis of ABM in this setting. The initial usual-care clinical assessment in the present study both missed and overdiagnosed ABM. A missed diagnosis (24% of ABM cases) is likely to have a major impact in a setting where diagnostic facilities and therapeutic options are limited, and the risk of death in an individual child is extremely high. Indeed, the mortality rate was 35.0% in children with ABM when the initial diagnosis of ABM was missed, compared with 23% when the initial diagnosis of ABM was correct. Although this difference was not statistically significant, it highlights that a missed diagnosis is associated with an outcome similar to that in a child with an initial diagnosis of ABM correctly made by the attending clinician. Indeed, it is possible that the consequences of a missed diagnosis of ABM in this study were attenuated by the empirical use of both artemether and antibiotic therapy, regardless of diagnosis. In addition, children with a missed diagnosis often had many of the clinical signs typical of meningeal irritation yet were not considered to have ABM. The discordant results do not appear to have occurred by chance alone. The detection of neck stiffness, Kernig’s sign and deep coma by study staff differed significantly from that of the local clinicians, perhaps reflecting a non-standardized approach to clinical examination outside a research setting. The PNG Department of Health recently changed the standard treatment guidelines for ABM. Chloramphenicol has been replaced by ceftriaxone and is now indicated only for severe pneumonia and some undifferentiated infections.9 One consequence of this change is that an attending clinician will now have to make an important practical decision for all severely unwell children, namely whether or not to prescribe ceftriaxone. The overdiagnosis of ABM in this study means that a high proportion of children without ABM receive ceftriaxone unnecessarily, resulting in increased costs to a health system that is already struggling with limited resources. Furthermore, widespread use of ceftriaxone will inevitably lead to an increased risk of antimicrobial resistance in hospital- and community-acquired bacterial pathogens. Our data contrast with similar studies performed in African children where up to 30% of proven ABM cases in Kenyan children would be missed if the diagnoses were based on clinical signs and CSF turbidity alone.16 However, these studies did not examine the use of an initial ABM diagnosis made independently as part of usual care. Large variation in clinical experience, local ABM and malaria epidemiology, and diagnostic facilities mean that our data may not be applicable to settings outside PNG. Nevertheless, the data from this study highlight difficulties for all clinicians when
J. Aipit et al.
References 1 Berkley JA, Versteeg AC, Mwangi I et al. Indicators of acute bacterial meningitis in children at a rural Kenyan district hospital. Pediatrics 2004;114:713–9.
3 Laman M, Manning L, Greenhill AR et al. Predictors of acute bacterial meningitis in children from a malaria-endemic area of Papua New Guinea. Am J Trop Med Hyg 2012;86:240–5. 4 Berkley JA, Mwangi I, Mellington F et al. Cerebral malaria versus bacterial meningitis in children with impaired consciousness. QJM 1999;92:151–7. 5 Muller I, Bockarie M, Alpers M, Smith T. The epidemiology of malaria in Papua New Guinea. Trends Parasitol 2003;19:253–9. 6 Lehmann D, Yeka W, Rongap T et al. Aetiology and clinical signs of bacterial meningitis in children admitted to Goroka Base Hospital, Papua New Guinea, 1989–1992. Ann Trop Paediatr 1999; 19:21–32.
10 Manning L, Laman M, Greenhill AR et al. Increasing chloramphenicol resistance in Streptococcus pneumoniae isolates from Papua New Guinean children with acute bacterial meningitis. Antimicrob Agents Chemother 2011;55:4454–6. 11 Michon P, Cole-Tobian JL, Dabod E et al. The risk of malarial infections and disease in Papua New Guinean children. Am J Trop Med Hyg 2007;76:997–1008. 12 Manning L, Laman M, Law I et al. Features and prognosis of severe malaria caused by Plasmodium falciparum, Plasmodium vivax and mixed Plasmodium species in Papua New Guinean children. PloS One 2011;6:e29203. 13 Waghdhare S, Kalantri A, Joshi R, Kalantri S. Accuracy of physical signs for detecting meningitis: a hospital-based diagnostic accuracy study. Clin Neurol Neurosurg 2010;112:752–7. 14 van de Beek D, de Gans J, Spanjaard L et al. Clinical features and prognostic factors in adults with bacterial meningitis. N Engl J Med 2004;351:1849–59.
7 Weber MW, Herman J, Jaffar S et al. Clinical predictors of bacterial meningitis in infants and young children in The Gambia. Trop Med Int Health 2002;7:722–31.
15 Oostenbrink R, Moons KG, Donders AR et al. Prediction of bacterial meningitis in children with meningeal signs: reduction of lumbar punctures. Acta Paediatr 2001;90:611–7.
8 Laman M, Manning L, Hwaiwhange I et al. Lumbar puncture in children from an area of malaria endemicity who present with a febrile seizure. Clin Infect Dis 2010;51:534–40.
16 Berkley JA, Mwangi I, Ngetsa CJ et al. Diagnosis of acute bacterial meningitis in children at a district hospital in sub-Saharan Africa. Lancet 2001;357:1753–7.
448
Downloaded from http://trstmh.oxfordjournals.org/ at Louisiana State University School of Veterinary Medicine Library on February 22, 2015
2 Peltola H. Burden of meningitis and other severe bacterial infections of children in Africa: implications for prevention. Clin Infect Dis 2001;32:64–75.
9 Paediatrics Society of Papua New Guinea. Standard Treatment for common illnesses of children in PNG. Port Moresby: Paediatrics Society of Papua New Guinea; 2011.
