Clinically based surveillance of invasive meningococcal disease in young children admitted to selected US hospitals between January 2000 and June 2009

Abstract

Invasive meningococcal disease (IMD) is under-reported in countries that do not employ polymerase-chain reaction for surveillance because culture-negative cases are omitted. To evaluate a clinically based, case-finding method, we developed case definitions for “probable,” “compatible with,” and “possible, but unlikely” IMD, respectively, based on supportive documentation (e.g., discharge diagnosis of meningococcal infection, culture-negative bacterial meningitis, petechiae/purpura, Gram-negative diplococci on Gram stain) and weight of clinical evidence, which we then applied to electronic health records for all children aged ≤5 y who were admitted to approximately 100 US hospitals between January 2000 and June 2009. Among 47,863 qualifying admissions, 16 children had culture-positive IMD, 5 had “probable” IMD, and 5 had illness “compatible with” IMD. Five additional children had disease considered “possible but unlikely” IMD. Our case-finding methods suggest that culture-based ascertainment may underestimate the number of IMD cases by 31–63%, supporting findings in other nations that culture-based reporting provides incomplete information on disease incidence and therefore underestimates the potential benefits of routine vaccination of young children against meningococcal disease.

Keywords: :

• Neisseria meningitidis
• PCR
• culture
• invasive meningococcal disease
• meningitis
• meningococcemia
• retrospective studies
• sepsis

Introduction

Invasive meningococcal disease (IMD) occurs suddenly in otherwise healthy persons, most frequently infants and young children, and causes permanent disability or death in a significant proportion of patients.¹ Many patients suffer from lifelong morbidity, and case fatality rates (CFRs) have remained unchanged in many developed countries despite advances in treatment.^2,3 In the US, the CFR for IMD as reported by the US Centers for Disease Control and Prevention (CDC) Active Bacterial Core surveillance system (ABCs) was 11.3% between 1998 and 2007.²

Due to the sudden onset of the disease, the best strategy for reducing disease burden is vaccination. Three vaccines have been approved for use in the US–2 conjugated vaccines, and one unconjugated vaccine. Compared with the unconjugated vaccine, conjugated vaccines have demonstrated superior efficacy against IMD and/or immunogenicity in older children and adults.¹ Two problems with meningococcal vaccines still must be overcome. First, an effective vaccine against serogroup B disease—currently, the highest incidence serogroup in the US—is not yet available. Second, immunogenicity in children younger than 2 y of age (where disease burden is greatest) is more variable relative to persons who are older. In the US, the Advisory Committee on Immunization Practice (ACIP) currently recommends routine vaccination (including a booster) of adolescents (ages 11–18 y) with conjugate vaccine. Vaccination is also recommended for persons aged 9 mo to 54 y at increased risk for meningococcal disease (complement component deficiencies, asplenia, HIV, microbiologists, or travelers to countries with epidemic/highly endemic meningococcal disease).⁴

In the US, epidemiology and surveillance of IMD is based on recovery of N. meningitidis from a normally sterile body site as part of the ABCs, which includes data from an overall surveillance population of approximately 40 million people in 10 states.² Appropriate primary care for young children presenting with fever of unknown source which may be due to invasive bacterial diseases, including IMD, may warrant parenteral antibiotics (e.g., ceftriaxone, cefotaxime), which are likely to render subsequent blood or cerebrospinal fluid (CSF) cultures negative. Further factors that can result in under-ascertainment of cases when relying on culture results in young children include resolution of the early bacteremic phase of illness prior to sample collection, small blood volumes, and technical errors. The incidence of culture-confirmed IMD among infants and young children reported by the CDC therefore represents a conservative estimate of the burden of IMD, an observation supported by passive surveillance data from the National Notifiable Diseases Surveillance System (NNDSS), which includes “probable” cases and reports proportionally more cases compared with ABCs.^1-3

In countries such as the United Kingdom, Spain, Brazil, and South Africa, polymerase-chain reaction (PCR) assays are used to detect N. meningitidis DNA in patients with clinically apparent IMD but sterile blood and/or CSF cultures. Recent studies focusing on PCR suggest that culture-based ascertainment can underestimate the number of cases of IMD by 39–100%.^5-8 PCR testing, however, is not customarily performed in the US and many other countries worldwide.

Alternative means of identifying probable cases of IMD are valuable to supplement established estimates of disease burden, and also to aid in establishing policies and practices for prevention and clinical management of IMD. We developed a case-finding method to identify possible cases of IMD and applied it to electronic medical records in the Cerner Health Facts Database. Specifically, the objective of our study was to develop a case-ascertainment system for IMD that allowed for detection of cases of culture-negative disease in the absence of PCR, and to estimate the burden of such cases in relation to cases of culture-confirmed disease among hospitalized infants and young children in the US.

Results

Source Population

In the Cerner Health Facts database, a total of 47,863 qualifying admissions were identified based on the criteria in Table 1.

Table 1. The source population initially assessed

Live-born admitted children 5 years of age or younger^* with one or more of the following
Diagnosis	ICD-9 code
Meningococcal infection	0.36XX
Bacterial meningitis	320.XX
Other meningitis	321.XX
Bacteremia	790.7
Septicemia	038.XX
Sepsis of newborns	771.81
Unspecified viral infection	079.XX
Fever	780.6
Systematic inflammatory response	995.91, 995.92
Tests performed
Lumbar puncture	NA
Microbiological CSF culture	NA
Any other test (culture, Gram stain, antigen test) for N. meningitidis	NA

* Children with cancer were excluded

Definite IMD cases

Sixteen patients with “definite” IMD were identified, all of whom had blood and/or CSF cultures positive for N. meningitidis. Six cases were CSF-confirmed, and 7 were blood culture-confirmed; 12 had evidence of a positive antigen test or Gram stain. Eleven of these 16 patients had an admission and/or discharge diagnosis of meningococcal infection (for ten patients, meningococcal infection was the principal diagnosis; for the other, it was “infection and inflammatory reaction due to nervous system device implant and graft”). Not all encounters had information about medications administered; however, 13 of 16 patients with culture-confirmed IMD had evidence of receipt of parenteral antibiotics with activity against N. meningitidis (e.g., ceftriaxone, cefotaxime) for at least 3 d (See Table 4).

Table 2. Clinical syndromes of interest among children in the source population who did not have culture-confirmed (“definite”) IMD

Clinical syndrome^*	Definition
Serious bacterial infection   (n=5800)	Continuous administration of parenteral antibiotics for 4 days or more –plus- Length of day of 5 days or more
Meningitis   (n=978)	Admitting or discharge diagnosis of meningitis (ICD-9 320.9, 322.9) –plus- Cerebrospinal fluid white blood cell count >10 based on earliest value within 24 hours of initial presentation
Sepsis   (n=8142)	Admitting or discharge diagnosis of volume depletion/dehydration/hypovolemia (276.5), sepsis (038, 038.X), shock without mention of trauma (785.5X) or multi-organ failure (995.92) -or- Receipt of vasoactive medications (e.g., dopamine) or dialysis in patients with no evidence of trauma or poisoning

^*Categories not mutually exclusive. Total identified was 12,156 from the initial source population of 47,863.

Clinical syndromes of interest

Of the remaining culture-negative admissions, 12,156 had evidence of at least one clinical syndrome of interest (Table 2): serious bacterial infection (n = 5800), meningitis (n = 978) and/or sepsis (n = 8142). We excluded 394 patients who underwent major surgery and 3721 who had an alternative infectious etiology (e.g., viral meningitis, positive cultures for other bacteria). Detailed case review (Table 3) was then performed for the 86 patients with the predefined findings of interest: petechiae/purpura (n = 72), Gram-negative diplococci on Gram stain (n = 14), and/or DIC (n = 32). Severity of illness was very high for all 86 cases chosen for detailed case review.

Table 3. Predefined hierarchical taxonomy and case definitions for definite or suspected invasive meningococcal disease

Category	Case definition
Definite IMD	Positive blood or cerebrospinal fluid culture for N meningitidis
Probable IMD^*	Clinical findings consistent with IMD (petechiae/purpura, low platelet count, elevated international normalized ratio, or prothrombin time)   -and- Clinical diagnosis of meningitis or meningococcemia
Compatible with IMD^*	Clinical findings consistent with IMD
Possible but unlikely^*	Some clinical findings consistent with IMD, but with additional clinical findings (e.g., hospital length of stay ≤3 days in a disease survivor) that made IMD less likely
Unlikely IMD^*	Disease inconsistent with IMD findings

* Not applied to children with culture-confirmed IMD; these definitions were applied to 86 cases of children with serious bacterial infection, bacterial meningitis, or sepsis and further findings of interest (e.g., petechiae/purpura, Gram-negative diplococci on Gram stain, and/or disseminated intravascular coagulation), and who were hospitalized for three days or longer.

Table 4. Clinical Data describing thirty-one patients meeting case definitions for definite or suspected invasive meningococcal disease (IMD)

Hierarchrical Taxonomy of IMD	Prinicpal diagnosis of IMD	Serious bacterial infection	Meningitis	Evidence of sepsis		Received agent of interest for ≥ 3 days	No other infectious etiology
				Clinical	Lab
Definite* N=16	10	11	11	3	11	13	16
Probable N=5	5	5	2	4	5	5	5
Compatible with N=5	0	5	1	5	5	5	5
Possible but unlikely N=5	0	3	2	4	5	3	3

* Diagnostic information was missing for 5 cases; information on medications received was missing for 3 cases

Classification of patients with clinical syndromes and predefined findings of interest

Following detailed case review of clinical and laboratory data for 86 patients with clinical syndromes believed to have high positive predictive value for IMD, we identified 5 additional cases with “probable” IMD, and 5 cases that were deemed “compatible with” IMD. Five additional patients had disease designated as “possible but unlikely” IMD. (see Table 4 and Appendix).

The 5 patients designated as having “probable” IMD each had specific, recorded diagnoses of meningococcal infection, with supportive evidence of clinical signs and/or laboratory evidence of sepsis. Each of these 5 children received a parenteral antibiotic with activity against N. meningitidis for at least 3 d, and had no evidence of any other infectious etiology.

All 5 patients with illness “compatible” with IMD received parenteral antibiotics with activity against N. meningitidis for at least 3 d and had both clinical signs of and laboratory support for a diagnosis of sepsis. None of these 5 patients had any other infectious etiology established.

The 5 patients who were designated as having “possible but unlikely” IMD had laboratory support for sepsis. Of these, 4 had clinical signs of sepsis, and 3 were treated for at least 3 d with parenteral antibiotics with activity against N. meningitidis. All 5 patients also had evidence of possible other underlying conditions, or had features not classically associated with IMD. For example, 2 of the 5 patients received antibiotic therapy for less than 3 d.

The Appendix contains a detailed summary of the clinical characteristics of the above-mentioned “definite,” “probable,” “compatible,” and “possible but unlikely” cases.

Sensitivity

Considering “probable” cases only, our estimate of the sensitivity of culture-based ascertainment is 76.2% (i.e., 16/21); considering both “probable” and “compatible” cases, we estimate sensitivity to be 61.5% (i.e., 16/26).

Discussion

Disease surveillance is an important means of monitoring public health. Optimal surveillance systems can minimize the impact of disease by identifying populations at relatively high risk, specific risk factors for disease, and trends in disease incidence over time. Results from epidemiologic surveillance inform allocation of public health resources and policy; accordingly, improvements in surveillance should lead to more informed planning and implementation of public health policy, including recommendations for prophylaxis, screening, and immunization.

Prevention and control of meningococcal disease is monitored worldwide by many groups, including the US CDC.^1,2,9 Some surveillance methods, like those used by the ABCs in the US, require culture confirmation from a normally sterile body site. These methods to measure the clinical burden of IMD were chosen to be specific, but probably result in missed cases. The choice for specificity is understandable; however, the number and type of additional cases is of interest for clinicians and policymakers. While culture-based surveillance is often regarded as a gold standard, countries in Europe, Africa, the Pacific Region and South America that include cases ascertained by PCR of body fluids report up to twice as many cases.^5-8

To provide more complete information about IMD cases in the absence of PCR, we developed clinical case definitions as a means of identifying culture-negative cases of IMD, based on the weight of available evidence (i.e., probable, compatible, possible but unlikely). In our review of records of hospital admissions of infants and young children to more than 100 acute-care facilities across the US, we identified a number of culture-negative cases that we deemed likely to have been IMD. The number of cases of “probable” and “compatible” IMD in our study (n = 10) represented a substantial proportion relative to culture-confirmed cases (n = 16), which also suggests that NNDSS estimates, although approximately 10% higher than those of the ABCs, also underestimate the true incidence of IMD. Our findings are more consistent with reports based on PCR indicating that >39% of IMD cases in the UK, South Africa, Brazil and Spain are culture-negative.^5-8

While we identified several additional culture-negative cases of IMD based on clinical diagnosis and management information, it is likely that our analysis was conservative. We limited our review of medical records to patients who were admitted to acute-care facilities in the Cerner Health Facts Database; therefore, we excluded all young children, even those who were culture-positive, who presented to the emergency room with IMD and died there, received treatment there and were sent home, or were transferred to another facility prior to admission. In addition, during detailed case review, we only examined cases for children with hospital stays of 3 or more days. Since IMD can be fatal within 48 h, our analysis would have excluded children who died within this window. While rare, to the extent that early antibiotic therapy obviated the need for hospitalization, we also would have excluded patients with mild disease who were treated exclusively on an outpatient basis.

Our case-finding method was based on the weight of available clinical evidence and known characteristics of IMD. We developed case definitions including clinical characteristics that are well-recognized as associated with IMD in this age group prior to undertaking any review of patient materials. In addition, once the source population was identified, we ruled out categories, such as children with cancer, that seemed unlikely to yield additional IMD cases. Finally, our case definition for probable IMD required a formal diagnosis by the attending physician.

A limitation of our study is that we did not have access to medical narratives and other records that could have provided added clinical detail. It is likely that a review of actual medical records and/or death certificates would have yielded additional clinical information that would have aided our clinical assessments. Further shortcomings, including errors in coding and missing information, are common to retrospective studies that use administrative databases. In our study, not all subjects had comprehensive laboratory values, which somewhat hindered our ability to assess various characteristics such as signs of sepsis. Another limitation of this study that also affects many retrospective studies that employ electronic databases is that the available data represent a convenience sample, as opposed to a random sample of all US hospitals. The hospitals that contribute information to the database may differ from those that are not represented, which could limit the generalizability of these records to all medical centers across the US.

A prospective study comparing PCR and culture-based diagnostic techniques would improve our understanding of the actual burden of meningococcal disease in the US. Only limited samples of PCR data are available in the US and these are insufficient to make a complete assessment.¹⁰ Unfortunately, PCR testing is not widely performed in the US; it also is not utilized in many other parts of the world, although its use is growing. Since our clinically based algorithms yield estimates of the burden of IMD that are consistent with previously published incidence estimates based on PCR, we believe our methods may prove useful in overcoming the limitations of culture-based surveillance for IMD in areas where PCR is unavailable.

Although young children are at greatest risk of IMD, persons of others ages also can develop the disease. Smaller peaks in incidence occur in persons aged 15–24 y and those ≥65 y of age.² While the current findings may not be generalizable to all persons, a similar method of case review, based on predefined parameters chosen to reflect the clinical characteristics of IMD in additional age groups, could provide added information about disease incidence.

We employed case definitions based on accepted parameters of clinical recognition to identify probable cases of IMD in infants and young children. Our findings suggest that, in the absence of PCR, careful consideration of clinical information represents a possible method for identifying likely cases of IMD. A fuller understanding of the extent of IMD can provide a more complete basis for assessing the potential value of routine vaccination of young children.

Materials and Methods

Data Source

We employed data from the Cerner Health Facts^® Database, an electronic data warehouse containing comprehensive anonymized clinical records from hospital admissions, emergency department encounters, and outpatient or clinic visits at more than 100 contributing acute-care, short-term hospitals in 26 states across the continental US. As no patient or provider contact was made and patient information was de-identified, Institutional Review Board (IRB) approval was not required.

Data for each hospital admission in the database typically includes: patient demographics (age, gender, race), hospital characteristics (geographic region, number of beds, teaching status), principal and secondary discharge diagnoses in ICD-9-CM format, length of stay, discharge status and disposition, drug therapy administered, laboratory information including culture reports and sensitivities, and in-hospital charges. Only partial data were available for admissions between January 2003 and March 2005. Clinical case narratives (history and physical, progress notes, discharge summary) and death certificates were not included for any cases. We considered only cases for hospitalized children; we did not consider patients treated only in emergency departments.

Source Population

We selected all hospital admissions between January 1, 2000 and June 30, 2009 among children aged ≤ 5 y with any of a set of prespecified discharge diagnoses and laboratory tests (Table 1). Children with cancer were excluded because many of the classic clinical signs of IMD (e.g., petechiae, purpura, thrombocytopenia) in these persons were deemed much more likely to be due to other causes (e.g., other infections, bone marrow toxicity).

Case Ascertainment

Case ascertainment began with the identification of all cases of definite, culture-positive IMD in the database. We next identified all culture-negative cases with clinical syndromes known to be consistent with IMD (See Table 2); exclusion criteria were applied to those with ≥ 1 such syndromes to eliminate patients who were unlikely to have IMD. A hierarchical taxonomy (See Table 3) was developed to reflect the relative likelihood of IMD, based on the weight of available clinical evidence. “Definite,” culture-confirmed, IMD was considered to be the top of the taxonomical hierarchy.

Identification of definite IMD cases

All patients in the source population with blood or CSF cultures that were drawn within 24 h of initial clinical presentation and positive for N. meningitidis were designated as “definite” IMD, consistent with the definition used by ABCs.²

Clinical syndromes of interest in patients without definite IMD

To identify cases of disease that would be suggestive of—or consistent with—IMD, we focused on hospital admissions with evidence of three “clinical syndromes”: serious bacterial infection, meningitis, and sepsis (See Table 2 for detailed definitions). These syndromes were selected because they were thought to represent clinical presentations consistent with IMD, and because the likelihood of IMD in children without any of these three syndromes was thought to be extraordinarily low.³

We eliminated patients who underwent major surgery or had evidence of alternative infectious etiology (i.e., microbiologically confirmed viral, fungal, other bacterial infection), since IMD in the context of these events was deemed highly unlikely. Among other characteristics, petechiae/purpura, Gram-negative diplococci on Gram stain, and/or disseminated intravascular coagulation (DIC) were chosen a priori as findings of interest for detailed case analysis, based on their well-known association with IMD (i.e. ref. 11); therefore, only patients with these findings were evaluated further. Patients with a hospital stay of less than three days also were excluded from detailed case review.

Classification of cases of suspected IMD by findings and weight of clinical evidence

Prior to case review, we developed a hierarchical taxonomy of categories ranging from probable IMD to unlikely IMD, based on clinical assessment of available information and the level and weight of clinical evidence (see detailed definitions in Table 3). The taxonomy was arranged in descending order of evidence, from “probable IMD” to “unlikely IMD” (“definite” [i.e., culture-confirmed] was at the top of the hierarchy). Detailed case review was undertaken independently by three authors (SJP, JK, JJS), who categorized each admission, following the prespecified definitions (Table 3). Inter-rater differences in classification were discussed by the three raters to achieve consensus.

Sensitivity

The sensitivity of culture to identify cases of IMD was evaluated by dividing the number of definite cases by a denominator that included: “definite” and “probable” IMD; or “definite,” “probable,” and “compatible with” IMD. We excluded cases deemed “possible but unlikely” to provide conservative estimates of sensitivity.

Appendix. Additional detail on admitted patients with “definite”, “probable”, “compatible”, and “possible but unlikely” IMD in multi-hospital database

	Comments
“Definite” IMD (n=16)
Patient #3575	Female, 4 years of age; CSF culture positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis
Patient #7966	Female, 2 years of age; blood culture positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis; also had culture finding of Gram-positive clusters
Patient #23828	Male, <1 years of age; CSF culture positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis
Patient #37517	Female, <1 years of age; CSF culture positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis
Patient #68445	Male, 3 years of age, CSF culture positive for N. meningitidis; antigen test positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis
Patient #238089	Male, 3 years of age, blood culture positive for N. meningitidis; Gram-negative diplococcus also found; patient transferred to another hospital after only 1 day
Patient #248667	Male, <1 years of age, blood culture positive for N. meningitidis
Patient #272475	Male, 1 year of age, blood culture positive for N. meningitidis
Patient #349197	Female, <1 years of age, CSF culture positive for N. meningitidis
Patient #360706	Male, 3 years of age, blood culture positive for N. meningitidis; all other information missing
Patient #420593	Female, 4 years of age, CSF culture positive for N. meningitidis; all diagnostic information missing
Patient #439722	Male, <1 years of age, CSF culture positive for N. meningitidis; all diagnostic information missing
Patient #539781	Female, <1 years of age, CSF culture positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis
Patient #554140	Male, 1 year of age, blood culture positive for N. meningitidis; Gram-negative cocci also identified; principal diagnosis of epidemic cerebrospinal meningitis
Patient #568217	Male, <1 years of age, blood culture positive for N. meningitidis; admitting diagnosis of meningococcal meningitis
Patient #577535	Male, 1 year of age, CSF culture positive for N. meningitidis; all diagnostic information missing
“Probable” IMD (n=5)
Patient #509567	Female, <1 years of age, discharge diagnoses of meningococcemia, sepsis, and petechiae; no cultures reported; LOS of 10 days
Patient #332824	Female, 5 years of age, discharge diagnoses of meningococcemia, epidemic meningitis, septic shock and DIC; LOS of 16 days
Patient #282256	Male, <1 years of age, discharge diagnoses of meningococcemia, epidemic cerebrospinal meningitis, DIC, and volume depletion; LOS of 6 days
Patient #215629	Male, <1 years of age, antigen test positive for N. meningitidis; principal diagnosis of epidemic cerebrospinal meningitis
Patient #100945	Female, 1 year of age, discharge diagnoses of meningococcemia, septic shock, severe sepsis, and corticoadrenal insufficiency; LOS of 11 days
“Compatible” IMD (n=5)
Patient #525825	Female, <1 years of age, admitting diagnosis of unspecified septicemia; discharge diagnoses of septic shock, DIC, grand mal seizure, rhabdomyolysis, and renal failure; LOS of 20 days
Patient #463542	Male, 2 years of age, discharge diagnoses of DIC, severe sepsis, unspecified septicemia, cerebral infarction, inhalation pneumonia, thrombocytopenia, and dehydration; LOS of 40 days
Patient #117345	Male, 1 year of age, discharge diagnoses of Down’s syndrome and severe sepsis
Patient #121755	Female, <1 years of age, discharge diagnoses of unspecified septicemia, septic shock, severe sepsis, DIC, thrombocytopenia, congenital hypoplasia of ascending aorta, and gonadal dysgenesis syndrome
Patient #275582	Male, <1 years of age, admitting diagnoses of meningitis due to Gram-negative bacteria and rash; LOS of 18 days
“Possible but unlikely” IMD (n=5)
Patient #34184	Male, 1 year of age, admitting diagnosis of pneumonia, LOS of 5 days, respiratory tract Gram stain revealed Gram-negative diplococci; respiratory tract culture of H. influenzae
Patient #20671	Female, 3 years of age, discharge diagnoses of DIC, shock, and thrombocytopenia
Patient #155466	Male, 4 years of age, discharge diagnoses of familial polycythemia and sepsis
Patient #186066	Male, 1 year of age, discharge diagnoses of neutropenia due to infection, thrombocytopenia, and petechiae; LOS of 6 days
Patient #341678	Male, <1 years of age, discharge diagnoses of unspecified viral meningitis and petechiae; LOS of 4 days; receipt of IV cetriaxone and vancomycin; CSF PCR for enterovirus negative

IMD: Invasive meningococcal disease; DIC: Disseminated intravascular coagulation; PT: Prothrombin time; INR: International normalized ratio

Acknowledgments

We wish to thank Valerie Hastings, M.S., Leiming Li, Kim Seefeld, M.S., and Elvira Ponce, MD, MSc., for contributions to study conduct. Lisa DeTora, PhD, made substantial contributions to the writing of the manuscript.

Financial Support

Funding for this research was provided by Novartis Vaccines and Diagnostics

Contributors

A. Berger, S. Pelton, J. Klein, J. Stoddard, J. Edelsberg, and G. Oster were involved in study design and conceptualization, data analysis and interpretation, manuscript preparation and/or critical review and approval.