Epidemiology of Meningococcal Disease, New York City, 1989-2000

Alexandre Sampaio Moura, Ariel Pablos-Méndez, Marcelle Layton, Don Weiss


Emerging Infectious Diseases. 2003;9(3) 

In This Article


The study included New York City residents who met the case definition for confirmed or probable meningococcal disease, as defined by the Centers for Disease Control and Prevention and the Council of State and Territorial Epidemiologists. Inclusion in the study as a confirmed case required a clinically compatible course with the isolation of N. meningitidis from a sterile site (e.g., blood, cerebrospinal fluid, joint fluid, or pleural fluid); inclusion as a probable case required a positive antigen test from cerebrospinal fluid or clinically described purpura fulminans.[4] The period of study was January 1989-December 2000.

We obtained the meningococcal disease cases from the New York City Department of Health and Mental Hygiene (referred to hereafter as NYC Department of Health) surveillance database of reportable diseases. Meningococcal disease is a national reportable disease; in New York City, all cases are required by health code to be reported to the NYC Department of Health. Physician reports, investigation forms, and laboratory reports were reviewed for all the meningococcal disease patients included in the NYC Department of Health database. All cases with evidence of the study criteria were included. Data on meningococcal disease used in this study were collected through routine passive surveillance, and serogroup identification was performed by the NYC Department of Health Public Health Laboratory. Antibiotic resistance profiles and pulsed-field gel electrophoresis results were available only for a subset of isolates after 1999 and are not included in this report. Archival data and population estimates (before 1989) were obtained from New York City Vital Statistics Annual Summary reports.

The database contained information on each cast-patient's age, sex, race, ethnicity, borough of residence, and death. Information on race and ethnicity was incomplete and therefore was not analyzed. When death information was missing, patient identifiers were submitted to the New York City Vital Records and Registry for a death certificate search, which was accomplished by searching by name and International Classification of Diseases (ICD) code. Using name search, staff in the New York City Vital Records and Registry department used visual inspection to search the New York City death certificates, looking for the name of each patient with an unknown cause of death in the 1-month period after the date of onset of the disease. We also conducted a search using the ICD codes that correspond to meningococcal disease (ICD-9 036.0-036.9 and ICD-10 A39.0-A39.9); the search identified all death certificates from 1989 to 2001 that included these codes. We used the information found through this second search method if the death certificate referred to a patient already in the database with an unknown outcome. We did not include death certificates with meningococcal disease ICD codes that referred to patients not previously included in the database (i.e., they had not been reported to the NYC Department of Health as having meningococcal disease) because of the lack of data to confirm the diagnosis. Patients whose names did not appear in the death certificate search file were considered survivors in the CFR calculation. This study was based on electronic data and surveillance records; we ensured confidentiality by excluding all identifying information from the active analysis database.

Incidence rates were calculated by using 1990 and 2000 population files from the U.S. Census Bureau. We used the Pearson chi-square test or Fisher exact test to assess the statistical significance of categorical variables and the Kruskal-Wallis test to assess continuous variables.

Time trend analysis was performed to detect an association between time (e.g., year or year group) and response variables (e.g., serogroup and outcome). We used the Spearman correlation test and chi-square test for linear trends to assess statistical significance. Logistic regression models were built to provide coefficients for significant trends.

Independence can be assumed from the data because most cases were sporadic throughout the study period; we considered the vast majority of cases to be unrelated. In addition, no patient had more than one episode of the disease during the study period, and the analyses were performed with the patients grouped into 3-year intervals to minimize any existing correlation between sequential years.[5] The SPSS (SPSS Inc., Chicago, IL) statistical software package and Epi Info 2000 (Centers for Disease Control and Prevention, Atlanta, GA) software were used to perform the statistical calculations.