Data sources
Case data were sourced from two databases of routinely collected information for this population-based descriptive study: the national notifiable disease surveillance database (EpiSurv) operated by The Centre for Environmental Science and Research Ltd. (ESR) and The National Minimum Dataset (NMDS) held by the NZ MoH.
EpiSurv contains data on notifiable diseases and should, theoretically, contain all cases of notifiable disease resulting in hospital admission, as well as cases managed in the community without hospitalization. ESR commenced standardized on-line reporting in 1997 and therefore 1997 was elected as the start point for data analysis. Laboratory definitive evidence for a confirmed case of listeriosis requires identification of L. monocytogenes from a normally sterile site, including the foetal gastrointestinal tract, by either isolation (culture) of L. monocytogenes or detection of L. monocytogenes nucleic acid [24]. Only invasive disease is notifiable. Where illness has occurred in a pregnant woman, foetus or infant aged ≤28 days, the mother is notified as the case and the disease is recorded as pregnancy-associated.
The NMDS is a national dataset of public and private hospital discharge information that describes clinical data for inpatients and day patients, and is coded according to international standards, including the International Classification of Diseases – Clinical Modification 9 (ICD-9-CM). The original NMDS was implemented in 1993 and back-loaded with public hospital discharge information from 1988. The NMDS (hospital events) dataset used in this study contains information on the number of episodes of disease resulting in hospital admission.
Data were requested from both databases for all listeriosis cases (ICD-9-CM code 027.0) from 1 January 1997 to 31 December 2016 for all women of childbearing age, defined as 15 to 45 years (inclusive), and all children < 15 years of age. Children < 15 years were included as this is the New Zealand cut off for pediatric hospital admissions. Data were included where listeriosis was either the principal diagnosis or another relevant diagnosis. Data were not stratified by, or excluded due to, listeria serotype, owing to variations and inconsistencies in data reporting.
Data from the NMDS were identifiable by patient National Health Index (NHI) number, a unique identifier that is assigned to every person who uses health and disability support services in NZ. All children were included, irrespective of their predicted immune status (i.e. children with known immunocompromising conditions, such as leukemia, were included and were identified using the ICD-9-CM codes: 204.01 - Acute lymphoid leukemia, in remission, and 204.00 - Acute lymphoid leukemia, without mention of having achieved remission). Women were considered pregnant if they had any ICD-9 diagnosis codes starting with V22, V23, V27, V28, or 630 to 679 (codes identifying pregnancy) in the NMDS, or if the notification in ESR was recorded as being pregnancy-associated or pregnancy was indicated in any of the text fields.
Data extracted from EpiSurv included: disease name, report date, sex, age, date of birth, ethnicity, District Health Board (of admission), mortality status (including if from confirmed listeriosis or not), meningitis status, septicaemia status, swab/culture site, recent overseas travel (confirmed or not) and current immunosuppressant drug use. For pregnancy-associated cases, data also included delivery date and pregnancy outcomes (gestation, preterm delivery, and death of foetus or infant).
The data extracted from the NMDS included: event date, admission type (acute or planned), date of birth, gender, ethnic group (prioritized), domicile, District Health Board domicile, facility (of admission) and diagnosis (according to ICD-9 code and including up to 30 diagnosis codes). For children, data also included admission weight, birth weight and gestation at birth.
For calculation of incidence rates, denominator data was sourced from the NZ MoH and Statistics NZ. The National Maternity collection, collated by the MoH, contains antenatal and postnatal event data obtained from primary maternity services and the NMDS. National counts of the number of women giving birth by maternal age and prioritized ethnicity, as recorded in the National Maternity collection, are available from the MoH website [25] (www.health.govt.nz; accessed 13 February 2019). Annual summaries of the estimated resident population (ERP) and number of live births, categorized by age group and total response ethnicity, were obtained from the Statistics NZ website (www.stats.govt.nz; accessed 13 February 2019).
Statistical analysis
EpiSurv and NMDS data were imported into the statistical software ‘R’ (R version 3.5.3 (2019-03-11)) for cleaning and analysis. A dataset of pregnancy-associated cases, with each case comprising information on the pregnant woman/mother and foetus/neonate, was created by matching EpiSurv and NMDS data primarily on the time (date of hospital admission, notification or disease onset) and geographical location (District Health Board) of the event as this information was most consistently recorded. A dataset of cases in children aged 29 days to < 15 years was created by matching EpiSurv and NMDS data by date of birth. Matched cases were confirmed by comparing patient NHI, or when NHI was not recorded or available, patient date of birth or date of delivery. Patient characteristics were derived by combining date of birth, gender, and ethnicity information from both datasets. For cases with multiple ethnicities, ethnicity was prioritized according to the order: Māori, Pacific, Asian, then European/Other. For example, a case was labelled as Māori if either EpiSurv or the NMDS recorded them as such. Patient outcomes were considered to have occurred if recorded in either EpiSurv or the NMDS.
Crude incidence rates were calculated by dividing the total number of pregnancy-associated cases or childhood cases across the 20 year period by the sum of the annual number of women giving birth or estimate resident population of children aged < 15 years respectively. Crude rates by age group and ethnicity were calculated with 95% binomial ‘Wilson’ confidence intervals (CI). Generalized linear models followed by a likelihood ratio tests (LRT) were used to assess differences across categories and calculate crude and age adjusted incidence rate ratios (IRR) with 95% CI.
Ethics approval
This study was approved by the Central Health and Disability Ethics Committee, ethics reference number 17/CEN/227. Data were de-identified once pregnancy status was confirmed/disproved (as above) and once cross-referencing was completed. Informed consent was not required due to the retrospective nature of data collection.