Historic notification and hospitalization data were used to calculate annual rates of campylobacteriosis in New Zealand during 1980–2009 for notifications and 1996–2009 for hospitalizations. A detailed descriptive analysis was then undertaken to examine the epidemiology of campylobacteriosis for the 12-year period 1997–2008 on the basis of notified and hospitalized cases.

Campylobacteriosis notification data are collated nationally by the Institute of Environmental Science and Research Ltd from notifications made by medical practitioners to their local public health service. During the study period, >96% of these notifications were culture-confirmed cases, with the remainder being epidemiologically linked to confirmed cases. Hospitalization data are collated by the Ministry of Health from information supplied by public hospitals. Analysis of hospitalized cases was based on patients with a principal diagnosis code for Campylobacter enteritis (International Classification of Diseases, 9th Revision, Clinical Modification, code 008.43, or International Classification of Diseases, 10th Revision, code A04.5). These data were further selected to exclude hospital transfers, readmissions within 30 days, and day cases (i.e., patients assessed in hospital for a short time but not requiring an overnight stay). Admissions to private hospitals were excluded because few patients with infectious diseases are admitted to such institutions and documentation is inconsistent.

In the detailed descriptive analysis, temporal trends in disease incidence and distribution were examined according to patient age, sex, socioeconomic status, ethnicity, urban versus rural dwelling, region (health board area), and season. Case-patients were assigned rurality and deprivation scores on the basis of their home domicile. For rurality assignment, we used a Statistics New Zealand classification system, which defines 7 grades of rurality on the basis of population size and employment address (18). Socioeconomic status was measured by deprivation scores assigned according to the New Zealand Deprivation Index, an area-based measure of socioeconomic position derived from the 5-year Census of Population and Dwellings (19).

The main descriptive analysis rates were calculated by using interpolated and extrapolated Census Usually Resident population data from 1996, 2001, and 2006. Rates for 2007 and 2008 (with 2007 being the transition year, on the basis of the gradual implementation of interventions) were compared with the average annual rates for 2 baseline periods (1997–2001 and 2002–2006). For the longer time-trend analysis, rates were calculated by using mid-year population estimates derived by Statistics New Zealand (20).

To examine the stability of the notification system for enteric diseases during the period of interest, we compared rates for campylobacteriosis notification and hospitalization with rates for 3 other notifiable enteric diseases (salmonellosis, yersiniosis, and cryptosporidiosis). Ethical approval for this study was obtained from the Multi-Region Ethics Committee, Wellington, New Zealand.

During March 2005–December 2008, C. jejuni isolates from human case-patients and environmental and food sources were collected in the Manawatu area and genotyped (sequence-typed) by using MLST (12,16). Food samples were collected from fresh meat (poultry, beef, lamb) in retail stores, and environmental water samples were collected from swimming locations in rivers. Sheep and cattle feces were sampled from farms adjacent to the catchments of these rivers.

Two models were used to apportion human cases to sources on the basis of sequence types: the modified Hald model and the Island model (12,15). The modified Hald model combines the prevalence of each C. jejuni sequence type among the sources with the observed number of human isolates of that type by using a Bayesian framework (15). This model includes source-specific and type-specific factors, and accounts for variation in the estimated prevalence. The source-specific factor gives a measure of the ability of a source to act as a vehicle for human infection, whereas the type-specific factor yields a measure of the ability of a particular sequence type to cause disease.

The Island model uses an evolutionary model to assign sequence types to a particular source “island” or population (12). Mutation, recombination, and migration rates for isolates within and between each island are estimated by using the source isolates, and the posterior distribution of these estimates are then used to infer the origin of human isolates (12,13). To account for variations in food-processing practices that may affect the likelihood of human infection from each food source, we further extended both models to examine whether changes had occurred over time in the relative contribution of different sources to human campylobacteriosis (dynamic modeling) (21).

Key informants (n = 12), including industry and food safety experts, were interviewed to obtain information on interventions implemented to reduce Campylobacter spp. contamination in poultry. We used information from these interviews together with a review of policy documents from NZFSA and the poultry industry to formulate a summary of the interventions implemented from 2006 through 2008.

The time-trend analysis of annual notification and hospitalization rates demonstrates a steady rise and then a marked decline in the incidence of campylobacteriosis (Figure 1). In the detailed descriptive analysis covering 1997–2008, the 2008 annual rate for campylobacteriosis notifications was 161.5/100,000 population, representing a 54% decline compared with the average annual rate of 353.8 for 2002–2006 (Technical Appendix). The 2008 campylobacteriosis hospitalization rate of 7.6/100,000 population represented a 56% decline compared with the average annual rate for 2002–2006 of 17.3/100,000 population (Figure 1).

Statistically significant declines in notifications were evident across all analyzed population subgroups, although the magnitude of the declines varied (Technical Appendix). Similarly, significant decreases were seen for most subgroups for campylobacteriosis hospitalizations (results not shown).

For the 2002–2006 period (before the decline), a trend for lower notification rates was shown among those residing in more rural areas compared with those living in main urban areas (rate ratios [RR] <1 where the reference is “main urban areas”) (Figure 2). In contrast, significantly higher notification rates were observed among those residing in more rural areas compared to those living in main urban areas in 2008 (Figure 2; Technical Appendix), indicating greater declines in incidence occurred in urban areas than in rural areas during 2007–2008.

The largest declines in campylobacteriosis notification rates between the average annual rate for 2002–2006 and the 2008 rate were seen in winter months (RR 0.38, 95% confidence interval [CI] 0.36–0.40), in urban populations (RR 0.42, 95% CI 0.41–0.43), in the age groups 20–29 years and 30–39 years (RR 0.40, 95% CIs 0.38–0.43 and 0.37–0.43, respectively) and in the Asian ethnic group (RR 0.26, 95% CI 0.22–0.31) (Technical Appendix).

Conversely, the smallest declines in notification rates comparing the 2008 rate with the average annual rate for 2002–2006 were seen in rural populations (RR 0.66, 95% CI 0.62–0.70), in the 0–4 and the >80 years-of-age groups (RR 0.63, 95% CI 0.59–0.67, and RR 0.61, 95% CI 0.53–0.70 respectively) and in Māori, the indigenous people of New Zealand (RR 0.49, 95% CI 0.44–0.55) (Technical Appendix).

Figure 3 shows the temporal relationship between campylobacteriosis notification rates for 1997–2008 and 3 other notifiable enteric diseases. The marked decline in campylobacteriosis notifications during 2007–2008 is evident, while over this same period, salmonellosis, cryptosporidiosis, and yersiniosis rates remained relatively stable

During the study period 2005–2008, 572 human C. jejuni isolates and 811 food and environmental isolates were collected (and had complete MLST profiles available). The estimated number of cases attributable to each source over time (based on the dynamic modified Hald model) is shown in Figure 4. These data show that compared with the baseline period (2005–2006), the number of cases in the Manawatu region attributed to poultry declined by 74% (95% credible interval 49%–94%) in 2008. No evidence was found for a decline in cases attributed to nonpoultry sources over the same period (p>0.5) (Figure 4). Similar results were obtained for the dynamic version of the Island model (results not shown).

Specific food safety and poultry industry interventions were implemented beginning in 2006, in line with NZFSA’s strategy for reducing the incidence of foodborne campylobacteriosis (Table). From April 2007, poultry processors monitored and reported to the NZFSA-administered National Microbiological Database Campylobacter spp. prevalence in poultry flocks by using presence/absence cecal testing and Campylobacter spp. contamination levels in poultry carcass rinsates at the end of primary processing (Table).

In April 2008, mandatory Campylobacter spp. performance targets were introduced based on enumerated levels of Campylobacter spp. contamination on poultry carcasses at the end of primary processing, with escalating regulatory responses if targets were not met (22). NZFSA has subsequently released an updated Campylobacter Risk Management Strategy (23).

Key informants noted that attention to detail with hygienic practices throughout production and primary processing and alterations to the immersion-chiller conditions were key areas in which improvements were made. Furthermore, the monitoring of Campylobacter spp. contamination levels in poultry carcass rinsates at the end of primary processing and setting mandatory Campylobacter spp. performance targets (rather than mandating specific interventions) were viewed by both industry and regulator informants as key facilitators of the strategy’s success.