Hostname: page-component-78c5997874-m6dg7 Total loading time: 0 Render date: 2024-11-05T04:20:59.750Z Has data issue: false hasContentIssue false

Severe weather events and cryptosporidiosis in Aotearoa New Zealand: A case series of space–time clusters

Published online by Cambridge University Press:  15 April 2024

Leah Grout*
Affiliation:
Department of Public Health, University of Otago, Wellington, New Zealand
Simon Hales
Affiliation:
Department of Public Health, University of Otago, Wellington, New Zealand
Michael G. Baker
Affiliation:
Department of Public Health, University of Otago, Wellington, New Zealand
Nigel French
Affiliation:
Tāwharau Ora, School of Veterinary Science, Massey University, Palmerston North, New Zealand
Nick Wilson
Affiliation:
Department of Public Health, University of Otago, Wellington, New Zealand
*
Corresponding author: Leah Grout; Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Occurrence of cryptosporidiosis has been associated with weather conditions in many settings internationally. We explored statistical clusters of human cryptosporidiosis and their relationship with severe weather events in New Zealand (NZ). Notified cases of cryptosporidiosis from 1997 to 2015 were obtained from the national surveillance system. Retrospective space–time permutation was used to identify statistical clusters. Cluster data were compared to severe weather events in a national database. SaTScan analysis detected 38 statistically significant cryptosporidiosis clusters. Around a third (34.2%, 13/38) of these clusters showed temporal and spatial alignment with severe weather events. Of these, nearly half (46.2%, 6/13) occurred in the spring. Only five (38%, 5/13) of these clusters corresponded to a previously reported cryptosporidiosis outbreak. This study provides additional evidence that severe weather events may contribute to the development of some cryptosporidiosis clusters. Further research on this association is needed as rainfall intensity is projected to rise in NZ due to climate change. The findings also provide further arguments for upgrading the quality of drinking water sources to minimize contamination with pathogens from runoff from livestock agriculture.

Type
Original Paper
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2024. Published by Cambridge University Press

Introduction

Cryptosporidiosis incidence has been associated with weather and climatic conditions [Reference Ikiroma and Pollock1, Reference Lal, Fearnley and Wilford2]. In Aotearoa New Zealand (NZ), cryptosporidiosis has been associated with rainfall [Reference Lal3Reference Lal5] and temperature [Reference Lal3, Reference Britton4, Reference Lake, Pearce and Savill6]. Heavy rainfall events can increase surface runoff of cryptosporidium oocysts in the environment [Reference Davies7Reference Lal10], leading to increased cryptosporidium pathogen loads in waterways [Reference Lake11Reference Young, Smith and Fazil13]. Cryptosporidium spp. are resistant to conventional water treatment techniques [Reference Coffey, Bergin and Cummins14], and increased pathogen loading due to heavy rainfall events can overwhelm potable water and wastewater infrastructure and lead to disease outbreaks [Reference Lal15]. Cryptosporidiosis cases are often linked to contaminated water [Reference McDaniel16Reference Castro-Hermida18], and Cryptosporidium spp. are one of the most reported causal agents for waterborne outbreaks of enteric disease in NZ [19]. Rainfall-associated runoff could exacerbate the risk related to the consumption of untreated drinking water, and higher cryptosporidiosis rates have been reported in areas with untreated or inadequate drinking water supplies in NZ [Reference Duncanson20]. Recreational water contact has also been identified as an important risk factor for infection [Reference Lake, Pearce and Savill6].

Space–time cluster detection methods have previously been used to investigate the clustering of a number of infectious diseases including listeriosis in the USA [Reference Sauders21], shigellosis in the USA [Reference Jones22], and cryptosporidiosis in NZ [Reference Lal and Hales23]. Specifically, the latter NZ study examined recurrent clusters of sporadic cryptosporidiosis cases for three time periods: 1997–2001, 2001–2004, and 2005–2008 using SaTScan software to identify locations and time periods of increased risk [Reference Lal and Hales23]. The study found that many of the statistically significant cryptosporidiosis clusters were in areas with high livestock land use and occurred during the spring [Reference Lal and Hales23]. The detection of space–time clusters of cryptosporidiosis may help to identify temporal and spatial risk factors, which could in turn inform disease prevention and control efforts. Therefore, this analysis aimed to (i) detect spatiotemporal foci in the incidence of cryptosporidiosis in NZ across a longer period (1995–2015), (ii) compare the timing and location of detected clusters to severe weather events recorded in the National Institute of Water and Atmospheric Research’s (NIWA) Historic Weather Event Catalogue, and (iii) compare detected clusters to reported outbreaks.

Methods

Data collection

All notified cases of cryptosporidiosis from 1997 to 2015 in NZ were obtained from the National Notifiable Disease Surveillance System, which is maintained by the Institute of Environmental Science and Research (ESR). The report date for cryptosporidiosis notifications was used for this study. Both sporadic cases, and cases identified as being part of known outbreaks, were included.

Census area unit (CAU; n = 1976) boundaries for the 2006 National Census and population data were obtained from Statistics NZ (the national official statistics agency). CAUs are non-administrative geographic units that vary in size, each with a population ranging from around 3,000 to 5,000 people. CAU centroids were calculated using ArcGIS Desktop v10.5.1 [24].

Daily weather data, including total precipitation (m) and maximum temperature (K), were obtained from the ERA-Interim database at the European Centre for Medium-Range Forecasts [Reference Berrisford25]. The weather estimates were originally provided at 0.25 degrees resolution but were summarized at the 2006 CAU level.

Space–time cluster detection

Cryptosporidiosis clusters from 1997 to 2015 in NZ were identified using the Kulldorff method of retrospective space–time permutation [Reference Kulldorff26] in SaTScan™ [Reference Kulldorff27]. The analysis was carried out at the CAU level with aggregated, unadjusted count data. This method automatically adjusts for purely spatial patterns that are constant over time, as well as purely temporal trends, such as seasonal patterns of disease. The space–time scan statistic is defined by a cylindrical window with a circular geographic base and with height corresponding to time. As the cylindrical window moves across the study area, it scans the specified time period and highlights potential spatiotemporal clusters of disease cases for each location. The size of the cylindrical window was limited to a radius of 25 km and a maximum of 30 days in order to identify smaller clusters (i.e. at least three cases) that may indicate localized exposure.

Monte Carlo replication was used to explore the statistical significance of detected clusters. A p-value was obtained by ranking the likelihood of an observed cluster in a data set over the maximum likelihoods acquired by generating 999 randomly produced data sets. The null hypothesis of ‘no cluster’ was rejected when the simulated p-value was less than 0.05.

SaTScan required the geographic coordinates for the CAU centroids corresponding to each case, but population data were not required when using a space–time permutation model. SaTScan provides HyperText Markup Language (HTML) files for temporal graphs and maps of clusters, but the output from SaTScan was exported and statistically significant clusters were mapped in ArcGIS.

Comparison to severe weather events

NIWA is the lead Crown Research Institute responsible for collecting data on climate and weather hazards in NZ. NIWA’s Historic Weather Events Catalogue (https://hwe.niwa.co.nz/) was searched for events with evidence of increased rainfall that occurred within the 21 days preceding each significant cryptosporidiosis cluster. A severe weather event can mobilize cryptosporidium oocysts in the environment through storm water runoff, and flooding can increase the contamination of surface waters during and after heavy rainfall [Reference Lal, Fearnley and Wilford2, Reference Young, Smith and Fazil13]. Cryptosporidium oocysts can retain viability and infectivity for more than 12 weeks in the environment [Reference Olson28], and while cryptosporidium oocyst numbers are typically reduced over time in water, they remain infective for the longest time in colder temperatures and can retain viability and infectivity after freezing [Reference Olson28Reference Masina30]. In addition, the average incubation period for cryptosporidiosis has been reported to be around seven days [Reference Gupta and Haas31Reference Horne34], with a minimum incubation period of around one to two days [Reference Gupta and Haas31Reference Mac Kenzie33]. Therefore, to account in part for high concentrations of cryptosporidium oocysts in surface waters following severe weather events that could retain viability and infectivity for longer periods, as well as the incubation period for cryptosporidiosis in humans, a period of 21 days (i.e. three times the average incubation period of seven days) was used to assess temporal alignment between severe weather events and significant clusters. Severe weather events also had to be reported to affect the region in which the cluster was detected to be considered spatially aligned. Alignment was also checked by exploring temporal patterns of weather data for the first listed CAU within each cluster (see ‘Comparison to daily total precipitation and maximum temperature’).

Comparison to daily total precipitation and maximum temperature

For the significant cryptosporidiosis clusters that aligned with severe weather events, the temporal patterns of daily total precipitation and maximum temperature were plotted prior to and for the duration of the detected clusters. Specifically, the weather variables were plotted for the identifying CAU for each relevant cluster (i.e. if there were multiple CAUs within the cluster boundary, weather data were only plotted for the first CAU listed by SaTScan).

Comparison to previously recorded outbreaks

For the statistically significant cryptosporidiosis clusters that aligned with severe weather events, if a case, or cases, within one of the detected clusters had been assigned an outbreak number, then information about the outbreak was requested from ESR for comparison to the cluster. Specifically, data on the timing, location, and number of cases were compared to the detected clusters and information on potential causes and modes of transmission was also examined. ESR defines the start date of an outbreak as the date of the onset of illness in the first reported case.

Results

Space–time cluster detection

A total of 15,822 cryptosporidiosis cases were notified in NZ from 1997 to 2015. SaTScan analysis detected 65 cryptosporidiosis clusters during the study period. A majority of those clusters (58.5%, 38/65) were statistically significant (p < 0.05). The statistically significant clusters were less likely to have occurred by chance than the clusters that were not statistically significant. In total, 645 (4.1%) notified cryptosporidiosis cases were identified by SaTScan as being part of significant clusters (Table 1). The number of cases for significant clusters ranged from 3 to 83 (mean = 17.0, median = 10).

Table 1. Statistically significant cryptosporidiosis space–time clusters (p < 0.05) identified by SaTScan in New Zealand, 1997–2015

Comparison to severe weather events

Around a third (34.2%, 13/38) of the 38 statistically significant cryptosporidiosis clusters showed temporal and spatial alignment with severe weather events from NIWA’s Catalogue (Table 2). There were approximately 236 severe weather events listed in the Catalogue for the study period (1997–2015), but only 13 of the 236 (5.5%) severe weather events showed alignment with significant cryptosporidiosis clusters. Several other clusters also showed temporal and spatial alignment with severe weather events but were excluded from further analysis because the weather events did not include heavy rainfall or because the cluster had a known cause unrelated to weather. The number of cases in the 13 clusters that aligned with severe weather events ranged from 3 to 55 (mean = 12.2, median = 10). Nearly half (46.2%, 6/13) of the 13 clusters that aligned with severe weather events occurred in the spring, three occurred in the winter, three occurred in the autumn, and one occurred in the late summer and early autumn (Table 2).

Table 2. Temporal alignment of statistically significant SaTScan-detected cryptosporidiosis clusters with severe weather events from NIWA’s Historic Weather Event Catalogue

Note: Clusters not listed in this table either (i) did not have temporal alignment with a weather event, (ii) did not have spatial alignment with a weather event, (iii) had a known cause unrelated to weather (e.g. direct exposure to infected calves at a university led to an outbreak among veterinary students), or (iv) there was no evidence of increased rainfall in NIWA’s Catalogue or in the total precipitation data for CAUs for the associated weather event(s).

a Dates are presented in yyyy/mm/dd format.

b Names of events were copied directly from NIWA’s Catalogue, but do not necessarily reflect the full extent of impacts.

c Associated with known outbreak (see Table 3 for more details).

Statistically significant clusters that align with severe weather events are indicated in Figure 1. The median time between the start of severe weather and the start of a significant cluster was 14 days (range = 3, 19).

Figure 1. Space–time clusters of cryptosporidiosis in NZ (1997–2015) at the CAU level as identified by spatial scan statistic in SaTScan that align temporally and spatially with severe weather events. The orange regions (with red dots at the CAU centroids) are CAUs with statistically significant space–time clusters (p < 0.05).

While NIWA Catalogue entries for five of the 16 severe weather events that occurred before the detected cryptosporidiosis clusters mentioned ‘sewage’ or ‘sewerage’ issues and one mentioned a boil water notice, none of the reported issues overlapped with the locations of the clusters. However, the NIWA Catalogue does not systematically record such issues and relies on media reports for such details of storm impacts.

Comparison with daily total precipitation and maximum temperature

Temporal patterns of daily total precipitation and maximum temperature prior to and during clusters were explored for the first CAU listed for each cluster. The mean highest total daily precipitation in the 21 days prior to the clusters was 24.92 mm (range = 6.06, 58.79). The mean lowest daily maximum temperature in the 21 days prior to the clusters was 283.16 K [10.01 °C] (range = 277.61, 288.01), while the mean highest daily maximum temperature in the 21 days prior to clusters was 291.01 K [17.86 °C] (range = 286.05, 295.81).

Cluster 2

There was a small increase in precipitation (circled in red, Supplementary Figure 1) in mid-March 1999 that aligned with the ‘March 1999 Kaikoura Rain’ severe weather event, which occurred approximately a week and a half before the start date of the cryptosporidiosis cluster.

Cluster 6

There was a slight increase in precipitation (circled in red, Supplementary Figure 2) in early October 2000 that aligned with the ‘October 2000 Manawatu-Whanganui and Wellington Flooding’ event, which directly preceded a large peak in total precipitation on 11 October (also circled in red) that aligned with the ‘October 2000 New Zealand Weather Bomb’ that occurred approximately two weeks before the cluster start date.

Cluster 11

There was a small increase in total precipitation on 12 March that aligned with the ‘March 2010 Wellington and Marlborough Storm’ that affected the top of the South Island six days before the start of the cluster (circled in red, Supplementary Figure 3). However, it should be noted that the cluster included six different CAUs and the total precipitation estimate is only for one of the CAUs. Therefore, it is possible that rainfall totals were higher in other areas.

Cluster 13

A period of precipitation (circled in red, Supplementary Figure 4) aligned with the ‘June 2013 New Zealand Storm’ that impacted the country ten days prior to the start of the detected cluster.

Cluster 14

A period of precipitation three days before the start of the cluster, circled in red in Supplementary Figure 5, aligned with the ‘October 1998 New Zealand High Winds and Flooding’ event.

Cluster 16

A small peak in total precipitation (circled in red, Supplementary Figure 6) aligned with the ‘September 2005 North Island Winds and South Island Snow’ event from 16 to 19 September. While the event primarily reported snow at higher elevations, it likely produced substantial rainfall at lower elevations as well.

Cluster 17

The large peak in total precipitation on 1 June 1997 (circled in red, Figure 2) aligned with the ‘June 1997 Northland Flooding and Hawke’s Bay Flooding’ event, which occurred around two weeks prior to the start of the cluster.

Figure 2. Patterns of daily total precipitation and maximum temperature in identifying CAU in the three weeks prior and for the duration of Cluster 17 (see Table 2).

Cluster 27

A small peak in total precipitation in late July and early August 2007 (circled in red, Supplementary Figure 7) aligned with the ‘July–August 2007 New Zealand Storm’ event, which occurred just over two weeks before the cluster start date. The storm event reportedly brought unsettled weather to the North Island.

Cluster 30

The large peak in total precipitation (circled in red, Figure 3) aligned with the ‘October 2005 Gisborne and Hawke’s Bay Flooding’ event, which brought heavy rain to the Bay of Plenty region.

Figure 3. Patterns of daily total precipitation and maximum temperature in identifying CAU in the three weeks prior and for the duration of Cluster 30 (see Table 2).

Cluster 31

A peak in daily total precipitation (circled in red, Supplementary Figure 8) aligned with the ‘November 2008 Wellington Wind’ event on 1 November 2008, which brought heavy rain to the West Coast, including Buller. Another later peak in total precipitation aligned with the ‘November 2008 South Island Cold Snap’ event from 4 to 7 November 2008, which brought both heavy rain and snow to the West Coast. Additional peaks in daily total precipitation were evident, but they did not coincide with any of the severe weather events included in NIWA’s Catalogue.

Cluster 33

A small peak in total precipitation in February (circled in red, Supplementary Figure 9) had a degree of alignment with the ‘February 2008 Upper North Island Storm’ event, which reportedly impacted the North Island from 22 to 23 February 2008. A large peak in precipitation in March also had a degree of alignment with the ‘March 2008 New Zealand Storm’ event, which impacted the country at the beginning of the month, in the week before the cluster start date.

Cluster 36

A large peak in total precipitation (circled in red, Supplementary Figure 10) aligned with the ‘April 2001 North Island Ex-Tropical Cyclone Sose’ event, which impacted many parts of the North Island from 12 to 14 April 2001. The storm arrived 12 days prior to the start of the detected cluster.

Cluster 38

A peak in total precipitation (circled in red, Supplementary Figure 11) aligned closely with the ‘October 2003 New Zealand Storm’ event, which brought heavy rain to much of the North Island from 3 to 5 October 2003. Another peak in daily total precipitation was observed approximately one week later, but the peak did not coincide with any of the severe weather events included in NIWA’s Catalogue.

Comparison to previously recorded outbreaks

Five of the 13 statistically significant cryptosporidiosis clusters that showed alignment with severe weather events had at least one case with an assigned outbreak number (Table 3; see Supplementary Text for additional details).

Table 3. Statistically detected space–time clusters of cryptosporidiosis and temporally and spatially overlapping recorded outbreaks of cryptosporidiosis in New Zealand, 1997–2015

a Dates are presented in yyyy/mm/dd format.

b Total cases include laboratory-confirmed, clinically confirmed, and probable cases.

c Outbreak end date was not reported, but the outbreak report date fell within the cluster date range.

d Outbreak end date did not fall within the cluster date range, but the report date for the outbreak fell within the cluster date range.

When the outbreaks were compared to the corresponding severe weather events, it became apparent that for four of the outbreaks (Table 3), the start date preceded the severe weather event start date (Table 2), while only one outbreak began after the severe weather event (outbreak number RO2003011; Table 3). However, severe weather events may lead to the further spread of an existing outbreak.

The other eight clusters did not match any cases that were associated with known outbreaks (Clusters 11, 14, 16, 17, 27, 30, 33, and 36; Table 3), which implies that statistically significant cryptosporidiosis outbreaks were undetected by the surveillance system during the period of this study.

Discussion

Main findings and interpretation

SaTScan analysis detected 38 statistically significant cryptosporidiosis space–time clusters. Approximately 4% of notified cryptosporidiosis cases from 1997 to 2015 were part of these clusters. By comparison, 9% of notified cases for the same period were associated with an outbreak that had been identified by routine surveillance. These findings suggest that most cryptosporidiosis cases in NZ are sporadic (i.e. cases with no statistically significant spatial and temporal relationship with other confirmed cases) and not associated with clusters (i.e. as operationally defined in this study as three or more confirmed cases with a spatial and temporal relationship). However, the space–time scan statistic was limited to a radius of 25 km in this study to indicate localized exposure. Therefore, it is possible that regional increases in cases (e.g. cases spread beyond 25 km) might not be considered a single cluster. For example, heavy rainfall could lead to a regional increase in cryptosporidiosis cases that are not linked to a single contaminated water supply, but rather to the contamination of multiple small water sources (e.g. bore wells) in the region.

Around a third (34.2%, 13/38) of the statistically significant cryptosporidiosis clusters showed temporal and spatial alignment with severe weather events from NIWA’s Catalogue. Of these, nearly half (46.2%, 6/13) occurred in the spring. Clusters detected during the spring months may be due to the transmission of the zoonotic Cryptosporidium parvum (C. parvum), rather than Cryptosporidium hominis (C. hominis), as previous research has indicated that there are seasonal strain-specific transmission cycles, with zoonotic transmission in the spring and anthroponotic transmission in the autumn [Reference Lal and Hales23, Reference Garcia-R35]. Spring peaks in cryptosporidiosis in rural areas may be associated with spring calving and lambing because newborn livestock can be an important source of this infection [Reference Snel36Reference Grinberg38]. Spring clusters of cryptosporidiosis in rural areas of NZ may also be associated with rainfall [Reference Britton4]. For example, a study in England showed that peaks in spring rainfall preceded peaks in cryptosporidiosis notifications [Reference Naumova39] and local rainfall has been reported as a determinant of cryptosporidiosis infection in children globally [Reference Lal, Fearnley and Wilford2]. Studies have also shown that heavy rainfall events can significantly increase surface runoff of cryptosporidium oocysts over land used for livestock agriculture [Reference Davies7Reference Lal10]. Furthermore, heavy rainfall events have been associated with increased cryptosporidium pathogen loads in waterways and increased cryptosporidiosis incidence [Reference Lake11Reference Young, Smith and Fazil13].

Only one of the statistically significant cryptosporidiosis clusters showing temporal and spatial alignment with severe weather events occurred in the summer (Cluster 33, Table 2). This suggests that recreational contact with contaminated waterways following heavy rainfall may be a less important transmission pathway than the consumption of contaminated drinking water in the development of cryptosporidiosis outbreaks. However, a clear pattern did not emerge when the daily maximum temperature was plotted for the 21 days prior to each cluster. Diffuse clusters or outbreaks may also involve other transmission routes. Person-to-person transmission may be particularly important with a number of cases linked to attendance at childcare centres or the changing of dirty diapers in NZ [Reference Snel36].

While 13 of the 38 (34.2%) significant cryptosporidiosis clusters showed alignment with severe weather events, it is important to note that only 13 of the 236 (5.5%) severe weather events showed alignment with significant cryptosporidiosis clusters. This finding indicates that only a small proportion of severe weather events may contribute to identifiable outbreaks of cryptosporidiosis. However, it would be beneficial to compare severe weather events to waterborne outbreaks of other gastrointestinal diseases (e.g. campylobacteriosis, giardiasis) to better ascertain the overall impact of such events on enteric disease outbreaks in NZ. Furthermore, the analysis of more recent disease data, especially following recent major weather events, would also be insightful.

When the daily total precipitation and maximum temperature were plotted for the identifying CAU for each cluster, it was clear that all the corresponding severe weather events caused increases in precipitation. However, for several clusters (Clusters 11, 14, and 16), the increase in precipitation in the identifying CAU was less than 10 mm per day. Cryptosporidiosis has been positively associated with rainfall in NZ [Reference Britton4, Reference Lal5], but evidence for the relationship between cryptosporidiosis and temperature has been mixed [Reference Britton4-Reference Lake, Pearce and Savill6]. Temperature, precipitation, and humidity affect the life cycle of Cryptosporidium spp. directly, as well as indirectly through ecological changes [Reference Ikiroma and Pollock1, Reference Patz40].

Waterborne or environmental transmission was reported for five outbreaks identified by routine surveillance (corresponding with Clusters 2, 6, 13, 31, and 38; Table 3). The start date for four of these (corresponding with Clusters 2, 6, 13, and 31; Table 3) preceded the start of the corresponding severe weather event. This timing indicates that the corresponding severe weather events (Table 2) cannot have initiated those outbreaks (or the clusters that they closely align with). However, it is possible that severe weather contributed to the spread of infection via faecal–oral transmission.

Strengths of this study

This is one of the first studies to detect space–time clusters of cryptosporidiosis in NZ. Lal et al. (2015) reported recurrent clusters of sporadic cryptosporidiosis cases for three time periods in NZ: 1997–2001, 2001–2004, and 2005–2008, while our study examined clusters over a longer time period (1997–2015). This is also the first study (to our knowledge) to compare detected clusters of cryptosporidiosis to severe weather events in NZ. Using SaTScan statistics in tandem with geographic information systems (GIS) techniques is useful for visualizing spatiotemporal patterns of disease risk [Reference Lal and Hales23, Reference Lian41]. While some of the clusters detected in this study may have represented previously identified outbreaks, the space–time clusters highlight potential spatial and temporal variation in the drivers of disease.

Limitations of this study

An important limitation of SaTScan is that when it is used over longer time periods, clusters may be detected because of underlying population changes in certain areas. The space–time scan statistic assumes that any change in the population at risk occurs uniformly across the study population [Reference Pfeiffer42], but demographic shifts were not considered in this study. Future studies may wish to consider the use of a discrete Poisson model, which would account for the underlying population at risk, or limit cluster detection to a single year at a time to prevent the risk of population shift bias. However, the benefit of using the space–time scan statistic in this study was that the model automatically adjusted for both purely spatial and purely temporal clusters (e.g. it accounted for seasonal patterns of disease).

Another potential limitation is the a priori choice of cluster size, but there are no clear guidelines for addressing this issue [Reference Pfeiffer42]. To address this potential limitation, we tested the space–time model using several different cluster sizes (e.g. varying maximum spatial radii and varying maximum temporal periods) before settling on the final parameters (i.e. 25 km maximum spatial radius and a 30-day temporal maximum). Additionally, it is possible that the use of a circular cluster window may result in the analysis missing non-circular clusters (e.g. those along a river) [Reference Pfeiffer42], although some studies have indicated that a circular cluster window can still successfully detect non-circular clusters of disease [Reference Kulldorff43].

There are also limitations associated with the use of routine surveillance data to assess disease distribution, given the low-case ascertainment of enteric infections [Reference Lal and Hales23, Reference Lian41] and errors in the geographic allocation of some that are identified [Reference Skelly44]. In NZ, disease notifications report a case’s location based on the person’s home address. However, a person’s home location may not be an accurate reflection of the location of the source of the infection or the location of the exposure due to commuting or travel patterns to work or recreational settings. However, this error is unlikely to impact the detection of space–time clusters because they must exhibit both spatial variation and temporal variation [Reference Kulldorff26]. It is also possible that the use of the onset date instead of the report date for notified cases could have improved the detection of disease clusters, as the onset date should be closer to the actual time of infection. However, onset dates were not available for all notified cases; therefore, the report dates for notifications were used in this study.

Another limitation was a change in the inclusion criteria for NIWA’s Catalogue in the 2010s [Reference Grout, Hales, Thornly and Wilson45]. This change may have led to the exclusion of some severe weather events that previously would have been included in the Catalogue prior to this decade. Additionally, the Catalogue generally excludes high rainfall events that have not been defined as storms or floods, but which may still have human health consequences. For example, a heavy rainfall event is thought to have led to the very large 2016 Havelock North campylobacteriosis outbreak [Reference Gilpin46], but this weather event was not recorded in NIWA’s Catalogue. Details on severe weather events in NIWA’s Catalogue are also limited. The Catalogue relies on newspaper and other media articles, which may selectively report examples of impacts. Therefore, the Catalogue may not capture impacts across all regions affected by the event. Furthermore, daily total precipitation and maximum temperature were only examined for the identifying CAU (i.e. the first CAU listed by SaTScan) for each cluster, but several detected clusters included multiple CAUs. Therefore, the plots may not be representative of conditions across the whole cluster area.

Research and policy implications

This study found that heavy rainfall and flooding events preceded several detected cryptosporidiosis clusters in NZ from 1997 to 2015. However, additional research is required to determine the contribution of extreme weather events to cryptosporidiosis incidence and similarly for other enteric diseases (e.g. campylobacteriosis, giardiasis) in NZ. There is a pressing need for such research as rainfall intensity is projected to keep rising in NZ due to climate change [Reference Lawrence, Hoegh-Guldberg and Wratt47, Reference Fischer and Knutti48]. Similar research should also be conducted in other countries or regions with a high level of reliance on non-reticulated water supplies to assess the likelihood of enteric disease outbreaks associated with severe weather events.

The results suggest that a number of potential cryptosporidiosis outbreaks may have gone undetected in NZ. The epidemiological significance of SaTScan-detected clusters that have no evidence of a relationship to a known outbreak also merits further investigation. Enhancement of the disease surveillance system using SaTScan or a similar cluster detection tool, in combination with field investigations and typing, could help better identify and determine the cause of outbreaks. As noted above, clusters detected during the autumn months may be due to the transmission of C. hominis rather than C. parvum, because of seasonal strain-specific transmission cycles. Strain-specific data were not available for this study. The typing of all cases and outbreaks could help refine the disease surveillance system, as C. hominis and C. parvum have different drivers and risk factors.

This study supports the need to protect drinking water supplies from microbial contamination, especially from runoff from land used for livestock agriculture. Notifications and outbreaks of cryptosporidiosis, as well as other enteric diseases, have been linked to contaminated water supplies [Reference Coffey, Bergin and Cummins14Reference McDaniel16]. For example, the 2016 waterborne Havelock North campylobacteriosis outbreak was attributed to the contamination of two untreated bore wells with sheep faeces and resulted in approximately 7,570 cases and four deaths [Reference Gilpin46].

Such outbreaks highlight the need to strengthen the regulation and management of drinking water supplies to protect public health. Around 4,077,000 individuals in NZ (82.9% of the population) are served by public drinking water supply systems, but approximately 868,000 of those individuals (23.1%) receive drinking water that does not fully meet protozoal standards [49]. Furthermore, approximately 840,000 people in NZ (17.1% of the population) rely on non-reticulated water supplies [49], which may be particularly vulnerable to contamination. Action in this area is highly consistent with the principles of drinking water safety for NZ articulated by the Government Inquiry into Havelock North Drinking Water [50]. These principles include the paramount importance of protecting source water; the need for multiple barriers against contamination; that change (such as heavy rainfall or flooding) precedes contamination; and the need to apply a preventive risk management approach. This inquiry initiated the largest overhaul of drinking water supply in NZ’s history and the creation of a new national water regulator, Taumata Arowai. It will be important for this regulator to consider these principles and the findings of this current study as part of its mission to protect and improve drinking water quality.

Conclusions

This study provides additional evidence that extreme weather events can play a role in the transmission of cryptosporidiosis in NZ. Many of the clusters that aligned temporally and spatially with severe weather events occurred in the spring and may be due to the transmission of the zoonotic C. parvum. Additional research is needed to assess the contribution of heavy rainfall and other environmental factors, such as livestock farming, to cryptosporidiosis cases and outbreaks, especially as rainfall intensity is projected to increase due to climate change.

List of abbreviations

CAU

census area unit

ESR

Institute of Environmental Science and Research

NIWA

National Institute of Water and Atmospheric Research

NZ

New Zealand

Supplementary material

The supplementary material for this article can be found at http://doi.org/10.1017/S095026882400058X.

Data availability statement

The disease notification data are available from ESR, but restrictions apply to the availability of these data, so they are not publicly available. Other data sets generated and analysed during this study are available from the corresponding author upon reasonable request.

Acknowledgements

The authors are very grateful to NIWA for developing and maintaining the Catalogue of the events used in part of this analysis.

Author contribution

All authors contributed to the conceptualization of the study. L.G. led the analysis and visualization of data, and all authors contributed to the interpretation of results. L.G. led the analysis, and all other authors contributed to the drafting and revision of the manuscript.

Funding statement

The authors thank the Health Research Council of New Zealand (Grant 19/104) for funding support. However, the funder had no role in the decision to perform this particular work, its content, or the decision to publish.

Competing interest

The authors declare that they have no competing interests.

Ethics approval and consent to participate

Ethics approval for the use of de-identified disease notification data was granted by the University of Otago Human Ethics Committee (HD17/019).

References

Ikiroma, IA and Pollock, KG (2021) Influence of weather and climate on cryptosporidiosis—A review. Zoonoses and Public Health 68, 285298.CrossRefGoogle ScholarPubMed
Lal, A, Fearnley, E and Wilford, E (2019) Local weather, flooding history and childhood diarrhoea caused by the parasite cryptosporidium spp.: A systematic review and meta-analysis. Science of the Total Environment 674, 300306.CrossRefGoogle ScholarPubMed
Lal, A, et al. (2013) Climate variability, weather and enteric disease incidence in New Zealand: Time series analysis. PLoS One 8, e83484.CrossRefGoogle ScholarPubMed
Britton, E, et al. (2010) The impact of climate variability and change on cryptosporidiosis and giardiasis rates in New Zealand. Journal of Water and Health 8, 561.CrossRefGoogle ScholarPubMed
Lal, A, et al. (2015) Environmental change and enteric zoonoses in New Zealand: A systematic review of the evidence. Australian and New Zealand Journal of Public Health 39, 6368.CrossRefGoogle ScholarPubMed
Lake, IR, Pearce, J and Savill, M (2008) The seasonality of human cryptosporidiosis in New Zealand. Epidemiology and Infection 136, 13831387.CrossRefGoogle ScholarPubMed
Davies, CM, et al. (2004) Dispersion and transport of cryptosporidium oocysts from fecal pats under simulated rainfall events. Applied and Environmental Microbiology 70, 1151.CrossRefGoogle ScholarPubMed
Davies-Colley, R, Nagels, J and Lydiard, E (2008) Stormflow-dominated loads of faecal pollution from an intensively dairy-farmed catchment. Water Science & Technology 57, 15191523.CrossRefGoogle ScholarPubMed
Tryland, I, et al. (2011) Impact of rainfall on microbial contamination of surface water. International Journal of Climate Change Strategies and Management 3, 361373.CrossRefGoogle Scholar
Lal, A, et al. (2013) Potential effects of global environmental changes on cryptosporidiosis and giardiasis transmission. Trends in Parasitology 29, 8390.CrossRefGoogle ScholarPubMed
Lake, IR, et al. (2005) Effects of weather and river flow on cryptosporidiosis. Journal of Water and Health 3, 469474.CrossRefGoogle ScholarPubMed
Curriero, F, et al. (2001) The association between extreme precipitation and waterborne disease outbreaks in the United States, 1948–1994. American Journal of Public Health 91, 11941199.CrossRefGoogle ScholarPubMed
Young, I, Smith, BA and Fazil, A (2015) A systematic review and meta-analysis of the effects of extreme weather events and other weather-related variables on cryptosporidium and giardia in fresh surface waters. Journal of Water and Health 13, 117.CrossRefGoogle ScholarPubMed
Coffey, R, Bergin, D and Cummins, E (2010) Use of meta-analysis to assess the effect of conventional water treatment methods on the prevalence of cryptosporidium spp. in drinking water. Human and Ecological Risk Assessment 16, 13601378.CrossRefGoogle Scholar
Lal, A (2014) Evaluating the environmental and social determinants of enteric disease in New Zealand. Doctoral Thesis, Department of Public Health, University of Otago, Wellington. Available at https://ourarchive.otago.ac.nz/handle/10523/4813.Google Scholar
McDaniel, CJ, et al. (2014) Humans and cattle: A review of bovine zoonoses. Vector-Borne and Zoonotic Diseases 14, 119.CrossRefGoogle ScholarPubMed
Cavirani, S (2008) Cattle industry and zoonotic risk. Veterinary Research Communications 32, S19S24.CrossRefGoogle ScholarPubMed
Castro-Hermida, JA, et al. (2009) Detection of cryptosporidium spp. and giardia duodenalis in surface water: A health risk for humans and animals. Water Research 43, 41334142.CrossRefGoogle Scholar
ESR (2018) Annual summary of outbreaks in New Zealand 2016 [Client Report FW17038]. Institute of Environmental Science and Research Ltd. (ESR). Available at https://surv.esr.cri.nz/surveillance/annual_outbreak.php?we_objectID=4741.Google Scholar
Duncanson, M, et al. (2003) Cryptosporidiosis: Drinking water quality and other risk factors in Aotearoa New Zealand. Epidemiology 14, S129S130.CrossRefGoogle Scholar
Sauders, BD, et al. (2003) Molecular subtyping to detect human listeriosis clusters. Emerging Infectious Diseases 9, 672.CrossRefGoogle ScholarPubMed
Jones, RC, et al. (2006) Use of a prospective space-time scan statistic to prioritize shigellosis case investigations in an urban jurisdiction. Public Health Reports 121, 133139.CrossRefGoogle Scholar
Lal, A and Hales, S (2015) Heterogeneity in hotspots: Spatio-temporal patterns in neglected parasitic diseases. Epidemiology and Infection 143, 631639.CrossRefGoogle ScholarPubMed
ESRI (2018) ArcGIS Desktop 10.5.1. Redlands, CA: Environmental Systems Research Institute.Google Scholar
Berrisford, P, et al. (2011) The ERA-Interim archive Version 2.0. ECMWF, Shinfield Park, Reading. Available at https://www.ecmwf.int/node/8174.Google Scholar
Kulldorff, M, et al. (2005) A space-time permutation scan statistic for disease outbreak detection. PLoS Medicine 2, e59.CrossRefGoogle ScholarPubMed
Kulldorff, M (2005) SaTScan 9.6. Boston, Massachusetts.Google Scholar
Olson, ME, et al. (1999) Giardia cyst and cryptosporidium oocyst survival in water, soil, and cattle feces. Journal of Environmental Quality 28, 19911996.CrossRefGoogle Scholar
Fayer, R and Nerad, T (1996) Effects of low temperatures on viability of Cryptosporidium parvum oocysts. Applied and Environmental Microbiology 62, 14311433.CrossRefGoogle ScholarPubMed
Masina, S, et al. (2019) Weather, environmental conditions, and waterborne giardia and cryptosporidium in Iqaluit, Nunavut. Journal of Water and Health 17, 8497.CrossRefGoogle ScholarPubMed
Gupta, M and Haas, CN (2004) The Milwaukee cryptosporidium outbreak: Assessment of incubation time and daily attack rate. Journal of Water and Health 2, 5969.CrossRefGoogle ScholarPubMed
Jokipii, L and Jokipii, AMM (1986) Timing of symptoms and oocyst excretion in human cryptosporidiosis. New England Journal of Medicine 315, 16431647.CrossRefGoogle ScholarPubMed
Mac Kenzie, WR, et al. (1995) Massive outbreak of waterborne cryptosporidium infection in Milwaukee, Wisconsin: Recurrence of illness and risk of secondary transmission. Clinical Infectious Diseases 21, 5762.CrossRefGoogle ScholarPubMed
Horne, S, et al. (2017) Cryptosporidium outbreaks: Identification, diagnosis, and management. British Journal of General Practice 67, 425426.CrossRefGoogle ScholarPubMed
Garcia-R, JC, et al. (2020) Species and genotypes causing human cryptosporidiosis in New Zealand. Journal of Parasitology Research 119, 23172326.CrossRefGoogle ScholarPubMed
Snel, SJ, et al. (2009) A tale of two parasites: The comparative epidemiology of cryptosporidiosis and giardiasis. Epidemiology and Infection 137, 16411650.CrossRefGoogle ScholarPubMed
Learmonth, J, et al. (2001) Seasonal shift in Cryptosporidium parvum transmission cycles in New Zealand. Journal Eukaryotic Microbiology 48, 34S35S.CrossRefGoogle Scholar
Grinberg, A, et al. (2005) The occurrence of Cryptosporidium parvum, campylobacter and salmonella in newborn dairy calves in the Manawatu region of New Zealand. New Zealand Veterinary Journal 53, 315320.CrossRefGoogle ScholarPubMed
Naumova, EN, et al. (2005) Effect of precipitation on seasonal variability in cryptosporidiosis recorded by the north West England surveillance system in 1990–1999. Journal of Water and Health 3, 185196.CrossRefGoogle ScholarPubMed
Patz, JA, et al. (2000) Effects of environmental change on emerging parasitic diseases. International Journal of Parasitology 30, 13951405.CrossRefGoogle ScholarPubMed
Lian, M, et al. (2007) Using geographic information systems and spatial and space-time scan statistics for a population-based risk analysis of the 2002 equine West Nile epidemic in six contiguous regions of Texas. International Journal of Health Geographics 6, 42.CrossRefGoogle ScholarPubMed
Pfeiffer, D, et al. (2008) Spatial Analysis in Epidemiology. New York, USA: Oxford University Press.CrossRefGoogle Scholar
Kulldorff, M, et al. (2004) Benchmark data and power calculations for evaluating disease outbreak detection methods. Morbidity and Mortality Weekly Report (MMWR) 53, 144151.Google ScholarPubMed
Skelly, C, et al. (2002) Disease surveillance in rural communities is compromised by address geocoding uncertainty: A case study of campylobacteriosis. Australian Journal of Rural Health 10, 8793.CrossRefGoogle ScholarPubMed
Grout, L, Hales, S, Thornly, L, and Wilson, N (2022) An exploration of the human health impacts associated with seven decades of severe weather events in Aotearoa New Zealand: The need for better data. Report, University of Otago Wellington, Wellington. Available at https://www.otago.ac.nz/wellington/departments/publichealth/research/heiru/otago841638.pdf.Google Scholar
Gilpin, BJ, et al. (2020) A large scale waterborne campylobacteriosis outbreak, Havelock North, New Zealand. Journal of Infection 81, 390395.CrossRefGoogle ScholarPubMed
Lawrence, J, et al. (2021) Chapter 11: Australasia. In Hoegh-Guldberg, O and Wratt, D (eds.), IPCC WGII Sixth Assessment Report. Geneva: Intergovernmental Panel on Climate Change (IPCC).Google Scholar
Fischer, EM and Knutti, R (2015) Anthropogenic contribution to global occurrence of heavy-precipitation and high-temperature extremes. Nature Climate Change 5, 560564.CrossRefGoogle Scholar
Ministry of Health (2020) Annual Report on Drinking-water Quality: 2018–2019. Wellington. Available at https://www.health.govt.nz/publication/annual-report-drinking-water-quality-2018-2019.Google Scholar
Government Inquiry into Havelock North Drinking Water (2017) Report of the Havelock North Drinking Water Inquiry: Stage 2. Auckland, December. Available at https://www.dia.govt.nz/Report-of-the-Havelock-North-Drinking-Water-Inquiry---Stage-2.Google Scholar
Figure 0

Table 1. Statistically significant cryptosporidiosis space–time clusters (p < 0.05) identified by SaTScan in New Zealand, 1997–2015

Figure 1

Table 2. Temporal alignment of statistically significant SaTScan-detected cryptosporidiosis clusters with severe weather events from NIWA’s Historic Weather Event Catalogue

Figure 2

Figure 1. Space–time clusters of cryptosporidiosis in NZ (1997–2015) at the CAU level as identified by spatial scan statistic in SaTScan that align temporally and spatially with severe weather events. The orange regions (with red dots at the CAU centroids) are CAUs with statistically significant space–time clusters (p < 0.05).

Figure 3

Figure 2. Patterns of daily total precipitation and maximum temperature in identifying CAU in the three weeks prior and for the duration of Cluster 17 (see Table 2).

Figure 4

Figure 3. Patterns of daily total precipitation and maximum temperature in identifying CAU in the three weeks prior and for the duration of Cluster 30 (see Table 2).

Figure 5

Table 3. Statistically detected space–time clusters of cryptosporidiosis and temporally and spatially overlapping recorded outbreaks of cryptosporidiosis in New Zealand, 1997–2015

Supplementary material: File

Grout et al. supplementary material

Grout et al. supplementary material
Download Grout et al. supplementary material(File)
File 1.1 MB