- Email: [email protected]
Spatial clusters of Clostridium difficile infection and an association with neighbourhood socio-economic disadvantage in the Australian Capital Territory, 2004–2014
+
MODEL
Infection, Disease & Health xxx (xxxx) xxx
Available online at www.sciencedirect.com
ScienceDirect journal homepage: http://www.journals.elsevier.com/infectiondisease-and-health/
Research paper
Spatial clusters of Clostridium difficile infection and an association with neighbourhood socio-economic disadvantage in the Australian Capital Territory, 2004e2014 Aparna Lal a,*, Ashwin Swaminathan b, Teisa Holani c a
Research School of Population Health, Australian National University, Acton, Australia General Medicine & Infectious Diseases Physician, Canberra Hospital, Garran, Australia c ANU Medical School, Acton, Australia b
Received 9 June 2019; received in revised form 18 August 2019; accepted 19 August 2019
KEYWORDS Geography; Health; Risk; Infection; Spatial; Maps
Abstract Background: In Australia, rates of Clostridium difficile infection (CDI) in all States and Territories have increased significantly since mid-2011, with rates of infection increasing faster in the community setting than within hospitals. Knowledge about the risk factors for CDI is essential to determine the risk of community outbreaks of CDI and to design interventions that reduce those risks. Methods: We examine the role of neighbourhood socio-economic disadvantage, demography and testing practices on spatial patterns in CDI incidence in the Australian Capital Territory (ACT). Data on all tests conducted for CDI, including postcode of residence, were obtained from January 2004eDecember 2014. Distribution of age groups and the neighbourhood Index of Relative Socio-economic Advantage Disadvantage (IRSAD) were obtained from the Australian Bureau of Statistics 2011 National Census data. A Bayesian spatial conditional autoregressive model was fitted at the postcode level to quantify the relationship between CDI and sociodemographic factors. To identify CDI hotspots, exceedance probabilities were set at a threshold of twice the estimated relative risk. Results: After controlling for spatial patterns in testing practices, area-level socio-economic advantage (IRSAD) (RR Z 0.74, 95% CI 0.57, 0.94) was inversely associated with CDI. Three postcodes had a high probability (0.8e1.0) of excess risk of diagnosed CDI. Conclusion: We demonstrate geographic variations in CDI in the ACT with a positive association of CDI with neighbourhood socioeconomic disadvantage and identify areas with a high probability of elevated risk compared with surrounding communities. These findings provide further evidence to inform a targeted response to reduce CDI risk.
* Corresponding author. National Centre for Epidemiology and Population Health, Australian National University, Acton, Australia. E-mail address: [email protected] (A. Lal). https://doi.org/10.1016/j.idh.2019.08.002 2468-0451/ª 2019 Australasian College for Infection Prevention and Control. Published by Elsevier B.V. All rights reserved.
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
2
A. Lal et al. ª 2019 Australasian College for Infection Prevention and Control. Published by Elsevier B.V. All rights reserved.
Highlights In the ACT, the proportion of positive CDI infections have increased between 2010 and 2014, compared with 2004. Neighbourhood socio-economic disadvantage was positively associated with infection. An excess risk of infection was observed in three postcodes, after adjusting for area-level factors. Such knowledge can inform the development of targeted strategies to reduce CDI in Australian communities.
Introduction Clostridium difficile is a leading hospital-acquired gastrointestinal infection globally [1], with a rise in infection rates in recent decades [2,3], with a number of strains in circulating in Australia [4]. While Clostridium difficile infection (CDI) can be asymptomatic, it is responsible for 10e25% of antibiotic-associated diarrhoea and majority of pseudomembranous colitis [5]. The infection can result in life-threatening conditions including intestinal perforation, toxic megacolon, sepsis and shock [6]. CDI is an under-diagnosed infection in the nonhospitalized community [2,3,7,8] with increasing rates of infection in populations previously not considered at-risk [9]. These groups include children [10,11], peri-partum women [12] and previously healthy persons unexposed to the hospital environment of antimicrobial therapy [9]. Recent studies have shown that approximately one-third of all CDIs in North America [2], Europe [3] and Australia [7,13] have been acquired outside hospital settings. The established view that CDI is largely an infection in healthcare settings is further challenged by the strong seasonal patterns in CDI observed worldwide [14] and the increasing number of studies that report CDI in animal reservoirs [15e17], soil [18], food and water sources [19e21]. These findings suggest that in addition to clinical risk factors such as antibiotic and proton pump inhibitor (PPI) [23] use and exposure to infected individuals, the social and demographic context is important for Clostridium difficile transmission [22]. There has been limited research into the area-level risk factors for CDI in the Southern Hemisphere with one 12month study in the United States showing proximity to a livestock farm and nursing home as being important risk factors while high rates of infection [24] and another in New Mexico showing an association with lack of health insurance and low educational attainment [25]. A recent spatial and temporal study in Queensland showed a summer peak in CDI and a positive association of CDI risk with rainfall [26]. This study did not take into account neighbourhood socio-economic deprivation or the underlying demography of the area, well-known risk factors for the spread of several infectious diseases [27]. In Western
Australia, public spaces such as lawns have high prevalence of toxigenic Clostridium difficile spores, indicating the potential for area-level factors to be important for transmission [28]. In addition, most studies that use laboratory surveillance or hospital data to describe spatial patterns in disease incidence [24,29] do not have access to the testing data. Therefore, spatial patterns and areas with increased risk may just represent community testing practices with hotspots relating to increased testing in those areas. We examine how geographic variations in CDI incidence are associated with neighbourhood socio-economic characteristics and population structure, after controlling for the spatial pattern in testing practices. The present study focuses on the Australian Capital Territory (ACT). Using data on CDI from 2004 to 2014, we develop a Bayesian conditional autoregressive model spatial model to assess whether neighbourhood socioeconomic disadvantage and population structure is associated with CDI risk, after controlling for the number of tests for CDI in each area and identify localized areas with excess risk.
Methods Study area The ACT is a land-locked territory that occupies an area of 2358 square kilometres located in the southeast of Australia (35 330 S, 149 180 E). Ethical approval for the study was granted by the Australian National University Human Research Ethics Committee (HREC) (2014/725) and ACT Health HREC (ETHLR.14.276).
Clostridium difficile infection Results for all tests for Clostridium difficile between January 1, 2004 and December 31, 2014 were acquired from ACT Pathology, which has five community and two public hospital-based collection centres throughout the ACT (at the time of the study) and with two 24-hour laboratories at Canberra Hospital (the main laboratory) and Calvary Hospital, Bruce. C. difficile tests sent to private pathology
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
Spatial patterns of clostridium difficile providers (including those servicing private hospitals) were not acquired. C. difficile testing at ACT Pathology has evolved over the period of the study. Prior to 2011, C. difficile was identified by testing for the C. difficile antigen, glutamate dehydrogenase (GDH), and toxin A through immuno-chromatography. From 2011, polymerase chain reaction (PCR) assays to detect the conserved C difficile gene target tcdB was added as an additional confirmatory test where there was conflicting GDH/toxin results. For our study, a C. difficile detected result indicates that the stool sample was positive for GDH and Toxin A and/or B or PCR positivity. For each patient tested for CDI, we had the date and time of specimen collection, the hospital, ward (if applicable), the setting (in-patient, emergency, outpatient, day unit), and the patient’s date of birth and postcode of residence. We analysed incident episodes of CDI only, disregarding repeat results from the same CDI episode or recurrent CDI. Of the 1256 patients that tested positive for C.difficile in the ACT from 2004 to 2014, 24 cases (1.5%) were excluded as they were missing a location of where the sample was taken. A further two cases were dropped due to missing age at diagnosis data. A total of 1232 positive CDI cases were used in the subsequent spatial analysis.
Geographic, socio-economic, and demographic data Geographical State and Territory and postal area boundaries were obtained from the Australian Bureau of Statistics (ABS) (www.abs.gov.au). The postal area boundaries produced by the ABS differ between each five-year census due to changes in population, postcode distribution areas and ABS methodology in defining postcode area boundaries. The ABS has not produced correspondence files allowing postcodes from 2001 to 2006 to be matched to 2011 boundaries. Therefore in this paper, data were mapped using the 2011 postcode boundaries. To determine area-based SES, we matched the postcode of residence for each patient to the corresponding ABS Socio-Economic Indexes for Areas (SEIFA) value from the 2011 Australian Census. We applied the score of Relative Socioeconomic Advantage and Disadvantage (IRSAD) for this analysis, in which higher scores represent the most advantaged populations. The IRSAD scores were modelled as a continuous variable. As one postal area can have more than IRSAD score, the category that applied to majority of the area was assigned to each postal area. Population data including age structure were based on the Estimated Resident Population (ERP) produced for the 2011 Census. Population data was used in three ways in the spatial analysis. Firstly, the total population in each postcode was used as an offset (i.e. cases proportional to the population at risk in each postcode). Secondly, as CDI is frequently reported in the elderly, the number of people over 65 years residing in each postcode was used as a covariate. Thirdly, the effect of population density was assessed by dividing the total population number in each postcode by the area in each postcode (obtained by calculating the area of 2011 postcode polygons in ArcGIS v10.3.1). Population density and the IRSAD scores were standardised to have mean 0 and standard deviation 1.
3
Statistical analysis Descriptive To describe the trend in CDI across the ACT from 2004 to 2014, the annual proportion of positive CDIs was calculated using the total number of tests performed in that year as the denominator. The average annual incidence risk was calculated by dividing the total number of C. difficile positive cases over the 11-year period (2004e2014) with an estimate of the population at risk by postcode in 2011 and converted to rates per 100,000 population. Visualization of rates by postcode was carried out in ArcGIS v10.3.1. As area-level rates can become unstable owing to small populations and data sparseness, we examined the geographic variation in the proportion of positive CDI as a function of the number of tests performed in each postcode. To visualise the geographic variation in the proportion of positive CDIs by postcode, we calculated the proportion positive by dividing the total number of positive CDIs over the study period by the total number of tests over the study period in each postcode. Spatial model We estimated postcode level risk of C. difficile using a Bayesian spatial model, constructed using the open source version of WinBUGS version 1.4.3 [30]. The number of positive CDIs in each postcode was specified using a Poisson model. The Poisson model estimates the risk of C. difficile in each postcode, relative to that expected if all areas had the same risk as the risk for the whole of the ACT. In the spatial model, the estimated risk in each postcode is influenced by its own data as well as by those of its neighbours (defined as those postcodes that share a boundary) by borrowing information. Moran’s I is a measure of spatial autocorrelation and we used this measure to assess how related the number of cases of CDI were based on the locations where they were measured. The individual data were aggregated by postcode with the observed number of cases (ys) for the sth Postcode (s Z 1 .. 27) following a Poisson distribution with mean ms . ys wPoissonðms Þ logðms Þ Z logðes Þ þ qs where es (expected number of cases in postcode s) is an offset to control for population size. The mean log relative risk (RR), qs for each covariate was modelled as qs Z a þ ðtestsÞb1 þ ðIRSADÞb2 þ ðpopdensityÞb3 þ ðover65Þb4 þ 4s þ vs where a is the intercept, b1 is the coefficient for the number of tests, b2 is the coefficient for the standardized IRSAD score of relative socio-economic advantage disadvantage, b3 is the coefficient for standardized population density, b4 is the coefficient for the total population over 65 years in postcode s, 4s is a spatially structured random effect with mean 0 and variance s2v , and vs is a spatially unstructured random effect with mean 0 and variance s2v . The spatially structured component 4s was modelled using a conditional autoregressive (CAR) prior structure, with spatial associations between postcodes modelled using an
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
4
A. Lal et al.
adjacency weights matrix [31]. Included in the model specification was a step function to calculate the exceedance probabilities for every postcode. Exceedance probabilities have been proposed as a Bayesian approach to hotspot identification. The exceedance probabilities for an area are the posterior probabilities that the spatial risk is greater than twice the estimated risk. Probabilities greater than 0.8 are regarded as strong evidence of excess risk. Model selection was based on several diagnostics, including trace and density plots for each of the estimated parameters and the Deviance Information Criterion (DIC). Bayesian model implementation Bayesian models use Monte Carlo Markov Chain (MCMC) simulation to combine the prior distribution with the data, leading to the posterior likelihood. Once all plausible values of the posterior distribution are thought to have been sampled, the model is said to have converged. We assessed convergence using GelmaneRubin plots [32]. We ran the model for 100,000 iterations after convergence on two independent MCMC chains (200,000 total) to provide mean parameter estimates and corresponding 95% credible intervals (95% CI) for all parameters in our models, including RR in each postcode. Maps were constructed in ArcGIS v.10.3.1.
Results The percentage of positive CDIs increased over the study period, 2004e2014, with an annual average of 3.9% (57/ 1439) in 2004 to 7.2% (277/3840) in 2014 (Fig. 1). C. difficile was identified in 1232 (5.7%) of the stool samples. The average annual rate of CDI was 38/100,000 population in the ACT over the 11 years of the study period. Since 2010, the proportion of positive CDIs has remained significantly higher than 2004 (Fig. 1) (z score 10.27, p < 0.001). Fig. 2 confirms the geographic variation in the average annual incidence risk of positive CDIs at the postcode level, with no obvious spatial clustering of patients with CDI by postcode of residence (Fig. 2A). An area with high rates of CDIs was observed in the southeast of the ACT. To examine whether these spatial patterns was influenced by small populations in some postcodes, the proportion of positive CDIs was also mapped at the postcode level (Fig. 2B). This
Figure 1
Proportion of positive CDIs in the ACT, 2004e2014.
map showed a similar spatial pattern, with the same area in the southeast with a high proportion of proportion of C. difficile positives highlighted. Moran’s I was significant at the alpha Z 0.05 level (p < 0.004). Based on these results, we rejected the null hypothesis that there is zero spatial autocorrelation present in CDI patterns. Table 1 shows the posterior estimates of relative risk (RR) and the corresponding 95% CI. The number of tests was showed a marginal positive association with CDI (RR Z 1.01, 95% CI 1.00, 1.02). The index of relative socioeconomic advantage disadvantage (IRSAD) was negatively associated with CDIs (RR Z 0.74, 95% CI 0.56, 0.94). Fig. 3A indicates the relative risk of CDIs estimated for each postcode (compared to the whole of the ACT) after neighbourhood socio-economic deprivation, the number of tests, population density and the population over 65 years in each area are taken into account. The high-risk areas (RR 1.5) are concentrated in the central parts of the ACT with an additional high-risk postcode in the west. The cluster with the highest risk of infection occurs in the southeast (RR 2.5e3.5). Fig. 3B shows the probability exceedance estimates for each postcode. Fig. 3B brings into starker relief the clustering from the map of relative risks (Fig. 3A). The values represent the probability of a postcode having a more than twice the estimated risk of C. difficile infection. Here, we can see a cluster of three postcodes in the ACT with a high probability (0.8e1.0) of exceeding twice the risk of infection estimated for that postcode.
Discussion This study found localised areas of excess Clostridium difficile infection risk compared with surrounding communities and a positive association with neighbourhood socioeconomic disadvantage, using laboratory data from the Australian Capital Territory over an 11 year observation period. Since 2010, the proportion of positive Clostridium infections in the Australian Capital Territory has remained consistently higher compared with 2004. This rise is in accordance with the literature, reflecting the increase in CDIs that has been observed over the past 15 years worldwide [2,3] with an inflexion point occurring around the turn of the decade. A stable increase in incidence of CDI since has been noted in other Australian jurisdictions [33]. A strength of this study is that spatial differences in the frequency of diagnostic testing were controlled for and the data were collected over 11 years from a stable population. Increased testing has been suggested as a potential reason for the regional increases in CDI across Australia since mid2011 [7]. Accounting for spatial patterns in testing practices over the study period provides support for the existence of localised infection clusters that do not simply reflect increased testing and are indicative of potential locality specific demographic and socio-economic factors important for transmission of CDI independent of hospital-associated drivers. Furthermore, by using the Bayesian exceedance probability to identify areas of high disease risk, we overcome a limitation of frequently used cluster detection methods like SatScan [34]), where the spatial autocorrelation structure of the exposures and health outcomes (a
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
Spatial patterns of clostridium difficile
5
Figure 2 Spatial patterns of CDIs by Postcode in the ACT, 2004e2014 (A) Average annual incidence per 100,000 population (B) Proportion positive as a function of total number of tests done.
Table 1
Regression coefficients from Bayesian spatial CAR model of CDIs in the Australian Capital Territory.
Variable
Posterior mean SD
Monte Carlo error
RR (95% CI)
Intercept Number of tests Socio-economic advantage Population density Population >65 years Heterogeneity Structured Unstructured
0.156 0.200 0.0006 0.0003 0.299 0.130 0.163 0.130 0.001 0.001
<0.01 <0.01 <0.01 <0.01 <0.01
1.01 0.74 1.17 1.00
0.455 0.162 0.190 0.294
<0.01 <0.01
particular feature of infectious disease data) are not taken into account [35]. Using Bayesian exceedance probability, our study has identified areas where the probability of the relative risk of CDI exceeds a specified threshold, after accounting for the spatial dependence in exposures and outcomes, thus more likely to represent clustering of cases that are not an artefact of spatial sampling bias. In the present study all cases of CDI were considered. Such an analysis does not allow the identification of risk factors specific to community or hospital acquired infection. A previous study in Queensland did not find an association of community-acquired CDI with population-level medication exposure in Queensland during 2008e2012 [36]. A related issue is the emergence of new strains [37e39]. It is likely that spatial patterns of CDI will be strain specific, which was beyond the scope of the present study. As our study is the first examination of spatial patterns of CDI
(1.00e1.02) (0.57e0.94) (0.91e1.53) (1.00e1.01)
across the ACT, consideration of distinguishing community and hospital acquired infection as well as strain-specific patterns should be the focus of future work. Neighbourhood socio-economic status is associated with a person’s general health quality, with several infections having a higher rate of incidence amongst persons of lower socio-economic status [40,41]. Neighbourhood socioeconomic status may be an indicator for household crowding, poorer house conditions and poorer hygiene that can contribute to the increased likelihood of exposure to infectious agents [42e44]. Of particular relevance for CDI, studies in Scotland and New Zealand have found that increasing socio-economic deprivation and household crowding is also consistently associated with increased rates of antibiotic prescribing [34,45]. Thus, our finding of a positive spatial association of CDI with socio-economic disadvantage in the ACT is likely to be a proxy for factors
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
6
A. Lal et al.
Figure 3 (A) Estimated Relative risks of CDI by 2011 Postcodes in the Australian Capital Territory (B) exceedance probabilities for CDI, with the exceedance probability threshold of twice the estimated risk.
including health care access, antibiotic prescribing practices and living conditions, which if focussed on, can help to identify the mechanisms that underlie the association showed here. Our findings raise interesting questions for future research which include: are public health messages be specifically tailored for high-risk communities (e.g. those residing in socio-economically deprived areas)? The present study did not find a relationship of CDI with the proportion of elderly that usually resided in a postcode. Advanced age is a significant risk factor for CDI development [46], mortality and recurrence [47]. Comorbidities, increased antibiotic exposure, decreased host immunity, age-related alterations to intestinal flora and increased healthcare facility exposure [48] are likely to contribute to increased risk of CDI in the elderly. Proton pump inhibitor (PPI) use has been demonstrated to be linked to an increased risk in CDIs [23], and its use in Australia is on the rise [49]. PPI use has been shown to be a significant risk factor for CDI in nursing home patients [50]. However, as this is a population-level analysis, no inferences can be made with respect to individual-level risk factors such as PPI use which continue to be an important risk factor for CDI. A potential reason for the increased proportion of positive CDI tests is the change in testing practices from ELISAbased test to a more sensitive PCR-based test [35]. Although the impact of increased sensitivity in testing as a result of the addition of PCR testing in 2011 cannot be excluded as a possible reason for the increased incidence, this change is unlikely to bias spatial patterns as collection centres for ACT Pathology are evenly distributed throughout the region and our dataset includes the only two emergency departments in the ACT. However, we cannot
exclude the possibility of false positives in our dataset, as patients may be colonized with C. difficile but not have symptomatic infection [2]. A recent study in Australian tertiary hospitals found different risk factors for asymptomatic and symptomatic patients colonised with Clostridium difficile [51].
Conclusions In the ACT, the proportion of positive Clostridium difficile infections have increased between 2010 and 2014, compared with 2004. The positive spatial association with neighbourhood socio-economic disadvantage, and localized areas with excess risk compared with surrounding communities are relevant for public health messaging and infection control in the ACT, as we have determined certain areas that have a higher risk of CDI. This knowledge will contribute to the development of targeted strategies to reduce CDI in Australian communities and reduce the costs on the Australian healthcare system.
Ethics Ethical approval for the study was granted by the Australian National University Human Research Ethics Committee (HREC) (2014/725) and ACT Health HREC (ETHLR.14.276).
Authorship statement AL conceived the study design and led the analysis. AS provided data from Canberra hospital and provided
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
Spatial patterns of clostridium difficile extensive input into all drafts of the manuscript. TH led the writing and re-drafting of the manuscript.
Conflicts of interest The authors declare no conflict of interest.
Funding There was no external funding required for this project. A Lal’s salaried position at the ANU enabled her involvement.
Provenance and peer review Externally peer reviewed.
Acknowledgements We thank all staff involved in collecting and processing of laboratory specimens.
References [1] Bartlett JG. Narrative review: the new epidemic of Clostridium difficile-associated enteric disease. Ann Intern Med 2006 Nov 21;145(10):758e64. [2] Khanna S, Pardi DS, Aronson SL, Kammer PP, Orenstein R, St Sauver JL, et al. The epidemiology of community-acquired Clostridium difficile infection: a population-based study. Am J Gastroenterol 2012 Jan;107(1):89e95. [3] Viseur N, Lambert M, Delmee M, Van Broeck J, Catry B. Nosocomial and non-nosocomial Clostridium difficile infections in hospitalised patients in Belgium: compulsory surveillance data from 2008 to 2010. Euro Surveill 2011;16(43). [4] Cheng AC, Collins DA, Elliott B, Ferguson JK, Paterson DL, Thean S, et al. Laboratory-based surveillance of Clostridium difficile circulating in Australia, September e November 2010. Pathology 2016;48(3):257e60. [5] Kelly CP, LaMont JT. Clostridium difficileemore difficult than ever. N Engl J Med 2008 Oct 30;359(18):1932e40. [6] Rupnik M, Wilcox MH, Gerding DN. Clostridium difficile infection: new developments in epidemiology and pathogenesis. Nat Rev Microbiol 2009 Jul;7(7):526e36. [7] Slimings C, Armstrong P, Beckingham WD, Bull AL, Hall L, Kennedy KJ, et al. Increasing incidence of Clostridium difficile infection, Australia, 2011e2012. Med J Aust 2014 Mar 17; 200(5):272e6. [8] Kuntz JL, Johnson ES, Raebel MA, Petrik AF, Yang X, Thorp ML, et al. Epidemiology and healthcare costs of incident Clostridium difficile infections identified in the outpatient healthcare setting. Infect Control Hosp Epidemiol 2012 Oct; 33(10):1031e8. [9] Centers for Disease C, Prevention. Severe Clostridium difficile-associated disease in populations previously at low riskefour states, 2005. MMWR Morb Mortal Wkly Rep 2005 Dec 2;54(47):1201e5. [10] Borali E, Ortisi G, Moretti C, Stacul EF, Lipreri R, Gesu GP, et al. Community-acquired Clostridium difficile infection in children: a retrospective study. Dig Liver Dis 2015 Jun 9; 47(10):842e6. [11] Schutze GE, Willoughby RE, Committee on Infectious D, American Academy of P. Clostridium difficile infection in infants and children. Pediatrics 2013 Jan;131(1):196e200.
7 [12] Garey KW, Jiang ZD, Yadav Y, Mullins B, Wong K, Dupont HL. Peripartum Clostridium difficile infection: case series and review of the literature. Am J Obstet Gynecol 2008 Oct; 199(4):332e7. [13] Wozniak TM, Rubin G, MacIntyre CR. The emergence of community-acquired Clostridium difficile in an Australian hospital. Healthc Infect 2015;20(2):72e7. [14] Furuya-Kanamori L, McKenzie SJ, Yakob L, Clark J, Paterson DL, Riley TV, et al. Clostridium difficile infection seasonality: patterns across hemispheres and continents e a systematic review. PLoS One 2015;10(3):e0120730. [15] Hensgens MP, Keessen EC, Squire MM, Riley TV, Koene MG, de Boer E, et al. Clostridium difficile infection in the community: a zoonotic disease? Clin Microbiol Infect 2012 Jul;18(7):635e45. [16] Himsworth CG, Patrick DM, Mak S, Jardine CM, Tang P, Weese JS. Carriage of Clostridium difficile by wild urban Norway rats (Rattus norvegicus) and black rats (Rattus rattus). Appl Environ Microbiol 2014 Feb;80(4):1299e305. [17] Bandelj P, Trilar T, Blagus R, Ocepek M, Rousseau J, Weese JS, et al. Prevalence and molecular characterization of Clostridium difficile isolated from European Barn Swallows (Hirundo rustica) during migration. BMC Vet Res 2014; 10:40. [18] Bauer MP, Kuijper EJ. Potential sources of Clostridium difficile in human infection. Infect Dis Clin N Am 2015 Mar;29(1):29e35. [19] Bakri MM, Brown DJ, Butcher JP, Sutherland AD. Clostridium difficile in ready-to-eat salads, Scotland. Emerg Infect Dis 2009 May;15(5):817e8. [20] Gould LH, Limbago B. Clostridium difficile in food and domestic animals: a new foodborne pathogen? Clin Infect Dis 2010 Sep 1;51(5):577e82. [21] Songer JG, Trinh HT, Killgore GE, Thompson AD, McDonald LC, Limbago BM. Clostridium difficile in retail meat products, USA, 2007. Emerg Infect Dis 2009 May;15(5):819e21. [22] Slimings C, Riley TV. Antibiotics and hospital-acquired Clostridium difficile infection: update of systematic review and meta-analysis. J Antimicrob Chemother 2014 Apr;69(4): 881e91. [23] Kwok CS, Arthur AK, Anibueze CI, Singh S, Cavallazzi R, Loke YK. Risk of Clostridium difficile infection with acid suppressing drugs and antibiotics: meta-analysis. Am J Gastroenterol 2012 Jul;107(7):1011e9. [24] Anderson DJ, Rojas LF, Watson S, Knelson LP, Pruitt S, Lewis SS, et al. Identification of novel risk factors for community-acquired Clostridium difficile infection using spatial statistics and geographic information system analyses. PLoS One 2017;12(5):e0176285. [25] Hudspeth WB, Qeadan F, Phipps EC. Disparities in the incidence of community-acquired Clostridioides difficile infection: an area-based assessment of the role of social determinants in Bernalillo County, New Mexico. Am J Infect Contr 2019;47(7):773e9. [26] Furuya-Kanamori L, Robson J, Soares Magalhaes RJ, Yakob L, McKenzie SJ, Paterson DL, et al. A population-based spatiotemporal analysis of Clostridium difficile infection in Queensland, Australia over a 10-year period. J Infect 2014 Jun 28;69(5):447e55. [27] Simonsen J, Frisch M, Ethelberg S. Socioeconomic risk factors for bacterial gastrointestinal infections. Epidemiology 2008 Mar;19(2):282e90. [28] Moono P, Lim SC, Riley TV. High prevalence of toxigenic Clostridium difficile in public space lawns in Western Australia. Sci Rep 2017;7:41196. [29] van Dorp SM, Hensgens MP, Dekkers OM, Demeulemeester A, Buiting A, Bloembergen P, et al. Spatial clustering and livestock exposure as risk factor for community-acquired Clostridium difficile infection. Clin Microbiol Infect 2019;25(5): 607e12.
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
8 [30] Lunn DJ, Thomas A, Best N, Spiegelhalter D. WinBUGS e a Bayesian modelling framework: concepts, structure, and extensibility. Stat Comput 2000;10:325e37. [31] Lawson AB, Browne WJ, Rodeiro CLV. Disease mapping with WinBUGS and MLwiN. John Wiley & Sons; 2003. [32] Gelman A, Rubin DB. Inference from iterative simulation using multiple sequences. Stat Sci 1992:457e72. [33] Worth LJ, Spelman T, Bull AL, Brett JA, Richards MJ. Epidemiology of Clostridium difficile infections in Australia: enhanced surveillance to evaluate time trends and severity of illness in Victoria, 2010e2014. J Hosp Infect 2016 Jul;93(3): 280e5. [34] Walls G, Vandal AC, du Plessis T, Playle V, Holland DJ. Socioeconomic factors correlating with community antimicrobial prescribing. N Z Med J 2014;128(1417):16e23. [35] Khanna S, Baddour LM, Huskins WC, Kammer PP, Faubion WA, Zinsmeister AR, et al. The epidemiology of Clostridium difficile infection in children: a population-based study. Clin Infect Dis 2013 May 15;56(10):1401e6. [36] Furuya-Kanamori L, Yakob L, Riley TV, Paterson DL, Baker P, McKenzie SJ, et al. Community-acquired Clostridium difficile infection, Queensland, Australia. Emerg Infect Dis 2016;22(9): 1659. [37] Eyre DW, Tracey L, Elliott B, Slimings C, Huntington PG, Stuart RL, et al. Emergence and spread of predominantly community-onset Clostridium difficile PCR ribotype 244 infection in Australia, 2010 to 2012. Euro Surveill 2015;20(10): 21059. [38] Lim SK, Stuart RL, Mackin KE, Carter GP, Kotsanas D, Francis MJ, et al. Emergence of a ribotype 244 strain of Clostridium difficile associated with severe disease and related to the epidemic ribotype 027 strain. Clin Infect Dis 2014;58(12):1723e30. [39] De Almeida MN, Heffernan H, Dervan A, Bakker S, Freeman JT, Bhally H, et al. Severe Clostridium difficile infection in New Zealand associated with an emerging strain, PCR-ribotype 244. N Z Med J 2013;126(1380):9e14. [40] Flory JH, Joffe M, Fishman NO, Edelstein PH, Metlay JP. Socioeconomic risk factors for bacteraemic pneumococcal pneumonia in adults. Epidemiol Infect 2009 May;137(5): 717e26.
A. Lal et al. [41] Huggan PJ, Wells JE, Browne M, Richardson A, Murdoch DR, Chambers ST. Population-based epidemiology of Staphylococcus aureus bloodstream infection in Canterbury, New Zealand. Intern Med J 2010 Feb;40(2):117e25. [42] Baker MG, Barnard LT, Kvalsvig A, Verrall A, Zhang J, Keall M, et al. Increasing incidence of serious infectious diseases and inequalities in New Zealand: a national epidemiological study. Lancet 2012 Mar 24;379(9821):1112e9. [43] Baker M, Das D, Venugopal K, Howden-Chapman P. Tuberculosis associated with household crowding in a developed country. J Epidemiol Community Health 2008 Aug;62(8):715e21. [44] Huang SS, Finkelstein JA, Rifas-Shiman SL, Kleinman K, Platt R. Community-level predictors of pneumococcal carriage and resistance in young children. Am J Epidemiol 2004 Apr 1; 159(7):645e54. [45] Covvey JR, Johnson BF, Elliott V, Malcolm W, Mullen AB. An association between socioeconomic deprivation and primary care antibiotic prescribing in Scotland. J Antimicrob Chemother 2014 March 1;69(3):835e41. [46] To KB, Napolitano LM. Clostridium difficile infection: update on diagnosis, epidemiology, and treatment strategies. Surg Infect (Larchmt) 2014 Oct;15(5):490e502. [47] Abou Chakra CN, Pepin J, Sirard S, Valiquette L. Risk factors for recurrence, complications and mortality in Clostridium difficile infection: a systematic review. PLoS One 2014;9(6): e98400. [48] Mizusawa M, Doron S, Gorbach S. Clostridium difficile diarrhea in the elderly: current issues and management options. Drugs Aging 2015 Aug;32(8):639e47. [49] Hollingworth S, Duncan EL, Martin JH. Marked increase in proton pump inhibitors use in Australia. Pharmacoepidemiol Drug Saf 2010 Oct;19(10):1019e24. [50] Al-Tureihi FI, Hassoun A, Wolf-Klein G, Isenberg H. Albumin, length of stay, and proton pump inhibitors: key factors in Clostridium difficile-associated disease in nursing home patients. J Am Med Dir Assoc 2005 Mar-Apr;6(2):105e8. [51] Furuya-Kanamori L, Clements ACA, Foster NF, Huber CA, Hong S, Harris-Brown T, et al. Asymptomatic Clostridium difficile colonization in two Australian tertiary hospitals, 2012e2014: prospective, repeated cross-sectional study. Clin Microbiol Infect 2017;23(1). 48.e1-.e7.
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
MODEL
Infection, Disease & Health xxx (xxxx) xxx
Available online at www.sciencedirect.com
ScienceDirect journal homepage: http://www.journals.elsevier.com/infectiondisease-and-health/
Research paper
Spatial clusters of Clostridium difficile infection and an association with neighbourhood socio-economic disadvantage in the Australian Capital Territory, 2004e2014 Aparna Lal a,*, Ashwin Swaminathan b, Teisa Holani c a
Research School of Population Health, Australian National University, Acton, Australia General Medicine & Infectious Diseases Physician, Canberra Hospital, Garran, Australia c ANU Medical School, Acton, Australia b
Received 9 June 2019; received in revised form 18 August 2019; accepted 19 August 2019
KEYWORDS Geography; Health; Risk; Infection; Spatial; Maps
Abstract Background: In Australia, rates of Clostridium difficile infection (CDI) in all States and Territories have increased significantly since mid-2011, with rates of infection increasing faster in the community setting than within hospitals. Knowledge about the risk factors for CDI is essential to determine the risk of community outbreaks of CDI and to design interventions that reduce those risks. Methods: We examine the role of neighbourhood socio-economic disadvantage, demography and testing practices on spatial patterns in CDI incidence in the Australian Capital Territory (ACT). Data on all tests conducted for CDI, including postcode of residence, were obtained from January 2004eDecember 2014. Distribution of age groups and the neighbourhood Index of Relative Socio-economic Advantage Disadvantage (IRSAD) were obtained from the Australian Bureau of Statistics 2011 National Census data. A Bayesian spatial conditional autoregressive model was fitted at the postcode level to quantify the relationship between CDI and sociodemographic factors. To identify CDI hotspots, exceedance probabilities were set at a threshold of twice the estimated relative risk. Results: After controlling for spatial patterns in testing practices, area-level socio-economic advantage (IRSAD) (RR Z 0.74, 95% CI 0.57, 0.94) was inversely associated with CDI. Three postcodes had a high probability (0.8e1.0) of excess risk of diagnosed CDI. Conclusion: We demonstrate geographic variations in CDI in the ACT with a positive association of CDI with neighbourhood socioeconomic disadvantage and identify areas with a high probability of elevated risk compared with surrounding communities. These findings provide further evidence to inform a targeted response to reduce CDI risk.
* Corresponding author. National Centre for Epidemiology and Population Health, Australian National University, Acton, Australia. E-mail address: [email protected] (A. Lal). https://doi.org/10.1016/j.idh.2019.08.002 2468-0451/ª 2019 Australasian College for Infection Prevention and Control. Published by Elsevier B.V. All rights reserved.
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
2
A. Lal et al. ª 2019 Australasian College for Infection Prevention and Control. Published by Elsevier B.V. All rights reserved.
Highlights In the ACT, the proportion of positive CDI infections have increased between 2010 and 2014, compared with 2004. Neighbourhood socio-economic disadvantage was positively associated with infection. An excess risk of infection was observed in three postcodes, after adjusting for area-level factors. Such knowledge can inform the development of targeted strategies to reduce CDI in Australian communities.
Introduction Clostridium difficile is a leading hospital-acquired gastrointestinal infection globally [1], with a rise in infection rates in recent decades [2,3], with a number of strains in circulating in Australia [4]. While Clostridium difficile infection (CDI) can be asymptomatic, it is responsible for 10e25% of antibiotic-associated diarrhoea and majority of pseudomembranous colitis [5]. The infection can result in life-threatening conditions including intestinal perforation, toxic megacolon, sepsis and shock [6]. CDI is an under-diagnosed infection in the nonhospitalized community [2,3,7,8] with increasing rates of infection in populations previously not considered at-risk [9]. These groups include children [10,11], peri-partum women [12] and previously healthy persons unexposed to the hospital environment of antimicrobial therapy [9]. Recent studies have shown that approximately one-third of all CDIs in North America [2], Europe [3] and Australia [7,13] have been acquired outside hospital settings. The established view that CDI is largely an infection in healthcare settings is further challenged by the strong seasonal patterns in CDI observed worldwide [14] and the increasing number of studies that report CDI in animal reservoirs [15e17], soil [18], food and water sources [19e21]. These findings suggest that in addition to clinical risk factors such as antibiotic and proton pump inhibitor (PPI) [23] use and exposure to infected individuals, the social and demographic context is important for Clostridium difficile transmission [22]. There has been limited research into the area-level risk factors for CDI in the Southern Hemisphere with one 12month study in the United States showing proximity to a livestock farm and nursing home as being important risk factors while high rates of infection [24] and another in New Mexico showing an association with lack of health insurance and low educational attainment [25]. A recent spatial and temporal study in Queensland showed a summer peak in CDI and a positive association of CDI risk with rainfall [26]. This study did not take into account neighbourhood socio-economic deprivation or the underlying demography of the area, well-known risk factors for the spread of several infectious diseases [27]. In Western
Australia, public spaces such as lawns have high prevalence of toxigenic Clostridium difficile spores, indicating the potential for area-level factors to be important for transmission [28]. In addition, most studies that use laboratory surveillance or hospital data to describe spatial patterns in disease incidence [24,29] do not have access to the testing data. Therefore, spatial patterns and areas with increased risk may just represent community testing practices with hotspots relating to increased testing in those areas. We examine how geographic variations in CDI incidence are associated with neighbourhood socio-economic characteristics and population structure, after controlling for the spatial pattern in testing practices. The present study focuses on the Australian Capital Territory (ACT). Using data on CDI from 2004 to 2014, we develop a Bayesian conditional autoregressive model spatial model to assess whether neighbourhood socioeconomic disadvantage and population structure is associated with CDI risk, after controlling for the number of tests for CDI in each area and identify localized areas with excess risk.
Methods Study area The ACT is a land-locked territory that occupies an area of 2358 square kilometres located in the southeast of Australia (35 330 S, 149 180 E). Ethical approval for the study was granted by the Australian National University Human Research Ethics Committee (HREC) (2014/725) and ACT Health HREC (ETHLR.14.276).
Clostridium difficile infection Results for all tests for Clostridium difficile between January 1, 2004 and December 31, 2014 were acquired from ACT Pathology, which has five community and two public hospital-based collection centres throughout the ACT (at the time of the study) and with two 24-hour laboratories at Canberra Hospital (the main laboratory) and Calvary Hospital, Bruce. C. difficile tests sent to private pathology
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
Spatial patterns of clostridium difficile providers (including those servicing private hospitals) were not acquired. C. difficile testing at ACT Pathology has evolved over the period of the study. Prior to 2011, C. difficile was identified by testing for the C. difficile antigen, glutamate dehydrogenase (GDH), and toxin A through immuno-chromatography. From 2011, polymerase chain reaction (PCR) assays to detect the conserved C difficile gene target tcdB was added as an additional confirmatory test where there was conflicting GDH/toxin results. For our study, a C. difficile detected result indicates that the stool sample was positive for GDH and Toxin A and/or B or PCR positivity. For each patient tested for CDI, we had the date and time of specimen collection, the hospital, ward (if applicable), the setting (in-patient, emergency, outpatient, day unit), and the patient’s date of birth and postcode of residence. We analysed incident episodes of CDI only, disregarding repeat results from the same CDI episode or recurrent CDI. Of the 1256 patients that tested positive for C.difficile in the ACT from 2004 to 2014, 24 cases (1.5%) were excluded as they were missing a location of where the sample was taken. A further two cases were dropped due to missing age at diagnosis data. A total of 1232 positive CDI cases were used in the subsequent spatial analysis.
Geographic, socio-economic, and demographic data Geographical State and Territory and postal area boundaries were obtained from the Australian Bureau of Statistics (ABS) (www.abs.gov.au). The postal area boundaries produced by the ABS differ between each five-year census due to changes in population, postcode distribution areas and ABS methodology in defining postcode area boundaries. The ABS has not produced correspondence files allowing postcodes from 2001 to 2006 to be matched to 2011 boundaries. Therefore in this paper, data were mapped using the 2011 postcode boundaries. To determine area-based SES, we matched the postcode of residence for each patient to the corresponding ABS Socio-Economic Indexes for Areas (SEIFA) value from the 2011 Australian Census. We applied the score of Relative Socioeconomic Advantage and Disadvantage (IRSAD) for this analysis, in which higher scores represent the most advantaged populations. The IRSAD scores were modelled as a continuous variable. As one postal area can have more than IRSAD score, the category that applied to majority of the area was assigned to each postal area. Population data including age structure were based on the Estimated Resident Population (ERP) produced for the 2011 Census. Population data was used in three ways in the spatial analysis. Firstly, the total population in each postcode was used as an offset (i.e. cases proportional to the population at risk in each postcode). Secondly, as CDI is frequently reported in the elderly, the number of people over 65 years residing in each postcode was used as a covariate. Thirdly, the effect of population density was assessed by dividing the total population number in each postcode by the area in each postcode (obtained by calculating the area of 2011 postcode polygons in ArcGIS v10.3.1). Population density and the IRSAD scores were standardised to have mean 0 and standard deviation 1.
3
Statistical analysis Descriptive To describe the trend in CDI across the ACT from 2004 to 2014, the annual proportion of positive CDIs was calculated using the total number of tests performed in that year as the denominator. The average annual incidence risk was calculated by dividing the total number of C. difficile positive cases over the 11-year period (2004e2014) with an estimate of the population at risk by postcode in 2011 and converted to rates per 100,000 population. Visualization of rates by postcode was carried out in ArcGIS v10.3.1. As area-level rates can become unstable owing to small populations and data sparseness, we examined the geographic variation in the proportion of positive CDI as a function of the number of tests performed in each postcode. To visualise the geographic variation in the proportion of positive CDIs by postcode, we calculated the proportion positive by dividing the total number of positive CDIs over the study period by the total number of tests over the study period in each postcode. Spatial model We estimated postcode level risk of C. difficile using a Bayesian spatial model, constructed using the open source version of WinBUGS version 1.4.3 [30]. The number of positive CDIs in each postcode was specified using a Poisson model. The Poisson model estimates the risk of C. difficile in each postcode, relative to that expected if all areas had the same risk as the risk for the whole of the ACT. In the spatial model, the estimated risk in each postcode is influenced by its own data as well as by those of its neighbours (defined as those postcodes that share a boundary) by borrowing information. Moran’s I is a measure of spatial autocorrelation and we used this measure to assess how related the number of cases of CDI were based on the locations where they were measured. The individual data were aggregated by postcode with the observed number of cases (ys) for the sth Postcode (s Z 1 .. 27) following a Poisson distribution with mean ms . ys wPoissonðms Þ logðms Þ Z logðes Þ þ qs where es (expected number of cases in postcode s) is an offset to control for population size. The mean log relative risk (RR), qs for each covariate was modelled as qs Z a þ ðtestsÞb1 þ ðIRSADÞb2 þ ðpopdensityÞb3 þ ðover65Þb4 þ 4s þ vs where a is the intercept, b1 is the coefficient for the number of tests, b2 is the coefficient for the standardized IRSAD score of relative socio-economic advantage disadvantage, b3 is the coefficient for standardized population density, b4 is the coefficient for the total population over 65 years in postcode s, 4s is a spatially structured random effect with mean 0 and variance s2v , and vs is a spatially unstructured random effect with mean 0 and variance s2v . The spatially structured component 4s was modelled using a conditional autoregressive (CAR) prior structure, with spatial associations between postcodes modelled using an
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
4
A. Lal et al.
adjacency weights matrix [31]. Included in the model specification was a step function to calculate the exceedance probabilities for every postcode. Exceedance probabilities have been proposed as a Bayesian approach to hotspot identification. The exceedance probabilities for an area are the posterior probabilities that the spatial risk is greater than twice the estimated risk. Probabilities greater than 0.8 are regarded as strong evidence of excess risk. Model selection was based on several diagnostics, including trace and density plots for each of the estimated parameters and the Deviance Information Criterion (DIC). Bayesian model implementation Bayesian models use Monte Carlo Markov Chain (MCMC) simulation to combine the prior distribution with the data, leading to the posterior likelihood. Once all plausible values of the posterior distribution are thought to have been sampled, the model is said to have converged. We assessed convergence using GelmaneRubin plots [32]. We ran the model for 100,000 iterations after convergence on two independent MCMC chains (200,000 total) to provide mean parameter estimates and corresponding 95% credible intervals (95% CI) for all parameters in our models, including RR in each postcode. Maps were constructed in ArcGIS v.10.3.1.
Results The percentage of positive CDIs increased over the study period, 2004e2014, with an annual average of 3.9% (57/ 1439) in 2004 to 7.2% (277/3840) in 2014 (Fig. 1). C. difficile was identified in 1232 (5.7%) of the stool samples. The average annual rate of CDI was 38/100,000 population in the ACT over the 11 years of the study period. Since 2010, the proportion of positive CDIs has remained significantly higher than 2004 (Fig. 1) (z score 10.27, p < 0.001). Fig. 2 confirms the geographic variation in the average annual incidence risk of positive CDIs at the postcode level, with no obvious spatial clustering of patients with CDI by postcode of residence (Fig. 2A). An area with high rates of CDIs was observed in the southeast of the ACT. To examine whether these spatial patterns was influenced by small populations in some postcodes, the proportion of positive CDIs was also mapped at the postcode level (Fig. 2B). This
Figure 1
Proportion of positive CDIs in the ACT, 2004e2014.
map showed a similar spatial pattern, with the same area in the southeast with a high proportion of proportion of C. difficile positives highlighted. Moran’s I was significant at the alpha Z 0.05 level (p < 0.004). Based on these results, we rejected the null hypothesis that there is zero spatial autocorrelation present in CDI patterns. Table 1 shows the posterior estimates of relative risk (RR) and the corresponding 95% CI. The number of tests was showed a marginal positive association with CDI (RR Z 1.01, 95% CI 1.00, 1.02). The index of relative socioeconomic advantage disadvantage (IRSAD) was negatively associated with CDIs (RR Z 0.74, 95% CI 0.56, 0.94). Fig. 3A indicates the relative risk of CDIs estimated for each postcode (compared to the whole of the ACT) after neighbourhood socio-economic deprivation, the number of tests, population density and the population over 65 years in each area are taken into account. The high-risk areas (RR 1.5) are concentrated in the central parts of the ACT with an additional high-risk postcode in the west. The cluster with the highest risk of infection occurs in the southeast (RR 2.5e3.5). Fig. 3B shows the probability exceedance estimates for each postcode. Fig. 3B brings into starker relief the clustering from the map of relative risks (Fig. 3A). The values represent the probability of a postcode having a more than twice the estimated risk of C. difficile infection. Here, we can see a cluster of three postcodes in the ACT with a high probability (0.8e1.0) of exceeding twice the risk of infection estimated for that postcode.
Discussion This study found localised areas of excess Clostridium difficile infection risk compared with surrounding communities and a positive association with neighbourhood socioeconomic disadvantage, using laboratory data from the Australian Capital Territory over an 11 year observation period. Since 2010, the proportion of positive Clostridium infections in the Australian Capital Territory has remained consistently higher compared with 2004. This rise is in accordance with the literature, reflecting the increase in CDIs that has been observed over the past 15 years worldwide [2,3] with an inflexion point occurring around the turn of the decade. A stable increase in incidence of CDI since has been noted in other Australian jurisdictions [33]. A strength of this study is that spatial differences in the frequency of diagnostic testing were controlled for and the data were collected over 11 years from a stable population. Increased testing has been suggested as a potential reason for the regional increases in CDI across Australia since mid2011 [7]. Accounting for spatial patterns in testing practices over the study period provides support for the existence of localised infection clusters that do not simply reflect increased testing and are indicative of potential locality specific demographic and socio-economic factors important for transmission of CDI independent of hospital-associated drivers. Furthermore, by using the Bayesian exceedance probability to identify areas of high disease risk, we overcome a limitation of frequently used cluster detection methods like SatScan [34]), where the spatial autocorrelation structure of the exposures and health outcomes (a
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
Spatial patterns of clostridium difficile
5
Figure 2 Spatial patterns of CDIs by Postcode in the ACT, 2004e2014 (A) Average annual incidence per 100,000 population (B) Proportion positive as a function of total number of tests done.
Table 1
Regression coefficients from Bayesian spatial CAR model of CDIs in the Australian Capital Territory.
Variable
Posterior mean SD
Monte Carlo error
RR (95% CI)
Intercept Number of tests Socio-economic advantage Population density Population >65 years Heterogeneity Structured Unstructured
0.156 0.200 0.0006 0.0003 0.299 0.130 0.163 0.130 0.001 0.001
<0.01 <0.01 <0.01 <0.01 <0.01
1.01 0.74 1.17 1.00
0.455 0.162 0.190 0.294
<0.01 <0.01
particular feature of infectious disease data) are not taken into account [35]. Using Bayesian exceedance probability, our study has identified areas where the probability of the relative risk of CDI exceeds a specified threshold, after accounting for the spatial dependence in exposures and outcomes, thus more likely to represent clustering of cases that are not an artefact of spatial sampling bias. In the present study all cases of CDI were considered. Such an analysis does not allow the identification of risk factors specific to community or hospital acquired infection. A previous study in Queensland did not find an association of community-acquired CDI with population-level medication exposure in Queensland during 2008e2012 [36]. A related issue is the emergence of new strains [37e39]. It is likely that spatial patterns of CDI will be strain specific, which was beyond the scope of the present study. As our study is the first examination of spatial patterns of CDI
(1.00e1.02) (0.57e0.94) (0.91e1.53) (1.00e1.01)
across the ACT, consideration of distinguishing community and hospital acquired infection as well as strain-specific patterns should be the focus of future work. Neighbourhood socio-economic status is associated with a person’s general health quality, with several infections having a higher rate of incidence amongst persons of lower socio-economic status [40,41]. Neighbourhood socioeconomic status may be an indicator for household crowding, poorer house conditions and poorer hygiene that can contribute to the increased likelihood of exposure to infectious agents [42e44]. Of particular relevance for CDI, studies in Scotland and New Zealand have found that increasing socio-economic deprivation and household crowding is also consistently associated with increased rates of antibiotic prescribing [34,45]. Thus, our finding of a positive spatial association of CDI with socio-economic disadvantage in the ACT is likely to be a proxy for factors
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
6
A. Lal et al.
Figure 3 (A) Estimated Relative risks of CDI by 2011 Postcodes in the Australian Capital Territory (B) exceedance probabilities for CDI, with the exceedance probability threshold of twice the estimated risk.
including health care access, antibiotic prescribing practices and living conditions, which if focussed on, can help to identify the mechanisms that underlie the association showed here. Our findings raise interesting questions for future research which include: are public health messages be specifically tailored for high-risk communities (e.g. those residing in socio-economically deprived areas)? The present study did not find a relationship of CDI with the proportion of elderly that usually resided in a postcode. Advanced age is a significant risk factor for CDI development [46], mortality and recurrence [47]. Comorbidities, increased antibiotic exposure, decreased host immunity, age-related alterations to intestinal flora and increased healthcare facility exposure [48] are likely to contribute to increased risk of CDI in the elderly. Proton pump inhibitor (PPI) use has been demonstrated to be linked to an increased risk in CDIs [23], and its use in Australia is on the rise [49]. PPI use has been shown to be a significant risk factor for CDI in nursing home patients [50]. However, as this is a population-level analysis, no inferences can be made with respect to individual-level risk factors such as PPI use which continue to be an important risk factor for CDI. A potential reason for the increased proportion of positive CDI tests is the change in testing practices from ELISAbased test to a more sensitive PCR-based test [35]. Although the impact of increased sensitivity in testing as a result of the addition of PCR testing in 2011 cannot be excluded as a possible reason for the increased incidence, this change is unlikely to bias spatial patterns as collection centres for ACT Pathology are evenly distributed throughout the region and our dataset includes the only two emergency departments in the ACT. However, we cannot
exclude the possibility of false positives in our dataset, as patients may be colonized with C. difficile but not have symptomatic infection [2]. A recent study in Australian tertiary hospitals found different risk factors for asymptomatic and symptomatic patients colonised with Clostridium difficile [51].
Conclusions In the ACT, the proportion of positive Clostridium difficile infections have increased between 2010 and 2014, compared with 2004. The positive spatial association with neighbourhood socio-economic disadvantage, and localized areas with excess risk compared with surrounding communities are relevant for public health messaging and infection control in the ACT, as we have determined certain areas that have a higher risk of CDI. This knowledge will contribute to the development of targeted strategies to reduce CDI in Australian communities and reduce the costs on the Australian healthcare system.
Ethics Ethical approval for the study was granted by the Australian National University Human Research Ethics Committee (HREC) (2014/725) and ACT Health HREC (ETHLR.14.276).
Authorship statement AL conceived the study design and led the analysis. AS provided data from Canberra hospital and provided
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
Spatial patterns of clostridium difficile extensive input into all drafts of the manuscript. TH led the writing and re-drafting of the manuscript.
Conflicts of interest The authors declare no conflict of interest.
Funding There was no external funding required for this project. A Lal’s salaried position at the ANU enabled her involvement.
Provenance and peer review Externally peer reviewed.
Acknowledgements We thank all staff involved in collecting and processing of laboratory specimens.
References [1] Bartlett JG. Narrative review: the new epidemic of Clostridium difficile-associated enteric disease. Ann Intern Med 2006 Nov 21;145(10):758e64. [2] Khanna S, Pardi DS, Aronson SL, Kammer PP, Orenstein R, St Sauver JL, et al. The epidemiology of community-acquired Clostridium difficile infection: a population-based study. Am J Gastroenterol 2012 Jan;107(1):89e95. [3] Viseur N, Lambert M, Delmee M, Van Broeck J, Catry B. Nosocomial and non-nosocomial Clostridium difficile infections in hospitalised patients in Belgium: compulsory surveillance data from 2008 to 2010. Euro Surveill 2011;16(43). [4] Cheng AC, Collins DA, Elliott B, Ferguson JK, Paterson DL, Thean S, et al. Laboratory-based surveillance of Clostridium difficile circulating in Australia, September e November 2010. Pathology 2016;48(3):257e60. [5] Kelly CP, LaMont JT. Clostridium difficileemore difficult than ever. N Engl J Med 2008 Oct 30;359(18):1932e40. [6] Rupnik M, Wilcox MH, Gerding DN. Clostridium difficile infection: new developments in epidemiology and pathogenesis. Nat Rev Microbiol 2009 Jul;7(7):526e36. [7] Slimings C, Armstrong P, Beckingham WD, Bull AL, Hall L, Kennedy KJ, et al. Increasing incidence of Clostridium difficile infection, Australia, 2011e2012. Med J Aust 2014 Mar 17; 200(5):272e6. [8] Kuntz JL, Johnson ES, Raebel MA, Petrik AF, Yang X, Thorp ML, et al. Epidemiology and healthcare costs of incident Clostridium difficile infections identified in the outpatient healthcare setting. Infect Control Hosp Epidemiol 2012 Oct; 33(10):1031e8. [9] Centers for Disease C, Prevention. Severe Clostridium difficile-associated disease in populations previously at low riskefour states, 2005. MMWR Morb Mortal Wkly Rep 2005 Dec 2;54(47):1201e5. [10] Borali E, Ortisi G, Moretti C, Stacul EF, Lipreri R, Gesu GP, et al. Community-acquired Clostridium difficile infection in children: a retrospective study. Dig Liver Dis 2015 Jun 9; 47(10):842e6. [11] Schutze GE, Willoughby RE, Committee on Infectious D, American Academy of P. Clostridium difficile infection in infants and children. Pediatrics 2013 Jan;131(1):196e200.
7 [12] Garey KW, Jiang ZD, Yadav Y, Mullins B, Wong K, Dupont HL. Peripartum Clostridium difficile infection: case series and review of the literature. Am J Obstet Gynecol 2008 Oct; 199(4):332e7. [13] Wozniak TM, Rubin G, MacIntyre CR. The emergence of community-acquired Clostridium difficile in an Australian hospital. Healthc Infect 2015;20(2):72e7. [14] Furuya-Kanamori L, McKenzie SJ, Yakob L, Clark J, Paterson DL, Riley TV, et al. Clostridium difficile infection seasonality: patterns across hemispheres and continents e a systematic review. PLoS One 2015;10(3):e0120730. [15] Hensgens MP, Keessen EC, Squire MM, Riley TV, Koene MG, de Boer E, et al. Clostridium difficile infection in the community: a zoonotic disease? Clin Microbiol Infect 2012 Jul;18(7):635e45. [16] Himsworth CG, Patrick DM, Mak S, Jardine CM, Tang P, Weese JS. Carriage of Clostridium difficile by wild urban Norway rats (Rattus norvegicus) and black rats (Rattus rattus). Appl Environ Microbiol 2014 Feb;80(4):1299e305. [17] Bandelj P, Trilar T, Blagus R, Ocepek M, Rousseau J, Weese JS, et al. Prevalence and molecular characterization of Clostridium difficile isolated from European Barn Swallows (Hirundo rustica) during migration. BMC Vet Res 2014; 10:40. [18] Bauer MP, Kuijper EJ. Potential sources of Clostridium difficile in human infection. Infect Dis Clin N Am 2015 Mar;29(1):29e35. [19] Bakri MM, Brown DJ, Butcher JP, Sutherland AD. Clostridium difficile in ready-to-eat salads, Scotland. Emerg Infect Dis 2009 May;15(5):817e8. [20] Gould LH, Limbago B. Clostridium difficile in food and domestic animals: a new foodborne pathogen? Clin Infect Dis 2010 Sep 1;51(5):577e82. [21] Songer JG, Trinh HT, Killgore GE, Thompson AD, McDonald LC, Limbago BM. Clostridium difficile in retail meat products, USA, 2007. Emerg Infect Dis 2009 May;15(5):819e21. [22] Slimings C, Riley TV. Antibiotics and hospital-acquired Clostridium difficile infection: update of systematic review and meta-analysis. J Antimicrob Chemother 2014 Apr;69(4): 881e91. [23] Kwok CS, Arthur AK, Anibueze CI, Singh S, Cavallazzi R, Loke YK. Risk of Clostridium difficile infection with acid suppressing drugs and antibiotics: meta-analysis. Am J Gastroenterol 2012 Jul;107(7):1011e9. [24] Anderson DJ, Rojas LF, Watson S, Knelson LP, Pruitt S, Lewis SS, et al. Identification of novel risk factors for community-acquired Clostridium difficile infection using spatial statistics and geographic information system analyses. PLoS One 2017;12(5):e0176285. [25] Hudspeth WB, Qeadan F, Phipps EC. Disparities in the incidence of community-acquired Clostridioides difficile infection: an area-based assessment of the role of social determinants in Bernalillo County, New Mexico. Am J Infect Contr 2019;47(7):773e9. [26] Furuya-Kanamori L, Robson J, Soares Magalhaes RJ, Yakob L, McKenzie SJ, Paterson DL, et al. A population-based spatiotemporal analysis of Clostridium difficile infection in Queensland, Australia over a 10-year period. J Infect 2014 Jun 28;69(5):447e55. [27] Simonsen J, Frisch M, Ethelberg S. Socioeconomic risk factors for bacterial gastrointestinal infections. Epidemiology 2008 Mar;19(2):282e90. [28] Moono P, Lim SC, Riley TV. High prevalence of toxigenic Clostridium difficile in public space lawns in Western Australia. Sci Rep 2017;7:41196. [29] van Dorp SM, Hensgens MP, Dekkers OM, Demeulemeester A, Buiting A, Bloembergen P, et al. Spatial clustering and livestock exposure as risk factor for community-acquired Clostridium difficile infection. Clin Microbiol Infect 2019;25(5): 607e12.
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002
+
MODEL
8 [30] Lunn DJ, Thomas A, Best N, Spiegelhalter D. WinBUGS e a Bayesian modelling framework: concepts, structure, and extensibility. Stat Comput 2000;10:325e37. [31] Lawson AB, Browne WJ, Rodeiro CLV. Disease mapping with WinBUGS and MLwiN. John Wiley & Sons; 2003. [32] Gelman A, Rubin DB. Inference from iterative simulation using multiple sequences. Stat Sci 1992:457e72. [33] Worth LJ, Spelman T, Bull AL, Brett JA, Richards MJ. Epidemiology of Clostridium difficile infections in Australia: enhanced surveillance to evaluate time trends and severity of illness in Victoria, 2010e2014. J Hosp Infect 2016 Jul;93(3): 280e5. [34] Walls G, Vandal AC, du Plessis T, Playle V, Holland DJ. Socioeconomic factors correlating with community antimicrobial prescribing. N Z Med J 2014;128(1417):16e23. [35] Khanna S, Baddour LM, Huskins WC, Kammer PP, Faubion WA, Zinsmeister AR, et al. The epidemiology of Clostridium difficile infection in children: a population-based study. Clin Infect Dis 2013 May 15;56(10):1401e6. [36] Furuya-Kanamori L, Yakob L, Riley TV, Paterson DL, Baker P, McKenzie SJ, et al. Community-acquired Clostridium difficile infection, Queensland, Australia. Emerg Infect Dis 2016;22(9): 1659. [37] Eyre DW, Tracey L, Elliott B, Slimings C, Huntington PG, Stuart RL, et al. Emergence and spread of predominantly community-onset Clostridium difficile PCR ribotype 244 infection in Australia, 2010 to 2012. Euro Surveill 2015;20(10): 21059. [38] Lim SK, Stuart RL, Mackin KE, Carter GP, Kotsanas D, Francis MJ, et al. Emergence of a ribotype 244 strain of Clostridium difficile associated with severe disease and related to the epidemic ribotype 027 strain. Clin Infect Dis 2014;58(12):1723e30. [39] De Almeida MN, Heffernan H, Dervan A, Bakker S, Freeman JT, Bhally H, et al. Severe Clostridium difficile infection in New Zealand associated with an emerging strain, PCR-ribotype 244. N Z Med J 2013;126(1380):9e14. [40] Flory JH, Joffe M, Fishman NO, Edelstein PH, Metlay JP. Socioeconomic risk factors for bacteraemic pneumococcal pneumonia in adults. Epidemiol Infect 2009 May;137(5): 717e26.
A. Lal et al. [41] Huggan PJ, Wells JE, Browne M, Richardson A, Murdoch DR, Chambers ST. Population-based epidemiology of Staphylococcus aureus bloodstream infection in Canterbury, New Zealand. Intern Med J 2010 Feb;40(2):117e25. [42] Baker MG, Barnard LT, Kvalsvig A, Verrall A, Zhang J, Keall M, et al. Increasing incidence of serious infectious diseases and inequalities in New Zealand: a national epidemiological study. Lancet 2012 Mar 24;379(9821):1112e9. [43] Baker M, Das D, Venugopal K, Howden-Chapman P. Tuberculosis associated with household crowding in a developed country. J Epidemiol Community Health 2008 Aug;62(8):715e21. [44] Huang SS, Finkelstein JA, Rifas-Shiman SL, Kleinman K, Platt R. Community-level predictors of pneumococcal carriage and resistance in young children. Am J Epidemiol 2004 Apr 1; 159(7):645e54. [45] Covvey JR, Johnson BF, Elliott V, Malcolm W, Mullen AB. An association between socioeconomic deprivation and primary care antibiotic prescribing in Scotland. J Antimicrob Chemother 2014 March 1;69(3):835e41. [46] To KB, Napolitano LM. Clostridium difficile infection: update on diagnosis, epidemiology, and treatment strategies. Surg Infect (Larchmt) 2014 Oct;15(5):490e502. [47] Abou Chakra CN, Pepin J, Sirard S, Valiquette L. Risk factors for recurrence, complications and mortality in Clostridium difficile infection: a systematic review. PLoS One 2014;9(6): e98400. [48] Mizusawa M, Doron S, Gorbach S. Clostridium difficile diarrhea in the elderly: current issues and management options. Drugs Aging 2015 Aug;32(8):639e47. [49] Hollingworth S, Duncan EL, Martin JH. Marked increase in proton pump inhibitors use in Australia. Pharmacoepidemiol Drug Saf 2010 Oct;19(10):1019e24. [50] Al-Tureihi FI, Hassoun A, Wolf-Klein G, Isenberg H. Albumin, length of stay, and proton pump inhibitors: key factors in Clostridium difficile-associated disease in nursing home patients. J Am Med Dir Assoc 2005 Mar-Apr;6(2):105e8. [51] Furuya-Kanamori L, Clements ACA, Foster NF, Huber CA, Hong S, Harris-Brown T, et al. Asymptomatic Clostridium difficile colonization in two Australian tertiary hospitals, 2012e2014: prospective, repeated cross-sectional study. Clin Microbiol Infect 2017;23(1). 48.e1-.e7.
Please cite this article as: Lal A et al., Spatial clusters of Clostridium difficile infection and an association with neighbourhood socioeconomic disadvantage in the Australian Capital Territory, 2004e2014, Infection, Disease & Health, https://doi.org/10.1016/ j.idh.2019.08.002