WEBVTT 1 00:00:01.230 --> 00:00:02.790 Welcome back. 2 00:00:02.790 --> 00:00:07.790 This lecture is a bonus presentation on disease clusters. 3 00:00:07.860 --> 00:00:11.220 I've done some research in the past on disease clusters, 4 00:00:11.220 --> 00:00:14.220 and I thought I could share some of that work with you. 5 00:00:14.220 --> 00:00:15.660 However, if you are not 6 00:00:15.660 --> 00:00:17.340 particularly interested in this topic, 7 00:00:17.340 --> 00:00:19.770 or find yourself very busy with other work, 8 00:00:19.770 --> 00:00:21.630 then you could definitely skip this lecture 9 00:00:21.630 --> 00:00:23.820 and just focus on the previous two lectures 10 00:00:23.820 --> 00:00:26.970 related to epidemiological study types 11 00:00:26.970 --> 00:00:29.703 and Hill's causal criteria. 12 00:00:31.620 --> 00:00:35.010 This slide shows an image from another cluster study 13 00:00:35.010 --> 00:00:36.900 that I was not involved in 14 00:00:36.900 --> 00:00:40.620 examining spatiotemporal clusters of cholera cases in Haiti, 15 00:00:40.620 --> 00:00:43.170 and I wanted to include this as another example, 16 00:00:43.170 --> 00:00:44.523 should you be interested. 17 00:00:47.970 --> 00:00:50.367 So let's quickly run through some basics. 18 00:00:50.367 --> 00:00:53.010 What are disease clusters? 19 00:00:53.010 --> 00:00:55.620 They're defined as a group of disease cases 20 00:00:55.620 --> 00:00:59.820 in time or in space, or both, 21 00:00:59.820 --> 00:01:03.150 in excess of the number that would normally be expected. 22 00:01:03.150 --> 00:01:05.040 So you could have temporal clusters, 23 00:01:05.040 --> 00:01:08.433 spatial clusters, or space-time clusters. 24 00:01:11.220 --> 00:01:13.140 So why do clusters occur? 25 00:01:13.140 --> 00:01:16.020 They can occur for a number of different reasons. 26 00:01:16.020 --> 00:01:18.750 They could occur due to the infectious spread of disease, 27 00:01:18.750 --> 00:01:21.000 the occurrence of disease vectors 28 00:01:21.000 --> 00:01:24.660 like mosquitoes or ticks in a specific location, 29 00:01:24.660 --> 00:01:27.660 the clustering of one or more risk factors, 30 00:01:27.660 --> 00:01:29.550 or the existence of health hazards 31 00:01:29.550 --> 00:01:31.683 like localized sources of pollution. 32 00:01:33.510 --> 00:01:36.510 Why is cluster analysis useful? 33 00:01:36.510 --> 00:01:39.000 Cluster detection methods can be used 34 00:01:39.000 --> 00:01:43.050 to identify high risk areas and time periods. 35 00:01:43.050 --> 00:01:45.300 They can also be used to link disease patterns 36 00:01:45.300 --> 00:01:48.060 to sources of contamination. 37 00:01:48.060 --> 00:01:50.580 They can highlight retrospective outbreaks, 38 00:01:50.580 --> 00:01:51.960 so we can look back in time 39 00:01:51.960 --> 00:01:54.030 and identify outbreaks of disease, 40 00:01:54.030 --> 00:01:58.350 or we can look forward and prospectively predict outbreaks 41 00:01:58.350 --> 00:02:00.870 using routine surveillance data. 42 00:02:00.870 --> 00:02:03.330 And generally cluster analysis can help to determine 43 00:02:03.330 --> 00:02:06.120 whether the clustering is statistically significant 44 00:02:06.120 --> 00:02:08.580 and worthy of further investigation, 45 00:02:08.580 --> 00:02:10.800 or whether it's likely to be a chance occurrence, 46 00:02:10.800 --> 00:02:15.120 or just reflects the distribution of the population at risk. 47 00:02:15.120 --> 00:02:18.570 Because in some cases, if you, for example, 48 00:02:18.570 --> 00:02:20.550 are looking at disease in children, 49 00:02:20.550 --> 00:02:23.490 maybe more children just happen to live in one area 50 00:02:23.490 --> 00:02:26.043 and that's why it looks like a cluster of disease. 51 00:02:27.990 --> 00:02:30.180 Now I wanna walk through a case study 52 00:02:30.180 --> 00:02:31.440 that's from my own research 53 00:02:31.440 --> 00:02:34.260 on the detection of space-time clusters 54 00:02:34.260 --> 00:02:37.590 of cryptosporidiosis in New Zealand. 55 00:02:37.590 --> 00:02:40.620 So let's start with a little bit of background. 56 00:02:40.620 --> 00:02:42.870 Cryptosporidium species are parasites 57 00:02:42.870 --> 00:02:45.150 that can cause gastrointestinal symptoms, 58 00:02:45.150 --> 00:02:47.490 and in very rare cases sometimes pulmonary 59 00:02:47.490 --> 00:02:49.383 or other symptoms in humans. 60 00:02:50.400 --> 00:02:52.890 Cryptosporidiosis cases are often linked 61 00:02:52.890 --> 00:02:54.515 to contaminated water 62 00:02:54.515 --> 00:02:58.770 or contact with infected animals or their feces. 63 00:02:58.770 --> 00:03:01.200 Animals that can carry cryptosporidium 64 00:03:01.200 --> 00:03:05.310 range from cattle, sheep, 65 00:03:05.310 --> 00:03:07.473 to other mammals as well. 66 00:03:08.730 --> 00:03:11.940 New Zealand generally has high rates of enteric diseases, 67 00:03:11.940 --> 00:03:14.820 and enteric disease just means infections 68 00:03:14.820 --> 00:03:17.370 occurring in the gastrointestinal system. 69 00:03:17.370 --> 00:03:19.770 But they have high rates of these diseases in general, 70 00:03:19.770 --> 00:03:22.350 and that includes cryptosporidiosis, 71 00:03:22.350 --> 00:03:26.190 especially compared to other high income nations. 72 00:03:26.190 --> 00:03:29.400 Cryptosporidiosis cases are, or rates are, 73 00:03:29.400 --> 00:03:33.150 almost 2.5 times higher in rural areas of New Zealand 74 00:03:33.150 --> 00:03:35.130 than in urban areas, 75 00:03:35.130 --> 00:03:37.950 and a dose response relationship has been detected 76 00:03:37.950 --> 00:03:40.110 between cases and rurality. 77 00:03:40.110 --> 00:03:44.820 So for each step 78 00:03:44.820 --> 00:03:48.180 towards more rural that you go, 79 00:03:48.180 --> 00:03:51.093 you see an increase in the number of cases. 80 00:03:52.440 --> 00:03:55.740 The highest incidence rates have been reported 81 00:03:55.740 --> 00:03:58.650 in infants and in children under the age of five, 82 00:03:58.650 --> 00:04:00.843 so they're a vulnerable population. 83 00:04:02.670 --> 00:04:05.430 So cluster detection may help to identify risk factors 84 00:04:05.430 --> 00:04:08.850 and inform disease prevention and control efforts. 85 00:04:08.850 --> 00:04:10.380 So here are some of the methods 86 00:04:10.380 --> 00:04:12.780 that were used in my research. 87 00:04:12.780 --> 00:04:13.950 I'll walk through them really quickly, 88 00:04:13.950 --> 00:04:17.220 but you don't really need to remember these details. 89 00:04:17.220 --> 00:04:19.240 So we use SaTScan software 90 00:04:20.160 --> 00:04:21.720 using a Kulldorff method 91 00:04:21.720 --> 00:04:24.570 of retrospective space-time permutation. 92 00:04:24.570 --> 00:04:28.783 So essentially we looked back in time from 1997 to 2015 93 00:04:30.000 --> 00:04:34.110 to detect clusters of cryptosporidiosis, 94 00:04:34.110 --> 00:04:37.170 both spatially and temporarily, in New Zealand. 95 00:04:37.170 --> 00:04:41.190 So this method scans the study area for each location, 96 00:04:41.190 --> 00:04:43.500 and in this case we looked at small areas 97 00:04:43.500 --> 00:04:46.380 called census area units within the country, 98 00:04:46.380 --> 00:04:48.630 and then it scans the specified time period 99 00:04:48.630 --> 00:04:51.663 and highlights potential clusters of disease cases. 100 00:04:55.530 --> 00:04:57.540 Here are a few of the key results. 101 00:04:57.540 --> 00:05:01.590 65 clusters were detected during the study period, 102 00:05:01.590 --> 00:05:04.800 and 38 of those were statistically significant 103 00:05:04.800 --> 00:05:07.063 with a P value of less than 0.05. 104 00:05:08.460 --> 00:05:12.780 However, only about 4.1% of the notified cases, 105 00:05:12.780 --> 00:05:16.050 of all notified cases from 1997 to 2015, 106 00:05:16.050 --> 00:05:18.150 so across the entire study period, 107 00:05:18.150 --> 00:05:20.460 were part of significant clusters. 108 00:05:20.460 --> 00:05:22.770 So that means that the majority of cases 109 00:05:22.770 --> 00:05:26.160 of this disease in New Zealand are actually sporadic, 110 00:05:26.160 --> 00:05:27.570 so they pop up here and there 111 00:05:27.570 --> 00:05:30.783 but they're exhibiting clusters. 112 00:05:34.200 --> 00:05:35.850 This map shows the locations 113 00:05:35.850 --> 00:05:38.550 of the 38 statistically significant clusters 114 00:05:38.550 --> 00:05:39.693 that were detected. 115 00:05:41.160 --> 00:05:43.350 You can see that there are clusters 116 00:05:43.350 --> 00:05:45.300 located throughout the country, 117 00:05:45.300 --> 00:05:46.710 but what this map doesn't show you 118 00:05:46.710 --> 00:05:48.990 is the timing of these clusters. 119 00:05:48.990 --> 00:05:50.880 What I can tell you from those results 120 00:05:50.880 --> 00:05:52.920 are that clusters in rural areas 121 00:05:52.920 --> 00:05:54.810 often occurred in the spring, 122 00:05:54.810 --> 00:05:56.340 while clusters in urban areas 123 00:05:56.340 --> 00:05:58.833 often occurred during autumn or winter months. 124 00:06:05.130 --> 00:06:06.630 So we asked ourselves, 125 00:06:06.630 --> 00:06:10.350 why might we be seeing clusters of cryptosporidiosis 126 00:06:10.350 --> 00:06:14.010 in rural areas of New Zealand in the spring months? 127 00:06:14.010 --> 00:06:17.220 What's happening in the spring in rural New Zealand 128 00:06:17.220 --> 00:06:20.790 that might explain the timing and location 129 00:06:20.790 --> 00:06:22.503 of these clusters of disease? 130 00:06:23.430 --> 00:06:24.810 Well, it just so happens 131 00:06:24.810 --> 00:06:29.250 that spring is calving and lambing season in New Zealand, 132 00:06:29.250 --> 00:06:32.250 and calves and lambs can carry 133 00:06:32.250 --> 00:06:35.460 high levels of Cryptosporidium parvum. 134 00:06:35.460 --> 00:06:37.800 It's a species of cryptosporidium 135 00:06:37.800 --> 00:06:42.060 that can spread from animal feces to humans 136 00:06:42.060 --> 00:06:46.410 through direct contact or through indirect contact. 137 00:06:46.410 --> 00:06:48.300 You could have direct contact with feces 138 00:06:48.300 --> 00:06:49.830 if you were out in a field, 139 00:06:49.830 --> 00:06:51.870 but it could also wash into waterways 140 00:06:51.870 --> 00:06:54.960 and contaminate water that you drink 141 00:06:54.960 --> 00:06:57.093 or water that you recreate in. 142 00:06:57.960 --> 00:07:01.320 So what we might be seeing is the zoonotic transmission 143 00:07:01.320 --> 00:07:03.810 of this particular species of cryptosporidium 144 00:07:03.810 --> 00:07:05.103 in rural areas. 145 00:07:09.000 --> 00:07:12.870 Some of our takeaways were that clusters in urban areas 146 00:07:12.870 --> 00:07:15.030 in autumn and winter months 147 00:07:15.030 --> 00:07:18.810 might be due to person to person transmission, 148 00:07:18.810 --> 00:07:22.413 and may be due to the species Cryptosporidium hominis, 149 00:07:23.430 --> 00:07:27.300 whereas clusters in rural areas in spring months 150 00:07:27.300 --> 00:07:30.030 might be subject to zoonotic, environmental, 151 00:07:30.030 --> 00:07:32.010 or waterborne transmission 152 00:07:32.010 --> 00:07:34.503 and the spread of Cryptosporidium parvum. 153 00:07:35.850 --> 00:07:37.702 This is in line with some of the previous research 154 00:07:37.702 --> 00:07:39.420 that's come out of the country, 155 00:07:39.420 --> 00:07:41.400 suggesting that there are seasonal, 156 00:07:41.400 --> 00:07:45.210 strain-specific transmission cycles. 157 00:07:45.210 --> 00:07:46.440 Going forward, 158 00:07:46.440 --> 00:07:49.950 the typing of all notified cases of cryptosporidium, 159 00:07:49.950 --> 00:07:51.900 and the typing of all outbreaks, 160 00:07:51.900 --> 00:07:54.810 would improve our understanding of the pathways 161 00:07:54.810 --> 00:07:58.500 and enable development of targeted disease control measures 162 00:07:58.500 --> 00:08:00.213 to help prevent this disease. 163 00:08:03.510 --> 00:08:06.990 Finally, we asked ourselves what the next steps would be 164 00:08:06.990 --> 00:08:09.162 in such an investigation. 165 00:08:09.162 --> 00:08:11.640 One of the things that we've actually done 166 00:08:11.640 --> 00:08:16.640 is to go back and compare the detected clusters of disease 167 00:08:16.680 --> 00:08:18.969 to known outbreaks that have been reported 168 00:08:18.969 --> 00:08:23.013 by the National Disease Surveillance System in New Zealand. 169 00:08:23.880 --> 00:08:25.320 That way we get a better sense 170 00:08:25.320 --> 00:08:28.180 of whether some outbreaks are going undetected. 171 00:08:28.180 --> 00:08:32.460 We also get a sense of potential transmission pathways 172 00:08:32.460 --> 00:08:35.373 from those previous outbreak investigations. 173 00:08:36.278 --> 00:08:38.760 The other thing that we did was to compare 174 00:08:38.760 --> 00:08:42.960 the location and timing of the clusters 175 00:08:42.960 --> 00:08:46.140 to potential environmental risk factors. 176 00:08:46.140 --> 00:08:49.560 For example, we can compare the clusters 177 00:08:49.560 --> 00:08:53.370 to spatial patterns of livestock density in the country. 178 00:08:53.370 --> 00:08:57.000 We can also compare them to spatiotemporal patterns 179 00:08:57.000 --> 00:08:58.950 of severe weather events, 180 00:08:58.950 --> 00:09:01.587 because for example, outbreaks, 181 00:09:01.587 --> 00:09:04.200 especially those that are waterborne, 182 00:09:04.200 --> 00:09:06.570 might be more common after heavy rains 183 00:09:06.570 --> 00:09:11.100 wash animal feces off of the land and into waterways, 184 00:09:11.100 --> 00:09:12.680 especially in rural areas 185 00:09:12.680 --> 00:09:16.410 where many people rely on drinking water sources 186 00:09:16.410 --> 00:09:18.720 that are not reticulated, 187 00:09:18.720 --> 00:09:21.510 so that is to say they aren't going through 188 00:09:21.510 --> 00:09:23.343 a drinking water treatment plant. 189 00:09:27.990 --> 00:09:30.450 So that was a very brief introduction 190 00:09:30.450 --> 00:09:31.650 to the type of research 191 00:09:31.650 --> 00:09:35.343 that can be done with cluster detection analysis. 192 00:09:36.450 --> 00:09:39.390 The key takeaways are that cluster analysis 193 00:09:39.390 --> 00:09:41.550 can help to do a number of different things, 194 00:09:41.550 --> 00:09:45.750 including identify those high risk areas and time periods. 195 00:09:45.750 --> 00:09:47.910 They can help you to link disease patterns 196 00:09:47.910 --> 00:09:50.940 to sources of contamination, 197 00:09:50.940 --> 00:09:53.082 and it can highlight both retrospective outbreaks 198 00:09:53.082 --> 00:09:55.950 and prospectively predict outbreaks 199 00:09:55.950 --> 00:09:58.560 using routine surveillance data. 200 00:09:58.560 --> 00:10:01.548 So that's something to bear in mind as you go forward. 201 00:10:01.548 --> 00:10:04.560 And if you're interested in more 202 00:10:04.560 --> 00:10:07.050 on cluster detection and analysis, 203 00:10:07.050 --> 00:10:09.900 I'm happy to provide you with additional resources 204 00:10:09.900 --> 00:10:11.500 that you might find interesting.