Infection Control & Hospital Epidemiology
265
Detecting Clostridium difficile outbreaks with ward-specific cut-off levels based on the Poisson distribution
Jon E. Edman-Wallér MD1,2 and Maria Werner MD, PhD1 1Department of Hospital Infection Control, Södra Älvsborg Hospital, Borås, Sweden and 2Center for Antibiotic Resistance Research (CARe), Department of
Infectious Diseases, Institute of Biomedicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden
To the Editor—Clostridium difficile infections (CDI) are among the most common healthcare facility-associated infections, and hospital outbreaks occur regularly. Outbreaks often occur either as a general increase in incidence on the hospital level, or as a cluster of cases within a single ward. For the latter type of outbreak, it is not entirely straightforward to determine the number of cases within a ward in a certain time period that ought to make one suspect an outbreak. In Ontario, Canada,1 notification thresholds for further investigation of possible outbreaks are set at 3 cases within a week or 5 cases within 4 weeks for wards with >20 beds. For wards with <20 beds, the threshold is set at 2 cases within 1 week or 4 cases within 4 weeks. Although this approach is sensible, it does not take into account that there is a variability of CDI background rates at different wards.2 At our hospital, the CDI rate varies 10-fold between high-incidence wards (eg, infectious diseases, gastroenterology) and low-incidence wards (eg, ortho- pedics, gynecology). For example, 3 cases within 4 weeks happen regularly at the former but very seldom at the latter. We would hereby like to propose a method for taking the variability in ward- specific rates in account when determining notification thresholds for CDI. We intend to validate the method scientifically, and in the meantime, we hope that this letter will inspire others to implement, test, and refine the method at their own hospitals. We welcome contact with other researchers interested in this issue. The Poisson distribution (Figure 1) describes the probability
for a certain number of independent events to occur within a specified period. Say, for instance, that you happen to bump into an old friend from your school years about once every year on average. You would not be surprised to have 0, 1, or 2 encounters during a year; 3 or 4 might raise your eyebrows, and if you encounter 5 or more schoolfriends in a single year, you may question the randomness of the situation. Perhaps your friends have talked to each other and decided to meet up, thus making the events dependent. Consulting the Poisson distribution, you could confirm your intuition: given an average of one encounter per year, random encounters of 5 old friends in a single year would be expected to happen only once every 326 years. The formula for the Poisson distribution for a given period is
PX=x ðÞ= λxeλ x !
where P(X=x) is the the probability for x events; x=0, 1, 2, 3... (ie, the number of events in question); λ is the mean number of events per time period over the long run; e is Euler´sconstant ≈2.718; and ! is the factorial sign. For example, 4!=4×3×2×1.
Author for correspondence: Jon E. Edman-Wallér, Department of Hospital Infection Control, Södra Älvsborg Hospital, Borås, Sweden. E-mail:
jon.edman@vgregion.se Cite this article: Edman-Wallér JE,Werner M. (2019). Detecting Clostridium difficile
outbreaks with ward-specific cut-off levels based on the Poisson distribution. Infection Control & Hospital Epidemiology 2019, 40, 265–266. doi: 10.1017/ice.2018.299
© 2018 by The Society for Healthcare Epidemiology of America. All rights reserved.
Fig. 1. The Poisson distribution for 3 different values of λ. Sporadic cases of CDI in a ward should theoretically follow the
Poisson distribution. In the event of spread within a ward, the cases are not independent and thus do not respect the distribution, similar to when your old friends talk to each other and decide to appear at the same time. An unlikely high number of cases for the ward in question should raise suspicion of spread within the ward. The method can be seen as a diagnostic test for CDI outbreaks, where there is a trade-off between sensitivity (ie, detecting most outbreaks) and specificity (ie, a low level of “false alarms”).Here,wedescribehow we have implemented the method at our secondary-care, 420-bed hospital and a neighboring 100-bed hospital in western Sweden. We determine the λ value for each ward at the hospital mea-
sured as cases permonth. In selecting the number of months to base λ on, there is a trade-off between having too few measure points such that chance impacts too much, and using data that is too old to be relevant. We used 36 months as a reasonable compromise. We then determined the variance (= standard deviation squared) of the distribution of cases per month, which should be equal to λ.Ifnot, it is not a Poisson distribution. At the wards in our hospital, most wards had a variance that was almost uncannily close to their λ, strengthening our assumption that sporadic CDI cases follow the distribution not only in theory but also in practice. A few wards, however, had variances that were substantially higher than their λ, suggesting that there may have been small outbreaks that went under our radar at the time. To not contaminate the background rates with possible outbreaks, we deleted any outliers (ie, months with an unusually high number of cases in a ward) from the sample until the variance approached λ. This measure could be discussed from a methodological standpoint, but as a safety measure, we deemed it appropriate. We chose a cutoff of 1% (ie, the number of cases in a ward that is expected to occur more infrequently than once every 100 months). This cutoff was set arbitrarily and will likely change as we test and refine the
method.Using this cutoff and
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124 |
Page 125 |
Page 126 |
Page 127 |
Page 128 |
Page 129 |
Page 130 |
Page 131 |
Page 132 |
Page 133 |
Page 134 |
Page 135 |
Page 136 |
Page 137 |
Page 138 |
Page 139 |
Page 140 |
Page 141 |
Page 142 |
Page 143 |
Page 144 |
Page 145 |
Page 146 |
Page 147 |
Page 148 |
Page 149 |
Page 150 |
Page 151 |
Page 152 |
Page 153 |
Page 154 |
Page 155 |
Page 156
