- Email: [email protected]
Surveillance of diarrhoea in small animal practice through the Small Animal Veterinary Surveillance Network (SAVSNET)
Accepted Manuscript Title: Surveillance of diarrhoea in small animal practice through the Small Animal Veterinary Surveillance Network (SAVSNET) Author: P.H. Jones, S. Dawson, R.M. Gaskell, K.P. Coyne, Á. Tierney, C. Setzkorn, A.D. Radford, P-J.M. Noble PII: DOI: Reference:
S1090-0233(14)00241-X http://dx.doi.org/doi:10.1016/j.tvjl.2014.05.044 YTVJL 4180
To appear in:
The Veterinary Journal
Accepted date:
31-5-2014
Please cite this article as: P.H. Jones, S. Dawson, R.M. Gaskell, K.P. Coyne, Á. Tierney, C. Setzkorn, A.D. Radford, P-J.M. Noble, Surveillance of diarrhoea in small animal practice through the Small Animal Veterinary Surveillance Network (SAVSNET), The Veterinary Journal (2014), http://dx.doi.org/doi:10.1016/j.tvjl.2014.05.044. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Surveillance of diarrhoea in small animal practice through the Small Animal Veterinary Surveillance Network (SAVSNET) P.H. Jones a, c, S. Dawson b, R.M. Gaskell a, K.P. Coyne a, Á. Tierney a, C. Setzkorn a, A.D. Radford a, P-J.M. Noble b, * a
University of Liverpool, Institute of Global Health, Leahurst Campus, Chester High Road, Neston, Cheshire, CH64 7TE, UK b University of Liverpool School of Veterinary Science, Leahurst Campus, Chester High Road, Neston, Cheshire, CH64 7TE, UK c National Consortium for Zoonosis Research, Leahurst Campus, Chester High Road, Neston, Cheshire, CH64 7TE, UK
* Corresponding author. Tel.: +44 151 795 6205. E-mail address: [email protected] (P-J.M. Noble).
Page 1 Page 1 of 21
20 21
Abstract Using the Small Animal Veterinary Surveillance Network (SAVSNET), a national
22
small animal disease-surveillance scheme, information on gastrointestinal disease was
23
collected for a total of 76 days between 10 May 2010 and 8 August 2011 from 16,223
24
consultations (including data from 9,115 individual dogs and 3,462 individual cats) from 42
25
premises belonging to 19 UK veterinary practices. During that period, 7% of dogs and 3% of
26
cats presented with diarrhoea.
27 28
Adult dogs had a higher proportional morbidity of diarrhoea (PMD) than adult cats (P
29
< 0.001). This difference was not observed in animals < 1 year old. Younger animals in both
30
species had higher PMDs than adult animals (P < 0.001). Neutering was associated with
31
reduced PMD in young male dogs. In adult dogs, miniature Schnauzers had the highest PMD.
32
Most animals with diarrhoea (51%) presented having been ill for 2-4 days, but a history of
33
vomiting or haemorrhagic diarrhoea was associated with a shorter time to presentation. The
34
most common treatments employed were dietary modification (66% of dogs; 63% of cats)
35
and antibacterials (63% of dogs; 49% of cats). There was variability in PMD between
36
different practices.
37 38
The SAVNET methodology facilitates rapid collection of cross-sectional data
39
regarding diarrhoea, a recognised sentinel for infectious disease, and characterises data that
40
could benchmark clinical practice and support the development of evidence-based medicine.
41 42
Keywords: Breed; Companion animal; Diarrhoea; Surveillance; SAVNET
43
Page 2 Page 2 of 21
44 45
Introduction Gastrointestinal (GI) disease commonly results in the presentation of pets to UK
46
veterinary surgeons, but few national statistics record the frequency or the diagnostic and
47
therapeutic approach to these cases. A study of dogs presented to veterinary practices in the
48
USA suggested that 2.2% of veterinary consultations were related to diarrhoea (Lund et al.,
49
1999). In UK, a survey using client questionnaires reported that up to 20% of dogs had mild
50
vomiting and up to 15% had mild diarrhoea over a 2-week period (Hubbard et al., 2007), and
51
using data from notes provided with referral cases, German et al. (2010) highlighted the high
52
levels of antibacterial drugs used to treat GI disease. However, that study lacked
53
denominators to set it in the context of all animals with GI disease and was based on a
54
comparatively small sample with low statistical power.
55 56
Radford et al. (2011) showed that the presence of GI disease increased the probability
57
that a veterinary surgeon would prescribe antibacterials in first opinion practice, although the
58
prevalence of diarrhoea was not reported. In comparison, data on human disease are much
59
more detailed, with studies of far larger populations coordinated by National Health Service
60
recording and surveillance systems alongside national auditing (O'Brien et al., 2010; Smith et
61
al., 2010).
62 63
It is clear that a more coordinated approach to diarrhoea surveillance in companion
64
animals is needed. As well as quantifying the disease burden, such an approach could identify
65
risk factors for disease susceptibility as well as determining outcome measures associated
66
with specific diagnostic and therapeutic approaches, a prerequisite for the development of
67
evidence-based medicine. Changes in the incidence of diarrhoea could act as a sentinel for
68
infectious disease outbreaks (Smith et al., 2010) and warn of the emergence of new GI
Page 3 Page 3 of 21
69
pathogens.
70 Disease surveillance schemes exist for livestock 1 and horses.2 The Small Animal
71 72
Veterinary Surveillance Network (SAVSNET) monitors disease in small animals attending
73
first opinion practice, using data collected from veterinary laboratories and near real-time
74
collection of consultation records from participating veterinary practices. In this novel study,
75
we used data gathered during pilot studies to establish the feasibility of SAVSNET
76
methodologies to profile the presentation, diagnostic approach and management choices for
77
dogs and cats presenting with diarrhoea to small animal practices in the UK.
78 79
Materials and methods
80
Data collection
81
Data were collected from practices using a compatible version of practice
82
management software (Premvet, Vetsolutions, v03.02.12) following a positive response to a
83
postal request. Seventy-four practices were approached, recruiting 16/59, 3/7, 0/6 and 0/2
84
practices in England, Wales, Scotland and Northern Ireland, respectively (in total 19 practices
85
comprising 42 premises). Data on GI disease were collected over a total of 76 days between
86
10 May 2010 and 8 August 2011. Data were only collected from consultations relating to sick
87
animals, and excluded vaccine consultations.
88 89
At the end of each consultation, the veterinary surgeon was asked whether the case
1
See: AHVLA, 2012. Veterinary Laboratories Agency: Veterinary Investigation Surveillance Report. http://www.defra.gov.uk/ahvla-en/publication/vida12/ (accessed 29 May 2014).
2
See: AHT, 2012.Animal Health Trust. DEFRA/AHT/BEVA Equine Quarterly Disease Surveillance Reports. Animal Health Trust. http://www.aht.org.uk/cmsdisplay/disease_surveillance.html (accessed 29 May 2014).
Page 4 Page 4 of 21
90
had presented for vomiting or diarrhoea. If the answer to this question was ‘no’, the
91
questionnaire terminated; if ‘yes’, the questionnaire was completed as shown (Fig. 1). The
92
questionnaire responses, signalment and demographic data and the free text consultation
93
record were collected and stored in the SAVSNET database. Data were excluded if the client
94
had opted out of study participation.
95 96 97
Statistical analysis Multiple visits for individual animals were not included in the analysis. For animals
98
that never presented with diarrhoea, data from the first consultation only were used. For
99
animals that presented with diarrhoea, only data from the first consultation for diarrhoea were
100
selected. Thus, the proportions of cases of diarrhoea approximated the proportional morbidity
101
of diarrhoea (PMD; Martin et al., 1987), where the total number of diseased animals was
102
approximated by the total number of animals presenting to participating veterinary practices
103
for sick animal consultations.
104 105
Univariable and multivariable logistic regression was used to model the presentation
106
of diarrhoea (as a binary dependent variable) and the resulting models were used to estimate
107
morbidity odds ratios (ORs), a statistic that can be interpreted as a relative risk on the
108
assumption that the morbidity rate for all other causes was unrelated to exposure to the risk
109
factor (Miettinen and Wang, 1981). The explanatory variables considered in the analysis
110
were species, breed, age and a combined gender-neutering variable that consisted of male-
111
entire, male-castrated, female-entire and female-neutered categories. Log-odds diarrhoea did
112
not show a clear linear association with age and, therefore, the continuous age variable was
113
categorised as young (< 1 year old), adult (1 to < 8 years old) or aged (≥ 8 years old).
114
Page 5 Page 5 of 21
115
Due to the very limited number of explanatory variables available, automated,
116
forwards and/or backwards stepwise algorithms for variable selection were not considered to
117
be appropriate and a more empirical approach was adopted. Based on a combination of
118
statistical and biological considerations, species and age were found to be the most important
119
explanatory variables. The effects of including additional variables or interaction terms were
120
assessed using likelihood ratio (LR) tests. The odds of animals presenting with diarrhoea
121
varied between practices and, therefore, following LR tests, the parameters of reported
122
models were recalculated using robust standard errors to account for intragroup correlation
123
within practices. The effects of each stage of the model building process are described in the
124
results. When testing many between-group comparisons using a single logistic regression
125
model, the overall Type I error (α) was controlled using the Bonferroni adjustment.
126 127
Many breeds were represented in the dataset. In cats, the vast majority of animals
128
were characterised as ‘domestic short-haired’, an unofficial breed that was most likely
129
applied indiscriminately to many cats, thereby limiting useful analysis. Breed was recorded
130
more reliably in dogs but many breeds and breed crosses were represented by only a few
131
individuals, limiting analysis of the whole dataset for breed associations with canine
132
diarrhoea. However, univariable logistic regression analysis was performed using a restricted
133
dataset consisting of breeds where there had been 10 or more cases of diarrhoea in adult
134
animals (≥ 1 year old).
135 136
Cases of diarrhoea were classified as either complicated (diarrhoea was haemorrhagic
137
and/or accompanied by vomiting) or uncomplicated. The time taken from the onset of clinical
138
signs to owners presenting sick animals to veterinary practices was recorded for each case.
139
For dogs and cats, the trend of odds of uncomplicated diarrhoea over each time category was
Page 6 Page 6 of 21
140
calculated using methods based on score statistics; the homogeneity of odds of haemorrhagic
141
diarrhoea across different gender-neuter status categories and, in dogs, across different
142
breeds, were similarly tested (StataCorp, 2007b). Descriptive statistics are presented to
143
illustrate the similarities and differences between diagnostic tests requested and the
144
treatments employed for cases of diarrhoea in dogs and cats. The comparison between dogs
145
and cats were made using univariate logistic regression.
146 147
Data were analysed using commercially available software (Excel, Microsoft and
148
Stata 10 IC, StataCorp, 2007). All proportions and logistic regression models were calculated
149
to allow for clustering within veterinary practices. In the case of large samples, confidence
150
intervals (CIs) were calculated directly, assuming a normal distribution of sample
151
proportions. When only small numbers were involved, CIs were estimated from logistic
152
regression models to avoid issues of error bars extending below zero or above 1. In all
153
analyses, statistical significance was defined as P < 0.05.
154 155
Results
156
Study sample
157
Individual consultation records (including repeated consultations for the same animal;
158
n=16,223) were recorded in the database in response to the ‘diarrhoea and vomiting’
159
questionnaire. Of these, 11,060 consultations (68%) were from dogs, 4,092 (25%) from cats,
160
387 (2%) from rabbits, 164 (1%) from guinea pigs, 416 (2.6%) from other species and 104
161
consultations where the species was not noted (Fig. 2a). Presentation for diarrhoea comprised
162
6% of canine consultations, 3% of feline consultations, 2% of rabbit consultations and 4% of
163
guinea pig consultations (Fig. 2b). Subsequently, data were analysed on a single visit per
164
animal basis, providing 9,115 and 3,462 unique records for dogs and cats, respectively.
Page 7 Page 7 of 21
165 166
As described, the odds of dogs presenting with diarrhoea were significantly different
167
between practices (likelihood-ratio χ2df=18=69.79; P < 0.001); the proportion ranged from 3-
168
13% (Fig. 3). In cats, the differences between practices were not statistically significant
169
(likelihood-ratio χ2df=18 = 14.73; P = 0.680; data not shown).
170 171
Species
172
On a single visit per animal basis, 7% of dogs and 3% of cats presented with
173
diarrhoea on at least one occasion. On univariate analysis, dogs were 2.2 (95% CI 1.8–2.6)
174
times more likely to present with diarrhoea than cats (P < 0.001; Table 1).
175 176 177
Age distribution The proportion of dogs and cats presenting with diarrhoea in each of the age
178
categories is shown in Table 1 and Fig. 4. The inclusion of age (and an interaction term)
179
significantly improved the fit of a model over one that contained species only. Species was
180
not associated with an increase or decrease in presentation with diarrhoea in puppies and
181
kittens (OR = 1.2, 95% CI = 0.7-1.9, P = 0.470). However, adult and aged dogs were 3.2
182
(95% CI 2.3-4.5) and 2.3 (95% CI 1.8-2.9) times more likely, respectively, to present with
183
diarrhoea than cats of similar ages (Table 1). In both dogs and cats, adult animals were
184
significantly less likely to present with diarrhoea than young animals (dogs, ORadult vs. young =
185
0.6, 95% CI = 0.5–0.7, P Bonferroni < 0.001; ORaged vs. young = 0.5, 95% CI = 0.4–0.6, P Bonferroni <
186
0.001; cats, ORadult vs. young = 0.2, 95% CI = 0.1–0.4, PBonferroni < 0.001, ORaged vs. young = 0.3,
187
95% CI = 0.2–0.4, PBonferroni < 0.001). In dogs and cats, no difference was observed in the
188
presentation with diarrhoea in aged and adult animals as calculated by the OR and 95% CI
189
(ORdogs = 0.9, 95% CI = 0.8–1.0, PBonferroni = 0.048; ORcats = 1.2, 95% CI = 0.8–2.0, P =
Page 8 Page 8 of 21
190
1.000; see Fig. 4).
191 192
Gender and neutering
193
The inclusion of the sex-neutering status variable (together with interaction terms)
194
significantly improved the fit of the model compared to one that contained species only (P <
195
0.001). However, the inclusion of the gender-neutering variable (main effects only) did not
196
significantly improve the fit of the model over one that contained age and species (P =
197
0.402), although including all two- and three-way interaction terms did have a significant
198
effect (P = 0.025). In order to facilitate the interpretation of multiple interaction terms, further
199
analyses were conducted in dogs and cats separately. In dogs, the inclusion of gender-
200
neutering status (plus interactions) significantly improved the fit of the model that contained
201
age only (P = 0.025). Young, neutered males had a significantly reduced odds of disease
202
compared with entire male animals of the same age (PBonferroni < 0.001); there were no other
203
significant associations. In cats, gender-neutering status was not associated with diarrhoea
204
when age was included in the model (P = 0.189).
205 206
Breed
207
Breed was significantly associated with presentation for diarrhoea in adult dogs. The
208
PMD (± 95% CI) in adult/aged dogs (≥1 year) is shown in Fig. 5. Miniature schnauzers had
209
the highest PMD (19%, 95% CI 5-33) whilst West Highland white terriers had the lowest
210
PMD (3%, 95% CI, 1-6).
211 212 213 214
Duration of illness and presence of complicating factors Most animals with diarrhoea (51%) presented with a history of illness over the previous 2-4 days. The presence of haemorrhagic diarrhoea and/or concurrent vomiting
Page 9 Page 9 of 21
215
(‘complicated diarrhoea’) occurred in 43% of cases (12% had vomiting, 25% had
216
haemorrhagic diarrhoea and 6% had both vomiting and haemorrhagic diarrhoea). In dogs, the
217
odds of uncomplicated diarrhoea showed a highly significant (P < 0.001) increasing trend
218
across categories of duration of illness (i.e. uncomplicated diarrhoea was more common in
219
dogs presenting later in the course of clinical signs; Fig.6). A similar pattern was seen in cats,
220
but the numbers involved were smaller and the association was less pronounced (P = 0.041).
221 222
The tests of homogeneity of odds of haemorrhagic diarrhoea across breeds and
223
gender-neuter status categories in dogs >1 year were not statistically significant (P = 0.658
224
and P = 0.218, respectively). Similarly, the test for homogeneity of odds of haemorrhagic
225
diarrhoea across age categories was not statistically significant (P = 0.983).
226 227
Diagnostic tests performed
228
Diagnostic tests were performed in 16% of cases with diarrhoea (118/729; 15% of
229
dogs; 24% of cats). Haematology/serum biochemistry, parasitology and bacteriology were
230
the most common tests used in 7% dogs and 15% cats, 7% dogs and 8% of cats and 6% of
231
dogs and 7% of cats, respectively (Fig. 7). Other testing modalities included diagnostic
232
imaging, GI function tests and biopsy.
233 234
Treatment of diarrhoea
235
Dietary modification was used in the majority of cases of diarrhoea (66% of dogs;
236
63% cats), as were antibacterials (63% dogs; 49% cats). Intravenous fluids and oral fluids
237
were used relatively infrequently (4% and 1%, respectively, for dogs; 5% and 0%,
238
respectively, for cats; Fig. 8a). In dogs and cats, the probability of using antibacterials was
239
increased in cases where haemorrhagic diarrhoea had been noted (ORDogs 3.7, 95% CI 2.0-
Page 10 Page 10 of 21
240
6.9, P < 0.001; ORCats 3.2, 95% CI 1.2-8.2, P = 0.018; Fig. 8b). Potentiated amoxicillin,
241
metronidazole and amoxicillin were the most commonly chosen antibacterials (Fig. 8c). Use
242
of antibacterials did not change the probability that a dietary modification would be used.
243 244 245
Discussion This study demonstrates the frequency of GI disease as a presenting complaint in
246
small animal practice using proportional morbidity, a measure that has previously been used
247
in circumstances where a population denominator data has not been available. Proportional
248
morbidity, as an epidemiological measure, has the limitation in that the denominator is
249
affected by the prevalence of other diseases. In addition, the ratio of proportional morbidity
250
for groups that have either been exposed or unexposed to a given risk factor can be
251
interpreted as the relative risk only when the total overall morbidity in exposed and
252
unexposed groups is equal (Miettinen and Wang, 1981). Consequently, an increase in the
253
morbidity rate for the disease of interest would need to be accompanied by an equivalent
254
decrease in morbidity for all other causes for the relationship to hold, a condition that is
255
unlikely to be realised in practice. Alternatively, however, the morbidity OR can be used as
256
an estimate of relative risk providing the morbidity rate of all other causes is equal in both
257
exposed and unexposed groups – a condition that is much more likely to occur in real-life
258
situations (Miettinen and Wang, 1981). In the current study, the significant associations of
259
risk factors were assessed using morbidity ORs.
260 261
The PMD was in broad agreement with that seen in other UK-based studies. Stavisky
262
et al. (2010) reported that 16/186 (8.60%; 95%CI 5.00–13.59) randomly selected dogs that
263
presented to veterinary practices in the UK (not including vaccination consultations) showed
264
signs of diarrhoea whilst German et al. (2010) reported that 2,058 of approximately 40,000
Page 11 Page 11 of 21
265
(5.15%; 95%CI 4.93–5.37) dogs seen at a referral hospital had originally presented with
266
diarrhoea. However, the PMD was different from that reported in vet-visiting dogs in the
267
United States. Lund et al. (1999) estimated that the prevalence of diarrhoea in the vet-visiting
268
populations of dogs and cats was 2.2% (95%CI 2.0–2.4) and 1.8% (95%CI 1.6–2.0),
269
respectively. Based on the information presented by Lund et al. (1999), the PMD can be
270
estimated by removing vaccination and pre-operative checks from the denominators and is
271
calculated to be 2.37% (95%CI 2.20–2.55) and 2.74% (95%CI 2.43–3.08) for dogs and cats,
272
respectively. The difference in PMD between the UK and the US does not necessarily reflect
273
differences in the prevalence of diarrhoea in the two countries, but could instead indicate that
274
animals in the US are presented more frequently to veterinary practices for conditions other
275
than diarrhoea.
276 277
In both dogs and cats, the PMD was significantly higher in animals <1 year of age,
278
potentially reflecting a higher incidence of infectious diarrhoea in younger animals
279
(Batchelor et al., 2008), or a greater tendency to present young animals with diarrhoea to a
280
veterinary surgeon. Our study is in agreement with findings in a previous study evaluating the
281
frequency of diarrhoea in four selected breeds of dogs at different age intervals (Saevik et al.,
282
2012). Adult and aged cats were significantly less likely to be presented to a veterinary
283
surgeon for diarrhoea than dogs of equivalent ages.
284 285
While causes of diarrhoea in cats and dogs have broad similarities (Cooper, 2011;
286
Hall, 2009), the difference in PMD could reflect key species differences. More specific
287
information about these differences could be more readily analysed by collecting a larger
288
sample of consultations with more follow-through of individual animals, thereby allowing
289
further assessment of disease characteristics including overall duration, response to therapy
Page 12 Page 12 of 21
290
and the number of repeat visits. It is also possible that the low PMD in cats might be because
291
the faeces of adult/aged cats are often unobserved. As a consequence, diarrhoea that is not
292
accompanied by other clinical signs (such as weight loss, malaise or inappropriate
293
defaecation in the home) could go unnoticed.
294 295
Neutering appeared to be protective for diarrhoea in young (< 1 year) male dogs only.
296
The explanation for this association is not immediately obvious, but it could reflect
297
differences in the attitudes of the owners of these dogs towards presenting their pets to
298
veterinary practices, rather than a true reduction in the prevalence of diarrhoea in this
299
subgroup of animals.
300 301
Analysing the data by individual breeds produced small sub-populations. Despite this,
302
some interesting patterns began to emerge, highlighting the potential utility of this data to
303
identify breed-related disease risk factors. For example, Yorkshire terriers >1 year of age had
304
one of the highest PMD for diarrhoea among adult dogs. Protein-losing enteropathy has been
305
reported in this breed (Waddell and Michel, 2000) and might be a severe manifestation of the
306
predisposition to diarrhoea suggested here.
307 308
National and regional statistics for breed numbers would have been useful to provide
309
a denominator for breed presentation in this study. National statistics of pet ownership are
310
available from the Pet Food Manufacturers Association,3 however, these would not allow for
311
partial national coverage of veterinary practice surveillance or for local variations in disease
312
prevalence. Overall, comparatively small numbers of breed-specific predispositions were
3
See PFMA, 2012. Pet Food Manufacturers Association: Dogs by breed. http://www.pfma.org.uk/dogs-by-breed/ (accessed 29 May 2014).
Page 13 Page 13 of 21
313
identified. With ongoing data collection through SAVSNET, substantially larger datasets will
314
become available, allowing for higher resolution analysis of the breed and age associations
315
that were hinted at in the current study.
316 317
Dietary modification was the most common therapy consistent with the accepted
318
position that the majority of cases of diarrhoea presented to small animal practice reflect
319
either dietary indiscretion (Stavisky et al., 2011), or diet-responsive disease. Antibacterials
320
were frequently used to treat diarrhoea, in keeping with previous findings that diarrhoea
321
increased the probability that antibacterials would be used in veterinary consultations
322
(Radford et al., 2011) and were the most common treatment used in dogs with diarrhoea prior
323
to referral (German et al., 2010).
324 325
A number of infectious agents are recognised to cause diarrhoea (Hackett and Lappin
326
2003; Parsons et al., 2010) but, in the current study, confirmation of infection using faecal
327
sampling was only performed in 6% of cases. Antibiotic-responsive diarrhoea is recognised
328
(Hall, 2011; Hostutler et al., 2004; Westermarck et al., 2005) but represents a minority of
329
cases. Enteropathogens are often present in faeces (Hackett and Lappin 2003; Marks et al.,
330
2011; Stavisky et al., 2011) but these are often found in the absence of diarrhoea (Westgarth
331
et al., 2009). The use of antibacterials may reflect a perception that animals with diarrhoea
332
often lose mucosal barrier function and our study showed increased usage when haemorrhage
333
was present. However, in the absence of sepsis, there is no evidence that outcomes are
334
affected by use of antibacterials in cases of haemorrhagic diarrhoea (Unterer et al., 2011).
335
Thus, the frequent use of antibacterials might reflect our incomplete understanding of the role
336
of suspected pathogens in the pathogenesis of diarrhoea.
337
Page 14 Page 14 of 21
338
SAVSNET has undertaken to communicate individual practice statistics to facilitate
339
benchmarking of prescribing behaviour in relation to anonymised peers. When combined
340
with expert opinion and longer-term follow-up of cases through SAVSNET, such
341
benchmarking might help practitioners review practice protocols and clinical decision-
342
making.
343 344
The practices recruited to the SAVSNET pilot project were widely distributed around
345
England and Wales. Consequently, variations in PMD in, for example, dogs from different
346
practices, could have reflected differences in diagnostic procedures among veterinary
347
practices, or might have indicated true geographical differences in disease prevalence. While
348
variations in PMD between different practices could provide a surrogate for geographical
349
variations in PMD in this pilot study, SAVSNET data also includes full postcode
350
information, which will allow true geographical mapping of disease as the project expands.
351 352
Ultimately, it will be valuable to establish baseline data that characterise the
353
occurrence of GI disease in companion animals. The approach taken by SAVSNET requires
354
the completion of short questionnaires by participating veterinary surgeons at the end of each
355
consultation. Questionnaire responses, together with signalment data and clinical notes, are
356
collated and stored on SAVSNET servers. SAVSNET will also have access to laboratory data
357
regarding the diagnosis of, for instance, infectious diarrhoea. This multi-component approach
358
allows for cross-correlation of data between different sources and parallels the approach
359
taken in large studies in human medicine (O'Brien et al., 2010).
360 361 362
Conclusions SAVSNET allows the collection of ethically-approved surveillance data in real-time
Page 15 Page 15 of 21
363
from veterinary practices. Using limited data from pilot studies, this study demonstrated the
364
ability to identify age and breed variation in PMD as well as inter-practice variation in PMD.
365
Factors were identified that could potentially influence clinical decision-making in practice.
366
Combined with expert opinion, data so gathered could be used to benchmark clinical
367
approaches in participating practices and identify targets for veterinary education and
368
research.
369 370 371 372
Conflict of interest statement None of the authors has any financial or personal relationships that could inappropriately influence or bias the content of the paper.
373 374 375
Acknowledgements During the study period (September 2008 - August 2011), SAVSNET was funded by
376
a consortium including Dechra, Defra, IntervetSP, Merial, Novartis, Pfizer, University of
377
Liverpool and Virbac, and was additionally supported by BSAVA. We are extremely grateful
378
to the 19 practices that participated in this study, and to our colleagues at Vet Solutions
379
without whose dedicated help and support, collection of these data would not have been
380
possible.
381 382
References
383 384 385 386 387 388 389 390 391 392 393
Batchelor, D.J., Tzannes, S., Graham, P.A., Wastling, J.M., Pinchbeck, G.L., German, A.J., 2008. Detection of endoparasites with zoonotic potential in dogs with gastrointestinal disease in the UK. Transboundary and Emerging Diseases 55, 99–104. Cooper, S., 2011. Recurrent diarrhoea in cats. In Practice 33, 272–281. Feldman, R.A., Banatvala, N., 1994. The frequency of culturing stools from adults with diarrhoea in Great Britain. Epidemiology and Infection 113, 41–44. German, A.J., Halladay, L.J., Noble, P.J.M., 2010. First-choice therapy for dogs presenting with diarrhoea in clinical practice. Veterinary Record 167, 810–814. Page 16 Page 16 of 21
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
Hackett, T., Lappin, M.R., 2003. Prevalence of enteric pathogens in dogs of north-central Colorado. Journal of the American Animal Hospital Association 39, 52–56. Hall, E., 2009. Canine diarrhoea: a rational approach to diagnostic and therapeutic dilemmas. In Practice 31, 8–16. Hall, E.J., 2011. Antibiotic-responsive diarrhea in small animals. Veterinary Clinics of North America: Small Animal Practice 41, 273–286. Hostutler, R., Luria, B., Johnson, S., Weisbrode, S., Sherding, R., Jaeger, J., Guilford, W., 2004. Antibiotic-responsive histiocytic ulcerative colitis in 9 dogs. Journal of Veterinary Internal Medicine 18, 499–504. Hubbard, K., Skelly, B.J., McKelvie, J., Wood, J.L.N., 2007. Risk of vomiting and diarrhoea in dogs. Veterinary Record 161, 755–757. Lund, E.M., Armstrong, P.J., Kirk, C.A., Kolar, L.M., Klausner, J.S., 1999. Health status and population characteristics of dogs and cats examined at private veterinary practices in the United States. Journal of the American Veterinary Medical Association 214, 1336–1341. Marks, S.L., Rankin, S.C., Byrne, B.A., Weese, J.S., 2011. Enteropathogenic bacteria in dogs and cats: diagnosis, epidemiology, treatment, and control. Journal of Veterinary Internal Medicine 25, 1195–1208. Martin, S.W., Meek, A.H., Willeberg, P., 1987. Veterinary Epidemiology: Principles and Methods. Iowa State Press, Ames, Iowa, USA, p75. Miettinen, O.S. and Wang, J-D., 1981. An alternative to the proportionate mortality ratio. American Journal of Epidemiology 114, 144-148. O'Brien, S.J., Rait, G., Hunter, P.R., Gray, J.J., Bolton, F.J., Tompkins, D.S., McLauchlin, J., et al., 2010. Methods for determining disease burden and calibrating national surveillance data in the United Kingdom: the second study of infectious intestinal disease in the community (IID2 study). BMC Medical Research Methodology 10, 39. Parsons, B.N., Porter, C.J., Ryvar, R., Stavisky, J., Williams, N.J., Pinchbeck, G.L., Birtles, R.J., et al., 2010. Prevalence of Campylobacter spp. in a cross-sectional study of dogs attending veterinary practices in the UK and risk indicators associated with shedding. The Veterinary Journal 184, 66–70. Radford, A.D., Noble, P.-J., Coyne, K.P., Gaskell, R.M., Jones, P.H., Bryan, J.G.E., Setzkorn, C., Tierney, A., Dawson, S., 2011. Antibacterial prescribing patterns in small animal veterinary practice identified via SAVSNET: the small animal veterinary surveillance network. Veterinary Record 169, 310–U91. Saevik, B.K., Skancke, E.M., Trangerud, C., 2012. A longitudinal study on diarrhoea and vomiting in young dogs of four large breeds. Acta Veterinaria Scandinavica 54, 8. Smith, S., Elliot, A.J., Mallaghan, C., Modha, D., Hippisley-Cox, J., Large, S., Regan, M., Page 17 Page 17 of 21
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
Smith, G.E., 2010. Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire, United Kingdom, June - July 2008. Euro surveillance: Bulletin Européen sur les Maladies Transmissibles. European Communicable Disease Bulletin 15, 19643. StataCorp. 2007. Stata Statistical Software: Release 10. StataCorp LP, College Station, TX, USA. StataCorp. 2007b. Stata 10 Survival Analysis and Epidemiological Tables. Stata Press, College Station, TX, USA, p75. Stavisky, J., Pinchbeck, G.L., German, A.J., Dawson, S., Gaskell, R.M., Ryvar, R., Radford, A.D., 2010. Prevalence of canine enteric coronavirus in a cross-sectional survey of dogs presenting at veterinary practices. Veterinary Microbiology 140, 18-24. Stavisky, J., Radford, A.D., Gaskell, R., Dawson, S., German, A., Parsons, B., Clegg, S., Newman, J., Pinchbeck, G., 2011. A case-control study of pathogen and lifestyle risk factors for diarrhoea in dogs. Preventive Veterinary Medicine 99, 185–192. Unterer, S., Strohmeyer, K., Kruse, B.D., Sauter-Louis, C., Hartmann, K., 2011. Treatment of aseptic dogs with hemorrhagic gastroenteritis with amoxicillin/clavulanic acid: a prospective blinded study. Journal of Veterinary Internal Medicine 25, 973–979. Waddell, L., Michel, K., 2000. Hypomagnesemia and hypocalcemia associated with proteinlosing enteropathy in Yorkshire Terriers: five cases (1992-1998). Journal of the American Veterinary Medical Association 217, 703–706. Westermarck, E., Skrzypczak, T., Harmoinen, J., Steiner, J., Ruaux, C., Williams, D., Eerola, E., Sundback, P., Rinkinen, M., 2005. Tylosin-responsive chronic diarrhea in dogs. Journal of Veterinary Internal Medicine 19, 177-186. Westgarth, C., Porter, C.J., Nicolson, L., Birtles, R.J., Williams, N.J., Hart, C.A., Pinchbeck, G.L., Gaskell, R.M., Christley, R.M., Dawson, S., 2009. Risk factors for the carriage of Campylobacter upsaliensis by dogs in a community in Cheshire. Veterinary Record 165, 526–530.
Page 18 Page 18 of 21
480
Table 1
481
Species associations with proportional morbidity of diarrhoea (PMD) stratified by age category
482 Age category Young (< 1 year) Adult (1-7.99 years) Aged (≥ 8 years) All ages 483 484 485 486
Cat
33
326
PMD (95% CI) a 0.092 (0.058, 0.13)
Dog
109
900
0.108 (0.087, 0.129)
1.2 (0.6, 1.8)
0.508
Cat
26
1166
0.022 (0.014, 0.030)
1.00
–
Dog
293
4066
0.067 (0.056, 0.079)
3.2 (2.2, 4.3)
0.000
Cat
49
1772
0.027 (0.020, 0.034)
1.00
–
Dog
219
3458
0.060 (0.050, 0.069)
2.3 (1.8, 2.8)
0.000
Cat Dog
108 621
3264 8424
0.032 (0.026, 0.038) 0.069 (0.059, 0.079)
1.0 2.2 (1.8,2.6)
0.000
Species Cases Controls
Odds ratio (95% CI) a 1.0
P –
95% CI, 95% Confidence interval a Confidence intervals calculated to allow for intragroup correlation within veterinary practice.
Page 19 Page 19 of 21
487
Figure legends
488 489
Fig. 1. The sequence of questions presented to the veterinary surgeon on completion of the
490
consultation is shown. * Indicates questions where multiple answer options were allowed.
491 492
Fig. 2. The study population. (a) Pie chart with segments representing proportions of different
493
species in the population questioned about diarrhoea. (b) Percentage of cases with diarrhoea
494
by species. Bars represent percentage of individuals presenting with diarrhoea for each
495
species (only species occurring in ≥100 consultations are shown).
496 497
Fig. 3. Proportional morbidity of diarrhoea (PMD) in dogs at individual veterinary practice
498
branches (anonymised). Bars represent the PMD for dogs presenting at individual practice
499
branches where the branch contributed >100 records to the study.
500 501
Fig. 4. The proportional morbidity of diarrhoea (PMD) in dogs and cats, segregated by age
502
group. Bars represent PMD of diarrhoea in dogs (grey bars) and cats (white bars) in indicated
503
age groups (± 95% confidence interval). Practice-adjusted odds ratios for groups with
504
common labels (a, b or c) were not significantly different (Bonferroni-adjusted P > 0.05).
505 506
Fig. 5. Breeds with diarrhoea. Proportional morbidity of diarrhoea (PMD) in individual
507
breeds is shown for adult/aged dogs (≥1 year). Bars represent PMD (± 95% confidence
508
intervals) for each breed. For the purposes of this study, the term ‘cross’ refers to a non-
509
pedigree animal where the breeds of one or more of the parents are recognizable and have
510
been recorded in the animal’s record. The term ‘crossbreed’ refers to a non-pedigree animal
511
where the breeds of the parents have not been recorded.
512 Page 20 Page 20 of 21
513
Fig. 6. The proportions of dogs and cats with uncomplicated diarrhoea presenting after given
514
durations of illness. Bars represent cases classed as uncomplicated (no vomiting or
515
haemorrhage) as a proportion of all cases of diarrhoea with a given duration of illness prior to
516
presentation. Strata representing 5-7 days and ≥8 days have been collapsed into a single
517
category to ensure sufficient sample size at each level. Grey bars represent dogs and white
518
bars represent cats.
519 520
Fig. 7. Diagnostic tests performed in cases of diarrhoea. Bars represent the proportion of
521
cases for which the given diagnostic test was performed. Grey bars represent dogs and white
522
bars represent cats. TLI, trypsin-like immunoreactivity; PLI, pancreatic lipase
523
immunoreactivity.
524 525
Fig. 8. Treatment choices in diarrhoea. (a) Frequency of use of different treatment classes.
526
Bars represent proportions of cases receiving a given treatment. (b) Frequency of use of
527
antibacterials in cases presenting with haemorrhagic and non-haemorrhagic diarrhoea, bars
528
represent proportion of cases (± 95% confidence intervals) treated with antibacterials (*
529
indicates proportion different to ‘no’ group, P < 0.05). (c) Antibacterials used in treating
530
diarrhoea. Bars represent the proportion of diarrhoea cases that were treated with
531
antibacterials that received each drug type. Grey bars represent dogs and white bars represent
532
cats.
Page 21 Page 21 of 21
S1090-0233(14)00241-X http://dx.doi.org/doi:10.1016/j.tvjl.2014.05.044 YTVJL 4180
To appear in:
The Veterinary Journal
Accepted date:
31-5-2014
Please cite this article as: P.H. Jones, S. Dawson, R.M. Gaskell, K.P. Coyne, Á. Tierney, C. Setzkorn, A.D. Radford, P-J.M. Noble, Surveillance of diarrhoea in small animal practice through the Small Animal Veterinary Surveillance Network (SAVSNET), The Veterinary Journal (2014), http://dx.doi.org/doi:10.1016/j.tvjl.2014.05.044. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Surveillance of diarrhoea in small animal practice through the Small Animal Veterinary Surveillance Network (SAVSNET) P.H. Jones a, c, S. Dawson b, R.M. Gaskell a, K.P. Coyne a, Á. Tierney a, C. Setzkorn a, A.D. Radford a, P-J.M. Noble b, * a
University of Liverpool, Institute of Global Health, Leahurst Campus, Chester High Road, Neston, Cheshire, CH64 7TE, UK b University of Liverpool School of Veterinary Science, Leahurst Campus, Chester High Road, Neston, Cheshire, CH64 7TE, UK c National Consortium for Zoonosis Research, Leahurst Campus, Chester High Road, Neston, Cheshire, CH64 7TE, UK
* Corresponding author. Tel.: +44 151 795 6205. E-mail address: [email protected] (P-J.M. Noble).
Page 1 Page 1 of 21
20 21
Abstract Using the Small Animal Veterinary Surveillance Network (SAVSNET), a national
22
small animal disease-surveillance scheme, information on gastrointestinal disease was
23
collected for a total of 76 days between 10 May 2010 and 8 August 2011 from 16,223
24
consultations (including data from 9,115 individual dogs and 3,462 individual cats) from 42
25
premises belonging to 19 UK veterinary practices. During that period, 7% of dogs and 3% of
26
cats presented with diarrhoea.
27 28
Adult dogs had a higher proportional morbidity of diarrhoea (PMD) than adult cats (P
29
< 0.001). This difference was not observed in animals < 1 year old. Younger animals in both
30
species had higher PMDs than adult animals (P < 0.001). Neutering was associated with
31
reduced PMD in young male dogs. In adult dogs, miniature Schnauzers had the highest PMD.
32
Most animals with diarrhoea (51%) presented having been ill for 2-4 days, but a history of
33
vomiting or haemorrhagic diarrhoea was associated with a shorter time to presentation. The
34
most common treatments employed were dietary modification (66% of dogs; 63% of cats)
35
and antibacterials (63% of dogs; 49% of cats). There was variability in PMD between
36
different practices.
37 38
The SAVNET methodology facilitates rapid collection of cross-sectional data
39
regarding diarrhoea, a recognised sentinel for infectious disease, and characterises data that
40
could benchmark clinical practice and support the development of evidence-based medicine.
41 42
Keywords: Breed; Companion animal; Diarrhoea; Surveillance; SAVNET
43
Page 2 Page 2 of 21
44 45
Introduction Gastrointestinal (GI) disease commonly results in the presentation of pets to UK
46
veterinary surgeons, but few national statistics record the frequency or the diagnostic and
47
therapeutic approach to these cases. A study of dogs presented to veterinary practices in the
48
USA suggested that 2.2% of veterinary consultations were related to diarrhoea (Lund et al.,
49
1999). In UK, a survey using client questionnaires reported that up to 20% of dogs had mild
50
vomiting and up to 15% had mild diarrhoea over a 2-week period (Hubbard et al., 2007), and
51
using data from notes provided with referral cases, German et al. (2010) highlighted the high
52
levels of antibacterial drugs used to treat GI disease. However, that study lacked
53
denominators to set it in the context of all animals with GI disease and was based on a
54
comparatively small sample with low statistical power.
55 56
Radford et al. (2011) showed that the presence of GI disease increased the probability
57
that a veterinary surgeon would prescribe antibacterials in first opinion practice, although the
58
prevalence of diarrhoea was not reported. In comparison, data on human disease are much
59
more detailed, with studies of far larger populations coordinated by National Health Service
60
recording and surveillance systems alongside national auditing (O'Brien et al., 2010; Smith et
61
al., 2010).
62 63
It is clear that a more coordinated approach to diarrhoea surveillance in companion
64
animals is needed. As well as quantifying the disease burden, such an approach could identify
65
risk factors for disease susceptibility as well as determining outcome measures associated
66
with specific diagnostic and therapeutic approaches, a prerequisite for the development of
67
evidence-based medicine. Changes in the incidence of diarrhoea could act as a sentinel for
68
infectious disease outbreaks (Smith et al., 2010) and warn of the emergence of new GI
Page 3 Page 3 of 21
69
pathogens.
70 Disease surveillance schemes exist for livestock 1 and horses.2 The Small Animal
71 72
Veterinary Surveillance Network (SAVSNET) monitors disease in small animals attending
73
first opinion practice, using data collected from veterinary laboratories and near real-time
74
collection of consultation records from participating veterinary practices. In this novel study,
75
we used data gathered during pilot studies to establish the feasibility of SAVSNET
76
methodologies to profile the presentation, diagnostic approach and management choices for
77
dogs and cats presenting with diarrhoea to small animal practices in the UK.
78 79
Materials and methods
80
Data collection
81
Data were collected from practices using a compatible version of practice
82
management software (Premvet, Vetsolutions, v03.02.12) following a positive response to a
83
postal request. Seventy-four practices were approached, recruiting 16/59, 3/7, 0/6 and 0/2
84
practices in England, Wales, Scotland and Northern Ireland, respectively (in total 19 practices
85
comprising 42 premises). Data on GI disease were collected over a total of 76 days between
86
10 May 2010 and 8 August 2011. Data were only collected from consultations relating to sick
87
animals, and excluded vaccine consultations.
88 89
At the end of each consultation, the veterinary surgeon was asked whether the case
1
See: AHVLA, 2012. Veterinary Laboratories Agency: Veterinary Investigation Surveillance Report. http://www.defra.gov.uk/ahvla-en/publication/vida12/ (accessed 29 May 2014).
2
See: AHT, 2012.Animal Health Trust. DEFRA/AHT/BEVA Equine Quarterly Disease Surveillance Reports. Animal Health Trust. http://www.aht.org.uk/cmsdisplay/disease_surveillance.html (accessed 29 May 2014).
Page 4 Page 4 of 21
90
had presented for vomiting or diarrhoea. If the answer to this question was ‘no’, the
91
questionnaire terminated; if ‘yes’, the questionnaire was completed as shown (Fig. 1). The
92
questionnaire responses, signalment and demographic data and the free text consultation
93
record were collected and stored in the SAVSNET database. Data were excluded if the client
94
had opted out of study participation.
95 96 97
Statistical analysis Multiple visits for individual animals were not included in the analysis. For animals
98
that never presented with diarrhoea, data from the first consultation only were used. For
99
animals that presented with diarrhoea, only data from the first consultation for diarrhoea were
100
selected. Thus, the proportions of cases of diarrhoea approximated the proportional morbidity
101
of diarrhoea (PMD; Martin et al., 1987), where the total number of diseased animals was
102
approximated by the total number of animals presenting to participating veterinary practices
103
for sick animal consultations.
104 105
Univariable and multivariable logistic regression was used to model the presentation
106
of diarrhoea (as a binary dependent variable) and the resulting models were used to estimate
107
morbidity odds ratios (ORs), a statistic that can be interpreted as a relative risk on the
108
assumption that the morbidity rate for all other causes was unrelated to exposure to the risk
109
factor (Miettinen and Wang, 1981). The explanatory variables considered in the analysis
110
were species, breed, age and a combined gender-neutering variable that consisted of male-
111
entire, male-castrated, female-entire and female-neutered categories. Log-odds diarrhoea did
112
not show a clear linear association with age and, therefore, the continuous age variable was
113
categorised as young (< 1 year old), adult (1 to < 8 years old) or aged (≥ 8 years old).
114
Page 5 Page 5 of 21
115
Due to the very limited number of explanatory variables available, automated,
116
forwards and/or backwards stepwise algorithms for variable selection were not considered to
117
be appropriate and a more empirical approach was adopted. Based on a combination of
118
statistical and biological considerations, species and age were found to be the most important
119
explanatory variables. The effects of including additional variables or interaction terms were
120
assessed using likelihood ratio (LR) tests. The odds of animals presenting with diarrhoea
121
varied between practices and, therefore, following LR tests, the parameters of reported
122
models were recalculated using robust standard errors to account for intragroup correlation
123
within practices. The effects of each stage of the model building process are described in the
124
results. When testing many between-group comparisons using a single logistic regression
125
model, the overall Type I error (α) was controlled using the Bonferroni adjustment.
126 127
Many breeds were represented in the dataset. In cats, the vast majority of animals
128
were characterised as ‘domestic short-haired’, an unofficial breed that was most likely
129
applied indiscriminately to many cats, thereby limiting useful analysis. Breed was recorded
130
more reliably in dogs but many breeds and breed crosses were represented by only a few
131
individuals, limiting analysis of the whole dataset for breed associations with canine
132
diarrhoea. However, univariable logistic regression analysis was performed using a restricted
133
dataset consisting of breeds where there had been 10 or more cases of diarrhoea in adult
134
animals (≥ 1 year old).
135 136
Cases of diarrhoea were classified as either complicated (diarrhoea was haemorrhagic
137
and/or accompanied by vomiting) or uncomplicated. The time taken from the onset of clinical
138
signs to owners presenting sick animals to veterinary practices was recorded for each case.
139
For dogs and cats, the trend of odds of uncomplicated diarrhoea over each time category was
Page 6 Page 6 of 21
140
calculated using methods based on score statistics; the homogeneity of odds of haemorrhagic
141
diarrhoea across different gender-neuter status categories and, in dogs, across different
142
breeds, were similarly tested (StataCorp, 2007b). Descriptive statistics are presented to
143
illustrate the similarities and differences between diagnostic tests requested and the
144
treatments employed for cases of diarrhoea in dogs and cats. The comparison between dogs
145
and cats were made using univariate logistic regression.
146 147
Data were analysed using commercially available software (Excel, Microsoft and
148
Stata 10 IC, StataCorp, 2007). All proportions and logistic regression models were calculated
149
to allow for clustering within veterinary practices. In the case of large samples, confidence
150
intervals (CIs) were calculated directly, assuming a normal distribution of sample
151
proportions. When only small numbers were involved, CIs were estimated from logistic
152
regression models to avoid issues of error bars extending below zero or above 1. In all
153
analyses, statistical significance was defined as P < 0.05.
154 155
Results
156
Study sample
157
Individual consultation records (including repeated consultations for the same animal;
158
n=16,223) were recorded in the database in response to the ‘diarrhoea and vomiting’
159
questionnaire. Of these, 11,060 consultations (68%) were from dogs, 4,092 (25%) from cats,
160
387 (2%) from rabbits, 164 (1%) from guinea pigs, 416 (2.6%) from other species and 104
161
consultations where the species was not noted (Fig. 2a). Presentation for diarrhoea comprised
162
6% of canine consultations, 3% of feline consultations, 2% of rabbit consultations and 4% of
163
guinea pig consultations (Fig. 2b). Subsequently, data were analysed on a single visit per
164
animal basis, providing 9,115 and 3,462 unique records for dogs and cats, respectively.
Page 7 Page 7 of 21
165 166
As described, the odds of dogs presenting with diarrhoea were significantly different
167
between practices (likelihood-ratio χ2df=18=69.79; P < 0.001); the proportion ranged from 3-
168
13% (Fig. 3). In cats, the differences between practices were not statistically significant
169
(likelihood-ratio χ2df=18 = 14.73; P = 0.680; data not shown).
170 171
Species
172
On a single visit per animal basis, 7% of dogs and 3% of cats presented with
173
diarrhoea on at least one occasion. On univariate analysis, dogs were 2.2 (95% CI 1.8–2.6)
174
times more likely to present with diarrhoea than cats (P < 0.001; Table 1).
175 176 177
Age distribution The proportion of dogs and cats presenting with diarrhoea in each of the age
178
categories is shown in Table 1 and Fig. 4. The inclusion of age (and an interaction term)
179
significantly improved the fit of a model over one that contained species only. Species was
180
not associated with an increase or decrease in presentation with diarrhoea in puppies and
181
kittens (OR = 1.2, 95% CI = 0.7-1.9, P = 0.470). However, adult and aged dogs were 3.2
182
(95% CI 2.3-4.5) and 2.3 (95% CI 1.8-2.9) times more likely, respectively, to present with
183
diarrhoea than cats of similar ages (Table 1). In both dogs and cats, adult animals were
184
significantly less likely to present with diarrhoea than young animals (dogs, ORadult vs. young =
185
0.6, 95% CI = 0.5–0.7, P Bonferroni < 0.001; ORaged vs. young = 0.5, 95% CI = 0.4–0.6, P Bonferroni <
186
0.001; cats, ORadult vs. young = 0.2, 95% CI = 0.1–0.4, PBonferroni < 0.001, ORaged vs. young = 0.3,
187
95% CI = 0.2–0.4, PBonferroni < 0.001). In dogs and cats, no difference was observed in the
188
presentation with diarrhoea in aged and adult animals as calculated by the OR and 95% CI
189
(ORdogs = 0.9, 95% CI = 0.8–1.0, PBonferroni = 0.048; ORcats = 1.2, 95% CI = 0.8–2.0, P =
Page 8 Page 8 of 21
190
1.000; see Fig. 4).
191 192
Gender and neutering
193
The inclusion of the sex-neutering status variable (together with interaction terms)
194
significantly improved the fit of the model compared to one that contained species only (P <
195
0.001). However, the inclusion of the gender-neutering variable (main effects only) did not
196
significantly improve the fit of the model over one that contained age and species (P =
197
0.402), although including all two- and three-way interaction terms did have a significant
198
effect (P = 0.025). In order to facilitate the interpretation of multiple interaction terms, further
199
analyses were conducted in dogs and cats separately. In dogs, the inclusion of gender-
200
neutering status (plus interactions) significantly improved the fit of the model that contained
201
age only (P = 0.025). Young, neutered males had a significantly reduced odds of disease
202
compared with entire male animals of the same age (PBonferroni < 0.001); there were no other
203
significant associations. In cats, gender-neutering status was not associated with diarrhoea
204
when age was included in the model (P = 0.189).
205 206
Breed
207
Breed was significantly associated with presentation for diarrhoea in adult dogs. The
208
PMD (± 95% CI) in adult/aged dogs (≥1 year) is shown in Fig. 5. Miniature schnauzers had
209
the highest PMD (19%, 95% CI 5-33) whilst West Highland white terriers had the lowest
210
PMD (3%, 95% CI, 1-6).
211 212 213 214
Duration of illness and presence of complicating factors Most animals with diarrhoea (51%) presented with a history of illness over the previous 2-4 days. The presence of haemorrhagic diarrhoea and/or concurrent vomiting
Page 9 Page 9 of 21
215
(‘complicated diarrhoea’) occurred in 43% of cases (12% had vomiting, 25% had
216
haemorrhagic diarrhoea and 6% had both vomiting and haemorrhagic diarrhoea). In dogs, the
217
odds of uncomplicated diarrhoea showed a highly significant (P < 0.001) increasing trend
218
across categories of duration of illness (i.e. uncomplicated diarrhoea was more common in
219
dogs presenting later in the course of clinical signs; Fig.6). A similar pattern was seen in cats,
220
but the numbers involved were smaller and the association was less pronounced (P = 0.041).
221 222
The tests of homogeneity of odds of haemorrhagic diarrhoea across breeds and
223
gender-neuter status categories in dogs >1 year were not statistically significant (P = 0.658
224
and P = 0.218, respectively). Similarly, the test for homogeneity of odds of haemorrhagic
225
diarrhoea across age categories was not statistically significant (P = 0.983).
226 227
Diagnostic tests performed
228
Diagnostic tests were performed in 16% of cases with diarrhoea (118/729; 15% of
229
dogs; 24% of cats). Haematology/serum biochemistry, parasitology and bacteriology were
230
the most common tests used in 7% dogs and 15% cats, 7% dogs and 8% of cats and 6% of
231
dogs and 7% of cats, respectively (Fig. 7). Other testing modalities included diagnostic
232
imaging, GI function tests and biopsy.
233 234
Treatment of diarrhoea
235
Dietary modification was used in the majority of cases of diarrhoea (66% of dogs;
236
63% cats), as were antibacterials (63% dogs; 49% cats). Intravenous fluids and oral fluids
237
were used relatively infrequently (4% and 1%, respectively, for dogs; 5% and 0%,
238
respectively, for cats; Fig. 8a). In dogs and cats, the probability of using antibacterials was
239
increased in cases where haemorrhagic diarrhoea had been noted (ORDogs 3.7, 95% CI 2.0-
Page 10 Page 10 of 21
240
6.9, P < 0.001; ORCats 3.2, 95% CI 1.2-8.2, P = 0.018; Fig. 8b). Potentiated amoxicillin,
241
metronidazole and amoxicillin were the most commonly chosen antibacterials (Fig. 8c). Use
242
of antibacterials did not change the probability that a dietary modification would be used.
243 244 245
Discussion This study demonstrates the frequency of GI disease as a presenting complaint in
246
small animal practice using proportional morbidity, a measure that has previously been used
247
in circumstances where a population denominator data has not been available. Proportional
248
morbidity, as an epidemiological measure, has the limitation in that the denominator is
249
affected by the prevalence of other diseases. In addition, the ratio of proportional morbidity
250
for groups that have either been exposed or unexposed to a given risk factor can be
251
interpreted as the relative risk only when the total overall morbidity in exposed and
252
unexposed groups is equal (Miettinen and Wang, 1981). Consequently, an increase in the
253
morbidity rate for the disease of interest would need to be accompanied by an equivalent
254
decrease in morbidity for all other causes for the relationship to hold, a condition that is
255
unlikely to be realised in practice. Alternatively, however, the morbidity OR can be used as
256
an estimate of relative risk providing the morbidity rate of all other causes is equal in both
257
exposed and unexposed groups – a condition that is much more likely to occur in real-life
258
situations (Miettinen and Wang, 1981). In the current study, the significant associations of
259
risk factors were assessed using morbidity ORs.
260 261
The PMD was in broad agreement with that seen in other UK-based studies. Stavisky
262
et al. (2010) reported that 16/186 (8.60%; 95%CI 5.00–13.59) randomly selected dogs that
263
presented to veterinary practices in the UK (not including vaccination consultations) showed
264
signs of diarrhoea whilst German et al. (2010) reported that 2,058 of approximately 40,000
Page 11 Page 11 of 21
265
(5.15%; 95%CI 4.93–5.37) dogs seen at a referral hospital had originally presented with
266
diarrhoea. However, the PMD was different from that reported in vet-visiting dogs in the
267
United States. Lund et al. (1999) estimated that the prevalence of diarrhoea in the vet-visiting
268
populations of dogs and cats was 2.2% (95%CI 2.0–2.4) and 1.8% (95%CI 1.6–2.0),
269
respectively. Based on the information presented by Lund et al. (1999), the PMD can be
270
estimated by removing vaccination and pre-operative checks from the denominators and is
271
calculated to be 2.37% (95%CI 2.20–2.55) and 2.74% (95%CI 2.43–3.08) for dogs and cats,
272
respectively. The difference in PMD between the UK and the US does not necessarily reflect
273
differences in the prevalence of diarrhoea in the two countries, but could instead indicate that
274
animals in the US are presented more frequently to veterinary practices for conditions other
275
than diarrhoea.
276 277
In both dogs and cats, the PMD was significantly higher in animals <1 year of age,
278
potentially reflecting a higher incidence of infectious diarrhoea in younger animals
279
(Batchelor et al., 2008), or a greater tendency to present young animals with diarrhoea to a
280
veterinary surgeon. Our study is in agreement with findings in a previous study evaluating the
281
frequency of diarrhoea in four selected breeds of dogs at different age intervals (Saevik et al.,
282
2012). Adult and aged cats were significantly less likely to be presented to a veterinary
283
surgeon for diarrhoea than dogs of equivalent ages.
284 285
While causes of diarrhoea in cats and dogs have broad similarities (Cooper, 2011;
286
Hall, 2009), the difference in PMD could reflect key species differences. More specific
287
information about these differences could be more readily analysed by collecting a larger
288
sample of consultations with more follow-through of individual animals, thereby allowing
289
further assessment of disease characteristics including overall duration, response to therapy
Page 12 Page 12 of 21
290
and the number of repeat visits. It is also possible that the low PMD in cats might be because
291
the faeces of adult/aged cats are often unobserved. As a consequence, diarrhoea that is not
292
accompanied by other clinical signs (such as weight loss, malaise or inappropriate
293
defaecation in the home) could go unnoticed.
294 295
Neutering appeared to be protective for diarrhoea in young (< 1 year) male dogs only.
296
The explanation for this association is not immediately obvious, but it could reflect
297
differences in the attitudes of the owners of these dogs towards presenting their pets to
298
veterinary practices, rather than a true reduction in the prevalence of diarrhoea in this
299
subgroup of animals.
300 301
Analysing the data by individual breeds produced small sub-populations. Despite this,
302
some interesting patterns began to emerge, highlighting the potential utility of this data to
303
identify breed-related disease risk factors. For example, Yorkshire terriers >1 year of age had
304
one of the highest PMD for diarrhoea among adult dogs. Protein-losing enteropathy has been
305
reported in this breed (Waddell and Michel, 2000) and might be a severe manifestation of the
306
predisposition to diarrhoea suggested here.
307 308
National and regional statistics for breed numbers would have been useful to provide
309
a denominator for breed presentation in this study. National statistics of pet ownership are
310
available from the Pet Food Manufacturers Association,3 however, these would not allow for
311
partial national coverage of veterinary practice surveillance or for local variations in disease
312
prevalence. Overall, comparatively small numbers of breed-specific predispositions were
3
See PFMA, 2012. Pet Food Manufacturers Association: Dogs by breed. http://www.pfma.org.uk/dogs-by-breed/ (accessed 29 May 2014).
Page 13 Page 13 of 21
313
identified. With ongoing data collection through SAVSNET, substantially larger datasets will
314
become available, allowing for higher resolution analysis of the breed and age associations
315
that were hinted at in the current study.
316 317
Dietary modification was the most common therapy consistent with the accepted
318
position that the majority of cases of diarrhoea presented to small animal practice reflect
319
either dietary indiscretion (Stavisky et al., 2011), or diet-responsive disease. Antibacterials
320
were frequently used to treat diarrhoea, in keeping with previous findings that diarrhoea
321
increased the probability that antibacterials would be used in veterinary consultations
322
(Radford et al., 2011) and were the most common treatment used in dogs with diarrhoea prior
323
to referral (German et al., 2010).
324 325
A number of infectious agents are recognised to cause diarrhoea (Hackett and Lappin
326
2003; Parsons et al., 2010) but, in the current study, confirmation of infection using faecal
327
sampling was only performed in 6% of cases. Antibiotic-responsive diarrhoea is recognised
328
(Hall, 2011; Hostutler et al., 2004; Westermarck et al., 2005) but represents a minority of
329
cases. Enteropathogens are often present in faeces (Hackett and Lappin 2003; Marks et al.,
330
2011; Stavisky et al., 2011) but these are often found in the absence of diarrhoea (Westgarth
331
et al., 2009). The use of antibacterials may reflect a perception that animals with diarrhoea
332
often lose mucosal barrier function and our study showed increased usage when haemorrhage
333
was present. However, in the absence of sepsis, there is no evidence that outcomes are
334
affected by use of antibacterials in cases of haemorrhagic diarrhoea (Unterer et al., 2011).
335
Thus, the frequent use of antibacterials might reflect our incomplete understanding of the role
336
of suspected pathogens in the pathogenesis of diarrhoea.
337
Page 14 Page 14 of 21
338
SAVSNET has undertaken to communicate individual practice statistics to facilitate
339
benchmarking of prescribing behaviour in relation to anonymised peers. When combined
340
with expert opinion and longer-term follow-up of cases through SAVSNET, such
341
benchmarking might help practitioners review practice protocols and clinical decision-
342
making.
343 344
The practices recruited to the SAVSNET pilot project were widely distributed around
345
England and Wales. Consequently, variations in PMD in, for example, dogs from different
346
practices, could have reflected differences in diagnostic procedures among veterinary
347
practices, or might have indicated true geographical differences in disease prevalence. While
348
variations in PMD between different practices could provide a surrogate for geographical
349
variations in PMD in this pilot study, SAVSNET data also includes full postcode
350
information, which will allow true geographical mapping of disease as the project expands.
351 352
Ultimately, it will be valuable to establish baseline data that characterise the
353
occurrence of GI disease in companion animals. The approach taken by SAVSNET requires
354
the completion of short questionnaires by participating veterinary surgeons at the end of each
355
consultation. Questionnaire responses, together with signalment data and clinical notes, are
356
collated and stored on SAVSNET servers. SAVSNET will also have access to laboratory data
357
regarding the diagnosis of, for instance, infectious diarrhoea. This multi-component approach
358
allows for cross-correlation of data between different sources and parallels the approach
359
taken in large studies in human medicine (O'Brien et al., 2010).
360 361 362
Conclusions SAVSNET allows the collection of ethically-approved surveillance data in real-time
Page 15 Page 15 of 21
363
from veterinary practices. Using limited data from pilot studies, this study demonstrated the
364
ability to identify age and breed variation in PMD as well as inter-practice variation in PMD.
365
Factors were identified that could potentially influence clinical decision-making in practice.
366
Combined with expert opinion, data so gathered could be used to benchmark clinical
367
approaches in participating practices and identify targets for veterinary education and
368
research.
369 370 371 372
Conflict of interest statement None of the authors has any financial or personal relationships that could inappropriately influence or bias the content of the paper.
373 374 375
Acknowledgements During the study period (September 2008 - August 2011), SAVSNET was funded by
376
a consortium including Dechra, Defra, IntervetSP, Merial, Novartis, Pfizer, University of
377
Liverpool and Virbac, and was additionally supported by BSAVA. We are extremely grateful
378
to the 19 practices that participated in this study, and to our colleagues at Vet Solutions
379
without whose dedicated help and support, collection of these data would not have been
380
possible.
381 382
References
383 384 385 386 387 388 389 390 391 392 393
Batchelor, D.J., Tzannes, S., Graham, P.A., Wastling, J.M., Pinchbeck, G.L., German, A.J., 2008. Detection of endoparasites with zoonotic potential in dogs with gastrointestinal disease in the UK. Transboundary and Emerging Diseases 55, 99–104. Cooper, S., 2011. Recurrent diarrhoea in cats. In Practice 33, 272–281. Feldman, R.A., Banatvala, N., 1994. The frequency of culturing stools from adults with diarrhoea in Great Britain. Epidemiology and Infection 113, 41–44. German, A.J., Halladay, L.J., Noble, P.J.M., 2010. First-choice therapy for dogs presenting with diarrhoea in clinical practice. Veterinary Record 167, 810–814. Page 16 Page 16 of 21
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
Hackett, T., Lappin, M.R., 2003. Prevalence of enteric pathogens in dogs of north-central Colorado. Journal of the American Animal Hospital Association 39, 52–56. Hall, E., 2009. Canine diarrhoea: a rational approach to diagnostic and therapeutic dilemmas. In Practice 31, 8–16. Hall, E.J., 2011. Antibiotic-responsive diarrhea in small animals. Veterinary Clinics of North America: Small Animal Practice 41, 273–286. Hostutler, R., Luria, B., Johnson, S., Weisbrode, S., Sherding, R., Jaeger, J., Guilford, W., 2004. Antibiotic-responsive histiocytic ulcerative colitis in 9 dogs. Journal of Veterinary Internal Medicine 18, 499–504. Hubbard, K., Skelly, B.J., McKelvie, J., Wood, J.L.N., 2007. Risk of vomiting and diarrhoea in dogs. Veterinary Record 161, 755–757. Lund, E.M., Armstrong, P.J., Kirk, C.A., Kolar, L.M., Klausner, J.S., 1999. Health status and population characteristics of dogs and cats examined at private veterinary practices in the United States. Journal of the American Veterinary Medical Association 214, 1336–1341. Marks, S.L., Rankin, S.C., Byrne, B.A., Weese, J.S., 2011. Enteropathogenic bacteria in dogs and cats: diagnosis, epidemiology, treatment, and control. Journal of Veterinary Internal Medicine 25, 1195–1208. Martin, S.W., Meek, A.H., Willeberg, P., 1987. Veterinary Epidemiology: Principles and Methods. Iowa State Press, Ames, Iowa, USA, p75. Miettinen, O.S. and Wang, J-D., 1981. An alternative to the proportionate mortality ratio. American Journal of Epidemiology 114, 144-148. O'Brien, S.J., Rait, G., Hunter, P.R., Gray, J.J., Bolton, F.J., Tompkins, D.S., McLauchlin, J., et al., 2010. Methods for determining disease burden and calibrating national surveillance data in the United Kingdom: the second study of infectious intestinal disease in the community (IID2 study). BMC Medical Research Methodology 10, 39. Parsons, B.N., Porter, C.J., Ryvar, R., Stavisky, J., Williams, N.J., Pinchbeck, G.L., Birtles, R.J., et al., 2010. Prevalence of Campylobacter spp. in a cross-sectional study of dogs attending veterinary practices in the UK and risk indicators associated with shedding. The Veterinary Journal 184, 66–70. Radford, A.D., Noble, P.-J., Coyne, K.P., Gaskell, R.M., Jones, P.H., Bryan, J.G.E., Setzkorn, C., Tierney, A., Dawson, S., 2011. Antibacterial prescribing patterns in small animal veterinary practice identified via SAVSNET: the small animal veterinary surveillance network. Veterinary Record 169, 310–U91. Saevik, B.K., Skancke, E.M., Trangerud, C., 2012. A longitudinal study on diarrhoea and vomiting in young dogs of four large breeds. Acta Veterinaria Scandinavica 54, 8. Smith, S., Elliot, A.J., Mallaghan, C., Modha, D., Hippisley-Cox, J., Large, S., Regan, M., Page 17 Page 17 of 21
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
Smith, G.E., 2010. Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire, United Kingdom, June - July 2008. Euro surveillance: Bulletin Européen sur les Maladies Transmissibles. European Communicable Disease Bulletin 15, 19643. StataCorp. 2007. Stata Statistical Software: Release 10. StataCorp LP, College Station, TX, USA. StataCorp. 2007b. Stata 10 Survival Analysis and Epidemiological Tables. Stata Press, College Station, TX, USA, p75. Stavisky, J., Pinchbeck, G.L., German, A.J., Dawson, S., Gaskell, R.M., Ryvar, R., Radford, A.D., 2010. Prevalence of canine enteric coronavirus in a cross-sectional survey of dogs presenting at veterinary practices. Veterinary Microbiology 140, 18-24. Stavisky, J., Radford, A.D., Gaskell, R., Dawson, S., German, A., Parsons, B., Clegg, S., Newman, J., Pinchbeck, G., 2011. A case-control study of pathogen and lifestyle risk factors for diarrhoea in dogs. Preventive Veterinary Medicine 99, 185–192. Unterer, S., Strohmeyer, K., Kruse, B.D., Sauter-Louis, C., Hartmann, K., 2011. Treatment of aseptic dogs with hemorrhagic gastroenteritis with amoxicillin/clavulanic acid: a prospective blinded study. Journal of Veterinary Internal Medicine 25, 973–979. Waddell, L., Michel, K., 2000. Hypomagnesemia and hypocalcemia associated with proteinlosing enteropathy in Yorkshire Terriers: five cases (1992-1998). Journal of the American Veterinary Medical Association 217, 703–706. Westermarck, E., Skrzypczak, T., Harmoinen, J., Steiner, J., Ruaux, C., Williams, D., Eerola, E., Sundback, P., Rinkinen, M., 2005. Tylosin-responsive chronic diarrhea in dogs. Journal of Veterinary Internal Medicine 19, 177-186. Westgarth, C., Porter, C.J., Nicolson, L., Birtles, R.J., Williams, N.J., Hart, C.A., Pinchbeck, G.L., Gaskell, R.M., Christley, R.M., Dawson, S., 2009. Risk factors for the carriage of Campylobacter upsaliensis by dogs in a community in Cheshire. Veterinary Record 165, 526–530.
Page 18 Page 18 of 21
480
Table 1
481
Species associations with proportional morbidity of diarrhoea (PMD) stratified by age category
482 Age category Young (< 1 year) Adult (1-7.99 years) Aged (≥ 8 years) All ages 483 484 485 486
Cat
33
326
PMD (95% CI) a 0.092 (0.058, 0.13)
Dog
109
900
0.108 (0.087, 0.129)
1.2 (0.6, 1.8)
0.508
Cat
26
1166
0.022 (0.014, 0.030)
1.00
–
Dog
293
4066
0.067 (0.056, 0.079)
3.2 (2.2, 4.3)
0.000
Cat
49
1772
0.027 (0.020, 0.034)
1.00
–
Dog
219
3458
0.060 (0.050, 0.069)
2.3 (1.8, 2.8)
0.000
Cat Dog
108 621
3264 8424
0.032 (0.026, 0.038) 0.069 (0.059, 0.079)
1.0 2.2 (1.8,2.6)
0.000
Species Cases Controls
Odds ratio (95% CI) a 1.0
P –
95% CI, 95% Confidence interval a Confidence intervals calculated to allow for intragroup correlation within veterinary practice.
Page 19 Page 19 of 21
487
Figure legends
488 489
Fig. 1. The sequence of questions presented to the veterinary surgeon on completion of the
490
consultation is shown. * Indicates questions where multiple answer options were allowed.
491 492
Fig. 2. The study population. (a) Pie chart with segments representing proportions of different
493
species in the population questioned about diarrhoea. (b) Percentage of cases with diarrhoea
494
by species. Bars represent percentage of individuals presenting with diarrhoea for each
495
species (only species occurring in ≥100 consultations are shown).
496 497
Fig. 3. Proportional morbidity of diarrhoea (PMD) in dogs at individual veterinary practice
498
branches (anonymised). Bars represent the PMD for dogs presenting at individual practice
499
branches where the branch contributed >100 records to the study.
500 501
Fig. 4. The proportional morbidity of diarrhoea (PMD) in dogs and cats, segregated by age
502
group. Bars represent PMD of diarrhoea in dogs (grey bars) and cats (white bars) in indicated
503
age groups (± 95% confidence interval). Practice-adjusted odds ratios for groups with
504
common labels (a, b or c) were not significantly different (Bonferroni-adjusted P > 0.05).
505 506
Fig. 5. Breeds with diarrhoea. Proportional morbidity of diarrhoea (PMD) in individual
507
breeds is shown for adult/aged dogs (≥1 year). Bars represent PMD (± 95% confidence
508
intervals) for each breed. For the purposes of this study, the term ‘cross’ refers to a non-
509
pedigree animal where the breeds of one or more of the parents are recognizable and have
510
been recorded in the animal’s record. The term ‘crossbreed’ refers to a non-pedigree animal
511
where the breeds of the parents have not been recorded.
512 Page 20 Page 20 of 21
513
Fig. 6. The proportions of dogs and cats with uncomplicated diarrhoea presenting after given
514
durations of illness. Bars represent cases classed as uncomplicated (no vomiting or
515
haemorrhage) as a proportion of all cases of diarrhoea with a given duration of illness prior to
516
presentation. Strata representing 5-7 days and ≥8 days have been collapsed into a single
517
category to ensure sufficient sample size at each level. Grey bars represent dogs and white
518
bars represent cats.
519 520
Fig. 7. Diagnostic tests performed in cases of diarrhoea. Bars represent the proportion of
521
cases for which the given diagnostic test was performed. Grey bars represent dogs and white
522
bars represent cats. TLI, trypsin-like immunoreactivity; PLI, pancreatic lipase
523
immunoreactivity.
524 525
Fig. 8. Treatment choices in diarrhoea. (a) Frequency of use of different treatment classes.
526
Bars represent proportions of cases receiving a given treatment. (b) Frequency of use of
527
antibacterials in cases presenting with haemorrhagic and non-haemorrhagic diarrhoea, bars
528
represent proportion of cases (± 95% confidence intervals) treated with antibacterials (*
529
indicates proportion different to ‘no’ group, P < 0.05). (c) Antibacterials used in treating
530
diarrhoea. Bars represent the proportion of diarrhoea cases that were treated with
531
antibacterials that received each drug type. Grey bars represent dogs and white bars represent
532
cats.
Page 21 Page 21 of 21