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
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