Abstract
Practicing veterinarians play an important role in detecting the initial outbreak of disease in animal populations. A pilot study was conducted to determine the feasibility of a veterinary-based surveillance system for the Ontario swine industry. A total of 7 practitioners from 5 clinics agreed to submit information from July 1, 2007 to June 30, 2008. The surveillance program was evaluated in terms of timeliness, compliance, geographic coverage, and data quality. Our study showed that the veterinary-based surveillance system was acceptable to practitioners and produced useful data. The program obtained information from 25% of pig farms in Ontario during this time period. However, better communication with practitioners, more user-friendly recording systems that can be adapted to each clinic’s management system, active involvement of the clinics’ technical personnel, and the use of financial incentives may help to improve compliance and timeliness.
Résumé
Les vétérinaires praticiens jouent un rôle un important dans la détection initiale d’une épidémie de maladie dans les populations animales. Une étude pilote a été menée afin de déterminer la faisabilité d’un système de surveillance vétérinaire pour l’industrie porcine ontarienne. Un total de 7 vétérinaires praticiens travaillant dans 5 cliniques ont accepté de soumettre des informations pour la période allant du 1er juillet 2007 au 30 juin 2008. Le programme de surveillance a été évalué en termes de d’opportunité, d’observance, de couverture géographique et qualité des données. Notre étude a démontré que le système de surveillance vétérinaire était acceptable pour les praticiens et fournissait des données utiles. Le programme a permis d’obtenir de l’information de 25 % des fermes porcines en Ontario durant cette période de temps. Toutefois, une meilleure communication avec les praticiens, des systèmes d’enregistrement conviviaux qui peuvent être adaptés au système de gestion de chaque clinique, l’implication active du personnel technique des cliniques et l’utilisation d’incitatifs financiers pourraient aider à améliorer observance et opportunité.
(Traduit par Docteur Serge Messier)
Introduction
There is a strong interest in developing reliable surveillance programs to provide “early-warning” systems due to the growing concern about emerging and re-emerging animal and zoonotic diseases, and their negative impacts on animal and public health, the environment, and economy (1,2). Practicing veterinarians are in close contact with livestock and producers, and consequently, they play an important role in detecting initial cases of novel infectious diseases or changes in the incidence of specific syndromes (such as respiratory disease or mortality) that may provide the first warning of a disease outbreak, and allow more rapid and efficient collection of diagnostic samples and implementation of disease control strategies (3). Veterinary-based surveillance systems have been implemented in some countries for cattle, but information on swine veterinary-based surveillance is limited or still under development (4–8).
The need for a disease surveillance program in Ontario has been clearly demonstrated by the events of the past 5 y. Over this time period 3 major disease problems occurred in the Ontario pig industry. Swine disease outbreaks associated with a triple-reassortant human-avian-swine influenza H3N2 virus, new strains of porcine respiratory and reproductive syndrome virus (PRRSv) and porcine circovirus type 2b were reported in 2004 to 2006 in the province (9–12). In addition, concern has been raised regarding the role that swine play with respect to worker safety and public health as a result of the outbreak of a novel H1N1 influenza virus that was first identified in humans in Mexico in the spring of 2009, and subsequently spread to swine in Alberta, Canada. The history of recent disease outbreaks together with the known willingness of the swine industry in Ontario to share health data, led to the decision of conducting a 12 mo pilot study to determine the feasibility of a veterinary-based syndromic surveillance system using practitioner data with the swine industry as a model.
The evaluation of a surveillance program’s performance is a necessary procedure to determine the reliability and usefulness of the results obtained, but few surveillance programs are actually evaluated (13,14). The basic aspects to be evaluated in an effective syndromic surveillance system include compliance and timeliness (13,15–19). Compliance is defined as good quality of participation in the surveillance system and takes into account both the frequency and duration of participation. Timeliness refers to the speed from information collection to dissemination (16,18,20); however, it is important to assess subsets of this measure. For instance, the time between disease occurrence and the availability of data for analysis by the surveillance program is often the most rate-limiting step in establishing real-time surveillance. Quality of data captured can be evaluated based on the representativeness or coverage with respect to the population, and based on the completeness or ability of each report to capture the information requested by the surveillance program (18,20). In addition, the effectiveness of various data collection tools used by the surveillance programs should be evaluated since timeliness, cost, acceptability, and even compliance may be influenced by the type of data collection tools or software used (21).
The objective of this study was to evaluate the Ontario Swine Veterinary-based Surveillance (OSVS) pilot program in terms of data quality, geographic coverage, timeliness of data submitted, effectiveness of data collection tools, and the level of compliance among participating veterinarians. The purpose of the evaluation was to highlight areas where the OSVS program could improve its performance in terms of acceptability to practitioners and speed and efficiency of data collection to improve the capacity of the program to identify outbreaks and the appearance of novel swine pathogens in Ontario. The report also highlights the most relevant feedback obtained from participating veterinarians to improve this surveillance system.
Materials and methods
Description of the pilot surveillance system
The OSVS pilot project, funded by the Ontario Ministry of Agriculture Food & Rural Affairs (OMAFRA) and the Ontario Animal Health Strategic Investment (AHSI) fund, was established primarily to determine the feasibility of a practiced-based swine health syndromic surveillance system for Ontario and to provide the OMAFRA with recommendations for the long-term establishment of this type of syndromic surveillance program. In addition, the OSVS project intended to record the occurrence/incidence of different syndromes and to identify increased rates of disease in space, time, and space-time. The proposed surveillance system was tested during a 12 mo pilot study from July 2007 to June 2008 and administered by 2 principal investigators (DP and RF) and 1 project coordinator (MRA) at the Department of Population Medicine, University of Guelph, Guelph, Ontario. Initially, key stakeholders were identified. Swine specialists from the University of Guelph, OMAFRA, and the Animal Health Laboratory (AHL) at the University of Guelph were contacted to gain from their experience with other surveillance programs and to obtain important feedback for the design of the data collection form.
Fifteen mixed (n = 9) and swine speciality (n = 6) veterinary clinics in Ontario that were registered with the Ontario Association of Swine Practitioners and known to service most of the swine industry in Ontario were visited and the swine veterinarians were asked to participate in the OSVS program. During these interviews, 18 veterinarians were asked to evaluate the data collection forms and respond to a series of open format questions that evaluated a variety of issues related to participation in the pilot program. In addition, during these visits financial support for additional diagnostic testing was offered to practitioners as an incentive to participate in the program. Seven veterinarians from 5 clinics recorded and transmitted data voluntarily during the pilot study.
During data collection, veterinarians were requested to record all their daily disease and non-disease related farm visits and calls, then submit these reports on a weekly basis to the OSVS project coordinator. Weekly e-mail contacts with the clinics were established to remind practitioners about the submissions of their records. Phone calls to individual practitioners were performed when submissions were not sent on a weekly basis. For each report, we recorded whether the visit/call was related to disease or routine management. For routine management or non-disease calls, only the date, name of veterinarian, farm code, and location of the farm were requested. Disease farm visits and calls were summarized according to the animal body systems affected and in what manner production was affected. Practitioners also recorded whether the visit/call was related to an incident or an ongoing situation. An incident case referred to a new disease condition or a situation where the farm had a history of the condition, but clinical signs had resolved. Epidemiological data related to the geo-location of the farm (county and postal codes) and actual estimates of farm size and animals affected were included in our initial form. All the information was collected at the farm level. At the beginning, practitioners rarely recorded counties and postal codes. An approach to obtain these data was discussed with practitioners and this action resulted in veterinarians producing a master list of codes for their clients’ farms with location information and the type of production of each farm. Several veterinarians felt that providing actual estimates concerning farm size and the number of animals affected was not practical based on the structure of many of their clients’ farms and the manner in which diseases are diagnosed by swine practitioners. Therefore, the questions related to actual estimates concerning farm size and animals affected were removed prior to the pilot study.
Three different collection tools were available to record data. The data were recorded either on carbon paper forms, personal digital assistants (PDA), or an Internet-based form. The questions were the same for all collection tools, but the layout was different based on the requirements of the specific media. The paper forms were transmitted to the project coordinator via fax or mail. The PDA version of the questionnaire was transmitted via e-mail as comma delimited files. The Internet version of the questionnaire allowed for the direct transmission of the data (Microsoft Excel; Microsoft, Redmond, Washington, USA) files. A database (Microsoft Access; Microsoft) was available to allow the manual entering of data from paper collection forms and the electronic addition of data obtained from PDAs and through the Internet. An electronic newsletter was distributed monthly to all clinicians within the surveillance network. The newsletter summarized the following: the overall distribution and trends of syndromes reported within the network; disease issues reported by clinicians in the network, government officials, industry, or the regional diagnostic laboratory; concerns and issues associated with the administration of the program; swine health research conducted at University of Guelph; and clusters of disease identified with their subsequent epidemiological validation. Automated “do-files” were created (Intercooled Stata 9 for XP, 2003; Stata Corporation, College Station, Texas, USA) to allow for the creation of summary graphs, tables, and the re-configuration of data for disease cluster analysis. Computer software (SaTScan Version 7; Kulldorff M and Information Management Services, National Cancer Institute, Bethesda, Maryland, USA) was used for the temporal, spatial, and space-time scan statistics for disease cluster detection analyses. The project coordinator conducted the daily administration of the pilot program, was responsible for entering non-electronic data as records were received, regularly reviewing the database for completeness of data, ran and interpreted statistical analyses, and contacted practitioners to follow up with farms that were located within statistical clusters to evaluate if they captured a disease outbreak of concern. In addition, the project coordinator met with stakeholders and practitioners, and created monthly reports.
Evaluation of the program
Compliance
In order to estimate the level of compliance among participating veterinarians, the OSVS project calculated the weekly ratio of the total OSVS submissions to submissions made by the same veterinarians to the regional diagnostic laboratory [Animal Health Laboratory (AHL), University of Guelph]. The weekly ratios of OSVS to AHL submissions were also calculated for each participating veterinarian to identify when compliance by individual practitioners was poor. The AHL offers a wide range of veterinary diagnostic tests and laboratory services in different specialty areas, such as pathology, bacteriology, and virology, in the field of animal health. The information from the AHL included the veterinarian and clinic code, and the date of submission. Only AHL submissions from farms made by the 7 practitioners participating in the OSVS program were used to calculate the ratios. A ratio less than 1 would reflect poor compliance since we required that submissions to OSVS include all calls/visits and not only those requiring laboratory diagnostics.
To determine if there was an association between compliance and season, a multilevel mixed effects logistic regression model was created. The months when the records were submitted were categorized as follow: summer — July, August, and September, 2007; fall — October, November, and December, 2007; winter — January, February, and March, 2008; and spring — April, May, and June, 2008. In this model, season was a fixed effect, veterinarian was modeled as a random intercept, and in order to account for interactions, the effect of season was allowed to vary by practitioner by including a random slope for season. Predicted means were computed and graphed to visually assess the compliance of each veterinarian within season.
Timeliness of data
In the present study, we limited our evaluation of timeliness to the time between disease occurrence and the availability of data for analysis by the surveillance program. The time in days to availability of the data was evaluated by calculating the difference between the date when the practitioner recorded the farm report and the date when the information was transmitted to the program coordinator (22,23). Means, medians, and standard deviations were calculated for each participating veterinarian, data collection tool (PDA versus paper), mode of transmission of the data (electronic versus non-electronic), and season (summer, fall, winter, and spring).
The association of time in days to availability of data with the type of collection tool (paper form versus PDA), the mode of transmission of the data (electronic versus non-electronic), the season when the data were recorded, and the veterinarian were analyzed using a mixed linear regression model. The transmission tool was coded as electronic (e-mail, fax) versus non-electronic (mail). Season was coded as previously described. A square root transformation was performed for the time to availability of data in order for the residuals to meet the model assumptions of normality and homoscedasticity (24).
In this model, collection tool and season were considered fixed effects and veterinarian was modeled as a random intercept, and a random slope was included for season to allow the term to vary by veterinarian. Linear prediction of the fixed portion and the contribution of the random effects were computed and graphed after fitting the model to visually assess the time to availability of data by veterinarian within season. The data were re-transformed by taking the square of the estimates of the fitted values to obtain the difference of date to availability of data. Residuals were visualized to determine the presence of unusual observations that would require further investigation.
Quality of data
Geographic coverage
The farm locations were geo-referenced based on postal codes and a geographic information system software package (ArcGIS version 9.2; ESRI, Redlands, California, USA), was used to visualize the locations. In order to determine the proportion of farms covered by the OSVS program by agricultural region, the number of swine farms visited by practitioners was compared to the total number of swine farms within each agricultural region in Ontario according to the Canadian Agricultural Census of 2006 (25). The first 3 digits of the postal codes of the farms were classified within agricultural region by cross-referencing the 2006 Census of Agriculture files and the file of Forward Sortation Area Maps that provides the reference to postal code boundaries in Canada. These data were obtained from the Geographic and Geospatial data of the Tri-University Data Resources Centre (TDR) at the University of Guelph.
Completeness of data
The completeness of the data was assessed by measuring the frequency of data entered for each of the form’s fields/elements of the questionnaire. The evaluation of each field was based on 2 contexts for which these fields should have been filled: i) field requested for all records (disease and routine management), or ii) field requested for a disease report only. Fields that were requested for all records included date of visit/call, farm code, county/city, postal code, whether the call was a farm visit or phone call, and if the visit/call was related to disease or a routine management procedure. Fields that were requested only in cases of a disease problem included the following: type of production, area/age affected in site, categories of body systems or production parameters affected, whether the samples from the affected farm were submitted to the AHL, whether a farm report was a new/incident case or an ongoing situation, and information related to treatment efficacy. The type of production affected included on the list of check boxes were: farrow-to-finish, farrow-to-wean, wean-to-finish, finishing, and nursery. The area/age affected on site included on the form were: breeding/gestation, farrowing, nursery, and finisher. The options included in the form for body systems affected were respiratory, digestive, nervous, reproductive, musculo-skeletal, and integument-senses. The options included in the form for production parameters affected included: increase in mortality, increase in morbidity, poor growth rate, poor farrowing rate, poor conception rate, and reduced litter size. Veterinarians were able to record additional options for these fields in a text box. A copy of the form is included in Appendix 1.
Statistical analyses
Descriptive statistics and quantitative statistical analyses of compliance, timeliness, and quality of data evaluation were conducted using computer software (Intercooled Stata 9 for XP, 2003; Stata Corporation). The xtmelogit and xtmixed commands were used for the mixed logistic and linear models, respectively.
Feedback from interviews
Veterinarians were interviewed at the beginning and the end of the study. During these formal interviews, practitioners were asked to respond to a series of open format questions to evaluate a variety of issues related to participation in the project including: aspects of the program that would maintain their motivation to participate; the type of feedback that would maintain their commitment to the program; conditions for sharing their data with various institutions; technologies preferred to record and transfer their data; what they would consider appropriate feedback from the program; appropriate incentives to continue participating in the program; and recommendations to improve compliance, timeliness, and the performance of the program.
Results
Compliance with the OSVS program
Five veterinarians from 4 clinics sent information on a routine basis to the program from July 1, 2007 to June 30, 2008. Two veterinarians from 1 clinic did not forward information onto OSVS until October, 2007. An additional practitioner sent information for 1 mo, but these data were not included in this evaluation. Based on the ratio of the total submissions to the OSVS and the AHL, the overall quality of compliance was poor for the first 3 mo of the project (Figure 1A). However, similar analyses for individual veterinarians indicated that compliance was variable among veterinarians and that there were intermittent periods when this ratio fell below 1 (Figure 1B and C).
Figure 1A,B,C.
Ratio of total farm reports (A) and ratio of 2 individual veterinarians (B,C) submitted to the Ontario Swine Veterinary-based Surveillance (OSVS) system and total submissions and submissions made by the same 2 veterinarians to the Animal Health Laboratory (AHL) from July 1, 2007 to June 30, 2008.
Ratio > 1 indicates a greater number of submissions to OSVS relative to the AHL.
For the individual analyses of three veterinarians, grey bars represent weeks that had submissions made only to OSVS (> 1) or only the AHL (< 1).
The multilevel mixed effects logistic regression model showed that the ratio of OSVS to AHL submissions was greater in the fall (P = 0.02), winter and spring (P = 0.04) compared to the summer based on the fixed effects in the model (Table I). However, the significant random slope suggested that the ratio of submissions to OSVS and AHL varied significantly by veterinarian and season. Although compliance improved in the fall, winter, and spring more variation of the predicted means was observed among veterinarians in the summer (Figure 2A).
Table I.
Mixed logistic regression model comparing the submissions from 7 veterinarians in Ontario to the Ontario swine veterinarian-based syndromic surveillance (OSVS) system and submissions to the regional diagnostic laboratory from July 1, 2007 to June 30, 2008
| Mixed logistic model with a random intercept for veterinarian and a random slope for seasona |
||||
|---|---|---|---|---|
| 95% confidence intervals |
||||
| Odds ratio | LCIb | UCIc | P-value | |
| Fixed-effects parameters | ||||
| Fall | 3.42 | 1.21 | 9.71 | 0.02 |
| Winter | 2.89 | 1.03 | 8.12 | 0.04 |
| Spring | 2.92 | 1.04 | 8.20 | 0.04 |
| Constant | 0.79 | 0.29 | 2.20 | 0.66 |
| Random-effects parameters | ||||
| Veterinarian | ||||
| Vard | 5.45 | 1.99 | 64.07 | < 0.001f |
| Cove | 1.03 | 0.59 | 1.77 | |
Using an exchangeable covariance structure.
Lower confidence interval.
Upper confidence interval.
Variance for fall, winter and spring.
Covariance for fall, winter and spring.
Summer is the referent season; OR > 1 indicated increased odds relative to referent; OR < 1 indicates decreased odds relative to referent.
The P-value refers to the overall significance of the random effects to the model.
Figure 2A,B.
Predicted probability of the compliance of veterinarians among different time intervals (A), and predicted time to availability of data for the OSVS system (B) from July 1, 2007 to June 30, 2008, obtained from the mixed logistic and linear regression models respectively with a random intercept for veterinarian and a random slope for season.
Each veterinarian is represented by a unique shape and shade.
Two practitioners did not send any data to OSVS until the fall.
Practitioners may have used multiple data collection tools within season, therefore their shape will appear two times.
Data collection tools and time to availability of data
Descriptive statistics
Three of 7 veterinarians used the paper forms. Four veterinarians used PDAs. One veterinarian switched from paper forms to PDA in the fall, another switched in the winter, and another in the spring. One veterinarian switched from PDA to paper in the spring (Figure 2B). The Internet-based link to the form was never accessed. A summary of the average and median times to availability by veterinarian, type of data collection tool, mode of record transmission, and season is included in Table II. Although practitioners were requested to transmit the data on a weekly basis, the average time to availability of data for all cases was 22.3 d. The time to availability of data from PDA users was shorter (3 d) and less variable than those completing paper forms (Table II). The time to availability of data for electronic transmission tools was also qualitatively shorter (5 d) and less variable than mail (Table II). The time to availability of data for fall, winter, and spring was shorter and less variable than summer (Table II).
Table II.
Summary statistics of time to availability of data to the Ontario Swine Veterinarian-based Surveillance (OSVS) pilot project (calculated based on veterinarian, collection tool used, and by mode of transmission) from July 1, 2007 to June 30, 2008
| Number of forms submitted | Mean days | Median days | Standard deviation | Min/Max | ||
|---|---|---|---|---|---|---|
| Veterinarian | Aa | 358 | 27.5 | 25.5 | 14.3 | 0/66 |
| Bb | 288 | 21.5 | 19 | 13.6 | 0/56 | |
| Cb | 255 | 23 | 21 | 14.3 | 0/82 | |
| Dc | 106 | 41.7 | 27.5 | 39.4 | 0/149 | |
| Ea | 180 | 25.8 | 22 | 17.4 | 0/69 | |
| Fa | 364 | 11.9 | 9 | 8.9 | 0/42 | |
| Gb | 59 | 10.6 | 7 | 9.7 | 0/35 | |
| Collection tool | Paper | 686 | 24 | 20 | 21 | 0/149 |
| PDA palm | 924 | 21 | 19 | 15.5 | 0/69 | |
| Mode of transmission | 625 | 25.5 | 21 | 21.4 | 0/149 | |
| Electronic | 985 | 20.3 | 19 | 15.8 | 0/69 | |
| Season | Summer | 199 | 35.3 | 29 | 29.7 | 1/149 |
| Fall | 441 | 21.6 | 19 | 15.7 | 0/69 | |
| Winter | 512 | 17.9 | 15 | 12.3 | 0/69 | |
| Spring | 458 | 22.5 | 20 | 16.5 | 0/69 |
No date of visit/call were recorded for 1 record.
PDA records.
Paper records.
Used paper form from July through December, 2007 and then switched to a PDA.
Practitioner G used a fax machine to transmit paper forms.
Statistical models
High correlations were observed between collection tools and modes of transmission of the data to the surveillance system (r = 0.92, P < 0.001); therefore, only collection tool, season, and veterinarian were included in the mixed linear model. In the multilevel mixed linear model, the effect of PDA versus paper was not significant and time to availability of data was significantly lower in winter compared with summer (Table III). No significant differences were observed among fall, winter, and spring seasons. However, the significant random slope in the model suggested that time to availability of data varied significantly by veterinarian by season. In summer, the predicted time to availability of data by veterinarians ranged between 11 and 88 d. Although the mean time improved for the following seasons, among veterinarians the predicted mean ranged from 5 to 47 d in the fall, 5 to 23 d in the winter, and 10 to 37 d in the spring (Figure 2B).
Table III.
Mixed linear models testing the time to availability of data in days, by collection tool and season, to the Ontario Swine Veterinarian-based Surveillance (OSVS) pilot project from July 1, 2007 to June 30, 2008
| Mixed linear model with a random intercept for veterinarian and a random slope for seasona |
||||
|---|---|---|---|---|
| 95% confidence intervals |
||||
| Coefficientb | LCIc | UCId | P-value | |
| Fixed-effects parameters | ||||
| PDA | 0.24 | −0.24 | 0.72 | 0.32 |
| Fall | −1.28 | −2.88 | 0.32 | 0.12 |
| Winter | −1.88 | −3.49 | −0.28 | 0.02 |
| Spring | −1.36 | −2.96 | 0.24 | 0.09 |
| Constant | 5.27 | 3.55 | 6.99 | < 0.001 |
| Estimate | LCIc | UCId | P-value | |
| Random-effects parameters | ||||
| Veterinarian | ||||
| Vare | 4.10 | 2.16 | 7.77 | < 0.001g |
| Covf | −0.13 | −1.41 | 1.15 | |
| Variance residuals | 1.98 | 1.84 | 2.12 | |
Using an exchangeable covariance structure.
Based on the square root of the difference in days between disease occurrence and the availability of data for analysis.
Lower confidence interval.
Upper confidence interval.
Variance for fall, winter and summer.
Covariance for fall winter and summer.
Paper and summer are the referent collection tool and season; coefficient > 0 indicates increased time to availability of data and coefficient < 0 indicates decreased time to availability of data.
The P-value refers to the overall significance of the random effects to the model.
Quality of data collected by the OSVS pilot project
A total of 2222 swine farms were enumerated in Ontario during the 2006 Agricultural Census. A total of 1226 farms were included in the master lists of all participating clinics (55.17%). Based on farms visited during the pilot project (n = 571), 524 (91.7%) of the farms were geo-located (Figure 3). These farms involved different types of production systems including: 142 farrow-to-finish (27.1%); 70 farrow-to-wean (13.3%); 11 wean-to-finish (2.1%); 128 finisher units (24.4%); 91 nurseries (17.3%); 11 gilt isolation units (2.1%); 8 boar studs (1.5%); 17 farrow-to-feeder (3.2%); 4 were visits to head offices (0.76%); and for 42 farms (8%) production type was not reported.
Figure 3.
Proportion of Ontario swine farms visited by 7 practitioners in the Ontario Swine Veterinary-based Surveillance (OSVS) pilot project, by agricultural region, from July 1, 2007 to June 30, 2008.
A total of 2222 were enumerated in the Canadian Agricultural Census of 2006. A total of 571 farms were visited however only 524 (23.6%) farms were geo-located.
*(Number of farms covered/Number of farms enumerated by the Canadian Agricultural Census of 2006) by region.
A total of 1610 records including disease and routine management reports were obtained during the study period from the participating veterinarians. Disease alone was reported in 754 of the records; however, in 200 records disease was identified in a routine management visit. The completeness of fields required for all records ranged between 95% and 100% for farm code, postal code, whether a visit was a farm visit or a phone call, or whether it was related to disease or routine management (Table IV). The field related to county or city was completed only 81.8% of the time. Although type of production and area affected were only requested for a disease report, the type of production was included in the master lists of 3 clinics and area affected was also recorded in several records regardless of the type of visit (Table IV). Practitioners consistently reported additional options for type of production such as boar studs, gilt development units, isolation, and artificial insemination units.
Table IV.
Summary of the completeness of data for fields required for all the records and fields required only for disease records of the Ontario Swine Veterinary-based Surveillance (OSVS) pilot project from July 1, 2007 to June 30, 2008
| Fields required for all the records sent to the OSVS (n = 1610 records) | ||
|---|---|---|
| Field | Number of records where field was reported | % completed |
| Farm code | 1588 | 98.6 |
| County/city ID | 1319 | 81.8 |
| Postal code | 1529 | 94.9 |
| Farm visit/phone call | 1609 | 99.9 |
| Disease/routine management | 1609 | 99.9 |
| Type of production | 1386a | 86.1 |
| Area(s) affected | 1112a | 82.9 |
| Fields required only for disease records (n = 954 records)b | ||
| Body system | 917 | 96.1 |
| Production parameter | 710 | 74.4 |
| Previous mortality | 68 | 7.1 |
| New mortality | 132 | 13.8 |
| AHL submissions (yes/no) | 212 | 22.2 |
| New versus ongoing disease problem on the farm | 923 | 96.7 |
| Treatment efficacy | 754 | 79 |
| Treatment used | 525 | 55 |
Some veterinarians included this information for reports that did not involve disease.
Total disease records n = 954 (754 disease alone record + 200 disease cases reported in a routine management visit).
The proportion of the fields completed for disease reports was based on the sum of all disease-only records and where a disease problem was reported during a routine management visit. Body systems affected were almost always recorded when the visit was related to a disease report (96.1%). However, the production parameters affected when a disease report was indicated were less frequently reported (74.4%; Table IV). For body systems and production parameters affected, the “other” option was commonly reported with additional options, such as urogenital, behavior, multi-systemic, increase return to heat, and reduced feed intake. In addition, other common clinical signs were recorded in the other option, such as: sudden deaths, fever, septicemia, wasting, abortions, off-feed, prolapse, tail biting, and ear necrosis. The field related to AHL submissions was recorded only in 22.2% of the disease records. Determining whether a case was a new/incident or an ongoing situation was reported in 96.7% of the disease records (Table IV). However, the veterinarians appeared to define these terms differently despite providing a document with the definitions for each element of the form. Veterinarians often appeared to record an event as an ongoing situation when the farm had not had a history of the condition, but clinical signs had been observed recently. In this particular case, we requested that the field should be recorded as an incident problem. Treatment efficacy was reported in 79% of the disease reports (Table IV). This field could be reported as effective, occasionally effective, and not effective. Type of treatment used in a disease record was an open question and the level of detail was highly variable. In some cases, only the route of treatment (such as water, feed, or injectable medication) was indicated, while in other instances specific drug names were provided.
Feedback from swine practitioners of issues related to the program
A total of 18 veterinarians from 15 clinics were interviewed at the beginning of the pilot study. In terms of maintaining participation to the program, most practitioners felt that timely feedback concerning the disease syndromes being studied, information concerning disease trends in their own practices, and information concerning disease trends from other sources, such as the AHL, would be important personal benefits for collecting data for the surveillance system. However, the clinics that were unwilling to participate stated that they were unwilling to provide the information required in a consistent manner because they were too busy. Some practitioners stated that they would consider participating only if they could be compensated at their hourly billing rate. No evident differences between participating and non-participating practitioners were observed. At the end of the pilot study, 7 veterinarians from 5 participating clinics were interviewed. Most of the practitioners expressed in the last interview their willingness to continue participating in the program and agreed that this type of veterinary-based surveillance system would be appropriate for detecting emerging or re-emerging pathogens. However, some practitioners highlighted their desire for obtaining information concerning the trends in swine health status in Ontario as a motivation to continue participating in the program. In addition, they expressed their expectation to obtain compensation for their time and having the program move towards an electronic real-time surveillance system that involved more clinics. Practitioners suggested that the OSVS recording form should be harmonized with their own medical records or billing forms to avoid the need to enter data twice. Most of the practitioners agreed that the OSVS monthly newsletter provided valuable information and expressed the desire of using the information for their own reports to clients.
Discussion
Methods to assess different aspects of veterinary-based surveillance programs must be established to determine which aspects of the surveillance performance have to be improved. Timeliness and compliance were key measurements to evaluate the effectiveness of our surveillance system. It is important to determine the potential of a surveillance system for real time surveillance so interventions can be applied in a timely and effective manner (26). The difference in days between recording a farm visit or phone call and the day when the record was received by the program was easy to calculate and appeared to be a useful way to measure timeliness in our pilot study. In addition, human studies have monitored time of compliance and the association of compliance with different variables using survival analysis (15,27). However, to our knowledge, there are no published studies that have measured the level of compliance among individual veterinarians involved in a disease surveillance system. As defined by Chauvin et al (15), a good quality of participation takes into account the duration and the frequency of participation. In our study all practitioners participated throughout the study, so we attempted to use a novel method to determine the level of participation of all and individual practitioners participating during the pilot study. The drop-offs of compliance (such as the periods within the pilot study when reports of visits/calls were not provided) could have an effect and complicate the interpretation of statistical disease cluster detection methods that we are investigating currently for the program. The ratio between the submissions of a surveillance program compared to submissions made to a commonly used diagnostic laboratory proved to be a useful method to identify periods of poor compliance. However, it has to be considered that although most of the participating practitioners have their disease submissions sent to the AHL, our ratios may be overestimated if veterinarians had made submissions to other laboratories in Ontario. Additional assessments may be required to measure submissions to additional diagnostic laboratories in Ontario to estimate compliance in a permanent program. Another limitation with these data is that a ratio of 1 indicates only that the same number of farm reports was sent to both AHL and the OSVS, but this would not tell us if these farm reports were different from each other. Linking the OSVS to AHL data will be required in the development of a permanent program.
Our results based on the mixed linear and logistic regression models showed that although time to availability of data and compliance improved after the first 3 mo of the project, both measurements varied widely by veterinarian and season. Some human practitioner-based surveillance studies reported that the length of time during which practitioners comply with a given surveillance protocol is highly variable (15,16,27,28). In addition, based on discussion with colleagues involved in other veterinary surveillance systems (3,5,6,29,30), we suspected that compliance and timeliness among veterinarians would be an issue for the success of the OSVS. The wide range of days to availability of data and the low compliance observed at the beginning of the program might be expected because the project started during common summer vacations. We also suspect that it took some time for veterinarians to adopt the recording and transmission of data to the surveillance program as part of their routine weekly activities, and this was confirmed by some veterinarians in the final interview. In addition, in the final interview, several practitioners confirmed that at times during the project, they were not able to record all the information requested by the program because of their busy schedules and, in particular, 2 veterinarians did not start to submit their data regularly until the fall. It is evident that improving this aspect of the program performance will be necessary. Although the continued e-mail and phone communication with practitioners, and the monthly reports likely had a positive effect on timeliness and compliance over time, veterinarians expressed that a permanent surveillance program may need to consider a financial compensation program to maintain their timely participation.
Compliance and timeliness are highly influenced by the type of methods used to record the data and may also be improved if user-friendly data collection tools or software are used (21). Ideally, a real time syndromic surveillance system would include an electronic recording system involving PDAs, web-based forms, or both (3,31) to allow the automatic transmission of data (2,17). Veterinary-based surveillance systems, such as Alberta Surveillance Veterinary Network (AVSN) and Veterinary Practitioner Aided Disease Surveillance System (VetPAD) in New Zealand, have used electronic systems that allow for the automatic transmission of data (3,4). In the final interview in our study, some veterinarians expressed that they would like to see the program moving towards an electronic-based surveillance system; however, this statement conflicted with the reporting behavior we observed during the project, because the Internet version of the form was never accessed and the PDAs did not result in data being transmitted significantly faster. Some veterinarians stopped using PDAs because they found it difficult to complete the free text fields, and practitioners were not willing to return to their offices and re-enter the data on the Internet based-form. Most veterinary clinics continue to use paper-based records for on-farm recording of data and requesting veterinarians to record data that they may consider outside their typical diagnostic database may limit the ability of a future program to maintain the timely cooperation of participants. The use of paper forms that harmonized the data needed for the surveillance system with the practitioners’ medical records, billing systems, or both may help to record information more accurately and in a timelier manner, if the forms could be sent on a routine basis. Involving animal technicians or other office staff in the regular transmission of each clinic’s data may also help to improve timeliness of data.
The coverage of a surveillance system is important to estimate because it reflects the representativeness of a surveillance program. The coverage of our program was estimated based on the distribution and number of farms involved in the OSVS surveillance system compared with Agricultural Census Data. However, surveillance coverage can also be compared with coverage of other networks or data sources, such as diagnostic laboratories or government institutions (2). Our program was able to obtain information from almost 24% of farms in Ontario during a short period of time and from a relatively small number of practitioners.
In general, practitioners did not have problems reporting general information requested for all visit/call records. In addition, no problems were observed with fields associated with type of production, area affected, body system, and the production parameter affected in a disease report. Practitioners consistently reported additional options for these elements. Adding more options for both, body systems, production parameters affected, or some specific clinical presentations, such as wasting and sudden death, should be considered.
Actual estimates of mortality were reported in different manners making it difficult to analyze these data. For syndromic surveillance systems for species like swine, it might be more relevant to record changes in production and reproductive parameters, such as increase in mortality, increase in morbidity, and poor farrowing rate, because an increase or decrease of these parameters may indicate the pre-symptomatic presentation of a new disease. Veterinary-based surveillance for livestock has been implemented mainly for cattle, where individual treatment of animals is commonly performed. However, it has to be recognized that for species such as pigs, the assessment of the herd is more important than that of individual animals (32). In addition, providing actual estimates of farm size and the number of animals affected may not be practical based on the structure of many swine farms and the manner in which diseases are diagnosed by swine practitioners.
Only 22% of farm reports indicated whether or not samples had been sent to the diagnostic laboratory. This would compromise proposed statistical cluster detection methods, since disease clusters would be difficult to validate with laboratory data. Whether a disease was a new/incident or an ongoing situation was noted in most of the disease records; however, this field proved to be difficult because differences in defining incident versus ongoing cases may create biases for analyses intended to capture incident as opposed to prevalent cases of disease. In addition, whether or not the disease responded to antimicrobials or other therapies was not listed in all disease records. This field was included to identify issues associated with treatment failure, including the emergence of antimicrobial resistance or the introduction of a novel pathogen. Further modification and education concerning the recording of this information is required. We may be required to extract these data based on the frequency and interval that health events are reported from a particular farm.
Our results showed that obtaining information from veterinarians can be a valuable, reliable, and a relatively inexpensive manner to collect data from a large proportion of Ontario’s swine farms. Most of the veterinarians participating in the program seemed to accept the program and agreed that this type of veterinary-based surveillance system is appropriate for detecting emerging or re-emerging pathogens. Currently, the timeliness of data transmission is too slow to identify disease trends on a weekly basis. In addition, drop-offs in compliance and difficulties defining incident cases may complicate the interpretation of statistical cluster detection methods. Regular communication with practitioners, user-friendly recording systems that can be adapted to each clinic’s management system, the involvement of clinic technical personnel in the transmission of data, and financial incentive programs may help to improve compliance and timeliness. This anonymized, yet temporally explicit, spatially informative, and quantitative demonstration of surveillance coverage, through regular and routine physical observation of animals on farms by licensed veterinarians, is very powerful data. It could potentially support regulatory authorities’ claims of freedom from serious diseases and demonstrates the ability to detect serious animal diseases. This in turn supports trade, the swine agri-food industry and veterinarians. A broader application of this mutually beneficial approach would be useful.
Acknowledgments
The authors thank the Agriculture and Agri-Food Canada; Food Safety Initiative (FSI); the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA); and the Ontario Animal Health Strategic Investment (AHSI) for funding. We also thank Dr. Beverly McEwen, a veterinary pathologist and disease surveillance specialist from the Animal Health Laboratory, Laboratory Services Division, and University of Guelph for providing the data associated with the swine submissions made to the AHL. We particularly thank the veterinarians that were willing to participate in the OSVS program.
Appendix 1. Data collection form used in the OSVS pilot study
References
- 1.Brown C. Emerging zoonosis and pathogens of public health significance — An overview. Revue Scient Tech (Office international des épizooties) 2004;23:435–442. doi: 10.20506/rst.23.2.1495. [DOI] [PubMed] [Google Scholar]
- 2.Department for Environment Food Rural Affairs. Assessing the utility of veterinary practitioner-derived surveillance data: Clinical examination of casualty slaughter cattle survey 2005. UK: DEFRA; [Last accessed July 12, 2010]. Project Code TS5018-G. [monograph on the Internet]. Available from http://www.defra.gov.uk/foodfarm/farmanimal/diseases/documents/ts5018-report160705.pdf. [Google Scholar]
- 3.McIntyre LH, Davies PR, Alexander G, et al. VetPAD-veterinary practitioner aided disease surveillance system. Proc of the 10th Int Symp Vet Epidem Econom. 2003 [Google Scholar]
- 4.Alberta Veterinary Service Network (AVSN) Alberta Government [homepage on the Internet] c 2005–2009. [Last accessed July 12, 2010]. Available from http://www1.agric.gov.ab.ca/$department/deptdocs.nsf/all/afs10440#what.
- 5.Bovine Syndromic Surveillance System (BOSSS), Australian Biosecurity, Cooperative for Emerging Infectious Diseases. AusVet Animal Health Services. [homepage on the Internet] c 2003. [Last accessed February 2007]. Available from http://www.ausvet.com.au/bosss/
- 6.Shephard RW. Doctor in Philosophy. Australia: Faculty of Veterinary Science, University of Sidney; 2006. The development of a syndromic surveillance system for the extensive beef cattle producing regions of Australia. [Google Scholar]
- 7.deGroot BD. The Rapid Syndrome Validation Project for Animals — Augmenting Contact with the Network of Accredited Veterinarians (RSVP-A) The National Animal Health Surveillance System (NAHSS) APHIS, USDA [monograph on the Internet] [Last accessed July 12, 2010]. c April 2005. Available from http://www.aphis.usda.gov/vs/ceah/ncahs/nsu/outlook/issue5/RSVP_A.pdf.
- 8.Davies PR, Wayne SR, Torrison JL, Peele B, de Groot BD, Wray D. Real-time disease surveillance tools for the swine industry in Minnesota. Vet Ital. 2007;43:731. [PubMed] [Google Scholar]
- 9.Gagnon CA, Tremblay D, Tijssen P, Venne MH, Houde A, Elahi SM. The emergence of porcine circovirus 2b genotype (PCV-2b) in swine in Canada. Can Vet J. 2007;48:811–819. [PMC free article] [PubMed] [Google Scholar]
- 10.Carman S, McEwen B, Josephson G, Fairles J. PRRSV outbreak in southwestern Ontario. [monograph on the Internet] [Last accessed June 30, 2010];Animal Health Laboratory (AHL) Newsletter. 2005 9:6. Available from http://www.uoguelph.ca/labserv/units/ahl/files/ANwsl9-4.pdf. [Google Scholar]
- 11.Karasin AI, Carman S, Olsen CW. Identification of human H1N2 and human-swine re-assortant H1N2 and H1N1 influenza A viruses among pigs in Ontario, Canada (2003 to 2005) J Clin Microb. 2006;44:1123–1126. doi: 10.1128/JCM.44.3.1123-1126.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Carman S, McEwen B, DeLay J, et al. Porcine circovirus-2 associated disease in swine in Ontario (2004 to 2005) Can Vet J. 2006;47:761–762. [PMC free article] [PubMed] [Google Scholar]
- 13.Dufour B, La Vieille S. Epidemiological surveillance of infectious diseases in France. Vet Res. 2000;31:169–185. doi: 10.1051/vetres:2000116. [DOI] [PubMed] [Google Scholar]
- 14.Dufour B, Hendrikx P, Toma B. The design and establishment of epidemiological surveillance systems for high-risk diseases in developed countries. Rev Sci Tech (Office international des épizooties) 2006;25:187–198. doi: 10.20506/rst.25.1.1659. [DOI] [PubMed] [Google Scholar]
- 15.Chauvin P, Valleron AJ. Participation of French general practitioners in public health surveillance: A multidisciplinary approach. J Epidem Comm Health. 1998;52:2S–8S. [PubMed] [Google Scholar]
- 16.Deckers JGM, Paget WJ, Schellevis FG, Fleming DM. European primary care surveillance networks: Their structure and operation. Fam Pract. 2006;23:151–158. doi: 10.1093/fampra/cmi118. [DOI] [PubMed] [Google Scholar]
- 17.Allport R, Mosha R, Bahari M, Swai E, Catley A. The use of community-based animal health workers to strengthen disease surveillance systems in Tanzania. Rev Sci Tech (Office international des épizooties) 2005;24:921–932. [PubMed] [Google Scholar]
- 18.Mandl KD, Overhage JM, Wagner MM, et al. Implementing syndromic surveillance: A practical guide informed by the early experience. J Am Med Inform Assoc. 2004;11:141–150. doi: 10.1197/jamia.M1356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Clothier HJ, Fielding JE, Kelly HA. An evaluation of the Australian sentinel practice research network (ASPEN) surveillance for influenza-like illness. CDI. 2005;29:231–247. doi: 10.33321/cdi.2005.29.22. [DOI] [PubMed] [Google Scholar]
- 20.Buehler JW Centers for Disease Control and Prevention (CDC) Review of the 2003 national syndromic surveillance conference-lessons learned and questions to be answered. MMWR Morbidity and mortality weekly report. 2004;53:18–22. [PubMed] [Google Scholar]
- 21.de Wal AH, Smith R, van der Werf G, Meyboom-de JB. Towards improvement of the accuracy and completeness of medication registration with the use of an electronic medical record (EMR) Fam Pract. 2001;18:288–291. doi: 10.1093/fampra/18.3.288. [DOI] [PubMed] [Google Scholar]
- 22.Jajosky RA, Groseclose SL. Evaluation of reporting timeliness of public health surveillance systems for infectious diseases. BMC Public Health. c 2004. [Last accessed June 30, 2010]. [monograph on the Internet]. Available from http://www.biomedcentral.com/content/pdf/1471-2458-4-29.pdf. [DOI] [PMC free article] [PubMed]
- 23.Overhage JM, Grannis S, McDonald CJ. A comparison of completeness and timeliness of automated electronic laboratory reporting and spontaneous reporting of notifiable conditions. Am J of Public Health. 2008;98:344–350. doi: 10.2105/AJPH.2006.092700. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dohoo I, Martin W. Veterinary Epidemiology Research. Charlottetown, Prince Edward Island, Canada: AVC Inc; 2003. [Google Scholar]
- 25.Statistics Canada. 2006 Census of Agriculture. c 2006. [Last accessed June 30, 2010]. [homepage on the Internet]. Available from http://www.statcan.gc.ca/pub/95-629-x/3/4182418-eng.htm.
- 26.Smith-Gagen J, Cress RD, Drake CM, Felter MC, Beaumont JJ. Factors associated with time to availability for cases reported to population-based cancer registries. Cancer Causes and Control. 2005;16:449–454. doi: 10.1007/s10552-004-5030-0. [DOI] [PubMed] [Google Scholar]
- 27.Chauvin P, Valleron A. Monitoring the compliance of sentinel general practitioners in public health surveillance: Which GPs persevere? Int J Epidemiol. 1997;26:166–172. doi: 10.1093/ije/26.1.166. [DOI] [PubMed] [Google Scholar]
- 28.Randall T. National morbidity data have their beginnings in offices of private practice physicians. J Am Med Assoc. 1990;263:2564–2566. [PubMed] [Google Scholar]
- 29.VetPAD, Veterinary Practitioner Aided Disease Surveillance System (VetPAD) c 2006. [Last accessed June 30, 2010]. [homepage on the Internet]. Available from http://www.vetpad.co.nz/Home.asp.
- 30.Vourc’h G, Bridges VE, Gibbens J, et al. Detecting emerging diseases in farm animals through clinical observations. Emerg Infect Dis. 2006;12:204–210. doi: 10.3201/eid1202.050498. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Zhang J, Calvo R, Shephard R, Jin C. Detecting diseases in farm animals with embedded system. Proc of the WSEAS International Conference on Computer Engineering and Application; Gold Coast, Australia. 2007. pp. 52–56. [Google Scholar]
- 32.Dewey CE, Straw BE. Herd examination. In: Straw BE, Zimmerman JJ, D’allaire S, Taylor DJ, editors. Diseases of Swine. 9th ed. Ames, Iowa: Blackwell Publ; 2006. pp. 3–13. [Google Scholar]