Medical registries are fundamental in understanding factors that may improve patient care or prognosis within health care systems. The usefulness of a given registry is dependent upon the quality of data collected; therefore, several authors have described methods for characterizing data quality within large databases.1–4 Clinical research registries including Trauma registries5 serve as collective sources of information to conduct studies that provide guidance toward improving patient care, clinical decisions, and treatment outcomes.6 The advent of a dedicated veterinary trauma research registry has provided guidance toward the future improvement of trauma patient care.1,3 While these registries may contribute toward the wealth of knowledge in medical literature, information accumulated within these research databases is prone to error and data misinterpretation.7 Assurance of data validation, quantification of errors, and an understanding of the influence of quality control measures on registry data is of fundamental importance for meaningful interpretation of the data provided.8 Thorough assessment of data quality in research involving health records tends to be expensive and time-consuming. Nevertheless, it is important that the necessary effort is made toward validating and refining research data, as erroneous data could lead to unfavorable health care outcomes. Although evaluations of the effects of errors on research implications have not been assessed, it is expected that error issues could lead to incorrect conclusions.1,3,7 The development of a standardized and streamlined approach to the assessment of the data quality in clinical research registries is an important step in the process of establishing a continuous improvement in data quality for research and relevant outcomes.
The Veterinary Committee on Trauma (VetCOT)9 has established a research registry with REDCap (Research Electronic Data Capture) that collects information from over 30 identified and verified Veterinary Trauma Centers (VTCs) primarily across the United States and Canada with 1 location in the United Kingdom. With the primary aim to improve trauma patient care and outcomes, the Veterinary Committee on Trauma (VetCOT) was formed to create a network of lead hospitals that seed the development of trauma systems. Information on over 50,000 canine and feline trauma cases has been logged since September 2013, and numerous research studies10–13 have been undertaken utilizing this data.
The purpose of this study was to generate a quality control assessment tool for the trauma database (VetCOT here) based on the following quality control parameters: accuracy, completeness, and consistency. The VetCOT data will be tested and compared to the Electronic Medical Records (EMR) collected from the veterinary centers. “Accuracy” of data quality exists when data conform with a verifiable source, “completeness” exists when all necessary data are provided, and “consistency” exists when data are logical across data points.14 Studies have demonstrated that clinical registries are prone to error,7 which is especially evident through assessments of accuracy, completeness, and consistency.4 By utilizing a quality control assessment, a standardized workflow may be developed so that a framework to efficiently assess data quality within the VetCOT trauma registry may be further implemented by other research studies. An additional focus in our analysis is on clinically relevant information. This includes data collected that is pertinent to patient prognosis as established by previous studies in the scientific literature.15–18 Coding algorithms developed in this project may be readily utilized by personnel with a basic familiarity with the R programming environment. The workflow has predefined parameters that are adjustable for a quality control assessment that is specific to individual research needs. There is information on certain parameters that are not collected in the EMR such as mGCS and ATT, however, are necessary for the VetCOT registry. Through the development of this analysis tool, RVetQual, it is expected that further quality control assessment of the VetCOT trauma registry may be streamlined so that characterization of the data quality produced by the registry may be done efficiently and through a process that may be modified for other research needs. A better understanding of the current state of data quality will allow for proper evaluation of research implications and continuous improvement of subsequent trauma case entries.
Methods
Analysis software was developed in the R programming language19 and named RVetQual. The program codes were published on the Github repository with the following link: https://github.com/cvmbsOne/VetQualPub.
The data from VetCOT and EMR were exported into Excel worksheets. The analysis workflow using RVetQual aimed to assess the following quality control dimensions pertaining to the data obtained from VetCOT trauma registry: (1) accuracy, (2) completeness, and (3) consistency.
Accuracy
The accuracy of the veterinary trauma registry was evaluated through comparison with data extracted from the VTC’s EMR for each case. Cases from the VetCOT trauma registry logged between April 2017 and December 2018 were extracted from REDCap for 6 participating university or commercial VTCs: Colorado State University (CSU), University of Minnesota (UMN), University of Pennsylvania (UPenn), The Ohio State University (OSU), MedVet Chicago, and Auburn University (Au). A random subset of these cases was selected using the R program (with Sample() function) to undergo manual review (357 cases, based on a sample size calculation to find a 50% error in the data discrepancy (50% or 0.5 is standard conservative practice for the probability of an event that represents the highest variability expected in a population when the variability is unknown) with a 95% confidence level, 5% margin of error). A spreadsheet in Microsoft Excel with 2 REDCap unique identifiers (ID and case number) per case was sent to a representative from each of the 6 participating VTCs. Using the unique identifier information (ID), 4 out of the 6 representative VTCs provided their EMR corresponding to the requested cases in a spreadsheet format. The information exported from REDCap and that compared to EMR data, both in free text format, included the following variables for all cases: species, age, breed, sex, body weight, type of trauma, cause of trauma, date/time of trauma, date/time presentation, operational canine (OpK9),20 prior (Doctor of Veterinary Medicine (DVM) care, prior non-DVM care, modified Glasgow Coma Scale (mGCS), Animal Trauma Triage (ATT) score, Intensive Care Unit (ICU) hospitalization, head or spinal injury present, surgical procedures, blood product administered, and outcome. Certain additional information was exported as well, for those cases having such extra information. The data in both REDCap and EMR were spot-checked for any dissimilarities in formatting or spell-checks. We termed such additional information as “Optional” data because it is data that is not always provided in each trauma center’s EMR. Therefore, it is the information that is recorded specifically for the VetCOT database, but not necessarily included in the original EMR of each case record. This includes the mGCS and ATT scores, which are mandatory in the VetCOT trauma registry, but not always recorded in the original EMR. This is a limitation of using the trauma center’s EMR because not all information provided in the trauma registry is mandatory in those records. The optional data was collected through a separate random subset of 30 cases that included all of the following optional data per case in REDCap: Abdominal Focused Assessment with Sonography for Trauma (AFAST) or Thoracic Focused Assessment with Sonography for Trauma (TFAST) performed, packed cell volume (PCV), total solids (TS), blood glucose (BG), blood lactate (BL), ionized calcium (iCa), and base excess (BE). Because mGCS and ATT scores were not mandatory EMR entries by most VTCs, the subset of comparisons for these variables included 30 cases.
With VetCOT exported Excel files and EMR data as input, RVetQual generated an output file in Excel format containing the following information: matches per variable, mismatches per variable, missing information from the EMR per variable, percent error (number of inaccurate values/total number of reported values*100), and measures of discrepancy (EMR value- VetCOT value). Matches per variable were defined as values that match between both REDCap and hospital data. Mismatches per variable account for the number of values that do not match between REDCap and hospital data when corresponding values are present. Missing values are also identified, which are defined as specific instances where either REDCap or a hospital provided no data on certain variables.
RVetQual software identified cases that may have been erroneously uploaded multiple times into REDCap for the same case presentation. It identified filtered cases with a repeated case number (representing a specific animal). If a case number is repeated, then a follow-up query determined whether the same presentation date had been entered for the multiple case entries. If the case number and presentation date were duplicated, then it indicated the data entry error. The most common discrepancy in case ID matching was the absence of VTC’s name as a prefix in the EMR. An additional step to add VTC’s name to the ID as a prefix in EMR was taken to try to ensure that the matching of unique IDs between REDCap and VTCs was correctly implemented. RVetQual calculated the total number of mismatches between REDCap and each VTC’s EMR per case. Furthermore, RVetQual output identified whether species and presentation date were mismatched between REDCap and the EMR. Cases were only excluded from this study if the unique ID (case number) on the EMR was a mismatch to that found on REDCap or if the case was present only in EMR or REDCap.
Discrepancy measures were applicable to continuous variables where the following calculations were made per variable: minimum error, maximum error, and median error. Minimum, maximum, and medians were also converted to percentages, and absolute values of the differences are calculated separately. Discrepancies comprised of calculations with and without the inclusion of absolute value standardization for potential identification of error trends.
Consistency
The consistency of the VetCOT trauma registry was evaluated through the comparison of multiple variables within each record. RVetQual attempts to screen for improbable scenarios between variables within each case. These screenings consisted of the following: (1) Date of presentation before the date of trauma, (2) Dates of the outcome before the dates of presentation, (3) small dog breeds >10 kg, (4) cats >10 kg, (5) all dogs >40 kg along with consideration with their breed, and (6) large dog breeds >3 months of age and <15 kg. For the breed-wise consistency analysis, only purebred dogs were considered. Mixed-breed canines were excluded from this analysis. Small dog breeds included 45 breeds characterized within the smallest purebred canine groups by the American Kennel Club (AKC)21 (https://www.akc.org/dog-breeds/). Large dog breeds included 38 breeds characterized within the largest purebred canine groups by AKC.
Completeness
Case completeness for records entered into the VetCOT registry was determined by measuring how many data points were recorded out of total possibilities per variable. RVetQual Excel output showed the total entries, total fields left blank and percent completeness (number of recorded values/total number of data points*100) per variable.
Clinical relevance
RVetQual identified clinically relevant differences of case data only from VetCOT that would dictate a worse prognosis as defined by the scientific literature.16–18,22 This included cases with the following information: mGCS score <18, ATT score >0, iCa <1.24 (mmol/L), BE < −6.6 (mmol/L), and AFAST Fluid score >2.
Validation of the RVetQual
The RVetQual tool was validated through a comparison between results acquired manually (described above in the Accuracy section) vs that automatically generated by the coding scheme. The manual assessment included a comparison of accuracy and all aforementioned calculations. The automatically generated results were compared to that acquired manually and adjustments to the tool were made until all comparisons matched.
Results
Data extraction
Exportation of REDCap data between April 2017 and December 2018 yielded all data collected on 5,000 trauma cases representing 6 VTCs during these time periods in Excel format. A randomly selected sample size of 357 of these cases was requested from all VTC’s EMRs for manual review and comparison. Of the 357 cases, 210 EMRs (= 58.8%) were received from only 4 VTCs on time during the period of study. Out of those 210 EMRs, 30 cases had optional variables. Data from the remaining 180 EMRs were manually entered on to an Excel spreadsheet alongside the corresponding information already exported from REDCap. However, the results from the 210 cases are presented (Tables 1–3).
Example output of accuracy results showing total number of comparisons between VetCOT and EMR data, as well as the matches, mismatches, percent mismatched and missing values
| Variable | Comparisons possible | Matches | Mismatches | Mismatched (%) | Missing values* |
|---|---|---|---|---|---|
| ID | 210 | 210 | 0 | 0 | 0 |
| Presentation date | 180 | 166 | 14 | 7.78 | 30 |
| Species | 180 | 180 | 0 | 0 | 30 |
| Canine breed | 152 | 123 | 29 | 19.08 | 58* |
| Cat breed | 28 | 26 | 2 | 7.14 | 182* |
| Canine age | 149 | 79 | 70 | 46.98 | 61* |
| Feline age | 27 | 16 | 11 | 40.74 | 183* |
| Sex | 180 | 152 | 28 | 15.56 | 30 |
| Canine weight | 149 | 101 | 48 | 32.21 | 61* |
| Feline weight | 26 | 14 | 12 | 46.15 | 184* |
| Prior presentation to DVM? | 180 | 167 | 13 | 7.22 | 30 |
| Part of operational K9 unit? | 180 | 180 | 0 | 0 | 30 |
| Prior treatment by non-DVM? | 180 | 172 | 8 | 4.44 | 30 |
| Trauma type | 174 | 144 | 30 | 17.24 | 36 |
| Blunt trauma type | 84 | 62 | 22 | 26.19 | 126 |
| Penetrating trauma type | 83 | 73 | 10 | 12.05 | 127 |
| Trauma date entered? | 180 | 144 | 36 | 20 | 30 |
| Trauma time entered? | 180 | 104 | 76 | 42.22 | 30 |
| Presentation time known? | 180 | 174 | 6 | 3.33 | 30 |
| Intensive care required? | 179 | 167 | 12 | 6.7 | 31 |
Missing values counts the number of fields without information entered, which includes incomplete data in species-specific entries. N = 210 cases. VetCOT = Veterinary Committee on Trauma. EMR = Electronic Medical Records. ID = Identifier. DVM = Doctor of Veterinary Medicine. K9 = Canine
Example output of the differences (min, max, and medians) between mismatches of continuous data, the absolute values of the percent differences.
| Variable | Min diff | Max diff | Median diff | Min % diff | Max % diff | Median % diff |
|---|---|---|---|---|---|---|
| Canine age (years) | −10.1 | 10 | 0.1 | 0.8 | 10,000 | 9.09 |
| Feline age (years) | −1 | 0.3 | 0.1 | 0.73 | 34.48 | 3.45 |
| Canine weight (kg) | −14.2 | 25.6 | 0.03 | 0.13 | 100 | 4.26 |
| Feline weight (kg) | −0.13 | 4.1 | 0.035 | 0.29 | 100 | 1.595 |
| Blood lactate (mmol/L) | −0.04 | −0.02 | −0.03 | 0.58 | 1.52 | 1.05 |
| Base excess (mmol/L) | −0.1 | −0.1 | −0.1 | 100 | 100 | 100 |
| Ionized calcium (mmol/L) | −0.01 | 0.01 | 0 | 0.71 | 0.78 | 0.745 |
| PCV (%) | −6 | 35 | 4 | 6.52 | 159.09 | 12.5 |
| TS (g/dL) | −3.6 | 2.1 | −0.05 | 5.33 | 52.78 | 18.02 |
| Blood glucose (mmol/L) | −141 | −19 | −55 | 16.52 | 83.43 | 49.55 |
Min diff/median diff/max diff are calculated based on the continuous variable’s units, whereas all percent columns are the calculated differences in percentage points. PCV = Packed cell volume. TS = Total solids.
Example output of consistency results from small dogs (filter applied: small dogs > 10 kg)
| Unique case ID | Dog breed code | Weight (kg) | Age (years) | Breed name |
|---|---|---|---|---|
| CN1 | 5192 | 10.3 | 5.1 | Pug |
| CN2 | 5134 | 10.7 | 3.5 | Italian Greyhound |
| CN3 | 5067 | 13.3 | 4.3 | Cocker Spaniel |
| CN4 | 5164 | 10.6 | 7 | Miniature Schnauzer |
| CN5 | 5192 | 10.9 | 12 | Pug |
| CN6 | 5009 | 11.3 | 1.3 | American Eskimo |
| CN7 | 5163 | 11.3 | 9.5 | Miniature Poodle |
| CN8 | 5041 | 11.3 | 6.1 | Boston Terrier |
| CN9 | 5208 | 11.2 | 1.4 | Shiba Inu |
| CN10 | 5135 | 11.8 | 9.6 | Jack Russell Terrier |
| CN11 | 5009 | 15.2 | 13 | American Eskimo |
| CN12 | 5207 | 11.8 | 14.3 | Shetland Sheepdog |
| CN13 | 5207 | 11.1 | 9.5 | Shetland Sheepdog |
| CN14 | 5041 | 15 | 4.3 | Boston Terrier |
| CN15 | 5067 | 19 | 0.9 | Cocker Spaniel |
| CN16 | 5208 | 20 | 14 | Shiba Inu |
| CN17 | 5192 | 12.8 | 10 | Pug |
| CN18 | 5204 | 14.6 | 13.5 | Scottish Terrier |
| CN10 | 5057 | 12.4 | 0.9 | Cavalier King Charles Span |
Accuracy
Output included the total number of comparisons between VetCOT and EMR data, as well as the matches, mismatches, percent error, and the number of variables without 2 data points to compare (ie, missing values; Table 1). The percent mismatch varied among all variables anywhere from 0% for ID to 46.98% for canine age. The age and weight of canines and felines and trauma time showed higher mismatches (47% and 41% for age; 32% and 46% for weight among canines and felines, respectively; 42% for trauma time). Additionally, the differences (min, max, and medians) between mismatches of continuous data are shown (Table 2). The absolute values of the percent differences are also shown (Table 2). These results are presented with a summary of all VTCs pooled together. The highest median percent difference was observed with BE, followed by blood glucose and then TS. Accidental duplicate case entries into REDCap were screened by unique identifier and presentation date. This led to the identification of 41/5,000 (0.82%) cases that were accidentally entered more than once into REDCap. For each unique case ID, there was an output of total variables screened, number of mismatches, and critical factors (species and presentation date). Lastly, comparison data of 1 variable as an example (PCV selected in this study) included an output of the mismatching values between REDCap and medical records alongside clinically relevant information (blood product use and outcome).
Consistency and completeness
Such information does not involve a comparison between REDCap and medical record data, it analyzes each case from each source. Outputs of case data included the variables assessed for consistency. Screening of cases by the criteria applied showed data being consistent across breed code, breed name, age, and weight. An example of the sample output from small dogs is presented (Table 3). Information regarding completeness included total fields, total entries per variable, total fields left blank, and a calculation of percent completeness is presented (Table 4). Among a total of 5,000 VetCOT case data percent completeness ranged from 95.82% for feline weight to 100% for ID, VTC, entry data, and species.
Example output of completeness results included total variables, total entries per variable, total fields left blank and percent completeness (15/102 variables) out of 5,000 cases.
| Variable | Total entries | Fields left blank | Percent completeness |
|---|---|---|---|
| ID | 5000 | 0 | 100 |
| Date of hospitalization | 4999 | 1 | 99.98 |
| VTC | 5000 | 0 | 100 |
| Entry date | 5000 | 0 | 100 |
| Case number | 5000 | 0 | 100 |
| Species | 5000 | 0 | 100 |
| Canine breed | 4198 | 12 | 99.71 |
| Feline breed | 781 | 9 | 98.86 |
| Canine age | 4203 | 7 | 99.83 |
| Feline age | 779 | 11 | 98.61 |
| Sex | 4994 | 6 | 99.88 |
| Canine weight | 4181 | 29 | 99.31 |
| Feline weight | 757 | 33 | 95.82 |
| Prior presentation to DVM? | 4991 | 9 | 99.82 |
| Part of operational K9 unit? | 4996 | 4 | 99.92 |
Clinically relevant information
An additional output identifies cases meeting the following criteria with VetCOT registry data: MGCS score <18, ATT score >0, ionized calcium <1.24 mmol/L, base Excess < −6.6 mmol/L, and AFAST Fluid score >2. The total number of cases filtered through such criteria indicating the total number of cases filtered by mGCS and ATT were 776 and 3972, respectively, is summarized (Table 5). Those cases that died or were euthanized ranged between 8.3% to 50%.
Summary of the total number of cases filtered by factors of lower prognosis (out of 5,000 cases).
| Filter criteria | Total cases filtered | Number euthanized or died in hospital | Percent nonsurvival (%) |
|---|---|---|---|
| mGCS <18 | 776 | 316 | 40.7 |
| ATT >0 | 3,972 | 421 | 10.6 |
| iCa <1.24 mmol/L | 240 | 20 | 8.3 |
| BE < −6.6 mmol/L | 175 | 53 | 30.3 |
| AFAST fluid score >2 | 18 | 9 | 50 |
Discussion
The quality analysis workflow designed in this study may allow researchers to efficiently assess the accuracy, consistency, and completeness of utilized VetCOT data. Because a secondary data source must be provided to analyze data matching, the assessment of accuracy may be limited by researchers due to an inability to access corresponding hospital trauma data. However, the assembly of case information and utilization of RVetQual can productively identify mismatching data and characterize the extent of difference between mismatching continuous data. Furthermore, RVetQual can identify duplicate case entries which can facilitate the removal of superfluous case data. Some human trauma registry studies have shown that the erroneous data vary among centers, leading to heterogeneity in data quality, and suggest that targeted, data-driven, and educational opportunities exist at the institutional level.2,4,7 Similarly, RVetQual could be used as a tool to check the data quality of the trauma database time-to-time to identify the key areas of improvement in data capture in VetCOT and EMR.
Accuracy is one of the most commonly studied quality control parameters3 and tends to be assessed through the evaluation of parallel sources of information.23 The analysis model established in this study allowed for this to be accomplished more efficiently, but still has its limitations. The number of samples was almost half of the required sample size of 357 due to the unavailability of EMR for all. Although the final sample size of 210 was below our sample size calculation, our goal was not to test a hypothesis on the error rate. Rather, it was to test and validate the RVetQual program. Accuracy could be quantified for about half of VetCOT variables, while the rest were not included due to an inability to ascertain this information through hospital medical records received. For instance, information collected by VetCOT such as the specific provider of non-DVM care before trauma presentation is typically unobtainable through a hospital record. Regardless, if access to both REDCap and hospital medical records is possible, there are multiple measures of accuracy that may be efficiently generated through the use of this workflow. Canine and feline age, weight, and trauma time were the most mismatched variables in the study indicating that data entry should be carefully performed and the person performing the data entry should be well-trained before the assignment. An advanced approach would be to allow data entry only in specific preassigned ways, in a specific format. For very important variables a double data entry could be required. Quantification of the level of differences between REDCap and hospital record data may allow for an understanding of the potential for such errors to affect research results. It is important to note that mismatches and differing continuous values are partial measures of the overall quality of the data sources, and identification of discrepancies between 2 sources does not necessarily identify which value is most accurate.
An assessment of completeness and consistency is readily provided through the RVetQual tool, as these require only REDCap data and not original medical record data. Consistency can be assessed by examining multiple variables in the same case.23 While analyses in this study address the consistency between variables such as age, weight, and breed, the appropriate analyses in other studies will vary depending on specific research questions. There were variables that showed a high proportion of missing values, which indicates high attention should be maintained during the data capture stage in VetCOT. RVetQual is freely modifiable so that the researchers familiar with R programming may modify it to fit research needs. The errors identified by the tool will provide the researchers an idea about their data quality and measure may possibly be taken to fix the errors. VetCOT is led by 1 of our co-authors K.H. The data from all trauma centers would benefit from the RVetQual program due to its quality check capacity. Understanding the completeness of the data collected is a measure of usefulness23 and this completeness is measured by the analysis tool. Significant data quality and heterogeneity issues remain in trauma registries.2 Part of it may be due to data submission by various trauma centers into these registries. All these centers have variable degrees of data quality including the types of variables included in their records. A similar finding was reported by the human registries.2,24 The fields in our study that showed the most incompleteness were canine and feline weight. Because the completeness assessment in our study spans all current REDCap variables, there is no need for modification of the analysis tool if working only with different REDCap exports.
Qualifiers for the screening of all case data pertaining to clinically relevant information were chosen based on factors indicative of a lesser prognosis of trauma patients.15,16,18,23,25–28 An analysis of these factors is readily accessible due to RVetQual.
In conclusion, an analysis tool “RVetQual” has been developed in the R programming environment to generate an assessment of the data quality of the VetCOT trauma database. This serves as an efficient methodology to generate an analysis of the quality of data collected between 2 databases. A careful review of such analysis may allow for the improvement of data quality through the removal of duplicated data and the identification of other data anomalies. Because this tool is modifiable, it may find uses for different research questions involving VetCOT data from REDCap and its’ assessment of accuracy, consistency, and completeness. The intended audience for RVetQual is primarily trauma database researchers. However, future avenues could be other registries that could follow a similar workflow to validate their registry data. Some of this analysis also requires the ability to access and assemble information from multiple hospital records. While limitations exist, it is hoped that utilization of this analysis tool will allow for an efficient assessment and, continuous improvements in the data quality of the VetCOT trauma registry, which can serve to improve the quality of information guiding trauma patient care.
Acknowledgments
The research on which this presentation is based used data from the Veterinary Committee on Trauma (VetCOT) Registry, and we are grateful to the Veterinary Trauma Centers that participated. The VetCOT assumes no responsibility for the interpretation of the Registry data. VetCOT Registry data are collected and managed using REDCap electronic data capture tools hosted at the University of Minnesota supported by Award Number UL1TR002494 from the National Center for Advancing Translational Sciences of the National Institutes of Health (NIH). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the NIH.29 The project was supported by grant funding from Colorado Clinical and Translational Research Institute (CCTSI).
References
- 1.↑
Hlaing T, Hollister L, Aaland M. Trauma registry data validation: essential for quality trauma care. J Trauma. 2006;61:1400–1407. doi:10.1097/01.ta.0000195732.64475.87
- 2.↑
Dente CJ, Ashley DW, Dunne JR, et al, GRIT Study Group. Heterogeneity in trauma registry data quality: implications for regional and national performance improvement in trauma. J Am Coll Surg. 2016;222(3):288–295. doi:10.1016/j.jamcollsurg.2015.11.035
- 3.↑
O’Reilly GM, Gabbe B, Moore L, Cameron PA. Classifying, measuring and improving the quality of data in trauma registries: a review of the literature. Injury. 2016;47:559–567. doi:10.1016/j.injury.2016.01.007
- 4.↑
Porgo TV, Moore L, Tardif PA. Evidence of data quality in trauma registries. J Trauma Acute Care Surg. 2016;80:648–658. doi:10.1097/TA.0000000000000970
- 5.↑
Moore L, Clark DE. The value of trauma registries. Injury, Int. J. Care Injured. 2008;39:686–695. doi:10.1016/j.injury.2008.02.023
- 6.↑
Hoque DME, Kumari V, Hoque M, Ruseckaite R, Romero L, Evans SM. Impact of clinical registries on quality of patient care and clinical outcomes: a systematic review. PLoS ONE. 2017;12(9):e0183667. doi:10.1371/journal.pone.0183667
- 7.↑
Goldberg SI, Niemierko A, Turchin A. Analysis of data errors in clinical research databases. AMIA Annu Symp Proc. 2008:2008:242–246.
- 8.↑
Lee CH, Yoon HJ. Medical big data: promise and challenges. Kidney Res Clin Pract. 2017;36(1):3–11. doi:10.23876/j.krcp.2017.36.1.3
- 9.↑
Hall K. VetCOT: the Veterinary Trauma Registry. Top Companion Anim Med. 2019;37:100365. doi:10.1016/j.tcam.2019.100365
- 10.↑
Bartling J, Whelan M, Sinnott-Stutzman V. Retrospective evaluation of elevator-related injuries in dogs (2015-2020): 13 cases. J Vet Emerg Crit Care (San Antonio). 2023;33(1):70–73. doi:10.1111/vec.13246
- 11.
Davros AM, Gregory CW, Cockrell DM, Hall KE. Comparison of clinical outcomes in cases of blunt, penetrating, and combination trauma in dogs: a VetCOT registry study. J Vet Emerg Crit Care (San Antonio). 2023;33(1):74–80. doi:10.1111/vec.13253
- 12.
Lee JA, Huang CM, Hall KE. Epidemiology of severe trauma in cats: an ACVECC VetCOT registry study. J Vet Emerg Crit Care (San Antonio). 2022;32(6):705–713. doi:10.1111/vec.13229
- 13.↑
Gregory CW, Davros AM, Cockrell DM, Hall KE. Evaluation of outcome associated with feline trauma: a Veterinary Committee on Trauma registry study. J Vet Emerg Crit Care (San Antonio). 2023;33(2):201–207. doi:10.1111/vec.13277
- 14.↑
Wang RY, Storey VC, Firth CP. A framework for analysis of data quality research. IEEE Trans Knowl Data Eng. 1995;7(4):623Y640. doi:10.1109/69.404034
- 15.↑
Hall KE, Holowaychuk MK, Sharp CR, Reineke E. Multicenter prospective evaluation of dogs with trauma. J Am Vet Med Assoc. 2014;244(3):300–308. doi:10.2460/javma.244.3.300
- 16.↑
Lisciandro GR, Lagutchik MS, Mann KA, et al. Evaluation of an abdominal fluid scoring system determined using abdominal focused assessment with sonography for trauma in 101 dogs with motor vehicle trauma. J Vet Emerg Crit Car. 2009;19(5):426–437. doi:10.1111/j.1476-4431.2009.00459
- 17.
Holowaychuk MK, Monteith G. Ionized hypocalcemia as a prognostic indicator in dogs following trauma. J Vet Emerg Crit Care (San Antonio). 2011;21(5):521–530. doi:10.1111/j.1476-4431.2011.00675.x
- 18.↑
Stillion JR, Fletcher DJ. Admission base excess as a predictor of transfusion requirement and mortality in dogs with blunt trauma: 52 cases (2007–2009). J Vet Emerg Crit Car. 2012;22(5):588–594. doi:10.1111/j.1476-4431.2012.00798
- 19.↑
R Core Team (2020). R: A Language and Environment for Statistical Computing, R version 3.6.2. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/
- 20.↑
Palmer L. Operational canine. Vet Clin North Am Small Anim Pract. 2021;51(4):945–960. doi:10.1016/j.cvsm.2021.04.011
- 21.↑
Dog Breeds. Accessed March 15, 2023. https://www.akc.org/Public-Education/Resources/General-Tips-Information/Dog-Breeds-Sorted-Groups/
- 22.↑
Ash K, Hayes GM, Goggs R, Sumner JP. Performance evaluation and validation of the animal trauma triage score and modified Glasgow Coma Scale with suggested category adjustment in dogs: a VetCOT registry study. J Vet Emerg Crit Care (San Antonio). 2018;28(3):192–200. doi:10.1111/vec.12717
- 23.↑
Wang YR, Strong DM. Beyond accuracy: what data quality means to data consumers. J Manag Inform Syst. 1996;12:5–33. doi:10.1080/07421222.1996.11518099
- 24.↑
Mann NC, Guice K, Cassidy L, et al. Are statewide trauma registries comparable? Reaching for a national trauma dataset. Acad Emerg Med. 2006;13:946e953. doi:10.1197/j.aem.2006.04.019
- 25.↑
Hall, KE, Sharp CR, Adams CR, Beilman G. A novel trauma model: naturally occurring canine trauma. Shock. 2014;41(1):25–32. doi:10.1097/SHK.0000000000000058
- 26.
Kolata RJ, Kraut NH, Johnston DE. Patterns of trauma in urban dogs and cats: a study of 1,000 cases. J Am Vet Med Assoc. 1974;164(5):499–502.
- 27.
Lecky F, Woodford M, Edwards A, Bouamra O, Coats T. Trauma scoring systems and databases. Br J Anaesth. 2014;113(2):286–294. doi:10.1093/bja/aeu242
- 28.↑
Steinmetz S, Tipold A, Loscher W. Epilepsy after head injury in dogs: a natural model of posttraumatic epilepsy. Epilepsia. 2013;54(4):580–588. doi:10.1111/epi.12071
- 29.↑
Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap) - a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377–381. doi:10.1016/j.jbi.2008.08.010
