Department of Clinical Sciences, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, PO Box 7054, SE-750 07 Uppsala, Sweden.
Acta Vet Scand. 2009 Dec 10;51(1):50. doi: 10.1186/1751-0147-51-50.
Computerized diagnostic information offers potential for epidemiological research; however data accuracy must be addressed. The principal aim of this study was to evaluate the completeness and correctness of diagnostic information in a computerized equine clinical database compared to corresponding hand written veterinary clinical records, used as gold standard, and to assess factors related to correctness. Further, the aim was to investigate completeness (epidemiologic sensitivity), correctness (positive predictive value), specificity and prevalence for diagnoses for four body systems and correctness for affected limb information for four joint diseases.
A random sample of 450 visits over the year 2002 (nvisits=49,591) was taken from 18 nation wide clinics headed under one company. Computerized information for the visits selected and copies of the corresponding veterinary clinical records were retrieved. Completeness and correctness were determined using semi-subjective criteria. Logistic regression was used to examine factors associated with correctness for diagnosis.
Three hundred and ninety six visits had veterinary clinical notes that were retrievable. The overall completeness and correctness were 91% and 92%, respectively; both values considered high. Descriptive analyses showed significantly higher degree of correctness for first visits compared to follow up visits and for cases with a diagnostic code recorded in the veterinary records compared to those with no code noted. The correctness was similar regardless of usage category (leisure/sport horse, racing trotter and racing thoroughbred) or gender.For the four body systems selected (joints, skin and hooves, respiratory, skeletal) the completeness varied between 71% (respiration) and 91% (joints) and the correctness ranged from 87% (skin and hooves) to 96% (respiration), whereas the specificity was >95% for all systems. Logistic regression showed that correctness was associated with type of visit, whether an explicit diagnostic code was present in the veterinary clinical record, and body system. Correctness for information on affected limb was 95% and varied with joint.
Based on the overall high level of correctness and completeness the database was considered useful for research purposes. For the body systems investigated the highest level of completeness and correctness was seen for joints and respiration, respectively.
计算机化的诊断信息为流行病学研究提供了潜力；然而，必须解决数据准确性问题。本研究的主要目的是评估与作为金标准的手写兽医临床记录相比，计算机化马临床数据库中诊断信息的完整性和正确性，并评估与正确性相关的因素。此外，目的还在于调查四个身体系统的诊断的完整性（流行病学敏感性）、正确性（阳性预测值）、特异性和患病率，以及四个关节疾病的受影响肢体信息的正确性。
从一家公司下属的 18 家全国性诊所中抽取了 2002 年一年的 450 次就诊的随机样本（nvisits=49,591）。选择了选定就诊的计算机化信息，并检索了相应的兽医临床记录的副本。使用半主观标准确定完整性和正确性。使用逻辑回归检查与诊断正确性相关的因素。
有 396 次就诊有可检索的兽医临床记录。总体完整性和正确性分别为 91%和 92%；这两个值都被认为是很高的。描述性分析表明，与随访就诊相比，首次就诊的正确性显著更高，与兽医记录中有诊断代码记录的病例相比，无代码记录的病例的正确性更高。无论使用类别（休闲/运动马、赛马和赛马）或性别如何，正确性相似。对于选择的四个身体系统（关节、皮肤和蹄子、呼吸、骨骼），完整性在 71%（呼吸）和 91%（关节）之间变化，正确性范围从 87%（皮肤和蹄子）到 96%（呼吸），而特异性对于所有系统均>95%。逻辑回归表明，正确性与就诊类型、兽医临床记录中是否存在明确的诊断代码以及身体系统有关。受影响肢体信息的正确性为 95%，并随关节而变化。
基于整体高水平的正确性和完整性，该数据库被认为可用于研究目的。对于所研究的身体系统，关节和呼吸的完整性和正确性分别达到最高水平。
