Bates D W, Cook E F, Goldman L, Lee T H
Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts.
Ann Intern Med. 1990 Oct 1;113(7):495-500. doi: 10.7326/0003-4819-113-7-495.
To develop and validate a model for the prediction of bacteremia in hospitalized patients, and to identify subgroups of patients with a very low likelihood of bacteremia in whom a positive blood culture has a low positive predictive value.
Prospective cohort study with clinical data on 1516 episodes collected from a random sample of all patients who had blood cultures done at one institution.
Urban, tertiary care hospital.
Derivation set: 1007 blood culture episodes sampled from all blood cultures done on patients at Brigham and Women's Hospital between October 1988 and February 1989. Validation set: 509 episodes, May 1989 to June 1989. The unit of evaluation was the episode, defined as a 48-hour period beginning after a blood culture was drawn.
True- and false-positive rates of blood cultures in the derivation set as assessed by independent reviewers were 7% (74 of 1007) and 8% (81 of 1007), respectively. Independent multivariate predictors of true bacteremia were temperature of 38.3 degrees C or higher, presence of a rapidly (less than 1 month) or ultimately (less than 5 years) fatal disease; shaking chills; intravenous drug abuse; acute abdomen on examination; and major comorbidity. In the low-risk group, defined by absence of these predictors, the misclassification rate of the model in the derivation set was 1% (4 of 303), and a positive blood culture had a positive predictive value of only 14% for true bacteremia. The model also identified a high-risk subset in which 16% (41 of 264) of episodes represented true bacteremia. The model was prospectively validated in 509 additional episodes, and the misclassification rate in the low-risk group was 2% (3 of 155).
None.
These findings provide a means of stratifying hospitalized patients according to their risk for bacteremia. If prospectively validated in other settings, this model may be helpful when deciding whether or not to do blood cultures or start antibiotic therapy and, when evaluating a positive blood culture, to determine whether or not it is a true-positive.
开发并验证一种用于预测住院患者菌血症的模型，并识别菌血症可能性极低的患者亚组，在这些亚组中血培养阳性的阳性预测值较低。
前瞻性队列研究，收集了从一家机构进行血培养的所有患者的随机样本中的1516次临床数据。
城市三级护理医院。
推导集：1988年10月至1989年2月期间在布莱根妇女医院对患者进行的所有血培养中抽取的1007次血培养事件。验证集：1989年5月至1989年6月的509次事件。评估单位是事件，定义为采血培养后开始的48小时时间段。
独立评审员评估推导集中血培养的真阳性率和假阳性率分别为7%（1007例中的74例）和8%（1007例中的81例）。真正菌血症的独立多变量预测因素为体温38.3摄氏度或更高、存在快速（不到1个月）或最终（不到5年）致命疾病、寒战、静脉药物滥用、检查时有急腹症以及严重合并症。在由这些预测因素不存在定义的低风险组中，推导集中模型的错误分类率为1%（303例中的4例），血培养阳性对真正菌血症的阳性预测值仅为14%。该模型还识别出一个高风险亚组，其中16%（264例中的41例）的事件代表真正菌血症。该模型在另外509次事件中进行了前瞻性验证，低风险组的错误分类率为2%（155例中的3例）。
无。
这些发现提供了一种根据住院患者菌血症风险进行分层的方法。如果在其他环境中得到前瞻性验证，该模型在决定是否进行血培养或开始抗生素治疗以及评估血培养阳性时，可能有助于确定其是否为真正阳性。
