[Consistency evaluation between patient-completed and physician-completed Caprini scores]

Objective: To evaluate the consistency of patient-completed and physician-completed Caprini scores. Methods: This study was a diagnostic study. We prospectively recruited 200 inpatients (including respiratory and critical care medicine, rheumatology and immunology, obstetrics and gynecology, and orthopedics). Clinical data of the recruited patients were collected. The Wechat applet was developed based on the Chinese version of the patient-completed Caprini score. Patient could enter the Wechat applet by scanning the QR code, and enter the height, weight and other contents to the Wechat applet. The applet could automatically calculate the score and make the risk stratification according to total score. At the same time, physicians would calculate the traditional Caprini score for the same patient and make risk stratification to evaluate the consistency of scores derived from the two methods. Results: The average age of these 200 patients was (59.6±13.9) years, 112(56.0%) of them were female and 184(92.0%) with high school education or above. There was no significant difference between the patient-completed and physician-completed scores (4.8±2.5 vs. 4.7±2.5,P=0.336). The time of physician-completed score was shorter than that of patient-completed score ((2.0±1.0) minutes vs.(2.4±1.2) minutes, P<0.000 1). There was no significant difference on the number of high-highest venous thromboembolism risk patients assessed by the patient-completed and the physician-completed scores: 84.5% (169/200) vs. 83.0%(166/200)(χ²=0.165, P=0.684).There was strong positive correlations between patient-completed and physician-completed scores (r=0.98, P<0.000 1). Cohen's ĸ evaluation showed that the patient-completed Caprini score was in excellent consistency with physician-completed Caprini score(κ=0.97,P<0.000 1). The result of Bland Altman method showed that only 3.0% (6/200) of the scores biased greatly, which was not within the 95% confidence interval, the result proved that the bias belonged to a small probability event. It was inferred that the scores of patient-completed were consistent with those of the physician-completed. Conclusions: The patient-completed Caprini score is in good agreement with the physician-completed Caprini score in this patient cohort.