Sakuma et al developed a logistic regression model to predict adverse drug events (ADE) in Japanese inpatients. The authors are from Kinki University, Brigham and Women’s Hospital and Harvard University using data from the JADE (Japanese Adverse Drug Events) study.
Patient selection: Japanase adult in the hospital
Parameters:
(1) physician
(2) scheduled operation
(3) dyspnea
(4) consciousness
(5) dementia
(6) hemiplegia
(7) cancer
(8) laxative use prior to admission
Parameter |
Finding |
Points |
physician |
resident |
0.43 |
|
post-resident attending |
0 |
scheduled operation |
no |
0 |
|
yes |
0.45 |
dyspnea |
no |
0 |
|
yes |
0.55 |
consciousness |
alert |
0.94 |
|
not alert |
0 |
dementia |
no |
0 |
|
yes |
1 |
hemiplegia |
no |
0 |
|
yes |
0.6 |
cancer |
no |
0 |
|
yes |
0.46 |
laxative use before admission |
no |
0 |
|
yes |
0.5 |
X =
= SUM(points for all 8 parameters) – 3
probability of an adverse drug event =
= 1 / (1 + EXP((-1) * X)
Interpretation:
• A probability >= 0.3 was considered high risk for an adverse drug reaction.
Performance:
• The area under the ROC curve is around 0.65 which is not very good but it may be helpful for screening.
Purpose: To determine the risk of an adverse drug event (ADE) in a Japanese adult inpatient using the model of Sakuma et al.
Specialty: Toxicology, Emergency Medicine, Critical Care
Objective: complications, complication detection
ICD-10: T88.7, Y40-Y59,