In a previous study, we developed a highly performant and clinically-translatable machine learning algorithm for a prediction of three-year conversion to Alzheimer’s disease (AD) in subjects with Mild Cognitive Impairment (MCI) and Pre-mild Cognitive Impairment. Further tests are necessary to demonstrate its accuracy when applied to subjects not used in the original training process. In this study, we aimed to provide preliminary evidence of this via a transfer learning approach.

We initially employed the same baseline information (i.e. clinical and neuropsychological test scores, cardiovascular risk indexes, and a visual rating scale for brain atrophy) and the same machine learning technique (support vector machine with radial-basis function kernel) used in our previous study to retrain the algorithm to discriminate between participants with AD (n = 75) and normal cognition (n = 197). Then, the algorithm was applied to perform the original task of predicting the three-year conversion to AD in the sample of 61 MCI subjects that we used in the previous study.

Even after the retraining, the algorithm demonstrated a significant predictive performance in the MCI sample (AUC = 0.821, 95% CI bootstrap = 0.705–0.912, best balanced accuracy = 0.779, sensitivity = 0.852, specificity = 0.706).

These results provide a first indirect evidence that our original algorithm can also perform relevant generalized predictions when applied to new MCI individuals. This motivates future efforts to bring the algorithm to sufficient levels of optimization and trustworthiness that will allow its application in both clinical and research settings.

We conducted a program of research to derive and test the reliability of a clinical prediction rule to identify high-risk older adults using paramedics’ observations.

We developed the Paramedics assessing Elders at Risk of Independence Loss (PERIL) checklist of 43 yes or no questions, including the Identifying Seniors at Risk (ISAR) tool items. We trained 1,185 paramedics from three Ontario services to use this checklist, and assessed inter-observer reliability in a convenience sample. The primary outcome, return to the ED, hospitalization, or death within one month was assessed using provincial databases. We derived a prediction rule using multivariable logistic regression.

We enrolled 1,065 subjects, of which 764 (71.7%) had complete data. Inter-observer reliability was good or excellent for 40/43 questions. We derived a four-item rule: 1) “Problems in the home contributing to adverse outcomes?” (OR 1.43); 2) “Called 911 in the last 30 days?” (OR 1.72); 3) male (OR 1.38) and 4) lacks social support (OR 1.4). The PERIL rule performed better than a proxy measure of clinical judgment (AUC 0.62 vs. 0.56, p=0.02) and adherence was better for PERIL than for ISAR.

The four-item PERIL rule has good inter-observer reliability and adherence, and had advantages compared to a proxy measure of clinical judgment. The ISAR is an acceptable alternative, but adherence may be lower. If future research validates the PERIL rule, it could be used by emergency physicians and paramedic services to target preventative interventions for seniors identified as high-risk.

To determine Canadian emergency physicians’ estimates regarding the safety and efficiency of chest discomfort management in their emergency department (ED), and their attitudes toward and perception of the need for a chest discomfort clinical prediction rule that identifies very low risk patients who are safe to discharge after a brief ED assessment.

300 members of the Canadian Association of Emergency Physicians (CAEP) were randomly selected to receive a confidential mail survey, which invited them to provide information on current disposition of patients with chest discomfort and their opinions regarding the value of a clinical prediction rule to identify patients with chest discomfort who are safe to discharge after a brief (~2 hour) assessment.

Of the 300 physicians selected, 288 were eligible for the survey and 235 (82%) responded. Only 5% follow discharged patients to measure safe practice. Overall, 165 (70%) felt the proposed prediction rule would be very useful and 43 (18%) felt it would be useful. Almost all (94%) believed a prediction rule would be useful if it identified patients safe for discharge without increasing the current rate of missed acute myocardial infarction (estimated at 2%). Most respondents (59%) believed that a clinical prediction rule should suggest a course of action, while 30% felt it should convey a probability of disease.

Canadian emergency physicians support the concept of a clinical prediction rule for the early discharge of patients with chest discomfort. Most believe that such a rule would be useful if it identified patients who are safe for discharge after a brief assessment, while maintaining current levels of safety. Future research should be aimed at deriving a clinical prediction rule to identify low risk patients who can be safely discharged after a limited emergency department evaluation.