Measuring Potentially Avoidable Acute Care Transfers From Long-Term Care Homes in Quebec: a Cross Sectional Study.

Background: Potentially avoidable emergency department transfers (PAEDTs) and hospitalizations (PAHs) from long-term care (LTC) homes are two key quality improvement metrics. We aimed to: 1) Measure proportions of PAEDTs and PAHs in a Quebec sample; and 2) Compare them with those reported for the rest of Canada. Methods: We conducted a repeated cross-sectional study of residents who were received at one tertiary hospital between April 2017 and March 2019 from seven LTC homes in Quebec, Canada. The MedUrge emergency department database was used to extract transfers and resident characteristics. Using published definitions, PAEDTs and PAHs were identified from principal emergency department and hospitalization diagnoses, respectively. PAEDT and PAH proportions were compared to those reported by the Canadian Institute for Health Information. Results: A total of 1,233 transfers by 692 residents were recorded, among which 36.3% were classified as being potentially avoidable: 22.8% 'PAEDT only', 11.6% 'both PAEDT & PAH', and 1.9% 'PAH only'. Shortness of breath was the most common reason for transfer. Pneumonia was the most common diagnosis from the 'both PAEDT & PAH' category. PAEDTs and PAHs accounted for 95% and 37% of potentially avoidable transfers, respectively. Among 533 hospitalizations, 31.3% were PAHs. These proportions were comparable to the rest of Canada, with some differences in proportions of transfers due to congestive heart failure, urinary tract infection, and implanted device management. Conclusions: PAEDTs far outweigh PAHs in terms of frequency, and their monitoring is important for quality assurance as they may inform LTC-level interventions aimed at their reduction.

Explainable Machine Learning Model to Predict COVID-19 Severity Among Older Adults in the Province of Quebec.

Context: Patients over the age of 65 years are more likely to experience higher severity and mortality rates than other populations from COVID-19. Clinicians need assistance in supporting their decisions regarding the management of these patients. Artificial Intelligence (AI) can help with this regard. However, the lack of explainability-defined as "the ability to understand and evaluate the internal mechanism of the algorithm/computational process in human terms"-of AI is one of the major challenges to its application in health care. We know little about application of explainable AI (XAI) in health care. Objective: In this study, we aimed to evaluate the feasibility of the development of explainable machine learning models to predict COVID-19 severity among older adults. Design: Quantitative machine learning methods. Setting: Long-term care facilities within the province of Quebec. Participants: Patients 65 years and older presented to the hospitals who had a positive polymerase chain reaction test for COVID-19. Intervention: We used XAI-specific methods (e.g., EBM), machine learning methods (i.e., random forest, deep forest, and XGBoost), as well as explainable approaches such as LIME, SHAP, PIMP, and anchor with the mentioned machine learning methods. Outcome measures: Classification accuracy and area under the receiver operating characteristic curve (AUC). Results: The age distribution of the patients (n=986, 54.6% male) was 84.5□19.5 years. The best-performing models (and their performance) were as follows. Deep forest using XAI agnostic methods LIME (97.36% AUC, 91.65 ACC), Anchor (97.36% AUC, 91.65 ACC), and PIMP (96.93% AUC, 91.65 ACC). We found alignment with the identified reasoning of our models' predictions and clinical studies' findings-about the correlation of different variables such as diabetes and dementia, and the severity of COVID-19 in this population. Conclusions: The use of explainable machine learning models, to predict the severity of COVID-19 among older adults is feasible. We obtained a high-performance level as well as explainability in the prediction of COVID-19 severity in this population. Further studies are required to integrate these models into a decision support system to facilitate the management of diseases such as COVID-19 for (primary) health care providers and evaluate their usability among them.

Taxonomy of Interventions to Reduce Acute Care Transfers From Long-term Care Homes: A Systematic Scoping Review.

OBJECTIVE: To develop a taxonomy of interventions aimed at reducing emergency department (ED) transfers and/or hospitalizations from long-term care (LTC) homes. DESIGN: A systematic scoping review. SETTING AND PARTICIPANTS: Permanent LTC home residents. METHODS: Experimental and comparative observational studies were searched in MEDLINE, CINAHL, Embase Classic + Embase, the Cochrane Library, PsycINFO, Social Work Abstracts, AMED, Global Health, Health and Psychosocial Instruments, Joanna Briggs Institute EBP Database, Ovid Healthstar, and Web of Science Core Collection from inception until March 2020. Forward/backward citation tracking and gray literature searches strengthened comprehensiveness. The Mixed Methods Appraisal Tool was used to assess study quality. Intervention categories and components were identified using an inductive-deductive thematic analysis. Categories were informed by 3 intervention dimensions: (1) "when/at what point(s)" on the continuum of care they occur, (2) "for whom" (ie, intervention target resident populations), and (3) "how" these interventions effect change. Components were informed by the logistical elements of the interventions having the potential to influence outcomes. All interventions were mapped to the developed taxonomy based on their categories, components, and outcomes. Distributions of components by category and study year were graphically presented. RESULTS: Ninety studies (25 randomized, 23 high quality) were included. Six intervention categories were identified: advance care planning; palliative and end-of-life care; onsite care for acute, subacute, or uncontrolled chronic conditions; transitional care; enhanced usual care (most prevalent, 31% of 90 interventions); and comprehensive care. Four components were identified: increasing human resource capacity (most prevalent, 93%), training or reorganization of existing staff, technology, and standardized tools. The use of technology increased over time. Potentially avoidable ED transfers and/or hospitalizations were measured infrequently as primary outcomes. CONCLUSIONS AND IMPLICATIONS: This proposed taxonomy can guide future intervention designs. It can also facilitate systematic reviews and precise effect size estimations for homogenous interventions when outcomes are comparable.

Do Weekly Alerts From a Mobile Application Influence Reading During Residency?

BACKGROUND: The benefits of "spaced education" have been documented for residents in highly focused specialties. We found no published studies of spaced education in family medicine. In this study, we report on the feasibility of delivering weekly alerts from a mobile application (app) developed for exam preparation, to increase the reading of clinical information in the family medicine residency. DESIGN: This is a 2-phase mixed methods study. Phase one is a quasi-experimental study of resident reading of information related to priority topics in family medicine. Reading was documented by page views in a noncommercial mobile app. PARTICIPANTS: All incoming first-year residents at two university training programs in Canada. The intervention group received one alert per week to priority topics on the app, beginning in their second month of residency. The control group was given access to the same app, but received no alerts. RESULTS: In this paper, we report the phase one preliminary findings. In the intervention group, 81 of 96 first year residents consented. At the control site, 79 of 85 residents consented. After 100 days, intervention group residents had viewed more pages of clinical information across all 99 priority topics (1,546 versus 900) and per topic (15.7 versus 9.1 pages, P < 0.0003). On average, each increase of one visit to the app following a weekly alert was associated with an increase of 3.2 visits to pages of clinical information in the app. CONCLUSION: A weekly alert delivered via mobile app shows promise with respect to reading in the family medicine residency.