Effective and rapid triaging from primary care into secondary care plays a pivotal role in providing patients with timely treatment and managing increasing demands for healthcare resources. Existing triaging methods from primary care to secondary care are labor-intensive processes that involve manually reviewing referral data from multiple sources and can cause long referral to treatment time. There has been no research using machine learning methods that automatically analyzes heterogeneous data including referral letters to recognize regularities to support the primary to secondary care triage. In this paper, we propose a heterogeneous data-driven hybrid machine learning model including Natural Language Processing (NLP) to improve hospital triage efficiency at the point of triage. The proposed model achieved a precision of 0.83, recall of 0.82, F1-Score of 0.83, accuracy of 0.82, AUC of 0.90 in identifying patients with non-inflammatory conditions (NIC) and inflammatory arthritis (IA) at the point of triage with explainable risk stratifications. Our model is piloted in a real-world trial in a large secondary care hospital in the UK to compare referral accuracy and time saved between our model and clinicians, and evaluate its acceptability by users. Our model achieved precision and recall of 0.83 and 0.81, compared with the precision and recall of 0.80 and 0.78 by clinicians. The research also shows that our model enabled decision support can save clinicians 8 h per week in assessing the referral assessment. This paper is the first study to streamline hospital triage from primary care to secondary care using machine learning.
Acute exacerbations of COPD (AECOPD) are associated with high morbidity and mortality and frequent readmissions.
What is the effectiveness of a COPD transition bundle, with and without a care coordinator, on rehospitalizations and ED revisits?
Two patient cohorts were selected: (1) the group exposed to the transition bundle and (2) the group not exposed to the transition bundle (usual care group). Patients exposed subsequently were randomized to a care coordinator. An AECOPD transition bundle was implemented in the hospital; patients randomized to the care coordinator were contacted ≤ 72 h after discharge. Six hundred four patients (320 to the care coordinator and 284 to routine care) who met eligibility criteria from five hospitals across three cities in Alberta, Canada, were exposed to the transition bundle, whereas 3,106 patients discharged from the same hospitals received the usual care. Primary outcomes were 7-day, 30-day, and 90-day readmissions, median length of stay (LOS), and 30-day ED revisits.
The transition bundle cohort were 83% (relative risk [RR], 0.17; 95% CI, 0.07-0.35) less likely to be readmitted within 7 days and 26% (RR, 0.74; 95% CI, 0.60-0.91) less likely to be readmitted within 30 days of discharge. Ninety-day readmissions were unchanged (RR, 1.05; 95% CI, 0.93-1.18). The transition bundle was associated with a 7.3% (RR, 1.07; 95% CI, 1.0-1.15) relative increase in LOS and a 76% (RR, 1.76; 95% CI, 1.53-2.02) greater risk of a 30-day ED revisit. The care coordinator did not influence readmission or ED revisits.
The COPD transition bundle reduced 7- and 30-day hospital readmissions while increasing LOS and ED revisits. The care coordinator did not improve outcomes.
The electronic Canadian Triage and Acuity Scale (eCTAS) is a real-time electronic triage decision-support tool designed to improve patient safety and quality of care by standardizing the application of the Canadian Triage and Acuity Scale (CTAS). The objective of this study is to determine interrater agreement of triage scores pre- and post-implementation of eCTAS.
This was a prospective, observational study conducted in 7 emergency departments (EDs), selected to represent a mix of triage documentation practices, hospital types, and patient volumes. A provincial CTAS auditor observed triage nurses in the ED pre- and post-implementation of eCTAS and assigned an independent CTAS score in real time. Research assistants independently recorded triage time. Interrater agreement was estimated with κ statistics with 95% confidence intervals (CIs).
A total of 1,491 individual triage assessments (752 pre-eCTAS, 739 post-implementation) were audited during 42 7-hour triage shifts (21 pre-eCTAS, 21 post-implementation). Exact modal agreement was achieved for 567 patients (75.4%) pre-eCTAS compared with 685 patients (92.7%) triaged with eCTAS. With the auditor’s CTAS score as the reference, eCTAS significantly reduced the number of patients over-triaged (12.0% versus 5.1%; Δ 6.9; 95% CI 4.0 to 9.7) and under-triaged (12.6% versus 2.2%; Δ 10.4; 95% CI 7.9 to 13.2). Interrater agreement was higher with eCTAS (unweighted κ 0.89 versus 0.63; quadratic-weighted κ 0.93 versus 0.79). Median triage time was 312 seconds (n=3,808 patients) pre-eCTAS and 347 seconds (n=3,489 patients) with eCTAS (Δ 35 seconds; 95% CI 29 to 40 seconds).
A standardized, electronic approach to performing triage assessments improves both interrater agreement and data accuracy without substantially increasing triage time.
The purpose of this study was to examine: a) how long and how frequently older hospitalized patients spend upright; b) whether duration and frequency of upright time change by time of the day, the day of the week, and during hospitalization; and c) whether these relationships differ based on the mobility level of patients at admission.
This prospective cohort study included 111 patients (82.2 ± 8 years old, 52% female) from the Emergency Department and a Geriatric Assessment Unit who were at least 60 years old and had an anticipated length of stay of at least three days. The main outcomes were accelerometer-measured total upright time and number of bouts of upright time during awake hours.
Patients were upright 15.9 times/day (interquartile range (IQR): 8.4–27.4) for a total of 54.2 min/day (IQR: 17.8–88.9) during awake hours. Time of day and day of week had little impact on the outcomes. Patients who walked independently at admission had 151.5 min (95% CI: 87.7–215.3) of upright time on hospital day 1 and experienced a decline of 4.5 min/day (−7.2 to −1.8). Those who needed personal mobility assistance or were bedridden had 29.5 min (−38.5–97.4) and 25 min (−48.3–100.3) of upright time on day 1, and demonstrated an increase of 3.6 (1.3–5.9) and 2.4 (0.05–4.5) min/day, respectively.
Hospitalized older adults spend only 6% of their awake hours upright while in hospital. Patients who can walk independently are more active but experience a decline in their upright time during hospitalization.