Leveraging Unlabeled Clinical Data to Boost Performance of Risk Stratification Models for Suspected Acute Coronary Syndrome.

The performance of deep learning models in the health domain is desperately limited by the scarcity of labeled data, especially for specific clinical-domain tasks. Conversely, there are vastly available clinical unlabeled data waiting to be exploited to improve deep learning models where their training labeled data are limited. This paper investigates the use of task-specific unlabeled data to boost the performance of classification models for the risk stratification of suspected acute coronary syndrome. By leveraging large numbers of unlabeled clinical notes in task-adaptive language model pretraining, valuable prior task-specific knowledge can be attained. Based on such pretrained models, task-specific fine-tuning with limited labeled data produces better performances. Extensive experiments demonstrate that the pretrained task-specific language models using task-specific unlabeled data can significantly improve the performance of the downstream models for specific classification tasks.

Pregabalin vs placebo to prevent chronic pain after whiplash injury in at-risk individuals: results of a feasibility study for a large randomised controlled trial.

ABSTRACT: There are few effective treatments for acute whiplash-associated disorders (WADs). Early features of central sensitisation predict poor recovery. The effect of pregabalin on central sensitisation might prevent chronic pain after acute whiplash injury. This double blind, placebo-controlled randomised controlled trial examined feasibility and potential effectiveness of pregabalin compared with placebo for people with acute WAD. Twenty-four participants with acute WAD (<48 hours) and at risk of poor recovery (pain ≥5/10) were recruited from hospital emergency departments in Queensland, Australia, and randomly assigned by concealed allocation to either pregabalin (n = 10) or placebo (n = 14). Pregabalin was commenced at 75 mg bd, titrated to 300 mg bd for 4 weeks, and then weaned over 1 week. Participants were assessed at 5 weeks and 3, 6, and 12 months. Feasibility issues included recruitment difficulties and greater attrition in the placebo group. For the primary clinical outcome of neck pain intensity, attrition at 5 weeks was pregabalin: 10% and placebo: 36% and at 12 months was pregabalin: 10% and placebo: 43%. Pregabalin may be more effective than placebo for the primary clinical outcome of neck pain intensity at 3 months (mean difference: -4.0 [95% confidence interval -6.2 to -1.7]) on an 11-point Numerical Rating Scale. Effects were maintained at 6 months but not 12 months. There were no serious adverse events. Minor adverse events were more common in the pregabalin group. A definitive large randomised controlled trial of pregabalin for acute whiplash injury is warranted. Feasibility issues would need to be addressed with modifications to the protocol.

Safety of rapid sequence intubation in an emergency training network.

OBJECTIVE: Rapid sequence intubation (RSI) is a core critical care skill. Emergency medicine trainees are exposed to relatively low numbers of RSIs. We aimed to improve patient outcomes by implementing an RSI checklist, electronic learning and audit, in line with current best evidence. METHODS: Prospective observational study of RSIs performed in the EDs of two Queensland hospitals between January 2014 and December 2016. Data collected included: first-pass success (FPS), predicted difficulty, indication for intubation, drugs used, positioning, number of attempts, checklist use and complications. Descriptive statistics and multivariable modelling were used to describe differences in FPS, and complications. RESULTS: Six hundred and fifty-five patients underwent RSI with FPS of 86.6%. Complications were reported in 15.9%, mainly hypotension (10.9%) and desaturation (4.0%). FPS improved with bougie use (88.9% vs 73.0% without bougie, P < 0.001) and video-laryngoscopy (88.2% vs 72.9% using standard laryngoscopy, P < 0.001). New desaturation was reduced with apnoeic oxygenation (2.0% vs 22.2%, P < 0.001), bougie use (2.8% vs 8.9%, P < 0.001), checklist use (2.3% vs 22.7%, P < 0.001) and achieving FPS (2.1% vs 16.3%, P < 0.001). Complications were reduced with checklist use (13.3% vs 43.2%, P < 0.001) and apnoeic oxygenation use (3.9% vs 31.1%, P < 0.001). Logistic regression found checklist use was associated with reduced desaturation (OR 0.1, 95% CI 0.04-0.27) and the composite variable of any complication (OR 0.39, 95% CI 0.17-0.89). CONCLUSIONS: Implementation of an evidence-based care bundle and audit of practice has created a safe environment for trainees to learn the core critical care skill of RSI. In our setting, checklist use was associated with fewer complications.

Does a single dose of intravenous dexamethasone reduce Symptoms in Emergency department patients with low Back pain and RAdiculopathy (SEBRA)? A double-blind randomised controlled trial.

OBJECTIVE: To assess the effect of a single dose of intravenous dexamethasone in addition to routine treatment on visual analogue scale (VAS) pain scores at 24 h in emergency department (ED) patients with low back pain with radiculopathy (LBPR). METHODS: Double-blind randomised controlled trial of 58 adult ED patients with LBPR, conducted in one tertiary and one urban ED. The intervention was 8 mg of intravenous dexamethasone (or placebo) in addition to current routine care. The primary outcome was the change in VAS pain scores between presentation and 24 h. Secondary outcomes included VAS pain scores at 6 weeks, ED length of stay (EDLOS), straight leg raise (SLR) angles and Oswestry functional scores. RESULTS: Patients treated with dexamethasone had a 1.86 point (95% CI 0.31 to 3.42, p=0.019) greater reduction in VAS pain scores at 24 h than placebo (dexamethasone: -2.63 (95% CI -3.63 to -1.63) versus placebo: -0.77 (95% CI -2.04 to 0.51)). At 6 weeks, both groups had similar significant and sustained decrease in VAS scores compared with baseline. Patients receiving dexamethasone had a significantly shorter EDLOS (median: 3.5 h vs 18.8 h, p=0.049) and improved SLR angle at discharge (14.7°, p=0.040). There was no difference in functional scores. CONCLUSIONS: In patients with LBPR, a single dose of intravenous dexamethasone in addition to routine management improved VAS pain scores at 24 h, but this effect was not statistically significant at 6 weeks. Dexamethasone may reduce EDLOS and can be considered as a safe adjunct to standard treatment. TRIAL REGISTRATION NUMBER: ACTRN12611001020976.

Intravenous 0.9% sodium chloride therapy does not reduce length of stay of alcohol-intoxicated patients in the emergency department: a randomised controlled trial.

BACKGROUND: I.v. 0.9% sodium chloride (normal saline) is frequently used to treat ED patients with acute alcohol intoxication despite the lack of evidence for its efficacy. OBJECTIVE: The study aims to compare treatment with i.v. normal saline and observation with observation alone in ED patients with acute alcohol intoxication. METHODS: A single-blind, randomised, controlled trial was conducted to compare a single bolus of 20 mL/kg i.v. normal saline plus observation with observation alone. One hundred and forty-four ED patients with uncomplicated acute alcohol intoxication were included. The study was conducted in one tertiary and one urban ED in Queensland, Australia. Primary outcome was ED length of stay (EDLOS). Secondary outcomes were treatment time, breath alcohol levels, intoxication symptom score, level of intoxication and associated healthcare costs. RESULTS: Both groups were comparable at baseline: blood alcohol content (BAC) was similar between treatment and control groups (0.20 % BAC vs 0.19 % BAC, P = 0.44) as were initial intoxication symptom scores (22.0 vs 22.3, P = 0.90). Both groups had a similar EDLOS (287 min vs 274 min, P = 0.89; difference 13 min [95% CI -37-63]) and treatment time (244 min vs 232 min, P = 0.94; difference 12 min [95% CI -31-55]). Change of breath alcohol levels, intoxication score and level of intoxication were not significantly different between the two groups. Patients in the treatment group had an additional healthcare cost of A$31.92 compared with control. CONCLUSIONS: I.v. normal saline therapy added to observation alone does not decrease ED length of stay compared with observation alone. Intoxication symptom scores and general state of intoxication were similar in both groups. The present study suggests that either approach is reasonable, but observation alone might be preferred as it is less resource intensive.