Current challenges in adopting machine learning to critical care and emergency medicine.

Over the past decades, the field of machine learning (ML) has made great strides in medicine. Despite the number of ML-inspired publications in the clinical arena, the results and implications are not readily accepted at the bedside. Although ML is very powerful in deciphering hidden patterns in complex critical care and emergency medicine data, various factors including data, feature generation, model design, performance assessment, and limited implementation could affect the utility of the research. In this short review, a series of current challenges of adopting ML models to clinical research will be discussed.

An Unsuspected Case of Euglycemic Diabetic Ketoacidosis With Twists.

Euglycemic diabetic ketoacidosis (EDKA) is defined by acidosis, ketones in serum and urine, and a high anion gap (AG) with a normal glucose level. Sodium-glucose cotransporter-2 (SGLT2) inhibitor has become one of the most common causes of EDKA. We present one of the lowest presenting blood glucose levels for the EDKA in the setting of SGLT2 inhibitor use. A 34-year-old female with a two-year history of type 2 diabetes mellitus (T2DM) presented after accidental ingestion of a metal clip and lack of oral intake for 28 hours. She reported a recent intentional weight loss of 60 lbs. She takes metformin 1000 mg twice daily and empagliflozin 25 mg daily. An urgent endoscopy was planned in the intensive care unit given the proximal location of the foreign body. The basic metabolic profile after the procedure demonstrated a glucose level of 75 mg/dL, bicarbonate level of 11 mmol/L, and AG of 17 mmol/L. The venous pH was 7.27 with a partial pressure of carbon dioxide of 30 mmHg. The urinalysis showed a glucose level greater than 500 µmol/L with a ketone level of 80 µmol/L. The blood and urine toxicology screening results were unremarkable. The patient was treated for EDKA with the administration of intravenous (IV) dextrose 5% in water with subsequent initiation of IV insulin. The assessment of her insulin reserve revealed a low C-peptide of 0.36 ng/mL, a high glutamic acid decarboxylase level greater than 250 IU/mL, and high zinc transporter 8 (ZnT8) antibodies of 42 U/mL, consistent with an undiagnosed transition to latent autoimmune diabetes in adults (LADA). The blood glucose levels for previously reported cases remain between 90 and 250 mg/dL. In this case, the combination of a low carbohydrate diet and prolonged starvation may have led to the impressively low glucose. Additionally, the SGLT2 inhibitor use in patients with LADA serves as one of the known risk factors for EDKA. EDKA poses a diagnostic challenge, especially in the ICU setting where there exists a myriad of causes for high AG metabolic acidosis. Additionally, many of the ICU patients are in a ketotic state brought on by prolonged starvation. Therefore, prompt diagnosis and treatment for EDKA require careful history taking and complete investigation for other causes of high AG metabolic acidosis.

A rare case of left dominant pulmonary edema in acute mitral regurgitation.

Unilateral pulmonary edema (UPE) due to cardiac causes is an unusual clinical finding and typically emerges on the right side. UPE poses a diagnostic challenge due to difficulty distinguishing infiltrative pneumonia from cardiogenic edema on chest imaging. Consequently, corrective clinical management is significantly delayed in UPE compared to bilateral cardiogenic pulmonary edema. We present a very rare case of left-sided cardiogenic pulmonary edema due to acute severe MR wherein a prompt cardiac evaluation for UPE led to successful corrective surgery and favorable outcome.