One of the problems clinicians face when a COVID-19 patient is hospitalized is opining what other complications this patient will present.  All too often, kidney failure is among the most common result.  (This held for  46% of the patients treated at Mount Sinai, with about 1/3 of them requiring dialysis).  The issue is getting the affected patients on dialysis before it’s too late to save their lives.

Employing the data garnered from 3235 COVID-19 patients that were hospitalized, folks from Mount Sinai Hospital in Manhattan (most were also associated with the Icahn School of Medicine there) employed artificial intelligence to develop an algorithm to predict who will develop acute kidney failure. (These were the critical data points: creatinine and potassium blood levels, heart rate, oxygen saturation, and patient age.) More importantly, the data necessary for this prediction is developed within the first 48 hours after admission to the hospital; this means treatment regimes can be prepared and ready to go before the patient manifests AKI (acute kidney injury) symptoms.

Some parameters were not included in developing the algorithm- even if they could prove valuable- because they were not always available.  Inflammatory markers (iL-6, ferritin, fibrinogen) were not ascertained early enough in the patient’s hospitalization, nor were urinalysis data always collected.  Other data (angiotensin receptor blockers, angiotensin converting enzyme inhibitors, statins, and non-steroidal anti-inflammatory drug use) that is not always available were also not included in the model’s development.

The paper’s lead author was Dr. Lili Chan, along with 62 other Mount Sinai (NY) physicians. “Machine Learning for Predication of Severe Acute Kidney Injury in Hospitalized Patients with COVID-19”  was presented at the Annual Meeting of the American Society of Nephrology that transpired from 19 to 25 October 2020.   (The virtual meeting was termed “ASN Kidney Week 2020 Reimagined,)

Patients who already manifested ESRD (end stage renal disease) prior to admission were not included in the model (for obvious reasons), as well as folks who were released from the hospital in less than 2 days and those who were missing laboratory and/or vital signs from their patient records.

The need for this algorithm becomes obvious once one realizes that those who develop AKI were far more likely to be transferred to intensive care, need vasopressor administration, or to be placed on mechanical ventilators.  (Of the 46% who manifested AKI in intensive care, 1/3 needed dialysis compared to some 11% who did not get transferred to the ICU.)  Moreover, some 44% of the survivors never recovered kidney function by the time of their release from the hospital- and of those manifesting AKI, about 3 in ten survived and recovered kidney function. (The model also predicts which patients are going to need chronic dialysis treatments.)

Let’s hope this model works.  Because this third wave (I’m guessing there will be two more) will involved a slew of folks hospitalized for COVID-19.  Getting the right treatment to them early will be critical for their survival.

Share this: