H.3 ICU Self Titrating Drips
The intensive care unit (ICU) houses the most critically ill and medically complex patients in the entire hospital. In general, admission into an ICU is warranted once the demand of a patient’s medical care exceeds what nursing staff can provide on a non-ICU floor. In other words, what differentiates ICU from non-ICU floors are nursing needs — which in an ideal world would allocate one nurse to one patient (vs one nurse to 3-4 patients in a non-ICU floor). Unfortunately, technology has only slowly made its way into the ICU to assist nurses and most patient care is still done manually.
One of the largest consumers of nursing time in an ICU are the real-time titration of medication drips — most commonly medications used to support blood pressure (called “vasopressors”). In this scenario, an arterial catheter is placed in a patients artery for real-time measurement of blood pressure. Based on these measurements, a nurse would adjust the dose of various vasopressor medications to attain a specific target blood pressure goal. This can often be a time-consuming process as fluctuations in blood pressure are frequent, and caused by a multitude of patient and non-patient factors. This takes valuable time away from nurses to be able to complete other tasks necessary for patient care, or to help assist with other patients. During the height of the COVID19 pandemic, nurses were often overwhelmed due to the number of drips that required manual titration per patient. If the titration of drips could be automated (or even augmented with automation), it would allow nurses to more efficiently take care of patients, improve clinical outcomes, and also be able to be a safety net against future pandemics when patient demands exceed nursing capacity.
Currently, there is no interconnectivity between blood pressure measurement devices and the pumps that administer medication drips (ie. they are not networked together and cannot exchange information). If this could be established, in scenarios where a nurse wanted more control over medication changes, the pump could simply give recommendations to the nurse as to what it believes the adjustment should be to achieve the target blood pressure goal. In disaster/emergency scenarios where an ICU is over-capacity, an algorithm could automatically make the adjustment on its own.
From an engineering perspective, in its simplest for, a simple adaptive filter would be a good initial first step, but there would likely be opportunities for the applications of AI/neural networks to be used in these recommendation algorithms.