The use of artificial intelligence and machine learning monitoring to safely administer a fluid-restrictive goal-directed treatment protocol to minimize the risk of transfusion during major spine surgery of a Jehovah’s Witness: a case report

Rarely is major surgery performed where the transfusion of blood products is not possible. But even under these circumstances, patients should not be deterred from receiving surgical intervention and treatment. Here, we present a case in which a goal-directed therapy based on machine learning hemodynamic monitoring provided a safe framework for conducting invasive spinal surgery in an older patient, who, as a committed Jehovah’s Witness (JW), refused all blood products. For this reason, patient blood management (PBM) played a pivotal role in perioperative management. Of note, pre-existing conditions and advanced age posed an elevated risk to the patient for intraoperative hypotension (IOH). Since IOH is a commonly encountered problem during anaesthesia and is associated with increased mortality and morbidity [1‐5], it is vital to stabilize cardiopulmonary function, even during phases of aggravated blood loss.

The patient gave her written consent to the anonymous description and publication of her case report.

We present the case of an 84-year-old Caucasian woman suffering from a progressive impression fracture of the 11th thoracic vertebral body and severe pain of the left flank of several weeks’ duration. Prior to the patient’s current hospitalization, as reported here, a decompression and stabilization of the lumbar and thoracic spine had been performed 2 months before in a different neurosurgical clinic. In view of the clinical results and continuing severe pain, an extension of the preexisting spinal fusion from the 4th lumbar vertebral body to the second sacral vertebral body, including a cage implantation, was recommended.

The patient presented multiple preexisting conditions, including high blood pressure, diabetes mellitus, diabetic polyneuropathy, moderate valvular aortic stenosis, and restless legs syndrome. Further, the patient introduced herself as a member of the community of JWs. For this reason, she was transferred to the local department of anesthesiology to discuss blood-saving techniques as well as the risks accompanying the procedure, as she was not willing to receive any form of blood transfusion.

After the patient’s decision to undergo spinal fusion surgery, an interdisciplinary plan based on the three pillars of PBM was made to provide maximum security for the patient’s outcome.

To our knowledge, this is the first time that a machine learning algorithm was used to safely utilize a fluid-restrictive GDT treatment protocol in a JW patient to minimize the risk of transfusion and to still guarantee hemodynamic optimization.

There are various ways to practice pre-, intra-, and postoperative PBM during high-risk surgical procedures. However, the inability to use blood products presents a uniquely challenging situation. In this case, the patient maintained a very strict view on the usage of blood products, leaving the anesthesiologist with a limited scope of therapeutic actions at their disposal. Due to the extensive nature of the surgery, it was necessary to focus not only on Hb- and iron compensation beforehand, but also on techniques that could optimize cardiopulmonary function respective to the corresponding oxygen delivery. Furthermore, it was vital to maintain a balanced hemodynamic status due to the patient’s age and various preexisting conditions. For this reason, the attending anesthesiologist chose to use an HPI-based hemodynamic GDT in order to treat hemodynamic instability and hypotensive events before their occurrence. This treatment approach led to a high degree of hemodynamic stability without putting the patient at risk of a too-liberal fluid therapy with consequent hemodilution. More specifically, in many GDT algorithms, fluid optimization is used by maximizing stroke volume, but the downside of this approach might be a fluid strategy that is too liberal: as with the last fluid bolus that does not significantly increase stroke volume, indicating that the patient is at top of its individual starling curve, additional fluid is infused, which actually decreases oxygen delivery by not further increasing cardiac output but leading to hemodilution. Such an approach might be especially detrimental in patients where blood transfusion is not an option. Therefore, our approach was to use the HPI monitoring protocol that would only indicate the need for hemodynamic optimization and fluid expansion in the case of imminent IOH, and would therefore offer a more fluid-restrictive strategy, minimizing the risk of hemodilution in this vulnerable patient.

HPI monitoring has been evaluated in several studies as regards the primary goal of reducing IOH, since it continues to present a common problem during anesthesia and is associated with increased mortality and morbidity [1‐5]. For example, mortality 1 year after anesthesia exposure is raised by 3.6% for every minute the systolic arterial blood pressure (SAP) remained below 80 mmHg. Furthermore, the mortality risk is 1.4 times higher when the mean arterial blood pressure (MAP) decreased under 55 mmHg for more than 10 min [2, 8]. Davies et al. outline a significant reduction of IOH by using an HPI-based GDT protocol [9]. Recently, our own study on patients undergoing hip replacement surgery showed that the use of HPI can significantly reduce the incidence and duration of IOH [6]. Further, Wijnberge et al. described the occurrence of less IOH in HPI-guided therapy compared with standard care [10]. Also, in the presented case, the duration of IOH (defined as MAP < 65 mmHg) could be reduced to just 8 min by strict adherence to the HPI-based GDT (Fig. 2). The data evaluation showed a total of 55 min with an HPI > 70% (of 425 min), resulting in 21 interventions that included colloid and noradrenaline infusions. The patient’s mean MAP over the total course of surgery was 88 mmHg. Further, cardiac function was preserved over the course of surgery despite the extended blood loss (2.4 L, Hb drop of 11.8 to 6.7 g/dl), and the patient never suffered any form of acidosis (Table 1).

PBM plays a pivotal role in high-risk bleeding surgeries, especially for someone who is a JW. The optimization of red blood cell production can easily be managed using EPO or iron substitutes before surgery. Also, surgical techniques have massively improved the provision of minimally invasive methods to help reduce blood loss, as has the use of tranexamic acid. However, in the opinion of the authors, hemodynamic optimization must be highlighted more in the context of PBM, since it not only reduces IOH and accompanying risks, but it can also keep patients hemodynamically stable, maximizing the tolerance of low Hb levels by directing individual volume demand. Most studies on the perioperative management of a JW patient concentrate on oral iron substitution and intraoperative normovolemic hemodilution, respectively, as well as the use of autotransfusion techniques. For example, without focusing on hemodynamic optimization, McCartney et al. were able to show that the abovementioned therapeutic actions were already associated with an improved clinical outcome after cardiac surgery of JWs [11‐13]. Also, several case reports have supported preoperative EPO administration and blood-saving surgical techniques [14]. Today, even though the clinical implementation of these strategies is commonly applicable, it should also be focused on individual hemodynamic stabilization.

In sum, our approach demonstrates how the use of an HPI-guided GDT in a transfusion-limited surgical context led to a high degree of hemodynamic stability while reducing the risk for dilutional anemia, and thus represents an extension of PBM both in general and for the JW community in particular.