The Journal of ExtraCorporeal Technology No 57-1 | Page 21

K . Gore et al .: J Extra Corpor Technol 2025 , 57 , 14 – 17 15

Materials and methods

Following IRB approval , a retrospective analysis of patient characteristics , comorbidities , and the incidence of RRT were studied from January 2017 to October 2023 in 129 patients receiving 159 MCS devices ( Table 1 ) at Ochsner Health- Jefferson Highway Campus in New Orleans , Louisiana . There were no patient exclusion criteria . Patient characteristics and recorded comorbidities ( Table 2 ) underwent machine learning to determine associations in the need for RRT [ 8 ].

Statistics

Baseline characteristics and comorbidities ( Table 2 ) underwent machine learning to explore these relationships for the need of RRT . The machine learning used in this prediction study included Decision-Tree ( Recursive Partitioning ), Bootstrap Forest , Boosted Tree , K Nearest Neighbors , Neural Support Vectors Machines , Discriminant , Fit Least Squares , Fit Stepwise , Logistic Regressions , Generalized Regression , Native Bayes , and Partial Least Squares [ 8 ]. P values for frequentist tests were set for statistical significance at < 0.05 . The statistical program , JMP Ò Pro 17.2 ( SAS Institute , Cary , NC ) was utilized for this study [ 8 ].

Results

In this study of patients requiring MCS , the incidence of RRT was 36 % with a 95 % confidence interval ( CI ) 29 – 44 %. The incidence of hospital mortality in patients requiring RRT was 79 % CI 66.7 – 87.5 % ( v 2 = 29 , P < . 0001 ) but was 35.3 % CI 26.7 – 44.9 % in MCS patients not requiring RRT . The types of MCS devices used in this study are shown in Table 1 . Baseline patient characteristics and recorded comorbidities in patients requiring MCS are shown in Table 2 . The baseline characteristics and comorbidities underwent machine learning associated with the outcome of interest , the need for RRT . Age , and two novel comorbidities , patients with a history immunomodulation , and patients with pacemaker / internal cardiac defibrillator ( ICD ) were statistically associated with the need for RRT ( v 2 = 44 , P = 0.0003 ; Table 2 ). The c-index statistic for this model was 0.81 . Based on the results of this model , two contingency tables were constructed to further explore the two novel comorbidities with the need for RRT ( Tables 3 and 4 ). Patients who had a history of immunomodulation were noted to have an incidence of 48 % in the need for RRT during MCS ( Table 3 ). Patients with a history of pacemaker / ICDs also had a high incidence ( 47 %) in requiring RRT during MCS ( Table 4 ). Patients with both comorbidities had a 66 % incidence in the need for RRT .

We further explored the role of anticoagulation used in the two novel groups when combined ( Interest groups ) and the results of that analysis are shown in Table 5 . In MCS patients receiving unfractionated heparin ( UFH ), a 43 % incidence in the need for RRT was observed in this cohort of patients . Four patients who did not receive anticoagulation therapy all required RRT , in contrast to three patients not requiring RRT when low molecular weight heparin ( LMWH ) was used

Table 1 . List of mechanical circulatory support devices .

( v 2 = 10.1 , P = 0.0064 ). While these findings were observed in a small subset of patients , the results warrant further investigation into anticoagulation practices used in this patient population .

Discussion

The use of MCS therapies is becoming an important component of supportive care in intensive care units [ 5 ]. Although initial support for patients frequently includes vasoactive support medications and / or mechanical ventilation , patients that continue to deteriorate or become refractory to medical therapy may require MCS [ 1 , 2 ].

Although the causes of cardiogenic shock are numerous , a low cardiac output state exists , that when unsuccessfully treated , results in end organ hypoperfusion [ 5 ]. In our study , we observed two novel preexisting risk factors for the need of RRT during MCS ; patients with preexisting inflammatory disorders requiring therapy , and patients with pre-existing pacemaker / ICDs . As we observed a higher incidence in the need for RRT in patients with these two disorders , this association suggests that an increased systemic inflammatory state exists that escalated the need for RRT [ 7 ], as in these two groups , the administration of UFH was not protective in reducing the need for RRT .

UFH is frequently used for anticoagulation during MCS which was based upon prior experience in procedures requiring cardiopulmonary bypass [ 9 ]. However with long-term UFH administration , heparin resistance and immune-mediated platelet activation leading to heparin-induced thrombocytopenia can develop [ 10 ]. In hypercoagulable states , such as observed in patients with renal failure , following major surgery , or histories of congestive heart failure , or autoimmune diseases , Kaur , Arsene , and colleagues recommend UFH should be used with caution [ 11 ]. Implantable cardiac devices have also been shown to generate an inflammatory response [ 12 ]. Taken together , the inflammatory components in UFH may contribute to the inflammatory state and increase the need for RRT . Studies with newer generation anticoagulants need to be conducted following development of bedside monitoring techniques to allow timely adjustment of anticoagulant therapy based upon realtime coagulation parameters [ 9 , 10 , 13 ].

Limitations

Limitations of retrospective studies suffer from completeness of medical record data . However , the strength of this study was the near 100 % data collection due to the recent development of electronic medical records . Another limitation of this