Detection of Hemodialysis Venous Needle Dislodgment Using Venous Access Pressure Measurements: A Simulation Study

Key Points circle Hemodialysis machine pressure alarms may not detect venous needle dislodgment when patients have changes in venous pressure. circle A cross-sectional analysis of hemodialysis treatment data identified the occurrences of venous pressure changes that would make it hard to trigger a machine alarm. circle A proof-of-concept use of a data analytic-derived algorithm for the detection of venous needle dislodgments was demonstrated. Abstract Background: In rare instances, hemodialysis venous needles may become dislodged, and when left undetected, this can lead to severe injury or death. Although dialysis machines have alarms to detect venous needle dislodgment (VND), their range of detection is limited. An understanding of the clinical conditions that may lead to missed needle dislodgments is needed for the development of more robust detection systems. Methods: We created a sham dialysis circuit with a Fresenius 2008K dialysis machine for in vitro simulation testing of machine alarm behavior under variable conditions. The circuit used a blood substitute and mimicked a patient's venous access site. We varied blood flow rate, venous pressure (VP), and upward drift in VP and analyzed the time to alarm for the machine and an improved alarm algorithm. We also performed a crosssectional retrospective study to identify the clinical occurrence of VP upward drift between September 1, 2016, and November 1, 2016, in patients on hemodialysis with an arteriovenous fistula. Results: Of 43,390 VP readings for 147 patients on hemodialysis, 16,594 (38%) showed an upward drift in VP (range 20-79 mmHg), with a mean6SD increase of 11618 mm Hg within 20614 minutes. A total of 19 VND simulations under different VP and blood flow parameters resulted in 19 (100%) algorithm alarm activations. Only eight simulations (42%) activated a machine alarm, and machine alarm activation time was longer than the algorithm activation time for all eight machine alarms (range 1-13 seconds). Conclusions: Patients can experience changes in VP during hemodialysis which may not trigger a machine alarm in the case of a VND. Our simulations showed that current dialysis machine alarm systems may not compensate for upward drift in VP, and improved algorithms for detecting needle dislodgment during hemodialysis are needed.