Data Matrix

Machine learning predicts factors associated with device infection during secondary procedures

By Dr Aswin Babu

Tarakji et al. (2022) recently published post hoc analysis of 2,803 control patients from the WRAP-IT trial who underwent secondary device procedures that received standard care with pre-operative antibiotics. Due to the low rate of infections within this cohort (1.2%), multivariate machine learning was utilised with least absolute shrinkage and selection operator (LASSO) regression analysis to detect patient and procedural factors associated with an increased risk of device infections.

81 variables were screened. 17 factors were associated with an increased risk of infection with 6 non-modifiable i.e. patient/device related and 11 modifiable i.e. procedure related factors.

Major non-modifiable risk factors included the number of previous procedures, history of atrial arrhythmia, lower BMI, and having a CRT-D implanted in comparison to pacemakers or ICD’s. Major modifiable risk factors include length of procedure time; use of a preprocedural glycopeptide (Vancomycin) instead of cephalosporin (Cefazolin); implantation of device in a non-left pectoral subcutaneous location and anticoagulant use at the time of procedure. However, it is important to note that being on Apixaban was associated with a decreased risk of infection. Additionally, chlorhexidine skin preparation in comparison to a povidone-iodine based products was associated with lower infection rates. Finally, an antibiotic pocket wash was linked to reduced infection rates when rivalled to either a non-antibiotic wash or no wash.

Although, limitations exist with the inherent biases associated with post hoc analysis, there are some important messages within this study. For example, it is imperative to minimise procedure time as well stopping anticoagulation for an appropriate time period before the day of implantation. Furthermore, although infection rates have now drastically reduced in secondary implantations as evidenced by both the WRAP-IT and PADIT trial (1.2-1.3%), it is still critical to counsel patients on the correlation between the number of procedures and the cumulative risk of infection.

Read the original article


  1. Tarakji, K.G., Krahn, A.D., Poole, J.E., Mittal, S., Kennergren, C., Biffi, M., Korantzopoulos, P., Dallaglio, P.D., Lexcen, D.R., Lande, J.D. and Hilleren, G., 2022. Risk factors for CIED infection after secondary procedures: insights from the WRAP-IT trial. JACC: Clinical Electrophysiology, 8(1), pp.101-111.