Practice patterns of midazolam administration in patients undergoing cardiac-related non-operating room procedures: analysis of a US anaesthesia database

Background Midazolam can be administered as an anxiolytic, a cardio-stable sedative, or an anaesthetic agent during non-operating room anaesthesia (NORA) procedures. Guidelines suggest avoidance in older or frail patients, and current practices are unknown. We hypothesised that there would be significant variation in midazolam administration for cardiac-related NORA procedures by attending and institutional characteristics, in addition to patient characteristics. Methods This is a retrospective observational analysis of a multi-institution US-centric electronic health record dataset including all adult patients undergoing cardiac-related NORA procedures in the Multicenter Perioperative Outcomes Group database from 2015 to 2022. The primary aim of the study was to identify the proportion of total variance in midazolam administration attributable to patient-, attending-, and institutional-level factors. The secondary aim was to identify practice pattern variations in the dosage and timing of midazolam administration. The primary outcome was any midazolam administration ≤1 h before anaesthesia start to 1 h after anaesthesia end. Secondary outcomes included total dose and timing of midazolam administration. Variance in midazolam administration attributable to patient, attending, and institutional factors was calculated with intraclass correlation coefficients following multilevel mixed-effects generalised linear models, in which a three-level nesting structure was assumed: patients nested within attendings and attendings nested within institutions. For the primary outcome, a binomial distribution and logit link function was specified. For the secondary outcome of total midazolam, the same three-level nesting structure was used, and covariates were modelled using a multilevel mixed-effects model specifying a Gaussian distribution and identity link function. Results A total of 246 918 cardiac-related NORA procedures with care provided by 5212 unique anaesthesia attendings across 63 institutions with a median case volume of 9 (inter-quartile range [IQR] 2–37) were included in the analysis. Midazolam was administered in 33.0% of patients ( n =81 602), including 25.5% of patients aged ≥65 yr ( n =40 469). Midazolam administration declined from 50.7% of patients in 2015 to 29.0% in 2022 ( P <0.001). The median dose was 2 mg (IQR 2–2), with 89.9% administered before procedure start. For transoesophageal echocardiography procedures, it was estimated that 37.0% of midazolam administration (30.4–44.3) was attributable to attending variation and 26.4% (19.2–35.2) to the institution, assuming that 36.6% of the variation was attributable to known patient factors. For transcatheter/endovascular procedures, attending, institution, and patient factors accounted for 45.7% (39.0–52.6), 27.7% (19.7–37.4), and 26.6%, respectively. Conclusions Although patient-level factors accounted for a meaningful proportion of variability, substantial variation in midazolam administration was also attributable to attending- and institution-level factors. Further studies on clinical outcomes and patient experience associated with this type of midazolam administration have the potential to inform future practice.