A replicable method for blood glucose control in critically Ill patients

CONTEXT:: To ensure interpretability and replicability of clinical experiments, methods must be adequately explicit and should elicit the same decision from different clinicians who comply with the study protocol. OBJECTIVE:: The objective of this study was to determine whether clinician compliance with protocol recommendations exceeds 90%. DESIGN:: We developed an adequately explicit computerized protocol (eProtocol-insulin) for managing critically ill adult patient blood glucose. We monitored clinician compliance with eProtocol-insulin recommendations in four intensive care units in four hospitals and compared blood glucose distributions with those of a simple clinical guideline at one hospital and a paper-based protocol at another. All protocols and the guideline used intravenous insulin and 80 to 110 mg/dL (4.4-6.1 mmol/L) blood glucose targets. SETTING:: The setting for this study was four academic hospital intensive care units. PATIENTS:: This study included critically ill adults requiring intravenous insulin. INTERVENTION:: Intervention used in this study was a bedside computerized protocol for managing blood glucose. MAIN OUTCOME MEASURE:: The main outcome measure was clinician compliance with eProtocol-insulin recommendations. RESULTS:: The number of patients was 31 to 458 and the number of blood glucose measurements was 2,226 to 19,925 among the four intensive care units. Clinician compliance with eProtocol-insulin recommendations was 91% to 98%. Blood glucose distributions were similar in the four hospitals (generalized linear model p = .18). Compared with the simple guideline, eProtocol-insulin glucose measurements within target increased from 21% to 39%, and mean blood glucose decreased from 142 to 115 mg/dL (generalized linear model p < .001). Compared with the paper-based protocol, eProtocol-insulin glucose measurements within target increased from 28% to 42%, and mean blood glucose decreased from 134 to 116 mg/dL (generalized linear model p = .001). CONCLUSIONS:: The 91% to 98% clinician compliance indicates eProtocol-insulin is an exportable instrument that can establish a replicable experimental method for clinical trials of blood glucose management in critically ill adults. Control of blood glucose was better with eProtocol-insulin than with a simple clinical guideline or a paper-based protocol.