Evaluating a digital sepsis alert in a multi-site hospital: a natural experiment

Abstract

Listen

Translate article

Abstract Background This study investigated the impact of a digital sepsis alert on patient outcomes in a busy London multi-site hospital. Sepsis is a serious illness and common cause of death, but rapid diagnosis and treatment improve patient outcomes. Digital health records allow algorithms to be embedded which ‘alert’ clinicians to patients who are at risk of developing sepsis. Despite the current promotion of ‘digital health’, evidence of the impact of algorithm driven alerts on patient outcomes is limited. Methods A retrospective natural experiment utilising the phased introduction of a digital sepsis alert into a large, multi-site hospital in England. Silent alerts (not visible to clinicians) acted as controls. Outcome measures were in-hospital all-cause mortality within 30 days of the alert, extended hospital stay (≥7 days) and timely antibiotics (≤60 minutes of the alert). Inversely weighted multivariable logistic regression was used to determine associations between alert and patient outcomes. Results In a sample of 21,183 inpatients, the mortality rate was 5.9%. The active, visible alert was associated with lower odds of death (Odds Ratio (OR):0.76; 95%CI:(0.70, 0.84)). In 9988 emergency department attendances ending in admission, 40.6% had an extended hospital stay and 41.5% received timely antibiotics. The active alert was associated with lower odds of extended hospital stay (OR:0.93; 95%CI:(0.88, 0.99)) and increased odds of receiving timely antibiotics (OR:1.71; 95%CI:(1.57, 1.87)). Conclusions This study demonstrates that a move to digital health, through an automated sepsis alert, embedded in digital health records, was associated with improved health outcomes. Further work is needed to identify the causal pathway, which is likely to include more rapid treatment with antibiotics, and possible unintended consequences. These findings support the ongoing roll out of digital alerting and provide a model for robustly evaluating their impact. Key messages The introduction of an automated sepsis alert associated with the use of improvement methodology was associated with improved process measures and patient outcomes. Introduction of digital health interventions can, and should, be robustly evaluated with appropriate statistical approaches.