From Variation to Standardisation: An Electronic Health Record Intervention to Improve Risk Stratification and Admission Decisions in Acute Lower Gastrointestinal Bleeding

BackgroundAcute lower gastrointestinal bleeding (LGIB) is a common surgical emergency with wide variation in management despite British Society of Gastroenterology (BSG) guidance.The Oakland score is validated to identify patients suitable for safe discharge, but uptake in routine practice is inconsistent.Where a validated score is not used, disposition reverts to unstructured judgement, risking both premature discharge of highrisk patients and unnecessary admission of low-risk ones.Against a guideline-derived standard of 100%, we aimed to improve documented risk stratification and the appropriateness of admission and discharge decisions. MethodsWe conducted a two-cycle retrospective closed-loop audit of consecutive adult LGIB presentations at a district general hospital, taking the bleeding episode as the unit of analysis (28 episodes in cycle 1, 22 in cycle 2).Between cycles, we introduced a multifaceted intervention comprising clinician education, ward posters, a departmental email, and an electronic health record SmartPhrase that assembles the seven components of the Oakland score at clerking, populating four automatically from the patient record.Primary outcomes were documentation of the Oakland score and concordance of the admission or discharge decision with the patient's Oakland risk category.Proportions were compared by Fisher's exact test using identical denominators in both cycles, with absolute differences and 95% confidence intervals by the Newcombe method. ResultsOakland score documentation rose from 6/28 (21.4%) to 19/22 (86.4%), an absolute increase of 64.9% (95% CI 38.2 to 79.2; p < 0.001).Decisions concordant with risk category rose from 21/27 (77.8%) to 20/20 (100%), an increase of 22.2% (95% CI 2.4 to 40.8; p = 0.031); the six cycle 1 discordances comprised three premature discharges (Oakland scores 9, 16 and 22) and three unnecessary admissions.Outpatient investigation was arranged in 9/22 (40.9%) versus 13/18 (72.2%), a difference of 31.3% (95% CI 0.5 to 54.7; p = 0.062).Transfusion fell within the restrictive threshold in 3/5 versus 8/8 transfused episodes (difference 40.0%, 95% CI -3.0 to 76.9; p = 0.128), with two inappropriate transfusions in cycle 1 and none in cycle 2. Anticoagulant management (12/15 versus 12/13; difference 12.3%, 95% CI -16.4 to 38.3; p = 0.60), in which the absolute number managed correctly was unchanged, and CT angiography for shock index >1 (2/4 versus 1/1; p = 1.00) showed no significant change. ConclusionsEmbedding risk stratification in the clerking workflow was associated with a substantial increase in documented Oakland score use, and no admission or discharge decision discordant with calculated risk was observed in the second cycle.The intervention was multifaceted and introduced over a short interval in a small single-centre cohort, so the contribution of the SmartPhrase cannot be isolated and causal attribution is not possible.The approach is low-cost, trainee-led and reproducible wherever a configurable electronic health record is used; confirmation that the effect is sustained will require a further cycle including patientlevel outcomes.