Electronic risk assessment for venous thromboembolism: Investigating physicians' rationale for bypassing clinical decision support recommendations

Objective: The underutilisation of venous thromboembolism (VTE) prophylaxis is still a problem in the UK despite the emergence of national guidelines and incentives to increase the number of patients undergoing VTE risk assessments. Our objective was to examine the reasons doctors gave for not prescribing enoxaparin when recommended by an electronic VTE risk assessment alert. Design: We used a qualitative research design to conduct a thematic analysis of free text entered into an electronic prescribing system. Setting: The study took place in a large University teaching hospital, which has a locally developed electronic prescribing system known as PICS (Prescribing, Information and Communication System). Participants: We extracted prescription data from all inpatient admissions over a 7-month period in 2012 using the audit database of PICS. Intervention: The completion of the VTE risk assessment form introduced into the hospital-wide electronic prescribing and health records system is mandatory. Where doctors do not prescribe VTE prophylaxis when recommended, they are asked to provide a reason for this decision. The free-text field was introduced in May 2012. Primary and secondary outcome measures: Free-text reasons for not prescribing enoxaparin when recommended were thematically coded. Results: A total of 1136 free-text responses from 259 doctors were collected in the time period and 1206 separate reasons were analysed and coded. 389 reasons (32.3%) for not prescribing enoxaparin were coded as being due to 'clinical judgment'; in 288 (23.9%) of the responses, doctors were going to reassess the patient or prescribe enoxaparin; and in 245 responses (20.3%), the system was seen to have produced an inappropriate alert. Conclusions: In order to increase specificity of warnings and avoid users developing alert fatigue, it is essential that an evaluation of user responses and/or end user feedback as to the appropriateness and timing of alerts is obtained.