Methods

Study design

A contextual inquiry methodological approach was used, including observations and semi-structured interviews informed by the Systems Engineering Initiative for Patient Safety (SEIPS) model.10 The SEIPS model provides a framework to understand work systems, processes and outcomes in health care and has been used extensively in patient safety and AMS research.11,12
This study was approved by the Hospital’s Human Research Ethics Committee (2020/ETH02859). The Consolidated Criteria for Reporting Qualitative Research (COREQ) Checklist13 was followed for the methodology and reporting of this study.

Setting

This study was performed at two metropolitan public teaching hospitals within the same local health district in New South Wales, Australia. Hospital A has 750-beds and Hospital B has 1000-beds. Both hospitals have an AMS team comprised of consultant and trainee infectious diseases (ID) and/or microbiology doctors and a senior clinical pharmacist. Restricted antimicrobials require approval from the AMS team.
Both hospitals use the same eMMS (Cerner Millennium®) with the exception of the ICU at Hospital B, which uses IntelliSpace Critical Care and Anaesthesia (Koninklijke Philips N.V©). The Cerner Millennium eMMS was implemented at Hospital A in May 2015 and Hospital B in September 2017. The hospitals have also implemented a clinical dashboard, Live AMS, into the Cerner eMMS in October 2017, with new features added in the subsequent years. Live AMS allows clinicians to view all patients on antimicrobials, including medication order information such as prescriber, indication and date of prescription. Patients can be filtered using these categories. AMS team representatives from the local health district were consulted during development of Live AMS and were involved in the implementation of the tool at their respective hospitals.

Participants

Purposive sampling was used to recruit participants for observations and interviews, with doctors and pharmacists involved in conducting AMS tasks invited by email. Study participation was voluntary and no compensation was provided. Written consent was obtained for in-person observations and interviews. Verbal consent was obtained for video-conferencing sessions.

Data collection

Data was collected between May 2021 and August 2022. During this time COVID-19 restrictions imposed by the government and hospital executives restricted researcher entry into hospitals. Consequently, observations were conducted through video-conference, and in-person once restrictions were lifted. Clinician(s) were observed for a maximum of 2 hours at a time. Multidisciplinary AMS team meetings were also observed. During the observations, the researcher would occasionally ask participants to explain what they were doing. The researcher took handwritten notes with a focus on all elements of the work system (i.e. SEIPS), including tools, tasks and people involved. Observations ceased when thematic saturation was reached.
Interviews were conducted after observations. A semi-structured interview guide (Appendix 1) was developed by the research team who have expertise in qualitative research, human factors, pharmacy and clinical informatics. The interview guide was piloted with doctors to ensure understanding. Interviews were conducted in-person or through videoconference, audio-recorded and transcribed verbatim.
Observations and interviews were conducted by a researcher (XX) with previous qualitative research experience. XX was not known to the participants prior to commencing the research and was not employed by the hospitals.

Data analysis

Researchers met periodically throughout data collection to discuss observations and interviews. Observation notes and interviews were analysed using an inductive content analysis approach as per the framework method.14 Two researchers (BV, MB) independently coded the first three interview transcripts and met to discuss and reach consensus on codes. The remaining interviews were coded by one researcher (BV) using this framework. Observation notes were coded by one researcher (BV). After inductive coding, themes from interviews and observations were deductively mapped to the SEIPS model. A second researcher (MB) reviewed all transcripts and coding for accuracy and consistency.