The Role of Computer Computer-based Clinical Decision Support Systems (CDSS) in Improving Antibiotic Management

Abstract

Background Inappropriate antibiotic prescribing is a key contributor to increasing antibiotic resistance. Despite the standard practices promoted through clinical practice guidelines (CPGs), treatment regimens are not always in accordance with these guidelines. In Australia, a significant proportion of inappropriate antibiotic prescriptions in hospitals and primary care is due to noncompliance with CPGs. This is further exacerbated by the difficulty faced by clinicians in integrating and managing multiple information streams at the point of care to inform evidence-based decision- making. There is increasing recognition that digital health interventions such as clinical decision support systems (CDSS) may assist in optimising antimicrobial management. CDSS provide treatment recommendations based on patient-specific risk factors and research evidence, allowing clinicians to provide personalised care. Many studies provide evidence of the potential of CDSS for promoting optimal antibiotic management; however, adoption of these systems in clinical practice remains low. In addition to this lack of effective system adoption, there is a high rate of clinicians’ ignoring or overriding the systems’ recommendations or only engaging with partial use of the systems’ features. These factors limit the efficacy of CDSS in improving antibiotic prescribing. Objective The objective of this thesis was to evaluate individual, organisational, and system-level factors that impact CDSS implementation for evidence-based antibiotic management. An understanding of the different aspects of CDSS implementation in Australia has been sought by bringing together the perceptions and experiences of different stakeholders. The project aimed to achieve this objective by i) synthesising the evidence on the efficacy of CDSS for antibiotic management; ii) understanding clinicians’ perceptions regarding CDSS use for optimal antibiotic prescribing; and iii) evaluating the challenges of integrating CDSS into the healthcare system. Methods To achieve the objectives outlined above, the thesis was divided into four studies: In study I, a systematic review and meta-analyses were conducted to evaluate the impact of CDSS implementation on various clinical and economic outcomes associated with antibiotic management. The study protocol was developed using the PRISMA-P checklist. Studies were selected using specific pre-defined study eligibility criteria. Studies providing sufficient data on the outcomes were included in the meta-analyses to calculate pooled effect estimates of the impact of CDSS implementation on antibiotic management. In studies II & III, a cross-sectional online survey was conducted in Australia. Clinicians directly involved in prescribing, administering, and managing antibiotics in hospital and primary care settings were invited to participate. We adopted the Unified Theory of Acceptance and Use of Technology (UTAUT) model to understand factors contributing to clinicians’ inappropriate antibiotic prescribing behaviour and their behavioural intent to adopt CDSS. Using this framework, we also examined the role of moderating factors such as gender, age, clinical experience, and care settings in shaping users’ behaviour in adopting CDSS. We used multivariate logistic regression models to investigate the association between these moderating factors and users’ perceptions regarding CDSS adoption. Finally, in study IV, we used a qualitative approach to conduct in-depth interviews with policymakers involved in the implementation and evaluation of CDSS in Australia. The focus of this study was to understand what is required to effectively scale-up CDSS implementation from pilot studies to a system-wide innovation. Participants shared their experiences and perceptions concerning the gaps and challenges in the Australian healthcare system for integration of CDSS into healthcare processes. The interview transcripts were thematically analysed to establish a contextual understanding of the system-wide challenges for CDSS implementation. Results Results from this research highlight that CDSS can help reduce the risk of inappropriate antibiotic prescribing by increasing compliance with prescribing guidelines. The findings further indicate that CDSS can improve antibiotic prescribing by reducing the volume of overall antibiotic use, duration of therapy, length of hospital stay and thereby decreasing the overall cost of therapy. However, most of the evidence included in our systematic review was from studies having moderate to low methodological quality. Non-randomised studies tended to overestimate the effect of CDSS on appropriate antibiotic management, compared to randomised studies. However, the direction of the effect was largely consistent across both study types and favoured the positive impact of CDSS for antibiotic management. There was also substantial statistical heterogeneity in the results across the included studies which can be explained by the large variability in CDSS adoption across studies. Findings from the survey with clinicians indicated that different individual and setting specific characteristics are important factors that influence clinicians’ perceptions regarding CDSS adoption and lead to variability in uptake across different clinicians. Experienced clinicians were more sceptical of using CDSS for clinical decision-making, potentially due to limited digital health literacy, mistrust in the information provided by CDSS and fear of compromising their professional autonomy. Similarly, in comparison to users, CDSS non-users were more likely to lack trust in CDSS recommendations and fear compromising their professional autonomy due to CDSS adoption. A lack of transparency and explainability in CDSS design, in which end-users are not aware of how systems have computed recommendations can reduce their trust in CDSS. Consistent with the context of primary care, primary care clinicians believed that time constraints and patient expectations were important drivers of CDSS adoption. These findings highlight that the efficacy of CDSS implementation may be limited by a lack of consideration of contextual factors such as clinical experience, setting of use, and users’ skills which impact the users’ behaviour to adopt CDSS. Targeted clinician engagement, digital health literacy and better communication of the reliability of information provided may assist with more successful implementation of CDSS at point of care. Interviews with Australian policymakers further explored system-level challenges and gaps that may impede successful CDSS implementation. The results show that the lack of shared vision between different stakeholders, and the fragmented infrastructure within the healthcare system are major barriers to the integration of CDSS within existing processes in the healthcare system. CDSS implementation needs to be supported by an effective governance structure that can establish clear roles, prioritise investment in health system capacity building and incorporate cross-discipline and inter-organisational collaboration for quality data sharing. The ability of CDSS to ensure coordinated and interoperable care by exchanging information across organisations requires mutually agreed data standards at a national level. There is a need to establish standards not only for generating data in a standardised format, but for semantic interoperability that allows data communication and interpretation across different systems. Notwithstanding the significance of standardisation to ensure interoperability in CDSS, our findings also highlight that this standardisation must be balanced with adequate flexibility in the CDSS design and implementation process, so that user and setting specific requirements can be incorporated to improve adoption. Conclusion In conclusion, our findings illustrate that CDSS reflects best practice for antibiotic management through evidence-based clinical decision making, integrating the knowledge base, and flagging medication errors. The integration of these systems in healthcare settings is, however, challenging due to the complex interaction between the system, organisational and human factors. The findings from our research suggests that individual and setting characteristics such as clinical experience, use of CDSS and the type of setting, influence the clinicians’ perception of CDSS role in antibiotic management. These characteristics provide a better understanding of why CDSS adoption varies across different clinicians and care settings. We also found that the lack of synergy evident between multiple stakeholders and organisations - who seem to have varying interests and objectives regarding CDSS implementation - is limiting the ability to develop a shared vision and collaborative action. These findings provide evidence firm foundation for policymakers for developing a holistic CDSS implementation framework that considers the interaction of the system within the context of organisational and human behavioural characteristics. Implementation processes need to be tailored to specific user and setting requirements for improved adoption and use of CDSS by clinicians. A better understanding of the clinical culture would support successful CDSS implementation, along with effective strategies to develop broader digital literacy, methods for sustaining clinicians’ engagement with the technology, and approaches to facilitating cross-discipline collaboration.