Can Advanced Analytics Transform High-Dose Antipsychotic Prescribing Safety?

What Are the Study Objectives and Key Research Questions?

The high-dose antipsychotic prescribing landscape in acute hospital settings across Scotland will be comprehensively examined in a new retrospective cohort study, using multiple national datasets to understand both the prevalence and impact of this common yet potentially risky prescribing practice. This research aims to address a critical knowledge gap in understanding how high-dose antipsychotics are used outside specialized psychiatric settings, where patients may be particularly vulnerable to adverse effects without specialized mental health oversight. The study will explore four key research questions: the prevalence and factors associated with high-dose prescribing, the clinical profiles of patients receiving these regimens, the short-term and long-term impacts compared to standard dosing, and the potential for machine learning models to enhance clinical decision-making in this context.

The study will analyze data from six Scottish health boards through the Hospital Electronic Prescribing and Medicines Administration (HEPMA) system, alongside twelve linked national health datasets including the Prescribing Information System (PIS) and Scottish Morbidity Records (SMR). This robust data linkage approach will allow researchers to track patients who received antipsychotics between January 2019 and December 2023, providing a comprehensive view of prescribing patterns, administration practices, and subsequent health outcomes. By examining both binary exposure (whether a patient ever received a high dose) and time-dependent continuous exposure (actual duration of high-dose therapy), the researchers will capture nuanced patterns of antipsychotic use that previous studies may have missed. The dual approach to defining "high dose" – using both chlorpromazine equivalents (>1000 mg/day) and British National Formulary percentage thresholds (>100% of maximum recommended dose) – further strengthens the methodology by allowing comparison across different classification systems.

Can High-Dose Antipsychotic Regimens Be Justified in Acute Care?

Antipsychotic medications represent a cornerstone of treatment for severe psychiatric disorders, yet their use at doses exceeding recommended maximums carries substantial risks. UK audit data indicates that over one-third of psychiatric inpatients receive high-dose antipsychotics, with approximately 43% prescribed multiple antipsychotics simultaneously. Even at standard therapeutic doses, these medications are associated with serious adverse outcomes including cardiovascular events, metabolic disturbances, and extrapyramidal symptoms. The risk-benefit calculation becomes particularly complex in acute hospital settings, where patients may have multiple comorbidities and where prescribers may have limited psychiatric expertise. The study's examination of both benefits (psychiatric disease control measured through hospitalization patterns and changes in adjunctive medication use) and harms (metabolic, cardiac, extrapyramidal, and cognitive effects) will provide a balanced assessment of current prescribing practices. Could these findings ultimately lead to more tailored approaches to antipsychotic prescribing that better account for individual patient characteristics and risk factors?

How Can Methodological Innovations Prevent Bias?

To address potential immortal time bias, the researchers will treat high-dose exposure as a time-dependent variable, ensuring patients contribute person-time as unexposed until they first receive a high-dose antipsychotic. This methodological approach is crucial for preventing bias that could arise if patients were classified as exposed from cohort entry before actually receiving a high-dose prescription. For patients receiving PRN (pro re nata) medications, only doses that were actually administered will be counted toward exposure measures, rather than relying solely on prescription data. This distinction is important as it ensures that exposure measures accurately reflect actual patient intake rather than intended prescribing, which may differ substantially in practice. The study will also employ appropriate methods to handle missing or incomplete data, with sensitivity analyses planned to evaluate the impact of missing data on study results.

Can Advanced Analytics Enhance Predictive Accuracy in Clinical Settings?

The researchers will employ sophisticated statistical methods to identify factors associated with high-dose prescribing and to assess outcomes. Beyond descriptive analyses characterizing the study cohort, they will use Kaplan-Meier survival curves and Cox proportional hazards models for time-to-event analyses. Machine learning algorithms including random forest classification and gradient boosting will enhance predictive accuracy and model generalizability. These advanced analytical techniques will help identify which patients might benefit from high-dose regimens versus those at greater risk of harm, potentially transforming clinical decision-making. Propensity score matching and weighting methods will address potential confounding when comparing dose groups, strengthening the validity of the findings. How might these predictive models be implemented in clinical practice to support real-time decision-making at the point of prescribing? Would integration into electronic prescribing systems provide the most effective pathway to clinical impact?

What Are the Strengths and Limitations of the Study Design?

The study design includes several methodological strengths, including the use of multiple national datasets to capture both prescription and administration data, the application of both binary and time-dependent continuous exposure measures, and the comparison of two different methods for calculating high doses. However, important limitations must be acknowledged. The absence of electronic data from psychiatric inpatient settings restricts analyses to acute hospital populations, potentially limiting the completeness of exposure data for patients who move between settings. Missing or incomplete data across linked datasets may introduce bias despite planned imputation strategies. Additionally, secular trends in clinical guidelines and antipsychotic availability between 2019 and 2023 may have influenced prescribing patterns in ways that complicate interpretation. Despite these limitations, the findings will likely provide valuable insights into current prescribing practices and their clinical implications. How might future research address these limitations to build a more comprehensive understanding of high-dose antipsychotic use across all healthcare settings?

How Might These Findings Inform Safer Antipsychotic Stewardship?

Beyond its immediate clinical relevance, this study may have broader implications for antipsychotic stewardship programs. By identifying patterns of high-dose prescribing and associated outcomes, the findings could inform the development of targeted interventions to promote safer, evidence-based prescribing practices. The machine learning algorithms developed through this research might eventually serve as clinical decision support tools, flagging high-risk prescribing scenarios and suggesting alternative approaches. The study's focus on both effectiveness and safety outcomes acknowledges the complex balance clinicians must strike when treating severe psychiatric conditions, particularly in medically complex patients. What organizational and systemic factors might influence the implementation of any recommendations arising from this research? How might barriers to change be addressed across different healthcare settings?

Who Will Benefit from the Disseminated Findings?

The researchers plan to disseminate their findings through peer-reviewed journals, conferences, policy briefs, and public outreach, maximizing the potential impact on mental health practices and policy. While patient and public involvement was not included in the study design, future iterations will involve patient groups in developing dissemination materials to improve the accessibility and relevance of findings. This comprehensive approach to knowledge translation recognizes the importance of engaging multiple stakeholders in improving prescribing practices. As antipsychotic medications continue to play a critical role in the management of severe psychiatric disorders, research that illuminates the real-world patterns of use and associated outcomes will be essential for optimizing patient care. What additional stakeholders might benefit from targeted dissemination of these findings, and how might their engagement enhance the translation of research into practice?

Summary

A new retrospective cohort study will comprehensively examine high-dose antipsychotic prescribing practices in acute hospital settings across Scotland, analyzing data from six health boards and twelve national health datasets between January 2019 and December 2023. The research addresses a critical knowledge gap regarding the prevalence, clinical impact, and safety of high-dose antipsychotic use outside specialized psychiatric settings, where patients may be particularly vulnerable to adverse effects. Using advanced methodological approaches including time-dependent exposure variables, dual dose classification systems (chlorpromazine equivalents and British National Formulary thresholds), and machine learning algorithms, the study will identify factors associated with high-dose prescribing and assess both benefits and harms of these regimens. The researchers will employ sophisticated statistical methods including Kaplan-Meier survival curves, Cox proportional hazards models, random forest classification, and gradient boosting to enhance predictive accuracy. Key strengths include the use of multiple national datasets capturing both prescription and administration data, while acknowledged limitations include the absence of psychiatric inpatient data and potential bias from missing data across linked datasets. The findings aim to inform safer antipsychotic stewardship programs and may eventually support clinical decision-making through predictive models integrated into electronic prescribing systems. Dissemination strategies will target multiple stakeholders through peer-reviewed publications, conferences, policy briefs, and public outreach to maximize impact on mental health practices and policy.