A Comprehensive Resource for Real-World Drug-Diagnosis Relationships

Understanding Real-World Drug Use in Healthcare

In the pursuit of personalized medicine, researchers and clinicians rely heavily on electronic health records (EHRs) to identify patterns and relationships between medications and diagnoses. However, the existing literature-based drug-disease maps often fail to accurately reflect real-world scenarios, particularly in the context of polypharmacy. This article presents a novel resource, curated from approximately 1.5 million individuals' primary and secondary healthcare data, to address this knowledge gap.

What is Polypharmacy?

Polypharmacy refers to the concurrent use of multiple medications by a patient. This practice is common, especially among older adults, and can lead to adverse drug reactions, interactions, and increased healthcare costs.

The Need for a Real-World Resource

Current literature-based drug-disease maps are based on mechanistic disease ontologies, which are not widely used in clinical settings. These maps also fail to align well with the International Classification of Diseases (ICD) system, which is the most commonly used healthcare classification system. To bridge this gap, our team developed a comprehensive resource describing real-world drug-diagnosis relationships.

Methodology

Our study utilized real-world primary and secondary healthcare data from approximately 1.5 million individuals. We identified:

1. Drug-Diagnosis Co-occurrences: We analyzed ~736,000 pairs of drug-diagnosis co-occurrences to understand the frequency of medication use in relation to various diagnoses.
2. Significant Associations: We identified 7,763 pairs of significantly associated drugs and diagnoses using statistical analysis.
3. Direct Drug Usage: We manually annotated the significant associations using expert curation by six independent clinicians (three in the UK and three in Denmark).

The Resource

The final resource comprises 7,763 associations with odds ratios [≥]3.5, manually annotated by our team of clinicians. To assess the confidence of each pair, we employed an Expectation-Maximization-based framework, providing a confidence score for each association.

Key Findings

Our resource reveals that the vast majority of significantly associated drugs and diagnoses in healthcare records are not due to direct treatment of the diagnosis. This finding highlights the importance of considering indirect relationships between medications and diagnoses in clinical practice.

Implications for Clinicians

Our resource provides a valuable tool for clinicians to better understand real-world drug-diagnosis relationships. By considering the indirect relationships between medications and diagnoses, clinicians can:

1. Improve Patient Safety: By identifying potential adverse drug reactions and interactions, clinicians can take steps to mitigate these risks.
2. Enhance Patient Care: By considering the complex relationships between medications and diagnoses, clinicians can develop more effective treatment plans.
3. Inform Healthcare Policy: Our resource can inform healthcare policy decisions by highlighting the need for more accurate and comprehensive drug-disease maps.

Limitations and Future Directions

While our resource provides a significant step forward in understanding real-world drug-diagnosis relationships, there are limitations to consider. The use of real-world data can introduce systematic differences among annotators, which we addressed using an Expectation-Maximization-based framework. Future studies should focus on validating our resource in diverse clinical settings and exploring its potential applications in precision medicine.

Conclusion

Our comprehensive resource provides a valuable tool for clinicians and researchers to better understand real-world drug-diagnosis relationships. By considering the indirect relationships between medications and diagnoses, we can improve patient safety, enhance patient care, and inform healthcare policy decisions.