TL;DR
This paper introduces a comprehensive dataset of polypharmacy patterns in the United States, derived from billions of prescription claims. It quantifies the incidence of multi-drug combinations, revealing that half of patients taking any prescription drug use two or more, and 5% use eight or more, with metabolic syndrome treatments being the most common combinations.
What Problem Does This Paper Address?
Polypharmacy, the concurrent use of multiple prescription drugs, is a growing concern, contributing significantly to drug-related morbidity and healthcare costs. Despite its prevalence, real-world polypharmacy patterns, particularly the specific combinations of drugs used, remain poorly characterized. This lack of detailed understanding hinders efforts to identify and mitigate the risks associated with these complex drug regimens, making it difficult to prioritize research into drug safety and efficacy.
Methods
The researchers developed a robust methodology to analyze outpatient prescription drug claims from the Truven Health MarketScan® Databases. Their approach involved:
- Scanning Prescription Claims: Individual prescription records were processed to determine drug ingredients, dates of service, and days of supply.
- Creating Discrete Exposure Windows: Prescription data was converted into non-overlapping 30-day exposure windows. This duration was chosen as it represents the most common prescription length, providing a natural timescale for assessing concomitant drug use.
- Summarizing and Counting Combinations: Within each 30-day window, the presence of drug ingredients and their corresponding Anatomical Therapeutic Chemical (ATC) drug classes were identified. The study then counted both "exact" combinations (where only a specific set of drugs was taken) and "at least" combinations (where a specific set of drugs was taken, potentially with others).
- Drug Mapping: A comprehensive mapping of National Drug Codes (NDCs) to RxNORM ingredient levels and then to second-level ATC drug classes was established. This allowed for analysis at both the ingredient and class levels.
- Overrepresentation Analysis: The study also quantified the overrepresentation of drug combinations beyond what would be expected by chance, based on individual drug frequencies.
Datasets
The primary dataset used in this study is the Truven Health MarketScan Research Databases (2007-2014). This dataset comprises health coverage records for approximately 100 million employees, dependents, and retirees in the United States, providing over 4 billion patient-months of observation.
The paper also references:
- DrugBank, RxNORM, and UMLS: Used for initial curation of drug ingredients.
- WHO Collaborating Centre for Drug Statistics Methodology: Used for extracting ATC drug class names.
No new benchmarks were introduced; instead, the paper focuses on creating a novel, comprehensive dataset of polypharmacy.
Key Insights and Ideas
- Ubiquitous Polypharmacy: The dataset confirms that polypharmacy is widespread, with half of all patients taking any prescription drug using two or more, and a significant 5% using eight or more drugs concurrently.
- Metabolic Syndrome Dominance: The most frequently observed multi-drug combinations are those used to treat manifestations of metabolic syndrome, highlighting a key area of complex drug management.
- Unique Combinations Exist: A notable 10% of exposure windows involved unique drug combinations not seen elsewhere in the vast dataset, suggesting a high degree of individualization or rare therapeutic approaches.
- Data-Driven Prioritization: The dataset provides a crucial resource for researchers to identify and prioritize common drug combinations for future safety and efficacy studies, moving beyond anecdotal evidence.
- Methodological Rigor: The development of discrete 30-day exposure windows and the distinction between "exact" and "at least" combination counts offer a nuanced approach to quantifying polypharmacy.
Key Results
- The dataset quantifies the incidence of multi-drug combinations across four billion patient-months of outpatient prescription drug claims from 2007-2014.
- 50% of patients taking any prescription drug were exposed to two or more drugs.
- 5% of patients were exposed to eight or more drugs.
- The analysis identified 220 million unique "exact" drug ingredient combinations and 39 million unique "exact" drug class combinations.
- Approximately 10% of exposure windows contained unique drug combinations.
- Table 3 shows that common 3-drug combinations can be associated with approximately 3-fold higher rates of Emergency Department visits.
Limitations and Future Work
- Data Source Limitations: The Truven Health MarketScan data is based on commercial claims and may not be fully representative of the entire US population.
- Adherence Bias: The study relies on filled prescriptions, not actual drug ingestion. Patients may not take their medications as prescribed, introducing bias.
- Over-the-Counter (OTC) Drugs: The analysis only includes prescription drugs, excluding OTC medications and supplements, which also contribute to total drug exposure.
- Non-Stationarity: The analysis assumes stable prescription patterns over time, which may not hold true given the observed increase in polypharmacy prevalence.
Future work could involve:
- Risk Stratification: Using the dataset to stratify patients by their risk of adverse health outcomes based on their prescription sets.
- Causality Exploration: While the dataset identifies associations, future research can explore causal links between specific combinations and health outcomes.
- Integration with Other Data: Combining this prescription data with other sources (e.g., electronic health records, patient-reported outcomes) to gain a more complete picture of drug use and its effects.
- Broader Drug Categories: Expanding the analysis to include OTC drugs and supplements.
Summary updated