High-Throughput Computing to Detect Harmful Drug-Drug Interactions in Older Adults: A Clinical Research Protocol (Preprint)
Bibliographic record
Abstract
BACKGROUND Drug-drug interactions (DDIs) are a major concern, especially for older adults taking multiple medications. While Health Canada and the US Food and Drug Administration (FDA) use population-based studies to identify adverse drug events, detecting harmful DDIs is challenging due to the millions of potential drug combinations. Traditional pharmacoepidemiologic studies are slow and inefficient, often missing important harmful DDIs. OBJECTIVE This protocol outlines a novel approach to efficiently identify harmful DDIs using administrative healthcare data. METHODS Using high-throughput computing, we will conduct multiple population-based, new-user cohort studies using Ontario's linked administrative healthcare data. The cohorts will be selected from the population of Ontario residents aged 66 and older who filled at least one oral outpatient drug prescription from 2002 to 2023. In each cohort, the exposed group will comprise individuals who are regular users of one drug (Drug A) who start a new prescription for a second drug (Drug B); the referent group will comprise regular users of Drug A not taking Drug B. We will evaluate 74 acute outcomes within 30 days of cohort entry, including hospitalizations, emergency department visits, and mortality. Propensity score methods will balance exposed and referent groups on 400+ baseline health characteristics. Modified Poisson and binomial regression models will estimate risk ratios and differences. To ensure findings are both statistically and clinically meaningful, we will apply pre-specified thresholds for effect sizes (e.g., lower bounds of 95% confidence intervals ≥1.33 for risk ratios and ≥0.1% for risk differences) and control the false discovery rate at 5% using the Benjamini-Hochberg procedure to address multiplicity. Subgroup and sensitivity analyses, including negative control outcomes and E-values, will assess robustness. RESULTS In a preliminary analysis, we identified approximately 3.8 million older adults who filled prescriptions for over 500 unique medications during the study period, and therefore, approximately 200,000 potential drug combinations will be available for study. The initial drug-pair cohorts had a median of 583 new users per cohort (interquartile range (IQR): 237- 2130); the median overlap in drug-pair prescriptions was 57 days (IQR 30-90). CONCLUSIONS This study will aim to identify credible signals of harmful DDIs in older adults in routine care. This study will use an innovative approach that leverages data from provincial administrative healthcare databases and integrates high-throughput computing and rigorous pharmacoepidemiologic methods to generate robust real-world evidence that can inform safer prescribing practices and regulatory decision-making.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.035 | 0.067 |
| Meta-epidemiology (narrow) | 0.003 | 0.002 |
| Meta-epidemiology (broad) | 0.003 | 0.005 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.004 | 0.002 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.004 | 0.004 |
| Research integrity | 0.004 | 0.005 |
| Insufficient payload (model declined to judge) | 0.120 | 0.039 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".