Consistent and reliable causality attribution at the case level is the cornerstone of confident signal detection. 

The current practice relies on study investigators to establish causal relationships based on their observations. The Sponsor (Company) can add their assessment based on additional information about the drug. The current industry standard, E2B (R3), accounts for multiple assessment methods and presents the data elements for each drug-event pair evaluated by multiple sources in a matrix. 

There are many causality assessment methods used within the industry, some universal, others more specialized. Most commonly used methods include WHO-UMC, Naranjo, Roussel-Uclaf (RUCAM) - to detect drug-associated liver injury, Karch and Lasagna, the French PV Algorithm, Bayesian Adverse Reactions Diagnostic Instrument (BARDI), MacBARDI, and Updated Logistic method. Expert judgment remains the most common method used. 

Serious challenges prevent the practical implementation of existing algorithms by the industry. Many of the algorithms cannot be applied rigorously because of missing data. Additionally, an accurate definition of clinical harm is often lacking (e.g., peripheral neuropathy, vasculitis). Brighton Collaboration Case Definitions partly address this component.

Algorithms do not consider medication errors and are not easy to use with interactions, contributory causation, or secondary harms. Information obtained from the reporter is usually insufficient to establish a causal relationship, and follow-up requests for information must be sent, often repeatedly. The result is a very high share of unassessable reports and poor internal consistency of existing assessments. 

I suggest modifying the ADE reporting to incorporate components enabling structured causality assessment directly by the reporting physician (postmarket) or investigator (clinical trials). Guiding questions would assist the reporting physician in determining causal relationships and facilitate algorithmic attribution upon submission: 

Temporal relationship is a key component of causality assessment. Safety databases routinely calculate latency and last dose latency that feed the algorithm. 

Dechallenge and Rechallenge represent key concepts in pharmacovigilance. This information is typically missing from reports. A series of questions regarding Outcome and Response (Action taken with drug) guide the reporting physician through a checklist for all suspect and interacting drugs, reliably and consistently calculating dechallenge/rechallenge for each drug-event pair. 

Biological plausibility is a complex component requiring knowledge of the drug and the patient's medical condition. The current process relies on extensive follow-up questions to the reporting physician about the patient's medical history and other confounding factors. A number of methods can be introduced to ensure relevant information is collected at this point, including drug-drug interactions or labeling. 

Finally, it is important to ask the reporting physician about any underlying diseases that could have contributed to the event. A clear answer to this question is an essential component of the causality assessment algorithms. 

The result is a rich, detailed, high-quality report with consistent, reproducible, and reliable causality assessment obtained directly from the primary source. Accurate, consistent attribution of causality is central to automated safety signal detection.