Client profile
BioWatch, a UK life sciences safety services company with 120 staff.
Challenge and scope
Screeners review 4,200 articles a month manually. Eighteen percent fall outside the review window in the baseline.
The same safety-relevant information can appear in a conference abstract, a journal article, or multiple database entries. Search results can also contain product name variants that need careful screening. The review record needs enough context for a safety specialist to understand how a potential case was identified.
Delivery workflow
Screen six databases for 22 product families. Apply approved inclusion rules, record screening decisions, and escalate potential individual case safety reports (ICSRs) for review.
Run the approved searches and log the source, date, and applicable product or term. Group duplicate publications with a reference to the retained record. Screen the title, abstract, or full text as required by the client’s criteria, then document the reason for inclusion, exclusion, or escalation.
Have reviewers check a sample of exclusions as well as possible case safety reports. Keep the search version and evidence trail traceable when a product list or search term changes. Give the client a handoff with the publication reference, relevant excerpt locations, and unresolved details for review.
Delivery team
Poland, about 9 FTE. 1 PV lead, 2 safety reviewers and 6 screening associates.
Systems and tools
Embase, PubMed, client safety database.
Client decision rights
The company owns causality assessment, signal evaluation, and regulatory reporting.
Early milestone
In week 4, screening time is 5.1 minutes per article.
Outcomes
Company-reported results, with the baseline, measurement window and supporting record described for each measure.
| Measure | Baseline | Result | Window | Supporting record |
|---|---|---|---|---|
| Screening time per article | 8.4 min | 3.0 min | 5 months | Safety database report |
| Articles outside review window | 18% | 1.2% | 5 months | Safety database report |
| Valid ICSR detection rate | 91% | 99.4% | 5 months | QA review sample |
Interpreting the results
The review time per article falls from 8.4 to three minutes while detection improves and fewer articles pass the review window. Specify the unit for screening counts when duplicates are present and use known cases to measure detection. Track extraction, review, and escalation timestamps for delays.
Recommendations for new clients
For life sciences companies with recurring literature queues, test against known positive articles and agree how to handle duplicate publications. Monitor detection rate alongside speed.
Provide approved search queries, product and synonym lists, and examples of prior positive and duplicate articles. Run a screening batch with known outcomes and ask the safety reviewer to review both inclusions and exclusions. Confirm how full text and non-English articles enter the client’s workflow.