FAERS Follow-Up Versions: Receipt-Date Lag, Cohort Age, and Extraction Rules
A review of 11,175,364 valid date pairs separates version number from receipt-date lag and measures right-censoring across recent cohorts.
The current FAERS extract contains 11,175,364 reports with valid first-receipt and latest-receipt date pairs. Version numbers identify 3,547,090 follow-up reports, or 31.7 percent of the cohort. Receipt-date lag exceeds 90 days for 1,702,168 reports and one year for 745,240 reports. The median lag is zero days, while the mean is 84.1 days and the 99th percentile is 1,512 days. Recent first-receipt cohorts show shorter observed lags because they have had less time to accumulate follow-up. These measurements require an extraction date and fixed first-receipt cohorts. They measure record updates, not event-to-report delay.
1. Cohort and counting rules
The cohort includes the latest available row for each report with nonmissing standardized first-receipt and latest-receipt dates. The latest date must not precede the first date. The analysis excludes 217 of 11,175,581 rows because they fail these date rules. Version 1 denotes an initial submission. A version number above one denotes at least one follow-up submission. Lag is the calendar-day difference between the two receipt dates.
2. Version numbers identify 3.55 million follow-up reports
Version 1 accounts for 7,628,274 reports. Version numbers above one account for 3,547,090 reports, or 31.7 percent. FDA states that an initial submission receives version 1, each follow-up receives the next version, and the latest version represents current case information. The extract therefore supports a follow-up classification without treating calendar dates as version identifiers.
3. Receipt-date lag has a long right tail
The latest receipt date equals the first receipt date for 7,676,411 reports. Another 988,326 reports have lags of 1 to 30 days. Lags exceed 90 days for 1,702,168 reports, including 745,240 above one year. The median is zero days because most rows have no positive lag. The mean of 84.1 days is pulled upward by the long tail and should not represent a typical report.
4. Same-day dates and version numbers are not interchangeable
Among 7,676,411 same-day records, 48,141 have version numbers above one. These cases received a follow-up without crossing a calendar-day boundary. Four version-1 rows have positive lags, which indicates exceptional source coding rather than a usable revision rule. Calling every same-day record “never revised” would therefore misclassify at least 48,141 reports.
5. Recent cohorts are right-censored
First-receipt cohorts from 2020 through 2025 show progressively shorter observed tails. The 90th-percentile lag declines from 278 days for 2020 reports to 82 days for 2025 reports. The share above one year falls from 8.0 to 0.5 percent. This pattern is expected when newer reports have less observation time. It does not establish faster case completion.
6. Stable analysis requires extraction and cohort dates
FDA states that FAERS queries can change because reports update and final versions vary. A reproducible trend should record its extraction date, assign records by first receipt, and preserve the same cohort limits during comparison. Later extracts can add follow-up versions to earlier cohorts. Analysts should report this maturation instead of silently replacing the earlier result.
7. Supported interpretation
The extract supports three conclusions. Nearly one-third of current report rows carry a follow-up version. Positive receipt-date lags have a long right tail. Recent cohorts are mechanically less mature than older cohorts. The data do not show why follow-up occurred, whether the clinical account changed materially, or when the underlying event was reported.
8. Limitations
The public quarterly record preserves the latest version rather than a complete public change history in this analysis. Version counts do not reveal which fields changed. Same-day follow-ups collapse to zero calendar days. Newer cohorts have shorter follow-up windows. Reports can be duplicated, corrected, delayed, or stimulated by publicity and litigation. Later extracts can change every reported total.