How many in‑custody deaths does ICE list for each calendar year from 2003 through 2025 when aggregated from its PDF and FY2018+ reports?
Executive summary
ICE publishes two complementary sources for in‑custody deaths: legacy PDF compilations covering October 2003 through mid‑2017 and post‑2018 death notices/FY reports on its website, but the documents provided here do not contain a single pre‑aggregated, year‑by‑year table spanning 2003–2025; publicly cited reporting shows 2025 tied or exceeded the 2004 high with 32 deaths according to multiple outlets sourcing ICE counts (The Guardian, POGO, Reuters) while ICE’s own death‑reporting page covers procedures and individual notices rather than a neatly aggregated calendar‑year list [1] [2] [3] [4] [5].
1. What the user is actually asking and why it’s hard to answer directly
The request seeks a simple annual tally from 2003 through 2025 “when aggregated from its PDF and FY2018+ reports,” which implies combining ICE’s older FOIA PDF lists (covering 10/2003–mid‑2017) with the agency’s FY2018 and later public death notices—but the materials provided here are fragments: an ICE FOIA PDF that logs deaths from October 2003 to June 2017 and ICE’s detainee death reporting page and newsroom notices that publish individual post‑2018 reports, none of which were supplied as an already‑aggregated calendar‑year table in the documents available to this review [1] [6] [2].
2. What can be stated with confidence from the supplied reporting
Multiple reputable outlets relying on ICE’s counts report that 2025 saw 32 in‑custody deaths, the most since 2004 and matching the earlier record, and that 2025’s December was especially deadly [3] [4] [5]. The ICE FOIA PDF exist and cover detainee deaths beginning October 2003 through mid‑2017, which is the likely source for year‑by‑year deaths from 2003–2017 if one extracts and aggregates the PDF entries [1] [6]. ICE’s detainee death reporting page documents the notification and posting process and hosts post‑2018 death notices, which are the necessary source for FY2018+ counts [2].
3. What cannot be produced from the provided sources alone
The precise calendar‑year counts for every year 2003–2025 cannot be enumerated here because the underlying PDFs and the full set of FY2018+ death notices—needed to aggregate the full series—were not provided in parsed, year‑by‑year form in the materials reviewed; therefore it would be improper to fabricate a complete table without extracting and reconciling every individual record from ICE’s FOIA PDFs and the FY2018+ notices [1] [2]. Any definitive annual series requires directly parsing ICE’s 2003–2017 PDF lists and the post‑2018 newsroom/detainee death entries.
4. How an accurate aggregation would be produced (and where the user can get it)
To produce the requested calendar‑year totals one must (a) extract each death date from ICE’s FOIA PDFs that cover 10/2003–mid‑2017 and assign them to calendar years, and (b) compile the ICE newsroom/detainee‑death notices from FY2018 onward and similarly tabulate by calendar year; ICE’s FOIA death lists and its detainee death reporting page are the official sources to be parsed to complete that aggregation [1] [2]. Independent trackers and media outlets (The Guardian, Reuters, Statista, POGO) have already used those sources to report selected milestones—most prominently that 2025 reached 32 deaths, the highest since 2004—but their summaries are not a substitute for a full, documented extraction of ICE’s own records [3] [5] [7] [4].
5. Context, competing narratives, and implicit agendas
Advocates and oversight groups emphasize preventability and under‑reporting, noting ICE’s past practices and calling for transparency, while ICE stresses procedural reporting requirements and multilayered reviews; media outlets often highlight spikes to argue for policy change, whereas ICE and DHS emphasize compliance with death‑reporting rules—readers should note these differing agendas when interpreting counts and causes [8] [4] [2]. Reported milestones—2025’s 32 deaths and that figure matching the 2004 peak—are consistently cited across several sources relying on ICE’s own tallies, but the user’s requested year‑by‑year breakdown requires the raw aggregation step that the supplied materials do not complete here [3] [4] [5].