How have death rates per 100,000 detainees changed across Bush, Obama, and Trump administrations, and what methodological caveats affect those comparisons?
Executive summary
A clear long‑run decline in measured mortality rates in U.S. immigration detention—when measured on a person‑year basis—was documented from the mid‑2000s through the 2010s, but pandemic and post‑pandemic metrics using different denominators show a large spike in 2020 and much lower rates afterward, and comparisons across Bush, Obama and Trump administrations are sharply constrained by inconsistent methods and incomplete reporting [1] [2]. The headline numbers therefore can be true and misleading at the same time: they reflect real changes in counted deaths but depend on how deaths and populations are counted, and on reporting practices that advocates say undercount certain cases [2] [3] [4].
1. What the published trends say: mid‑2000s to 2014, then 2018–2023 numbers
A peer‑reviewed analysis of deaths in ICE detention from 2003–2015 found the death rate expressed per 100 person‑years fell from 0.146 in 2004 to 0.018 in 2014—a decline that, converted to the more familiar per‑100,000 person‑years metric, equates roughly to 146 deaths per 100,000 person‑years in 2004 falling to about 18 per 100,000 person‑years in 2014 [1]. More recent work focused on fiscal years 2018–2023 reports a pandemic‑era high of 10.833 deaths per 100,000 admissions in FY2020 and lower per‑100,000‑admissions rates of 3.251 in FY2021, 0.939 in FY2022, and 1.457 in FY2023 [2]. Those later figures capture the sharp COVID disruption and partial recovery but use “per admissions” denominators rather than person‑years, producing different scale and interpretation [2].
2. Why “Bush vs Obama vs Trump” is harder than it sounds: denominator and timeframe problems
Directly attributing a single administration to higher or lower “deaths per 100,000 detainees” depends on choosing a denominator (admissions, average daily population, person‑years, or a one‑day custody snapshot) and the timeframe; the academic literature often uses person‑years to adjust for length of stay while government tables sometimes use admissions or a one‑day census, and those choices change the magnitude of rates markedly [1] [5]. The 2004–2014 decline is reported in person‑years (useful for risk per unit time), while FY2020–2023 figures in the recent update are per 100,000 admissions (useful for per‑intake risk), so simple comparisons across administrations without harmonizing the denominator are misleading [1] [2] [5].
3. Small numbers, spikes, and pandemic distortion
Deaths in detention are relatively rare events in any given year, so rates can swing widely when denominators are small or when extraordinary events occur; FY2020’s spike to 10.833 per 100,000 admissions reflects the COVID‑19 pandemic’s outsized mortality effect rather than a steady policy‑era trend [2]. Statistical guidance in criminal justice reporting warns explicitly to “interpret with caution” rates based on small counts or single‑day custody snapshots, because small absolute changes produce large relative shifts [5].
4. Reporting gaps and practices that bias comparisons
Independent investigations and advocacy groups document practices and data gaps that complicate cross‑administration comparisons: watchdog reports note that ICE sometimes releases people immediately prior to their deaths—cases that may never be counted in ICE’s in‑custody death totals—and that investigatory reports can omit facts that would implicate facilities or agency oversight failures [3] [4]. Advocacy organizations also emphasize systemic care failures and rising detention populations in recent years, which affect absolute counts and risk mixes even when rates per denominator decline [6] [3].
5. What the data reliably show — and what they don’t
Reliable cross‑period conclusions are limited: measured mortality per person‑year fell substantially from the mid‑2000s through the mid‑2010s [1], the COVID pandemic produced a measurable spike in FY2020 when calculated per admissions [2], and post‑2020 per‑admission rates have been lower though not immune to year‑to‑year variation [2]. What these numbers do not, by themselves, prove are the precise causal effects of any single president’s policies, because changes in detention population size, average length of stay, pandemic exposure, reporting choices, and possible exclusions of near‑death releases all shape the published rates [1] [2] [3] [5]. Analysts must harmonize denominators, adjust for person‑time, and account for potentially excluded cases before asserting that one administration was categorically “safer” or “deadlier” than another.