Do Health Through Housing results leave out residents' drug use and overdoses after move-in?

1. Pass Title matches the evidence

   The title asks whether results leave out residents' drug use and overdoses after move-in. The evidence (ev-missing, ev-health, ev-elsewhere) shows the program's reports publish housing, ER, hospital and jail results, and no drug use or overdose figure after move-in. The only overdose counts are countywide or from a 2023 newspaper, and the article says so. No number or scope in the title goes beyond the evidence.

2. Pass No obvious contradictions

   Finding True, the answer, and the conclusion ('The claim is true as rated') agree. Key point 4 (225 countywide overdose deaths in supportive housing) is consistent with the conclusion, which says the countywide count is not reported for this program. Deaths are described consistently as a one-line count (10 in 2025, 15 in 2024). The 'all eight operators' reading in obj-operators is in tension with the ev-elsewhere limitation, but the wording 'latest public reports' makes this a judgment call, not an obvious contradiction.

3. Pass Plausible justification

   The conclusion ties the True rating to the absence of any published after-move-in drug use or overdose figure, while crediting the program's ER and hospital results and noting the lack of a control group. It also gives the comparison points (Oxford House abstinence figures, countywide overdose deaths) that show such measures are collectible. A reader would accept this as the reason for the rating.

4. Pass Sources are relevant

   The sources are the program's annual reports, dashboard and plan, the county's MIDD and overdose reviews, the council staff report and motion, the RFP abstract, two operator annual reports, a Seattle Times story, and FactHub analyses. Each is on topic for the claim it supports.

5. Pass Common questions answered

   There are nine objections. Each response addresses its question: countywide overdose deaths, the 2023 newspaper count and lapsed motion, the health results and their limits, whether ER figures include overdoses, sobriety as a goal, the 93% retention figure, operator reports, privacy, and the evaluation's scope.

6. Pass Headline and verdict agree

   Finding True gives short answer 'Yes'. The sentence a reader takes away is: 'Do Health Through Housing results leave out residents' drug use and overdoses after move-in? Yes.' That means the program's results do leave these out, which matches the claim (not reporting drug use or overdoses after move-in), the finding and the answer. The title's positive wording 'leave out' is not negated, so polarity is correct.

Coverage: I audited the article payload only. I did not open sources, snapshots or the web. Each source takeaway in the sources block is marked 'not yet checked', so I did not treat any takeaway as established. I judged from the article's own text: the key points (marked established), the conclusion, the evidence summaries and limitations, the objections, and the cited quotes shown in the payload. I did not check the accuracy of the quoted figures against the sources. Noted but not rated: obj-operators and the conclusion say the latest reports of all eight operators were read, while the ev-elsewhere limitation says two operators publish no annual report and the Salvation Army's 2024 and 2025 reports were not online. The method also says Seattle's Human Services Department was not checked. obj-eval describes an outside evaluator's scope that no cited takeaway covers. These are judgment calls that the deep reviewers should weigh. The title's compliance with check 6 was checked against the claim, finding and answer only.

1. Pass Rates the real claim

   The claim says the program "reports residents' housing, hospital and jail results but not their drug use or overdoses after they move in." In circulation, critics such as We Heart Seattle (named in context) put a looser version: high housing retention doesn't show whether people get healthier, get sober or stay alive. That version can slide into "the county doesn't track" overdoses. The article rates the published-results version, which matches the claim's wording ("reports"). It explicitly does not reach the stronger "doesn't track" reading (key point 1: "Whether it tracks them internally is unknown; we didn't ask"; editorial note). Deaths, which critics also raise, are handled openly: "Deaths get one line." The conclusion stays within the program, its public reports, and the period through November 2026. It does not claim the program fails or succeeds. No mismatch.

2. Pass Rating follows from the evidence

   I formed my own rating from the evidence before reading the finding. ev-missing reports a full reading and keyword search of the dashboard, the data appendix and all four annual reports. It found 11 resident measures, none on drug or alcohol use after move-in, sobriety or overdoses. ev-health confirms the housing, hospital and jail results the program does publish. The near-misses are each handled: the at-move-in addiction share, the 60% behavioral-health access figure, the all-cause ER drops and a bare death count. None is a post-move-in drug use or overdose figure. On this evidence I would rate True. The article's finding (True), its answer ("no figures on residents' drug use or overdoses after move-in") and its conclusion agree with each other and with the evidence.

3. Pass Confidence matches the evidence

   High confidence fits a documentary claim about what one program's public reports contain, checked by reading those reports in full. The article states the limitations that matter: - Not published doesn't mean not tracked (limitation:ev-missing, key point 1). - The ER and hospital figures come from a before-and-after design with no comparison group, and cover only residents who consented to share data (limitation:ev-health). - Not every filter of the interactive views was read, Seattle's Human Services Department was not checked, and the full call for evaluators is behind a supplier login (limitation:ev-elsewhere). - The evaluation is due December 31, 2026 and may add measures (limitation:ev-plan, editorial note). The title is a neutral question. The summary ("None gives drug use or overdose figures after move-in") matches the evidence without overstating it. I would also give High.

4. Pass Strongest objections answered

   The strongest objections to the article's reading are all answered with evidence: - **The program does report health results** (obj-health). The ER and hospital drops are given, with the no-control-group caveat, and the 60% figure is shown to mix mental health and addiction care and to measure access only. - **The ER drops already include overdoses** (obj-er-overdose). These are all-cause counts with nothing broken out by cause, and overdoses that never reach a hospital are missed. - **Sobriety isn't the program's goal** (obj-goal). The article agrees for sobriety but notes overdoses measure safety under any approach, and that the same department measures abstinence for its Oxford House vouchers. - **Privacy explains the gap** (obj-privacy). Totals are already published from Medicaid data, and no county statement cites privacy. - **The program never promised these measures** (ev-plan, key point 5). The article concedes this. - **Others publish the figures** (obj-operators, obj-news, obj-od-count). - **The coming evaluation will fill the gap** (obj-eval). Answered from the announced scope. One possible remaining objection is that the county might release the figures on request. It is acknowledged as a limitation ("we didn't ask"), which is appropriate because the claim concerns what is published. No material gap.

5. Pass Evidence is independent

   ev-missing and ev-health both draw on the program's own 2025 annual report and dashboard. They are not offered as independent confirmations of one fact: one documents what is absent, the other what is present, and for a claim about what a program publishes the program's own reports are necessarily the primary record. The other two items add genuinely separate material: - ev-plan uses the implementation plan and a council staff report. - ev-elsewhere uses Public Health's overdose fatality reviews, MIDD reports, the regional authority, the Seattle Office of Housing, eight operators, the auditor, council motions and a newspaper. These bear on whether the figures appear anywhere else. Method step 2's claim of "more than one source" is fair.

6. Pass Lead questions hold up

   The only lead question is obj-od-count: "Does King County count overdose deaths in supportive housing at all?" Short answer: "Yes, countywide: 225 in 2024." It passes all five tests. 1. **Surprising.** After a verdict saying no overdose figures are published, a reader would likely guess "no." 2. **Not the headline.** The title and answer don't mention the countywide count. 3. **Settled as worded.** Counting overdose deaths in supportive housing is a plain factual question that the Overdose Fatality Review settles with a number. The response limits scope clearly (all supportive housing, not this program). 4. **Neutral.** It is not one side's talking point. It informs readers on both sides. 5. **Stands alone.** The first sentence ("Yes: 225 people died of overdoses in 2024 in King County's homes with on-site staff for formerly homeless people (supportive housing), the health department reports.") uses everyday words, puts the technical term in parentheses and points back to nothing.

7. Pass Rating states its terms

   The finding is True, so part (a) does not apply. One reading question is in play: "reports" meaning publishes versus tracks internally. Key point 1 states the article's reading ("publish no figure") and the result under the other reading ("Whether it tracks them internally is unknown; we didn't ask"), where a reader sees it next to the rating. The other boundary, at move-in versus after move-in, is built into the claim and the answer, and key point 3 gives the at-move-in figure (half had an addiction). Nothing further is needed.

Coverage: I examined the full article payload: claim, finding, answer, confidence, summary, key points, context, conclusion, all four evidence items with their methods and limitations, all nine objections, the method steps and the source list. I took each evidence item as stated and did not read source snapshots. Every takeaway in the sources block is marked "not yet checked"; none is marked established. My judgments therefore rest on the takeaways as the article states them and should be treated as conditional on the source audit. The chart, the dataset analyses (the FactHub records behind tk-ds-measures, tk-ds-none and tk-ds-places) and the related articles were not examined beyond their descriptions.