SAN FRANCISCO · OFFICIAL RECORDS · JULY 2025

911 calls at supportive housing

Explore monthly incidents for 155 sites tabulated by the Department of Emergency Management and Department of Homelessness and Supportive Housing. January 2020–June 2025.

2025 is a half-year, not a full year

Correction · October 18, 2026

John Burton Housing’s “all incidents” row appears cumulative, despite the source table being labeled monthly. The printed values never decrease across 66 months: 80 in December 2023, 104 in December 2024, and 129 in May 2025. Our earlier version summed these entries as monthly counts, incorrectly producing 1,073 for 2024 and placing this building first.

The disputed row is now excluded from combined-incident rankings, map sizing, and aggregate totals. Its printed values are preserved for inspection and marked as disputed in downloads. Police, EMS, and fire figures remain separate. The 2024 combined total across the remaining 154 source rows is 31,511 recorded incidents. This excludes an unresolved building row and is not a corrected complete-portfolio total.

If the row is cumulative, its increases would be 24 during 2024 and 25 during January–May 2025; these are hypotheses, not verified replacement counts. No inferred monthly figures have been substituted. DEM confirmation is needed. John Burton was the only “all incidents” row with 16 or more numeric entries that never decreased in this check; that screen is not a full source-data audit.

Recorded incidents · filtered sites
Sites with numeric observations
Monthly cells reported

Highest recorded counts

Compare monthly patterns

● Selected site● Comparison

Monthly values

All filtered sites

Click a site to inspect its monthly pattern. Counts include only numeric cells; blank months are excluded. *John Burton’s disputed combined row is excluded from totals and rankings; its raw series remains available in the monthly comparison.

Dorsey’s request: June 17, 2025

San Francisco Board of Supervisors · Item 38, Roll Call for Introductions · recorded excerpt, approximately 3 minutes 28 seconds.

Open or download the video

Video hosted on Bluoz; an internet connection is needed for playback. The dashboard’s embedded data works independently.

Meeting transcript and recording · Official written request summary

Transcript supplied for this excerpt

Automated transcript, supplied by the publisher; paragraph breaks added for readability. Apparent recognition errors and repetitions—including “Mayor Lohrey,” “10 o drug OED,” and the repeated “nine one one” wording—are retained. Figures and characterizations are attributed to Dorsey’s remarks and have not been independently verified here.

Thank you, supervisor Chen. Supervisor Dorsey?

Matt Dorsey, Supervisor (District 6):

Thank you, madam clerk. Colleagues, I am today submitting a letter of inquiry that seeks from our Department of Emergency Management and Department of Homelessness and Supportive Housing, a comprehensive analysis of nine one one calls for service at all site based permanent supportive housing addresses in San Francisco. I am asking that this analysis evaluate data from 01/01/2020 to the present to get a longitudinal perspective on the public safety and medical emergency demands of permanent supportive housing or PSH and how they may be changing over time. I also ask that each PSH site, that's listed include information identifying its nonprofit contractor and property manager.

Now, the topic of public safety issues in and around PSH facilities isn't new. It is a subject we took up at a hearing last year, in fact. And while we have seen improvements at some locations in District 6, we have also seen new problems emerge at some new PSH locations.

This need for better data comes in the context of a letter I sent this morning to the Tenderloin Neighborhood Development Corporation, which, disappointingly, opted to oppose Assembly Bill two fifty five legislation by Assemblymember Matt Haney to fund drug free supportive recovery housing, which this board unanimously endorsed endorsed twice, and which enjoys similarly strong support from Mayor Lohrey and our entire legislative delegation in Sacramento.

Right now, we thankfully have comprehensive data on drug overdose deaths and permanent supportive housing, and those numbers are increasingly alarming. In a year's time, the proportion of drug OD fatalities in PSH has grown from about twenty one percent last year, accounting for about one in every five drug deaths citywide, to thirty percent so far this year, represent near representing nearly one in three drug overdose deaths citywide.

This is a disturbing trend that I believe drug free and recovery oriented PSH options would help reverse. As it happens, TNDC ranks third worst among PSH providers this year with 10 o drug OED fatalities in just four months. We also know that TNDC managed properties, count for about five of the top 100 most responded to addresses for police calls for service this year. This these, five facilities generated, actually, last year, more than sixteen hundred nine one one calls just for police emergencies alone, not counting nine one one calls for medical emergencies or fire. That means that each of these buildings is averaging around nine one one nine one one call per day for just police emergencies.

Now, I believe, anyone nobody should blame TNDC for the loss of life or predictable public safety challenges we see at these drug tolerant sites. Still, to my mind, these problems make all the more troubling this organization's decision to oppose needed progress and to defend the adequacy of a status quo that no one should find acceptable. We need better data to assess public nuisances and public safety challenges, to more fully inform our policy decisions, and to match the high quality data we now have on drug overdose fatalities. So I thank the clerk for facilitating this request, and I look forward to seeing the analysis and the rest I submit.

How HSH obtained the data—and public access

Documented turnaround: 15 calendar days. Supervisor Matt Dorsey’s inquiry was received June 17, 2025. The joint response from Emily Cohen, HSH Deputy Director, and Robert Smuts, DEM Deputy Director, is dated July 2, 2025 and includes the portfolio-wide monthly tables.

Source external to HSH: the letter states that DEM’s data team pulled monthly call data for each address associated with site-based PSH (PDF page 44). These emergency-dispatch counts are a separate source from nonprofit providers’ critical incident reports. The response demonstrates that HSH can obtain building-level data through interdepartmental cooperation; it does not establish routine updates or direct HSH database access.

Bulk work is documented; delivery details are not. The tables cover the PSH portfolio across several years, showing a coordinated extraction. The letter does not specify whether this involved one query, one transfer, electronic delivery to HSH, or when HSH contacted DEM.

The posted public route differs. DEM’s public dispatch-record request form accepts requests at dem.records@sfgov.org, provides incident-location and date fields, and offers delivery by postal mail or pickup. It says records cannot currently be sent by email and directs city/county employees not to use that form. Form reviewed October 18, 2026; posted procedures may change.

What this comparison supports: a portfolio-wide analysis was included in the joint response within 15 days, while the posted public form describes a less convenient delivery route. A comparable public bulk-data request has not been tested here. These documents do not prove that public access to a bulk electronic extract is prohibited, that a public request would take longer, or that the form guarantees an exhaustive address-history search.

Connection to oversight: these counts could help identify buildings and periods for comparison with HSH critical incident reports. Dispatch calls and reportable critical incidents have different definitions; a difference in counts alone does not establish underreporting.

Sources: joint HSH–DEM letter, PDF pages 41–44, and DEM’s linked public request form.

Grand Jury connection and timeline

The 2025–2026 Civil Grand Jury’s At Scale, At Risk: Upgrading Data and Oversight to Improve Homelessness Services, released June 23, 2026, discusses building-level police calls alongside its analysis of resident safety and critical incident reporting.

“Just one PSH facility alone generated 654 police calls for service in a single year — averaging nearly two calls per day. A sample of 15 PSH sites showed 5,872 police calls for service during that same time frame.”

Source and attribution: report printed page 22. The Jury is quoting factual recitations in legislation considered by the Board’s Public Safety and Neighborhood Services Committee on April 23, 2026. These are figures quoted by the Jury from that legislation, not totals independently calculated for this dashboard. The passage does not establish which period or selection of sites matches the July 2025 tables.

On printed page 23, the Jury calls for a process linking overdose deaths to PSH addresses, including residents who die elsewhere, and identifies coordination among HSH, DEM, DPH, and the Office of the Chief Medical Examiner as a way to do this. Its subsequent critical incident reporting discussion addresses weaknesses in aggregation, analysis, and systemwide safety monitoring.

EventDate
Dorsey requests portfolio-wide PSH call dataJune 17, 2025
2025–2026 Civil Grand Jury term beginsJuly 1, 2025
Joint Cohen–Smuts response with monthly tablesJuly 2, 2025
Response appears in Board packetJuly 15, 2025
Legislation cited by the Jury for police-call figuresApril 23, 2026
Jury releases At Scale, At RiskJune 23, 2026

What was almost concurrent: the joint response was dated one day after the Jury’s annual term began. The Jury’s published findings followed nearly a year later. The exact start date of this particular investigation is not established here.

What the report establishes: by publication, the Jury knew of building-level police-call statistics and identified DEM as a potential partner in improved safety-data coordination. A review of the report’s text, methodology, and bibliography found no explicit identification of the July 2 Cohen–Smuts letter, Dorsey’s June 17 inquiry, or the complete monthly tables. That absence does not prove the Jury was unaware of them or did not review them. Whether it saw this particular extraction, especially when it first convened, remains unconfirmed.

References: Grand Jury report (official attachment), printed pages 22–27, methodology pages 42–43 and Works Cited; Board file 260730 and report attachments; Superior Court description of July–June jury terms. The uploaded report supplied for this project was used for the text review.

Source, scope, and limitations

Open the original Board packet. Response dated July 2, 2025, from Emily Cohen (HSH) and Robert Smuts (DEM), responding to Supervisor Matt Dorsey’s June 17 inquiry. Context: PDF pages 41–44. All incidents: 45–46; police: 47–48; EMS: 49–50; fire: 51–52. Page numbers refer to the PDF, not printed agenda pagination.

  • The letter describes 156 sites. The supplied tables contain 155 rows in each category; this dashboard reproduces those 155 rows.
  • All incidents are unduplicated across responding services. Police, EMS, and fire overlap and must not be added together. The source notes merged calls may retain multiple incident numbers within service categories.
  • Blank cells remain blank (shown as —), not zero. They do not establish whether a site was open, closed, or had no calls. Source shading is not reproduced.
  • *Unit counts are current PSH units, or maximum units for closed sites, rather than historical occupied units. Rates are exploratory and are not adjusted for occupancy, resident needs, or time open. Mixed buildings may include incidents involving non-PSH units.
  • These are address-level incidents, not unique residents, arrests, crimes, or confirmed medical diagnoses. This dataset alone does not provide a comparable non-PSH control group.
  • 2025 contains January–June only. Use matching month ranges when comparing years. Missing months can also make site comparisons incomplete.
  • Site names, population labels, and unit counts follow the tables. No inferred addresses or operator affiliations have been added.