a/justice· 31 October 2026 · 7 min read

Why do we send armed officers to people in crisis, and who gains when we do?

Cover of Protecting Communities in Crisis (Omxus Press)

In 2012 Deborah Danner, a woman in the Bronx who lived with schizophrenia, sat down and wrote six pages about it. She wrote about Eleanor Bumpurs, a mentally ill woman killed by police in the Bronx in 1984. "They (the police) used deadly force to subdue her because they were not trained sufficiently in how to engage the mentally ill in crisis," she wrote. "This was not an isolated incident." Four years later, a neighbour called about her.

Who got sent

On 18 October 2016, a police sergeant went into Danner's apartment. She was 66. She'd been holding scissors, then a baseball bat. He shot her. A court later acquitted him of all charges. The police commissioner said the next morning: "What is clear in this one instance, we failed. That's not how it's supposed to go. It's not how we train" (ABC7 New York, 2018; Communities United for Police Reform, 2016; InsideHook, 2016).

Whatever you think of that one night, look at the step before it. A neighbour was worried about a sick woman. The only people that call could summon carried guns. That part isn't an accident, and it isn't one sergeant's choice. Someone in a car park having a breakdown, someone asleep in a doorway, neighbours arguing over a fence, someone alone on a bridge at 2 am: in most places all of them get the same answer.

What most calls are

A multicity 911 analysis by the Vera Institute found that most police time goes on work other than enforcement: noise, welfare checks, traffic, disputes and mental health crises (Neusteter et al., 2020). The training doesn't match. The average US academy runs about 840 hours. Roughly 168 go to firearms, self-defence and use of force. Mental health crisis response gets about 10 (Reaves, 2016). Former sheriff Sue Rahr calls this the "warrior" model, in which any encounter might turn into a fight (Rahr & Rice, 2015).

We train people for combat, send them to welfare checks, then express surprise when the outcome is combat.

The Washington Post recorded every person shot dead by on-duty US police from 2015. We counted the decade to 2024: 10,430 people. About one in five, 19.7%, showed signs of mental illness, and half of those had no gun. The yearly count didn't fall. It rose from 995 to 1,175. Those were years of reform, and body-worn cameras spread everywhere, but a large randomised trial in Washington DC found cameras made no measurable difference to officers' use of force (Yokum et al., 2019). The camera records the encounter. It doesn't change who gets sent.

Two people in a van

There was another answer, and it was older than most of these officers' careers. In 1989 the White Bird Clinic in Eugene, Oregon, started sending two-person teams, a medic and a crisis worker, to these calls. No weapons. No power of arrest. It called the service CAHOOTS.

The clinic's own figures, which no outside body audited, had CAHOOTS taking around 24,000 calls a year and asking police for backup fewer than 150 times, for about $2 million a year against about $90 million for the city's police (White Bird Clinic, 2020). The clinic also says nobody was killed during a CAHOOTS response in all those decades. A death on a call would leave a police report and a news story, and nobody has produced one. Eugene's own police analysts added a caution: most CAHOOTS calls were welfare checks, transport and public assistance, so it mostly added help rather than taking over police work on dangerous calls (Eugene Police Department Crime Analysis Unit, 2020).

The best outside evidence comes from Denver, which copied the idea in 2020 with a van called STAR. Two Stanford economists compared the areas it covered with similar areas it didn't. Reports of the minor crimes it was sent to, things like trespassing and public disorder, fell by 34%. Serious crime didn't rise (Dee & Pyne, 2022). The calls were still there. They were just answered as what they were. The "crime" had been distress all along, written down in the wrong column.

CAHOOTS isn't radical. It's boring. It's two people in a van doing the obvious thing. The radical act is continuing to pretend that the obvious thing is dangerous while the dangerous thing is normal.

Then the money ran out

You'd expect a story like that to end with a statue. On 7 April 2025, CAHOOTS stopped serving Eugene. The city faced a budget shortfall. Its contract covered about 40% of what the service cost, the clinic's interim director said, and the federal grants paying for much of the rest had become harder to hold. White Bird said it no longer had the money, and laid off many of its crisis workers (Hansen-White, 2025). A state-funded county service stepped in, billing Medicaid, without the outreach CAHOOTS did, and people have been working to bring CAHOOTS back (Wilk, 2025).

Nobody shut it down because it failed. It's just that nothing in the system was built to notice it worked.

That's the design. Our systems provide wanted attention for unwanted results. Police departments get counted on calls answered, arrests made, cases opened. An officer who talks someone down from a bridge and connects them to a counsellor has done extraordinary work, and the department has no way to count it. An officer who arrests the same person produces paperwork, and maybe repeat business. So an institution whose failures generate funding goes on, and a van whose success generates nothing has to beg. Nobody designed it to hurt people. Nobody has had a reason to redesign it.

The same idea in Australia

Aboriginal communities have run their own version for decades. Night patrols, often staffed by Elders, walk the streets, calm disputes and get people home safely. They speak language and carry no weapons. A review for the Commonwealth found patrols across the Northern Territory, Western Australia and Queensland doing work that would otherwise become a police call (Blagg & Valuri, 2003).

Canberra has tested another answer for when harm has been done. In the RISE experiments, young people who'd committed violent offences went at random either to court or to a face-to-face conference with the person they'd hurt. The conference group went on to commit 38 fewer offences per 100 people a year (Sherman et al., 2000).

Nobody loses their job

The objection people raise first is the workforce. So keep it. Quick response, fitness and staying calm under pressure are what paramedic teams, fire services and crisis teams need. Change the job, keep the people, and count what the public actually wants: lives saved, crises resolved, people connected to help. A nine-agency study of 911 calls warns that moving work away from police takes real money and planning (Lum et al., 2022). True. Eugene just showed what happens when it doesn't get it.

Deborah Danner saw this coming and wrote it down. We shouldn't need another six pages.

What you can do

  1. Find out who comes if you call about someone in a mental health crisis where you live. Ask before you need to know. In danger, call 000 regardless. For crisis support in Australia, Lifeline is on 13 11 14.
  2. Ask your state MP or council what they count. If nobody tracks crises resolved without force, ask them to start.
  3. If your area has a crisis team or a night patrol, find out who funds it and when that runs out. Good things die quietly, at budget time.

Sources

ABC7 New York. (2018, February 15). NYPD sergeant who shot mentally ill woman found not guilty. https://abc7ny.com/post/sergeant-who-shot-mentally-ill-woman-found-not-guilty/3087715/
Blagg, H., & Valuri, G. (2003). An overview of night patrol services in Australia. Commonwealth Attorney-General's Department.
Communities United for Police Reform. (2016, October 19). NYPD admits 'we failed' in police shooting of mentally ill Bronx woman. https://www.changethenypd.org/media/nypd-admits-we-failed-police-shooting-mentally-ill-bronx-woman
Dee, T. S., & Pyne, J. (2022). A community response approach to mental health and substance abuse crises reduced crime. Science Advances, 8(23), eabm2106. https://doi.org/10.1126/sciadv.abm2106
Eugene Police Department Crime Analysis Unit. (2020). CAHOOTS program analysis. City of Eugene. https://www.eugene-or.gov/DocumentCenter/View/56717/CAHOOTS-Program-Analysis
Hansen-White, R. (2025, April 8). White Bird, city end CAHOOTS in Eugene over contract, funding issues. KLCC. https://www.klcc.org/health-medicine/2025-04-08/white-bird-city-end-cahoots-in-eugene-over-contract-funding-issues
InsideHook. (2016, November 11). Respect must be paid: The story of Deborah Danner. https://www.insidehook.com/culture/tragic-death-schizophrenic-deborah-danner
Lum, C., Koper, C. S., & Wu, X. (2022). Can we really defund the police? A nine-agency study of police response to calls for service. Police Quarterly, 25(3), 255–280. https://doi.org/10.1177/10986111211035002
Neusteter, S. R., O'Toole, M., Khogali, M., Rad, A., Wunschel, F., Scaffidi, S., Sinkewicz, M., Mapolski, M., DeGrandi, P., & le Roux, D. (2020). Understanding police enforcement: A multicity 911 analysis. Vera Institute of Justice.
Rahr, S., & Rice, S. K. (2015). From warriors to guardians. Harvard Kennedy School and National Institute of Justice.
Reaves, B. A. (2016). State and local law enforcement training academies, 2013 (NCJ 249784). U.S. Department of Justice, Bureau of Justice Statistics.
Sherman, L. W., Strang, H., & Woods, D. J. (2000). Recidivism patterns in the Canberra Reintegrative Shaming Experiments (RISE). Australian National University.
The Washington Post. (2025). Fatal Force: Police shootings database, version 2. https://github.com/washingtonpost/data-police-shootings
White Bird Clinic. (2020). CAHOOTS media guide.
Wilk, N. (2025, June 3). Eugene after CAHOOTS: The end of the crisis response program and the efforts to bring it back. KLCC. https://www.klcc.org/health-medicine/2025-06-03/eugene-after-cahoots-the-end-of-the-crisis-response-program-and-the-efforts-to-bring-it-back
Yokum, D., Ravishankar, A., & Coppock, A. (2019). A randomized control trial evaluating the effects of police body-worn cameras. PNAS, 116(21), 10329–10332. https://doi.org/10.1073/pnas.1814773116

Keep reading