DURHAM, N.C. — When the murder of George Floyd in 2020 sparked widespread demand for police reforms, some cities moved to announce major overhauls of law enforcement. One of the most promising—and popular—proposals was to develop mental-health response units, staffed by social workers or crisis counselors, who could respond to mental-health incidents in lieu of armed police.

The idea grew out of frustration with police over a slew of horrifying and high-profile incidents when law-enforcement officers, dispatched to deal with a person in crisis, ended up killing the very person they were called to help. In a widely watched 2020 case, officers in Rochester, New York, killed Daniel Prude, a Black man whose brother had called for help because he was acting erratically. A Washington Post analysis found that one-quarter of people fatally shot by U.S. police in 2015 were mentally ill or in emotional crisis. Many police officers, meanwhile, have expressed frustration about spending time handling mental-health calls for which they feel unequipped, rather than fighting crime.

“We’re not about defunding—we’re about funding and providing the additional services you need beyond someone with a gun strapped to their shoulder or to their hip,” President Joe Biden said while discussing crime in New York on February 3.

San Francisco, Los Angeles, and New York are among the cities that rolled out new mental-health or crisis response teams in the past 16 months. Many cities have looked to CAHOOTS (Crisis Assistance Helping Out on the Streets), a program in Eugene, Oregon, that sends unarmed responders to deal with mental-health calls, drug issues, minor injuries, and the like, as a model for diverting calls, and funds, from the police.

The idea is relatively simple; executing it is more complicated. In Durham, a liberal city with a strong activist community and a very progressive city council, a proposal to establish such a system seemed likely to sail through. But the city’s leadership wanted to design a unit that would actually meet the city’s needs. As they discovered, figuring out what those were was not so easy.

Durham officials wanted to understand just how many 911 calls the city was getting that could be diverted to someone other than police, so they asked the research organization RTI International to analyze 1 million calls for service over three years to assess how much work a mental-health unit would have, part of a review involving seven cities. The results stunned them. RTI found that operators had categorized only 1 percent of the calls as mental-health-related at the time of dispatch.

“When we saw those numbers, and there was such a low number of them that were identified as mental-health calls—I was surprised,” Steve Schewel told me. Schewel, who retired in December after two terms as mayor, had often been on ride-alongs with police officers, and had seen that many of the calls—far more than 1 percent—were obviously mental-health-related. Durham police officers in an RTI focus group, convened as part of the research, reported that the overwhelming majority of the crisis calls they answered involved people struggling with their mental health.

The researchers concluded that the 1 percent figure was a substantial undercount of the mental-health calls reaching Durham 911. In too many cases, calls related to people in crisis were going unrecognized until police were already on the scene.

Other cities that have studied their own data have identified similar undercounts. This gap in the data is a major challenge to the whole enterprise of diverting calls from armed officers. For mental-health response units to respond to the right incidents, emergency dispatchers need to be able to decide what counts as a mental-health call, and they need to be able to do it from the incomplete information provided during a 911 call made in the heat of a crisis. The best alternative responders or co-responders in the world can help only if emergency dispatchers know when to deploy them. If first responders don’t recognize that a call is mental-health-related until cops arrive, it’s too late.

Over the past few years, leaders around the country have reported increases in mental-health calls and pointed to struggles in identifying them. RTI reanalyzed calls for service in Burlington, North Carolina, and found that just 42 percent of mental-health calls had been correctly coded. A study of the Gresham, Oregon, police department, conducted by Portland State University researchers, found that mental-health calls might occupy anywhere from 7 to 12 percent of calls, though they can take up almost a quarter of officers’ time. In a survey, Gresham officers overestimated the portion of mental-health-related calls, at nearly 70 percent, a sign of the stress they cause for officers. The problem is we don’t really know what we don’t know: Only limited analyses of call data have been conducted, and they may not be representative of the national picture or across agencies of different sizes.

The apparent glut of misidentified calls is an unexpected side effect of the spread of 911. The standardized number for emergencies first launched in 1968, though Congress didn’t mandate the current system nationwide until 1999. It is a miracle of modern technology: Almost any American, anywhere in the country, can pick up any phone and with just three numbers summon prompt help to wherever they are.

But adding that supply induced new demand. With an easy way to reach emergency responders, Americans began using 911 for lots of things that weren’t arrestable offenses or acute medical emergencies. Maybe someone was standing on their corner in a way they didn’t like, or maybe someone simply didn’t look right, or maybe the caller was just confused and needed help. How often police are responding to these noncrime calls is a mystery, because 911 is not really a system but about 6,000 individual call-answering points, each with its own approach to collecting and sharing data.

Many 911 operators are underpaid and overworked. Depending on the community and the time of day, call volume can be so high that operators have little time to get information from a member of the public who is calling—assuming that person has good information to begin with. As a general rule, criminologists estimate that 5 percent of a city’s blocks account for half of its 911 calls, and local police come to know certain people they encounter frequently on the beat or on calls. Local knowledge can be essential to guessing what a call is about and what kind of response is best, but many call centers have high turnover, so the people taking the call don’t have time to build up that knowledge. Durham’s 911 center has been so overwhelmed in the past few months that calls have been redirected to other local centers and response times have lagged.

When someone calls 911, an operator typically answers, asking whether the caller needs police, fire, or medical help, and enters that information into a standardized form. Then a dispatcher assigns a first responder, turning to firefighters and EMTs to handle fires and obvious medical emergencies and routing everything else to police. A 2021 Pew Charitable Trusts survey of 911 centers found that staff are seldom trained on mental-health issues and have few options for mental-health responses. Many centers use computer-aided dispatch systems that require each call to be assigned a primary code, but many calls that are related to mental-health might not be identified as such, instead getting tagged with other, non-mutually-exclusive labels such as “public disturbance” or “welfare check.” That means mental-health calls are hidden within the data.

Some communities have experimented with staffing call centers with additional responders, including nurses who can advise on nonemergency medical problems or social workers to talk with people in crisis or those around them. In early 2021, Austin, Texas, changed its 911 protocol so that operators ask, “Are you calling for police, fire, EMS, or mental-health services?” If the answer is mental health, the call is transferred directly to a clinician. In fiscal year 2021, Austin transferred almost 4,500 calls to clinicians, 3,600 of which didn’t wind up requiring any police response, Lieutenant Ken Murphy, who runs Austin’s 911 operations, told me. Murphy predicts that this fiscal year, the new protocol will divert about one in 10 calls made to Austin 911, avoiding needless police encounters, providing better-targeted services, and saving taxpayers as much as $9 million.

These experiments are promising, but they remain exceptions: Murphy said he believes that Austin is the only 911 center in the country that proactively offers callers mental-health services, and just a few centers are staffed with clinicians. Even CAHOOTS, with years of experience and deep roots in the community, entirely diverts only 5 to 8 percent of calls to the Eugene dispatch center, according to a 2020 analysis by the Eugene Police Department.

In jurisdictions without such programs, the mantra remains “When in doubt, send them out.” Regardless of the input, if it’s not a fire or medical emergency, the likely outcome is the same: a cop in a squad car heading to the scene. Sending police is an easy answer, but it carries big risks.

“It creates an untenable situation for police and the communities they serve,” Rebecca Neusteter, a University of Chicago researcher who studies 911, told me. “Police are being sent out on far too many calls, most of which don’t have to do with public-safety emergencies. It can create real tensions. It creates a lot of discord and issues around morale for police officers.”

Police officers generally have few good options for dealing with a mental-health crisis. They might ask someone to move along, which may or may not work. They can just let someone go, which may be the best option but means that the person in crisis won’t get much help, and the citizens who called 911 may be frustrated by a lack of response. In a very few cases, officers may be able to involuntarily commit someone, but the bar is high for that. Or they can find some grounds to arrest a person, which will get them off the street temporarily but in practice just adds to or creates a rap sheet without solving the underlying problem. (Mental health can sometimes be an underlying cause in crimes, rendering some of these distinctions murky.)

Once police officers do intervene, many of those suffering from a psychiatric crisis are funneled into other institutions—courts and jails—that do little for their mental health while also giving them the stigma, and real consequences, of a criminal record. “You couldn’t go out of your way to invent a system that was less useful,” says Mike Gleason, who as Eugene’s city manager helped launch CAHOOTS in 1989.

For policy makers in liberal cities, the strongest allure of alternative-responder programs may be the opportunity to defuse the “Defund the police” bomb. From the police perspective, they eliminate some of the most intractable and time-consuming work for officers, and free them up to fight violent crime at a time when many departments are struggling to hire and retain officers. Officers in the RTI focus group said they often didn’t have good information on what to expect until they arrived on scene, and felt unprepared when dealing with mental-health crises. “One can’t expect every person to be the best mental-health-crisis responder, and the person kicking down the door, and several other police responsibilities, all at the same time,” one officer said.

When it works, diversion helps reduce interactions between citizens and armed police officers and reduces the reach of the police department, potentially setting the stage for budget reductions as well. In June 2021, the city of Durham decided to create a Community Safety Department. “The best way out of the political tension is not to pretend that this is a decision, that you have to decide between one option or the other—that we need both,” Ryan Smith, the head of the new department, told me. “The question is, who is the right responder for which type of situation?”

The size and direction of the department will depend on choices by the city council, including a new mayor, and candidates for mayor and council sparred over how best to deal with rising crime rates during last year’s election. But by design, it’s starting slowly with a staff of just 15 and a set of pilot programs, including a small team of alternative responders and a mental-health responder embedded at the 911 center. The goal, Smith told me, is to figure out what works best for Durham before making big decisions—a rare instance, perhaps, of a call on public safety made with good data.

By Percy