The City of Madison, Wisconsin, exemplifies how mobile crisis response analytics can inform operational planning, communicate value, and facilitate regional expansion. The city’s CARES (Community Alternative Response Emergency Services) program has become a model for communities nationwide, thanks in part to its commitment to structured, ongoing analysis.
Madison CARES data analyst Dan Petty illustrates how a thoughtful approach to data can lead to better outcomes, more efficient deployment, and more substantial cross-agency alignment.
A data-driven foundation
Madison CARES pairs a Madison Fire Department community paramedic with a crisis worker from Journey Mental Health. Together, they respond to behavioral health–related 911 calls, welfare checks, and non-emergent community needs that do not require law enforcement. Last year alone, CARES responded to roughly 3,500 calls.
From the beginning, CARES made data a core operational asset. Petty joined the program after serving as a data analyst for the Tempe Fire Department in Arizona, where he supported analytics for EMS and community paramedicine. That experience shaped his approach in Madison, where he now helps tell the story of CARES through data, including the types of calls the team responds to, areas where demand is growing, the frequency of independent CARES responses, and how outcomes compare to those of EMS or police responses.
Petty describes his role as helping people understand how the program works and how it improves community care. “My job is to tell the story,” he says. “How we do what we do, where we do it, and what results we’re seeing.”
Integrating data into existing EMS systems
Madison CARES documentation is captured within the same Fire/EMS electronic patient care reporting (ePCR) environment used across the department. This shared infrastructure gives CARES immediate access to consistent clinical data, dispatch timestamps, response times, district information, and other key elements typically captured by EMS units.:
- Standardized EMS documentation fields
- Dispatch timestamps and response times
- Location-based data at the district level
- Consistent reporting formats
Because the Fire Department bills for ambulance services, it is considered a HIPAA-covered entity, which allows CARES to operate under the same privacy structure. Journey Mental Health maintains its own client data, which cannot be merged due to privacy rules. For reporting, Petty relies primarily on the Fire/EMS dataset to maintain a single source of truth.
Although the EMS data schema was not explicitly designed for crisis response, CARES adapted it to reflect behavioral health workflows better. For example, phone-only engagements, common in crisis response, needed to be represented. The team modified documentation expectations rather than building an entirely new system, proving that a Fire-based mobile crisis unit can operate successfully within existing EMS infrastructure.
Using mobile crisis response analytics to understand demand
One of the most valuable operational insights came from asking dispatchers to flag when a CARES unit would have been assigned to a call but was unavailable at the time of assignment. This small workflow change allowed Petty to answer a critical question:
When and where is demand not being met?
The findings were clear. During the week, unmet demand clusters around 9:00 a.m., before CARES’ staggered shifts fully overlap. On weekends, missed calls are more frequent and more evenly distributed throughout the day. This pattern helps leadership consider how to adjust coverage without immediately moving to a 24-hour model, which the data does not currently support.
In other words, rather than relying on anecdotal comments like “we’re always missing calls overnight,” Madison uses mobile crisis response analytics to pinpoint actual needs and make targeted staffing decisions.
Evaluating model changes with data rather than anecdotes
Mobile crisis programs often evolve rapidly, and program leaders must assess how these changes impact service delivery. In Madison, CARES consolidated multiple units into one downtown location. Almost immediately, concerns surfaced: Would response times worsen? Would west-side calls be underserved?
Using heat maps and district-level comparisons, Petty found no evidence that centralization hurt response. In fact, the proportion of west-side calls actually increased. These findings helped settle concerns before they became accepted narratives.
Similarly, Petty reviewed both average and 90th-percentile response times before and after the change. The trend remained remarkably stable. Despite relocating to a single location, CARES continued to respond reliably across the city.
This illustrates one of the most powerful functions of mobile crisis response analytics: distinguishing perception from reality.
Understanding how CARES resolves calls
Madison’s disposition profile looks very different from EMS. Among patients CARES engages with in person, only about a quarter are transported. Roughly 20% are resolved by phone alone. Police transport is rare, occurring in around 1–2% of cases, and often involves protective custody rather than law enforcement action.
Compared with EMS units, which transport the majority of patients to the hospital, CARES demonstrates greater diversity in dispositions. This confirms that mobile crisis response programs offer a meaningful alternative to automatic ED transport, helping reduce emergency department utilization and connecting people to more appropriate resources.
By consistently tracking these outcomes, CARES can demonstrate to community partners that it provides high-touch crisis care while alleviating pressure on hospitals and law enforcement.
Working independently and knowing when to bring in partners
Another consistent finding is that CARES handles most of its work alone. Across multiple years of data, around two-thirds of CARES calls are managed without any other emergency units on scene at any point. Even more striking, once CARES arrives, they complete the call independently about 93% of the time.
This reinforces a central value proposition: a mobile crisis team can safely manage behavioral health incidents without requiring police, fire, or ambulance support on most calls. For communities evaluating whether mobile crisis response can alleviate pressure on traditional systems, this is a compelling data point.
Using mobile crisis response analytics to guide expansion
In early 2024, CARES expanded into Sun Prairie, a neighboring community. Uptake required an initial adaptation period, but participation grew rapidly. Approximately half of the Sun Prairie-related CARES activity occurred in the most recent month measured. As awareness increased, volume followed.
This model provides a blueprint for counties or regions considering collaboration on mobile crisis response. Smaller municipalities can adopt an existing program without having to build a new system from scratch. Staff can be shared. Data remains unified. Training is consistent. Communities benefit from faster implementation and lower startup costs.
The emerging role of AI
Looking ahead, Petty expects the most immediate value of artificial intelligence to center on documentation and data analysis. Several EMS vendors already offer features that generate draft patient narratives based on structured fields such as impressions, vital signs, and medications. These tools allow clinicians to edit rather than write from scratch, improving clarity and reducing administrative burden.
In Tempe, Petty also used machine learning to identify individuals likely to call 911 again within 30 days. That helped proactively assign follow-up for community paramedicine. While similar modeling may not translate perfectly into crisis response, it reflects the broader future: using data to anticipate needs rather than react to them.
A model built on continuous learning
The CARES program has grown quickly, yet data remains its anchor. Analytics guide scheduling, demonstrate equity across districts, help track outcomes, and support measured expansion.
Instead of reacting to pressure for more hours or more units, Madison grounds its decisions in clear evidence. As Petty explains, his goal is to prevent broad claims, “we’re seeing more of this,” or “things are getting worse,” from becoming the narrative without data to support them.
Mobile crisis response analytics enable Madison to communicate its value, refine its model, and reinforce system trust. That is not just a reporting benefit. It is what sustains momentum.
Conclusion
Madison’s CARES program shows how mobile crisis response analytics can transform raw activity into operational knowledge. By operating within the existing Fire/EMS data environment, CARES established a robust foundation that captures the complete story of each encounter. With dedicated analysis, the program has fine-tuned deployment, demonstrated service equity, reduced unnecessary transports, and strengthened partnerships across health and public safety.
Communities planning to launch or expand mobile crisis programs can apply Madison’s lessons immediately: build within existing infrastructure, define a single source of truth, adapt EMS schemas thoughtfully, track unmet demand, and let the data, not assumptions, shape the model.
The result is a more reliable, more accountable, and more sustainable mobile crisis response system, one designed to meet people where they are and deliver the proper care at the right time.
