How a Kansas home health agency taught its clinicians to trust predictive care guidance

Decisions made on last week's data

Douglas County VNA had a solid foundation in MatrixCare and a clinical team that knew its patients well. What it did not have was time.

Care planning ran through a full-team meeting where each discipline arrived with its own estimate of what a patient would need and for how long. The conversations were thorough. They were also built on judgment alone, with no shared view of where an episode was actually heading.

The data that could have informed those conversations arrived too late to use. Reports on utilization, rehospitalization risk, and length of stay took five to seven days to reach the team — long enough that by the time a problem showed up, the window to do anything about it had usually closed.

Predictive insight on top of the record they already keep

Douglas County VNA was not looking to replace MatrixCare, and it was not going to ask clinicians to document anywhere new. It needed something the existing reporting could not produce: a current view of LUPA exposure, inpatient transfer risk, and where each episode was trending.

Mosai Pulse reads the clinical and visit data the agency’s clinicians are already entering in MatrixCare — structured assessments and the narrative in their notes — and returns a prioritized view of the census. MatrixCare stays the system of record. Pulse refreshes against it every three hours, so what a manager sees reflects what a clinician documented that morning.

Pulse is built to augment clinical judgment, not to replace it. It shows the team where to look and what is driving the risk. The clinicians decide what to do about it.

A prioritized census, reviewed every day

Start of care

When the start-of-care note closes, Pulse returns a visit recommendation for the certification period. The team shares it across the full care team through MatrixCare Link, and it becomes the starting point for the case conference rather than a replacement for it. If Pulse recommends 15 visits, the conversation is whether the team agrees and what each discipline contributes to that plan.

It also changed a habit nobody planned for. Clinicians close their charts faster now, because the recommendation does not arrive until they do.

Daily census review

Each morning the team opens a prioritized list instead of hunting through reports. Scheduled and potential LUPAs surface early enough to adjust a visit plan. Patients whose condition has declined are flagged from the documentation itself, in time for someone to ask what changed. Gaps in care longer than seven days are visible before they turn into something worse.

End of episode

Beginning on day 45, Pulse analyzes the episode and returns a recommendation: recertify, review further, or discharge. For a team that had always made that call on clinical instinct, it does what the visit recommendation does at the start of care — it opens the conversation. Clinicians bring their discharge planning to the table, compare it against what the data shows, and decide together.

Alongside the recommendation, Pulse generates an AI summary of the episode drawn from what the team has already documented, with each section citing the source data behind it. It doesn't make the call or write the justification — it surfaces what is already in the record, so the clinician documenting a recertification decision spends less time hunting for it.

What changed

The most measurable change is speed. Information that used to take five to seven days now reaches the team within three hours of being documented, which moves most decisions from review to intervention. In the second quarter of 2026 the agency averaged 12.2 visits per period, with the visit plan for each patient set against a recommendation rather than an estimate.

The second change is harder to put a number on and probably matters more. Predictive guidance did not arrive to a receptive audience.

What changed the team’s mind was not a demonstration. It was recognizing their own documentation coming back to them. The model reads what clinicians write, which means the guidance carries their assessment rather than overriding it — and once that landed, adoption followed. Care planning now runs in smaller, faster case conferences focused on one patient at a time, and the agency reports the discussions are more productive for it.

In the first fully reported period since Pulse went live, the agency averaged a 5.3 percent LUPA rate against a national average of 7, and 12.7 visits per episode against a national average of 15 to 17. Fewer episodes fall below the visit threshold that costs an agency its full period payment, and visit volume is matched to what patients actually need rather than to an estimate made in a meeting.

Recertification decisions have held steady over the same period, and that is the point. The team is not recertifying more often. It is recertifying on evidence rather than instinct, with the day 45 analysis and the episode summary in front of them when they make the call.