The Thirty-Day Evaluation of Gia Health Analytics

Thirty days, your own data, hosted by Scienza, Scienza's fee waived. This is what happens, step by step, and what you have at the end to decide with.

What happens during the thirty-day evaluation of Gia Health Analytics?

The evaluation exists because a slide is not evidence. The question a medical director and an analytics lead actually have is: what does this model say about our population, from our records, and does it agree with what we already know? Thirty days answers that on your own data.

You send a prepared export of up to fifty members. The foundation model for aging runs in Scienza's secure US cloud, and your data is never used to train it.

Here is the sequence, in four steps.

Step one: choose the cohort and the question

Before anything runs, your team names a cohort and a question. Members over 75 with no cognitive assessment in the record. Everyone discharged in the last 30 days. One site, one region, or the whole book.

The question is usually some version of: where is cognitive change likely to concentrate in the next year, and does that match what our care managers see? Writing it down at the start is what makes the result at the end mean something.

This step also settles who is in the room at the end. The evaluation is for a clinical lead and an analytics lead together, because the report only becomes useful when someone who knows the members reads it beside someone who knows the data. Name both people before the export, and put the walk-through on their calendars.

Step two: send a prepared export

Scienza hosts Gia Health Analytics in our secure US cloud. You subscribe through the AWS Marketplace or Google Cloud Marketplace, compute is included, and Scienza's fee for the evaluation is waived.

Your team prepares an export of up to fifty members from what you already hold: coded conditions, encounters, medications and assessments. There is no new data collection to start.

Your information security team will ask what the model receives, what it stores and what it sends back. The answers are: the prepared export, only what the analysis and your agreement require, and aggregate reports to you. Your data is never used to train the model. On-premises deployment is coming soon.

Step three: the report comes back as aggregates

The released module produces an aggregate one-year dementia report for your cohort: eligibility and scoring counts, a cohort mean score and full provenance. Members are never named. Nothing in the report identifies a person, which is what makes it readable by a medical director and an analytics lead in the same meeting.

Falls and hospital return run on your cohort now as well, and Scienza states the evidence for each outcome separately. Five outcomes passed a pre-registered sealed evaluation. Falls replicated on a second look. Hospital return is trained, without a sealed-test result. Ask for the evidence standing of each outcome in the walk-through, and expect a straight answer about which results have been through a sealed test and which have not.

Step four: the walk-through, then your decision

Scienza walks your clinical and analytics leads through the result. The useful conversation is the comparison: does the cohort picture match what your nurses already see, and where does it point that nobody was looking?

Every output is an option for your clinicians to review. The report is population analytics for clinical and operational review, not a diagnosis of anyone. What a clinician does with a flagged cohort, through an encounter, is the clinician's decision.

At day thirty you have a report from your own records, a comparison against your own team's judgment, and a decision to make. If the next step is reaching the members the report points to, Gia, the AI co-clinician, makes those calls by phone or video, in 44 languages, and hands each result to a clinician with Proactive Decision Actions. Options, never orders. That is a separate decision, and the report stands on its own without it.

If the answer is no, you have spent thirty days and one prepared export, and nothing else. If the answer is yes, you have a cohort, a report and a team that has already agreed on what it means.

Where to start

Name the cohort and the question this week. Ask your analytics team to prepare the export. Put the walk-through on the calendar for the clinical lead and the analytics lead together.

See how ACOs and health systems run the evaluation, then start the thirty-day evaluation.