Xsolis

The Care Level Score is the engine of the core module. Xsolis describes it as a numerical representation of the likelihood of appropriate status for each patient, produced by real time predictive analytics and machine learning over the full record, and the utilization review workflow prioritizes cases by it. Continuous scoring of every patient does not exist without the model.

The rest of what Xsolis sells carries value on its own. Dragonfly also provides the utilization review, concurrent authorization, discharge planning and physician advisor workflows, analytics and reporting, and a shared case review framework with 850+ payer connections, and the company sells staffed Physician Advisor Services and Denial Services alongside the platform. Remove the model and a hospital still has a workflow platform, a payer network and a clinical services team.

The platform produces a score, not a determination, and the people around it are described. The score and the supporting clinical data go to utilization review nurses, who prioritize cases by it, and escalations go to physician advisors and peer to peer conversations with the payer; the solutions page describes decisions built on data backed insights and expert oversight.

The boundary is not published. The Baylor Scott & White study evaluated the score at a defined threshold, and Xsolis markets a score above 75 on an observation stay of two midnights as a missed opportunity for conversion, but nothing published says what score routes a case to a physician advisor, what a reviewer must document when they depart from the score, or whether a reviewer can finalize a status on the score alone. An older company post says the platform can support medical necessity decisions without additional criteria, which puts more weight on that missing rule.

The approach is described in general terms and nothing in it is identified. Xsolis says the Care Level Score is a numerical representation of the likelihood of appropriate status for each patient, built on evidence based medicine and powered by real time predictive analytics and machine learning over the full record, and a newer capability uses generative AI to draft a clinical synopsis for physician advisors before payer conversations. No model type, input feature set, training data, confidence interval or mapping from score to status recommendation is published, and no versioning or update practice.

The peer reviewed studies show how the score performs, not how it works.

The architecture is described and no provider is named. The Care Level Score is Xsolis's own predictive model reading the hospital record, and the newer clinical synopsis capability uses generative AI without saying whose model drafts it or where it runs. No hosting arrangement or subprocessor list is published.

The party that most changes the chain is a customer. The platform is sold as a shared framework in which health plans, through 850+ payer connections, see the same clinical view as the hospital for concurrent authorization, so a payer sits in the path between a patient's record and a determination by design. That is a different relationship to the record than the one a hospital's release process was built around.

Peer reviewed evaluation in live utilization management at three named health systems, each run by the health system's own clinical and research teams, with journals a buyer can find. At Baylor Scott & White (Baylor University Medical Center Proceedings) the real time Care Level Score classified inpatient designation correctly 86 percent of the time at a defined threshold and predicted better than traditional commercial screening tools. At Yale New Haven Health (Journal of Doctoral Nursing Practice) the observation discharge rate fell from 16.69 percent to 12.75 percent after the tool was integrated into a nurse led department. At Mayo Clinic Health System (Journal of Clinical and Translational Science) diagnosis related group prediction reached 81 percent accuracy and length of stay was predicted within 0.14 days. The first study answers the product's headline claim directly.

Each study is at a single system; the Yale result is a before and after comparison without a concurrent control; the announcement does not give study sizes or say whether Xsolis staff took part; and none reports accuracy by payer, admission source, diagnosis group or demographics. The customer figures Xsolis publishes (5 to 7 times average return on investment, 95 percent client retention) are vendor reported.

Externally validated security controls exist and no stewardship position is published. Xsolis announced HITRUST r2 certification for the platform in March 2022 and described it as a proactive approach to protecting health information; the website privacy policy covers site visitors only. Nothing published states how long clinical data drawn from hospital records is kept, whether it is de identified, or whether it trains the Care Level Score.

The architecture raises a sharper question. The platform is sold as a shared framework in which a health system and its payers work from the same clinical evidence, with 850+ payer connections, so a plan sees live clinical documentation for admitted patients through the platform rather than by request. Nothing says what the payer side can see, whether that view is limited to the minimum needed for the determination at hand, whether access is logged and reviewable by the hospital, or whether data from one hospital's patients improves the score used at others.

No HIPAA status is published. The website privacy policy covers site visitors only, and the home, solutions and research pages carry no HIPAA claim, business associate statement or agreement. The company's 2022 announcement of HITRUST r2 certification describes it as a proactive approach to protecting health information; a security certification is not a statement of HIPAA status, and nothing says who the contracting business associate is.

The contracting question is layered because of where the company sits. The platform is shared between a health system and its health plans working from common clinical data, so Xsolis would hold business associate relationships with two parties who are counterparties in the determination it informs.

Xsolis announced in March 2022 that its platform, then CORTEX and now Dragonfly, had earned HITRUST risk based, two year (r2) certified status. The r2 is a tailored, risk based assessment validated by an authorized external assessor, with the certification decision made by HITRUST itself and interim testing in between.

Nothing is retrievable: no current certification letter, validity dates, trust center or report request process is published. The currency question is real: a two year term from March 2022 would have needed renewal in 2024 and again in 2026, and the company's site announces no renewal. Nothing says whether the generative synopsis capability sits inside the certified scope. Third party software directories attribute a long list of other standards to the product; none is stated by Xsolis.

No device authorization is claimed. The Care Level Score predicts the appropriate level of care, inpatient or observation, and supports status and concurrent authorization decisions that utilization review nurses and physician advisors make; that is a prediction feeding an administrative and payment classification.

The regime that governs this work is accreditation and Medicare's rules for utilization review. URAC and NCQA both accredit utilization management, and URAC's version 8.0 standards added requirements on algorithmic transparency, bias testing and oversight of clinical decision support tools; Xsolis states no such accreditation. It sells Physician Advisor Services in which its own physicians take part in reviews, which puts it closer to the accreditable side of that line. Its own older material says the platform complies with Medicare's Conditions of Participation for utilization review and can support medical necessity decisions without additional criteria.

The model has been tested in production, and nothing is published on how it is governed. Three peer reviewed studies at Baylor Scott & White, Yale New Haven and Mayo Clinic Health System evaluated it in live utilization management, which is testing of the model's behavior, but no governance framework, monitoring practice, change control or subgroup result is published, and the company describes oversight only as expert review of the score.

The gap matters because of the decision the score informs. A medical necessity score shapes whether a patient is admitted as an inpatient or placed in observation. That distinction determines what the hospital is paid, and for Medicare patients it also determines the patient's own cost sharing and whether later skilled nursing care is covered. The same score also runs on the payer side of the platform. A systematic tilt would move money and coverage in the same direction repeatedly, for patients who never see the score.

The company's own framing sharpens the obligation: the score is sold as objective and as removing subjectivity from status determination, a claim about fairness.

Mechanisms let a person challenge the score before it becomes a decision, and nothing stands behind it. The platform gives utilization review nurses and physician advisors the score with the supporting clinical data, and the status decision, peer to peer conversation and appeal stay with people. Nothing published commits Xsolis to anything when the score is wrong.

There is a published accuracy figure, but not yet in a form a buyer can hold the company to. The Baylor Scott & White study reports that the Care Level Score classified inpatient designation correctly 86 percent of the time at a defined threshold, which means roughly one patient in seven was classified wrongly at that threshold; the company's announcement gives neither the number of patients nor the split between patients wrongly pushed toward observation and those wrongly pushed toward inpatient. Those two errors land on different people: one on the patient's Medicare coverage and cost sharing, the other on the payer. Xsolis also markets a conversion metric counting observation stays of two midnights or more with a high score as missed opportunities, which shows which direction the product is sold to correct.

Integration is claimed and nothing is named. Xsolis says the Care Level Score continuously monitors the full spectrum of patient data directly from the electronic medical record, and the platform is described as shared with payers for concurrent authorization, with 850+ payer connections and 615+ hospitals. No record system, interface standard, marketplace listing or write back is named on the company's site, so none of it can be checked.

The shared clinical access between a health system and its payers through Dragonfly is a distinct integration pattern, since both parties read the same record view.

Nothing published says where the platform runs or where hospital clinical data rests. The website privacy policy covers site visitors and says the site is based in the United States; the home, solutions and research pages, and the 2022 certification announcement, name no hosting provider, region or tenancy model for the platform that reads hospital records in real time.

The HITRUST r2 certification announced in 2022 implies the environment was assessed, but a certification scope is not a residency disclosure. Tenancy matters here because of the shared access model: where a health system and a health plan both work inside the same platform against common clinical data, nothing says how the two customers are separated, what sits in a shared layer, or whether the historical determination patterns behind the score are drawn from one relationship or pooled across many.

No public pricing. The offering is a hybrid: alongside the platform the company sells Physician Advisor Services covering night and weekend gaps or operating fully outsourced, so the buy can be software, staffed service, or both. A clinical staffing component behaves very differently from a license.

Precisely bounded to hospital utilization management and the workflows around it: admission and status determination, continued stay and concurrent authorization, discharge readiness and length of stay, physician advisor escalation, and denials and appeals, used by 615+ hospitals and, on the payer side, by health plans and provider sponsored plans for concurrent authorization.

Validation stands behind part of that. Peer reviewed studies at Baylor Scott & White, Yale New Haven and Mayo Clinic Health System support the hospital side; nothing published evaluates the score as payers use it.