Medication risk, recomputed against the patient in front of you.
MosaicRx is designed to surface the medication risks that matter most for a specific patient, conditioned on that patient's current physiology, and show the reasoning behind every determination.
Elevated anticholinergic burden
HighCumulative ACB score 7 across 3 active meds - associated with cognitive slowing & falls.
Prescribing cascade signal
WatchDiuretic started 14 days after NSAID initiation - temporal pattern consistent with cascade.
Fall-risk phenotype: high
HighOrthostatics + sedating load + recent gait change. Phenotype match 0.89.
Illustrative example. Not a real patient.
The most expensive medication errors don't look like errors.
They look like new diseases - until a patient falls, becomes delirious, or lands in the ED. By then, the chain is invisible.
A cascade starts when a medication side effect gets read as a new condition. The new condition gets treated. The patient now carries both drugs, and the second one often carries its own side effect. Nobody made an error at any single step, which is why these are so hard to catch by looking at one prescription at a time.
The same medication list is not the same risk in every patient.
Standard burden scoring treats each drug as carrying a fixed weight. A patient with stage four kidney disease clears those drugs differently, so the identical list carries different risk. MosaicRx is designed to adjust both the burden score and the threshold it is measured against, as functions of the patient's current physiology.
Nominal
- Anticholinergic burden: 3.0
- Threshold: 3.0
- Status: Within range
Conditioned on stage four CKD
- Anticholinergic burden: 4.2
- Threshold: 3.0
- Status: Elevated
Illustrative example. Not a real patient.
Nothing about the prescription changed. The patient did.
A burden of three in stage four kidney disease is not a three.
Top 3 risks · why they matter · what's driving them.
MosaicRx is designed as a prioritized, patient-specific medication-risk review at the prescribing moment, not another alert. Clinical utility will be evaluated in the pilot.
- Phenotype-specific, not population-generic
Risks are intended to be scored against the patient's longitudinal trajectory, not a static drug-drug table.
- Glass Box explainability
Every signal is designed to expose its evidence trail: contributing meds, timing, and the questions worth reviewing.
- Clinician-in-the-loop, always
MosaicRx is designed to surface risks. The clinician decides. Nothing auto-prescribes, nothing auto-stops.
- It tells you what it ruled out.
MosaicRx is designed to show when it evaluates a hypothesis and rejects it, with the work: the hypothesis considered, the pharmacologic and temporal reasoning behind the rule-out, and the agents examined and excluded with grounds. A tool that speaks only when it has something to flag gives you no way to know whether it looked.
- Every response teaches the engine.
When a clinician responds to a determination, the response is intended to be captured as acted on, rationale documented, not relevant, or not actionable today. The distribution across those four states is how we plan to find where the engine is wrong.
Cumulative ACB ≥ 3 in adults >75 is associated with cognitive decline, urinary retention, and falls. Patient's score: 7.
- →Is amitriptyline still needed for sleep, or can it be tapered?
- →Substitute oxybutynin with mirabegron?
Three steps. No new workflow.
Context capture
EHR data, longitudinal medication history, labs, and real-world inputs where available - assembled into a patient-specific picture.
Risk engine
Phenotype-aware models rank by relevance, not completeness. Built explicitly to avoid alert fatigue.
Glass Box output
Top 3 risks with evidence trace, contributing meds, timeline, and questions to review. Clinician decides.
The model never makes the clinical claim.
The intended architecture requires every clinical assertion MosaicRx produces to come from a deterministic rules engine with a complete decision log. The language model renders the explanation in readable form. It cannot generate, alter, or extend a clinical claim.
That separation is designed to make every output reproducible and traceable to a specific rule and a specific source. The pilot will evaluate whether running the same patient twice gives the same answer, with the same reasoning, every time.
Three ways a tool can approach medication risk.
Not an alert. Not a scribe. Not a polypharmacy calculator. MosaicRx is designed as a prioritization layer at the prescribing moment, around clinical judgment.
Flags known drug-drug interactions from a fixed table, without regard to who the patient is. This is what most EHR alerting does, and it is the reason alert fatigue exists.
Estimates the likelihood of a future event from a trained model. Useful for triage, opaque under questioning, and bounded by the population the model was trained on.
Starts from what is already happening to the patient, works backward to the medication that most plausibly explains it, and conditions the entire calculation on current physiology. This is where MosaicRx is designed to sit.
For the clinician making the call, and the team carrying the risk.
MosaicRx has to earn its place with the clinician and the team carrying risk. These are the needs the pilot is designed to evaluate.
For the clinician
You did not go into medicine to fight your inbox or to second-guess a medication list at the end of a long Friday.
- It watches so you can decide. When an older patient is on ten or more medications, you already sense one of them might be the problem. Finding which one in a fifteen-minute visit is the hard part. MosaicRx is designed to support that review and surface the few things most worth your attention.
- It stays in the passenger seat. Nothing auto-prescribes and nothing auto-stops. You see the reasoning, you weigh the questions it raises, and you make the call.
- It gives you something to point to. Every signal is intended to carry its own evidence, so you can review the reasoning with a worried daughter or document your thinking in the chart.
- It is designed around the moments that matter most. Supporting safer medication review and more time at home are the goals. Whether MosaicRx contributes to those outcomes remains to be evaluated.
The pilot will evaluate whether this approach supports clinical review without adding noise to your day.
For the value-based care leader
When your group carries risk on older, complex patients, medication-driven harm is one of the largest costs you cannot see coming.
- It targets medication-related cost of care. Falls, delirium, ED visits, and readmissions can affect medical loss ratio. MosaicRx is designed to surface medication-related risks; effects on clinical outcomes and cost have not been established.
- It is designed with medication-safety quality measures in mind. MosaicRx is intended to address the risks behind the medication-safety measures your plans are scored on by the Healthcare Effectiveness Data and Information Set (HEDIS): Use of High-Risk Medications in Older Adults, Polypharmacy: Use of Multiple Anticholinergic Medications in Older Adults, Polypharmacy: Use of Multiple Central Nervous System-Active Medications in Older Adults, and the Potentially Harmful Drug-Disease Interactions in Older Adults measure that flags sedating medications in patients with a history of falls. These are measures where a lower rate means better performance, and MosaicRx is designed to surface those risks with an evidence trail. Any effect on HEDIS or Stars performance remains to be evaluated.
- It aims to focus your scarce care-management time. MosaicRx is designed to prioritize relevant medication risks rather than flood pharmacists and care managers with noise. The pilot will evaluate usefulness in practice.
- It starts with measurable questions. A focused pre-pilot workflow can begin without a deep EHR build. Clinical utility, clinician response, and repeat use are planned evaluation endpoints, not proven results.
Clinical outcomes and financial impact require follow-up beyond the initial pilot.
For specialists managing the whole list
Nephrologists already do this, for the kidney. Renal function governs nearly every drug eventually, which is why nephrology is the specialty most willing to question a prescription written by someone else. MosaicRx is designed to apply that same reasoning across the whole medication list, conditioned on the number you are already watching.
The family feels it too.
Behind most older patients on a long medication list is a family member tracking the pillbox, noticing the new confusion, and quietly wondering why their mother seems worse. As care moves into the home, that family member has become part of the care team.
In 2024, about 59 million U.S. family caregivers provided an estimated $1.01 trillion in unpaid care, a workforce now larger in value than all Medicaid spending (AARP, Valuing the Invaluable 2026).
- It aims to turn a silent change into a conversation. When a clinician adjusts a medication, the family usually hears only that something changed. MosaicRx is designed to give the clinician an explainable reason to review and share, so the daughter at the bedside can understand the why.
- It is designed with families in mind. More time at home and fewer medication-related crises are the intent. These outcomes have not yet been demonstrated for MosaicRx.
- It aims to lighten an invisible load. The burden of a complication extends to family caregiving, missed work, and worry. The pilot begins the evidence-gathering needed to understand any effect on that burden.
MosaicRx stays a clinical tool used by clinicians. Its potential impact on families remains to be evaluated.
What the pilot measures.
The first deployment is planned as a community clinic pilot with a December readout. Planned endpoints are cascade interception rate, change in conditioned medication burden, the distribution of clinician adjudication across the four response states, repeat use, and clinician time recovered.
Outcome validation, meaning whether intercepted cascades translate into fewer falls and fewer avoidable admissions, requires longitudinal follow-up that this pilot is designed to begin rather than to conclude. We will publish what we find, including the parts that do not work.
Built to generate the evidence, not only the intervention.
For health economics and outcomes research, the question is whether an intervention produces measurable, defensible value in the real world. MosaicRx is designed to generate that evidence as a byproduct of normal use. The planned evidence pipeline would log surfaced risks, clinician responses, and downstream outcomes to support evaluation of medication burden, cognition, and fall risk in older adults.
- Anticholinergic and psychoactive burden change. Change in average anticholinergic burden and CNS-active and sedative medication load per patient, the exposures most tied to cognition and falls in older adults.
- Longitudinal clinical outcomes. Change in medication-related falls, delirium, and cognitive and functional decline, along with the avoidable emergency visits and readmissions those events drive.
- Phenotype-stratified outcomes. Results stratified by the dimensions that actually govern medication risk in this population: cognitive burden, fall risk, activity and functional status, and psychoactive exposure. Physiologic factors that affect drug handling adjust the underlying score but are not the lens.
- Modeled cost avoided per 1,000 patients. A conservative total-cost-of-care estimate sized for risk-bearing populations.
The intent is to evaluate MosaicRx as both a point-of-care tool and a source of real-world evidence for value demonstration and budget-impact modeling. No clinical or economic benefit has yet been established.
Pilot MosaicRx in a community clinic.
Designed for risk-bearing outpatient groups managing complex older adults. Start with a focused workflow - integration may follow evaluation.
- A clinical champion and 3-10 prescribing clinicians
- Synthetic or de-identified data for the first phase (no integration required)
- Defined patient cohort (e.g., 65+ with ≥5 active meds)
- Optional EHR integration scoped after pilot validation
- Cascade interception rate
- Clinician adjudication across four response states
- Repeat use per clinician
- Time recovered per encounter
- Explainability engagement rate
The team building MosaicRx.
A clinician-aware, builder-first team backed by senior clinical and commercial advisors across Stanford, community medicine, and value-based care.
Operator-clinician translator. Builds at the intersection of value-based care, medication safety, and prescribing workflow.
Chemistry, B.S. Biotechnology and pharmaceutical expertise. Founder of Prasad Marketing Company.
Engineering leader. Built mission-critical civic tech at scale. Owns architecture, security, and the EHR integration path.
The questions clinical and economic buyers actually ask.
Request a demo or start a pilot conversation.
Tell us a bit about your team. We'll follow up within two business days.

