A blood test can suggest a treatment is working while the person taking it continues to struggle. Their biomarker improves, but fatigue still makes work difficult. Their medical record shows a stable condition, yet symptoms change between appointments. Both accounts can be true, and the distance between them exposes a significant gap in how medicines are researched.
Pharmaceutical companies have never had more information available to them. Genomic datasets, clinical records, imaging and advances in artificial intelligence are expanding what researchers can investigate. But more data is not automatically a more complete account of a patient’s life.
The challenge is continuity: understanding how people respond to illness and treatment over time, including the days when they are not in a hospital, clinic or trial visit.
Healthcare systems are designed to record clinical encounters. A hospital record is invaluable for diagnoses, tests, prescriptions and admissions. It is less consistent at describing the everyday changes that may determine whether a treatment is genuinely tolerable and useful.
Consider two people receiving the same medicine for a chronic condition. Their laboratory results may look similar, while one remains active and the other experiences persistent fatigue, sleep disruption or a gradual decline in function. If a research dataset captures principally laboratory results and hospital visits, the comparison is incomplete.
Patient-reported outcomes can reveal symptoms and daily functioning. Appropriate wearable measurements may provide additional context on activity, sleep or physiology. Treatment histories and clinical measurements remain essential. The opportunity lies in connecting these sources thoughtfully, rather than treating any one of them as the whole story.
This has implications well beyond patient monitoring. Better longitudinal evidence could help researchers understand the burden of disease, design studies around outcomes patients value, investigate why treatment responses differ and examine what happens after medicines enter routine care.
AI can help analyse relationships across these different information sources. Yet an algorithm cannot reliably recover experiences that were never measured. Larger models do not remove the need for consent, data quality, careful scientific design and testing across diverse patient populations.
There is also an important commercial distinction. Collecting large quantities of information is not, by itself, a durable advantage. A useful research platform must be able to establish where its data came from, what it represents, what is missing and which conclusions the evidence can support.
London-based Sanius Health is building a patient intelligence platform designed to connect clinical information with what patients report and experience over time. Its core capabilities include direct patient engagement, longitudinal data collection, research-grade patient-reported outcomes and the integration of relevant clinical and wearable-device measurements.
Through programmes in rare diseases, haematology, oncology and other complex conditions, Sanius works to capture disease burden, symptoms, treatment experience and changes in health beyond isolated appointments. The objective is to turn those observations into appropriately governed evidence that pharmaceutical and healthcare researchers can investigate and use.
Founder and CEO Orlando Agrippa sees the opportunity as much larger than a digital health application. Sanius is developing AI capabilities alongside its data platform, including Patient Find, intended to help identify people whose clinical characteristics warrant further assessment for under-recognised conditions, and models investigating whether longitudinal patterns can help predict disease complications. Any clinical application depends on condition-specific validation and professional oversight.
At its centre is a repeatable model: engage patients, connect their experiences with relevant clinical data, establish the quality and provenance of those observations, and apply rigorous research and analytics. That combination, rather than a series of disconnected tools, is the core of the Sanius proposition.
Drug development is becoming more computational, and regulators are scrutinising how real-world evidence is generated and used. That makes scientific credibility central. Data provenance, representativeness, missingness, privacy and appropriate validation are not background technical issues; they determine whether insights can be trusted.
No longitudinal dataset can replace randomised trials for the questions those trials are designed to answer. Nor should consumer-device signals be mistaken for clinical diagnoses. But when collected and analysed appropriately, information from everyday life can complement conventional evidence and make research questions more relevant to patients.
The next competitive advantage in pharmaceutical AI may therefore be less about having the biggest algorithm and more about connecting the right information to the right question, consistently and responsibly.
Medicine has become remarkably precise at measuring what happens in molecules. Understanding what happens to people between appointments may be the next equally important step.
.png)