Clinical data is any information about a patient’s health that is collected during the course of medical care or clinical research. This includes everything from blood pressure readings and lab results recorded during a routine checkup to adverse events tracked during a drug trial. It forms the foundation of medical decision-making, drug approvals, insurance billing, and increasingly, the algorithms that predict health outcomes before symptoms appear.
Where Clinical Data Comes From
The most familiar source is the electronic health record, or EHR. Every time you visit a doctor, a nurse takes your vitals, or a lab processes your bloodwork, that information enters your EHR. These records contain diagnoses, medications, allergies, imaging results, surgical histories, and clinician notes generated during routine care.
But EHRs are only one piece. The FDA recognizes several other major sources: medical claims data (the billing records submitted to insurers for payment), product and disease registries (databases that track everyone with a specific condition or everyone using a specific device), and patient-generated data from home settings, including wearable health trackers. All of these fall under the umbrella of “real-world data,” which the FDA defines as data relating to patient health status or the delivery of health care routinely collected from a variety of sources.
Then there’s clinical trial data, which is collected under tightly controlled conditions. During a drug trial, researchers record specific endpoints: how patients respond to treatment, what side effects emerge, how long it takes for symptoms to improve or worsen. Phase I trials focus primarily on toxicity and adverse events to find a safe dose. Phase II trials measure treatment response. Phase III trials compare the new treatment against existing options using larger groups. The data from these trials is what regulatory agencies review when deciding whether to approve a new drug.
What Gets Recorded
Clinical data spans a surprisingly wide range of information. The core categories include:
- Diagnostic data: lab values, imaging scans, biopsy results, genetic tests
- Treatment data: prescriptions, procedures performed, therapies administered
- Physiological measurements: blood pressure, heart rate, oxygen saturation, weight
- Administrative data: admission and discharge dates, billing codes, insurance claims
- Patient-reported outcomes: pain levels, quality of life scores, symptom diaries, health behaviors
Patient-reported data has become an increasingly important category. These are measures that come directly from you rather than from a lab or imaging machine. They cover five main areas: health-related quality of life, functional status (physical, cognitive, and sexual function), symptoms like fatigue and pain intensity, health behaviors such as exercise, diet, and substance use, and your experience of the care itself. Clinicians use this information to identify areas for intervention and to track how well treatments are actually working from the patient’s perspective, not just on paper.
Clinical Records vs. Claims Data
One important distinction that often causes confusion is the difference between clinical records and insurance claims data. Both contain health information, but they serve different purposes and capture different levels of detail.
Clinical records are created by the provider caring for you. They contain the full picture: what the doctor observed, what they suspected, what they tested for, and what they found. Claims data, by contrast, exists because someone needed to get paid. It contains diagnostic codes and procedure codes submitted to insurers, but it often misses conditions that were noted clinically but weren’t relevant to the bill. Research comparing the two sources found that agreement on diagnoses ranged from 65% to over 90% depending on the condition. Cardiovascular disease, for instance, was recorded in clinical data but completely absent from insurance claims in 31% of prostatectomy cases and 17% of cholecystectomy cases. Claims data tends to undercount less severe conditions that don’t drive the visit’s billing.
This gap matters because many large-scale health studies rely on claims data since it’s easier to access at scale. Knowing that it systematically misses certain conditions helps explain why clinical records remain the gold standard for accuracy.
How Clinical Data Gets Shared
One of the biggest challenges in healthcare is getting clinical data to move between systems. Your primary care doctor, your specialist, your hospital, and your pharmacy may all use different software. For data to flow between them, it needs to be structured in a common format.
The current leading standard is called FHIR (Fast Healthcare Interoperable Resources), developed by the health data standards organization HL7. FHIR uses web-based technology to let different health systems exchange information in a standardized way. When paired with standardized medical vocabularies that assign consistent codes to conditions and symptoms, FHIR allows a diagnosis entered in one system to be understood by another. It has emerged as the best candidate for achieving interoperability, replacing older and more rigid data exchange formats that were harder to implement.
Real-World Evidence and Drug Regulation
Clinical data collected outside of controlled trials has taken on a new role in drug regulation. The FDA now draws a clear line between real-world data and real-world evidence. Real-world data is the raw information pulled from EHRs, claims, registries, and digital health devices. Real-world evidence is what emerges when you analyze that data to draw conclusions about a medical product’s benefits or risks.
This distinction matters because the FDA increasingly uses real-world evidence to support regulatory decisions. Rather than relying solely on clinical trials, which are expensive and study carefully selected patient populations, regulators can examine how a drug performs across millions of real patients with messy, complicated health profiles. This approach helps catch safety signals that trials might miss and can support new uses for existing medications.
Clinical Data and Predictive Algorithms
Hospitals and health systems are increasingly feeding clinical data into machine learning models that predict patient outcomes. These models analyze sequential patterns in vitals, lab results, and other measurements to forecast events like hospital readmission or disease progression. In some cases, machine learning has outperformed traditional statistical methods. One study found that these approaches predicted abnormal artery wall thickness in patients with type 2 diabetes more accurately than conventional models.
Image-based clinical data is another major application. Algorithms trained on large collections of medical images can learn to distinguish healthy tissue from abnormalities, assisting in tasks like tumor detection and disease classification. The key ingredient in all of these applications is volume: models improve as they’re trained on more clinical data, which is why large health systems with millions of patient records are at the center of this work.
Wearables and the Clinical Data Boundary
Consumer wearables like smartwatches and fitness trackers generate enormous amounts of health-related data, but not all of it qualifies as clinical data. The distinction hinges on reliability, evidence, and integration into the medical record. Healthcare professionals have raised concerns that composite scores generated by wearables, like sleep quality ratings, are built on proprietary algorithms never designed for diagnosis.
For wearable data to cross the threshold into clinical use, health systems are working to define specific care pathways: which devices have evidence behind them, what clinical thresholds trigger action, and at what point that data gets sent to the EHR and falls under health privacy protections. Until those guardrails are in place, wearable data occupies a gray zone, useful for personal health awareness but not yet treated as clinical-grade information in most care settings.