Real-world evidence (RWE) is clinical evidence about how a product performs in routine care, built from data collected outside a research clinic. For a digital health company, it is the bridge between a promising pilot and the proof that buyers, regulators and investors will pay for. It only works when the data, the study design and the clinical question fit together, and that fit is a clinical leadership decision as much as a data science one.
This guide is written for healthtech founders, healthcare SaaS leaders and the investors who back them. It covers what the terms mean, which data sources are worth using, how FDA frames fitness for use, and how to judge whether a company's evidence will survive scrutiny.
TL;DR
- Real-world data (RWD) is raw material. Real-world evidence is the analysis of that data, and the FDA treats the two as different things.
- Every source has a built-in blind spot: claims lack clinical detail, EHRs lack breadth. Pick the source for the question, not the other way around.
- Credible RWE is designed like a trial: prespecified population, outcomes, time zero and analysis plan, written before outcome data are reviewed.
- Digital health technologies need verification and validation before the data they generate can support a claim.
- Investors should ask whether the evidence matches the product's intended use and the buyer's decision, not how large the dataset is.
Table of Contents
- What real-world data and real-world evidence mean
- Choosing real-world data sources
- From data to evidence: a fit-for-purpose process
- Making digital health data count: verification and validation
- Where RWE fits across the product life cycle
- What FDA expects when RWD supports a device claim
- How investors and boards should read a company's RWE
- Our perspective
- Working with The StartUp MD on your evidence strategy
- FAQ
- Sources
What real-world data and real-world evidence mean
The FDA defines real-world data as data on patient health status or the delivery of care that is routinely collected from a variety of sources. Real-world evidence is the clinical evidence about a product's use and its potential benefits or risks that comes from analyzing that data. The same page lists the sources: electronic health records (EHRs), registries, administrative claims, device-generated data, patient-generated data, public health surveillance data, clinically annotated biobanks, device data repositories and billing data.
The distinction matters for a simple reason. A pile of usage logs or claims lines is not evidence. An ASCO Post commentary makes the point directly: variables can be missing, incompletely captured or misclassified, and analysis cannot fully compensate for information that was never collected. What turns data into evidence is curation, validation and a study design that answers a defined question.
How it differs from clinical-trial evidence
A randomized trial controls who gets the intervention, who is enrolled and how outcomes are measured. RWD, by contrast, is not collected in a special research clinic or setting. That makes it closer to how your product is actually used, with messier patients, variable adherence and real workflows. It also means you inherit the biases of whoever recorded the data.

RWE is not limited to retrospective analysis. The FDA notes it can come from non-interventional studies and from designs such as point-of-care trials and externally controlled trials, where the control arm is drawn from RWD sources. For digital health, that range is useful: a SaaS product embedded in a health system can often support a pragmatic design that a classic trial cannot.
Choosing real-world data sources
The first strategic decision is which data to use. Each source answers some questions well and others poorly. The FDA's guidance on EHR and claims data for drug and biologic studies notes that claims exist to support payment and may not accurately reflect a disease or how it is managed, and that EHR data are generated for clinical care and limited to what one health system captures. The Sentinel Initiative's claims versus EHR comparison fills in the other side of the ledger.
| Source | What it does well | Where it falls short |
|---|---|---|
| Administrative claims | Longitudinal, time-stamped care across settings; quality supported by reimbursement use | Misses non-prescription and out-of-pocket care, demographics and detailed clinical information; can change during adjudication |
| EHR data | Clinically rich elements such as labs and vitals | Smaller, geographically restricted samples; a missing value may mean no event or no recording |
| Registries | Disease-specific, often curated outcomes | Narrower populations and enrollment dependent |
| Device and patient-generated data | Continuous, patient-level signals from wearables and apps | Needs verification and validation before it can stand in for a clinical measure |
Most useful studies link two or more of these. A practical 2026 guide to implementing RWE describes linking claims, EHR and registry data through tokenization or probabilistic matching, using HIPAA Safe Harbor or Expert Determination de-identification. Linkage quality then becomes part of the evidence.
If your team is comparing candidate datasets for relevance, reliability, completeness and provenance, a healthcare and life sciences research data platform can help you survey clinical trial, filing and academic sources in one place before you commit to a design.
From data to evidence: a fit-for-purpose process
The same practical guide describes an iterative fit-for-purpose and feasibility assessment. In plain terms, you test whether the data can answer the question before you lock the protocol. Here is that routine as a founder would run it.
- Define the clinical question and the decision it supports. Is it a payer asking about utilization, a health system asking about readmissions, or a regulator asking about safety? The decision sets the evidence bar. Where this goes wrong: teams start from the dataset they happen to have and then search for a claim it supports.
- Check that the data contain what you need. The guide lists the population, exposures, outcomes, follow-up and covariates. If you cannot see who used your product and when, you cannot estimate an effect.
- Assess completeness, timing, provenance and linkage. The same assessment covers misclassification, sample size and informative missingness, meaning data that are absent for reasons related to the outcome. Any of these can send you back to change the data source, design, endpoint or estimand (the precise quantity the study is trying to estimate).
- Design it like a trial. ISPOR's authors argue that reliable and relevant data are not enough, because confounding and time-related biases are common. They recommend prespecified eligibility, treatment strategies, outcomes, follow-up, causal contrasts, an analysis plan and a defined time zero.
- Finalize the protocol before looking at outcomes. FDA's updated device RWE guidance, as described in a CDRH town hall transcript, requires the protocol and analysis plan to be created and finalized before outcome data are reviewed, with a relevance and reliability assessment in the study report.
- Report limits plainly. Transparency about how key variables were collected, derived and validated is what lets a reviewer or buyer trust the result.
A Duke Health Policy white paper adds a useful reminder: RWE is not simply anecdotes drawn from RWD. It requires curation, validation and standardization, and the data-hygiene tools built for claims research can help close method gaps for EHR and patient-generated data.
Our guide on clinical evidence building for founders walks through how this sequence fits alongside early validation work.
Making digital health data count: verification and validation
A sensor or app is not automatically a measurement tool. FDA's August 2026 paper on digitally derived measures says a digital health technology used to generate a digitally derived measure in a clinical investigation should be verified and validated to be fit for purpose. It breaks this into three layers:
- Verification: the technology measures the defining parameter accurately and precisely.
- Analytical validation: the technology assesses the clinical event or characteristic in the target population.
- Clinical validation: the measure reflects a meaningful aspect of health and tracks changes in disease or clinical status.
The paper also says patients, caregivers and clinicians should help determine what counts as meaningful, and it points to a risk-based approach that tailors evidence to intended use. For a SaaS company, the practical lesson is to settle what your product's data represent clinically before you build the evidence story on top of them.
Where RWE fits across the product life cycle
RWE is not only a late-stage regulatory tool. A 2024 framework for digital therapeutics proposes four milestone-driven phases: design, development, testing and monitoring, with RWD incorporated throughout. Adapted for most healthtech teams, that looks like this:
- Design: use RWD to size the unmet need and understand current care patterns before you commit to a feature set.
- Development: use it to shape the intervention and define the outcomes that matter to clinicians and patients.
- Testing: use it to compare your users against a credible comparator, including historical controls where appropriate.
- Monitoring: use post-deployment data to track safety, effectiveness and durability of results.
A 2022 review of digital therapeutics evaluation shows the pattern in practice. It notes that RWD can supplement randomized trials, and cites retrospective claims analyses of reSET and reSET-O in 351 patients with opioid use disorder for cost-effectiveness, along with retrospective cohort studies of digital therapeutics in type 2 diabetes and hypertension.
For broader sequencing, see our clinical strategy guide for digital health companies and the clinical product development process.
What FDA expects when RWD supports a device claim
The FDA's CDRH page on RWE and its town hall materials show how seriously the agency treats data quality. The updated approach adds examples of when an investigational device exemption may be required under 21 CFR 812 for clinical studies using RWD, and says data routinely collected from devices authorized under an emergency use authorization may count as RWD if relevant and reliable.
The town hall transcript also explains the two tests reviewers apply:
- Relevance: data availability (detail, longevity, continuity of care), linkages across sources, timeliness, and whether the sample generalizes to the intended-use population.
- Reliability: data accrual and data quality and integrity, including quality control processes, completeness, accuracy, consistency, patient privacy protections and prior demonstration that RWE can be generated from the source.
The 2026 practical guide frames the first hurdle for RWE as a primary basis for approval or label expansion as acquiring a dataset that meets FDA fitness-for-use standards: relevance, reliability, fitness for purpose, transparency and provenance. Software teams working through related regulatory questions may also find our pieces on clinical decision support documents and medical device data systems useful.
Privacy sits inside this too. A 2025 review of prescription digital therapeutics notes that HIPAA protects EHR data but has no explicit provisions for prescription digital therapeutics, and that 42 CFR Part 2 adds protections for substance use disorder records. Build your data governance for the stricter regime from the start.
How investors and boards should read a company's RWE
For a venture or private equity team, the question is whether the evidence is decision-grade. Use the same logic regulators use, translated into diligence questions:
- Does the evidence match the intended use? A retention study does not prove a clinical outcome, and an outcome study in one population does not prove value in another.
- Is the data source right for the claim? Claims-based findings about clinical severity, or EHR findings from a single system presented as national, deserve a closer look at the source's limits.
- Was the design set before the results? A prespecified protocol, analysis plan and time zero signal discipline. This is the standard ISPOR and FDA both point to.
- Is the measurement validated? If the evidence rests on device-generated data, ask for verification and validation work for the specific measure.
- Can the buyer act on it? Payers, health systems and regulators each need different outcomes. Evidence that answers none of their decisions is a cost, not an asset.
- Is there a monitoring plan? Ongoing post-deployment data show whether results hold as the customer base grows.
These are also the questions a founder should be able to answer before a Series A or a health system procurement review. A physician leader who can read the study design, not only the headline result, closes most of the gaps quickly.
Our perspective
Over 25 years in medicine and business, I have seen the same pattern repeat. A healthtech team builds a good product, collects a lot of data, and then discovers that the data cannot answer the question a buyer, regulator or investor is asking.
Real-world evidence is a clinical discipline as much as an analytic one. Someone has to decide which patients the product is meant for, which outcome matters to the clinician at the bedside, and whether a number in a dashboard reflects health or just activity. That is where a physician leader changes the result.
This is the gap a fractional Chief Medical Officer is built to close. We work with healthcare SaaS and digital health companies on clinical strategy, product evaluation, regulatory guidance and investor support, and we bring both a clinical and a business view to the evidence plan. The earlier that view enters, the fewer studies get rebuilt.
If something about your evidence feels misaligned with your next milestone, I would welcome a conversation.
Paul Bergeron MD, MBA
Working with The StartUp MD on your evidence strategy
The StartUp MD is a healthcare advisory firm led by Paul Bergeron, MD, MBA. We offer Fractional Chief Medical Officer and advisory services for healthcare SaaS and healthtech companies, including clinical strategy development, product evaluation, regulatory guidance, market readiness assessment and investor support. Founders use that leadership to decide which evidence to build first. Investors use it to assess portfolio companies' clinical claims and readiness.

To discuss your evidence plan, start at thestartupmd.com.
FAQ
Is real-world data the same as real-world evidence?
No. Real-world data is the routinely collected information, such as EHR entries, claims or device readings. Real-world evidence is the clinical evidence about a product's use, benefits or risks that comes from analyzing that data with an appropriate design. The FDA's definitions keep the two separate.
Can real-world evidence replace a randomized trial for a digital health product?
Sometimes it can support or complement one. FDA describes designs such as externally controlled trials and point-of-care trials that use RWD, and notes that RWE has supported regulatory decisions. Whether it is enough depends on the intended use, the data's relevance and reliability, and a prespecified analysis plan. Agree the approach with your regulatory and clinical advisers early.
Which data source should a healthtech startup start with?
Start with the question. Claims are strong for utilization and cost across settings, EHR data for clinical detail such as labs and vitals, and registries for disease-specific outcomes. Many credible studies link two sources, so plan for de-identification and linkage quality from the beginning.
What do investors look for in a company's real-world evidence?
They look for a match between the evidence and the product's intended use, a data source suited to the claim, a protocol set before results were reviewed, validated measures and outcomes a buyer can act on. A clear plan for ongoing monitoring adds confidence that results will hold at scale.
Sources
- CDRH and Real-World Evidence, FDA
- Realizing the Promise of Real-World Evidence, FDA
- Real-World Data: Assessing Electronic Health Records and Medical Claims Data, FDA
- Claims and EHR Data, Sentinel Initiative
- A Practical Guide to Implementing Real-World Evidence, PMC
- Advancing Patient Health Through Regulatory-Grade Real-World Evidence Generation, ISPOR
- Transcript for CDRH Town Hall Real World Evidence, FDA
- A Framework for Regulatory Use of Real-World Evidence, Duke Health Policy
- Key Considerations for the Development and Use of Digitally Derived Measures, FDA
- The Digital Therapeutics Real-World Evidence Framework, PMC
- Clinical Evaluation of Digital Therapeutics: Present and Future, PMC
- Therapeutic Evaluation and Utilization Analysis of Mental Health Prescription Digital Therapeutics, PMC
- Real-World Data Are Not Necessarily Real-World Evidence, ASCO Post
