What do electronic health records and registries actually give a researcher, and where do they fall short? The plain answer is that they can supply real world data, which can then be analyzed into real world evidence about how a medical product is used and what outcomes follow. Their value depends on what they capture, how they are structured, and whether their limits are stated clearly.
I think this is where many people blur two different things. Real world data is the raw material. Real world evidence is the claim built from that material after analysis. An EHR or a registry does not speak for itself. It becomes useful only when a researcher can see what is recorded, what is missing, and how the records relate over time.
An EHR, or electronic health record, is the digital chart created in routine care. It may include diagnoses, prescriptions, lab results, and visit notes. A registry is a planned collection of data about a disease, a treatment, or a group of patients. Both are built from care that already happened, not from a trial designed for one study question.
That matters because these sources have a kind of ordinary strength. They follow patients over time. They can show when a treatment started and what happened later. That temporal order is useful when someone is trying to judge possible benefit or harm. If the record shows a drug was prescribed in January and a hospital visit came in March, the sequence is visible. A causal claim is still not automatic, but the order is there to inspect.
They also help with scale. A database that covers many patients can show how often an outcome occurs in a treated group. It can also show a comparison group. That makes risk estimates possible in a way that is much harder with spontaneous reports alone, where the total number of exposed patients is often unknown.
Another strength is the range of background detail. EHRs and registries often hold age, sex, diagnoses, other medicines, and other conditions. These are the kinds of factors that can confuse an analysis if they are not seen. A patient on one medicine may also have diabetes, kidney disease, or another treatment. Those facts matter when a researcher tries to explain an outcome.
Here is a small example. Imagine a registry for people with rheumatoid arthritis. It records which patients start a new drug, when they start it, and whether they later report a serious flare or a hospital stay. A librarian or researcher reading that registry knows several things at once. They know the group under study, the time order, and the outcome pattern. They still need to ask what the registry does not record. Maybe it misses symptoms that were never entered. Maybe it has only one clinic’s patients. Maybe follow-up stops too soon. Those limits change how far the findings can be trusted.
That is the real habit to build with these sources. Do not begin by treating them as complete. Begin by asking what kind of record they are. An EHR is built for care, not for one research design. A registry is built for a defined purpose, but its scope is still narrow. Neither source captures every fact about a patient’s life. Neither source gives a final answer on its own.
For libraries and information work, this is an access question as much as a method question. A source is only as useful as its documented coverage. Who is included? What years are present? Are notes searchable, or only coded fields? Are visits linked across time, or stored as separate events? These are not technical side issues. They shape what a user can actually recover from the record.
The search side matters too. A database may offer structured fields, full-text search, or both. Structured fields are useful when a researcher needs a diagnosis code or a drug name. Full-text search can catch narrative notes, but it can also return many weak matches. A good search strategy depends on knowing which parts of the source are indexed and which parts are not. I care about that before any claim about insight or completeness.
There is also a larger point about real world evidence itself. Evidence is not the same as raw capture. It is the result of asking a specific question of a specific data source. The question may concern safety, effectiveness, risk, or burden of disease. The answer will depend on whether the source is rich enough for that question. A registry may be excellent for a narrow disease group and thin for broad population comparison. An EHR may be broad and messy at the same time. That is not a flaw in the abstract. It is the shape of the source.
The practical lesson is simple. EHRs and registries are valuable because they show what happens in ordinary care, with real patients and real follow-up. They can support estimates of risk, comparison groups, and studies of confounding factors. They do not erase the need for careful reading of scope, coding, and missing data. The evidence is only as sound as the record behind it.
With that in view, a reader can now tell the difference between a care record and a research claim, and can judge what an EHR or registry can reasonably support. That is the kind of plain judgment The Source List tries to keep alive: one digital source worth knowing, one search tip, and one honest limitation.