Definition of Real-World Data (RWD)
Real-World Data (RWD) is data that is relevant to a patient’s health status or data related to the delivery of healthcare.
RWD can come from different sources such as electronic health records (EHR), medical claims and billing data, patient-generated data such as wearables and health apps, and data from product registries and disease registries.
Definition of Real-World Evidence (RWE)
Real-World Evidence (RWE) is information derived from analysis of RWD that can be used in the regulatory processes and private or public decision making related to a medicinal product, medical device or any other kind of treatment.
RWE is clinical evidence obtained from the analysis of RWD that provides information about usage, risks, and benefits of the treatment method derived from sources other than traditional randomized clinical trials (RCT).
Depending on the country, retrieving data which originates from hospital databases is relatively “cheap and fast”.
For example, in the UK it can take about 4 – 6 months to fetch data from a database for a single hospital or group of connected hospitals – also known as a Hospital Information System, e.g. UK Hospital Episode Statistics (HES) database. The cost of data retrieval depends on the administrative costs of the registry or database holder.
The “cheap and fast” here is meant to be sarcastic.
In the current world where data is being collected about everything into electronic databases and APIs are a standard (albeit relatively unknown one), it should not take months of waiting and days of administrative work to retrieve data.
The hospital databases can include inpatient diagnoses, procedures performed, medical devices and drugs prescribed and dispensed, equipment and supply fees, imaging, inpatient laboratory results, discharge location, etc. Additional information can include for example the costs of the treatment, medication or supplies used.
The obvious limitation of data collected in a hospital environment is that the full details are available on inpatient stay only. If the records are from an outpatient facility, there might be only partial data available.
Electronic Medical Records
Depending on the country or region and the local healthcare system, electronic medical records can be more expensive to acquire, and the process might take longer, for example 6 – 9 months.
Electronic medical records contain data used to capture and record an accurate and complete patient health record. An electronic medical record is the digital version of a paper chart.
The strength of EMRs is that they include full details of patient-healthcare professional interaction, including laboratory results, pathology, disease staging, disease severity,etc. EMRs usually contain both structured data (age, gender, location, dates, etc.) and unstructured data (treatment notes, patient-reported outcomes, lifestyle, behavior, risk factors, etc.) In some situations, the EMRs also include administrative and billing data.
If the electronic medical records contain structured data and data management practices at the source organization are good, the extraction of desired information can be relatively quick.
Unstructured data requires natural language processing (NLP) or some other advanced extraction method to put it into usable form. This area of linguistics and artificial intelligence is advancing quickly and one of the key factors for the growth of business related to NLP is the potential for structuring health data. Based on the recent developments with OpenAI’s tools, one can be hopeful that there is soon a large new dataset available from old unstructured EMRs.
Electronic medical records are often limited to the inpatient and outpatient setting. This means that EMRs may miss encounters outside the electronic medical records network. In regards to the use of medicinal products, EMRs reflect intent to prescribe rather than patients going to the pharmacy to get the prescribed medication or actually using the product.
Another downside of electronic medical records is that, unfortunately, there is little to none harmonization between the systems used to create and manage EMRs among hospitals, hospital regions, healthcare regions, cities or countries. When combining data from different areas, data cleaning and manipulation is required.
EMRs can be used to create RWE related to physician prescribing patterns, treatment effectiveness, natural history of disease and course of treatment, risk factors/predictors of disease outcomes, development of standard-of-care cohort for comparison with single-arm clinical trial, etc.
Setting up a new Registry
If there are now usable databases or registries available and clinical trials are difficult or too expensive to perform, creating a new registry is an option. However, setting up a registry is also expensive and it can take about 12–18 months to do.
Registry is an organized system that uses observational study methods to collect uniform data (clinical and other) to evaluate specified outcomes for a population defined by a particular disease, condition, or exposure, and that serves a predetermined scientific, clinical, or policy purpose.
Most important part about registry studies is that they provide more information on patient subpopulations that are not necessarily included in clinical trials and they can provide a possibility for long-term follow-up. Registry design can be customized to suit one’s needs.
This is also the main challenge: in traditional settings, the usefulness of collected data for research purposes depends on advanced planning and strong vision for the registry. Patient enrollment for the registry can be challenging depending on the disease area. The study sites may not collect all the requested data or the data quality can be low. In short, building a new registry is a lot of work, but sometimes it is worth it.