There are several sources of real-world data, and payers should identify the most pertinent questions about a disease state or population in order to use each of them appropriately, according to speakers at AMCP 2022.

In 2016, the 21^st Century Cures Act placed new emphasis on real-world data for regulatory decision-making. Additionally, during the pandemic, real-world data was used to evaluate new COVID-19 therapies coming to market and identify patients for enrollment in clinical trials.

“Life science companies can use this [data] to get innovations out faster by not having to always do randomized controlled trials,” said Myla Maloney, MBA, BCMAS, chief commercial officer, Premier Inc.

Real-world data are derived from several places including electronic health records, claims and billing activities, product and disease registries, patient-generated data, and other sources such as mobile devices, Ms Maloney said.

These data are used for a variety of purposes, including differentiating real-world outcomes from those in highly controlled trials, recruiting patients for clinical trials, making tools and algorithms that support decision-making, and guiding value-based contracts.

“When you’re evaluating a real-world evidence data source, you have to inquire about their HIPAA strategy,” Ms Maloney said. “[It’s] important to make sure we’re always protecting patient information that should remain private.”

Identifying information can include but is not limited to names, geographic data, email addresses, account numbers, web URLs, health plan beneficiary numbers, full face photos, and vehicle identifiers and serial numbers. To access and use patient data, stakeholders must deidentify the information.

Ms Maloney reviewed 2 ways of deidentifying data. The safe harbor method is used to remove all identifiers from the data, but she cautioned this method may eliminate data that could be useful for analysis.

Statistical deidentification is the preferred method, Ms Maloney said. This method allows for data to retain some HIPAA identifiers, with an expert certifying there is a very low statistical risk that a patient could actually be identified with the remaining information.

Sources of Real-World Data

In an administrative billing or chargemaster database, data is sourced from hospital cost accounting and billing systems. The information includes all inpatient and emergency room visits in the acute setting, as well as most outpatient visits when patients have same day surgeries or seek care in hospital clinics.

Administrative billing/chargemaster databases are beneficial for defining utilization in combination with detailed cost data. However, this type of database is not useful for understanding the full patient journey, and stakeholders also gain no insight into mortality that occurs outside of the health system.

The use of claims data, on the other hand, allows payers to view the patient’s journey across a spectrum of health care visits. Stakeholders can understand drug utilization rates and view data from all medical visits.

Claims databases also have limitations, however. They lack information on specific device utilization and the full extent of utilization for patients admitted to hospitals, and can cause delays in data collection.

Electronic health records (EHRs) are another rich source of data, said Jim Sianis, BS, PharmD, MBA, senior director of real-world evidence, Premier Inc. EHRs provide detailed patient information not available from other sources, as well as data on treatments and related outcomes. But EHR data is mostly unstructured and requires normalization.

Registry data are collected from several sources for a specific patient population. These data can include clinician- and patient-reported outcomes and external sources like claims, pharmacy and lab databases, and output from medical devices.

“What’s unique about a registry dataset…is that a lot of the information is predetermined. They’re going to gather information from a lot of different sources,” Dr Sianis said.

Data from registries use structured and predetermined elements and offer longitudinal information about a defined population. However, registry data may only be focused on 1 specific disease or condition, which limits information availability on comorbidities and may not be a true representation of a population if the registry primarily contains patients with more severe disease.

Dr Sianis also reviewed patient-generated data, which is gathered through wearables, surveys, smart phones, social media, and other similar sources. These data are comprised of daily steps or physical activity, weight measurements, sleep patterns, vital signs like heart rate or blood pressure, and patient-reported data on pain and mood.

Patient-generated data are gathered in real-time and allow patients to track their own health, but it may be difficult to integrate these data into current systems being used to track and analyze clinical data, and patient privacy is also a large concern.

Selecting the Appropriate Data Source

There is no such thing as a perfect dataset, Dr Sianis said. But tokenization makes data linking possible, giving stakeholders the opportunity to use richer combinations of data sources to answer complex questions while making formulary and contracting decisions.

“[Tokenization] allows you to link disparate datasets and maintain patient confidentiality. It’s really opened up some exciting opportunities from a research perspective,” Dr Sianis said.

Real-world data can be used to understand outcomes in an episodic or continuous sense, according to James Kalus, PharmD, FASHP, director of pharmacy, Henry Ford Health System. Episodic data are used to identify medication use opportunities and support quality improvement projects, while continuous data can be used to create quality and workload metric dashboards or understand medication safety measures.

Deidentified benchmarking data, including drug utilization data, drug spend data, and patient data, can be used to identify opportunities for interventions, such as those aimed at improving cost savings and adherence.

While choosing which data sources to use, payers should identify the most appropriate questions about populations and treatments, seek databases that are most relevant to those questions, and ensure data are from reliable and valid sources, Dr Sianis said.