Data with Dan

‍

High-Density Data: 
The Foundation for Minimizing Bias in Non-Interventional Research

‍

Non-interventional studies can answer meaningful, complex research questions, but the fragmented U.S. healthcare system creates challenges for capturing the data needed to generate reliable evidence. Methods that rely on real-world data from siloed sources often result in incomplete clinical datasets with missingness that introduces biased research outcomes. This is evident in site-based chart abstractions, which are limited to care received within that site or network, and aggregated EHR datasets, which are anchored in a specific environment with only contributing partners. As a result, these datasets provide limited visibility into all points of care along the patient journey that occur outside of those systems.

‍

Advances in data retrieval technology are now making it possible to capture more complete patient journeys that address these limitations. Once just a technical consideration, the method of data retrieval has become the foundation for generating reliable real-world evidence based on curated, comprehensive data.

‍

‍

Record Availability & Journey Representation

‍

Patient-mediated medical record retrieval and multi-site data retrieval are two methods enabling more complete patient journey capture. Patient-mediated retrieval allows researchers to request medical records on behalf of the patient with their consent, without adding burden to them. Multi-site automated retrieval continuously gathers information from across all healthcare encounters by pulling from both electronic data networks, such as health information exchanges, and direct-to-facility record requests, regardless of EHR vendor or payer coverage. 



‍

When used in combination, these methods yield greater visit density for every patient, resulting in greater data density and breadth. Both are critical dimensions of data completeness, ensuring that complete data on key exposures, outcomes, and covariates are captured from the full spectrum of a patient’s healthcare interactions.

‍
‍

Greater Data Density Starts With Greater Visit Density


‍

However, not all “advanced data retrieval” methods produce the same degree of data completeness. Compared to a large health data platform company, PicnicHealth’s advanced, best-in-class retrieval method demonstrates greater visit density with respect to time, frequency, and types of routine care settings. This provides a more longitudinal, continuous, and complete medical history for each patient enrolled in the study.

Figure 1: Compares PicnicHealth’s data retrieval to a large health data company’s retrieval for patients with paroxysmal nocturnal hemoglobinuria

Reducing Bias and Confounding Through Completeness


‍

More complete data reduces the risk of selection bias, information bias, attrition bias, and confounding, all of which can undermine the validity of non-interventional research.

‍

Selection bias often arises when data is only collected from specific sites or EHR systems, limiting the diversity of the patient sample. For example, if medical records are only pulled from dermatology clinics, researchers may overrepresent severe cases while missing patients with milder conditions who are treated by primary care providers. Advanced retrieval methods that capture records across multiple care settings ensure a more representative patient population and improve the generalizability of findings.

‍

Information bias occurs when exposures or outcomes are misclassified due to missing data. This may happen when critical study variables, such as treatments, diagnostic biopsy, or emergency room visits, are recorded across multiple provider records at different timepoints. Automated, longitudinal retrieval mitigates this by:


‍

• Continuously collecting data from every provider a patient sees
• Validating and cross-referencing information to ensure accuracy critical to determine eligibility
• Adjusting for risk factors and confounders
• Ascertaining whether or not a clinical outcome occurred

‍

Attrition bias and residual confounding are also mitigated as more complete longitudinal data provides continuity and fills in missing variables. Learn more about how attrition bias is reduced and confounding is addressed in the full whitepaper “Reducing bias in non-interventional research with greater visit and data density.”


‍

‍

Strengthening Robust Non-Interventional Research

‍

Greater visit density, achieved through advanced medical record retrieval that combines patient-mediated and comprehensive multi-site approaches, enables researchers to collect more complete and reliable datasets. This, in turn, improves study design through stronger baseline adjustments, exposure classification, and outcome ascertainment while reducing the risk of bias and missingness that can compromise non-interventional research.