No more Feasibili…tease OR how we do Protocol Feasibility.
One of the most important things for the implementation of a clinical trial is a feasibility study. A feasibility study is an analysis of how successfully a project can be completed, accounting for various factors that could affect the outcome. It answers the question “What is needed for this clinical trial to be successful?” This helps determine recruitment, budgets, and timelines needed to implement the trial.
Setting realistic expectations for the prevalence of a disease is one of the corner stones of feasibility. One of the most common ways an epidemiologist estimates the availability of eligible patients is by taking prevalence rates of the disease and applying it to the general population. The problem is that a general population prevalence rate is not targeted and may lead to an overestimation of the eligible population. Additionally, using any further rates (medication, procedure, or demographic rates etc.) on the estimated eligible population can be misleading. A more targeted approach is by using current EHR data to find patients with relevant diagnosis codes, medications, labs, and other needed criteria.
Diagnosis classifies a patient as having or not having a particular disease or illness. Traditionally, a diagnosis was regarded as the primary guide to treatment, and overlying condition (“what is the likely outcome”), and is still considered the core component of clinical practice. Initiatives such as “meaningful use” and the introduction of more specific coding types, such as ICD-10 and SNOMED, help to regulate diagnoses. When using EHR queries for disease prevalence, the medication, procedure, and/or demographic criteria can be applied to the eligible patient population for a more accurate eligibility rate.
ePatientFinder uses a sample size of over 42 million de-identified patients to run protocol feasibilities. We take inclusion criteria as well as exclusion criteria and through a series of analyses come up with an estimated eligible patient population. We problem solve how to find the most accurate eligibility rate, as well as recommendations on how to strategically broaden the scope of the analysis to include patients that might not have the exact diagnosis. For example, we found that many doctors document the symptoms of a disease rather than the underlying condition. For one clinical trial for arthritis of the knee, we looked at the broader diagnosis of knee pain to see if there was any indication of correlation; using a survey and chart review, we found that of those with a more general diagnosis of “knee pain,” roughly 42% had arthritis of the knee.
Using EHR data to determine an eligible population cuts out the guesswork of how diagnosis, medication, procedure, and demographic rates will interact with each other. For an implementation projection, this process develops a much clearer picture of the eligible population which can define more efficient ways to target patient recruitment and resources.
About the Author:
Ashton Dixon is a Data Scientist at ePatientFinder. With over seven years in research and public health, she has a passion for understanding the whole picture medical data creates. She holds a Master’s degree in Epidemiology with a Biostatics minor from The University of Texas Health and Science Center. She also rock climbs, kayaks, and creates woodburns and watercolors.