Whitepaper

OM1’s Data Automation in Prospective Studies and Registries

OM1's Data Automation in Prospective Studies and Registries

Research White Paper | February 2025

OM1, Inc. 2112 James Ave Suite #1210 Boston MA, 02116 +1 (888) 224 2666 info@om1.com www.om1.com

Introduction: Addressing the Challenges of Traditional Clinical Research

Clinical research, particularly post-marketing and safety studies, has traditionally been a labor-intensive endeavor, often burdened by manual effort, high costs, long timelines, and low satisfaction among sites and patients.

The 21st Century Cures Act, signed into law in 2016, aimed to accelerate medical product development and bring innovations to patients faster. In response, the FDA established a framework for a Real-World Evidence (RWE) Program to evaluate the use of RWE to support new drug indications and post-approval study requirements. This legislative context has created the growing need for efficient and high-quality real-world data (RWD) for regulatory decision-making. The challenge lies in transforming the traditional, often inefficient research paradigm to meet these demands.

This booklet outlines how OM1's innovative approach, leveraging data automation and Artificial Intelligence (AI), addresses these fundamental problems by streamlining data collection and processing while maintaining traceability and auditability, ensuring regulatory compliance, and enabling more cost-effective, insightful, and patient-centric prospective studies and registries.

OM1's Transformative Approach: Automation and AI

OM1's strategy fundamentally shifts the paradigm of prospective research by employing automation to make studies more predictable and realistic, reduce marginal costs, and achieve massive cumulative savings. This is achieved by combining passive data collection (using existing systems of record) with active data collection (for data not routinely captured).

Passive Data Collection

OM1 emphasizes the direct, consented retrieval of Electronic Health Record (EHR) data. OM1 acts as a primary source of RWD, containing a patient's medical history, diagnoses, treatment plans, immunization dates, allergies, radiology images, pharmacy records, and laboratory results. OM1 connects to central EMRs and ancillary systems of record, such as Laboratory Information Management Systems (LIS), Radiology Information Management Systems (RIMS), and Pathology Systems. Additional linked data sources include claims data including medical and pharmacy claims, social determinants of health (SDOH), and mortality data may also be obtained by leveraging common token systems. This extensive data linkage increases the breadth and depth of information on individual patients over time.

Active Data Collection

When essential data elements are not routinely collected through passive or ancillary data sources, OM1 employs an active data collection to the minimize the burden, making studies and registries simpler and more cost-effective for both sites and patients:

Electronic Clinical Outcome Assessments (eCOAs)
Electronic Patient-Reported Outcomes (ePROs)
Specialized clinical and laboratory assessments
Virtual follow-up assessments through a virtual or telehealth maintained by an OM1 Virtual Center

In other words, the most efficient and least burdensome model for data collection in prospective studies and registries leverages passive data collection of existing data sources (rather than having sites re-enter data into OM1 directly interfacing with health records) and AI-based automated unstructured and integrated active/structured data collection for those data elements not routinely captured; unless a site uses an existing EHR-integrated ePROs or specialized tests. In use cases of rare diseases where a specialized facility is involved, data can be collected through a virtual center where a patient provides consent for OM1 to retrieve their data from any healthcare facility that may visit (or follow) improving this model.

Leveraging AI for Data Automation

AI plays a crucial role in enhancing data completeness and study efficiency.

OM1 uses AI to map and process structured data in multiple formats from any number of clinical centers with different EHR systems. This is technically very difficult as having a common data model with common terminology must be maintained as OM1 suggest to maintain traceability. AI also aids in quality control of these complex data pipelines.

OM1 extensively leverages AI-enabled and clinician-validated text extraction to extract data from unstructured sources like physician notes, images, and radiology/pathology reports, converting them into structured data. This automates collection of the significant amount of key clinical data often residing in unstructured formats and quality that would typically require manual abstraction by a clinical site.

Digital Phenotyping (PhenOM®), powered by proprietary AI models, automatically mines clinical data, including EMR and specialized third party laboratory RWD data to identify patient phenotypes and disease states. This study fully leverages automated data collection and processing of unstructured clinical management, lab reports, and outcomes data.

Meeting Regulatory Standards with Automated Studies and Registries

The FDA's RWE Program requires that RWD be "fit-for-use" in regulatory decision-making, emphasizing relevance and reliability. Relevance pertains to the availability of data for key study variables (exposure, outcomes, covariates) and sufficient numbers of representative patients. Reliability encompasses accuracy, completeness, provenance, and traceability. OM1's automated study and registry programs are designed to meet these stringent requirements:

Data Quality and Standardization: OM1-managed studies collect structured and predefined data elements, offering longitudinal, curated data about defined patient populations. The use of common data elements promotes standardization, using industry-standard data extraction, facilitating comparison and linkage with other sources. This approach ensures conformance with FDA requirements for submitting study data in applicable drug submissions.

Curation and Transformation: Data curation applies standards to source data, such as coding for adverse events or disease progression. Automated processes for data transformation, extraction, validation, and integration into a Common Data Model (CDM) are meticulously documented, including justifications for approaches used to reconcile challenges like inconsistent coding, changes in code lists (e.g., ICD-9 to ICD-10 codes), or missing information. This documentation is crucial for FDA review.

Traceability and Auditability: OM1's systems maintain traceability of data from analysis results back to source data (source records, often with electronic signatures), which is vital for verification during FDA inspections.

Prospective Planning and Consultation: Sponsors are encouraged to consult with the appropriate FDA review division early in the process and submit protocols and statistical analysis plans before conducting studies that include registry data. This pre-planning process ensures that the registry's design, including target population definition, data elements, linkage strategies, and outcome validation methods, aligns with regulatory expectations.

Cost Efficiency and Diverse Purposes of Registries

Cost Reduction Curve Based on OM1 Studies

One of the most significant advantages of data automation is its impact on cost. As shown in OM1's studies, the cost per subject decreases exponentially as the study scales up using data automation. This efficiency is a direct result of automating processes that traditionally require manual effort, such as file processing and data integration. For example, studies with hundreds of thousands or a million subjects become far more cost-effective with automation compared to traditional methods.

**Figure 1** below shows a curve, comparing total number of subjects in OM1 studies versus the average cost per subject. Study sizes range from 1500 patients to more than one million. As shown, as the number of patients of subjects increases, the cost costs per subject reduces dramatically. All numbers below 1000 patients, costs are closer to traditional methods but site satisfaction is highest as manual effort is still a major component of investigator site burden. In these smaller studies, OM1 achieves these cost efficiencies by leveraging its automation approach to streaming medical records as well as supporting active data collection where feasible.

Figure 1: Cost-Efficiency of Data Automation

Chart description: A scatter plot with a downward-sloping curve showing the relationship between the number of subjects per study (x-axis, labeled "# SUBJECTS PER STUDY", with tick marks at 0, 200,000, 400,000, 600,000, 800,000, and 1,000,000) and the total cost per subject in dollars (y-axis, labeled "Total Cost Per Subject ($)", ranging from $0 to approximately $1,800). The chart demonstrates that as the number of subjects increases, the cost per subject decreases dramatically. Title: "Cost-Efficiency of Data Automation" with subtitle "Total Cost Per Subject vs # Subjects." An annotation box states: "Cost per subject decreases exponentially as the study is scaled up using data automation."

At ~1,500 subjects: cost per subject approximately $1,400–$1,800
As subjects increase to ~200,000: cost drops significantly toward ~$200
At ~400,000 subjects: cost per subject approaching very low levels (~$100 or below)
At ~1,000,000+ subjects: cost per subject near minimal/near zero

The second advantage of reducing manual effort is a significant increase in site satisfaction with sites participating in automated studies and registries. OM1 measures Net Promoter Score (NPS) in studies to asses site research coordinators. As shown in some of the case examples below, OM1 automated registries and studies have NPS typically above 70 which indicates an extremely high satisfaction level.

Diverse Purposes of Prospective Studies and Registries

Programs managed by OM1 leveraging automation and AI serve multiple critical purposes throughout the medical product lifecycle, from early development to post-market surveillance. OM1's automated approach enhances the ability for all these applications.

Safety: Post-approval safety studies and registries are invaluable for measuring or monitoring safety and harm associated with specific products and treatments, including conducting comparative assessments of safety. They can systematically collect data on adverse events and their incidence, addressing limitations of spontaneous reporting systems. For instance, a registry can evaluate safety signals identified from other sources or assess factors affecting risk like dose and timing.

Effectiveness: Post-market studies and registries help determine clinical effectiveness or cost effectiveness in real-world clinical practice. They can link subjects to evaluate drugs received during routine medical practice and provide information on long-term efficacy outcomes. They can address gaps in generalizability from clinical trials by including more diverse and representative patient populations (e.g., older patients, different ethnicities, those with comorbidities) who might be underrepresented in traditional trials.

Natural History Studies: They are common platforms for natural history studies aimed at understanding disease progression, identifying demographic, genetic, environmental, and treatment variables that correlate with disease development and outcomes.

Supporting Clinical Trials and Regulatory Needs: OM1 provides these capabilities for external controls for interventional trials, to support or satisfy post-approval study requirements, or to provide a framework for trials embedded within registries. They also enable a cost-effective approach to compare follow-up studies, including for small populations (e.g., rare diseases) with long data streams (e.g. gene and cell therapies). Some sponsors are beginning to explore using OM1 systems for automated collection of passive data as a routine part of randomized clinical trials to reduce manual entry and costs.

OM1 Key Use Cases and Accomplishments

OM1's data automation and AI capabilities support a wide range of use cases, demonstrating significant advantages over traditional methods. The following are examples from OM1 experience over the last few years:

1) A very large, multi-center real world evidence study for a new label indication in oncology diagnostics: OM1 has processed structured and unstructured data from 32 major health systems with over 920,000 cases now submitted to the Food and Drug Administration for a new label indication. The processing of these cases leverages 37 AI algorithms for processing unstructured data elements that were validated as part of the submission including clinical history, biopsy results and so forth. The data underwent 19 post-submission review steps by FDA as the first step in submission and slated to be further reviewed by FDA.

Large scale comparative study for cancer screening test for label expansion

Challenges:

Retrospective collection of existing data to evaluate a cancer screening test from a limited number of major health systems to 32 health systems
Collect and integrate data from disparate sources in multiple formats (EHR, pathology reports, radiology systems, laboratory reports, medical histories, diagnoses, comorbidities, oncology criteria)
Evaluate the change in performance against expected findings from the label

Solution:

Large, nationwide cohort of patients receiving test submitted for label expansion to the FDA
Rapidly matching of data as well as collection of data across 32 health systems
Reference validation package

2) A large, multi-center prospective comparative effectiveness and safety study evaluating a neurological disease in 10,000 patients eligible for infusion therapy: OM1 is collecting EMR data (including imaging reports and clinical notes) and patient reported outcomes data about neurological testing (using correlational values) with five years follow-up. Patients are enrolled or referred by specialists and over 50% are followed through an OM1 virtual center.

3) A multi-center, prospective study leveraging 30+ centers and 10,000+ patients to inform 'when' OM1 central site is responsible for clinical outcomes assessments, PRO, and EMR data collection and analysis. This study specifically evaluates when to use central vs. site-based collection for clinical outcomes assessments, PROs, and EMR data collection, and prospectively tracks a dementia condition where patients need to be evaluated prior to the initiation of treatment.

4) A multi-center, retrospective and prospective registry in multiple sclerosis where data is collected using automation from EMRs, clinicians and patients including diary and ePROs. Both clinicians and patients access their own dashboard with summary and data streaming information.

5) A multi-center retrospective comparative effectiveness study of a pharmaceutical product used in post-transplant with approximately 5,000 patients. OM1 is collecting data, including EMR and specialized third party laboratory RWD data (e.g. rare diseases) to produce registry reports. This study fully leverages automated data collection and processing of unstructured clinical management, lab reports, and outcomes data.

6) A multi-center retrospective and prospective study identifying and following ~100 patients with a rare pulmonary disease (cystic fibrosis), including obtaining genomic data and clinical narratives for voice application.

Describing burden of disease in cystic fibrosis

Challenges:

CF is associated with a significantly lower quality of life and a high disease burden
Evaluate the change in patient-reported and clinician-reported disease burden after initiating CF treatment

Solution:

Conducted an observational study to evaluate the real-world effectiveness of treatment options and their impact on quality of life
PROs and Caregiver Reporter Outcomes were collected every 3 months
Quantified select clinical outcomes after initiating 27 treatment options

Figure data — study design elements:

Targeting ~100 subjects
Retrospective & Prospective
3 study

7) A large, multi-center FDA-mandated post-marketing commitment (PMC) study following 50,000 patients tested for colorectal disease. Data from EMRs and lab/or values collected by OM1 is submitted quarterly to the FDA. Site satisfaction is very high.

Support past market commitments and expand label for a lower age group

Challenges:

Accessing, processing and extracting complex and unstructured data from multiple health systems and diagnostic labs. OM1 created a real-world data operations foundation for analysis. OM1 is evaluated for assessing the effectiveness of a screening device for colorectal cancer across the population.

Method:

We are conducting an observational, prospective multicenter real world data monitoring
Collecting patient pathology and cytology, and diagnostic data
From integrated Systems Networks (IDNs), hospital systems and large medical practices
OM1 submits quarterly reports on an expected indicator

Figure data — process flow elements:

400,000 results collected
~5 years of follow-up
~8 months until first FDA submission

8) A multi-center breast cancer screening registry leveraging more than 40 facilities and one million patients for collection with retrospective and prospective data collection. Data is collected from electronic health records, case management systems, and cancer registries with 8 years of follow-up. The study has focused on effectiveness of population-based screening and publications from it have directly impacted guidelines.

9) A pilot study using digital phenotyping to identify patients with a rare disease for study inclusion. The aim of combining EMR records with an automated health system are currently collecting data on thousands of patients with screening at high risk for specific lysosomal storage disease and recruited for enrollment into the study.

Conclusion

The landscape of clinical research is undergoing a fundamental transformation, driven by the imperative for more efficient and reliable evidence generation, particularly from real-world sources. The traditional, manual approaches are no longer sustainable for meeting the demands of modern medicine and regulatory requirements.

OM1's approach, integrating data automation and AI, offers a robust solution by dramatically reducing manual effort and costs, enabling scalable and high-quality data collection. By seamlessly combining passive data from existing systems with active, patient-centric data collection like PROs and eCOAs, these automated studies and registries generate rich, longitudinal datasets. This comprehensive data capture, coupled with stringent quality control and documentation practices, ensures that the data meets the relevance and reliability standards required for FDA regulatory submissions. The demonstrated exponential reduction in cost per subject as studies scale up further underscores the economic viability of this new paradigm.

Ultimately, this revolution in prospective studies and registries means a future where every clinical trial can have a low-cost, real-world, long-term follow-up study, leading to improved time to data and enhanced patient outcomes. The continued evolution of data automation and AI will further solidify their role as indispensable tools in advancing medical product development and patient care.