Welcome to the inaugural edition of PicnicHealth’s quarterly customer newsletter!
At PicnicHealth, we’re committed to continuously expanding our offerings to better serve you. In this newsletter, you’ll find our latest product features, research updates, and innovations designed to simplify observational research.
This year, we celebrate a major milestone — our ten-year anniversary. Reflecting on the past decade, I am inspired by how PicnicHealth has grown to support research in new ways and improve patient outcomes. Your collaboration has been crucial to our journey, and we are grateful for your continued partnership.
Please don't hesitate to reach out to your life sciences study team to discuss how we can further support your research goals. I look forward to continuing to work together.
Noga Leviner
Co-Founder and CEO at PicnicHealth
What's New at PicnicHealth
Introducing the PicnicHealth Virtual Site
Site-based approaches to observational research can be burdensome, time-consuming, and costly for sponsors, patients, and site staff. Last month, to address these challenges, we introduced the PicnicHealth Virtual Site.
PicnicHealth is the industry leader in capturing medical record data to power observational research. Now with the PicnicHealth Virtual Site, we offer faster, more cost-effective study execution while also capturing patient data beyond medical records, enabling a wider range of study designs. The PicnicHealth Virtual Site can be used for both hybrid and fully virtual studies.
Virtual Site Services
The PicnicHealth Virtual Site offers services that allow sponsors to capture data endpoints that are not captured as a part of patients’ routine care. These services enable PicnicHealth to support a broader range of research needs, including long-term follow-up and safety adjudication. These services include:
PicnicHealth’s providers can schedule virtual visits with study participants to conduct assessments required by the study protocol. Using clinical expertise, these assessments help evaluate participants' symptoms, overall health, and functional ability.
The PicnicHealth care team can order specific diagnostic tests, such as labs or imaging, if they weren't part of the patient's routine care. This ensures that sponsors have all the necessary data to address their unique research questions.
PicnicHealth’s clinical team can provide support to ensure appropriate safety reporting. This includes monitoring for safety events to support safety adjudication.
The PI of the PicnicHealth Virtual Site provides clinical oversight to ensure appropriate study conduct, including assessing whether the study is following study protocol, meeting compliance with regulatory standards and good clinical practice guidelines, collecting data accurately, and maintaining documentation and producing progress reports as required.
With the PicnicHealth Virtual Site, we are simplifying observational research study designs. To learn more, read our blog.
Visit BlogProduct Corner
Introducing LLMD, PicnicHealth’s large language model for more efficient research
Last month, PicnicHealth released a preprint on LLMD, our state of the art medical large language model (LLM). LLMD achieves human-level accuracy when extracting clinical insights from medical records, and outperforms the most powerful general and industry-specific LLMs. This model highlights how AI can realize the potential of real-world data (RWD) and offers life sciences companies a more efficient, powerful way to gain valuable clinical insights.
The use of AI in research is becoming increasingly critical as more life sciences companies recognize the value of RWD. In 2021, 96% of FDA approvals included a real-world evidence (RWE) study, up from 75% in 2019. However, working with RWD is challenging—80% of healthcare data is stored in unstructured formats within medical records, making it difficult to extract useful insights. Researchers must also analyze data from multiple patient records to form a complete, longitudinal view of the patient’s journey.
PicnicHealth’s AI simplifies this process by sifting through vast amounts of data and guiding clinical data abstractors and research coordinators to the most relevant information. This significantly streamlines the collection and organization of large volumes of patient data with human-level accuracy, while reducing the time, costs, and effort required for observational research.
LLMD outperforms other general and industry-specific LLMs because it is trained specifically to work with medical records, making connections across patient data over time and across different facilities. LLMD is trained on over 350 million clinician abstractions, covering 40+ disease models and 30 million notes from 750,000+ providers across 100,000+ care sites.
With LLMD, PicnicHealth is uniquely positioned to transform RWD into actionable clinical insights tailored to specific research needs. By providing a scalable alternative to manual data abstraction, LLMD can complete 30 person-years' worth of work in just two days, all with human-level accuracy. PicnicHealth’s advanced technology simplifies the handling of large volumes of unstructured data, allowing researchers to focus on generating insights that lead to meaningful advancements in healthcare.
To learn more about PicnicAI and how we built LLMD, contact us.
Data with Dan
Generating fit-for-purpose evidence: The value of collecting data from all sites of care
Creating a fit-for-purpose research dataset often requires stitching together clinical encounters over time, often from multiple care sites or data types.
In a recent analysis, we compare multiple sclerosis (MS) patient histories, as retrieved by PicnicHealth, with simulated datasets created from visits sourced at a single network of care (the Mayo Clinic). Results demonstrate the value of collecting data from multiple sites of care in order to effectively capture longitudinal disease history.
In comparing the simulated datasets with the data that can be collected by PicnicHealth, we asked:
- How many years of data and neurology visits are we able to capture?
- What clinical elements about the patient journey do we miss after a patient drops out of a single network?
- What doesn’t a single network “see” from before and after a patient is in that system?
To be considered for analysis, the patient must have had at least one visit in the targeted healthcare system post-diagnosis to ensure that the simulated dataset would “know” about their diagnosis and disease trajectory. Secondly, a certain proportion of inpatient/outpatient (not lab) visits must have been in the target healthcare system to ensure that patients are using that system during some of their history for care (not a one-off hospitalization or urgent care visit). Patients in the analysis were further restricted to those who have visits in only one of the target healthcare systems of interest. This ensured that estimates of in-system observation time were not inflated.
When we compared the number of median visits found by PicnicHealth data compared to our simulated single network, PicnicHealth had nine years of visits and 13 neurology visits, while the single network would have observed a median of four visits each.
If we assume that the data collected in PicnicHealth is the “gold standard,” the simulated single network would have observed only 47% of MS relapses, 81% of treatment start dates, 70% of Expanded Disability Status Scale (EDSS) measurements, and 60% of neurology magnetic resonance imaging (MRIs).
Lastly, to illustrate the impact of observing data across multiple sites versus a single care network, we displayed what data would have been lost if a patient is lost to follow-up in a traditional site-based study.
Completeness of real-world data is critical for accurate insights. Our analyses demonstrate that missing information is more likely in traditional site-based methods or analyses limited to single care networks than in the PicnicHealth methodology. This can lead to misclassification and selection bias, which will lead to incorrect insights.
PicnicHealth in the News
MedCity News
featuring Dan Drozd, MD, MSc
Applied Clinical Trials
featuring Troy Astorino
Clinical Researcher
featuring Dan Drozd, MD, MSc
Real World
MedAdNews featuring Dan Drozd, MD, MSc
The Clinical Trial Vanguard, featuring
Dan Drozd, MD, MSc
PharmaExec featuring Noga Leviner
PicnicHealth’s Latest Publications
Employee Spotlight
Colleen Goldberg, BSN, RN, OCN, Clinical Research Nurse
Please share what your role is at PicnicHealth.
In my role, I leverage my clinical expertise to support the design and implementation of products that enhance our data capture capabilities. I help evaluate new product concepts and provide clinical feedback to assess their feasibility for integration into our existing abstraction interface. Additionally, I act as a clinical subject matter expert within the organization, offering clinical insights and context to support various initiatives and decision-making processes.
Can you share your background prior to PicnicHealth?
I have seven years of specialized nursing experience in oncology, spanning a range of clinical settings from urban hospitals to rural clinics. My roles have included inpatient oncology nursing as well as various outpatient positions, such as infusion, care navigation, and clinical research. This diverse experience has deepened my understanding of the healthcare system and reinforced my commitment to improving patient care.
What does your day-to-day at PicnicHealth look like?
At PicnicHealth, I continue to pursue my passion for helping others on a broader scale. I work closely with the Data Production team, applying my clinical expertise to develop and refine product solutions that improve our data abstraction interface and workflows.
With a deep understanding of the challenges patients face in navigating the healthcare system, I take pride in being part of a company that leverages technology to create impactful solutions for improving patient lives. My current projects focus on enhancing record completeness for participants and developing strategies for abstracting complex clinical data. Through these initiatives, we are advancing our ability to collect high-quality data and provide valuable insights to our customers.
FUN FACT: Colleen is a “failed” travel nurse — she did one assignment in Boston and never left the northeast. She was born and raised in Vermont and loves it there!
In Case You Missed It
Considerations for the use of real-world evidence in non-interventional studies: Comments for the FDA
Earlier this year, the FDA published a guidance document titled “Real-World Evidence: Considerations Regarding Non-Interventional Studies for Drug and Biological Products”, that provides study design and analysis recommendations to sponsors looking to use real-world data (RWD) in observational research. This guidance, which emphasizes the potential for real-world evidence (RWE) to meet regulatory standards for evidence generation, underscores the FDA's recognition of RWE as a valuable component in drug and biological product evaluation.
This guidance is very helpful in providing clarity on how to use RWD and RWE and how it can be used for regulatory decision-making. However, based on the latest market trends and advancements in data collection, we suggest a few areas where it would be beneficial for the FDA to expand their guidance around the use of RWE. Below, we share four of our recommendations.
Embracing AI and Machine Learning in RWE Studies
With the rapid advancement of artificial intelligence (AI) and the critical role it can play in observational studies, we suggest further guidance on the use of AI in collecting data and generating RWE. Given the transformative potential of AI and machine learning (ML) in data interpretation and analysis, it is crucial for the FDA to address how these technologies can be evaluated and validated. This includes transparency in the use of AI throughout study design, conduct, and reporting, ensuring quality metrics are established to address bias and model drift.
Broadening the Definition of RWD
The volume and types of RWD have been growing, so it will be important to expand the definition of RWD to include a wider range of data types, such as primary data collected from patients, caregivers, and healthcare providers. This broader definition will better accommodate current and future data sources.
Addressing Novel Data Sources
As data collection methods evolve, it is essential to maintain flexibility in the guidance for incorporating various data sources beyond traditional claims and electronic health records (EHRs). With new direct-to-patient approaches, it is now easier to collect primary data from patients and caregivers than ever before for observational research. Because of this, it will be helpful for life sciences companies to have guidance from the FDA on how to provide transparency in data curation, particularly when utilizing innovative approaches. This will support the integration of diverse data types and enhance the robustness of RWE.
Harmonizing with Global Standards
With the rapid advancement of AI and the critical role it can play in observational studies, we suggest further guidance on the use of AI in collecting data and generating RWE. Given the transformative potential of AI and machine learning (ML) in data interpretation and analysis, it is crucial for the FDA to address how these technologies can be evaluated and validated. This includes transparency in the use of AI throughout study design, conduct, and reporting, ensuring quality metrics are established to address bias and model drift.
The latest FDA guidance provides great insights on the role and use of RWE in observational studies, and we are excited to see greater clarity on this topic. By incorporating some of these recommendations, the FDA can further enhance the flexibility, transparency, and global alignment of its guidance, ultimately improving the quality and acceptance of RWE in regulatory decision-making.