APHL 2022 POSTER ABSTRACTS
Informatics testing based on authoritative data in Protect ( see https :// bit . ly / 3rvapq7 )
• The CDC can dynamically map new LOINC codes for new COVID-19 tests to capture the full range of COVID-19 testing data , such that changing LOINC codes do not result in data gaps
• The lessons the US Government has learned , which can be applied to modernize laboratory systems and data and analytics for public health response , include :
• Standard pipelines ( including connectors and health checks ) significantly reduce reporting burden by making it easy for new labs to onboard and share data with the US Government
• With granular-level permissions , data at the facility , local , state , and national level can be made available to authorized users ( e . g ., federal users ) within the bounds of a rigorous privacy framework
• Feedback loops and direct engagement with individual labs is critical to maintaining data quality and trust over time
• Out-of-the-box analytics and dashboards directly on top of highquality shared data enable rapid reporting of information to senior leaders and other stakeholders
In the future , the US Government and its partners can use such commercial tools to collect new types of laboratory and other data with minimal effort and in a timely manner to accelerate crossgovernment collaboration on emerging threats .
Presenter : John Napoli , Palantir Technologies , jnapoli @ palantir . com
Anomaly Detection for Public Health Electronic Lab Reporting ( ELR )
J Hightower , Ruvos
A major issue with any source of data is an awareness of quality issues , how they may impact models , and the subsequent impact on decision-making strategies . Under the stress of COVID-19 Public Health Agencies are experiencing a rise in data quality issues including batch sending records , delays in ELR delivery , and missing records . The key problem is that public health agencies are trying to prioritize very limited resources in containing the COVID-19 pandemic , while maintaining regular epidemiological day-to-day operations , on poor quality data . One of the solutions Ruvos has developed through the Integrity Platform™ is Anomaly Detection notification and reporting for data ingestion of ELR to state departments of health . Through this cloud-based solution pipeline data are ingested directly from labs , filtered through a validation protocol , and are exposed to a real-time analytics model to predict if the inbound ELR message count exceeds safe thresholds determined by the anomaly detection engine . Several Anomaly Detection algorithms were tested and weighted based on identifying crucial historical anomaly markers , flexibility , scalability , and computation cost . The chosen model is an AutoRegressive Integrated Moving Average ( ARIMA ) trained uniquely per data sender ( i . e ., lab , hospital , clinic ) and updated daily . The training mechanism has been automated , including hyperparameter fitting , such that the users of Anomaly Detection never have to manually adjust settings as the model auto updates daily . Unique models are trained per data sender with an output of several threshold bands that indicate high or low quantity volume anomalies . As these thresholds are crossed , notifications and visualizations are immediately generated to allow end users to identify problematic senders and irregular data volume . One of the major advantages of this system is how anomaly detection is allowing for a paradigm shift from reactive to proactive response to poor data quality . We have found a strong positive correlation between a critically low daily ELR volume preceding an extremely high volume in the subsequent days . This Anomaly Detection identification and reporting mechanism allows users to get ahead of the error before it impacts the community as a whole in the form of false spikes or dips in daily cases and the subsequent response from policy makers . The second advantage of Anomaly Detection is the ability to be an early detection mechanism for true , unexpected surges in cases .
Presenter : Jake Hightower , Ruvos , jhightower @ ruvos . com
Expanding Participant Engagement and Equity in Clinical Trials through Improved Communication , Consent , and Patient-reported Data
D Bhadra 1 , M Ravelo 2 ; 1 The MITRE Corporation , 2 Ruvos
The purpose of the Alliance Participant Engagement Portal ( PEP ) initiative is to enable clinical trial teams to better engage with patients who participate in oncology clinical trials , with the aim to improve outreach and longitudinal connection to trial participants independent of participant or trial site resources . The initiative will adapt the capabilities of the Sara Alert TM system to enhance the clinical trial experience for trial participants by 1 ) connecting participants to trial details , consent , and trial updates ; collecting patient-reported demographic data and social determinants of health to give researchers insight into equity in trials and allow adaptation to improve equity and inclusion ; and 3 ) enabling future collection of additional patient-reported information . The goal of the effort is for participants to have a unified , more informed experience during the trial and to inform broader and more equitable participation in current and future trials . The PEP initiative is a partnership , led by Alliance Data Innovation Lab with technology and infrastructure support by MITRE and Ruvos .
PEP ’ s key components include the following .
• Anteroom – Website that contains information on the trial that is relevant to trial participants . Anteroom data includes but is not limited to trial purpose , description of required elements , research team , trial updates , initial results , etc .
• Participant Engagement Management – Web-based tool that allows clinical trial teams to communicate directly with the participant community and provides tools to manage patient responses .
• Participant Surveys – Surveys via text , email , web , phone that enable participants to share self-reported data .
• Dashboard and Reporting – Capability to view dashboards at the site or trial level and report / export data for further analysis .
This project is in the early stages of planning and development , and with launch targeted for 2022 .
Presenter : Diane Bhadra , The MITRE Corporation and Manny Ravelo , Ruvos , dbhadra @ mitre . org & mravelo @ ruvos . com