Periodic Reporting for period 1 - PREPARE (PERSONALIZED REHABILITATION VIA NOVEL AI PATIENT STRATIFICATION STRATEGIES)
Reporting period: 2023-06-01 to 2024-11-30
Health rehabilitation is defined by the World Health Organization (WHO) as “a set of interventions designed to optimize functioning and reduce disability in individuals with health conditions in interaction with their environment”. Rehabilitation is a key strategy for achieving United Nations’ Sustainable Development Goal 3: “Ensure healthy lives and promote well-being for all at all ages”.
PREPARE aims to advance rehabilitation care for patients with chronic noncommunicable diseases by developing, validating, and implementing robust, clinically relevant, and data-driven computational prediction and stratification tools. Current predictive models for these pathologies are generally based on small datasets, lack sufficient validation, or fail to predict outcomes, making it difficult for health care professionals and patients to select the optimal therapy strategy. We will combine real-world clinical datasets in a federated way, including key sociodemographic, living conditions, and behavioral information. Exploiting the latest advances in clinical, socio-behavioral and public health research, data science, and advanced statistical and AI learning methods, we will pave the way to more personalized, reliable, and holistic rehabilitation and care that considers external circumstances and patient factors to improve quality of care and life.
PREPARE has the following concrete objectives, which are categorized in three types:
The Scientific and Innovation objectives are: SO1) Define design criteria for AI prediction models, SO2) Establish a common infrastructure for assessing and analyzing multiple observational data sources, and SO3) Design automatic AI tools to convert unstructured/structured health care data.
The Technical objectives are: TO1) Develop an innovative and effective stakeholder’s liaison platform for managing models’ results, and TO2) Develop prediction and stratification machine-learning strategies for rehabilitation medical data.
The Demonstration, Dissemination & exploitation objectives are: DO1) Validate the prediction models via demonstration studies in chronic non-communicable diseases, DO2) Disseminate and communicate outcomes to raise awareness amongst the health stratification community, patients, and public; and exploit synergies with other EU projects, enhancing societal impact, and DO3) Roadmap steps towards certification approach of AI tools applicable beyond PREPARE lifetime.
The developed tools and the PREPARE platform will be evaluated and validated through 9 pilot cases for the 9 pathologies that constitute the most dominant causes for rehabilitation in Europe and worldwide. We will use existing databases at the disposal of the partners, kickstarting the development of the prediction and stratification models.
PREPARE will develop a data management platform to share results on classification of profiles and prediction of health outcomes. Clinical researchers across Europe may use these models for their own needs. Knowledge from available databases can be shared on this platform as well as re-used in other studies thus, clinical researchers will use effective health data integration solutions for classifying clinical phenotypes.
PREPARE will give the opportunity to health care professionals to use robust/validated data-driven computational tools to successfully stratify patients. This will be achieved by offering the developed classification and prediction algorithm via an online platform allowing the clinicians to use them. Furthermore, interactive tools will assist the clinicians in using the algorithms and discussing the results with the patients for advanced care planning.
PREPARE will provide a roadmap detailing appropriate steps towards MDR compliance helping in the approval of computer aided patient stratification strategies to enable personalized diagnosis and/or personalized therapy strategies by regulatory bodies.
PREPARE through the development of an online infrastructure/tools with guidelines where clinicians can interact with the prediction models will help health care professional to adopt evidence-based guidelines for stratification-based patient management superior to the standard of care.
PREPARE aims to advance rehabilitation care for patients with chronic noncommunicable diseases by developing, validating, and implementing robust, clinically relevant, and data-driven computational prediction and stratification tools. Current predictive models for these pathologies are generally based on small datasets, lack sufficient validation, or fail to predict outcomes, making it difficult for health care professionals and patients to select the optimal therapy strategy. We will combine real-world clinical datasets in a federated way, including key sociodemographic, living conditions, and behavioral information. Exploiting the latest advances in clinical, socio-behavioral and public health research, data science, and advanced statistical and AI learning methods, we will pave the way to more personalized, reliable, and holistic rehabilitation and care that considers external circumstances and patient factors to improve quality of care and life.
PREPARE has the following concrete objectives, which are categorized in three types:
The Scientific and Innovation objectives are: SO1) Define design criteria for AI prediction models, SO2) Establish a common infrastructure for assessing and analyzing multiple observational data sources, and SO3) Design automatic AI tools to convert unstructured/structured health care data.
The Technical objectives are: TO1) Develop an innovative and effective stakeholder’s liaison platform for managing models’ results, and TO2) Develop prediction and stratification machine-learning strategies for rehabilitation medical data.
The Demonstration, Dissemination & exploitation objectives are: DO1) Validate the prediction models via demonstration studies in chronic non-communicable diseases, DO2) Disseminate and communicate outcomes to raise awareness amongst the health stratification community, patients, and public; and exploit synergies with other EU projects, enhancing societal impact, and DO3) Roadmap steps towards certification approach of AI tools applicable beyond PREPARE lifetime.
The developed tools and the PREPARE platform will be evaluated and validated through 9 pilot cases for the 9 pathologies that constitute the most dominant causes for rehabilitation in Europe and worldwide. We will use existing databases at the disposal of the partners, kickstarting the development of the prediction and stratification models.
PREPARE will develop a data management platform to share results on classification of profiles and prediction of health outcomes. Clinical researchers across Europe may use these models for their own needs. Knowledge from available databases can be shared on this platform as well as re-used in other studies thus, clinical researchers will use effective health data integration solutions for classifying clinical phenotypes.
PREPARE will give the opportunity to health care professionals to use robust/validated data-driven computational tools to successfully stratify patients. This will be achieved by offering the developed classification and prediction algorithm via an online platform allowing the clinicians to use them. Furthermore, interactive tools will assist the clinicians in using the algorithms and discussing the results with the patients for advanced care planning.
PREPARE will provide a roadmap detailing appropriate steps towards MDR compliance helping in the approval of computer aided patient stratification strategies to enable personalized diagnosis and/or personalized therapy strategies by regulatory bodies.
PREPARE through the development of an online infrastructure/tools with guidelines where clinicians can interact with the prediction models will help health care professional to adopt evidence-based guidelines for stratification-based patient management superior to the standard of care.
During the first eighteen months of the PREPARE project identified the user requirements and functional specifications of the PREPARE solution. It achieved these goals by performing a state-of-the-art analysis, user requirements and gap analysis. Furthermore, PREPARE developed the early reference architecture of the PREPARE platform by defining the technological requirements and federated datasets management. PREPARE planned the pilot cases and their implementation methodology and validation. Finally, PREPARE defined the testing methodology/plan focused on the various building blocks of the system and produced an integration plan to govern the integration/adaptation and prototype testing with the aim to optimize and shorten the delivery cycle of the integrated platform.
PREPARE developed the initial version of the tools that can convert and merge unstructured/structured health care data, the initial infrastructure and tools for assessing multiple observational data sources using the OMOP CDM, and mapped the data for four clinical cases, deployed an initial version of tools though a VM to validate and test a federated learning infrastructure based on the ODHSI tools for the AI supported patient stratification, and start developing the initial tools of the PREPARE platform. PREPARE also defined the data, metrics and application architecture of the patient analysis engine for three of the clinical cases. Furthermore, a preliminary bias assessment tool was developed using dummy data to serve as a proof of concept for assessing bias in patient stratification models.
PREPARE developed the initial version of the tools that can convert and merge unstructured/structured health care data, the initial infrastructure and tools for assessing multiple observational data sources using the OMOP CDM, and mapped the data for four clinical cases, deployed an initial version of tools though a VM to validate and test a federated learning infrastructure based on the ODHSI tools for the AI supported patient stratification, and start developing the initial tools of the PREPARE platform. PREPARE also defined the data, metrics and application architecture of the patient analysis engine for three of the clinical cases. Furthermore, a preliminary bias assessment tool was developed using dummy data to serve as a proof of concept for assessing bias in patient stratification models.
The PREPARE data platform will have the ability to manage big data through a federated access / learning infrastructure. It will also develop AI driven tools that will convert unstructured health care and socio behavioral data to a digital form using language models, AI supported stratification by extending the OHDSI cohort definition and predictive tools, and AI clinically relevant rehabilitation outcomes prediction models. All the above will highly contribute to clinical researchers using effective health data integration solutions for classifying clinical phenotypes, health care professional using robust/validated data drive computational tools to successfully stratify patients and adopt evidence-based guidelines for stratification based patient management superior to the standard of care.
Finally, the development of a roadmap including the relevant steps through the MDR certification of the PREPARE platform as well as similar platforms in other medical fields will highly contribute to regulatory bodies approving computer aided patient stratification strategies to enable personalized diagnosis and/or personalize therapy strategies.
Finally, the development of a roadmap including the relevant steps through the MDR certification of the PREPARE platform as well as similar platforms in other medical fields will highly contribute to regulatory bodies approving computer aided patient stratification strategies to enable personalized diagnosis and/or personalize therapy strategies.