FOA Title: 
Notice of Special Interest (NOSI): Computational and Statistical Methods to Enhance Discovery from Health Data
Grant Type: 
NOT-LM-19-003
Primary IC: 
NLM
Release Date: 
Mar 19 2019
Expiration Date: 
N/A
AC Source: 
N/A
Purpose: 
Notice Special Interest NOSI): Computational Statistical Methods Enhance Discovery Health Data Notice Number: NOT-LM-19-003 Key Dates Release Date: March 19, 2019 Related Announcements PAR-18-896 Issued National Library Medicine NLM) Purpose National Library Medicine issuing Notice highlight interest receiving grant applications through NLM Research Grants Biomedical Informatics Data Science R01 Clinical Trial Optional) PAR 18-896), focused research reduce mitigate gaps errors health data sets. Background Recent successes the of data-centric artificial intelligence AI) methods such deep learning stimulating interest the promise harnessing large complex digital health data sets advance goals precision medicine. Applying AI methods large health data sets promises provide new powers discovery, diagnosis, prediction, decision support aimed improving health outcomes reducing healthcare costs. Numerous public datasets human non-human data available, a rich array specialized tools platforms be used studies applications. However, recent work identifying addressing systematic biases blind spots data, in AI systems derived that data, highlighted array potential problems fairness, accuracy, safety, reproducibility inferences conclusions. Work bias incompleteness health data sets includes studies find poor representation minority groups, seniors, women. See, example, https://www.eurekalert.org/pub_releases/2016-10/uoms-nsr100716.php, or https://datasociety.net/output/fairness-in-precision-medicine/?utm_sourc...). recent Wall Street Journal article https://www.wsj.com/articles/a-crucial-step-for-avoiding-ai-disasters-11...) noted computational tools developed a diverse team help avoid bias algorithms. Beyond problems biases other gaps data, research using health data humans requires special care protect sources the data see https://www.ncbi.nlm.nih.gov/pubmed?term=Barocas%2C%20Solon%5BAuthor%5D ). All Us Research Program https://allofus.nih.gov/) aims develop unbiased, representative health data resource, there many health data sets already use being constructed. Tools developed using biased incomplete data sets contribute erroneous analyses. Statistical fallacies representational errors unrelated the research question hand introduce systematic errors. core questions understanding mitigating and problems health data research are: can done, computationally and/or statistically, reduce mitigate gaps errors data sets used health research?" and, can improve tools used discovery, understanding, visualization health data sets their analyses?" Whether problem due incomplete health data inadequate tools, approaches needed strengthen reproducibility applicability data-centered research the etiology, epidemiology treatment health conditions. Research Objectives NLM invites research grant applications propose state the art methods approaches address problems large health data sets tools used analyze them, whether data drawn electronic health records public health data sets, biomedical imaging, omics repositories other biomedical social/behavioral data sets. Areas interest include are limited 1) developing testing computational statistical approaches applied large and/or merged health data sets holding human non-human data, a focus understanding characterizing gaps, errors, biases, other limitations the data inferences based the data; 2) exploring approaches correcting biases compensating missing data, including introduction debiasing techniques policies the of synthetic data; 3) testing new statistical algorithms other computational approaches strengthen research designs use specific types biomedical social/behavioral data; 4) generating metadata adequately characterizes data, including provenance, intended use, processes which was collected verified; 5) improving approaches integrating, mining, analyzing health data preserve confidentiality, accuracy, completeness overall security the data. Applicants should address ethical issues might arise their proposed approach. Application Submission Information Applications response this Notice must submitted through NLM’s funding opportunity announcement, PAR-18-896: NLM Research Grants Biomedical Informatics Data Science R01 Clinical Trial Optional). instructions PAR-18-896 must followed. Submissions should indicate they in response NOT-LM-19-003 Field 4.b the SF 424 R&R form. Program Directors/Principal Investigators PDs/PIs) planning submit applications this topic strongly encouraged contact scientific contact listed this Notice advice the appropriateness a potential application alignment NLM’s program priorities. Inquiries Please direct inquiries to: Alan Vanbiervliet, PhDNational Library Medicine/Extramural ProgramsTelephone: 301-594-4882Email: alan.vanbiervliet@nih.gov