The Grand Challenges India sixth thematic call was announced on 3rd July 2018 on ‘“ki data Challenge for Maternal and Child health- Data Science Approaches to Improve Maternal and Child Health in India”, for 45 days to support projects focused on data science approaches, to address challenges that are faced in improving the health of mother and child in India and comparable geographies.
This unique call diversified the scope of data analytics which can be explored to improve the lives of mother and child and is consequent to the Bill & Melinda Gates Foundation’s Healthy Birth, Growth, and Development Knowledge Integration (HBGDki) initiative, India. HBGDki-India seeks to develop a deeper understanding of the risk factors contributing to poor maternal and child health outcomes with a focus on reducing the global burden associated with three complex and interrelated outcomes: Preterm birth, physical growth faltering, and impaired neurocognitive development.
The overall goal of the ki data challenge program is to support innovative data analytics and modeling approaches applied to ki India data repository or other relevant existing health and social data sets arising from multiple sources that applicants can access to accelerate maternal and child health research and develop improved solutions and help to formulate public health recommendations that are data-driven and cost-effective.
The mandate also encouraged engaging a broad spectrum of collaborators - including research and clinical scientists working with data scientists, bioinformaticians, statisticians, epidemiologists, engineers, and computer programmers.
This program sought proposals that encouraged solutions that translated leading and best practices and solutions developed and validated by data science approaches required to be potentially translatable, results from which would be used to inform policy decisions related to maternal and child health as well as design subsequent related challenges
This Request for Proposals (RFP) was specific to Indian researchers and was synergized with the Grand Challenges calls from Brazil and later with Africa.
The grant for developing data analytics solutions was to fund at a maximum of $100,000 per project, for 12-18 months. The grantees were required agreement for data access and contribution and were meant to provide an opportunity to test particularly bold ideas, including applying approaches from outside the field or that bridge fields to refine and rigorously test approaches.
The platform is not only providing grantees with the tools and on-going support that they need to ensure the success and sustainability of their projects, also the ki team is facilitating the collaboration opportunities for the India and Brazil grantees to discuss their projects, challenges, analysis methods and more with the teams. Towards this the use of tools such as synapse and cognitive city is encouraged encourages collaboration between teams, countries & thus has been made available for the teams to use widely
Summaries of the 7 funded projects under Grand Challenges India: Data Science Approaches to improve Maternal and Child Health in India
- Preterm birth risk in pregnant women – prediction using machine learning models
Organisation: Translational Health Science and Technology Institute; in collaboration with, Indian Institute of Technology, Madras
The study proposes on pregnant women (Garbh-ini) cohort, a multidimensional longitudinal dataset purposely designed to study preterm birth. The study will apply data-driven machine learning approaches to develop an accurate and clinically useful model to predict the risk of preterm births. It will use multiple models for classification, with better objective functions and misclassification penalties that will aid in a higher rate of accurate predictions, and resampling of the data to avert biases arising from class imbalance. The primary deliverable will be dynamic prediction models that can predict, at different periods of gestation, the PTB risk using the clinical, epidemiological and imaging data.
- Exploring risk factors of adverse maternal and child health outcomes using machine learning and other advanced data analytical approaches
Organisation:IIT Delhi; in collaboration with, Society for Applied Studies, Delhi
Combining multiple data sets from HBGDki using ML tools for prediction, classification and topic discovery may yield new insights for adverse birth outcomes and intermediate outcomes of interest. The study is based on a set of epidemiological, clinical and biochemical variables risk stratification algorithms for various adverse outcomes with practical applicability in health programme, and clinical settings may be feasible to develop using ML tools.ML can be used to suitably impute/bin missing values within datasets and merge variables from multiple datasets using robust data triangulation algorithms.
- Size matters: Predicting personalized risk of SGA
Organisation: Indian Institute of Science Education and Research, Pune,
The project aims to solve the problem of a large number of low-weight births in the Indian population and the inability to predict the risk reliably in the antenatal period. The study plan goes beyond finding a single metric or rule aiming to describe the risk of SGA in all women who plan to use clustering with factor-selection on the various available data-sets. They will use algorithm development and Bayesian modelling. At the individual level, an identification of possible causal factors behind the prediction of SGA, which will aid the clinician in determining the appropriate intervention.
- A data science approach to develop growth cut-offs for graded care of malnutrition
Organisation:St. John’s Research Institute, Bangalore; in collaboration with, Society for Applied Studies, Delhi
The study aims to calculate cut-offs using data provided by HBGDki and datasets with SAS, SJRI where weight, height, and age are available for children below five years in combination with other outcomes such as death, morbidity or hospitalization. Using WHO standards, weight for height, height for age and weight for age will be calculated, and these metrics will be used as determinants for risk of death, morbidity/hospitalization. A finer categorization of malnutrition based on the risk of mortality or significant morbidity can be used to develop and then deliver tailored optimized therapeutic options for what is essentially a far more eclectic group than what is captured by a three-category classification MAM, SAM and others.
- Child undernutrition in India: Exploration of nutritional gap based on distal and proximal factors,
Organisation:St. John’s Research Institute, Bangalore
The project study intends to adopt unconventional data analytical techniques to explore the multiple dimensions of child undernutrition in India, utilizing existing national surveys and the HBGDki database. The novel and sophisticated analytical methods that will be used include geospatial analysis, data triangulation through statistical matching, and multilevel modeling. The idea is unconventional because India does not have a single granular and multidimensional Health and Nutrition survey that can be analyzed at the district and sub-district level to provide precision insights to public health policy.
- Understanding the effects of initiation of complementary feeding at four months compared to six months on growth and infection among Indian children
Organisation: Christian Medical College and Hospital
Datasets within the purview of HBGDki with data on duration of breastfeeding, anthropometric measurements, and socio-demographic characteristics will be extracted, compiled and harmonized. WAZ, WHZ and HAZ scores will be calculated for each child. The study may describe the breastfeeding patterns in individual studies using a survival analysis approach and overall, through a meta-analytic approach.
- Developing district-level forecasts of vaccine coverage and inferring vaccine confidence across India using large public health datasets
Organisation: IIT Delhi; in collaboration with, INCLEN Trust, JNU & Imperial College, UK
The study aims to explore regional trends and variations in vaccine uptake, uncover relationships to other socioeconomic, demographic, and public health indicators, and develop a predictive model of the state of vaccine confidence in different parts of India. This will infer local-level confidence in vaccines by identifying areas with good access to healthcare infrastructure. The main goal of the proposal is to develop a prototype coverage monitoring and forecasting system across districts by using Gaussian process methods.