Professor Rhema Vaithianathan joined ISSR in 2019 as Professor of Social Data Analytics to establish the Australian site of the Centre for Social Data Analytics (CSDA) at ISSR (UQ). CSDA was founded at Auckland University of Technology, New Zealand, in 2016 and is an international leader in data science research for social good. The Centre is highly regarded for its human-centred approach to applying machine learning methods to high stakes decision-making.
Rhema brings vast experience in the application of data analytics and machine learning techniques for social good and her predictive analytics work focuses on the methodologies for, and implementation and implications of, predictive risk modelling in health and social services settings. Her approach to predictive risk modelling is underpinned by a set of guardrails that ensure the development of trusted, ethical and transparent tools.
Rhema is internationally renowned for developing the Allegheny Family Screening Tool (AFST), an advanced algorithm used by call centre staff in the Allegheny County Department of Human Services in Pennsylvania (US), to help them triage child maltreatment allegations. This tool is recognised as a world first in the careful and effective creation and use of risk modelling tools for child protection.
Rhema is at the forefront of international research and implementation in data science that harnesses the potential of data, using machine learning for social good.
The successful implementation of the AFST in 2016 opened opportunities for Rhema and CSDA to explore predictive analytics solutions for a diverse range of public sector use cases in Allegheny County and beyond. As a result, Rhema has also led the development of a predictive risk modelling tool to support the County’s new Hello Baby child harm prevention program. Hello Baby is designed to engage and support all county families with new-born babies, with the predictive risk modelling tool ensuring that the most intensive support services are offered to the highest need families.
Rhema’s team also recently deployed their first machine learning tool in the homelessness domain in Allegheny County. The Allegheny Housing Assessment (AHA) tool uses data to understand a client’s risk of experiencing harms associated with homelessness, accurately prioritising requests for housing help. This tool was subject to a detailed independent ethical review that informed its development and deployment, helping to ensure that the AHA is accurate and does not introduce any disadvantage for vulnerable groups.
In the last two years Professor Vaithianathan and her team have also developed and deployed child welfare call screening tools for two Colorado counties (Douglas County and Larimer County), which build upon the approach taken with the AFST.
With positive results from the independent impact evaluation for the AFST, promising preliminary impact evaluation results for the Douglas County tool and an impact evaluation of the Larimer County tool underway, Rhema and her team are focused on extending the application of predictive analytics for positive impact in the child welfare domain. In a new child welfare use case for data analytics, CSDA is working with the Children’s Data Network at the University of Southern California to develop and implement a predictive risk modelling supervision tool for the Los Angeles County Department of Children and Family Services. Rhema is leading the CSDA team responsible for developing and deploying a child welfare risk stratification model to estimate investigation complexity and risk. This tool will allow supervisors to better incorporate data into their supervision practices with frontline staff, while also providing opportunities to implement continuous quality improvement protocols during investigation.
Rhema’s recent publications include research in Jama Pediatrics that establishes a link between a child’s risk of placement out home and their risk of hospitalisation injury and an article in the Future Healthcare Journal that shares her work as part of an international team that established an association between the daily mood of trainee doctors and their intention to quit. There is significant potential for both sets of findings to inform practice – the first article confirms the value of predictive risk modelling and raises important new research questions about racial disparity, and the second article demonstrates the potential for systematic workplace changes to help with staff wellbeing and retention.
In Queensland, Rhema is currently exploring the potential for predictive analytics to help identify children with critical illness who are at risk of poor educational outcomes; investigating options for predictive risk modelling that can help agencies better serve families and children; and supporting the prioritisation of primary health care services for clients in need of intensified support.
Rhema believes the future of predictive risk modelling for government agencies lies in the public sector leading the way by adopting human-centred data science.