Closing the immunization equity gap: Using predictive analytics powered by machine learning to identify & target high-risk urban populations
Enabling real-time identification of children likely to drop out of the immunization schedule for early intervention and inclusive planning
Describe what you intend to do and how you'll do it in one to two sentences (required 350 Characters)
IRD aims to scale a Predictive Analytics (PA) algorithm powered by Artificial Intelligence (AI) with Government’s Digital Immunization Registry (DIR) in Pakistan. AI will help to identify children at risk of drop-out and allow programs to accurately identify, locate and provide early intervention to at-risk individuals and communities.
Explain the innovation (2,500 characters)
Metropoles in LMICs contain clusters of un- and under-immunized children, who often reside in densely populated, demographically unstable slum areas. Policy-makers lack the technology to model and visualize risk on large, evolving datasets, thus having neither the ability nor the evidence to identify and target these vulnerable clusters. Population mobility and diversity further compound the problem.
Our innovation provides a universally compatible technology which identifies at-risk populations and delineates socio-demographic predictors of under-immunization even in dynamic contexts. Using machine learning (ML), it accurately forecasts future immunization outcomes of groups and individuals based on existing data. It is a standalone module, which can be linked to any existing DIR to help identify defaulters in complex urban settings. As soon as child’s data is entered into the registry, the algorithm classifies it as high or low risk through modelling based on similarity analytics. Risk classification of groups and individuals is not only visible to the vaccinator, who can provide on-spot counselling, but is also accessible to program managers and policy makers, who can preempt drop-outs among high-risk groups through targeted planning. As more data is captured over time, the AI system continues to self-learn from accumulated records, responding to changing immunization correlates in the context of rapid urbanization.
Primary beneficiaries include children under 2 who initiate the immunization schedule but are unlikely to complete. Targeted efforts will be made to track them and prevent them from defaulting following their categorization as high-risk. Furthermore, marginalized populations will benefit as a whole, as policy-planners will become aware of their characteristics (ethnicity, occupation, location, etc.), and will target them with tailored interventions. Vulnerable populations identified by PA are likely to include children residing in urban slums who live in makeshift homes and belong to migrant families. With over 1.6 million children enrolled in our target area’s DIR, we expect our intervention to impact thousands of families.
A pilot of PA algorithm with 49,000 records showed that despite significant data variation, the accuracy of the prediction was high (78%). With ML, accuracy would gradually increase to over 99% within a year. Since PA can be linked to any DIR, it will help accelerate immunization efficiencies in other Gavi countries.
This image shows how the Predictive Analytics Algorithm can be utilized practically and the value that it creates at multiple levels: caregivers (adequate counselling for at-risk children and reduction in morbidity related to vaccine-preventable diseases), vaccinators (real time alerts for at-risk children prompting delivery of on-spot counselling and other targeted
interventions), and supervisors (identification and tracking of at-risk populations).
Which part(s) of the world does this innovation target?
Our idea targets Sindh province of Pakistan. Home to almost 48 million people, Sindh is characterized by rapid urbanization, large number of slums, with cities expanding exponentially. As LMICs are shifting towards digitization, and many have implemented DIRs, PA can also be deployed in other LMICs
Who will work alongside your organization in the project idea? (1,000 characters)
Our implementing partners include Government of Sindh’s Expanded Program for Immunization (EPI), the Indus Health Network in Pakistan (IHN) and Interactive Health Solutions (IHS). IRD has been directly engaged with EPI at the provincial and national level in Pakistan since 2000, and has supported EPI in various projects. As a software development organization, IHS leverages modern technologies to aid health service planning and delivery, and has worked alongside IRD for over a decade. The two organizations have collaboratively developed and deployed several mHealth projects, including development, pilot and scale-up of provincial DIR over 5 years. We have a close partnership with IHN, which consists of 14 primaries, secondary and 10 tertiary care health facilities, encompassing one of the largest networks of tertiary care programs in Pakistan.
How is your idea unique? (750 characters)
Traditionally AI-based initiatives have focused on providing diagnosis or decision support, whereas our approach uniquely focuses on predictive analytics to identify vulnerable children and delineate risk factors for under-immunization in real time.
As an organization which has successfully leveraged technological innovations to address health delivery gaps, with over a decade’s worth of experience working with EPI, we are well-positioned to implement this idea. Moreover, due to interoperability of the algorithm module with a standard DIR, it fits well into the broader ecosystem, and adds value to a multitude of existing global immunization systems.
What is the name of your organization
Explain your organization (250 characters)
IRD is a not for profit independent global health delivery and research organization working in over 15 countries. The IRD team leverages process and technology innovations to address global health delivery gaps in low resource environments.
Type of Submitter
We are a registered NGO or Non-Profit Organization
Women’s health/rights focused organization
Gender and Diversity (500 characters)
IRD is an equal employment opportunity employer and proactively recruits applicants with diversified candidature. HR policies focus on nurturing a women-friendly culture and providing universal career advancement opportunities, resulting in women comprising of 44% of the overall workforce. A strict workplace harassment policy is enforced to ensure universal protection of rights, and all programs implemented by the organization undergo review to ensure gender-awareness within their protocols.
Organization Location (less than 250 Characters)
IRD is headquartered in Singapore, and our team is based in Karachi, Pakistan. In addition to Pakistan and Singapore, IRD has country offices in the UAE, South Africa, Bangladesh, Indonesia, Phillipines, and Vietnam.
Size of organization (number of employees):
Scale of organizational work
Global (within 2 or more global regions)
Tell us more about you
Our team consists of Subhash Chandir (doctorate in Global Disease Epidemiology and Control with certificate in Vaccines Science & Policy from Johns Hopkins University), Danya Arif (Masters in Development Management from the London School of Economics), Mehr Munir (Masters in Humanities from New York University), and Ali Habib (Master of Engineering Management from Duke University). Our connection with local partners are already established, and our experience of developing and deploying mhealth interventions on a large scale makes us the right team to implement this project. In our work, we prioritize ethical responsibility towards our subjects and beneficiaries.