AI-ML Platform for prevention of Antimicrobial Resistance in Neonates

A machine learning based digital health application for early detection of Neonatal Sepsis and Optimization of Antibiotics among newborn babies in low resource regions.

Dr. Hyma Goparaju, PhD is the Co-founder and CEO of Avyantra Health Technologies founded in 2017 in India. She has industry experience of close to 20 years. 

India has a high neonatal mortality rate (NMR) of 20 per 1000 live births. Rural NMR is two times that of urban NMR. Neonatal sepsis including antimicrobial resistance (AMR) causes mortality of nearly 6 lakh babies annually in India. In India, close to 300,000 deaths due to AMR and 1 million associated with AMR are reported annually. Primary factors for high neonatal mortality rate (NMR) in India are delayed diagnosis. Culture reports, the current gold standard diagnosis requires a blood volume between 0.5 ml – 1 but getting this volume from a newborn baby is very difficult. Also, culture assessment for microbial presence takes anywhere between 48 and 72 hours. These issues have led to the practice of prescribing empirical antibiotics which is contributing to antibiotic resistance occurring due to unnecessary / overuse of antibiotics. Especially in preterm babies, it is hard to make a suspicion of infections, and antibiotic administration is common, leading to higher rates of mortality and morbidities when administered during the first seven days among newborns. These challenges call for other novel methods for early diagnosis of neonatal sepsis to pave the way for optimal administration of antibiotics and deliver effective treatment. 

The target beneficiaries for our application are:

a. Frontline health volunteers / Rural Clinics / Primary healthcare professionals – Low resource and infrastructural constraints hamper the smooth delivery of effective healthcare services in rural India. Our platform will use algorithms developed from clinical data and non-clinical inputs to put in place evidence-based practice for early diagnosis and treatment of newborn babies.

 b. Tertiary hospitals - are loaded due to increasing patient flow. Prescription of empirical antibiotics is a widely followed practice due to which antibiotic resistance is on the rise among newborns leading to mortality and future health hazards. Our integrated digital health platform helps these centers to provide timely and optimal treatment aided by continuous evaluation of babies with the help of machine learning-based digital health applications.

 c. Mothers and Children – Young mothers and infants find it hard to commute an average distance of 100 km to receive quality healthcare from an urban health center. Our platform helps in addressing this issue by helping primary centres handle a good portion of their patient load instead of transferring them to tertiary centres. This move can significantly reduce 40% of deaths of newborn babies caused due to transport delays.

PreSco is a Digital Public Good (DPG) status. The Digital Public Goods Alliance (DPGA) is a multi-stakeholder initiative whose mission is to accelerate the attainment of Sustainable Development Goals (SDGs) in low- and middle-income countries by facilitating discovery, development, use of, and investment in DPGs. We have developed a comprehensive cloud-based Neonatal Sepsis prediction application with open-source technologies, and advanced Machine Learning algorithms. It is built by processing historical & prospective data on sepsis among neonates. Since our solution is an asset-light model using open-source technologies for building it, this makes our solution cost-effective and interoperable. Additionally, by connecting low-resource areas of developing countries using cost-effective technologies, we hope to achieve improved access to quality healthcare sustainably and continuously.

Clinical decision support systems CDSS systems can bring down the usage of antibiotics significantly by at least 50% and the rate of infections by 84% with consistent use and reduce the cost of treatment, antibiotic resistance, and mortality. Results from studies and experiments on maternal and child health in developing countries from Africa have reported that innovative interventions demonstrate long-term sustainability and effectiveness with optimal use of digital health decision-making tools. These studies have also demonstrated that in a period of five years from implementation, digital interventions result in significant improvement of maternal and child indicators in low- and middle-income countries. Recent reports from India have stated high mortality rates, close to 300,000 directly due to AMR and 1 million associated with AMR.

Third-party research - 

• Bhutta, Z. A., Das, J. K., Bahl, R., Lawn, J. E., Salam, R. A., Paul, V. K., & Walker, N. (2014). Can available interventions end preventable deaths in mothers, newborn babies, and stillbirths, and at what cost?. The Lancet, 384(9940), 347-370.
• Divall, P., Camosso-Stefinovic, J., & Baker, R. (2013). The use of personal digital assistants in clinical decision making by health care professionals: a systematic review. Health informatics journal, 19(1), 16-28.

We plan to scale the impact in the next year through the completion of testing and validation at an additional 4 - 5 hospitals, integration of application in ABDM digital initiative in India, the addition of new features such as image processing, EDGE enablement device for low connectivity regions,  enrolment and training of doctors, nurses, and ASHA workers for the application, performing diagnostic tests though the application and improving the performance of the sepsis application through feedback and finetuning of algorithms. In the next 3 years, we plan to increase the coverage of application across the country and collaborate with the government and international organizations for the adoption of application in healthcare centers in the country which will lead to a significant impact on early diagnosis of neonatal sepsis, rationalization of antibiotic usage and long-term impact on AMR. PreSco is a scalable platform that can easily be upgraded to other healthcare conditions. Our patient health record system enables tracking of a baby's health records, treatment protocols, antibiotic usage, resistance, etc., right from the time of birth to its growth. This system would ensure equitable and informed decision-making for mothers and health professionals alike.  

Key metrics of Risk Predictor Scores such as Sensitivity, Specificity, Accuracy, and AUROC – We aim at a significant improvement in sensitivity to a consistent80% and above in the prediction of neonatal sepsis risk score generation. We expect the application to record clinical and non-clinical data of babies and track their antibiotic usage and administration frequencies.

Effectiveness of the application i.e. Rapid diagnosis of Sepsis – Consistency in score generation for rapid diagnosis with clinical data inputs entered into the application to less than 2 mins (compared to 48 – 72 hours for gold standard testing of culture tests).

Ease of Deployment in low-resource settings - The application should have the ease to be accessed from anywhere, anytime. In rural areas, the application should be accessed from any mobile or desktop device having internet connectivity.

Positive Feedback on Performance & User-Friendly Features from Users – Expect positive feedback from users.

Regulatory / Legal Barriers – Currently, there are no specific regulatory policies for the commercialization of digital health apps in India. However, we expect to follow the guidelines laid down by the National Digital Health Mission (NDHM) for building and implementing digital solutions.

Cultural and Education Barriers – Digital health apps slowly gaining traction in India and more awareness programs need to be conducted for healthcare practitioners. We expect to address this issue with a demonstration of the application in health centers, and health exhibitions etc conduct training in the usage of the application.

Infrastructural Barriers – The performance of cloud-based digital health apps could be affected in low internet connectivity regions. We will also explore the possibility of ML-on-the-edge solutions for low internet connectivity regions.

Financial Constraints – Our project needs continuous funding support to gain the traction required. We are actively pursuing multiple funding opportunities for taking forward the validation of the application and eventual commercialization of the application. 

We are applying to The Trinity Challenge as it is one of the most popular innovation challenges globally. We believe that winning the challenge will help us tremendously in scaling our current work. The Challenge will also help us in making significant strides towards commercialization. The fund size and networking support provided by The Trinity Challenge will give a boost to our team and help in our quest towards building sustainable innovation for addressing global challenges. We are confident that winning the challenge will not only help us in overcoming the various barriers but also aid us in taking gaining global visibility and traction for our solution.

For the B2B model, we will increase our partnerships with hospital networks.

We intend to tap connects formed as part of our collaboration with the Unicef India Office.

We are proposing to integrate our solution with the Government of India’s Universal Health Program, Ayushman Bharat Digital Mission (ABDM) which we strongly believe will help us in expanding our B2B as well as B2C footprint.

We look to collaborating with MIT Solve’s associated institutions in the allied area to expand our solution’s presence and reach.

We are also looking at collaborating with associated publications for sharing the generated evidence and results through a publication in a peer-reviewed journal.

 In our continued effort to enhance our Digital Public Goods, we would be keen to collaborate with like-minded professionals and associations who would be willing to contribute to our solution.