AI-driven predictive solutions for Early diagnosis of Sepsis in the Indian healthcare industry.

Time of India

a day ago

AI, the game changer: COMPOSER as a global benchmark
Live Events
Indian institutions leading innovation:
Indian and global startups taking sepsis predictions forward:
Doze.
Avyantra Health Technologies:
Prognosis:
AlgoDx:
Dascena:
Challenges:
Sepsis, in 2017, was known as a global health priority by the World Health Assembly and WHO, considering its highly ranked morbidity and mortality rates around the world. On a global scale, sepsis has affected over 49 million people each year, contributing to approximately 11 million deaths each year, significantly pressing down on the low- and middle-income countries, not exempting India.The condition occurs as a result of the body's (immune system, to be specific) extreme reaction to an infection, eventually leading to tissue damage, organ failure, and even death. This condition stands to be so fatal because of how difficult it is to be detected immediately. For the disease is easily mistaken for other illnesses, and its variability is based on individual uniqueness. Sepsis is quite a challenge in India, with delayed detection and treatment being common obstacles, similarly to most of its global counterparts.Therefore, the interference of AI has claimed its position as a powerful tool in forecasting sepsis risk, allowing healthcare professionals to assess early intervention, fostering a chance at saving a larger number of disease-stricken patients.AI-predictive models exist with the ability to analyze huge amounts of clinical data, such as vital signs, lab results, and electronic health records (EHR), at a much faster rate than a human being would have been able to, according to traditional approaches. These algorithms are deemed to improve the accuracy and timeline of sepsis prediction in a clinical setting. AI prediction technology is particularly beneficial in ICUs, where sepsis is prevalent amongst a large number of patients.Although the exact mechanism varies from model to model, a common factor that persists amongst them all is that they examine key variables that trigger the development of sepsis. These variables include vital signs, demographics, lab results, comorbidities, and administrative data like the duration of stay in the ICU . Simply put, the algorithm, upon analyzing the vast clinical data, generates a pattern with common factors that correlate to the conception of sepsis within the individual's body. Moreover, the models work by constantly monitoring a patient's EHR data and assessing the previously mentioned variables in real-time to detect the likelihood that sepsis will happen. Eventually, enabling it to predict the onset of sepsis hours prior.With such capabilities, healthcare professionals are served a chance at garnering precise sepsis prediction, allowing time-efficient interventions associated with the administration of antibiotics. This efficiency was proved in last year's 'Npj Digital Medicine' study, which evaluated the effectiveness of 'COMPOSER' (Conformal Multidimensional Prediction Of Sepsis Risk), a deep learning model designed for early sepsis prediction. The study assessed the effectiveness of the deep learning model by having it predict the same in two emergency departments.There were a total of 6,217 patients in both emergency departments combined between January 1st, 2021, and April 30th, 2023, all of whom satisfied the eligibility requirements for sepsis, according to the Third International Consensus Definitions for Sepsis and Septic Shock, published in 2016 (Sepsis-3 Consensus). This group of individuals was used to test if the deep learning model would be able to flag these patients earlier than the clinicians. This testing proceeded further, with an already granted guarantee that the patients do in fact suffer from sepsis, so if the model were to misclassify them, it would be made obvious, making it a retrospective validation.Interestingly, they were accurately classified by the deep learning model; 5065 patients were identified in the pre-intervention phase and the remaining 1152 in the post-intervention phase. Henceforth, leading to a 17% reduction in mortality rates.While COMPOSER itself may not be a commercial product, it serves as a powerful means to stand as a global benchmark for what AI can achieve in sepsis prediction.Building on the same framework of data-driven, predictive intervention, a new wave of AI startups, both in India and the global arena as a whole, is working to utilize these capabilities at scale, across hospitals and community settings. These startups are transforming the insights from models like COMPOSER into accessible and usable solutions that aim to save lives in the real world.Indian research institutes and hospitals have begun to actively adopt this approach as well, considering the urgent need for the same in healthcare. Amrita Vishwa Vidyapeetham University in Kochi has developed an AI model to detect sepsis via its predictive capabilities. This model, too, works the same way as the above-mentioned 'COMPOSER.'Amrita Vishwa Vidhyapeetham University uses a model that works with a 'vital-SEP' system that helps doctors spot early patterns of sepsis by analyzing the patients' vital signs over time. These vital signs include heart rate, temperature, and blood pressure. The system breaks down each patient's vitals into 6 time points before the onset of a potential sepsis event, then it extracts smart 'features' from them, which consist of the average of the values ('baseline'), how much these values change ('deltas'), how quickly these values change ('change of rate'), and how much they vary ('variances').These 'features,' up to 240 per patient, are then turned into data values that a computer can comprehend. By conducting a comparative analysis between the patient who developed sepsis and those who didn't, the Vital-SEP system identifies patterns such as rapid changes in heart rate or temperature as the early warnings of a potential onset of sepsis.Apollo Hospitals, on the other hand, uses data analytics and AI-driven infection control surveillance systems to monitor patient data and detect infections, including sepsis. This methodology helps the hospital reduce sepsis mortality rates by allowing faster clinical interventions.Amrita Vishwa Vidhyapeetham University and research initiatives like COMPOSER have set a benchmark, demonstrating AI's clinical potential in controlled settings, welcoming startups to productize and scale these innovations to reach as many hospital bedsides as possible. These ventures are curating usable platforms that inculcate daily workflows and address real-world gaps like ICU staffing, rural demographics' health access, and the misuse of antibiotics. Below are some of the key players leading this transformation:1.'Dozee' is an AI-powered contactless patient monitoring platform founded by Mudit Dandwate and Gaurav Parchani in 2015. Dozee started after the co-founder Mudit Dandwate was hospitalized due to an unforeseen injury. There he understood firsthand how delayed vitals monitoring contributed to missed early warning signs. The startup's flagship product is a non-invasive sensor placed under a mattress that tracks vitals such as temperature, heart rate, respiration, and sleep patterns constantly.By means of this sensor, the machine learning algorithms involved can analyze data in real-time, thereby detecting early signs of clinical deterioration before the symptoms become visible to the naked eye. The platform sends out smart alerts to nurses and doctors via the apps the model is attached to, allowing the hospitals to act early.2.Avyantra Health Technologies builds AI solutions for maternal and neonatal healthcare for the rural populations. It was founded by Dr. Rajeev Kumar in 2020 in Hyderabad in collaboration with pediatricians and neonatologists, aimed at helping accredited social health activists (ASHAs), midwives, and primary health workers take early preventive actions in rural areas. Its flagship innovation, 'PreSco,' is a cloud-based tool that predicts neonatal sepsis risk using data such as maternal history, birth weight, temperature, and other kinds of vitals.Avyantra moves forward with its goal to reduce unnecessary antibiotic use and improve neonatal outcomes by enabling real-time and data-driven decisions in the first week of life.3.Prenosis is an FDA-authorized AI tool designed for sepsis diagnosis. It was founded by Bobby Reddy and Jason Paragas in Chicago. Prenosis focuses on early, accurate, and lab-independent sepsis diagnosis. Its flagship product, ImmunoScore, uses a mix of 22 clinical and immunological biomarkers, including vital signs, bloodwork, and inflammation levels. The machine learning models involved, trained on over 100,000 patient encounters, then process the mentioned data.Prenosis became the first company to receive FDA authorization for an AI diagnostic for sepsis in 2024. It is now being used across hospitals in the US, improving early diagnosis rates and reducing unnecessary ICU admissions.4.AlgoDx is an AI-driven clinical decision support system founded by Daniel Persson and Ola Winqvist in Sweden. The AI-powered tool offers automated sepsis prediction using real-time hospital EHR (electronic health records) data. AlgoDx's product, NAVOY CDS, runs quietly in the background of hospital systems, alerting the staff when a patient's risk for sepsis rises based on subtle trends in vitals, labs, and stay duration. In 2024, NAVOY CDS also received an FDA 510(k) clearance, confirming its clinical effectiveness in reducing sepsis mortality.5.Dascena is created by a machine learning model trained to predict sepsis hours prior to the traditional methods of diagnosis. It was founded by Dr. Ritankar Das in 2017 in California. InSight is an EHR-based sepsis prediction algorithm, processing high-frequency clinical data including vitals, comorbidities, and recent interventions to generate real-time risk scores.This model was validated in peer-reviewed trials and used in hospitals across the US before the company was officially acquired by CirrusDx.Though these innovations in the healthcare industry provide promising advantages for sepsis predictions, they come with several challenges. For instance, these models ask for access to high-quality and digitized data. Something remains in limitation in many Indian hospitals where EHR systems either do not exist or are inconsistently implemented.They must also integrate seamlessly with busy clinical environments. Apollo Hospitals use data analytics for infection surveillance, but they also highlight the requirement for clinician-friendly tools that ask for a lessened manual load rather than increasing it altogether.Lastly, it touches upon sensitive patient data and influences clinical decisions; it also raises questions of concern about privacy, transparency, and safety. All of these concerns were recognized globally, not exempting the developers of models like COMPOSER, thereby emphasizing the need for reliable and uncertainty-aware AI in clinical decision-making.Despite the hurdles, the interweaving of AI and healthcare is opening new gateways in the fight against sepsis in India. With technology of this kind, Indian hospitals can diagnose sepsis early, dramatically improving patient outcomes and saving lives.