Artificial intelligence-driven fraud detection has emerged as a key focus under the Ayushman Bharat Pradhan Mantri Jan Arogya Yojana, with officials saying anti-fraud systems already deployed under the scheme have prevented fraudulent claims worth around ₹630 crore. Additional penalties and recoveries imposed on hospitals were estimated at around ₹200 crore, according to officials of the National Health Authority.
AI Tools to Detect Health Scheme Fraud
Officials said AI-driven fraud detection systems developed through a national healthcare hackathon could help prevent large-scale misuse of public funds under the Ayushman Bharat scheme. The hackathon was organised in collaboration with the IndiaAI Mission and the Indian Institute of Science, Bengaluru.
FCRF Academy Launches Premier Anti-Money Laundering Certification Program
Addressing the finale of the two-day event in Bengaluru, Sunil Kumar Barnwal, Chief Executive Officer of the National Health Authority, said the growing use of AI-generated fake medical records and forged claims had become a serious challenge for public health insurance systems.
The event focused on developing solutions that could help detect document forgery, deepfakes and fraudulent medical claims. Officials said such systems could also speed up claims processing and improve patient care.
Fake Records and Forged Claims Under Scanner
Jyoti Yadav, Joint Secretary, AB-PMJAY, said AI-generated clinical notes, discharge summaries and diagnostic reports were now being submitted. She described the trend as still being at an early stage, adding that authorities wanted to develop solutions before the problem became too large to handle.
She said hospitals submit scanned medical records, diagnostic reports and patient documents while raising claims under AB-PMJAY. These records are vulnerable to manipulation through AI tools, including alteration of images, removal of watermarks, copy-pasting of reports and changes in formatting or alignment.
The hackathon sought solutions to identify forged or manipulated documents, correctly classify medical records and verify whether submitted reports actually supported the treatment claimed by hospitals. Officials said several innovative solutions emerged during the event, including one fraud-detection model built entirely on mathematical techniques rather than conventional AI systems.
Data Analytics to Track Suspicious Patterns
Officials said fraud patterns under the scheme ranged from inflated billing to fabricated treatment records. Ms. Yadav cited cases where patients diagnosed with minor ailments such as fever were shown as having undergone major procedures like total knee replacement.
In some cases, photographs of one ICU patient were allegedly reused across multiple insurance claims by changing names and patient details. Under AB-PMJAY, hospitals receive higher payments for ICU admissions compared to general ward admissions, creating incentives for misuse, she said.
Officials also said machine-learning systems were being trained to check whether laboratory parameters in diagnostic reports genuinely supported procedures such as dialysis. If creatinine levels did not match medical guidelines for dialysis treatment, claims could be automatically flagged.
Data analytics was also helping authorities identify unusual treatment patterns across States. Officials said systems could detect suspicious trends where multiple beneficiaries from one region travelled to a particular hospital in another State for the same treatment shortly after obtaining Ayushman cards.
Ms. Yadav said individual cases may not appear suspicious, but patterns emerge when data is analysed collectively. AI systems are also being used to identify irregularities in chemotherapy cycles, where hospitals allegedly bill for treatment sessions at medically impossible intervals.
She added that healthcare fraud detection requires a careful balance because overly rigid automated systems can also affect genuine patients. Health cannot function entirely on fixed rules, she said, as there are always exceptional cases.
