Claims Leakage Analytics in Indian Health Insurance

Claims leakage is the avoidable share of claim cost: amounts paid that, with better information or process, would not have been paid or would have been paid at a lower amount. Indian health insurers commonly experience leakage at 12 to 18% of incurred claims, with the variance driven by retail-versus-group mix, hospital network composition, and the maturity of the insurer or TPA's controls.

Leakage is not the same as fraud. Fraud is a deliberate deception, typically by a provider, member, or organised ring. Leakage includes fraud but also covers:

• abuse: services billed within technical compliance but inappropriate to the case
• waste: unnecessary admissions, length-of-stay padding, low-value tests
• error: billing mistakes, coding errors, eligibility gaps
• process slippage: failure to apply policy exclusions, sub-limits, or co-payments correctly
• network leakage: claims paid at out-of-network rates that should have been steered

A leakage programme that focuses only on fraud captures a third of the addressable pool. The remaining two-thirds sit in abuse, waste, and process, where the loss is real but the intent is not adversarial.

Indian health claims flow through a TPA or in-house claims function, with hospital empanelment driving cashless authorisations and reimbursement claims handled separately. Leakage sources differ by stage.

Pre-authorisation stage: provider over-coding (higher acuity than warranted), inflated estimates, medically unnecessary admissions, length-of-stay padding, and add-on services that do not match the diagnosis.

Discharge and billing stage: line-item upcoding, double billing of consumables, charges for services not rendered, unbundled charges for procedures that should be billed as a bundle, and pharmacy markups beyond policy norms.

Adjudication stage: failure to apply sub-limits (room rent, ICU charges, consumables), failure to apply waiting periods on pre-existing diseases, mis-classification of diagnosis groups, and missed deductibles or co-payments.

Reimbursement stage: member-side fraud through inflated or fabricated bills, duplicate submissions across multiple insurers, and policy-eligibility gaps not detected in time.

An effective analytics programme covers all four stages. Programmes that focus only on adjudication recover a portion of leakage but miss the larger pool at pre-authorisation, where the right intervention prevents the spend entirely.

Pre-pay analytics is the highest-impact form of leakage control. It flags claims before payment and routes them to enhanced review. The Indian implementation challenges differ from US markets because the TPA holds operational control and the insurer's analytical signal must move through the TPA workflow.

A working pre-pay analytics architecture includes:

• rule-based flagging on documented red flags (admission for a diagnosis typically managed outpatient, length of stay exceeding clinical norm by a defined margin, charges out of pattern for the network and city)
• provider-pattern scoring identifying hospitals whose admission rate, length of stay, or charge profile diverges from peers serving similar populations
• member-pattern scoring identifying repeated admissions, multiple-insurer overlap, or claim patterns inconsistent with member age and history
• clinical review by a medical officer or peer-reviewer where rules and scores produce a high-suspicion case
• feedback from review outcomes into the rule and model parameters

Hospitals account for the majority of leakage volume in Indian health insurance, both through deliberate abuse and through systemic billing patterns that exceed clinical justification. Provider-pattern analytics is now the highest-priority workstream for most insurers and TPAs.

Features worth modelling:

• admission rate for elective procedures compared to peer hospitals in the same city and tier
• average length of stay by diagnosis, controlled for case-mix
• average billed amount by procedure, controlled for insurer-negotiated rates
• complication coding patterns: hospitals that bill complication-driven uplifts at materially higher rates than peers
• discharge patterns: hospitals that discharge precisely on day-3 versus those with realistic clinical distributions
• pharmacy and consumables share of total billing, often a fast-rising leakage source

Provider-pattern scoring outputs feed into network management decisions: enhanced contractual terms for outlier providers, audit visits, suspension of cashless pre-authorisation for specific diagnosis groups, and ultimately de-empanelment. Indian insurers and TPAs are increasingly publishing provider scorecards internally and using them in network reviews.

Post-pay audit recovers leakage after claims have been paid. Recovery rates in Indian health are typically lower than in mature US markets because hospital relationships are commercial-sensitive and recovery requires hospital cooperation that is not always forthcoming.

A practical post-pay programme structure:

• statistical sampling of paid claims, with stratification by hospital tier, claim size, and procedure type
• focused audits on outlier hospitals identified by pre-pay or provider-pattern signals
• member audits on reimbursement claims with statistical red flags
• billing review of large group accounts where the employer or broker has standing to challenge
• recovery negotiation with hospitals, typically through the TPA's network team

Recovery rates of 2 to 5% of paid claims are common at well-run programmes, with the upper end achieved through credible threats of network suspension and through state-level coordination with regulators. The economics of post-pay audit are good (recovery materially exceeds audit cost), but post-pay alone cannot substitute for pre-pay; the leakage that has already gone through the cashless channel is hard to recover once the patient has been discharged and the hospital has been paid.

Organised fraud in Indian health insurance ranges from individual document forgery to large network-driven rings involving hospitals, medical professionals, and intermediaries. Network analytics, in the graph-theory sense, surfaces patterns that simpler analytics misses.

Useful network signals include:

• shared phone numbers, addresses, or bank accounts across members, agents, and providers
• agent-driven clusters where one agent's portfolio shows abnormal claim frequency or amount
• doctor-driven clusters where one doctor's prescriptions or referrals appear disproportionately in suspect claims
• geographic clusters where multiple unrelated members from one location present the same claim profile
• time-pattern clusters where claims appear in suspicious sequences after specific events (policy issuance, sum-insured increase)

Indian insurers with mature fraud programmes increasingly run dedicated Special Investigation Units (SIUs) backed by graph analytics tooling. The IRDAI's 2024 guidance on fraud control supports this direction and expects insurers above a defined scale to maintain an SIU function with documented escalation paths to law enforcement under PMLA and IPC provisions.

A capable claims-leakage analytics programme is not a tool purchase; it is an operating discipline. Five elements distinguish programmes that capture material leakage from those that produce reports without recovery.

First, leadership alignment. The chief executive, chief actuary, and TPA leadership must agree that leakage control is a measurable priority with recovery targets in the operating plan, not a side project.

Second, data infrastructure. Pre-pay analytics requires sub-hour data flow from the TPA system to the analytics platform. Post-pay analytics requires complete claim history with provider, member, and case attributes. Many programmes stall because the data foundation is incomplete or fragmented.

Third, clinical capability. Rules and models surface suspicion; humans confirm. The insurer or TPA needs medical officers and clinical reviewers with authority to confirm, deny, or refer cases, with workload calibrated to the flagging volume.

Fourth, provider-management muscle. Analytics surfaces outlier providers; reducing their leakage requires contract revision, audit visits, and willingness to suspend or terminate problem providers. Programmes that flag but do not act produce frustration, not recovery.

Fifth, continuous feedback. Every confirmed leakage case should feed back into rules, models, and provider scorecards. The signal-to-noise ratio improves only through systematic learning, which requires investment in the operational discipline that closes the loop.