Developments in AI, coupled with advanced machine learning are at last aiming at the core of the problem of medical inflation, while also addressing down-stream concerns such as elevated customer experience at the insurance moment-of-truth – claims.
The solution lies in being able to mitigate two frictions – one is pure intelligence at the processing stage and the second is the ability to counter misuse or abuse. A third, higher order outcome that AI enables is real-time decision at both transaction level (decision) and portfolio level (product design, pricing and network management).
Losses in health insurance industry and its victims
‘Misuse’ and ‘control’ have always co-existed. At the fundamental level, an insurance company is structured to provide for losses in case of any eventuality. This contract is threatened by ‘bad actors’ – and necessitates that insurers constantly evolve their approach to minimise or eliminate this risk.
Emerging technologies increasingly provide the tools for this to be achieved.
Insurance ‘misuse’ or ‘fraud’ are mainly attributed to the following reasons:
Health insurance claims leakage comes in three forms: Claim for services not performed, wrongly calculated claims or those submitted for higher than actual amount. The immediate impact of the leakage is on margins and profitability – but with serious and lasting downstream consumer impact by reduced access and/or higher premium costs.
In markets like India an estimated $3.5bn (8% of premium revenues) is lost annually to fraud, waste and abuse with 70%+ is attributed to fraud documentation & fake bills. In the US, healthcare fraud is estimated at 10% (~$300bn) of the US’ annual health spending annually. Closer home, frauds in APAC region average 10%+ – with the insurance industry losing ~$2bn annually to fraudulent medical claims.
Anyone who has taken a health insurance can be victim of health insurance fraud, with some groups more vulnerable than others.
- Policyholders can become unwitting victims if they are unknowingly enrolled in a fraudulent insurance plan or due to identity theft.
- Healthcare providers can be unwittingly complicit (and ultimately victims) in fraudulent billing for services that were never provided or unnecessary procedures
- Government healthcare programmes are often victims of health insurance fraud – leading to significant financial losses and ultimately reduced access to healthcare services
Preventing and detecting fraud is critical to the integrity of the healthcare system.
Technology to the fore
ML driven and AI supported technologies have emerged as tools to deliver material claims automation and FWA-detection. Such cognitive systems take over the administrative pressure of screening and decisioning at speed and scale – leaving claims assessment teams to focus on higher order decisions – especially those that stem from medical decisions more than policy decisions.
We have representative implementation case studies from the ASEAN region to illustrate the working environment we have experienced in the implementation, the pitfalls and opportunities and finally the impact and outcomes.
Claims management today?
A typical APAC insurer with ~3m members can expect to receive 500,000+ claims annually. The task of claims verification requires a small army of assessors (in-house/outsourced), with a smorgasbord of supporting technology – and a huge first mile data debt in terms of digitally useful input data. Our experience across different health insurers has shown roughly 1 in 10 claims tend to be materially incorrect/fraudulent.
However, the process of identifying and challenging the claim is convoluted. Up to 90% claims tend to get flagged ‘suspect’— i.e. as potentially incorrect— based on the insurer rule book. Each claim then needs to be checked manually – and basis tribal knowledge, claims decisions are made, with varying degrees of decision quality.
However, the actual interventions/decisions can be for as low as 20% of the claims – in other words, the balance 70% of claims are flagged but without much actual decision making possible on them.
The intervention goal is therefore to reliably flag claims for which intervention is likely to pay off – for both the insurer and the provider – with the strategic motto to intervene only when required.
Incorrect claims value in our experience is in the high single digit percentages in Malaysia, with this rising to material double digit values in countries such as Vietnam and Indonesia.
AI-led claims management with digitalised inputs
The most critical – and often missing – step in making a smart decision is eliminating first-mile data debt. Most organisations struggle to eliminate this, with most efforts requiring manpower and technology solutions that deliver sub-optimal outcomes. Our research demonstrates that low-tech digitalisation efforts focused on digitalising analogue claims tend to miss a large mass of fraud that comes in the form of incorrect or fake documentation at the claim initiation stage. Technology is available that can eliminate such fraud at the claim initiation stage itself, saving substantial amounts before the claim reaches the assessor.
After intelligent digitalisation, comes intelligent categorization of the line items – this is critical to ensure there is no ‘garbage-in’ going to the later adjudication process. This is perhaps the most complex and as yet under-appreciated step – because claims line items change almost every day, depending on evolution of treatment protocols, addition of providers or even changes to the brands in use.
Our experience shows that – left unchecked – this can quickly balloon into fatal errors in terms of claims assessment, leading the system to effectively seize up and go back to requiring large scale manual interventions in assessment decisions.
Effective smart decisions
Correct implementation relieves the assessor from the need to make multiple time-sensitive interventions – freeing up capacity for higher order decisions and provider engagements, where needed. This is especially relevant when intelligent decision matrices are able to provide not only the decision flag, but also with guidance on how to approach the intervention.
The result is a simpler, faster claims management process – up to and including the intervention itself.
The final critical-to-quality (CTQ) factor is time. In our experience, perfectly turned out decisions delivered several hours (or sometimes days) after the point of decision becomes difficult to reconcile with the business imperative of actually recovering excess payments or with the customer experience imperatives. The final value is to ensure outcomes are in near real time – allowing the customer to walk away in a manner of minutes after the medical intervention is completed.
Getting the implementation right
The deployment of the right smart claims system is predicated on having the right underlying conditions:
- Valid underlying datasets to aid planning and structure of outcomes: A workable database generally encompasses several thousand data records with precise, consistent entries on the billing of individual cases (patient information, diagnoses, claims data) as well as related audit results.
- Flexible IT models: It is almost impossible to adapt new technologies to legacy IT landscapes. Our implementation experience has led to the design of an API led system that can easily graft onto systems, with no disruption to the core of the platforms in use.
- Deep partnering between client and provider: Our experience has involved 100s of successful implementations – and also a handful that didn’t work. One common factor in implementations that got bogged down was a traditional mindset of being required to play a ‘vendor role’ – which multiplies implementation risks when cutting edge technologies are being deployed. In an overwhelmingly large number of engagements, client teams have tended to develop trust quickly in the pilot/sandbox phase – setting the stage for successful full scale implementation.
Benefits for health insurers
Embedding AI in the process of hospital claims management offers multiple benefits at once, not just for insurers but also for patients, given the saving potential. In short, the shift away from claims management based on rigid rule books in favour of smart algorithms leads to greater efficiency and valid decisions –thus relieving the burden on all stakeholders and delivering savings. A
Mr Pramod Veturi is CEO of Perfios.ai, the AI analytics and real-time decision leader, with its headquarters in Kuala Lumpur, Malaysia. More can be found at
www.perfios.ai
| Case study: A $2 investment for a $2m saving
How intelligent ML models can nip medical inflation in the bud
One of SEA’s leading insurers zeroed in on the perfios.ai solution to help deliver digital transformation.
After validation by the country regulator and formally approved for adoption, all cashless and reimbursement claims go through perfios.ai Acclaim.
All elements of the claim are digitally extracted, intelligently categorised and run against policy validation and NCI rules.
The client deployed the Perfios tool to enable STP Auto Adjudication of both cashless and reimbursement claims, in < six months. Claims decision TAT dropped to < 15 minutes.
Critically, auto-adjudication capabilities delivered direct savings of over $2m over the implementation – with a per transaction investment of ~$2.
|
|
Intelligent categorisation is key
Digitalisation only captures what is there – it cannot attribute value to what it is not there – for example a medication is a medication but for the system to determine whether it is the right medication for the context of the claim is intelligence. No OCR tool generate this intelligence.
|