Type to SearchView Tags

Analytics use cases in Disability Claim management
Veeraraghavan Kazhiyur Mannar Business Manager- Insurance Practice, HCL Financial Services | July 23, 2020
143 Views

Disability claim management is a complex area in the claims value chain. Short-Term Disability (STD) and Long-Term Disability claims (LTD) constitute the two most important types of disability claims. These are living benefits and tied to time duration. The benefit must be paid until the disability ceases and a period review must be conducted for closure. 

There are many aspects the insurers must deal with to manage these claims effectively to reduce cost and increase profitability. Some of them are tackling claims that are paying for longer duration, preventing paying claims which are not eligible, getting the true state of claimants receiving disability benefit etc. With the huge number of claims and claim documents, it will not be possible for the claim analysts and the investigating team to get a full view of these.

In the insurance industry, insurers have started investing in Predictive Analytics in all these aspects to increase the efficiency in operations and reduce cost that would drift away due to lack of insights.

Insurers have started investing in Predictive Analytics in all these aspects to increase the efficiency in operations and reduce cost that would drift away due to lack of insights.

Let us look at some of the use cases to get better insights on the claims processed and to predict some of the risks that the claims may pose.

Use of Predictive Analytics

Generally, insurers use traditional ways to analyze their claims and manage them effectively. For this, they are mostly dependent on the Analyst working on the claim. They have teams like Special Investigation Unit (SIU) to take care of fraudulent claims and investigations by accessing the reports and flagging them based on analyst’s judgment. 

It is manually unrealistic to handle the sheer volume of data generated nowadays and to make sense of it. Fortunately technologies which have also evolved that makes it possible to handle this data avalanche. Insurers have started to use predictive analytics to take informed decisions based on insights from multiple sources of data across all their value chains.

According to recently survey conducted by Willis Towers Watson. , approximately 70 percent of large insurers are deploying predictive analytics in the areas of individual life and group life. By mid-2021, 90 percent of large insurers will use predictive analytics for individual life, and a full 100 percent will use for group life. 

Disability claims are offered under both individual and group employee   benefits insurance in all geographies. The use cases, which we will discuss, are more relevant to the US geo employee benefits program but can be applied to other geographies and other individual disability insurance, as well.

Before we move on to some useful use cases, let us look at the disability claim management process and the need for predictive analytics.

Disability Claim Management Process

The diagram below represents the various stages in the disability claim process. We will further discuss some of the use cases for better management of the disability claims, later in this document.

disability claims

  • Claim Intake Process – Completion or denial for want of information
  • Claim Assignation - Work flow process (manual or automated)
  • Claim Adjudication - Analyst evaluation, coordination with medical examiners, physiotherapists and other medical practitioners as required, further evidence records from doctors, medical history records hub, fraud detection. Examples are ReleasePoint, MIB etc.
  • Claim Decision - Approval, denial, litigation consequences, decision support tool
  • Claim Payment - Weekly or monthly, identity theft fraud, check, or EFT
  • Claim Correspondence - Physical, personalization through technology

Use Cases:

1. Improving claim decisions through ‘right claim to right analyst’

With most insurers, claim assignment to analyst is based on the line manager's discretion and availability of analysts rather than the skill set and the experience of the analyst. Due to this, there is a greater risk of delayed decision or less accurate analysis of complex claims leading to escalations, overpayments, losing insights of fraud, etc. Using data analytics we can assign right claims to right analyst and right time.

Data requirements:

  • Historic data of past claims right from assignment to claim closure.
  • Special aspects like fraud, overpayment, and analyst notes that will have occurred on some of the claims.
  • Historic data of the analyst profile to whom each claim was assigned. Usually this will be available out of the claim processing system.
  • Data related to change of users assigned if any on each claim.

Modelling criteria:

The following parameters of the claim can be used to rate the complexity of the claim. The categories can be simple, medium, complex, very complex and so on.

  • The diagnostic codes provide most of the information about the disability.
  • Any co-morbid conditions will have more influence on the disability.
  • The age of the claimant at disability. 
  • Time taken to decide (number of days between assign date and decision date).
  • Duration of past claim to reach closure.
  • Any litigation implication on the data set.

Factors related to the analyst:

  • Number of claims settled by each analyst decision-wise.
  • Current workload.
  • Education level and work experience in years of the analyst.
  • Medical knowledge index of the analyst.
  • Special communication skills (analyst may be required to speak to the claimants).

The data can be obtained from the insurer claim and human resource databases. Using these factors, a model can be built to score the claims based on complexity level and analysts can be categorized based on their expertise such as junior, senior, expert and finally match the score of the claim with the expertise and allocate the claim based on workload.

Benefits:

2. Predicting return to work from disability

Claim duration is a major concern affecting the performance of the insurer. This affects both STD and LTD claims. With respect to STD, the predictability of them ending before the disability period is possible. When the benefit period in STD ends and the claim is unrecovered, from the disability insurer will set up an LTD claim as part of this rollover (if the member has elected for LTD).

Most insurers have automated this conversion process based on the benefit end date. But there are some challenges to in this process. The insurer may not have full information about the current disability condition. So, at times, the insurer may not be required to set up an LTD claim if the claim was returned to work . This further affects claim assignment to the analyst causing unnecessary waste of time on initial scrutiny for LTD eligibility.

Another challenge on claim duration is with respect to claims that would have been paying out for more than two or three years. These will be LTD claims as STD claims are usually limited to 6 months. They are called matured claims and need a different treatment for return to work or, in difficult cases, a settlement of lumpsum benefit in lieu of continued payment for future. 

Predictive models can be developed to identify the STD claims that may roll over to LTD. The time and effort required to recall LTD claimants to work, or in difficult cases, whether any settlement can be offered to claimants to bring the claim to a closure.

Data requirements:

  • History of claim data and member data of both STD and LTD claims.
  • History of intervened cases both for return to work or settlement for treating the model along with medical records and visits data during the disability period; this will shed light on the current level of treatment or claimants’ response to the treatments

Modelling criteria;

  • Past history of the claim data will provide extensive insights about the duration it took for the claimants to return to work, the diagnosis, the age and various medical and physical support that were given to the claimant during the course of the disability to successfully return to work during the benefit paying period.
  • This will also provide insights in cases where the claimants took longer than usual to return to work or cases which are still being paid due to prolonged disability. 
  • Factors like age, cause, diagnostic codes, primary and secondary, educational level will help rate claims into categories based on return to work possibilities. Some claims will be sure of returning, and some may or may not return to work in the expected time and some will be of longer duration. So, the insurer can target the cases where they may return to work as predicted.
  • It is usual for insurers to approach claimants who have a permanent disability with some settlement options where they may be persuaded to accept a lumpsum amount equivalent to a certain percentage of the future payments. Some may decline the offer and some may accept. Analysis of this data can provide insight on the parameters that were prevalent on either of these outcomes.
  • The dependent variable thus found out from the analysis can be used in the predictive model to treat the current claim data.
  • The output of the model can be populated on a dashboard with actionable claims for supporting the claimants to return to work and claimants who are predicted to accept one-time settlements.  

Benefits:

  • The insurer can support the claimants with appropriate medical, physical, psychological factors to bring them back to work. 
  • Facilitating return to work help helps the employer to regain the employee.
  • The insurer can support the STD claim to return to work avoiding rolling over to LTD. For assured cases that will move to LTD, the insurer can plan for allocation of reserve for the LTD claim and plan .
  • Helps the insurer to close the claim and release the reserve amount

3. Predicting claims that can be fraudulent

Frauds in the insurance industry is still a major concern. In disability claims, the general types of frauds include:

  • ID theft and misrepresentation
  • Works and receives benefits
  • Recovered and receives benefits
  • Filing ineligible claims
  • Forged documents

In a report titled State of Insurance Fraud Technology, 2019 conducted by the Coalition Against Insurance Fraud and the SAS Institute, 90% of the respondents said they use technology primary to detect frauds in claims in the insurance industry.

Predictive analytics can help identify many fraudulent cases to take corrective actions. Let us discuss frauds related to pregnancy claims. The fraud here is that the claimant opted for pregnancy coverage after they became pregnant. At the time of claim, it would appear as a normal pregnancy claim but the analyst on scrutiny of the enrollment records, may find that the difference between the date of enrollment for the pregnancy cover and the date of child-birth to be shorter than nine months. Therefore, the analyst may need to flag it and send the claim to the special investigation unit for further investigation.

Data requirement:

  • Historic STD and LTD claims data where pregnancy was the primary diagnosis. 
  • Member data and enrollment data at coverage level.
  • Medical records related to the pregnancy to confirm the disability date (in addition to the documents submitted by the claimant and the physician, data can be accessed from the third party medical hubs).

 Modelling Criteria:

The modeling criteria can be straightforward to find the shortfall of days required for pregnancy claims. However, from past pregnancy fraud claims, the variables that were dominant in those claims will provide better insight to train the model with more accurate algorithms. Sometimes with the connivance of physicians, miscarriages are also reported to benefit from leave and benefit, if the policy is entitled. So, correlating with the medical history records from companies like MIP or Release Point or any provider that the insurer in contract with, we can obtain clear data patterns to train the model.

Systems that identify anomalies in a database m be used to develop algorithms that enable an insurer to automatically stop claim payments. An insurance technology expert said that this approach has produced 20 to 50% reductions in fraud loss for some insurers – Insurance Factbook 2019

4. Predicting Claims that are prone to litigation

When claims involve litigation, the cost goes up, extends the cycle time, and more resources at the insurance office are involved in the claim. In most cases, the attorney’s involvement happens on denied claims. Claims related to workers compensation and claims of long-term disability in nature, may have risk of litigation due to their complexity. In addition, if attorneys get involved in such claims, the cost for the insurers may increase.

The insurance industry has recently been leveraging analytics to predict claims that may become litigious. so that they intervene early or settle with the claimant to avoid litigation and learn upfront the probable cost of litigation.

Over the six-year period studied, attorneys were involved in 12% of all claims (including medical-only cases) and 80% of permanent disability claims

-- the California Workers’ Compensation Institute

Data requirements:

  • Past count of claims that went into litigation.
  • Claim history of the litigative claimant.
  • Insurer attorney data and case judgments.
  • Opponent lawyers record of accomplishment.
  • Social media data associated with the claimant.

Modeling criteria:

Based on the analysis of past claim data with litigation, the following factors on their own or in correlation with the others may highly influence the claimant to seek legal assistance when the claim is denied.  

  • Age, marital status, pre-disability earnings, litigation history and social connect of the claimant.
  • Standard and non-standard data from the legal diaries of past litigation claims also will help.