Individual Insurance Chap 4: Managing Anti-Selection Flashcards
Introduction
Key to managing antiselection is having an accurate model to predict the impact of antiselection on blocks of policies. Occurs when insureds make choices to maximize the value they receive in return for what they pay
The Three Faces of Antiselection
- External
- Internal
- Durational (Cumulative)
1. “Antiselection is that annoying tendency people have of doing what’s best for themselves”
2. Antiselection can be specific (an optional rider to cover maternity) or general (someone perceives himself as unhealthy)
- This chapter focuses on general antiselection
3. Antiselection can occur when the person is first issued, while they are insured, and as they make decisions about when to end the contract
- These are called external antiselection, internal antiselection, and durational antiselection respectively
4. External
- When the person is first becoming insured or is first seeking insurance. E.g. when a risky person
knows they’ll likely need coverage and seeks insurance
5. Internal
- While the person is insured. E.g. when the prices of coverage change, the healthy will choose the cheaper option and the impaired risks will choose the more expensive option (better coverage)
6. Durational (Cumulative)
- When the person makes a decision about whether or not to end the contract. E.g. when a large
rate increase causes some to leave the coverage and the higher risk individuals to stay
Underwriting the Individual Risk (External Antiselection)
- Mechanisms to control Antiselection
- Types of Underwriting
- ACA Consideration
- Types of Underwriting Tools
- Privacy
- Analysis of Issues
- Underwriting Actions
1. Controlled by insurers through a variety of mechanisms including
- Individual (medical) underwriting before issue
- Policy provisions that exclude or limit coverage due to pre-existing conditions
- Requiring an enrollment mechanism that doesn’t permit antiselection
▪ Example: minimum participation percentages for associations
2. Underwriting begins with the sales representative who initially contacts the applicant
- Some coverages rely on sales rep to qualify the prospect
- The internet is a new distribution portal, but eliminates the opportunity for the sales force to act as an underwriter
3. Types of underwriting (primarily used in the life insurance environment):
a. “Medical” underwriting means an examination by a physician as a condition to receive insurance
b. “Paramedical” underwriting means the exam can be by a paramedic
c. “Nonmedical” means there are no exams, just medical questions
d. However, in the health insurance world, “medical underwriting” means there are medical questions on the app
4. Under ACA, underwriting and pre-existing condition exclusions are not allowed for major medical insurance
a. Guaranteed issue
b. Rates cannot vary based on the health status of the applicant
5. Underwriting tools
a. The individual application
▪ Ask for identifying information, financial information (especially for disability coverage), and medical history questions
▪ Typical questions to include asking whether the insured has ever:
* Been diagnosed with a list of particular conditions
* Had particular symptoms
* Seen a doctor or had medication prescribed for particular conditions
▪ May be additional questions relevant to specific coverages
▪ Under ACA, insurers are not allowed to ask about medical conditions any more
b. Attending physician statement (APS)
▪ Any application which isn’t entirely “clean” (no conditions or symptoms) requires follow- up by the insurer
▪ Most common way to obtain further information is to require an APS from physicians on the application
c. Commercial databases
▪ Prescription drug, medical information bureau (MIB), and other databases are tools for identifying undisclosed conditions
▪ Final underwriting usually cannot be based on database alone
d. Internal data
▪ Sometimes insurer’s own records (especially previously submitted applications) be a source for additional information
e. Telephone interviews
▪ Often used in place of APS or other third party information sources
▪ Insureds are much less willing to mislead an insurer’s representative on the phone than on a written application, so this is a good way to gain additional information
f. Inspection reports
▪ Information obtained from the applicant
▪ Includes telephone interviews, but also in person and through other sources
▪ Most commonly used in life and disability
g. Lab testing
▪ Some insurers use blood, saliva, or urine testing to identify use of tobacco, illegal drugs, and some medical conditions
▪ Too expensive to be used on all applicants, but may be used with certain subsets (demographic, geographic, etc.)
h. Medical exams
▪ High cost, but still used in some situations such as for high amount disability coverage
i. Tax returns
▪ Financial information given by insured can be hard to evaluate, so tax returns of individual and business are probably the best source
o Pre-existing condition provisions
▪ Not used by medical underwriters, but rather to provide blanket protection from antiselection
▪ The pre-existing condition provision states that any condition for which there was treatment or symptoms for up to 6-12 months prior to the application will not be covered for 12-24 month after issue
▪ Can be used in place of more rigorous underwriting (hospital indemnity) or as a supplement to traditional underwriting (individual major med)
▪ Consult statutory limitations on ability to define and use pre-ex clauses
6. Privacy
a. HIPAA contains strict provisions regarding security and privacy of individual health data
▪ This includes underwriting information
b. May require re-vamping of information flows and construction of physical and software protections
7. Analysis of Information
a. Once information is obtained, tools are needed to put information together in a meaningful way
▪ Debit manuals
▪ Conditions are assigned points, and the overall point value of the applicant guides the underwriting action
▪ Some conditions may be broken down further
* A mild cancer 5 years in remission may result in no load, while stage four cancer currently in treatment is uninsurable
▪ Remember that some conditions may already be accounted for through age rating
b. Genetic testing
▪ Politically controversial and not presently done in health insurance underwriting
c. Predictive models
▪ Any model that provides a prediction of future claims
▪ Becoming commonplace and much work is being done to improve and measure accuracy
▪ Examples include:
* Credibility formulas that blend expected claims with past experience
* Modified credibility formulas based on a linear combination of past subsets of claims; For example, inpatient, outpatient, other
* Diagnosis-based risk adjusters
* Debit manuals
▪ Risk adjusters continue to improve, but they cannot replace medical underwriters as they
are not sophisticated enough to handle potential antiselection of an individual
8. Underwriting Actions
a. After accumulating information, a decision needs to be made on whether to offer coverage
b. One extreme is to offer full coverage with no restrictions and another extreme is to decline coverage completely
c. However, there are also several responses in between, including:
▪ Offer coverage at a higher premium rate
* Usually a multiple of the standard premium
* Can be temporary or permanent
* Most carrier will consider removing later if risk profile improves
▪ Offer a standard policy, but exclude coverage for that specific condition or affected body system
* Accomplished through a “waiver”, “impairment”, or “exclusion” rider
* Useful for defined and isolated conditions, but not for systemic conditions (like obesity)
* Viewed negatively by regulators because it excludes coverage for which insured has greatest need, but may provide opportunity for insurance that might
otherwise not be available at all
▪ Offer a different policy or plan than the one applied for
* Some carriers have separate pools for substandard risks
* Blue Cross plans often use this approach
* Other carriers offer limited benefit plans
▪ Offer a different plan of benefits than the one applied for
* Used where applicant has a low cost condition (i.e. asthma)
* Common in disability or long term care to offer longer elimination period or shorter benefit period
d. Limitations of underwriting action
▪ For med supp, there is a six month open enrollment period for new enrollees of Medicare Part B
* No declines are allowed
* Different prems based on health status are not permitted in some states
▪ For major medical, HIPAA puts requirements on states to find a way to insure those who lose coverage
e. Pre-ex investigation, material misrepresentation, and rescission
▪ Most carriers have rigorous investigation process to uncover cases where applicant has not disclosed conditions at time of application
▪ Criteria to be used in deciding when to conduct an investigation:
* Timing
- Insurer has a limited amount of time after issue to rescind because of bad information on app
- Exception is made if insurer can prove fraud, but this is harder to prove
- If limit has passed, insurer may be less likely to investigate claims
- Conditions
- Some conditions can be excluded from being pre-existing (accidents, for example)
- Size
- Insurer would not want to investigate claims if cost of investigation exceeded cost of claim
- However, insurer may not forego acting on knowledge that a claim was pre-existing, as this may result in losing legal right to rescind later
- Sentinel conditions or procedures
- Some conditions related to others lend themselves to antiselection
- Example: A PCP claim may be cause for an investigation of preexisting HIV
▪ Possible outcomes of investigation include
- Reformation
- Reissue retroactively under terms which would have been applied had insurer been aware of condition
- Rescission
- Declare a policy void from the beginning
- May lead to public relations and litigation risks
- ACA prohibits rescissions of health insurance unless insurer proves fraud or material misrepresentation.
▪ Material misrepresentation occurs when claimant provides false information on the app
* Some jurisdictions requires claimant must know information is false
▪ Drug history is often a key avenue to obtain undisclosed medical information
Changes in Plan (Internal Antiselection)
1. When policyholders are provided opportunities to antiselect, they will generally do so
2. Premium leakage
a. Insurers allow policyholders to change to higher deductible at renewal
b. Higher risk policyholders are much less likely to opt for higher deductible plans; they will keep the richer benefit, creating antiselection
c. Premium leakage = the difference between (the expected premium if a random sample of the population migrated to the leaner plan) minus (the actual premium that results from internal antiselection)
2. The buy-down effect
a. Common with coverage with high or frequent rate increases
b. Insurer never receives the full value of the premium increases they put in because
▪ Older policyholders with higher premiums go to Medicare and drop coverage
▪ Policyholders look for a way to reduce premium and switch to lower cost plan
c. Due to antiselection, claims generally do not drop as much as premium
3. Modeling and managing internal antiselection
a. Best tool to measure is a partitioning model – separating the population into subsets based on perceived health status
b. Monitor how different groups change their benefit plans over time
c. Under ACA, individuals who develop medical conditions can change to richer benefits at next open enrollment without additional underwriting. Increases potential internal antiselection
Antiselection Upon Lapse (CAST/Durational Antiselection)
1. More persisting policyholders will be high risk than low risk because high risk individuals are:
a. Less likely to find coverage elsewhere (no longer true post-ACA)
b. Less likely to want to become uninsured
c. Emotionally less willing to change their current insurance situation
* Large rate increases that increase lapse rates will also increase antiselection
* Result is the Cumulative Antiselection Theory (CAST) model, which adjusts for mix shift towards more high risk policyholders
* This effect can be measured by looking at experience over the durational dimension, rather than over the
calendar dimension
Modeling Antiselection
1. The partition model
a. Can be thought of as having population get into a long line from the lowest to highest cost members
▪ Note that this model is an exaggeration in that members cannot perfectly predict their claims
b. Once line-up occurs, we partition into healthy and unhealthy insureds
c. But where do we draw the line between healthy and unhealthy?
▪ Drawing the line in a CAST model
* Cutoff is chosen so the ratio of the unhealthy to healthy claim cost is a fixed multiple
* Often a higher multiple (like 5-10) is chosen
▪ Drawing the line in a MNAM model
* Model tried to find boundary conditions on antiselection that might occur in a specific situation and then use the boundary conditions to estimate antiselection expected under different scenarios
▪ Drawing the line in internal antiselection
* After line-up and partition, we choose or derive low risk and high risk subsets, priced separately
* Involves several assumptions deduced from results
▪ Deterministic vs. stochastic models
* The typical use of an antiselection model is to take history, define it in detail, and project it into the future using deterministic methods
* A true picture requires we work with distributions, not the expected values themselves provided by stochastic models
▪ Markov processes
* There is no reason why a partition need to be limited to two subsets
* If further partitions are done, model lends itself to being treated as a Markov process
2. Projecting the model
a. Projection model requires assumptions about how each of the following variables will act on the population over time
▪ Trends in claim costs
▪ Lapsation
▪ Movement between populations
▪ Time value of money
▪ Premium rate increases
b. CAST model
▪ Provides an important tool for understanding behaviors
▪ May not be a good fit in these circumstances:
* The first 3-4 durations, when wear-off of underwriting overwhelms the CAST effect
* Later duration where only a fraction of the original population remains
* All durations when a rate spiral is severe and volatile
▪ The specific factors used when applying the CAST model are very important, but can be difficult to calibrate
The Future
More work needs to be done to refine modeling of antiselection, but much progress has been made and existing models are far superior to prior method used by actuaries of “Let’s just assume X% antiselection”
Antiselection Under the ACA
- The ACA prohibited most traditional techniques insurers have used to control antiselection in individual major medical. The following were all prohibited starting in 2014:
a. Underwriting, including offering alternative coverage or denying coverage; o Health status rating;
b. Pre-existing condition exclusions;
c. Exclusionary riders;
d. Lifetime or annual dollar limits;
e. Limiting benefit coverage or imposing very high cost sharing designed to attract healthier risks;
f. Rescissions, except in cases of fraud or intentional misrepresentation; and
g. Marketing practices that discourage unhealthy risks from signing up. - In the absence of further protections, individuals would:
a. Wait until they need care to purchase insurance (external antiselection)
b. Switch whenever they chose to the plan that provided the best coverage for their conditions (internal antiselection)
c. Drop coverage as soon as they were well again (cumulative antiselection in its most extreme form)
- ACA measures to control antiselection:
a. Coverage mandates (in the form of tax penalties) and premium subsidies to encourage participation
b. Aligning market rules on and off the Exchanges
c. Open enrollment periods
d. Minimum benefit levels
e. Premium stabilization programs (known as “the three R’s”). Each measure is examined in more detail below
- Coverage Mandates and Premium Subsidies
a. ACA requires large employers (with 51 or more employees) to offer affordable insurance coverage meeting a minimum coverage level. Employers that fail to do so may have to pay a significant penalty
b. ACA requires all individuals to obtain insurance that provides minimum essential coverage or pay a penalty on their tax returns. This is known as the individual mandate. Many forms of coverage qualify as minimum essential coverage:
▪ Medicare
▪ Most Medicaid plans, including CHIP
▪ Certain other public programs such as TRICARE (military coverage)
▪ Employer sponsored insurance that meets minimum value
▪ Individual major medical insurance
▪ Certain other coverage recognized by CMS
c. Requiring all Americans to obtain coverage greatly reduces antiselection in the individual market, but only if
▪ Individuals believe the penalty is significant in relation to the cost of coverage,
▪ The requirement applies to a broad enough population segment, and
▪ It is enforced
d. Individual mandate penalty – for each tax filing household, the penalty is the greater of a flat per person penalty or a percentage of all income over the filing threshold for the household
e. The flat amount is limited to a maximum of three times the individual per-person amount. The total penalty for a household cannot exceed the national average premium for a bronze qualified health plan
f. There are a variety of exemptions from the individual mandate.
▪ Individuals who have incomes below the tax filing threshold are exempt, as are individuals who have suffered a hardship
g. The mandate allows individuals to have a gap in coverage of less than three months during a year without having to pay the penalty. This could lead to selection where healthy individuals drop coverage for a few months each year and then sign up again in the next open enrollment
h. The premium subsidies to lower income individuals help reduce antiselection by making coverage more affordable
- Aligning rules on and off exchanges
a. Most insurers have both ACA-compliant and non-compliant policies on their books
▪ ACA-compliant policies may either be on-exchange or off-exchange
▪ Non-compliant policies may be either grandfathered or transitional
b. ACA-compliant plans
▪ Effective Jan 1, 2014 or later which comply with all ACA rules
▪ Sold either through a public exchange or directly by the insurer or its agent
▪ To mitigate selection risk
* An insurer’s identical plans must have identical rates on and off the exchanges
* Insurers must participate in the risk adjustment mechanism
* Same commissions to brokers and agents on and off exchanges
* The exchange fee must be spread across the entire single risk pool, including off-change policies
* Carriers must offer at least one gold and one silver level plan on the exchange.
There is no equivalent to this requirement off-exchange, which could lead to selection
* Prohibit marketing intended to discourage unhealthy individuals from signing up
* Open enrollment periods are identical on and off the exchange
c. Grandfathered policies
▪ Were in existence when the ACA was signed March 2010
▪ New grandfathered policies cannot be sold
▪ Exempt from most ACA requirements
d. Transitional policies
▪ Were sold after ACA passed, but before major reform changes took effect in 2014
▪ Exempt from many ACA requirements and they are rated separately from compliant plans
▪ Do not participate in the risk mitigation programs
▪ States can allow an extension of transitional policies up to renewals of October 2016 and potentially another year beyond that
e. Healthy policyholders are more likely to keep the non-compliant coverage, while unhealthy have an incentive to switch to the ACA-compliant pool
- Open enrollment periods
a. Limits opportunity for antiselection by requiring members to enroll during a set time each year
b. Once a plan is selected, it cannot be changed until the next year’s enrollment period
c. Exceptions are granted for life events such as a birth, adoption, or loss of other coverage
- Minimum benefit levels
a. In the individual and small group markets, policies must cover all essential health benefits at least at bronze actuarial value (AV). Catastrophic plans can be les rich than bronze, available to those to qualify
b. In the large ER market, plans must meet a minimum value of 60%
c. Minimum value = similar to AV because estimates the percentage of claim costs covered by the ER plan. Differs from AV because large ERs are not required to cover all essential health benefits
- The 3 R’s
a. 3 premium stabilization programs apply to ACA-compliant individual and small group markets
b. 2 programs are temporary (transitional reinsurance and risk corridors)
c. Risk adjustment program is permanent
- Federal
a. Temporary from 2014 through 2016
b. Reimburses insurers in ACA-compliant individual market for a portion of high claimants’ costs
c. Funds collected using a per capita fee primarily from the group major medical market (including self-funded employer groups)
