Lists Flashcards
Application of Risk Adjustment
(PROBE)
Provider profiling
Risk management
Outcomes of health care measurement
Budgeting (Pricing)
Efficiency measurement
Characteristics of a good model
VARIABLE CRISP CARD
Valid
Adequately documented
Rigorous
Inputs to parameter values appropriate
Arbitrage free
Behaviour reasonable
Length/expense of run not too long/high
Easy to understand
Communicable re workings & output
Reflects risk profile of contracts modelled
Independent verification of outputs
Sensible joint behaviour of variables (dynamics)
Parameters allow for sign features
Clear results
A range of implementation methods
Refineable
Developable
Assumptions Required
RIM PINT CREW
Risk discount rate
Investment returns
Mortality / Morbidity rates
Profitability requirements
Inflation
New business volume & mix (demographic, financial)
Tax rates
Commission
Reserving basis
Expenses
Withdrawals
Product Design & Pricing - Go to
CRUDS CA
C - Customer acceptability (clear in benefit, amount and variability of prems)
R - Regulator requirements
U - Underwriting methodology
D - Distributors needs
S - Systems and other internal constraints
C - Company culture in style & price —> consistency with other products
A - Adequate profitability/ return on capital
GLM: Suitability of explanatory variable
- Relationship with the response variable
- Quality of Data
- Quantifiable in a discrete way
GLM: Disadvantages of One-way analysis
- looks at effect on frequency and severity of each rating factor separately
- it ignores correlation and interaction effects between variables, for e.g. age and disease
- as a result, may underestimate/ overestimate the effect of variables correlated to one another
GLM: Assumptions of classical linear models
- error terms are independent and come from a normal distribution
- mean is a linear combination of the explanatory variables
- error terms have a constant variance
GLM: Drawbacks of the normal model for multiple linear regression
- assumes that the response variable, Y, has a Normal distribution which may be not true for the variable being modelled
- assumes constant variance, which may not be appropriate for the variable being modelled
- adds together the effect of different explanatory variables, seldom practical
- with more than two explanatory variables, a manual solution becomes increasingly long winded.
GLM: GLM addresses some of the linear modelling shortcomings/ Benefits of GLM
RL SCAAH
- Response variable can take any distribution from exponential family
- Link function introduced:
- removes the assumption that the effects of different variables must simply be added together.
- Stable transitions between levels of risk
- Control over interactions considered
- Assess different combinations of explanatory variables
- Accounts for the effects of other explanatory variables in calculation of effects sizes
- Handles risk cells with small volumes, uses all data
GLM: Characteristics of a Link Function
- It must be both differential and monotonic
- Usually include; log, logit and identity functions
- Log link function is of particular interest in pricing —> used for results in a model where the effect of different rating factors are multiplied together
GLM: Properties of the exponential family
- Distribution is completely specified in terms of mean and variance
- the variance of Yi is a function of its mean
GLM: The Tweedie distribution
Direct modelling of pure premium or incurred loss data for PMI business is problematic
A typical pure premium distribution will consist of a large smile (i.e. a point mass) at zero (where policies have no claim)
And then a wide range of amounts (where policies have had claims)
Tweedie dist. can handle this well
GLM: Pitfalls of using GLM vs One-Way
- if influential points affect coefficients, the impact spreads beyond the single cell that the influential point lies in
- potential for model error if not specified correctly
-requires some statistical understanding to be able to use
-might not capture correct shapes of relationships - one-way may call out areas of concern that might be overlooked
GLM: Methods for testing the appropriateness of model
Deviance residuals - measures the distance between the observation response & the fitted values
Standard Pearson - the difference between the observation response & the predicted value, adjusted for the standard deviation of the predicted value & the leverage of the observed value.
Residual plots - plots of residual against fitted values, which should be symmetrical about the x - axis & fairly constant across the width of fitted values & average residual of zero
Cook’s distance - used to test the influence of a data point in models results, if cook’s distance >=1 should be investigated —> excluded or capped
What determines the cover limit
(SCR T)
Size of scheme - no. of lives covered
Compulsory or Voluntary membership
Required take up rate/ proportion taking up cover if voluntary
Total & average sum insured
Data sources
PROMOTAR
P - population data (provided by government)
R - reinsurer’s data
O - own data of company
M - Market data (including insured lives data and published returns)
O - overseas data
T - trade magazines
A - actuarial consultant’s data
R - rate table software
Data Challenges
- Outdated
- Contain errors
- Incomplete - missing entries
- Not sufficient to be divided into credible homogenous groups (SA, Occupation, location)
- Not enough data points
- Not available by various types of risk cells
Aims of Managed Care
- Manage claims costs (Reduce it)
- Maintain/ improve quality health care
- Provide insurer with some control over HC service providers through provider networks
- Reduce unnecessary utilisation
- Manage high risk members
- Ensure medical services are provided in an appropriate setting
Risks of managed care
- Provider networks restricts access to care (lower utilisation)
- Providers may resent external parties imposing clinical protocols on them and influencing the way they practice medicine
- Protocols used unfairly:
—> May compromise the quality of care provided by encouraging under servicing of clients - The use of formularies and other financial based managed care incentives may result in additional costs being transferred to ph with no overall cost reduction
- Biases and discrimination ( selecting healthy and young lives)
- Price, intensity, severity, frequency & actuarial risk due to managed care.
Examples of managed care techniques/ strategies
TRRRP HCF
T - Treatment protocols —>governs the treatment & medicines that a member gets access to for certain conditions
R - Reimbursement methods
R - Risk sharing i.e co-payment
R - Referrals (GP to Specialist) —> will only cover the full specialist fee if referred by GP to reduce unnecessary specialist visits
P - Provider networks
H - Hospital pre-authorisation
C - Cases and disease management
F - Formulary medicines (generic medicine)
Characteristics of provider networks
Negotiated prices
Insurer has better control on procedures taken I.e. treatment protocols
Able to monitor data
Able to check efficiency of each provider
Minimise fraud
Able to incentivise provider for not being wasteful/ for being under budget
Policyholders may not have access to provider networks, leading to a decrease in sales
Value add services
Considerations when deciding on the risk factors:
Data easily obtainable
Objectives of factors chosen
Verifiable data
Not politically sensitive
Cognisant of particular features of the country
Parsimonious (captures as much information as possible in a few as possible features)
Balance between demand-side and supply side factors
Relevant & up-to-date
Inexpensive to collect relevant data