Objective 4

Question 1

Risk factors that indicate whether a person may have high claims

Answer 1

1. Inherent risk factors: age, sex, race
2. Medical condition-related factors
3. Family history for inheritable conditions
4. Lifestyle risk factors: smoking, exercise, nutrition
5. External risk factors: industry, location, education

Question 2

Types of medical management interventions

Answer 2

1. Care coordination (system): case management, discharge planning, in-hospital care coordination
2. Condition management (patient): disease management, risk factor management
3. Provider management (provider): provider profiling, pay-for-performance, ACOs

Question 3

Areas where condition-based models are used in healthcare financial applications

Answer 3

1. Program management: ID high-risk individuals, financial modeling and resource allocation, program evaluation (savings)
2. Provider or health plan reimbursement: normalizing populations (for payment or evaluation), profiling providers
3. Actuarial and underwriting functions: pricing, underwriting, projecting future claim costs

Question 4

Types of predictive models that are not based on medical conditions

Answer 4

1. Age/sex
2. Prior cost
3. Combo of age/sex and prior cost

Question 5

Sources of data for developing risk factors

Answer 5

1. Claims data
2. Self-reported data
3. External data

Question 6

Risk factors identified by a health risk assessment

Answer 6

1. Personal disease history
2. Family disease history
3. Health screenings and immunizations
4. Drinking
5. Smoking
6. Sunscreen
7. Nutrition
8. Exercise
9. Weight management
10. Injury prevention
11. Stress and wellbeing
12. Women’s health
13. General health assessment
14. Functional health status
15. Mental health status

Question 7

Types of data used for predictive modeling

Answer 7

1. Physician referral/chart (high reliability / low practicality)
2. Enrollment (high / high)
3. Claims (medium / high)
4. Pharmacy (medium / high)
5. Lab values (high / low)
6. Self-reported (low / low)

Question 8

Drawbacks of using data from medical charts

Answer 8

1. Don’t cover out-of-network services or drugs prescribed by out-of-network providers
2. Don’t record patient compliance
3. Transcribing / transferring to uniform format is time consuming and requires highly-trained staff
4. Lack of uniformity in physician coding of conditions and severity
5. Typically unavailable to health plan or actuary

Question 9

Advantages and disadvantages of using diagnosis codes for identifying member conditions

Answer 9

Advantages
1. Almost always present on medical claims
2. Uniform
3. Useful for identifying conditions
Disadvantages
1. Generally only primary and secondary are populated
2. Coding errors
3. Upcoding
4. Different physicians follow different coding practices

Question 10

Drawbacks of using survey data

Answer 10

1. Must be commissioned, budgeted, executed
2. Isn’t updated as medical events occur - can become stale
3. Response bias
4. Untruthful answers

Question 11

Questions to answer when building a clinical identification algorithm

Answer 11

Set of rules applied to claims to identify conditions

1. Where are diagnoses?
2. Source of diagnosis?
3. If claims, what claims to consider (IP/OP/Lab)?
4. If more than one diagnosis, how many to consider?
5. Over what time span and how often must diagnosis appear?
6. What procedures for identifying severity?
7. What drugs for identifying conditions?

Question 12

Challenges when constructing a condition-based model

Answer 12

1. Large number of procedure and drug codes
2. Severity level at which to recognize condition
3. Impact of co-morbidities for conditions often found together
4. Degree of certainty
5. Extent of data
6. Underlying benefit design

Question 13

Definitions of sensitivity and specificity

Answer 13

1. Sensitivity - percent of true positives

2. Specificity - percent of true negatives

Question 14

External sources of clinical identification algorithms

Answer 14

1. HEDIS from NCQA
2. DMAA (for chronic diseases)
3. Grouper models
4. Literature

Question 15

Reasons for using commercially-available grouper models

Answer 15

1. Lots of work to build from scratch
2. Must be maintained to add new codes - even more work
3. Commercially-available models are accessible to many users. May be a requirement for providers and plans

Question 16

Common features of Medicare prospective payment systems

Answer 16

1. System of averages
2. Increased complexity
3. Relative weights
4. Conversion factors
5. Outlier payments
6. Updates
7. Focus on access and quality

Question 17

Challenges with patient classification systems based on coding systems

Answer 17

1. Need for new DRGs
2. ICD coding
3. Upcoding
4. New coding systems

Question 18

Factors for choosing the right predictive model

Answer 18

1. Correlation structure
2. Purpose of analysis
3. Nature of available data
4. Characteristics of outcome variable
5. Distribution of outcome variable
6. Functional relationship
7. Complex decision model

Question 19

Steps of the data warehousing process

Answer 19

1. Identify patients to include
2. Identify data elements to merge on
3. Create flags
4. Attach derived variables and flags

Question 20

Characteristics for assessing the quality of a model

Answer 20

1. Parsimony
2. Identifiability
3. Goodness of fit
4. Theoretical consistency
5. Predictive power

Question 21

Statistics for determining whether a model is good

Answer 21

1. R^2 and adjusted R^2 = 1 - (1 - R^2) * (N - 1) / (N - k - 1)
2. Regression coefficients (signs / significance)
3. F-Test
4. Statistics for logistic models:
   a) Hosmer-Lemeshow
   b) Somers’ D
   c) C-statistic
5. Multicollinearity
6. Heteroscedasticity
7. Autocorrelation

Question 22

Re-sampling methods for validating a model

Answer 22

1. Bootstrap (sampling with replacement)
2. Jackknife (leave out 1 observation at a time)
3. Cross-validation (subsets held out)
4. Permutation test (rearrange labels)

Question 23

Factors used in developing risk scores in the CMS-HCC risk model

Answer 23

HCC = Hierarchical condition category

1. Demographics - higher for duals
2. Disabled
3. Separate models for LTC / ESRD
4. New enrollees - separate age/gender only factors
5. Prospective risk-adjustment methodology
6. Calibration every 2 years
7. Health status factors from ICD-9 codes grouped into HCCs

Question 24

Central features of Massachusetts health care reform

Answer 24

1. Exchange
2. Employers must establish Section 125 accounts
3. Large subsidies for families under 300% FPL
4. More limited plan available for families over 300% FPL
5. Individual mandate
6. Funding through federal funds for safety net hospitals / uncompensated care

Question 25

Steps for developing predictive models for care management programs

Answer 25

1. Choose a disease or condition
2. Rank conditions based on intervenability
3. Identify population
4. Plan the intervention
5. Perform economic modeling of program
6. Develop the predictive model
7. Test actual outcomes against predictions, modify model and program as needed

Question 26

Metrics that should be recognized in the Risk Management Economic Model

Answer 26

1. Number and risk-intensity of members targeted
2. Types of interventions used
3. Number of nurses / staff, program costs
4. Methodology for contacting and enrolling members
5. Rules for integrating the program with the rest of the care management system
6. Timing and number of contacts, enrollments, interventions
7. Predicted behavior of target population without intervention, predicted effectiveness of intervention at modifying