Building a Healthcare Model

It is important to take a deliberate approach to building, validating and analyzing a health economic model. Here are recommended steps.

What are you trying learn from your health economic model?

You cannot build a useful and accurate model if you don’t fully understand your goals.
Make sure you are clear on all the points below before you start build a model.

  • Population: who is affected by the health issue?
    • People diagnosed with a specific type of cancer.
    • Women with high risk pregnancies.
  • Interventions: what are our treatment and/or diagnostic options available?
    • Radiation vs surgery vs radiation + surgery.
    • Screening options for breast cancer.
    • Frequency of colonoscopies.
  • Perspective: who’s perspective are we considering in our decision?
    • Healthcare payer vs societal vs patient-only.
  • Time horizon: how long should we follow patients?
    • Short-term vs lifetime.
  • Outcomes: what outcomes will we use to make our decision.
    • Usually cost per QALY for cost-effectiveness analysis.
    • Are there additional outcomes we want to report?
      • Deaths or infections avoided.
      • Healthy births.
      • Cases detected.

What kind of model do you want to build?

Choose the model type that best matches with your goals.

  • Decision Tree: for short-term outcomes.
    • ER visit.
    • Orthopedic surgery.
  • Markov Model: for long-term disease progression with important events.
    • Cancer treatment with adverse events.
    • Asthma treatment.
  • Partitioned Survival Model: for long-term disease progression based on survival curves.
    • Cancer treatment.
  • Patient Simulation: for disease progression with heterogeneity and important patient history.
    • Colorectal cancer with screening based on prior screening results.
    • Subgroup analysis on patient characteristics.
  • Discrete Event Simulation: patient simulation based on event time sampling.

Build Patient Pathways for Each Health Strategy

To generate accurate results, your model will mimic the the patient experience.

  • Decision & Strategies: your model will start with a decision node with a branch for each health strategy (treatment, diagnostic, etc.).
    • Each strategy is built and analyzed separately to determine the average outcomes per patient.
  • Patient Pathways: patients pathways represent the full set of events and health statuses that a patient could experience.
    • Patient pathways split at chance nodes with probabilities representing the likelihood a patient would proceed down each branch.
  • Markov Model: structure consists of health states, events and cycles that represent disease progression over time.
    • The overall time horizon is broken down into time cycles to reuse the structure into the future.
    • Health states represent the current status of a patient.
    • From each health state, a branching set of events represent what can happen in a cycle, ending with a return to the health states for the next cycle.
  • Partitioned Survival Model: structure consists of survival curves that define health state membership.
    • Typically two survival curves (PFS, OS) are defined by survival functions describing declining state membership over time.
  • Patient Simulation: run simulated patients through the same structure as a Markov model.
    • This provides greater flexibility through the addition of patient characteristics and patient history to impact disease progression.
  • Discrete Event Simulation: structure is similar to Markov models except movement through events is driven by sampled event times.

Add Data Inputs

Your model requires data for disease progression, costs, utilities, etc. to generate results. In TreeAge Pro, you reference input data specifically where it should be applied in the model.

  • Disease Progression: this data informs your model on how the disease manifests itself in the target population.
    • Markov: transition/event probabilities.
    • Partitioned Survival: survival and/or hazard functions.
    • Discrete Event: event time distributions.
  • Costs: costs associated with treatment, diagnosis, hospitalization, etc.
  • Utilities: indication how healthy you are in each health state.
  • Others: test sensitivity/specificity, hazard ratios, etc.

Validate Your Model

You need to validate the accuracy of your model before you can rely on its results.

  • Visualize Disease Progression: examine disease progression over time to ensure it is consistent with clinical data.
  • Validate Outcome Calculations: examine how each outcome is accumulated over time.
  • Patient Trace: Examine patient trace at the cohort level and at the patient level for simulation models.
  • Debug Complex Calculations: review complex calculations to ensure validity.

Validate Your Model

You need to validate the accuracy of your model before you can rely on its results.

TreeAge Pro Healthcare Features