Validating Health Economics Models in R Exercise

Reference Papers

The Chemotherapy Model for "Value of Information for Health Economic Evalulations", Heath, Kunst and Jackson (2021).

This model is structured based on "A Need for Change! A Coding Framework for Improving Transparency in Decision Modeling" by Alarid-Escudero, F., Krijkamp, E., Pechlivanoglou, P., Jalal, H., Kao, S., Yang, A. and Enns, E.

This model has been adapted from a model presented in "Calculating the Expected Value of Sample Information Using Efficient Nested Monte Carlo: A Tutorial" by Heath, A. and Baio G. This will be used as the running example throughout the book.

Instructions and Code File Structure

For R users, after downloading the entire folder, double-click the Rproj file named Chemotherapy_Model_book_mistakes.Rproj. This will set up your working environment automatically via the Rproj file.

The folder structure is as follows:

• 01_data_raw: Contains raw CSV data files.
• 02_data: Contains model inputs derived from raw data and distribution parameter assumptions.
• 03_R: Contains scripts for miscellaneous functions and model functions.
• 04_analysis: Here, we run the chemotherapy model, conduct analyses, and create plots using 01_model_run.R.

Output directories:

• 05_output: Stores output files.
• 06_figs: Stores figures.
• 07_tables: Stores tables.

Model Inputs Tables: R_HTA_Tables.pdf

Errors

There are 8 errors in the model located within the scripts:

• 02_data/02_Assumption_Inputs.R
• 03_R/01_misc_functions.R
• 03_R/02_model_functions.R.

Please note that there's no error or warning messages will appear if you run the model using the script 04_analysis/01_model_run.R.

Hints

Hints and code are located in 04_analysis/02_hints_to_detect_errors.R.

• General Hints:

  # Command to check generated PSA parameters:
  apply(m_params, 2, mean)
  round(apply(m_params, 2, sd), 3)
  
  # Command to check the result dataframe:
  apply(df_e, 2, mean)
  apply(df_c, 2, mean)

• Hint 1: Compare definitions in 02_data/02_Assumption_inputs.R with R_HTA_Tables.pdf.

• Hint 2: Beta distribution's parameters, alpha and beta, should both be greater than 0.

• Hint 3: Set u_recovery to 1 and rerun the model.

  m_params$u_recovery <- 1

• Hint 4: Please check results' reproducibility by repeating 01_model_run.R.

• Hint 5: Check the transition probability matrix in 03_R/02_model_functions.R.

  Two functions you might consider (from the darthtools package):

  # Checks if each row of the transition matrices sums to one.
  check_sum_of_transition_array(MM.mat)
  # Checks if transition probabilities are within [0, 1].
  check_transition_probability(MM.mat)

• Hint 6: If there are potential problems with the transition probability matrix, try to check each probability element of the transition matrix.

• Hint 7a: Check the PSA scatter plot output plot.

  plot_psa(l_psa, txtsize = 13)

• Hint 7b: Also, try setting c_treatment_2 to another value and check the model's output.

  l_params_all$c_treatment_2 <- 350000