Line 323 was corrupted (missing '= ' in variable assignment)
causing: NameError: name 'mortality_undiscountedcohort_size_by_age'
Should be: mortality_undiscounted = cohort_size_by_age * mortality_rate
- Bug: Formula (1-exp(-r*L))/r gives 0/0=NaN when r=0
- Fix: Check if |r| < 1e-10, return undiscounted life expectancy
- Tested: All discount rates from 0% to 100% work correctly
- Specifically verified r=0.1 (10%) works as expected
- Updated cancer_treatment_cost default across all modules
- Changed from PRIME 2017 value () to updated estimate ()
- Reflects ~39% increase to account for:
- Inflation from 2017 to 2024
- Healthcare cost increases beyond GDP deflator
- More current treatment cost estimates
- Updated in: demo.py, turkey.py, ui_app.py, test files, validation
Changes:
- Remove Turkey-specific data info banner at top
- Fix current_year to 2024 (constant, not user input)
- Fix oldest_year to 1931 (constant, removed from UI)
- Move 'Number of doses' selector to top section (with vaccine type)
- Change layout from 3 columns to 2 columns in top section
- Remove duplicate dose selector from cost section
- Cleaner, more focused UI for Turkey-specific analysis
Change filename from:
analysis_TUR_2020_20250127_003045.xlsx
To:
PRIMEpy_TUR_4vHPV_age10_20250127_003045.xlsx
More descriptive and includes key parameters (vaccine type, vaccination age).
- Add hpv_distribution parameter to DemoParams (0.0-1.0, None=auto-load)
- Auto-load from data.hpv_distribution by default based on country and vaccine
- Add UI controls: checkbox for 'Use default' + manual input
- Default values respect vaccine type: 4vHPV=77.5%, 9vHPV=92.0% for Turkey
- Add to parameters output table in Excel export
- Log which source is used (explicit vs loaded from data)
HPV distribution represents proportion of cervical cancer cases
attributable to vaccine-covered HPV types. Higher coverage = more
cases preventable by the vaccine.
Fix cost-effectiveness assessment: restore 3-tier WHO thresholds
- Add defensive checks for 'Link' and 'Source' columns
- Fallback to constructing source display from 'Year' if needed
- Prevents crash when sources_df doesn't have Link column
Revert "Simplify cost-effectiveness threshold to 1×GDP only"
- Change from 3-tier system (Very CE / CE / Not CE) to 2-tier (CE / Not CE)
- Cost-effective if ICER < 1×GDP per capita
- Not cost-effective if ICER ≥ 1×GDP per capita
- Remove 3×GDP threshold reference
- Update both UI display and Excel export logic
Remove remaining Turkish comments from cost_effectiveness.py
- Replace Turkish comments with English abbreviations
- YLL = Years of Life Lost (was: Kurtarılan yaşam yılı)
- YLD = Years Lived with Disability (was: Önlenen ölümcül olmayan DALY)
- ICER = Incremental Cost-Effectiveness Ratio (was: Kurtarılan DALY başına ICER)
- Replace data.global_data fallback logic with Turkey constant
- GDP now correctly shows ,325 in UI instead of ,000
- Matches prime repo implementation
Fix circular import: move TURKEY constants to demo.py
RATIONALE: TÜİK 2024 population data only covers 2007-2024. With current_year=2024
and next_n_years>1, future cohorts (2025+) have no population data, causing them
to be skipped. This leads to incorrect ICER calculations.
CHANGES:
- Removed 'Next N years' input widget from UI
- Fixed next_n_years to 1 throughout the application
- Updated turkey.py default from 10 to 1
- Updated info box to clarify single-cohort analysis
- Removed next_n_years from parameter summary table
RESULT: Cost-effectiveness now calculated for single 2024 cohort only, ensuring
all data is available and ICER is accurate.
- Added vaccinated_birth_cohorts list to iterate over multiple birth cohorts
- Changed from single-cohort to multi-cohort logic to match prime repository
- This enables proper next_n_years handling for multi-year vaccination programs
- Cohorts with missing population data are now skipped (cohort_size=0) instead of using placeholder
- Changed all prepare_cohort_outcomes calls to use disability_weights (dict) and disability_durations (dict)
- Added prevalence_rate parameter to all calls
- Matches the updated function signature from YLD/YLL fixes
Add Turkey 2024 TÜİK data defaults and convenience functions
- Created turkey.py module with create_turkey_params() and run_turkey_simulation() convenience functions
- Defaults to TÜİK 2024 data: population (2007-2024) and GDP per capita (15,325 USD)
- Set current_year default to 2024 as basis for all experiments
- Added info box in UI highlighting Turkey-specific TÜİK 2024 data
- Exported Turkey convenience functions in __init__.py for easy access
Usage:
from primepy import create_turkey_params
params = create_turkey_params(vaccine='9vHPV', coverage_vac=0.95)
Fix YLD and YLL calculations to match R PRIME model exactly
- Fixed YLD calculation: Implemented R's 4-phase approach (diagnosis, control, metastatic, terminal) instead of simplified formula
- Fixed YLL discounting bug: Use undiscounted mortality in YLL calculation to avoid double-discounting
- Results now match R PRIME model within 0.001% for all metrics (cases, deaths, YLL, YLD, ICER)
- Validated against R model with detailed logging showing perfect alignment
The bug was in the aggregation logic across multiple birth cohorts. When summing
outcomes across N years of vaccination, we were:
- SUMMING health outcomes (incidence, mortality, etc.) ✓
- NOT SUMMING vaccinated counts ✗
This caused ICER to scale with next_n_years because vaccination costs stayed constant
while benefits scaled by N.
Fix: Sum ALL columns during cohort aggregation, including cohort_size, vaccinated,
and immunized. Different birth cohorts have different population sizes (different
vaccination years = different cohort sizes at vaccination age). For multi-year
program CE, both costs and benefits must scale proportionally.
Also updated UI help text to clarify that next_n_years represents years of
vaccination program, not just birth cohorts.
Remove 'Use 2024 Turkish Population Data' toggle - auto-enable when available
- Removed the checkbox toggle for 2024 Turkish population data
- App now automatically uses TÜİK 2024 data when available
- Shows caption indicating which population source is being used
- Cleaner UI with fewer user decisions required
Change default coverage to 90% and add parameters+CE results to Excel export
- Updated default routine vaccination coverage from 80% to 90%
- Added 'parameters' sheet to Excel export with all input parameters
- Added 'cost_effectiveness' sheet with CE results summary
- Both sheets now appear first in the Excel workbook for easy reference
Update GDP per capita to 15325 USD from TÜİK 2024 data
