Skip to main content
HEOR Africa

Approach and quality

Evidence you can trace, check and reproduce

Speed only matters if the evidence holds up. Our methods are designed so that a technical reviewer can follow every number back to its source.

01

A reproducible, version-controlled pipeline

Models and analyses are written in R and kept under version control. Each result is generated by code from a recorded set of inputs, so any figure in a report can be regenerated, and every change between versions is logged.

  • Analysis code, parameter tables and outputs held in one repository per engagement
  • Tagged versions for every report that leaves the firm
02

Every parameter traceable to a source

Each model input sits in a parameter table with its value, distribution, source, and the reason it was chosen. Where local data are not available, the proxy used is stated and tested in sensitivity analysis.

  • Full citation or data provenance for every input
  • Explicit notes on adjustments such as inflation and currency conversion
  • Uncertainty ranges and distributions documented alongside point values
03

Independent quality control

Before results are released, a second analyst who did not build the model checks it against a written checklist covering inputs, logic, code and outputs. Issues are logged and resolved, and the QC log is available on request.

  • Input checks against the parameter table and sources
  • Logic and extreme-value tests of the model
  • Verification that report tables match model outputs
04

CHEERS 2022-aligned reporting

Reports follow the Consolidated Health Economic Evaluation Reporting Standards (CHEERS 2022), so that reviewers find the information they expect in a familiar structure. A completed CHEERS 2022 checklist is supplied with every report.

05

Expert sign-off at three gates

Work cannot move to the next stage until a senior expert signs off the scope and model structure, the inputs and sources, and the results and their interpretation. No deliverable leaves the firm without expert sign-off.

06

Confidentiality and data protection

Client and project data are handled under written confidentiality terms and in line with Rwanda's Law No. 058/2021 relating to the protection of personal data and privacy. We collect the minimum data needed, prefer aggregate or de-identified data, and restrict access to named team members.

  • Non-disclosure agreements as standard, signed before scoping
  • Encrypted storage and access limited to the engagement team
  • Agreed retention and deletion at the end of each engagement

Method

How we work

Every engagement runs through the same reproducible pipeline, with three points where a senior expert must sign off before work moves on.

  1. 1

    Scoping

    Agree the decision problem, population, comparators, perspective and outcome measure.

    Sign-off gate 1: Scope and model structure

    A senior health economist signs off the scope and model structure before evidence work begins.

  2. 2

    Evidence

    Identify and appraise clinical, epidemiological and economic evidence in a documented search.

  3. 3

    Local inputs

    Source local costs, resource use and epidemiology, each traced to a cited source.

    Sign-off gate 2: Inputs and sources

    A clinical or subject-matter expert signs off the parameter table and its sources.

  4. 4

    Modelling

    Build or adapt the model in R, with version-controlled, reviewable code.

  5. 5

    Independent QC

    A second analyst checks code, inputs and outputs against a written checklist.

    Sign-off gate 3: Results and interpretation

    Results and their interpretation are signed off before anything is released.

  6. 6

    Reporting

    A CHEERS 2022-aligned report, parameter table and briefing for decision-makers.

Validation

How we validate models

Before any result is used, the model is validated in four ways. Where a published source model exists, we rebuild it and confirm that our code reproduces its reported results before adapting it. We test face validity with a clinical or subject-matter expert, who reviews the structure and inputs against local practice. We run technical verification: extreme-value, logic and consistency tests, and independent recalculation of key outputs. Finally, we compare results with other published evidence and explain any differences. Each step, and any discrepancy found, is documented in the technical report.

  • Replication of published source models before adaptation
  • Face validity review by a clinical or subject-matter expert
  • Technical verification: extreme-value, logic and consistency tests
  • Comparison with other published evidence, with differences explained

Want to see our QC checklist?

We are happy to walk through our methods, checklists and a sample parameter table on a short call.