Main Principles
===============

Input Data
------------

The epidemiology module is based on three main data sources:

  * The **IHME** database published on the web tool **Epi Visualization - IHME** :cite:p:`InstituteforHealthMetricsandEvaluationIHME`. This tool provides estimates for incidence, remission, fatality and prevalence rate for over 300 conditions in 195 countries and territories from 1990 to 2016.
  * The **GBD Results Tool** :cite:p:`InstituteforHealthMetricsandEvaluationIHMEa`. Among various data, this tool provides estimates for incidence and mortality rates for cancers.
  * The **IARC** database :cite:p:`IARC2018`. Provides estimates for incidence, mortality, prevalence at 1, 3 and 5 years for cancers.


Modelling framework
--------------------

The model simulates disease pathway (incidence, fatality and remission) through events at the individual level. The epidemiology module is based on two main models: 
one for cancers (:numref:`modelling_cancers`) and one for all the other chronic diseases (**generic disease model** see :numref:`disease_pathway`). 
The latter is largely derived from the framework of DisMod-MR 2.0 [#f1]_ , a tool developed for GBD analyses (:cite:p:`Flaxman`) (:numref:`disease_pathway`). Other models are variations from those two models. See :ref:`moddeling_specific_diseases` for 
details.

.. _residual_mortality:

Fatality and residual mortality
``````````````````````````````````

The cause-specific mortality is removed from total mortality to compute residual mortality
using the following formula:

.. math::

   \begin{eqnarray}
        \lambda_r = \lambda - \sum_{i \in disease}P_i \cdot f_i - \sum_{j \in cancers}m_j
   \end{eqnarray}

Where :math:`\lambda` (resp. :math:`\lambda_r`) is the hazard ratio for the total (resp. residual)
mortality, :math:`f_i` and :math:`P_i`, the fatality and the prevalence of disease i (except cancers) and :math:`m_j` the mortality of cancer j.


Disability weights
``````````````````````````````````

Disability weights are used to measure the overall disease burden. Every disease is attached to a single disability weight shown below in :numref:`dw_table`. Methodology is described below.

.. _dw_table:
.. csv-table:: Disability weights for diseases in the OECD microsimulation model
   :file: tables/dw_table.csv
   :widths: 60, 25
   :header-rows: 1
   :stub-columns: 0
   :align: center


Disease specific disability weights
************************************

The OECD-SPHeP model uses the methodology described in the technical appendix of :cite:p:`GBD2016DiseaseandInjuryIncidenceandPrevalenceCollaborators2017`. Every disease coincides with one or 
multiple health states which are associated with a disability weight (DW). These were taken from WHO :cite:p:`Mathers2017a`. 


Where multiple health states were associated with a single disease, a severity split was used to weight the different DWs into a single measure. 
The summary disability weight table is shown in :numref:`dw_table` and :numref:`dw_section` provides the calculation of DWs for specific diseases. 
Where DWs with and without treatment were provided, “with treatment” was used. 

Long-term disability weight
******************************

For acute diseases (e.g. stroke, injuries), another DW was used for the longer, post-recovery phase, reflecting the long-term disability associated with the 
disease. It applies until the death of the individual. These weights are described as ‘long-term’ DWs and shown in :numref:`sequela_dw_table`. (Conversely, classic disability weights are described as ‘short-term’). 

For chronic conditions, there is no distinction between short and long term DWs (e.g. osteoarthritis). 
If the individual is afflicted by the disease, the DW will be applied to their health status. 
If the individual recovers, the DW no longer applies. 

.. _sequela_dw_table:
.. csv-table:: Long term disability weights associated with diseases sequelae in the OECD microsimulation model
   :file: tables/dw_sequela.csv
   :header-rows: 1
   :stub-columns: 0
   :align: center



Individual disability weight
********************************

The :math:`DL` is a yearly measure which represents the disability-adjusted quality of life of an individual. 
It is computed along the life of an individual and takes into account changes in health status 
(incidence, recovery, etc...). In case of multi-comorbidities, it is assumed that DWs interact in a 
multiplicative fashion. More formally, the following formula is used:

.. math::
   :label: eq_residual_mortality
   
   DL = 1 - \prod_d (1 - DW_d)

Where :math:`DL` is the disability-adjusted quality of life of a person with a given set of comorbidities,
*d* is one of the comorbidities affecting that person
and :math:`DW_d` is the disability weight for a comorbidity d.







.. [#f1] DisMod-MR 2.0 is a Bayesian mixed-effects meta-regression modelling tool developed for GBD analyses. It provides estimates for incidence, remission, fatality and prevalence rate for over 300 conditions in 195 countries and territories from 1990 to 2016. This tool is used to generate the IHME database (Epi Visualization | Viz Hub) available at https://vizhub.healthdata.org/epi/