Investigation of National and Regional Mortality and Morbidity Risks for the Air Health Indicator (AHI).

NOTICE OF PROPOSED PROCUREMENT (NPP)


Solicitation #:  1000173235         Closing Date:  July 10, 2015         Time:  2 p.m. EDT

This requirement is for the department of Health Canada.

Title: 
Investigation of National and Regional Mortality and Morbidity Risks for the Air Health Indicator 

Scope:
Health Canada has recently developed a new methodology, the Air Health Indicator (AHI), for assessing the effects on daily mortality of short-term exposure air pollution for warm season from April to September, as they may vary dynamically over space and time in response to changes in air quality, and has been assessing its performance in simulation studies and applying it to 26 years (1984-2009) of data from 24 major Canadian Census Divisions (CDs). Hundreds of time-series studies of daily mortality have now been published worldwide, and are critical components of the scientific evidence supporting a causal relationship between air pollution and public health. The AHI provides time trends in annual risks at city-specific, regional and national levels such as an increasing, decreasing or constant trend over a time period of interest. The AHI can be used in policy analysis, with potentially important applications to the assessment of the public health impacts of air quality regulation. 

For the AHI, health outcome (daily mortality) data from 24 Canadian cities was used and estimation of between city heterogeneity has been problematic due to the small number of cities involved. Applying least squares estimation (LSE), negative estimates that are meaningless have sometimes been obtained. To improve the estimation, a Bayesian MCMC approach has been employed to estimate the heterogeneity as well as national risks. In previous contracts we investigated the effects of prior distributions on parameters on the risk estimation and have found the use of non-informative prior distributions to be desirable. We have investigated region specific risk and variability for several different region definitions, which needs further exploration. 

Estimation of short-term (or acute) risk due to air pollution is the main topic of the AHI and has been subject to much scrutiny in the recent contract. We have demonstrated a source of bias in current estimates of population health risk due to air pollution based on natural spline smoother of time and proposed a new smoother (Slepian smooth functions ) to reduce the compensates of this bias, resulting in estimates with more accurate interpretation. 

In previous contracts we have developed a model for aggregating multiple lags of a pollutant together so as to provide a single estimate of risk. This estimate can be thought of philosophically as similar to a distributed lag effect, representing the acute risk due to the air pollutant across a number of lags. This approach called synthetic lag model offers a satisfactory and stable way of synthetically lagging pollutants to give estimates of risk which encapsulate multiple lags in one series. 

So far the AHI has developed for mortality only on a single pollutant. It is desirable to consider multiple pollutants at a time as the population are always exposed to multiple pollutants. The main pollutants for the AHI are ozone and fine particulate matters <2.5 µg/m3 (PM2.5) but their availability is quite different. Ozone is available for 1980-2009, whereas PM2.5 for 2000-2009. For trend detection purpose, investigation over a long time period is more benefit to model development, and thus a 2-pollutant model for ozone and nitrogen dioxide (NO2) seems a good start. The next model would be for another 2-pollutant model for ozone and PM2.5 and possibly for 3-pollutant model for ozone, NO2 and PM2.5.

In addition to expanding from single to multiple pollutants, another expansion on response variable is desirable to include morbidity. Air pollution could make people sick, which leads to hospitalization, and thus mortality and morbidity attributable to air pollutants can be investigated together as bivariate responses. 

Further work needs to be done to incorporate a new smoother (Slepian smooth functions), synthetic lags and multiple pollutants into a bivariate response model.  

Objectives of the Requirement
This work is supposed to provide information on the following questions: 

(1)	How to best model single, synthetically lagged pollutant risk estimates at regional and national levels?
(2)	How to best model warm season model minimizing bias in risk estimates from (1)?
(3)	How to best model two pollutant risk estimates based on (1) and (2) above? In particular, for two-pollutant models for ozone & NO2.
a.	What is the 1st pollutant specific risk?
b.	What is the 2nd pollutant specific risk?
c.	What is the risk in common between the two pollutants, which cannot separate one pollutant from the other? 
(4)	How to best model multiple pollutant models and short temporal (single- or several-year) blocks for bivariate responses? 
(5)	Can the method of (3) be extended to include PM data?
(6)	How best to deal with pre-2000 PM data
(7)	How best to model the interaction between temperature and pollutants?


Estimated Value: 
The total amount of funds available for this contract is $25,000, plus applicable taxes.  A one-year contract will be awarded for $25,000, plus applicable taxes, with two option years, each worth $25,000 plus applicable taxes, for a total potential contract value of $75,000, plus applicable taxes.  The option years will only be picked up if funding becomes available and the contractor’s work is deemed satisfactory.  The winning bidder will be selected based on highest technical score within this budget.


Ownership of Intellectual Property:
The Contractor will own the Intellectual Property, while the Crown will have an irrevocable, royalty-free licence to use the IP.


Security Requirement: 
There is no security requirement.


Mandatory Requirements: 

M1.  The bidder’s project leader must have a PhD from a recognized university with specialization in Statistics with experiences in Bayesian hierarchical models, Spectral analysis, and smoother functions.  

M2.  The bidder’s project leader must demonstrate that within the last 5 years, they have undertaken at least one project on Canadian population health for environmental issues. 

M3.   The bidder’s project leader must have at least one peer-reviewed publication on Canadian population health in a scientific journal listed in the Science Citation Index Expanded (within the last 5 years). 

M4.  The bidder’s project leader must show that they have experience working with Canadian environmental databases on air pollution and climate.

M5.  The bidder’s project leader must have experience working with generalized additive model and/or generalized Poisson model. 

M6.  The bidder’s project leader must show that they have experience working with statistical software such as R. 

		
Selection Methodology
The contract will be awarded to the bidder with the  highest technical score within the budget of $25,000 for each fiscal year (one year contract with two option years).


Enquiries regarding this Request for Proposals are to be submitted in writing to: 

Robert Merrick
Contracting Authority
E-mail:   Robert.Merrick@hc-sc.gc.ca