Population Health Impact of Disease in Canada (PHI)

Workbooks and microsimulation

• Workbooks
• Microsimulation tool
• FAQs

Two electronic tools will help users understand and manipulate results of the PHI. Excel workbooks will provide estimates of the health-adjusted life years (HALYs) lost to about 200 diseases and injuries. A microsimulation tool will provide more sophisticated estimates and intervention scenarios. The following sections describe these tools and discuss their potential uses, relative strengths and limitations.

Workbooks

The workbooks provide comparable information about the relative importance of about 200 major diseases in Canada in terms of years of life lost to premature mortality and to reduced functioning. Input data from multiple sources is assembled to generate summary measures of health: years of life lost to premature mortality (YLL), year-equivalents lost to reduced functioning (YERFs) and the sum of these, health-adjusted life years (HALYs) lost to the disease.

There is one workbook per disease. Each workbook contains several worksheets that calculate and document YLLs, YERFs, and HALYs lost to the disease for the reference year 2001.

The workbooks combine Canadian data on incidence, remission, duration, and fatality associated with the different stages of the disease; when Canadian data are not available, the best alternative sources are researched and used. Whenever possible, estimates use 2001 data. Most of this information is stratified by age group and sex. Mortality data use 1999 mortality rates applied to the 2001 population distribution. The workbooks also incorporate health state preference scores measured in the Canadian population.

Users can modify the workbooks to suit their own requirements. They can apply discount rates, choose a different life table (Canadian period life table, Canadian cohort life table or the standard life table from the World Health Organization), or use a different population such as a Canadian province. Users can modify most parameters, for instance, incidence or remission rates, to estimate the potential effect of such changes on HALY estimates.

Workbooks are also provided to calculate the contribution of selected risk factors to some of the diseases. These workbooks combine data on population prevalence of risk factors from Canadian health surveys, relative risks from the literature, and HALYs for the disease of interest. Population attributable fractions, YERFs, YLLs and HALYs are provided for several risk factors.

[return to top]

Microsimulation tool

The microsimulation simulates one individual life at a time, from their age at initialization to death; the process is repeated until all simulated lives in the run have been completed. A typical run would be for one million individuals; larger runs are used to model rarer events or to provide more precise estimates.

The microsimulation is initialized from a cross-sectional representation of the Canadian population: the Canadian Community Health Survey 2000/01. At initialization, individuals are assigned demographic characteristics, such as age, sex and province of residence; chronic conditions or diseases; and a risk factor profile that includes tobacco use, body mass index, alcohol consumption, physical activity, and fruit and vegetable consumption, amongst others. The microsimulation also attaches a utility that describes the individual's functional health: either 1 for full health or less than 1 depending on what diseases and health conditions they have. This score will change when the individual moves to a new health state.

The microsimulation events start January 1, 2001. As the individual ages, their risk profile changes based on trends observed in the longitudinal National Population Health Survey (1996-2002) and the Canadian Heart Health Surveys (1986-1992). Exposure to risk factors is combined with age-specific incidence rates to place individuals at risk of developing diseases. When a disease is assigned, information on the treatment, duration and case-fatality are used to model the disease progression.

The way in which decisions are made in the simulation is often referred to as a stochastic or Monte Carlo process. In effect, the microsimulation generates random numbers that are compared with probabilities of occurrence to decide, for instance, if the individual will develop a disease in their current year of life. Random numbers are also used in the calculation of survival times from diagnosis of a disease to subsequent disease events. Essentially, the random number is used to determine a point on the survival curve of the subsequent event, which corresponds to a specific time to wait before the event will occur.

The microsimulation has several advantages over the workbooks that may lead to more realistic estimates of summary measures of health. First, it can model two or more diseases that occur together and allocate portions of the health impact to each. Second, it can take into account the various health states that individuals experience just prior to diagnosis, rather than assuming that all individuals of a given age experience the same health state. It is important to consider this heterogeneity of health states because individuals with certain risk profiles will be at higher risk of being diagnosed with more than one disease, and thus their functional health score will be lower than that of the general population for that age group. For example, smokers are at higher risk for respiratory disease and lung cancer. The microsimulation takes into account that, among those diagnosed with lung cancer, a certain proportion also have respiratory disease and thus a preference score lower than the average score for their age group. Third, the microsimulation is a true incidence approach: the calculation of HALY is based on incidence cases in 2001 and the deaths arising from those cases in future years; by contrast, the workbooks use the incidence and deaths in 2001.

The microsimulation is designed to easily implement "what-if" scenarios. A simple change in input parameters provides a new estimate of the summary measures of health. For instance, we could estimate the population attributable fraction of heart disease arising from obesity, by eliminating obesity in a "what-if" scenario. More realistic policy relevant scenarios can also be done to evaluate which interventions would lead to the greatest impact on health outcomes.

Finally, the microsimulation is suited to making projections in future years: it monitors changes to the population continuously through time, starting in 2001 until all the end of the simulated life. Thus the simulation can report health outcomes (number of new cases, deaths, HALYs, YLLs and YERFs) in any future year and at any level of detail (by age group, sex, disease, high risk groups). Coupled with "what-if" scenarios, it provides a powerful tool to evaluate the impact of various intervention strategies.

Results from the microsimulation can be exported to Excel.

This microsimulation is an offspring of the Population Health Model (POHEM). More information about microsimulation is available on the Statistics Canada website .

[return to top]

FAQs

What are the advantages of presenting the data in workbooks?
The workbooks provide documentation of data sources and calculations so that users can see exactly where the estimates come from. In addition, they provide flexibility for adapting them to another population.

Users will be able to customize the workbooks by choosing the underlying life table, reference population and discount rates. The incidence and remission rates, durations, health state preference scores, proportion of cases that receive treatment, case-fatality, and mortality rates can also be modified to estimate the potential effects of these changes.

Similarly, the workbooks that calculate population attributable fractions of risk factors are easy to manipulate so users can perform their own basic “what-if” scenarios by changing the prevalence of risk factors and their associated relative risks.

How are the results of the microsimulation models different from the workbooks?
A 2001 incidence run will answer the same questions as the workbooks but the results will not be identical because of the following value added from the microsimulation. First, microsimulation can model two or more diseases that occur together. Second, microsimulation allows for more complex pathways of causality: a risk factor can lead to more than one disease, and sometimes one disease is a risk factor for another. Microsimulation helps sort out these overlapping pathways of risk, providing a more "realistic" estimate of HALYs lost to the disease, injury or risk factor. Third, microsimulation is an incidence-based approach, whereas the workbooks combine incidence and prevalence approaches. This changes the estimate of the mortality component of the HALY, and will be more pronounced for diseases such as breast cancer that have longer survival and changing incidence and survival patterns over time.

How does the microsimulation handle individuals diagnosed with more than one disease in one year?
The microsimulation keeps track of changes in functional health as measured by preference scores and continues to account for further reductions (or increases) throughout the life of the individual. When two diseases occur together, the preference score for having both is estimated by

u _1,2 = u ₁ * u ₂

where u ₁ and u ₂ are the mean preference scores for each of the two diseases respectively and

u _1,2 is the mean preference score associated with having both diseases. The cumulative impact was empirically confirmed using the Canadian Community Health Survey 2000-01 and the Health Utilities Index Mark 3 (HUI3).

How does microsimulation account for complex causality of diseases?
Complex causality of diseases refers to risk factors that lead to more than one disease or diseases that are risk factors for other diseases. The microsimulation models changes in risk factors and disease incidence in continuous time so that these pathways develop causally over a simulated lifetime. The following examples illustrate these complex causal pathways.

Example 1: Smoking is a risk factor for both lung cancer and heart disease (Figure 1). We could calculate how much lung cancer is due to smoking and how much heart disease is due to smoking; or conversely, how much lung cancer or heart disease could be avoided if no one smoked. The total impact of removing smoking from the population would not be the total of those developing lung cancer and heart disease since the two diseases may occur in the same individuals.

Figure 1
Modeling risk factors that lead to multiple diseases

Example 2: Diabetes is a risk factor for heart disease and obesity is a risk factor for both (Figure 2). With workbook calculations it would be difficult to sort out the contribution of obesity to heart disease via diabetes; in the microsimulation we can easily report on these individuals with obesity who will develop diabetes.

Figure 2
Modeling disease as a risk factor

How is the impact of risk factors calculated?

In addition to measuring the population health impact of disease in terms of YLLs, YERFs, and HALYs, the impact of the major risk factors for each disease is calculated using standard formulae for population attributable fraction (PAF). ¹ The PAF represents the proportion of YLLs, YERFs, or HALYs attributable to a given risk factor for a disease. It is calculated using the population prevalence of the risk factors and relative risks for the disease.

For a given risk factor, the population attributable fraction is estimated by age group (a), sex (s) and cancer site (c):

PAF_a,s,c = Σ i [ Pe_a,s,i * (RR _a,s,i,c -1) / ( 1 + Pe_a,s,i * (RR _a,s,i,c -1) ) ]

where Pe is the proportion of the population exposed to the risk factor, RR is the relative risk of developing or dying of the disease due to the exposure, and index i represents the risk category.

Data and data sources are provided in electronic workbooks.²

What are the limitations of the microsimulation?
Although microsimulation tools can be readily adapted for users (for instance, using provincial data), they are not readily accessible to all users. Creating meaningful intervention scenarios requires an understanding of the underlying data, assumptions and methodology of the model. Thus the microsimulation is sometimes seen as a black box, unlike an Excel workbook where all the calculations can be easily checked or manipulated by the user.

These microsimulation models are based on chronic disease modeling, and thus have major shortcomings for infectious diseases. For example, they do not model transmission rates between individuals. Infectious models are currently being developed elsewhere at Statistics Canada and the Public Health Agency of Canada.

1. Jekel JF, Katz DL, Elmore JG. Epidemiology, Biostatistics, and Preventive Medicine, second edition. Philadelphia (PA): WB Saunders Company; 2001.
2. Flanagan W, Boswell-Purdy J, Le Petit C, Berthelot J-M. Estimating summary measures of health: a structured workbook approach. Population Health Metrics 2005; 3(1):5.

[return to top]