To perform a forecast on iFBAD, you first will need to create an account. This account allows you to access, download and delete saved forecasts.
After creating an account, you may perform forecasts. The forecast requires specification of several inputs including population projections, mortality rates and incidence rates. You may use the default inputs or you may provide your own inputs by uploading data files.
Our model allows for two stages of disease (stage 1 and 2). If you are interested in assessing the impact of potential interventions to delay disease onset and/or progression to stage 2 disease, then you are required to indicate the year when the intervention would be introduced to the population and how the intervention would impact disease incidence and/or the probability of transitioning to stage 2 disease.
The results of the forecast include several tables and figures that summarize the findings in addition to a text file that includes calendar year, gender and stage specific estimates of prevalence and incidence. This text file can be downloaded and used to create additional tables and figures by the user.
In the text that follows, we provide detailed descriptions of how to create an account, how to perform a forecast including the required format of data files that the user may provide, a summary of the results provided and then finally how to navigate through your user account.
Prior to performing a forecast, you must first set-up an account. Creating an account will allow you to save, access at a later time, download, and delete results from forecasts. To set-up an account follow the instructions below:
1. Select "Forecast" from the menu on the left hand panel of the website.
2. A "Login" screen will open and you should select "Need an account?"
3. A "Create Your Account" screen will open and you should provide a username, password, your full name and email address. Submit!
NOTE: Your username must be 5-12 characters. The first character must be a letter and the remaining characters can be letters, numbers or underscores.
NOTE: Your password must be 8-12 characters. The characters can be letters, numbers or underscores.
NOTE: Your email address should be entered as user@domain.
4. Your account information (excluding the password) will be displayed and then you are ready to forecast, select "Go to the inputs form".
5. At the top of the inputs page, you will see options for your account including "Upd Acct" (update your account information), "User Files" (saved results from prior forecasts), and "Logout". See Navigating your user account for a description of the options within your account.
NOTE: On subsequent sessions in iFBAD, you will simply select "Forecast" and on the "Login" screen provide your username and password to start forecasting.
To perform a forecast, you must specify a series of inputs. The required inputs are described in detail below:
Projections from 2001 to 2050 are estimated for Africa, Asia, Europe, Latin America / Caribbean, North America and Oceania (Australia, New Zealand, Melanesia, Micronesia, and Polynesia) using United Nations population projections.
1b. Forecast the burden of Alzheimer's disease for the U.S.
Projections from 2001 to 2050 are estimated for the U.S. using U.S. Census Bureau population projections.
1c. Forecast the burden of Alzheimer's disease for a specific region or country.
You may obtain projections from 2001 to 2050 for a specific region or country by providing user-defined population projections.
You are required to upload male and female population projections for the region or country of interest. The male and female population projections must be saved as a text file, either tab or space delimited, and each line of the file should end in a carriage return/line return. The files should NOT contain any row or column titles. Each file should contain 42 rows representing ages 59 to 100 and 51 columns representing the calendar years 2000 to 2050. The values in the text file should be the population size in 1,000s of people (i.e. 507 represents 507,000 persons). For an example, see the male population projections for the U.S.
The default background mortality rates are gender, age (60 to 100) and calendar year (1959 to 2090) specific modeled U.S. mortality rates. We obtained historic U.S. mortality rates from 1959 to 2002 from The Human Mortality Website. We then used regression models to predict future mortality rates based on recent historic trends; specifically, for each year of age, we fit a regression model to the U.S. mortality rates in years 1988 to 2002 to obtain the annual percent change in mortality. These estimated annual percent change in mortality were applied to the historic data to project mortality rates from 2002 to 2090.
Alternatively, the user may supply their own background mortality rates.
If you wish to upload male and female mortality rates, each file must be a text file; either tab or space delimited, and each line of the file should end in a carriage return/line return. The files should NOT contain any row or column titles. Each file should contain 41 rows representing ages 60 to 100 and 132 columns representing the calendard years 1959 to 2090. The values in the text file should be the mortality rate expressed as the percent per year (i.e. 0.02 represents 2%). For an example, see the estimated male mortality rates for the U.S.
The default incidence rates are based on a systematic review of the world's literature on the incidence of Alzheimer's disease (reference) . We assume that male and female incidence rates for Alzheimer's disease are: incidence rate (% per year) = 0.117 x exp(0.127 x (age - 60)).
Alternatively, the user may supply their own age-specific incidence rates of Alzheimer's Disease.
If you wish to upload male and female incidence rates, each file must be a text file; either tab or space delimited, and each line of the file should end in a carriage return/line return. The files should NOT contain any row or column titles. Each file should contain 41 rows representing ages 60 to 100 and 1 column representing the incidence rate. The values in the text file should be the incidence rate expressed as a percent per year (i.e. 0.012 indicates incidence is 1.2 percent per year). For an example, see the default incidence rates.
The default value is 0.167 corresponding to an approximate mean duration of 6 years in the early stage of disease (i.e. 1/6 = 0.167).
Alternatively, you may provide a value of the transition probability. This value must be a real number between 0 and 1 with at most 5 decimal places.
Note: if you want to assume a one-stage model of disease, set this annual transition probability to 0, i.e. no persons ever transition to stage 2 disease.
Select either an additive or multiplicative model for the impact of Alzheimer's disease on mortality.
In the additive model, the population mortality rates change by an additive constant, which depends on both the stage of disease and gender (i.e. if the mortality rate is 5 deaths per 100 persons among healthy persons aged 60 and the effect of stage 1 disease is to increase the death rate by 1 death per 100 persons, then the mortality rate is 6 deaths per 100 persons among persons aged 60 with stage 1 disease. The model would be specified as additive and the additive constant is 0.01).
You may also assume a multiplicative model where the population mortality rates change by a multiplicative factor, which depends on both the stage of disease and gender (i.e. if the mortality rate is 5 deaths per 100 persons among healthy persons aged 60 and the effect of stage 1 disease is to increase the death rate by 20 percent, then the mortality rate is 6 deaths per 100 persons among persons aged 60 with stage 1 disease. The model would be specified as multiplicative and the multiplicative constant is 1.2).
Our review of the published literature of survival rates with Alzheimer's disease suggests an additive model in which background mortality is not increased during stage 1 disease but increased by 11% per year once persons progress to stage 2 disease (regardless of gender) (reference).
An intervention is introduced that delays disease onset, that is, modifies the incidence rate. You must specify the relative risk of disease onset with the intervention and the calendar year when the intervention is introduced. We assume that the entire population will have access to the intervention and that the intervention is 100% effective for the entire population.
For example, suppose that in 2015 there will be an intervention that reduces the incidence of stage 1 disease by 35%. Then the relative risk of disease onset is 0.65 and the calendar year is 2015. A 35% reduction in the incidence of stage 1 disease with the introduction of the intervention translates to prolonging the onset of stage 1 disease by an average of 3 years. Under our incidence rate model (0.132 x exp(0.121 x (age - 60))), we estimated several relative risks corresponding to mean delays of onset of disease. The table below displays several options according to our incidence rate model:
| Mean delay of onset (years) | Relative risk of onset |
| 1 | 0.88 |
| 1.5 | 0.83 |
| 2 | 0.77 |
| 2.5 | 0.72 |
| 3 | 0.65 |
| 3.5 | 0.63 |
| 4 | 0.59 |
| 4.5 | 0.56 |
| 5 | 0.52 |
An intervention is introduced that delays progression of disease from stage 1 to 2, that is, modifies the annual transition probability. You must specify the mean delay (in years) of stage 1 to stage 2 disease and the calendar year during which the intervention will be introduced. We assume that the entire population will have access to the intervention and that the intervention is 100% effective for the entire population.
For example, suppose that in 2015 there will be an intervention that delays the progression of stage 1 disease by 1 year. Then the mean delay in progression to stage 2 disease is 1.00 year and the calendar year is 2015.
Disability adjusted life years (DALYs) are calculated using the methods of Murray and Lopez (reference). The DALYs are the sum of years of life lost due to premature death from disease and years living with disabilty due to disease. The years of life lost are calculated using life expectancy of Japanese women (based on 2007 life tables). The years living with disability requires a measure of disease severity or disability weight.
In our application, we define stage 1 disease as mild/moderate disease and stage 2 disease to be severe disease where persons require the equivalent of nursing home care. The default disability weights are set to 0.45 and 0.94 for stage 1 and 2 disease, respectively (reference).
The user may supply disability weights for persons living with stage 1 and stage 2 disease. These should be values between 0 and 1 and with at most 2 decimal places.
The user must provide the estimated costs for stage 1 and 2 Alzheimer's patients. Costs may be defined in many ways, for instance, treatment or total care. The cost must be an integer up to 6 digits.
In addition, the cost projections can be for the current year (2009) or in future dollars by specifying an inflation factor. For instance, the user may assume that costs will rise by roughly 3 percent per year, so the user would provide 1.03 as the inflation factor.
The default is the current calendar year (2009). The user may provide a calendar year between 2006 and 2050.
The following output is produced for each forecast:
1. A series of summary tables is created within a html file.
1a. If you perform a global forecast, the following tables are created (outputtables.htm):
1b. If you perform a US or region/country forecast, the following tables are created:
2. A html file containing the inputs that were used to perform the forecast (saveinput.htm).
3. A series of figures summarizing the results of the forecast.
3a. If you perform a global forecast, the following figures are created:
3b. If you perform a US or region/country forecast, the following figures are created:
4. A complete listing of the results is generated so that the user may perform additional calculations or generate figures.
4a. If you perform a global forecast, the following additional files are available:
4b. If you perform a US or region/country forecast, the following additional file is available:
To access your user account, go to the Forecast page and login.
You will see a banner welcoming you with the following options:
1. Upd Acct
This form may be used to change your password, name, or email address. Your current password is always required.
2. User Files
This form allows you to view saved files (Saved Files: default page) and also manage uploaded datasets that you used for your forecast (Uploaded Files).
2a. Saved Files
This page provides you will the ability to view, download or delete saved forecasts.
2b. Uploaded Files
This page provides you with a list of files that you uploaded during forecasts. These files may be viewed by selecting the filename, saved by right-clicking the filename or deleted. You will be asked to verify that you want to delete the file.
Here we describe how to obtain a forecast for a particular state within the US.
We will use the state of Maryland as an example.
1. Obtain population projections for Maryland:
The US Census Bureau provides population projections for 2004 to 2030 by gender and age (single year for 60 - 84 and combined for 85+) for each state within the US (reference, see File4 under Detailed Data).
You are required to provide population projections by gender for ages 60 to 100 (40 rows) and years 1959 to 2050 (51 columns) in 1,000s of persons. In this case, we set the population projections to zero for the calendar years and ages that were not provided within the US Census Bureau data. In addition, all values of the projections were divided by 1,000.
The male and female formatted population files used in the forecast can be viewed and downloaded here: Male Female
2. AD incidence rates:
The population projections only provide data by year of age up to 85. However the program requires age (60-100) specific incicence rates. Therefore, we need to create inputs for incidence rates.
We utilized the default incidence rate model and replaced the age 85 incidence rate with the incidence rate that corresponds to the incidence rate for the average age among persons at least 85 for each gender. We obtained the average age among persons at least 85 by using the default US population projections (calendar year 2009, 89 and 90 for males and females, respectively). We set the incidence rate to zero for all ages 86-100.
The male and female formatted incidence rate files used in the forecast can be viewed and downloaded here: Male Female
3. Costs:
We forecast the cost of formal care for persons with early and late stage AD utilizing data published by Fox et al. Formal care includes medical care and social services. We assumed that early stage of disease equates to community residents and late stage disease is similar to disease requiring institutional care. Fox et al. report that in 1998 average total costs (formal plus informal costs) for community residents was $64,741 and for institutionalized residents was $71,1149. The authors estimated that roughly 27% and 88% of the total costs were formal costs for community and institutionalized residents, respectively. Therefore, in 1998, the average formal costs were $17,480 and $62,611 for community and institutional residents, respectively. We inflated the 1998 costs to 2009 costs using the Bureau of Labor Statistics inflation calculator to obtain our estimates of $23,160 and $82,957 for early and late stage disease, respectively.
4. Impact of interventions
We ran the forecast three times. First we assumed no intervention. Then we introduced a preventative intervention in 2015 that would delay the onset of disease on average by 1 or 2 years. We set the "Relative risk of onset" to 0.88 and 0.77, corresponding to an average of 1 or 2 year delay for the onset of disease, respectively.