Prevent user group

Welcome to the webpage of the Prevent user group

Here you'll find an overview of problems encountered and possible solutions for those problems.

Please also visit the webpage on prevent designed by Jan Barendregt - who developed Prevent.

 Topics:

• Prevent user group 
• How to model web of risk factors
• How to model increases rather than decreases in risk factor prevalence?
• How to specify risk functions for continuous variables?
• What do the error messages mean?
  
• Literature on Prevent
  

* Prevent User Group

Q: We are currently looking into using the PREVENT model. It would be very useful for us to share our experience with other users. Do you know of any other groups currently using the PREVENT model and who would be best to ask about detailed questions regarding populating the database tables for the model?

• A: we keep a list of people that we know are (planning to) use Prevent. Their data can be accessed through this (password protected) link.

* Web of risk factors

Q: How do I model a web of risk factors? E.g. Physical activity influences disease directly, but also through mediating BMI, which is a risk factor as well?

• A:  There are several things you have to keep in mind. You will need prevalence data on both risk factors (in this example physical activity and BMI). In the access datafiles you will need to do the following:
  
  
• in the tab DiseasesAndRiskfactors you need to indicate that BMI is both a disease and a risk factor. You do that by checking the box Riskfactortoo (and if you have a continuous variable and/or have historical data you also check the box continuous and/or use the lookback field). For physical activity you don't need to change anything
• in the tab DiseaseRiskfactorRelation you specify the relation between physical activity and BMI (BMI being 'disease')
• in the tab ConrfRelRisk or CatrfRelRisk (depending wether you have physical activity as continuous or categorical risk factor) you specify the change in BMI upon changes in physical activity. Make sure that in the field 'Disvar' you change prevalence (there are no incidence data on BMI, only prevalence!). When you run Prevent make sure the changes in BMI are in the direction you expect! 
• As always, make sure variable names are spelled consistently throughout your file
• Now you should be able to run Prevent

* Increasing prevalence of risk factors

• A:  In the tab Conrfinterventions, you have to specify for a lognormal (or normal) distribution that the mu (mean) and/or sigma (sd) are increasing. You can do this by specifying a negative number. A number in MPar1 and FPar1 between 0 and 1 indicates a certain % shift in mu towards the optimum distribution, a negative number indicates a shift away from this distribution for the mu. Similarly, for the sigma, you can specify a negative number: sigma (the spread) then increases, and the extreme exposures become more prevalent.
  You can always check the effects of an intervention by running PREVENT with the intervention and inspecting the prevalence data outcomes.

* Regarding the ConrfRelRisk tab

How to specify the risk functions for continuous risk factors? You have to use the tab ConrfRelRisk. First, based on other sources, get information regarding the shape and strength of the risk function.

When you have a disease that occurs only in one sex, then place RR=1 for the other sex, otherwise Prevent will not run (do not use 0!).

Then, the most often used risk functions are the linera and PerunitRR options. Here a short description of these functions:

Linear: describes risk function of shape y=alpha+beta*x

• Par1: intercept (alpha) 
• Par2: slope parameter (beta)
• Par3: not in use, place 0 (zero) in cells

PerunitRR: described changes in risk per X unit(s) increase in prevalence

• Par1: RR per unit
• Par2: level for which RR=1 (attention! Par2 cannot be zero, make it a small number if needed but NOT zero!)
• Par3: number of units for which information is given (when per 1 unit, use 1; when e.g. per 5 units, write 5)
  N.B. When you have a protective factor and use a PerunitRR, then specify -1 per unit (in Par3). You cannot specify a RR<1, so you have to specify that the RR increases per unit decrease of prevalence.
  

* Error messages

Unfortunately, Prevent does not give very clear error messages. Below we list a number of error messages and possible causes of the message. This list might not be complete, if you have additions, please let us know!

• "Access violation": this means there is a problem in the specification, for example a risk function in conrfrelrisk that cannot exist (negative values in a logit function, etc)
• "Floating point division by zero": Prevent detects that somewhere, data are missing:
  - this might be a diseaseriskfactor relation
  - missing data for a certain age group (if not available, assume prevalence or make it 0)
  - if in GeneralTab you project beyond the years for which you have projected population data (or if in the interventions you project beyond those years)
• "Range check error": this means that somewhere there are differences in the specification, probably in the age groups, check that age groups are specified equally in the exposure, relative risk and intervention tabs, and that all age groups are present everywhere.

Literature on Prevent

Literature on previous versions:

• Gunning-Schepers L. The health benefits of prevention: a simulation approach. Health Policy 1989;12(1-2):1-255.
• Gunning-Schepers LJ, Barendregt JJ, Van Der Maas PJ. Population interventions reassessed. Lancet 1989;1(8636):479-81.
• Gunning-Schepers LJ, Barendregt JJ. Timeless epidemiology or history cannot be ignored. J Clin Epidemiol 1992;45(4):365-72.
• Naidoo B, Thorogood M, McPherson K, Gunning-Schepers LJ. Modelling the effects of increased physical activity on coronary heart disease in England and Wales. J Epidemiol Community Health 1997;51(2):144-50.
• Gunning-Schepers LJ. Models: instruments for evidence based policy. J Epidemiol Community Health 1999;53(5):263.
• Mooy JM, Gunning-Schepers LJ. Computer-assisted health impact assessment for intersectoral health policy. Health Policy 2001;57(3):169-77.
• Brønnum-Hansen H. How good is the Prevent model for estimating the health benefits of prevention? J Epidemiol Community Health 1999;53(5):300-5.
• Brønnum-Hansen H, Juel K. Estimating mortality due to cigarette smoking: two methods, same result. Epidemiology 2000;11(4):422-6.
• Brønnum-Hansen H. Predicting the effect of prevention of ischaemic heart disease. Scand J Public Health 2002;30(1):5-11.
• Barendregt JJ, Oortmarssen GJ, van, Hout BA, van, Bosch JM, van den, Bonneux L. Coping with multiple morbidity in a life table. Mathematical Population Studies 1998;7(1):29-49.
• Barendregt JJ, Bonneux L, van der Maas PJ. The health care costs of smoking. N Engl J Med 1997;337(15):1052-7.
• Barendregt JJ, Bonneux L, Maas PJ, van der. When does nonsmoking save health care money? The many answers to a simple question. In: Jeanrenaud C, Soguel S, editors. Valuing the cost of smoking. Assessment methods, risk perception and policy options. Boston, Dordrecht, London: Kluwer Academic Publishers; 1999. p. 75-91.
• Barendregt JJ, Oortmarssen GJ, van, Vos T, Murray CJL. A generic model for the assessment of disease epidemiology: the computational basis of DisMod II. Population Health Metrics 2003;1(1):4.
• Barendregt JJ, Ott A. Consistency of epidemiologic estimates. European Journal of Epidemiology 2005;20(10):827-832.