chjackson/healthchecks

scenarios

Opened this issue · 0 comments

JDWoodcock commented

implement the scenarios

HC model – scenarios
31 May 2016

The following is a list of scenarios that we could simulate with the current HC model, plus some information on what is needed to run the scenarios. For 2-7, one idea to simulate scenarios would be to do a 'parameter sweep', ie. Changing the parameter in question in increments from a low to a high level within a loop and recording the relevant Health Outcome, displaying results as a dose-response curve.

For 1, the scenario is different as parameter is binary – I would suggest running combinations of settings for CVD and diabetes registers, such as () CVD==1 and diabetes==1, () CVD==1 and diabetes==0, etc, until (*) CVD==0 and diabetes==0, which should be the status quo to compare with.

Note: Parameters are currently changed by invoking the self.SetUncertainParameter() function.

  1. Including people with CVD and on diabetes registers

Introducing new variable array self.include_registers to explicitly define whether HC-eligible individuals can be on registers/have CVD or not
[already implemented]
2. Change age thresholds for eligibility

currently done by changing elements in parameter array up or down
[starting at age 50]

  1. Overall uptake rate increased to 'best practice'

currently done by changing parameter self.up_HC_takeup from currently 0.488 to 0.85, as per Anna's recently sent figures, drawing on Robson 2015.
[ready to implement]

  1. Increased uptake of people in deprived areas

currently done by changing elements parameter self.up_SES_vec up or down – last element in this vector is RR for uptake by most deprived people
20% increase among those in most deprived quintile, 10% increase those 2nd most deprived quintile

  1. Increased uptake of smokers

currently done by changing second element in parameter vector up_smoker_vec up or down (RR of smokers to attend)
[same chance of attending for smokers as non smokers]

  1. Increased uptake of high risk individuals

currently done by changing elements in parameter vector up_Qrisk_vec up or down
20% relative among high risk individuals (qrisk 15+)

  1. Increased treatment and compliance rates to 'best practice'

currently done by adjusting up_Statins_comp parameter (compliance for Statins) and 'up_HC_statins_presc_Q20minus'/'up_HC_statins_presc_Q20plus' parameters to according levels
Anna to recirculate doc from May from Robson from TH. Just uptake no change in compliance for now.

Extra suggestion from Nick

  • Scenarios; to include a specific scenario around targeting those who do not attend for HC (I think we suggested we might have three scenarios:

  1. the probability of attending at subsequent health checks depends on past attendance;
    70% for those who attend & 30% for those who didnt.

  2. stochastic i.e. probability of attending a health check is x% and is fixed and independent of previous attendance, or risk factors – this is currently how the model runs;

  3. those who did not previously attend are specifically targeted… to a certain extent (2) achieves this, but we may want to go further) (Oli to look at data on existing screening programmes re: probability of re-presenting for screening)
    30% for those who attend & 70% for those who didn't.

Result reporting:

  1. Results per million HC
  2. total gains