Shared Flashcard Set


Building a Decision support Model
developing a graphic Model of the causal link that allow the programme to achieve its objectives

Additional Management Flashcards





learning objectives

  •  formulate & build a qualitative model to address a decision making problem in managing community HCP
  • simplify and refine the model structure
  • id and gather relevant data
  • build a spreadsheet model
  • use the model to predict the impact of different options in terms of the number of lives saved
  • carry out a sensitivity analyses
  • evaluate the model
Step1:Draw up an initial graphic model

Think about what managers are trying to achieve &what the model needs to predict

1. drawing an influence diagram showing relevant factors for one disease.

2.start with your outcome variable pn the right of sheet of paper and work upstream to the left.

3.identify direct or "proximal influences (PIs)" then more distal factors that affect PIs.

4. id control variables

5. define treatment impact factor

Step3: Refine the graphic Model

Refine model in light of what has been learned about the system and any concerns raised so far;

  • leave factors out that are expected to be unchanged over model's time scale
  • Add more detail: variable costs (incr trx volumes lowers cost/pt) & fixed costs.
  • forecast how exogenous variables will vary
  • include how the balance of numbers of lives saves with different numbers treated
  • derive a formulae of how outcome variable calculated from control/exogenous variable
Step3: Identify and gather relevant data
  • data on treated and untreated mortality (studies in literature)
  • data on variable and fixed costs (local information)
  • think about the data's reliability and whether a sensitivity analysis is needed
  • improve model by repeating model dvt cycle;i.e. re-examining your theory and cillecting new data
Step 5: use model yo decide on the strategic balance
Step 6: carry out a sensitivity analyses (SA)
  • trying different values of the parameters to se how much they alter yiur conclusions
  • investigates the effects of error in the data and changes in the underlying assumptions on the results.
  • multiple criteria and judgement; 
  1. one objective thus no judgement involved
  2. In reality more objectives (e.g. improve sx) and equity issues
  3. thus, best option will be a matter of opinion thru "consensus" e.g weight-avoiding death vs any other benefits from trx.
  4. when >1 criterion no objectively best option unless; one option is equal to or better than ALL the others on ALL criteria
Step 7: evaluate the model
  • transparency: simple spreadsheet allow managers to examine formulae and test model-i.e. assess face validity
  • predictive validation: to a degree as results can be tested for plausibilirt (whether they are in the right kind of range) but not for correctness (i.e. case fatality rates)
  • accreditation:  models published in scientific literature should have been subject to peer review . Publication: corroboration btw differ models =incr confidence

Cl: 1st series of cycles of DM; unable to use model validated against empirical data in local context. Thus rely on 1/consensus, 2/internal logic/3 data from elsewhere.

Supporting users have an ad free experience!