Term
| Simple tree models -useful for questions about routes that flows of patients may take as they move thru the system Measures outcomes -useful if outcomes occur within a narrow band at fixed time ( specific time horizon) after the intervention. If not one approach 1/is to divide into sequence of slices (recursive tree model) -but becomes unwielding v.quickly if many branches & time slices. Another approach is 2/Markov model w/c avoid this proliferation by assuming memory less property But if outcome spread out over time/ occur more than once -use Markov models |
|
Definition
| -Using tree models to address a policy issue a) sensitivity analysis to account for the uncertainty in the distribution in mean values . Monte Carlo technique -Monte Carlo method: approach to modelling that involves using values of individual attributes and for time intervals that are sampled from distributions rather than fixed. Each run if the simulation gives a different result and running the simulation many times gives distributions of results . B) Estimating health benefit to enable direct comparison with other possible health care programmes . Requires an estimation of the number of life saved -requires gross assumptions - apply discounting I.e years of live saved that arrive earlier in a programme are valued more highly than and so are discounted less -consider differences associated with personal attributes ( gender ,age ): cost-effectiveness of program dp on the local distribution of the population |
|
|
Term
Simple tree model
Limitations |
|
Definition
1-Benefits discounted
2-smoking habits subject to continuous change ( quit, relapse) & years of life saved will dp on whether would have quit in future w/o program
3-program not offered only once
Application analyse prenatal screening |
|
|
Term
| Modelling a longer period using chains of recursive trees |
|
Definition
| -Tree models r misleading; if times from intervention to a relevant outcome is 1/uncertain, 2/ some pts experience the intervention several times over or undergo several changes of state -overcome by 1/dividing period during w/c relevant events might occur into a series of shorter periods/cycles 2/modelling distribution of outcome events by stringing together a series of trees, each tree showing the number of events during a defined period 3/ each cycle should be short enough to allow the assumption that each individual will experience at most one event per cycle Note: different sets of probabilities and mortalities can be used in each period Application : evaluating a cyclical prevention programme |
|
|
Term
| Markov property Memoryless property means that the probability that an object moves from one state to another in a given time period depends ONLY on what the initial & final states are |
|
Definition
| -Based on assumption that the three groups of people still smokers ( & ex smokers) at end of 1st period are interchangeable -need only one smoker tree in subsequent cycle instead of three -thus all you need from each in each period is the total number of outcomes of each type -implications of this assumption; 1/ when u merge all the streams at the end of each cycle you lose information or forget -about route taken they the series of cycle trees up till that point ( memmory less model ) 2/ probability of moving from state C to Y in a given cycle can be treated the same for everyone who was in state X at beg if the cycle 3/and independent of what happened in the past I.e -number of people who started in the cycle in state X - the states that they gave been thru in previous cycles -Markov chain; a) is when the number of states are finite and transition probabilities do not change over time b) can be solved by matrix algebra to give expected survival times in each state under different Cdx |
|
|
Term
| Pros & Cons of Markov models |
|
Definition
Advantages:
1-the effects of repeated interventions and repeated changes of state can be modelled
2-interventions ( implying costs ) and changes of state ( implying benefits ) take place within given time periods so that costs and benefits can be discounted
3-risks & probabilities can change with he passage if time or with the ageing of cohorts , imp in context if health care
-they can be built using standard spreadsheet software & add ins are available (@RISK), w/c allow parameter values to be sampled from distributions rather than remaining fixed
Disadvantages;
-Model may need many time periods if the events involved are not rare
-memory less property ; overcome by disaggregation
-models flow of groups and as flows rather than the sum of experiences of individuals |
|
|
Term
| Feedback loop; -Causal loops formed when flow rates out of a process or state influence flow rates into it -which determine whether the response to disturbance is stable or unstable behaviours.whereas in system dynamics modelling the aim is to understand this behaviour and take account of it in graphic version |
|
Definition
| -systems where causal links work both ways -elective surgery example; although the level of the tank is affected by the inflow , the inflow is also affected by the level of the tank ( feedback) -here negative feedback ( level down leads to inflow up) results in a stable system, whereas positive feedback destabilises the system -system dynamics :key feature of such models is the recognition that the behaviour of some parts of the system are affected by the state of other parts . Understanding feedback is imp as it determines whether the response of the system as a whole to change will be stable or unstable |
|
|
Term
| Building a system dynamic model |
|
Definition
1st step; draw up a graphic version that shows how the behaviour of some parts of the system are affected by other parts
-the system boundary : separates what is considered to be inside the model ( endogenous variables ) from what is outside |
|
|
