A2:Assessing and Interpreting Results from Health State Models

Weighting: 30%
Due? Date:?Monday 15 July 2024 4pm AEST
Group work:No
Type: Technical Report
Length:? up to 1500 words + excel sheet

Assessment Description

This task will focus on applying an existing model implemented in excel to conduct projections and scenario analysis relating to the material discussed in Modules 2 and 3 (weeks 3 to 6 of the course).This project is designed to assess students technical understanding, critical thinking and attention to detail.

Specificallyin this assessment, students will need to complete three tasks relating to a simplified health-state model of smoking, capturing survival of non-smokers, current smokers and former smokers.

Background

You have been provided with an excel sheet that implements the model described briefly here. This is a health-state model stratified by single year of age and by sex, with only age groups 15 to 100+ included. The model considers smoking uptake as an age-dependent risk????(a), which transfers people from the Never Smoker groupN(a) to the Current Smoker group C(a). We also allow people in the Current Smoker group to quit at an age-dependent rate????(a). In this simplified model, quitting is considered to permanently transfer people from the Current Smoker group to the Former Smoker groupF(a)(i.e. Former Smokers do not re-start smoking). In each state, there is also an age and sex-specific mortality rate that further varies by smoking status but does not take account of smoking duration. See Figure 1 below for the model schematic:

Even though the model is relatively simple, it still needs to be supported by >10 parameters which are roughly calibrated to Australian smoking prevalence data. These parameters are shown below in Table 1. If we start at the top, you will see that we include different relative mortality risks (described as death ratios in the tutorials) for the Current Smokers and Former Smokers in comparison to Never Smokers. Youll also see that we assume standard 100,000 initial cohort populations and in addition that there are no current or former smokers entering the cohorts at age 15.

The other parameters, however, are sex and age-specific so that we can roughly capture the rapid increase in current smoking prevalence in younger people and rise in former smokers with age. The age-specific risks are applied here in a dichotomous way for the age groups 15-21 and 22+, with these also differing by sex to reflect the differing prevalence in men and women. Quit rates are applied in a similar way with different rates in age groups 15-21 and 22+ and also by sex.

Note, that the model described has been implemented in the excel sheet using the parameters described in Table 1.

Task 1: 3-state Markov model of smoking prevalence (5 marks)

Background

For this task, please refer to the tab 3-state model in the excel worksheet.

Assessment items - Firstly, graph age-specific mortality risks for the three cohorts by sex and then similarly graph the cohort sizes by age and sex for each of the Never, Former and Current smoker groups. Briefly describe your results.

Task 2: Comparisons of model interventions (15 marks)

Background

For this task, you will need to modify the model presented on the tab "3-state model". It may be most convenient for you to create one or more new tabs with copies of the model where the interventions are applied. If you do choose to do this, take care to ensure that you have copied the model correctly, so that cell-referencing remains correct on your new tabs.

Consider that you have choice of implementing one of the simplified interventions in Table 2 below. Ive focused here on interventions that can be introduced just by changing a single parameter at a time. These values all represent an approximate halving or a doubling of the relevant rate or ratio (for mortality, 1.9 is halfway between 1 and 2.8 and 1.2 is halfway between 1 and 1.4).

Assessment items

(a) The first part of this task is to sequentially (i.e. one at a time!) apply each of the 6 interventions listed in Table 2. This means changing the relevant parameters, making a copy of the relevant results and crucially changing parameters back to base-case before applying the next intervention.

As part of applying the interventions correctly, remember that the interventions either change the proportions of people in each smoking state or decrease mortality rates in the Current or Former smoker groups. This situation is not reflected in the mortality risks provided for the overall cohort. Therefore, you will need to recalculate the overall cohort survival as part of evaluating each intervention. Important here to recognise that none of the interventions alter mortality risks for the Non-Smoking group (i.e. these should be kept fixed in all scenarios).

In terms of outcomes to compare the performance of interventions on, my suggestion would be to use measures based around the total cohort life-years to live at age 0 (the T0 statistic we used in tutorial 1). You could just calculate this for the overall cohort, but if you want to also give some consideration to the size of current and/or former smoking cohorts in your comparison, then you should also calculate the same statistic for each of these columns (i.e. T0F and T0C). To remind yourself of details of calculating this statistic, look back at tutorial 1 and the Key formulae pdf for Module 1.

(b) In your written report, please provide a brief description of the approach you took to comparing the interventions, supported by a table or graph summarising the results (e.g. the values for your T statistics for each intervention) and your interpretation of the most effective intervention based on these results. In addition, please provide the cohort graphs for your favoured intervention in comparison to no-intervention.

(c) Finally, I would like you to briefly discuss which of these interventions you would prioritise to apply immediately in the Australian population, taking into account your results but also thinking through how quickly benefits would accrue over time. Note Im not expecting you to do any further analysis as part of this but just to reflect on some of what weve discussed in terms of how cohort effects translate to population models.

Submission details - Written presentation

It is perfectly fine to provide the written presentation as a series of tasks, with a written description andresults under each task heading. A narrative presentation is not advised.

For each task, note the following.

1. Graphs need to be appropriately presented with sufficient detail to stand alone. As a guide to this,all graphs need captions, axis titles and typically a legend unless only one data series is shown.
2. Results need to be briefly discussed in the text and any important limitations need to be clearlystated.

There is no formal need for referencing to complete the paper but if information from outside the coursematerials is included, this should be appropriately referenced in-text and included in a reference list.

Marking criteria

• You will need to complete three tasks relating to a model ofsmoking prevalence and mortality.
• The assessment is worth 30 marks in total, with 5 marks allocated to Task 1, 15 marks to Task 2, 5 marks to Task 3 and 5 marks to overall quality of writing and presentation.