# Table 2 Model assumptions

Assumption Implication for modelling
6-month behaviours from the trial are assumed to represent year 1 behaviours in the model. May over-estimate the duration of U@Uni behavioural treatment effects by 6 months.
Health Survey for England complete case data represents behavioural patterns in the general population that is represented in the ONS Life Tables. Complete cases from Health Survey for England may be atypical and not representative of the general population in some way (unclear effect on results)
Probability of death aged >100 years = 1 May under-estimate life expectancy increases as a result of U@Uni.
Assumes health behaviours from Health Survey for England when aged over 90 are equal to those when aged 90 Health behaviours of people aged over 90 may be different from those of people aged 90 in some way (unclear effect on results).
Assumes relationship between alcohol and mortality risk is linear and between physical activity or fruit and vegetables and mortality risk is logarithmic (to avoid negative values for hazard ratios) The shape of the true relationship between the health behaviours and mortality risk may have a different functional form (unclear effect on results).
Hazard ratio for 0 portions fruit and vegetables = 1.6 May under- or over-estimate the increased mortality risk associated with eating no fruit and vegetables (unclear effect on results).
Hazard ratio for 0 minutes of physical activity = 1.6 May under- or over-estimate the increased mortality risk associated with doing no physical activity (unclear effect on results).
Except for due to the effect of U@Uni an individuals’ fruit and vegetable consumption, physical activity, and alcohol consumption measured 6-months after university are assumed to stay on the same age-specific percentile rank from the general population throughout their lifetime Individuals’ health behaviours could be expected to vary more than this over a lifetime (unclear effect on results).
Except for due to the effect of U@Uni an individuals’ smoking status measured 6-months after university is assumed to stay fixed throughout their lifetime Individuals’ smoking status could be expected to vary more than this over a lifetime (unclear effect on results).
Health behaviour change decays linearly up to the year of the maximum length of the treatment effect Behaviour change may decay non-linearly (unclear effect on results).
Hazard ratios for the effect of health behaviours on mortality risk are age independent The relative effect of health behaviours on mortality risk may vary with age (unclear effect on results).