Physical activity exposure and impact

Individual-level MMETs

Using trip sets and the synthetic population, total_mmet computes total MMETs per person as the sum of walking MMETs per day and cycling MMETs per day, which are scaled up to a week via the PA scalar, and work/leisure MMETs per week.

We calculate the proportion of people who do no work/leisure PA according to the confidence in the PA survey. First, we calculate the raw probability that a person in demographic group completes no non-travel PA. Then, we set the probability to use as the raw value if our confidence is 1. If it's less than, we map the confidence to parametrise a Beta distribution.

Finally, for each person in the population, we sample non-travel MMETs as zero with the calculated probability and from the raw non-zero density of their demographic group.

PA--disease dose--response relative risk

gen_pa_rr uses each person's MMETs per week to compute their relative risk of six diseases, using curves found from meta analysis. Each disease (except type 2 diabetes) has a threshold beyond which there is no further change in relative risk. (Total cancer, breast cancer, colon & rectum cancer, endometrial cancer, & coronary heart disease: 35; lung cancer: 10; stroke: 32; all cause: 16.08.) For all other diseases, we use mortality.

The relative risk for each person for each disease is calculated in PA_dose_response using the curve selected, which is an interpolated mean or a sample of functions. See https://www.overleaf.com/read/mrjtkhffzfzr for an example.

Combining PA and AP relative risks

combined_rr_ap_pa combines the relative risks of gen_ap_rr and gen_pa_rr through multiplication. Not all diseases have a dose--response relationship for both AP and PA. Just stroke, lung cancer, and IHD have both. For the other diseases, only one of the RRs is different from one.