NCD Risk Factor: Physical Inactivity

Overview

This repository provides a translation of the risk factor NC_RFInactivity from Spectrum. It serves as a foundation for future modifications of the risk factor and its associated interventions as new model releases become available.

Data Components

Risk Factors analysis requires three types of data:

1. Prevalence
2. Relative Risks
3. Intervention Impacts (Impact Factors)

Data Sources

Data Type	File Path	Worksheet Name
Prevalence	./data/GBD_Country_Data.xlsx	PIA_Guthold_2018
Relative Risks	./data/RiskFactorData.xls	RR
Impact Factors	./data/RiskFactorData.xls	ImpactFactors

Accessing the Data

The Spectrum data is available in the data/ directory of this repository.

For a computer-friendly version of this data:

1. Clone this repository
2. Run the ./extract_data.sh script

Prerequisites:

• Bash shell
• Python (accessible via python command)

Interventions

P1: Primary Care Integration

Intervention: Provide physical activity assessment, counselling, and behaviour change support as part of routine primary health care services through the use of a brief intervention.

Implementation:

• Brief advice as part of routine care (95% coverage)

P2: Population-wide Communication Campaigns

Intervention: Implement sustained, population-wide, best practice communication campaigns to promote physical activity, with links to community-based programmes and environmental improvements to enable and support behaviour change.

Implementation:

• Awareness campaigns to encourage increased physical activity (95% coverage)

How the models work?

TLDR; The models work by modifying the incidence rate of a disease. In the model image above, you can see a flow from yellow, to blue, to orange. The final node in orange, NC_RFStroke-PAF-Marginal stores the calculated marginal population attributable fraction, which modifies incidence. Therefore, this node would be multiplied against a transition rate in another model (e.g. Stroke Incidence in CVD) to modify it.

Here are the steps:

Step 1: Calculate the Baseline Population Attributable Fraction (PAF)

• the baseline coverage of the risk factor is loaded. We store this as RFInactivity-BaselinePrevalence
• The relative risk of the risk factor and the disease is loaded into RFInatcitivty-RR-Stroke
• The "excess risk" of this relative risk is calculated by subtracting 1 from the relative risk. This is stored in RFInactivity-RR-Stroke-ExcessRisk
• We multiply the prevalence by the excess risk, and store this in the node RFInactivity-Stroke-prev_rr_baseline
• We add this value to a node called NC_RFStroke-PAF_Denom_Baseline. This already has a value of 1, and the expectation is that all risk factors that impact stroke will have their "prev_rr" values added to this. Currently, as each model only has one risk factor, this is only done once.
  • "denom" here is referring to a "denominator". We are going to use this as a denominator in the next calculation.
• The baseline PAF (NC_RFStroke-PAF-baseline) is calculated as RFInactivity-Stroke-prev_rr_baseline / NC_RFStroke-PAF_Denom_Baseline

Step 2: Calculate the Impact of the Intervention on Prevalence

• the effect of the intervention is loaded in and stored in NC_RFPhysicalActivityAwreness-Impact-Prevalence
• the baseline coverage is set in NC_RFPhysicalActivityAwreness-BaselineCoverage and the target coverage is set in NC_RFPhysicalActivityAwreness-Target-Coverage
  • We set the baseline coverage as a negative number, which will be relevant in the next step
• We calculate the difference in coverage by adding both coverages together (remember, baseline is a negative)
• The marginal impact is calculated by multiplying the difference in coverage by the effect size. This is stored as NC_RFPhysicalActivityAwreness-Impact-MarginalImpact
• This is added to a node called Summed_Marginal_Impacts which is kept in case several interventions are being considered.

Step 3: Calculate the Target/Current Population Attributable Fraction

This follows the exact same steps as Step 1, except instead of just using RFInactivity-BaselinePrevalence we are using that value, then modifying it by multiplying it against the (1 - Summed_Marginal_Impacts).

Calculate the Marginal Population Attributable Fraction

• We are left with NC_RFStroke-PAF-baseline and NC_RFStroke-PAF-Current.
• The marginal paf (NC_RFStroke-PAF-Marginal) is simply Current minus baseline.

Contributing

[Notes on how to contribute]

License

[Insert appropriate license information here]

Contact

For questions or further information, please contact [insert contact information].