Background Image

Methodology

Read the NCD-RisC methodology

NCD RisC work has two key components. The first is data curation and management, which involves the sourcing, harmonisation, and organisation of studies that contain data on NCDs and their risk factors. The second is statistical analysis, which involves the use of bespoke statistical models to obtain estimates of patterns and trends in the data. The key aspects of our data curation and management are outlined below. For the details of our statistical analysis please see each specific studies' publications.


  • Inclusion/exclusion criteria - read more
    To ensure only high quality studies that are representative of general population are used, all studies in the NCD-RisC database are assessed against inclusion and exclusion criteria.

  • Data cleaning - read more
    NCD-RisC have developed a data cleaning procedure which removes data with values that are possible data entry errors. This process is done in three stages:
    1. Checking all variables against their respective univariate cleaning criteria;
    2. Checking that all data meet relevant multivariate constraints (e.g., systolic blood pressure should be higher than diastolic blood pressure);
    3. Detecting multivariate outliers by pairs of related variables (e.g., BMI and waist circumference) based on the relationship of the relevant variables in the entirely of dataset.

  • Conversion to metrics with standard definition
    For a small number of data sources in the NCD-RisC database, only a subset of metrics of interest are available. In such cases, conversion regression models have been developed to estimate other metrics of interest from the available data.
    1. BMI - read more