Detect, Protect and Perfect (DPP)

The NHS Long Term Plan set out the aim of identifying and supporting patients at risk of stroke. Atrial fibrillation (AF) is a key risk factor for stroke and has become a focus for transformative interventions.

During 2022/23 and 2023/24, c.£26 million was invested in a funding programme to expand access to direct oral anticoagulants (DOACs) - recommended by NICE as the most effective treatment for preventing clots that can cause strokes in patients with AF. The programme had three pillars:

• Detect: reduce incidence of stroke by diagnosing more patients with AF

• Protect: ensure patients diagnosed with AF are offered anticoagulation, where clinically appropriate

• Perfect: ensure patients with AF are on the correct dose of the best value DOAC where clinically appropriate.

About this repo

All data used in this analysis is within the public domain. Details regarding funding applications have been withheld.

The recommended approach to re-create the analysis is:

• Install dependencies,

• Obtain matching data,

• Obtain outcome data,

• Run the analysis files.

Details on how to perform each step are provided below.

Installing dependencies

Run the get_dependencies.R script to identify the packages used within this analysis. NB, this file requires at least renv, janitor and dplyr to first be installed on your system to run.

Matching data

This analysis uses a Propensity Score Matching Differences-in-Difference (PSM-DiD) approach. In this method, GP practices which are part of the intervention group are matched with GP practices that are not part of any DPP project (the control group).

The variables used to match practices are grouped into clinical and socio-economic factors:

Clinical factors

Matching variable	Rationale	Data source
Proportion of the practice population aged 65 years or older	The risk of developing AF doubles with each progressive decade and exceeds 20% by age 80 years	GP registered population data by NHS Digital
Prevalence of obesity at the practice	Obesity (defined as BMI >= 30) is an independent risk factor for the development of AF	Fingertips
Prevalence of diabetes at the practice	Diabetes mellitus is an independent risk factor for AF, especially in young people	Fingertips
Prevalence of hypertension at the practice	People with hypertension have a 1.7-fold higher risk of developing AF compared with people with blood pressure in the normal range	Fingertips
Gender proportion	Age-adjusted incidence, prevalence and lifetime risk of AF is higher in men compared with women	GP registered population data by NHS Digital

Socio-economic factors

Matching variable	Rationale	Data source
Registered practice population per full time equivalent (FTE) clinician	The number of patients per clinical WTE can be a marker for a range of factors such as funding available to the practice, presence of other clinical staff, clinical needs of the local population and changes in the local population	GP workforce data by NHS Digital
Weighted deprivation score per practice	Greater socio-economic deprivation is associated with poorer outcomes across a range of health issues, including AF	Fingertips
Rural-Urban classification of the practice	Distance can be a barrier to accessing healthcare. Rurality can be a proxy measure for ease of healthcare access and has implications in the care of people with AF	Rural-Urban classification based on practice postcode

These data for these variables need to be downloaded.

• Get a list of GP practice details from the ODS API using get_gp_details.R

• Get monthly details of GP practice registered populations from NHS Digital using get_gp_registered_patients.R

• Process this data to extract out relevant details using get_gp_population_summary.R

• Get matching data from Fingertips using data_fingertips.R

• Download GP workforce data for April 2023

• Process GP workforce information using get_gp_workforce.R

• Process all the matching data using get_matching_variables_df.R

Outcome data

The outcome data was obtained from the CVDPREVENT audit via their API.

Use the get_cvd_data.R script to download the data.

Impute missing matching data

• Investigate the amount of missing data using dpp_matching_variables_missingness.qmd

• Perform multiple imputation to ‘fill-in’ missing matching data using dpp_impute_missing_matching_variables.qmd

Assign GP practices to projects

GP practices are assigned to each funded intervention based on the geography specified as part of funding applications.

• Get a lookup from GP practice to PCN, CCG, ICB and NHS Region from NHS Digital. Name this file intervention_counterfac_lu.csv.

• Get a list of funded initiatives and their respective start / end dates using get_dpp_project_final_list.R NB, this requires a copy of the funding application spreadsheet, which is not currently publicly available.

• Link funding applications with practices using link_grants_with_practices.R NB, this step is not possible without the above list of funding applications.

Some hospital-based projects have a unique list of practices or use a method of practice assignment which is based on dominant provider of care based on MSOA geography. To assign practices based on this approach you need to:

• Download a lookup of hospital catchment areas produced by Office for Health Improvement and Disparities (OHID). NB, this is the ‘2022 Trust Catchment Populations_Supplementary MSOA Analysis.xlsx’ file.

• Download a lookup from postcode to MSOA from the Open Geography Portal.

• Use get_gps_per_hospital.R to produce catchment areas based on the dominant provider of care in each MSOA, which are then linked with GP practices based on postcode to MSOA lookup.

Matching and impact analyses

The process of matching intervention GP practices with control GP practices and the impact analyses are done at national, case-study (aggregate) and case-study (individual) levels.

• The national-level analysis can be reproduced using dpp_programme_report_template.qmd.

• The case-study (aggregate) analyses can be reproduced using dpp_overall_report_template.qmd.

• The individual case-study reports can all be reproduced using dpp_gp_report_template.qmd. Which project the template produces a report for is controlled by the parameters set within the yaml header of the document:

  • project_id controls which project is reported, using the ‘P’ nomenclature, where ‘P1’ is the ID for project 1.

  • alternate_data is used where there are multiple ways of assigning GP practices to the intervention area. Setting this to FALSE will mean the default GP practice assignment method defined in link_grants_with_practices.R is used. Setting to TRUE (default) will use any alternative method of assignment, as detailed in the dpp_gp_report_template.qmd script.

  • alternate_data_details is used to specify which of multiple GP practice assignment methods are to be used. Commonly used parameters include “emergency” and “af”. “emergency” is used for hospital-based projects to assign GP practices based on MSOAs where the hospital is the dominant provider of emergency care. “af” is used for a single project in which only GP practices which undertook activities to improve care of atrial fibrillation are to be counted.

  • save_did is used to control whether the underlying data is saved for future use with the calc_percentage_difference.R script.

---
params:
    project_id: "P20"
    alternate_data: TRUE
    alternate_data_details: ""
    save_did: TRUE
---