In md0u80c9/SSNAPStats: Stroke Sentinel National Audit Project (SSNAP) Statistics Calculations

knitr::opts_chunk$set(collapse = TRUE,
                      comment = "#>",
                      fig.width = 7,
                      fig.height = 7,
                      fig.align = "center")

About the SSNAP R packages

SSNAPStats is a package designed to produce statistics from national audit data. It is specifically designed around Stroke Sentinel National Audit data (see http://www.strokeaudit.org), but recent developments have made it sufficiently generic to be potentially applicable to other audits in future.

SSNAPStats is part of a group of packages for processing SSNAP data:

DiagrammeR::DiagrammeR("graph TB;
  A>Raw SSNAP Data]-->B(SSNAPInterface);
  B-->C(SSNAPStats);
  C-->E(SSNAPReports);
  D(SSNAPCharts)-->E;
  D-->F((SSNAP Dashboard));
  C-->F")

• SSNAPInterface loads raw CSV file data.
• SSNAPStats processes loaded SSNAP data, organises the data into cohorts and prepares aggregated data for reporting.
• SSNAPCharts provides generic report outputs for reports.
• SSNAPReports produces reports from SSNAP data.
• SSNAPDashboard is a Shiny-based application for realtime SSNAP data analysis.

How does SSNAPStats analyse audit data?

SSNAPStats breaks down the audit processing into a set of key steps:

DiagrammeR::DiagrammeR("graph TB;
  A(cohorts)-->D{create_cohort};
  B(audit_measure)-->C(audit_output_spec);
  C-->D")

The first step is to set up rules to produce a definition of the cohorts you want to use. Cohorts are time series-orientated collections of patient records.

The second stage is to set up rules to produce audit measures. These rules produce either discrete results (ie. counts of a particular value, or proportions of patients meeting or not meeting a condition,) or they may be continuous (ie. to create median values).

Different outputs of results from an audit require different levels of detail: the more detail you request the longer it takes to process. An audit output specification is used to produce a list of the audit measures you want, alongside how much detail from each measure in the list, without having to re-write the same measure multiple times to get different levels of detail for different purposes.

Finally we use the create_cohort() function to apply the cohort definition with the audit output specification to produce the results.

Shortcuts to make routine SSNAP data analysis easier

So far the process we have described is applicable to all audits. With SSNAP, there are predefined cohort definitions, and predefined lists of audit measures that are used for routine published analysis.

To make things simpler, there are a set of lists and functions, all prefixed with ssnap_, that make handling routine data easier. These use pre-built cohorts and measures within the package. There are also functions to build a 'ssnapScores' table - which takes aggregated data from multiple cohorts and produces the scored reports seen in published SSNAP data. These functions are all built using the lower level generic functions described earlier.

• ssnap_field is a named list of 'record level' expressions (such as time differences). These are inserted into audit measures to make the code easier to read and/or reduce duplication.
• ssnap_measures is a named list of audit measures used in SSNAP.
• ssnap_cohort_definitions is a named list of cohort definitions.
• The ssnap_audit_output_specs.R file defines a set of audit output specifications for cohorts used in SSNAP analysis. Each one of these is a tibble.
• ssnap_cohorts is a simplified version of the create_cohort functions. It creates cohorts using the pre-defined cohorts and measures.

SSNAP reporting periods have changed over time from three monthly to four monthly and back again; therefore there are also definitions for the reporting periods and annual report periods for time series analysis and filtering.

md0u80c9/SSNAPStats documentation built on Feb. 11, 2020, 11:43 a.m.