Accurately identifying variation in performance between GP Practices allows poor performers to improve by prioritising areas for action.
It has taken three years and we have been down a few rabbit holes but I think my team at Trayned Insight have found a way of using open source data to evaluate performance of GP practices across England in a totally new and effective way. Our approach does not tell commissioners or regulators what to do, but it very clearly identifies variation in performance in a host of indicators in a way that is statistically valid and that forms the basis for action. We have identified very large opportunities for savings in both money and by reducing avoidable presentations at A&E or admissions to hospital.
Existing benchmarking tools rely mainly on national averages to assess GP practices but this approach makes no allowance for “environmental” population characteristics that affect outcomes. Our approach, which we are calling Trayned Pioneer, identifies which uncontrollable factors affect the outcomes of each indicator, e.g: demographics, disease prevalence, deprivation. It then sifts all characteristics and pinpoints which parameters are most closely shared between GP practices. It then creates custom peer groups of the 99 most similar practices (nearest 99 or “N99”) for every primary care practice across England on the basis of the shared characteristics that are most important to each indicator. In this way about 8,000 peer groups are created for every indicator – hundreds of thousands of groups in total. The indicator value for each GP practice may then be compared to its closest peers for that indicator and each practice in a peer group of 100 is ranked according to the indicator values and divided into ten equal groups – deciles – for ease of analysis and visualisation.
The output displays the practices in a grouping, typically within a given CCG, in a simple bar graph that shows where they sit within the deciles measured against their peers. Thus, they are not being compared to each other, but against the level of performance that they might have been expected to achieve, all things being equal. The convention we have adopted is that deciles to the left of the graph always represent poor performance or ranking. This approach allows commissioners to Identify real opportunities for improvement and real examples of best practice so that they can decide where to prioritise their efforts to make improvements in care and cost savings.
The second output is a simple graph that enables a user to hover over any practice in the grouping to display performance within that practice against all of the indicators. Again this is displayed in a bar graph that shows the relative ranking for that practice within its peer group for each indicator. The practice can then decide where to focus improvement activity internally, but can also identify practices within its grouping that score more highly, and work with them to identify what they are doing differently to achieve higher performance.
We are building much more information into Trayned Pioneer, including more indicators and a tool that allows performance to be linked to specific therapy areas according to BNF chapters, and then to individual drugs and their cost. But in this first edition we are inviting users to tell us which indicators they would like to see displayed. You can see an open version of Trayned Pioneer on our website that includes 20 indicators and allows closer view of two practises per CCG.