Challenge brief

<section> <p><strong>Corrected dataset now available.</strong></p> </section> <section> <h2>UK and Global Health System Performance, 2018 to 2024</h2> <p>How did health systems in 32 countries absorb the pandemic shock, and who recovered? In this challenge you step into the role of a health system performance analyst reporting to the board of an international policy unit. Your job is to find the patterns in the data, measure them, and turn them into a clear, evidence-led report.</p> <p>This is a corrected reissue of the November challenge dataset. It replaces the version first published this month, so please build on the new file only. If you entered on the earlier file, you can submit again on the new data.</p> </section> <section> <blockquote> <p>This dataset is synthetic. It was simulated for a training challenge and does not report the real performance of any country, health system or organisation. Country names and reference attributes (WHO region, World Bank income group, OECD and EU membership, population) come from public sources so the scenario feels familiar. Every performance, workforce and spending figure is simulated, and the health system regions are fictional. Where a published country figure exists (for example hospital beds, health spending, doctors and nurses per 1,000), the simulated values are centred on it so the scale is realistic. Do not cite these figures as real-world evidence.</p> </blockquote> <p><strong>Please add this short footer to your charts:</strong> Synthetic data, DataDNA November 2026 challenge.</p> </section> <p><!-- Keep the existing Dataset Download button/form block here. Do not remove or recreate the form. --></p> <section> <h2>What You Will Analyse</h2> <ul> <li><strong>The pandemic shock in acute care:</strong> ICU occupancy, emergency department 4-hour performance and recorded respiratory illness across the scenario’s pandemic phases, and how the change differs by country.</li> <li><strong>The elective backlog and who recovered:</strong> elective waiting times in 2019 and 2021, overall and country by country. In this scenario you also follow waits in Sweden, Norway, Denmark, Finland and Germany after 2021, and compare the United Kingdom with its own 2019 level and with the EU member countries.</li> <li><strong>Health spending:</strong> in this scenario, health spending per person in the United Kingdom compared with the EU member countries.</li> <li><strong>The workforce under strain:</strong> the staff burnout index year by year, the link between burnout and vacancy rates, and, in this scenario, how vacancy rates and burnout in the United Kingdom compare with other countries.</li> <li><strong>Long-running disease trends:</strong> recorded cardiovascular disease and diabetes in each WHO region, and, in this scenario, recorded cancer incidence in high-income and low-income countries.</li> </ul> <p>The brief in the download tells you where to look, not what you will find: measuring each pattern, and saying clearly when you find none, is the challenge.</p> </section> <section> <h2>The Dataset</h2> <ul> <li>5 tables in a star schema: <code>dim_country</code> (32 countries), <code>dim_date</code> (28 calendar quarters), <code>dim_health_system</code> (212 fictional health system regions), <code>fact_health_performance</code> (5,936 rows, one per region per quarter) and <code>fact_workforce</code> (1,484 rows, one per region per annual survey).</li> <li>Period: 2018-01-01 to 2024-12-31, by calendar quarter, with each quarter labelled Pre-Pandemic, Acute Pandemic, Recovery or Post-Pandemic.</li> <li>Currency: USD.</li> <li>Included in the download: <code>README.md</code>, <code>CHALLENGE_BRIEF.md</code>, <code>docs/DATA_DICTIONARY.md</code> and the CSV files.</li> </ul> </section> <section> <h2>How Entries Are Judged</h2> <ul> <li><strong>Analytical accuracy:</strong> findings match the data and each headline number states the measure, filter and period behind it.</li> <li><strong>Coverage and depth:</strong> the analysis areas are explored with the right cuts, including country-level variation, not only overall averages.</li> <li><strong>Modelling and method:</strong> a sound star schema, averages rather than sums, correct comparisons between one country and a group, and trend slopes that are measured, not just drawn.</li> <li><strong>Visual communication:</strong> clear, well-labelled charts carrying the synthetic data footer.</li> <li><strong>Insight and judgement:</strong> recommendations framed for the scenario, honest about what the data cannot show, including saying so when a pattern is not there.</li> <li><strong>How the reviewer applies these:</strong> an entry is judged on whether it finds the patterns described in the dataset context, with the right direction, measured with a sound and stated method. The size of an estimate is judged by the method behind it; a reasonable method that gives a somewhat different number is not marked down, while a wrong direction or a pattern read from the wrong cut is. Entries need not cover every area: depth and correct measurement count for more than breadth. Reporting, with the cut used, that a relationship is not there is a valid finding.</li> <li>Caveats that the data is synthetic or not real-world evidence are correct and are never penalised. Findings about named countries should be framed as the scenario; simulated figures presented as real-world evidence, or outside statistics presented as findings from this dataset, are not credited.</li> <li>A participant who reported that no pattern could be found in the withdrawn November file is not penalised: that file has been replaced and the finding was correct.</li> </ul> <p><strong>Deadline:</strong> 24 November 2026</p> </section>