"Quality Issues: Collection, Coding, Metadata - some NHS experiences" Philip Johnston NHS Scotland, Information Services Division Quality Issues: Collection, Coding, Metadata - some NHS experiences • about Information Services Division (ISD) • about ISD data • ISD processes, and some examples • some statistics on quality about ISD • part of National Services Scotland, and NHS • 40+ years old • around 650 staff across Scotland • Scotland's national organisation for health information, statistics and IT services • 17 work programmes – eg waiting times, mental health about ISD data • around 60 data sets, evolving • variety of data providers – eg: hospitals, NHS boards, GP practices, voluntary sector, private sector • per year: ~60 publications, ~3000 information requests, ~500 Parliamentary Questions • Part of National Statistics • “Scotland has some of the best health service data in the world” ISD data : some context • some uses: clinical governance, planning, epidemiology, performance management, setting policy • some stakeholders: NHSScotland, Scottish Executive, Scottish Parliament, local authorities, Audit Scotland, GRO, researchers, media, public, political parties • some influences: health policy, devolution, National Statistics, Freedom of Information, data protection, patient involvement, IT developments (eg web), media awareness ISD data : SMR01, an example dataset • patient/episode -based • acute inpatients / day cases • ~100 data items, including clinical codes (ICD, OPCS) • ~1.2m records/year SMR01 data collection • electronic, secure (encrypted) • varying timetables for submission • data for ~1000 locations, from ~17 IT systems • ~100 validation files, with ~3m records • ~1500 error/query checks Main condition accuracy at 3-digit level Scot land TEACHING HOSPITALS West er n Inf ir mar y/ Gar t navel Gen & Beat son Hospit als Royal Inf ir mar y of Edinbur gh, Cit y & PMR Or t ho. Hospit als Glasgow Royal Inf ir mar y & Canniesbur n Hospit al Aber deen Royal Inf ir mar y & Woodend Hospit al West er n Gener al Hospit al, Edinbur gh Ninewells Hospit al LARGE GENERAL HOSPITALS Cr osshouse Hospit al St John' s Hospit al at Howden SMR01 quality : accuracy Vict or ia Hospit al, Fif e Vict or ia Inf ir mar y, Glasgow Queen Mar gar et Hospit al, Fif e Inver clyde Royal Hospit al Raigmor e, Cait hness & Belf or d Hospit als Bor der s Gener al Hospit al St obhill Hospit al Per t h Royal Inf ir mar y The Ayr Hospit al Falkir k and Dist r ict Royal Inf ir mar y Wishaw Gener al Hospit al • Accuracy of coding of main condition, 200002, by hospital St ir ling Royal Inf ir mar y Hair myr es Hospit al Monklands Hospit al Sout her n Gener al Hospit al Royal Alexandr a Hospit al Dumf r ies & Galloway Royal Inf ir mar y SMALL GENERAL HOSPITALS St r acat hr o Hospit al Vale of Leven Dist r ict Gener al Hospit al Dr Gr ay' s Hospit al, Elgin Lor n & Islands Dist r ict Gener al Hospit al CHILDREN' S HOSPITALS Royal Hospit al f or Sick Childr en, Edinbur gh Royal Hospit al f or Sick Childr en, Yor khill Royal Aber deen Childr en' s Hospit al 0 10 20 30 40 50 60 70 80 Percentage accuracy (target 90%) 90 100 SMR01 quality : timeliness SMR01 quality : other measures • • • • • • • validity completeness fitness for purpose relevance coherence comparability data ‘sign off’ SMR01 quality – some factors • conflicting local/national priorities • IT system issues (eg purpose / specification / delivery) • timeliness vs accuracy • quality of source data (eg case notes) • training and definitions • changes in service delivery SMR01 : example of accuracy versus timeliness SMR01 : some metadata • definition? • Data Dictionary (web-based; ~1500 items), including definitions and recording manuals • statistics on quality of data • data audit controls • Change Control process • ‘health warnings’ on outputs • statistical imputation / suppression other data collection : some examples • census of Allied Health Professionals (non-NHSnet; developmental) • Scottish Birth Record (national repository) • the future, eg: clinical datasets, electronic health record, SNOMED, data warehouses Quality Issues: Collection, Coding, Metadata - some NHS experiences – in summary • range of complex, interdependent factors affect data quality • ISD applies considerable resource to data quality • ISD ability to measure data quality, and to understand influences, is essential and is much valued by users