A Team Approach to Data Quality A despicable data story! Presented by: Shane Downey & Sue Gardiner, Mater Health Services Outpatients: Accelerating Patient Flow and Improving Service Integration Change Champions Seminar: 19 October 2015
The Mater Story Founded in 1906 by the Sisters of Mercy Values-driven Collocation of public & private facilities Several hospitals, health centres, a medical research institute, pathology & pharmacy businesses all with one aim to provide Exceptional Care
Some Fast Facts Source: Mater Health Services 2014 Annual Review
Source: Mater Health Services 2014 Annual Review
An Episode of Care through Ambulatory Services
Introduction of Activity Based Funding In 2012 the Australian Government announced changes to hospital funding that required the submission of accurate and complete patient data, a significant change to the aggregated totals that were reported at the time. Introduction of the ABF model represented a real challenge to an organisation using best-of-breed health solutions 6 month project to investigate the impacts to the business, systems and processes Revealing results - unfunded services, inconsistencies in capture and reporting of information Identification of new data elements required for the national reporting requirements and gaps in existing solutions Recommendations written - including the need for a better approach to meeting our statutory reporting requirements
Solution environment Mater are a best of breed solution adopter Each Clinical area has its own Patient Administration System: Mater Adults uses ipm Mater Private uses PractiX Allied Health uses TAHDIS (an in-house solution) Mater Cancer Care Centre uses CHARM Each system has different capabilities for reporting, and different data structures for storing and representing data Not every system captures the required data
The challenge With data and information being captured in several key systems (i.e. separate PAS for public and private patients, theatre booking and management systems, pathology, pharmacy, and oncology) - how could we meet this new requirement to provide accurate Patient level administrative, demographic and scheduling data within the required timeframes?
Current state information architecture ipm is the source of truth for Patient demographics and the Master Patient Index, but not necessarily for service event and schedule management Current reporting involves manually running system reports and SQL queries, loading data into Excel and then massage into something useful Our BI team report the same data but use different processes and often there is misalignment between results due to data having changed between activities Our Casemix team were providing a third report (subset of the same data) using their own processes and so again we often see a misalignment of data The problem of misaligned data and reporting is not specific to Healthcare! All 5 systems are integrated via the Mater Clinical Data Repository (MCDR); an integration hub based on InterSystems Healthshare Foundations
The Vision To create an extensible solution that centralises all of the data required for statutory and corporate reporting in relation to outpatient service events, based on robust and transparent business rules and error management
Solution design Adopted a Master Data Management approach: MDM principles of Governance, Intelligence, Integration & Security Utilised agile principles Established a combined business and technology working party including key decision makers Frequent build and test cycles Minimal documentation
Project Governance Recommendation from the ABF gap analysis report Aligned teams represented in the centre Supported by the information management lifecycle Working party supported by the PADG Committee
Data Governance Access to data must be authorised Requests must satisfy acceptable use criteria in accordance with both Mater policy and State and Federal laws Data is validated on the way in Any errors are sent to the relevant teams including line managers on a nightly basis via email All reported errors must be corrected within 4 business days Errors not corrected within that time are flagged on a weekly & monthly KPI report that is sent to the Ambulatory Director for follow up Supported by our Information Management Framework, and Information Security & Control Policy
Data Integration Data is warehoused internally based on either HL7 message triggers or overnight data pull Data is stored based on a simple star schema An interface provides service activity data to our Casemix team for costing and billing This same interface is used to send data to Queensland s Department of Health to meet our statutory reporting requirements
Data Flow
Business process change Data quality was used to drive business process review and change Education to staff around the importance of getting the data right the first time A net resulting decrease in data quality errors by 80%!
Daily trending
Monthly trending
Key metrics KPIs July 2012 to Jan 2015 454,239 Outpatient Appointments since July 2012 62,822 Outpatient Appointments that contain errors 14% Error rate since July 2012 13% Invalid Outcome 236 Clinics with 100% error rate Feb 2013 Feb 2014 Feb 2015 Outpatient 9,874 11,060 (+11%) 13,827 (+29%) Appointments Outpatient 3,867 736 (-81%) 559 (-86%) Appointments that contain errors Error Rate 39% 7% 4% Invalid Outcome 38% 6% 3% Clinics with 100% error rate 18 3 (-83%) 2 (-89%)
Key metrics (cont.) Total Data Quality Cost 32,589 Outpatient appointments that contain errors 2 mins Average time to fix an appointment $62,822 Hidden cost of repairing all outpatient appointments with errors ($228 per day) 276 Work days spent correcting data quality errors The cost of doing nothing in terms of potential lost revenue. $13,129,798.00 Operational 46 Administration FTE (59 staff) 235 Outpatient Clinics $0 Total capital investment in the solution
The New Vision To accelerate patient care through robust data management resulting in an increase to patient attendances to 95%, and a reduction of data errors to fewer than 2%
Future State Continue to proactively manage FTA (Fail to Attend) rates through a more patient-centred approach Introduce emailing of appointment letters to consenting patients Establish a Data Quality Assurance capability Develop a worklist model for managing errors
Worklist Model Real time error management through a web based workflow Each time an error is detected it goes onto an individualised queue Staff correct the error in the source system and the error drops off the list Staff are able to escalate an error to a Data Quality Officer if they are unable to correct it Supervisors and Managers will be able to view stats Individuals work queues should be empty when staff go home each day Rewards-based approach aggregate points for zero errors
Lessons learned Administrative data quality will get you every time. wash, rinse & repeat! Start measuring the data quality before you make any changes. You need to quantify how bad is bad and have a baseline to benchmark changes against Use the data to drive business process change. Look for bright spots and focus on areas with the highest errors first Service review just because we can doesn t mean we should Resource effectively a second data analyst would have significantly reduced the overall delivery time
Lessons learned (cont.) MDM is a great approach but is hard going unless you have a project to drive it Bring front of house and back of house together to collaborate on the rules and the data management processes The data is always right - If it doesn t look right check your processes back to source Data quality begins at the point of capture - Everything that happens after that is waste
Team award http://iaidq.org/ Winner Data Quality Award 2014 Team Project Role Person Manager Data Services / Information Architect Shane Downey Practice Manager Susan Gardiner Quality & Education Officer Skye Ring Senior Data Analyst Specialist Fiona King Senior Data Integration Specialist Andy Richards