Let Hospital Workforce Data Talk A Data Visualisation Exercise Health & Biosecurity Yang Xie yang.xie@csiro.au HIC, 08-Aug-2017 THE AUSTRALIAN E-HEALTH RESEARCH CENTRE
Healthcare Marketplace: the big picture Source: Coursera, Healthcare Marketplace, by Professor Stephen T Parente
Fixing hospitals with data
What is the particular gap this work wants to fill?
There is evidence of correlation between staffing level with patient outcomes
Creation of useful workforce metrics is important Most current studies looking at workforce metrics rely on: Costly Bias Put sensors on doctors? Who dare?
Payroll data is an effective indicator of the hospital workforce allocation Available Low cost Objective Easy to scale
Input Data processing Output By processing the payroll data, we acquire the workforce metrics Data 34M payroll record 3M inpatient separation 1.1M patient Site 30 public hospitals in QLD Period Apr. 2013 Dec. 2015 2.75 years Step 1 Reshape and aggregate payroll data Step 2 Reshape separation records Step 3 Link workforce data to patient demand Absolute Counts Hourly counts of doctors, nurses and other staff in each hospital Supervision level Staff-to-staff ratios Link to patient demand Staff-to-patient ratios
Calendar heat map shows daily counts of all staff, weekends are usually short staffed Sun 2013 Sat 2014 2015 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Daily counts of all staff of a public hospital in the studied period
Absolute staff counts show vast variations in workforce pattern between hospitals Principal Referral Hospital Doctors Nurses Public acute Group C Hospital Mon Tue Wed Thu Fri Sat Sun Mon Tue Wed Thu Fri Sat Sun Noticeable afterhour/weekends differences Noticeable peer-group differences
Staff ratio reflects the supervision level of doctors and nurses Principal Referral Hospital Doctors ratio Nurses ratio Public acute Group C Hospital Mon Tue Wed Thu Fri Sat Sun Mon Tue Wed Thu Fri Sat Sun Percentage of threelevel doctors/nurses among all doctors/nurses Eliminate size effect Less senior staff in after-hours
Staff to patient ratio links the supply and demand Principal Referral Hospital Doctors/ Patient ratio Nurses/ Patient ratio Public acute Group C Hospital Mon Tue Wed Thu Fri Sat Sun Mon Tue Wed Thu Fri Sat Sun Number of doctors/number of patients Less medical staff per patient in after hours
Merits of this methodology Overcome bias and cost issues of conventional approaches Improve insight into staff dynamics & workforce variations Reliability Insight Applicabili ty Scalability Applicable to studies which needs evidence to support policy Automate to scale and apply to health informatics community
Potential applications Example 1 Evaluating impact on patient outcome associated with legislated changes to nurse-topatient ratio
Potential applications Example 2 Assessing whether the time of care makes any difference to patient outcomes.
Key takeaways Linking payroll data to patient admission creates useful metrics to model hospital service delivery. This methodology overcomes drawback of conventional approaches and improves insight into staff dynamics and workforce variations. It can be automated to scale and applied to a broad range of studies, especially where evidence is needed to support policy.
Thank you For more information, please contact: Yang Xie Postdoctoral Fellow T: +61 7 3253 3649 E: Yang.Xie@csiro.au W: www.aehrc.com Norm Good Senior Experimental Scientist T: +61 7 3253 3600 E: Norm.Good@csiro.au W: www.aehrc.com Sankalp Khanna Research Scientist T: +61 7 3253 3629 E: Sankalp.Khanna@csiro.au W: www.aehrc.com Justin Boyle Research Scientist T: +61 7 3253 3606 E: Justin.Boyle@csiro.au W: www.aehrc.com THE AUSTRALIAN E-HEALTH RESEARCH CENTRE