Modelling patient flow in ED to better understand demand management strategies. Elizabeth Allkins Sponsor Supervisor Danny Antebi University Supervisors Dr Julie Vile and Dr Janet Williams
Contents Background Literature review Statistical analysis Simulation model What-if scenarios Going forward..
Aims Gain insight into the functioning of the Emergency Department in the Royal Gwent Hospital Explore the effect on the system of actions to redistribute demand, reduce overcrowding and long waiting times
Background
Background Welsh A&E waiting time target missed again Penarth Times, 19th August 2013 Concern over 24-hour A&E waits in Wales BBC News, 30th August 2013 Catastrophic mistakes and shortcomings have been identified that should not - and must not - be dismissed by those in charge Andrew Davies, Leader, Welsh Conservatives Ambulance response targets missed across Wales BBC News, 24th April
Two complementary approaches Reduce attendance at ED Improve flow through ED
This project Masters project with Cardiff University, sponsored by ABCi. Using Operational Research and Statistics to provide mathematical insight.
This project Literature review Statistical analysis Simulation modelling What-if scenarios
Lit Review Operational Research is relatively new to healthcare. Overcrowded emergency departments and long waiting times are a widespread issue. It is difficult to implement academic recommendations in the real world without continued clinician support (a champion).
Stats Analysis: Attendance Breaching the 4 hour target is not related to the number of ED attendances Hourly pattern Split of Majors by source of service Average attendance by hour 20.00 Monday 18.00 Tuesday 16.00 Wednesday 14.00 Thursday 12.00 Friday 10.00 8.00 6.00 4.00 2.00 0.00 Saturday Sunday
Proportion of attendance to Majors by source of service request Emergency services Self referral
Stats analysis: Length of Stay Category (severity) accounts for 6% of the variation in LOS Age accounts for 8%; older people stay longer
Stats analysis: Regression model Regression model 50% of attendance can be explained by these factors in this equation. Attendance = 230 + (27 * Monday) + (21 * Sunday) - (20 * December) - (12 * Friday) + (7 * March) (16 * November) - (16 * January) - (8 * School Holiday) - (8 * February)
Stats analysis: Introduction of CDU in 2013 Impact on length of stay Average LOS in ED (minutes) Year Total Majors Minors Paeds 2012 195 295 135 155 2013 193 288/251 143 156 Change in use of MAU 540 fewer patients admitted to the MAU in 2013 than 2012 Average time spent in MAU has increased from 798 to 980 minutes
Misplaced patients Minors Majors Category 1, 2 or 3 Category 4, 5 or 6
Simulation model DES model built in Simul8
Simulation model: Structure
Simulation model: Parameters
Simulation model: Resources Staff Nurses Doctors Call handler Receptionist Beds Beds for Majors and wards Rooms for Minors Xray machine
Validation and Verification 999 call Non-mathematical Visual Clarity Logic Mathematical Comparison of key statistics between model and real data Other (e.g. Nursing home GP referral Ambulance Walk ins Walk ins Call handler Paeds Majors Minors Registration within half an hour Registration within half an hour Registration within half an hour Triage: ENP Triage: ENP? Another hospital Wait for bed Initial assessment: junior doctor or ENP A&E clerking Triage: ENP GP/GP out of hours Treatment or investigation: ENP Treatment A&E review by junior doctor Seen by ENP (about two thirds) Further tests Surgical SAU Medical MAU Seen by doctor (about one third)? Referred CDU Other? Referral to speciality Tests Medical clerking Admitted ACP Review Admitted Discharged Review Treatment Discharge
Final model
What-if scenarios CDU Use as a fast-track stream Use as an additional ward with 12 extra beds. Comparative average LOS for discharged Majors patients 2012 baseline Introduction of CDU Earlier routing to CDU Split beds (50%<1 day, 50% <3 days) Increase routing to CDU 190 200 210 220 230 LOS (minutes) 240 250
What-if scenarios WAST pre-hospital streaming Streaming ambulance patients direct to the MAU GP trial Streaming GP referrals to a bed in the MAU Reduction in WAST conveyance rates Reduction of 10% reduced waiting times
Limitations A computer model of a very complex system! Data limitations Human behaviour
Conclusions Detailed analysis of ED data Cost saving of CDU Reduction in Majors LOS Simulation Demonstrated the power of modelling Explored scenarios to improve waiting times Built a solid foundation for future research
Any questions? Elizabeth.Allkins@gmail.com Julie.Vile@wales.nhs.uk