Using Data to Assess and Improve Operational Performance in the Emergency Department Jody Crane, MD, MBA Chuck Noon, PhD
Objectives - Web and Action Summarize demand/capacity issues within the ED with respect to nursing and physicians hours and bed demand Determine the best capacity plan by hour of day based on characteristics of patient demand and determine the ideal schedule of staff Create an operational plan to drive future management decisions Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Objectives for Today To begin to understand basic concepts behind ED operations To provide an overview of analysis with EDucate © To orient participants to the data necessary to populate EDucate Inputs © To provide guidance for collecting and validating data for EDucate Inputs ©
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 What is EDucate ©? The Emergency Department Utilization and Capacity Assessment Tool in Excel © A tool to give Emergency Departments an accurate assessment of their current operational state to help guide improvement and operational redesign It is the most comprehensive, data-driven assessment tool currently available to help EDs focus on targets for operational improvement
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 EDucate © Summary EDucate © can be an effective tool for ─Fast Track and Triage planning ─Matching demand to capacity throughout your process flow ─Determining the correct team size ─Determining the impact of holds, diversions, and walkouts on staffing, throughput, and revenue ─Creating “what-if” scenarios for future projects
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Concepts Behind EDucate © EDucate © was developed based on ─A recognition that ED demand is random yet varies predictably throughout a day ─A view that ED operations can be modeled as a network of queues, with each queue having its own arrival and service characteristics ─The belief that understanding current demand/capacity misalignment is the first step in operational improvement
Server A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Person (MD, nurse, tech, transporter, housekeeper, etc.) or Resource (bed, scanner, equipment, etc.) A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Customer Arrivals A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Customer Arrivals Person (patient, call, etc.) or Thing (lab sample, soiled room, tray to be picked, etc.) A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Customer Departures A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Queue (waiting line) Customer Arrivals Customer Departures A Simple Queue Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Queue (waiting line) Customer Arrivals Customer Departures A Simple Queue Service Rate ( Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Server Queue (waiting line) Customer Arrivals Customer Departures A Simple Queue Arrival Rate ( Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
A Simple Queue Server Queue (waiting line) Customer Arrivals Customer Departures Arrival Rate ( Service Rate ( Avg Number in Queue ( L q ) Avg. Wait Time in Queue ( W q ) Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example 1 Suppose we have a triage operation staffed by a single nurse. Patients arrive and wait in the waiting area if the triage nurse is busy triaging other patients. When a patient is seen by the triage nurse, the triage activity occurs in a single encounter. Data was gathered and, on average, 6 patients arrive per hour. The average time it takes to triage a patient is 12 minutes. So, will there be any waiting? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example 1 Suppose we have a triage operation staffed by a single nurse. Patients arrive and wait in the waiting area if the triage nurse is busy triaging other patients. When a patient is seen by the triage nurse, the triage activity occurs in a single encounter. Data was gathered and, on average, 6 patients arrive per hour. The average time it takes to triage a patient is 12 minutes (equivalent to rate of 5 per hour). So, will there be any waiting? Most definitely. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example 2 Suppose we have a triage operation staffed by a single nurse. Patients arrive and wait in the waiting area if the triage nurse is busy triaging other patients. When a patient is seen by the triage nurse, the triage activity occurs in a single encounter. Data was gathered and, on average, 4 patients arrive per hour. The average time it takes to triage a patient is 12 minutes (again, a “service rate” of 5 patients/hour). So, will there be any waiting? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example 2 (cont’d) It depends. Let’s look at two situations. Situation 1 – A patient arrives exactly every 15 minutes and the time to triage each patient is exactly 12 minutes. Will there be any waiting? If so, how much? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example 2 (cont’d) It depends. Let’s look at two situations. Situation 1 – A patient arrives exactly every 15 minutes and the time to triage each patient is exactly 12 minutes. Will there be any waiting? If so, how much? There will be NO waiting. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example (cont’d) Situation 2 – Patients arrive according to a random arrival process (a Poisson process) at an average rate of 4 patients per hour. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Poisson Arrival Examples Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
From a sample of arrival data Arrival data from a California hospital. Mondays, 2pm-6pm. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example (cont’d) Situation 2 – Patients arrive according to a random arrival process (a Poisson process) at an average rate of 4 patients per hour. Triage (service) times are distributed according to an Exponential distribution with a mean of 12 minutes. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Exponential Distribution of Triage Times Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
From a sample of ED Triage Times. Average = 5.06 Std.Dev. = 4.97 Time study data from MWH. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example (cont’d) Situation 2 – Patients arrive according to a random arrival process (a Poisson process) at an average rate of 4 patients per hour. Triage (service) times are distributed according to an Exponential distribution with a mean of 12 minutes. Will there be any waiting? If so, how much? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
How can we estimate the waiting? We can use steady-state formulas for certain queuing interfaces. For example, in cases with Poisson arrivals, Exponentially distributed service times, and 1 server, the estimated wait in front of triage can be calculated using the formula: Alternatively, we could use any one of a number of computer simulation packages. Or, we can use a spreadsheet like QueueCalc which uses the steady-state formulas. Wq = / ( - ) Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
With the Coefficient of Variations set to 1, we are Assuming maximal variation with Poisson arrivals and Exponentially distributed service times. Values of 0 would imply no variation. Infinite Queue Approximation Worksheet INPUTS KEY OUTPUTS
Triage Example In the triage example, we had an arrival rate of 4/hour and a service rate of 5/hour (average triage time of 12 minutes). Assuming maximal variation, we can use QueueCalc to estimate steady-state values for: 1. Server Utilization __80%__ 2. Average Wait Time (Wq) _.8 hour___ 3. Average Line Length (Lq) _3.2___ Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example Revisited In the triage example, we had an arrival rate of 4/hour and a service rate of 5/hour. If the number of visits to the ED grew by 15%, what would happen to the waiting times? 1. Server Utilization _______ 2. Average Wait Time (Wq) ________ 3. Average Line Length (Lq) ________ Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Triage Example Revisited In the triage example, we had an arrival rate of 4/hour and a service rate of 5/hour. If the number of visits to the ED grew by 15%, what would happen to the waiting times? 1. Server Utilization __92%__ 2. Average Wait Time (Wq) _2.3 hours___ 3. Average Line Length (Lq) _10.6___ Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Queuing Interface Principles For queues with the arrival rate less than the service rate, and full-time server availability,… ─…there will be no waiting if there is no variation ─…variation can be present in the arrival process and/or the service process ─…as variation increases, waiting increases ─…in queues with considerable variation, waiting will increase dramatically as utilization increases and, hence, alignment of capacity with demand is key. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Typical Queue Behavior Waiting Time (minutes) Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
The average rate of arrivals varies fairly predictably throughout a day… Based on one year of arrival data for a California hospital with 58,000 visits. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
… but the actual count of arrivals for any given hour or day can vary considerably. This is arrival variation. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Medium sized 24-hour Emergency Department with 95 visits per day, on average. Example Problem - Situation Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Example Problem - Situation Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Medium sized 24-hour Emergency Department with 95 visits per day, on average. Physicians can treat 2.5 patients per hour, on average. So, 38 (computed as 95/2.5) physician- hours required per day. Therefore, staffing 2 physicians on duty 24-hours per day should provide sufficient capacity cushion (estimated utilization of 79%, computed as 38/48). Example Problem - Situation Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Example Problem - Situation Average Capacity Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Makes sense, but how will it really perform? Simulate for one year with assumed Poisson arrivals and Exponentially distributed service times. Example Problem Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Example Problem – Performance (average waiting time by hour of arrival) Average patient wait is 74 minutes Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Longest lines and waits occur at 9-11pm Average patient wait is 74 minutes (average waiting time by hour of arrival) Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Example Problem - Situation But that is not where the problem was caused. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Example Problem - Resolution Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Example Problem – Performance (average waiting time by hour of arrival) Average patient wait is 43 minutes Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Typical Queue Behavior Giving This Getting This Waiting Time (minutes) Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
An ED is a network of queues. Balance is crucial.
Network of Queues ED operations can be viewed as a network of queues, each of which must be properly configured to align capacity with demand. The completion through one server creates an immediate arrival for the next server if handoffs are clear & direct. S1S1 S2S2 S4S4 S3S3 Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 1 Nurse1 Doctor N2N2 4 Beds Treatment Times Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds Treatment Times 9 min. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Treatment Times 9 min. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Treatment Times 9 min. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Simulation Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times 9 min. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times 9 min. Total LOS of 5.0 HOURS Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Treatment Times 8 9 min. Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times Hours2.6 Hours 9 min. Total LOS of 4.0 HOURS Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times Hours2.6 Hours 9 min. Total LOS of 4.0 HOURS Caution! This assumes no increase in service times!
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times Hours2.6 Hours 9 min. Total LOS of 4.0 HOURS Let’s include a slight increase in service times
D Computer Simulation Results N1N1 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times Hours3.1 Hours 9 min. Total LOS of 6.8 HOURS! Let’s include a slight increase in service times.
D Computer Simulation Results N1N1 9 min. 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Treatment Times Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
D Computer Simulation Results N1N1 9 min. 27 min. 1 Nurse1 Doctor N2N2 9 min. 4 Beds 30 min. Average time between arrivals Waiting Room Patient Averages 3.3 hours in Waiting Room Patient Averages 1.7 hours in Exam Room Treatment Times 1.4 Hours1.6 Hours Total LOS of 3.0 HOURS Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
What do You Have to Tell the Tool?
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Inputs
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Inputs
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Inputs
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Inputs
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Inputs
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? Assume all 30 minutes in first hour? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? Unrealistic! Assume all 30 minutes in first hour? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? 10 Minutes in first hour 10 Minutes in second hour 10 Minutes in third hour Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? 10 Minutes in first hour Better, but not exactly the way we do things! 10 Minutes in second hour 10 Minutes in third hour Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? 15 Minutes in first hour 6 Minutes in second hour 9 Minutes in third hour Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour MD Demand 3-hour Length Of Stay MD spends 30 minutes per patient, but when? 15 Minutes in first hour EDUCATE © allows you to specify how MD (or RN) time is spent within a patient’s LOS. 6 Minutes in second hour 9 Minutes in third hour 50% in first 33%20% in second 33%30% in third 33% Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Inputs
Sources of Data Arrivals data (monthly, daily, hourly) – Most registration systems Walkouts (daily, time of day) – Most registration systems Lab performance (order to collect, collect to receive, receive to results) – Most hospital ordering systems Radiology performance (order to procedure start, proc start to proc end, proc end to results signed) – Most radiology systems © 2007, Jody Crane, MD, MBA
ED Intervals Door to triage Triage start to triage end Triage end to bed placement Bed placement to MD MD start to MD dispo MD dispo to discharge Door to Doc Interval In-ED Time © 2007, Jody Crane, MD, MBA
ED Intervals Most EDIS can give you this information © 2007, Jody Crane, MD, MBA
ED Intervals What if you don’t have an EDIS? You have the following options: ─Patient Tracking Infrared RFID Ultrasound Wideband ─Retrospective chart review Time consuming ? Accuracy (depends on written documentation) ─Sampling/Observation © 2007, Jody Crane, MD, MBA
Sampling There are multiple forms and methods of sampling in the ED Some require lengthy period of observation Some require some math and smarts In many cases, a mix can be used to develop the inputs © 2007, Jody Crane, MD, MBA
Time Observation Sheet © 2007, Jody Crane, MD, MBA
Patient Based Tracking Patient-completed tool Accompanied by timer Some minimal reward upon completion Starbuck’s card Money ($5) Most will cooperate, especially if you let them know it might help their future visits © 2007, Jody Crane, MD, MBA
Sampling for Intervals Take snapshots with the following information: ─Number of patients waiting to be triaged ─Number in triage ─Number triaged waiting to be seen ─Number in beds ─Number awaiting discharge © 2007, Jody Crane, MD, MBA
Sampling for Intervals Once you have taken multiple snapshots you will have sufficient data for the following: ─Take your overall LOS ─Take the average number of patients in each area ─Divide this number by the total in the department ─You will then have % distribution ─Multiply this by the overall LOS to arrive at the length of each interval © 2007, Jody Crane, MD, MBA
Sampling for Intervals - Example Your sampling results in the following averages: Overall LOS – 180 minutes ─# to be triaged 5 ─# in triage 1 ─# to be seen10 ─# in beds20 ─# awaiting discharge 5 © 2007, Jody Crane, MD, MBA
Sampling for Intervals - Example Your sampling results in the following averages: Overall LOS – 180 minutes ─# to be triaged 512% ─# in triage 1 4% ─# to be seen1024% ─# in beds2049% ─# awaiting discharge 512% Total41 © 2007, Jody Crane, MD, MBA
Sampling for Intervals - Example Your sampling results in the following averages: Overall LOS – 180 minutes ─# to be triaged 522min ─# in triage 1 4min ─# to be seen1044min ─# in beds2088min ─# awaiting discharge 522min Total41 180min © 2007, Jody Crane, MD, MBA
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 What can EDucate Tell You? Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Information on Daily ED Arrivals and Admissions Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Arrivals by Various Modalities (Triage, EMS, FT) Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
ED Size/Annual Visits per Bed Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour Bed Demand 3-hour Length Of Stay
1 hour Bed Demand 3-hour Length Of Stay Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour Bed Demand 3-hour Length Of Stay Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour Bed Demand 3-hour Length Of Stay Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour Bed Demand 3-hour Length Of Stay Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
1 hour Bed Demand 3-hour Length Of Stay Bed Demand Profile Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Bed Demand/Capacity Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Bed Demand/Capacity (Boarders)
RN and MD to Bed Ratios Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
RN and MD Staffing vs. Patient Demand Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
Graphical representation of Main ED and FT MD and RN demand relative to capacity by hour of day RN and MD Staffing vs. Patient Demand (Boarders)
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 RN and MD Staffing vs. Patient Demand (Boarders+Lunches)
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 RN to MD Ratios
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Admissions / Hospitalist Demand/Capacity Matching
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Inpatient Capacity and Admit Process Assessment
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 The Effect of Boarding on RN FTEs
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 The $$$ Related to LWOBS and Diversions
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 The Right Combo of Bed,RN,MD
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 Summary Page
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008
EDucate © Summary EDucate © can be an effective tool for ─Fast Track and Triage planning ─Matching demand to capacity throughout your process flow ─Determining the correct team size ─Determining the impact of holds, diversions, and walkouts on staffing, throughput, and revenue ─Creating “what-if” scenarios for future projects
Copyright Jody Crane, MD, MBA, Chuck Noon, PhD 2008 EDucate © Summary There are 4 main components that have to be aligned in order to achieve a highly functional ED that flows: Triage, MDs, RNs, and Beds (related to ancillary TAT) This tool will help identify where and when demand/capacity may be mismatched Garbage in/garbage out ─Try to get exact data ─If you don’t have exact data, that’s ok, but make your best estimates and understand the subsequent limitations