De-Identification
Download
Report
Transcript De-Identification
Emergency Department Load Estimation
Off line and on line load monitoring (and More)
Boaz Carmeli, IBM Haifa Research Laboratory & the Technion
[email protected]
[email protected]
®
Agenda
Emergency Department Crowding:
Consensus Development of Potential Measures
Measuring and Forecasting Emergency Department
Crowding in Real Time
Real Time ED Monitoring and Control System
Initial thought
The Rambam, Technion and IBM Open Collaborative Research
Project
(creative project name – anyone ??)
IBM Haifa Research Laboratory
2
Emergency Department Crowding:
Consensus Development of Potential Measures
®
Paper Summary
Authors
Leif I. Solberg, MDF; HealthPartners Medical Group and Clinics Minneapolis, MN;
Brent R. Asplin, MD, MPH Department of Emergency Medicine, Regions Hospital, St. Paul, MN
Robin M. Weinick, PhD; Agency for Healthcare Research and Quality, Rockville, MD;
David J. Magid, MD, MPH; Colorado Permanente Medical Group, Denver, CO;
What is already known on this topic
Although emergency department (ED) crowding is a topic of increasing public and professional
concern, there is no standardized definition of it and little agreement on what factors may
contribute to it
What question this study addressed
To use a broad-based and thorough expert process to identify all measures of ED and hospital
workflow that may be useful in understanding, monitoring, and managing crowding
What this study adds to our knowledge
A panel of 74 national experts assessed 113 measures, and chose 38 through a discussion and
rating process
How this might change clinical practice
The 38 measures should serve as a resource for research to determine which ones
are related to crowding, and eventually to develop tools to predict and modify
crowding
IBM Haifa Research Laboratory
4
ED Conceptual Model
Input
Throughput
Output
Patient Arrive
at ED
Ambulatory
care
system
Emergency Care
• Seriously ill and injured
patients from the community
• Referral of patients with
emergency conditions from
other providers
Ambulance
Diversion
Triage and room
placement
Unscheduled urgent care
• Desire for immediate care
• Lack of capacity for unscheduled
care in the ambulatory care
system
Safety net care
Demand for
ED Care
Diagnostic evaluation
and ED treatment
ED boarding
of inpatients
Leaves
without
treatment
complete
Patient
disposition
Transfer to
other
facility
Admit to
hospital
• Vulnerable populations (eg,
Medicaid beneficiaries, the
uninsured) care
• Access barriers (eg, financial,
transportation, insurance, lack of
usual source of care)
IBM Haifa Research Laboratory
5
ED Conceptual Model
The model is based on engineering principles from queuing theory
and compartmental models of flow, dividing ED function into input,
throughput, and output stages
The input-throughput-output model permits most factors affecting
use and crowding to be grouped into 1 of these 3 stages
Input or demand for ED services depends on the volume of ill and injured
people in the community and the capability of the rest of the health care system
to address the needs of individuals not requiring emergency care
Throughput includes factors that affect the efficiency of an ED to cope with its
input, ranging from ED beds and staffing to the efficiency of ancillary services
and consultant access
output factors include the ability of the inpatient system to admit patients
requiring hospital care and of the ambulatory care system to provide timely
post-discharge care
IBM Haifa Research Laboratory
6
The KPI* Selecting Process
Core investigators (Authors and additional 6 people)
in response to a request for task order proposals from the Agency for Healthcare
Research and Quality
Request to “select a group of content experts with expertise representing clinical
care, data, emergency medical services, ED staff, hospital administration,
information technology, and other relevant areas”
A group of expert reviewers with varied expertise and experience
Independent of the core investigator group
The final group of reviewers includes experts from 58 organizations in 21 states
The majority (72%) are emergency physicians
*KPI – Key Performance Indicators
IBM Haifa Research Laboratory
7
The KPI Selecting Criteria
1. Feasibility
How feasible would it be for operational staff to collect the data needed for this measure
routinely (or as frequently as would be needed) in the rater’s ED system or in one known to
the rater?
2. Early warning potential
How well would this measure provide warning about impending capacity problems within
the next 2 to 24 hours?
3. Planning value
How well would this measure provide information about trends and changes in ED business
and crowding throughout a period of weeks to months?
4. Cost-efficiency
How affordable would the data collection be for this measure?
5. Summary rating of operational usefulness
According to a combination of the above criteria, how useful would this measure
be for clinical and administrative operations?
6. Usefulness for research
Entirely apart from the aforementioned criteria, how much would this measure help
to improve our general understanding of the causes and consequences
of ED crowding?
IBM Haifa Research Laboratory
8
KPI Categories
To clarify their purpose, the KPI have been grouped within each stage
by the main concept they represent
1. Patient demand (6 items)
The volume of patients presenting to the ED for medical care
2. Patient complexity (3 items)
Patient factors such as the urgency and potential seriousness of the presenting
complaint, the stability of the clinical condition, and the baseline medical and
psychosocial burden of illness
3. ED capacity (6 items)
The ability of the ED to provide timely care for the level of patient demand
according to the adequacy of physical space, equipment, personnel, and the
organizational system.
IBM Haifa Research Laboratory
9
KPI Categories
(Cont.)
4. ED workload (6 items)
The demand and complexity of patient care that is undertaken by the ED within
a given period
5. ED efficiency (3 items)
The ability of the ED to provide timely, high-quality emergency care while
limiting waste of equipment, supplies, and effort
6. Hospital capacity (6 items)
The ability of the hospital to provide timely inpatient care for ED patients who
require hospitalization according to the adequacy of physical space, equipment,
personnel, and the organizational system
7. Hospital efficiency (8 items)
The ability of the hospital to provide timely, high-quality inpatient care
while limiting waste of equipment, supplies, and effort
IBM Haifa Research Laboratory
10
The Rating Process
This revised set of measures was then rated by 56 of the 64 reviewers and
the core investigators on an Internet Web site by using a magnitude
estimation technique.
This technique permits averaging of ratings across many raters on a ratio
level scale by asking each respondent to provide a relative score from 0 to
infinity for each item in comparison with a measure used as a standard.
The standard score was set at 100.
For example, if the feasibility of a measure was believed to be twice that of
the standard in the mind of a reviewer, a score of 200 would be assigned.
Likewise, if it were half as feasible, the reviewer would assign a score of
50.
Theoretically and empirically, the distribution of scores from a magnitude
likelihood task are log linear, and thus the geometric mean rather than the
more common arithmetic mean is the appropriate measure of central
tendency, which results in much less clustering of scores than often occurs
with rating scales using the more traditional Likert scale.
It also makes it easier to interpret the ratings because a rating of 200 for a
measure means that the reviewers as a group thought that the measure
was literally twice as good as one that was rated 100 in the same category.
IBM Haifa Research Laboratory
11
Input KPIs
Input Measure
Concept
Operational
Definition
1. ED patient volume,
standardized for bed hours
Patient
demand
Number of new patients registered within a defined
period (hour, shift, day) ÷ number of ED bed hours
within this period
2. ED patient volume,
standardized for annual
average
Patient
demand
Number of new patients registered within a defined
period ÷ annual mean number new patients
registered within this period
3. ED ambulance patient
volume, standardized for bed
hours
Patient
demand
Number of new ambulance patients registered within
a defined period ÷ number of ED bed hours within
this period
4. ED ambulance patient
volume, standardized for
annual average
Patient
demand
Number of new ambulance patients within a defined
period ÷ annual average of new ambulance patients
registered within this period
5. Patient source
Patient
demand
Time, arrival mode, reason, referral source, and usual
care for each patient registering at an ED in a defined
period (hour/shift/day)
* Leave without treatment complete includes those patients who leave without being seen,
leave before being finished, and leave against medical advice.
IBM Haifa Research Laboratory
12
Input KPIs
Input Measure
Concept
Operational
Definition
6. Percentage of open
appointments
Patient
demand
Percentage of open appointments at the beginning of
a day in ambulatory care clinics that serve an ED’s
patient population
7. Percentage of patients who
leave without treatment
complete*
ED capacity
Number of registered patients who leave the ED
without treatment complete ÷ total number of
patients who register during this period
8. Leave without treatment
complete severity*
ED capacity
Average severity of patients who leave the ED without
treatment complete within a defined period
(shift/day/week)
9. Ambulance diversion
episodes
ED capacity
Number and duration of all diversion episodes at EDs
within the EMS system within a defined period
(week/month/year)
10. Ambulance diversion
requests denied and forced
openings
ED capacity
Number of diversion requests denied or forced
openings within a defined period (week/month/year)
* Leave without treatment complete includes those patients who leave without being seen,
leave before being finished, and leave against medical advice.
IBM Haifa Research Laboratory
13
Input KPIs
Input Measure
Concept
Operational
Definition
11. Diverted ambulance
patient description
ED capacity
Chief complaints and final destination of diverted
EMS patients within a defined period
(week/month/year)
12. Average EMS waiting time
ED efficiency
Total time at hospital for ambulances delivering
patients to ED during a defined period
(shift/day/week/month) ÷ number of ambulance
deliveries within that period
13. Patient complexity as
assessed at triage
Patient
complexity
Mean complexity level as assessed at triage (using
local criteria) for all
14. Patient complexity as the
percentage of ambulance
patients
Patient
complexity
Percentage of patients registering at an ED in a
defined period (shift/day/week/month) who
arrived by ambulance
15. Patient complexity as
assessed by coding
Patient
complexity
Mean complexity level as coded at the end of the
visit for all patients completed in a defined period
(shift/day/week/month)
* Leave without treatment complete includes those patients who leave without being seen,
leave before being finished, and leave against medical advice.
IBM Haifa Research Laboratory
14
Throughput KPIs
Throughput
Measure
Concept
Operational
Definition
1. ED throughput
time
ED efficiency
Average time between patient sign-in and departure
(separately for admitted vs discharged patients) within a
defined period (day/week/month)
2. ED bed placement
time
ED efficiency
Mean interval between patient sign-in and placement in a
treatment area within a defined period
(shift/day/week/month)
3. ED ancillary
service turnaround
time
ED efficiency
Average time between physician order and result report
(separately for each service area) within a defined period
(shift/day/week/month)
4. Summary
workload,
standardized for ED
bed hours
ED workload
Summary of (patients treated ׳acuity) in a defined period
(shift/day/week) ÷ number of ED bed hours within this
period
5. Summary
workload,
standardized for
registered nurse
staff hours
ED workload
Summary of (patients treated ׳acuity) in a defined
period(shift/day/week) קtotal ED staff registered nurse
hours within this period
IBM Haifa Research Laboratory
15
Throughput KPIs
Throughput
Measure
Concept
Operational
Definition
6. Summary
workload,
standardized for
physician staff hours
ED workload
Summary of (patients treated x acuity) in a defined period
(shift/day/week) ÷ total ED staff physician hours within
this period
7. ED occupancy rate
ED workload
Total number of ED patients registered at a defined time ÷
number of staffed treatment areas at that time
8. ED occupancy
ED workload
Total number of patients present in the ED at a defined
time ÷ number of staffed treatment areas at that time
9. Patient disposition
to physician staffing
ratio
ED workload
Number of patients admitted or discharged per staff
physician during a defined period (shift/day/week)
IBM Haifa Research Laboratory
16
Output KPIs
Output Measure
Concept
Operational
Definition
1. ED boarding time
Hospital
efficiency
Mean time from inpatient bed request to physical departure
of patients from the ED overall and by bed type within a
defined period (shift/day/week)*
2. ED boarding time
components
Hospital
efficiency
Mean time from inpatient bed request to physical departure
of patients from the ED by bed type by component (bed
assignment, bed cleaning, transfer arrival) within a defined
period*
3. Boarding burden
Hospital
efficiency
Mean number of ED patients waiting for an inpatient bed
within a defined period ÷ number of staffed ED treatment
areas
4. Hospital admission
source, standardized
Hospital
efficiency
Number of requests for admission within a defined period
(shift/day) overall and by admission source ÷ annual mean
requests for admission during that period and adjusted for
day of week and season of year†
5. ED admission
transfer rate
Hospital
efficiency
Number of patients transferred from ED to another facility
who would normally have been admitted within a defined
period ÷ number of ED admissions within this period
*Bed type=ICU/telemetry/psychiatry/ward.
†Admission source=ED/operating room/catheterization laboratory/outpatient/other.
IBM Haifa Research Laboratory
17
Output KPIs
Output Measure
Concept
Operational
Definition
6. Hospital discharge
potential
Hospital
efficiency
Number of inpatients ready for discharge at or within a
defined period ÷ number of hospital inpatients at that time
7. Hospital discharge
process interval
Hospital
efficiency
Mean interval from discharge order to patient departure
from a unit in a defined period (shift/day/week/month)
8. Inpatient cycling
time
Hospital
efficiency
Mean amount of time required to discharge an inpatient
and admit a new patient to the same bed within this period
9. Hospital census
Hospital capacity
Mean number of inpatient beds available by bed type at a
defined time ÷ number of staffed inpatient beds by bed
type*
10. Hospital
occupancy rate
Hospital capacity
Number of occupied inpatient beds overall and by bed type
÷ number of staffed inpatient beds overall and by bed
type*
*Bed type=ICU/telemetry/psychiatry/ward.
†Admission source=ED/operating room/catheterization laboratory/outpatient/other.
IBM Haifa Research Laboratory
18
Output KPIs
Output Measure
Concept
Operational
Definition
11. Hospital
supply/demand
status forecast
Hospital capacity
Forecast of expected hospital admissions and discharges as
reported daily at 6 AM and compared with hospital census
12. Observation unit
census
Hospital capacity
Mean number of available ED observation beds at a defined
time ÷ number of staffed ED observation beds
13. ED
volume/hospital
capacity ratio
Hospital capacity
Number of new ED patients within a defined period
(shift/day) ÷ number of available hospital beds at the
beginning of analysis period overall and by bed type*
14. Agency nursing
expenditures
Hospital capacity
Registered nurse agency nursing expenditures (ED/overall)
within a defined period ÷ total nursing expenditures
(ED/overall) within this period
*Bed type=ICU/telemetry/psychiatry/ward.
†Admission source=ED/operating room/catheterization laboratory/outpatient/other.
IBM Haifa Research Laboratory
19
Measuring and Forecasting Emergency Department
Crowding in Real Time
®
Paper Context
Authors – Vanderbilt University Medical Center, Nashville, TN
Nathan R. Hoot, MS; Department of Biomedical Informatics
Chuan Zhou, PhD; Department of Biostatistics
Ian Jones, MD; Department of Emergency Medicine & Biomedical Informatics
Dominik Aronsky, MD, PhD; Department of Biomedical Informatics & Biomedical Informatics
What is already known on this topic
In the absence of an accepted definition of emergency department (ED) crowding, multiple
scores have been proposed to measure this phenomenon
What question this study addressed
How 5 metrics for measuring current and impending ED crowding fared in predicting
ambulance diversion status during an 8-week period in a single adult ED
What this study adds to our knowledge
All measures performed reasonably well at measuring crowding in real time, but none
outperformed the simplest measure, ED occupancy level. None of the measures was
particularly useful as a short-term warning system for future crowding
How this might change clinical practice
This study will not change clinical practice but suggests that ED occupancy,
the simplest metric for measuring ED crowding, performs just as well as
more complex methods
IBM Haifa Research Laboratory
21
Paper Summary
Study objective:
To quantifying the potential for monitoring current and near-future emergency
department (ED) crowding by using 4 measures:
the Emergency Department Work Index (EDWIN)
the National Emergency Department Overcrowding Scale (NEDOCS)
the Demand Value of the Real-time Emergency Analysis of Demand
Indicators (READI)
the Work Score
Methods:
Study calculated the 4 measures at 10-minute intervals during an 8-week study
period (2006)
Ambulance diversion status was the outcome variable for crowding
Occupancy level was the performance baseline measure
Evaluation of discriminatory power for current crowding was calculated by the
area under the receiver operating characteristic curve (AUC)
To assess forecasting power, activity monitoring operating characteristic curves
was applied, which measure the timeliness of early warnings at various false
alarm rates
IBM Haifa Research Laboratory
22
Paper Summary
(Cont.)
Results:
7,948 observations were recorded during the study period.
The ED was on ambulance diversion during 30% of the observations
The AUC (Area Under Curve) was:
0.81 for the EDWIN
0.88 for the NEDOCS
0.65 for the READI Demand Value
0.90 for the Work Score
0.90 for occupancy level
In the activity monitoring operating characteristic analysis, only the occupancy level
provided more than an hour of advance warning (median 1 hour 7 minutes) before
crowding, with 1 false alarm per week
Conclusion:
The EDWIN, the NEDOCS, and the Work Score monitor current ED crowding with high
discriminatory power
None of them exceeded the performance of occupancy level across the range
of operating points
None of the measures provided substantial advance warning before crowding
at low rates of false alarms
IBM Haifa Research Laboratory
23
Emergency Department Work Index
EDWIN was calculated by
Offered load ??
EDWIM = Σniti / (Na X (Bt – Pboard))
Number of physicians
where
ni – number of non-boarding patients in triage category i
ti – reversed triage category i, where
5 denotes the sickest patients and
Spare treatment cycles
1 denotes the least sick patients
Na – number of attending physicians on duty
Bt – number of licensed treatment beds in the ED
Pboard – number of boarding patients
IBM Haifa Research Laboratory
24
National Emergency Department OverCrowding Scale
NEDOCS was calculated by
Patient Indexnumber of patients
within ED
NEDOCS = (Pbed ⁄ Bt) X 85.8 + (Padmit ⁄ Bh) X 600 + (Wtime X 5.64)
+ (Atime X 0.93) + (Rn X 13.4) – 20
Admitted Index –
number of patients
within the hospital
Where
Pbed – number of patients in licensed beds and overflow locations, such as
hallway beds or chairs
Registration Time – time
from registration to triage
Bt – number of licensed treatment beds
Padmit – number of admitted patients
for the last patient
Bh – number of hospital beds
Wtime – waiting time for the last patient put into bed
Atime – longest time since registration among boarding patients
Rn – number of respirators in use, maximum of 2*
Admission Time –
the longest time an
admitted patient is
staying at the ED
Number of Respirators
– an indication for
(non-linear)
* The respirator variable (Rn) did not generalize to the study setting, because patientsadditional
ill enough to require
mechanical ventilation are stabilized and transferred immediately to a critical care unit. As a surrogate,
loadthe
number of trauma beds was used in place of the number of respirators.
IBM Haifa Research Laboratory
25
NEDOCS Nomogram
IBM Haifa Research Laboratory
26
Real-time Emergency Analysis of Demand Indicators
The Demand Value of the Real-time Emergency Analysis of Demand
The demand for care –
Indicators (READI) was calculated by
DV = (BR + PR) X AR
BR = (Ptotal + Apred – Dpred) ⁄ Bt
AR = Σniti ⁄ Ptriage
PR = Ahour ⁄ΣPPH
Where
patients, care givers
and acuity
Patient Index – number of
(expected) patients within
the hospital
Provider Ratio – the number of patients a
provider is treated in an hour
DV – Demand Value, BR – bed ratio, PR – provider ratio, AR – acuity ratio;
Ptotal – number of ED patients, Apred – number of predicted arrivals,
Dpred – number of predicted departures, Bt – number of licensed treatment beds;
ni – number of patients in triage category i, ti – reversed triage category i,
Ptriage – number of patients in the ED with an assigned triage category,
Ahour – number of arrivals in the past hour,
PPH – average patients seen per hour for each attending physician
and resident on duty.
IBM Haifa Research Laboratory
27
READI Calculation – Additional Info
The predicted number of arrivals (Apred) and departures (Dpred) for
each hour of the day was calculated using 9 months of ED data
The original READI instrument used a 4-level triage system, so the
5-level Emergency Severity Index was condensed into 4 categories
by combining the 2 least severe acuity levels.
The number of patients treated per hour was calculated for
residents at each level of training and for attending physicians who
treated patients without a resident, using 9 months of ED data
IBM Haifa Research Laboratory
28
Work Score
The Work Score was calculated using the following formula:
Work Score = 3.23 X Pwait ⁄ Bt + 0.097 Σniti ⁄ Nn +
10.92 X Pboard ⁄ Bt
where
Pwait – number of waiting patients
Bt – number of licensed treatment beds
ni – number of patients under evaluation in triage category I
ti – triage category I
Nn – number of nurses on duty
Pboard – number of boarding patients
IBM Haifa Research Laboratory
29
ED Occupancy Level
The ED occupancy level was used as a control measure for baseline
comparison
The occupancy level was calculated using the following formula:
Occupancy level = 100 X Pbed ⁄ Bt
Where
Pbed – number of patients in licensed beds and overflow locations such as
hallway beds or chairs
Bt - number of licensed treatment beds
IBM Haifa Research Laboratory
30
Calculation Results
Time series plots of the crowding measures during the study period
The plots shown here are smoothed using cubic splines
Episodes of ambulance diversion are marked by the shaded areas.
IBM Haifa Research Laboratory
31
Receiver Operating Characteristic Curves
IBM Haifa Research Laboratory
32
Activity Monitoring Operating Characteristic Curves
IBM Haifa Research Laboratory
33
Further Research
Future research should focus on improving the forecasting power of
crowding measures
The use of historical data to predict changes in the next few hours may
allow for substantial improvements in the performance of an early warning
system
Advanced modeling techniques such as neural networks, applied specifically
for the purpose of forecasting, may result in improved forecasting power
The development of a good forecasting model for ED crowding will pave
the way to studying intervention policies, which may allow researchers to
identify ways of sustaining health care quality and access in the face of
crowding
Other researchers have discussed strategies including the use of reserve
physicians and nurses and deferring care of low-acuity patients
either of which could be initiated, given a few hours of advance
warning before crowding
IBM Haifa Research Laboratory
34
Paper Summary
The findings demonstrate the feasibility of implementing 4
measures for real-time monitoring of ED crowding
Occupancy level showed discriminatory power similar to or greater
than the 4 other measures for measuring current ED crowding
In terms of timely forecasting, none of the measures showed a
clear advantage over occupancy level
These findings suggest new directions for the measurement and
management of ED crowding.
IBM Haifa Research Laboratory
35
Real Time ED Monitoring and Control System
Work in (its very early stages but in) progress
®
Real Time ED Monitoring and Control System
Improving the Forecasting Power of Crowding Measures
Data Collection
Adding RFID based location tracking system for Physicians, Nurses, Patients and
other relevant personnel
Collect real-time relevant information from hospital IT systems such as PACS,
EHR, ADT, LAB etc
Better utilize historical EHR and operational data from existing IT systems
within the hospital
Data Visualization
Operational dashboard
Provide sophisticated data
Analysis Techniques
Machine learning - neural networks
Mathematical models – service engineering
Other ??
IBM Haifa Research Laboratory
37
System Architecture
Data Collection
Analysis
Data Visualization
ED Simulator
• Based on observation
• Will be used, mainly, for
design phase e.g. to mimic the
RFID system
RFID based Location
Tracking
• Low level location tracking for
patients and care personnel
• Technology dependent
capabilities
Hospital IT systems
Real Time
Event Processing
Network
Rule Based Analysis
Machine Learning
Algorithms
Analysis of Historical
And Real-time Data
Mathematical Models
e.g. Queuing Theory
• Admit, Discharge, Transfer
• Electronic Health Records
• Lab request/results
• Picture Archive and
Communication System (PACS)
IBM Haifa Research Laboratory
38
OCR project – Next Generation Hospital
Rambam/Technion/IBM open collaborative research
®
What is OCR?
An IBM funded program to support open collaborative research between IBM and universities
in Computer Science (including related disciplines in Electrical Engineering and Math) and its
applications
Implements the Open Collaboration Principles established under IBM’s leadership in 2005 - IP openly
published or available in royalty free “public commons”, software available as open source
Choose a limited number of strategically defined topics where open collaborative innovation would benefit IBM
and the world at large – endorsed by Research area strategists and VP strategists
Subject to approval in advance by the OSSC
Piloted in US in 2006
Research topics and universities: Software Quality (Rutgers, UC Berkeley), Privacy & Security Policy
Management (CMU, Purdue), Clinical Decision Support (Columbia, Georgia Tech), Mathematical Optimization
(CMU, UC Davis)
Joint announcement and publicity with universities 12/14/2006
Expanded in 2007, including outside US
Research topics and universities: Accessibility for an Aging Population (Dundee, Miami), NewGen Hospital
(Technion, Rambam Hospital), Service Professionals’ Social Network (Indian School of Business), Privacy &
Security Policy Management (Imperial College)
What makes it work?
Multi-year, so that faculty can take on new students and obligations
Collaborative, allowing IBM and university participants to forge deep relationships
Open, providing maximum opportunity for others to build on the results
Challenging, research requiring considerable innovation
Well-funded, large enough to make a difference (average $150K)
IBM Haifa Research Laboratory
40
OCR Overview
Joint research project between Rambam hospital, Technion, and IBM
Leverage Technion’s relationship with Rambam hospital
Goal: Combined multi-dimensional improvement of patient care process
Clinical
Operational
Financial
Multi-disciplinary approach:
Medical (Rambam)
Statistics (IBM, Technion)
Operations Research (IBM, Technion)
Healthcare informatics (IBM, Rambam)
Process improvement (IBM, Technion)
Human factors engineering (Technion)
Financial (Rambam)
Domain specific knowledge in above areas – IBM & Technion
Participation
Rambam hospital: Top management including hospital general manager, Prof. Rafi Bayar,
and ER manager, Dr. Dagan Schwartz
Technion: Prof. Avishai Mandelbaum, Prof. Danny Gopher, Prof. Avi Shtub, Prof. Eitan Naveh
IBM: Pnina Vortman, Segev Wasserkrug, Boaz Carmeli, Ohad Greenshpan and Sergey Zeltyn
IBM Haifa Research Laboratory
41
Processes at the Healthcare Domain
Business & Operational
processes
financial, HR, assets aspects and
indicators)
Clinical processes
(Mostly health aspects and
indicators)
Organization
centric
Interested role: Management
Interested role: Care Personnel
Example: Procurement, training
Example: Protocols and procedures
Patient centric
Interested role: Patient &
Management
Interested role: Patient & Care
Personnel
Example: Obtaining
reimbursement for medical
procedures
Example: Arrival and treatment at
ER
IBM Haifa Research Laboratory
42
OCR Approach and Status
Approach
Pick four high intensity department
ER
Operating room
Neonatal
Trauma
Map patient centric processes from various dimensions
Focus and implement specific research projects based on initial analysis
HRL work mode
Hands on (joint work)
Mentoring student projects at the Technion
Collaborating work carried out by Technion graduate and undergraduate
students
IBM Haifa Research Laboratory
43
Current status
Work carried out In ER
Some initial discussions with second department (OR)
Metrics: Detailed Metrics document was developed collaboratively
between the 3 parties
Longitudinal observations: Around 200 observations were taken
Horizontal observations (Work sampling)
Compared to 2001 observations (later to 2004-2007/8)
Data analysis and improvement ideas (Lean Manufacturing in Healthcare)
Observations were used to develop the following analytical views:
Process Maps (Activities, Resources, Information)
Demo of the Monitoring Console – Command & Control
Forecasting Model - forecast arrival flow to the ED, based on short term
historical data
Online Statistics – integrating Technion’s SEE-Stat tool, which takes an
Operations Research viewpoint
IBM Haifa Research Laboratory
44
Current status
(Cont.)
Layout planning
Simulation-based analysis of the temporary ER, continuing into Future ER
Layout of new trauma unit
Flow from ER to wards
Projects on Fairly moving patients from the ED to the Wards (Fairness Table)
RFID: Highly accurate measurement of actual patient flow using
RTLS is being discussed
Pilots with three companies
Additional options being considered
Focused research topics:
Measure, forecast, predict, and optimize intraday performance
Combination of Forecasting, OR models, and Cognos
ED vs. ER
Cognos based measurement, forecasting and improvement pilot
being planned
IBM Haifa Research Laboratory
45
Intraday measurement, forecast and optimization
Process:
Measure multi dimensional
metrics
Forecast near term future
Enable optimization and decision
making using advanced analytics
Based on:
Cognos BI platform
ER patient process simulator
created by Technion
Jointly created load forecasting
models
Jointly created optimization
algorithms
IBM Haifa Research Laboratory
46
Thank You
®