The Joint Commission Journal on Quality and Patient Safety 2017; 43:275–283

PERFORMANCE IMPROVEMENT

Using Lean to Rapidly and Sustainably Transform a

Behavioral Health Crisis Program: Impact on Throughput
and Safety
Margaret E. Balfour, MD, PhD; Kathleen Tanner, MA, LSSBB; Paul J. Jurica, PhD; Dawn Llewellyn, BA;
Robert G. Williamson, MD; Chris A. Carson, MD, MBA

Background: Lean has been increasingly applied in health care to reduce waste and improve quality, particularly in fast-
paced and high-acuity clinical settings such as emergency departments. In addition, Lean’s focus on engagement of frontline
staff in problem solving can be a catalyst for organizational change. In this study, ConnectionsAZ demonstrates how they
applied Lean principles to rapidly and sustainably transform clinical operations in a behavioral health crisis facility.
Methods: A multidisciplinary team of management and frontline staff defined values-based outcome measures, mapped
the current and ideal processes, and developed new processes to achieve the ideal. Phase I was implemented within three
months of assuming management of the facility and involved a redesign of flow, space utilization, and clinical protocols.
Phase II was implemented three months later and improved the provider staffing model. Organizational changes such as
the development of shift leads and daily huddles were implemented to sustain change and create an environment support-
ive of future improvements.
Results: Post-Phase I, there were significant decreases (pre vs. post and one-year post) in median door-to-door dwell time
(343 min vs. 118 and 99), calls to security for behavioral emergencies (13.5 per month vs. 4.3 and 4.8), and staff injuries
(3.3 per month vs. 1.2 and 1.2). Post-Phase II, there were decreases in median door-to-doctor time (8.2 hours vs. 1.6 and
1.4) and hours on diversion (90% vs. 17% and 34%).
Conclusions: Lean methods can positively affect safety and throughput and are complementary to patient-centered clin-
ical goals in a behavioral health setting.

L ean is an organizational philosophy developed to trans-

late the successes of the Toyota Production System to auto
manufacturing in the United States and has since been
which they work, with support from and collaboration with
leadership.8 In Lean organizations, management functions
to support the staff in problem solving, and shifts from asking
adapted to a wide variety of industries.1 Lean has been in- “Why didn’t staff do their job?” to “Why couldn’t staff do
creasingly applied to health care settings to achieve quality their job?”9 Although Lean employs many tools for data anal-
goals.2,3 An important focus of Lean is the reduction of waste, ysis and outcome measurement, it differs from traditional
which is defined as anything that is “non-value added” to research methods in that the focus is on continuous im-
the customer, such as time spent waiting.4 This is naturally provement rather than proving a hypothesis.10 Improvement
appealing to fast-paced health care settings, and thus many efforts are rapid and iterative, and methods change quickly
implementations of Lean methods have been in emergency and often as outcomes are continually monitored and new
departments (EDs) and operating rooms.5 problems are identified and addressed.
In addition to the impact on quality outcomes, Lean can The purpose of the study reported in this article is to (1)
be a catalyst for broader organizational culture change in demonstrate how Lean principles can be applied to achieve
health care organizations.6 Fundamental principles of Lean rapid transformation of clinical operations, (2) describe strat-
include the challenge of continuous improvement and respect egies for sustaining change and promoting ongoing
for the teams of people performing the work.7 This leads to improvements, and (3) identify special considerations for be-
an approach different from that found in traditional top- havioral health settings.
down management structures; rather, frontline staff are
empowered to engage in improvement of the processes in METHODS
Study Setting and Population
1553-7250/$-see front matter This work was performed at the Crisis Response Center
© 2017 The Author(s). Published by Elsevier Inc. on behalf of The Joint (CRC), a freestanding behavioral health facility providing
Commission. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
crisis services and emergency psychiatric care to adults and
http://dx.doi.org/10.1016/j.jcjq.2017.03.008 children in Pima County, Arizona. The CRC was created
276 Margaret E. Balfour, MD, PhD, et al Using Lean to Rapidly Transform a Behavioral Health Crisis Program

in 2011 to reduce the numbers of people with behavioral statement: “To meet the immediate needs of people in be-
health needs booked into jail or boarding in hospital EDs. havioral health crisis in a safe and supportive environment
Management of the CRC was transferred to our organiza- in collaboration with community partners.” When there was
tion (ConnectionsAZ) in April 2014 to address concerns unanimity among management, frontline staff could be
related to long waits, poor patient experience, and repeat- engaged in improvement efforts with consistent support and
ed safety incidents. The CRC serves 11,000 adults and 2,200 guidance from their managers.
children annually. This study includes the adult popula-
tion served in the urgent care clinic and 23-hour observation Defining Values-Based Outcomes. As described
unit. Approximately 45% are brought directly from the field elsewhere,13 we used a quality improvement tool called a
by law enforcement, 10% are transferred from outside medical Critical-to-Quality (CTQ) Tree14 to translate our core values
EDs, and the remainder arrive via walk-in. The CRC is staffed into outcome measures: Crisis Reliability Indicators Sup-
24/7 by behavioral health medical professionals (BHMPs: porting Emergency Services (CRISES). We defined our values
psychiatrists, nurse practitioners, physician assistants), nurses, as Timely, Safe, Accessible, Least Restrictive, Effective,
behavioral health technicians, crisis workers (social services Consumer/Family Centered, and Partnership. These are con-
staff with various credentials), and peer supports. (Peer sup- sistent with the Institute of Medicine’s six aims for quality
ports are staff with their own lived experience with mental health care: Safety, Effectiveness, Equity, Timeliness, Patient-
illness and/or substance use who have received training in Centeredness, and Efficiency,15 while also focusing attention
how to use that experience to engage with patients in a on goals unique to the behavioral health crisis setting. This
manner different than that of traditional clinical staff.11) The work informed the selection of the primary outcome mea-
CRC is located on the Banner—University Medical Center sures for the process improvement initiatives described in
campus in Tucson and is a training site for residents and the current study, specifically the following:
medical students at the University of Arizona. The Region- • Timely: door-to-door dwell times, door to assessment
al Behavioral Health Authority is the major source of funding by a BHMP (door-to-doctor time)
supporting the operations of the CRC. • Safe: staff injuries
• Least Restrictive: calls to security to assist with behav-
Study Design ioral emergencies
This retrospective, observational, pre-post interventional study • Partnership: time on diversion
compared outcome measures related to throughput and safety
before and after the implementation of process improve- Interventions
ments. Our organization assumed management of the facility Phase I Intervention: Reengineering the Triage
and its staff on April 1, 2014. Phase I interventions were Process (July–December 2014). A multidisciplinary
implemented on July 1, 2014, and Phase II interventions group composed of executive leadership, management, and
were implemented on October 1, 2014. We included all adults frontline staff (including peer supports who had previously
presenting for services during the study period of January been patients at the CRC) mapped the baseline triage process,
1, 2014, to December 31, 2015. The 2014 data were divided which was as follows: Each walk-in patient arrived via an
into pre- and postintervention intervals and compared. We unlocked waiting room (WR) where he or she was checked
then compared the preintervention interval to the follow- in and received a brief medical screen by a behavioral health
ing year (January 1, 2015, to December 31, 2015) to assess technician. The patient was then called into 1 of 12 clinic
whether the improvements had been sustained.12 We ana- assessment rooms where a behavioral health technician per-
lyzed administrative data from our electronic health record formed a safety search and inventoried the patient’s property,
(EHR) but not individual patient records. As a quality im- which was locked up for the duration of the visit. Then the
provement effort, this project was exempt from Institutional patient met with a crisis worker for a screening assessment
Review Board review and did not require informed consent. and waited in that room for further evaluation. Another crisis
worker performed a more extensive assessment, then dis-
Preintervention Activities (April–June 2014) cussed the case with the BHMP, who may either perform
Engagement, Information Gathering, and Values his or her own assessment, direct the crisis worker to dis-
Development. Executive leadership began by engaging charge the individual, order medication or a period of
managers, frontline staff, and patients via town-hall meet- observation then have the individual wait for reassessment
ings, rounding, and working clinical shifts to gain firsthand in a clinic assessment room, or write admission orders to the
knowledge of operations and stakeholder concerns. All agreed 23-hour observation unit. The observation unit is adjacent
that the adult triage process was the highest priority, but there to the clinic and is comprised of an open area where pa-
was disagreement on the desired outcomes of that process, tients are visualized at all times and, unlike the clinic, meets
stemming from a lack of consensus regarding the mission inpatient safety standards for anti-ligature design. Police and
and values of the CRC. Leadership convened the manage- ambulance arrivals entered through a gated sally port (SP).
ment staff to define values and develop a new mission The process for these patients was the same except they were
Volume 43, No. 6, June 2017 277

assessed for medical issues by a nurse in the SP then taken there is a compelling reason to do so for a given individu-
directly to an assessment room in the clinic. al. Patient flow is tracked using a visual management tool
The team identified opportunities for improvement in dwell (Appendix 2, available in online article) comprised of a white-
times and staff injuries as well as processes that put pa- board with magnets and colors to indicate patient status. The
tients and the organization at risk. For example, it was not process is the same for SP arrivals except that the nurse meets
clear which patients required assessment by the BHMP vs. individuals in the SP, directs them to one of three desig-
social services staff only. Individuals with acute symptom- nated assessment rooms, and determines the risk level based
atology were often held for hours or even overnight in clinic on the same triage tool described above. A value stream map16
rooms that were not ligature-safe or amenable to constant illustrating the old and new processes is shown in Appendix 3
visualization due to inconsistent criteria for admission to the (Appendix 3, available in online article).
observation unit or delays in finding a BHMP to write ad-
Phase II Intervention: Addition of a Behavioral
mission orders. Staff were spread out over a large area and
Health Medical Provider in Triage (October–
often unable to proactively attend to the needs of acutely
December 2014). After the Phase I interventions, there
psychotic or intoxicated patients and prevent episodes of ag-
continued to be long waits to BHMP assessment, and the
itation or violence. Consequently, security was often called
facility was frequently on diversion. To address this, we added
to assist. Other processes led to suboptimal patient experi-
an additional 12-hour BHMP shift assigned to the clinic be-
ence, such as seclusion/restraint of high-acuity patients in
ginning October 1, 2014. The clinic BHMP focuses on newly
close proximity to low-acuity patients and their families who
arrived patients, whether discharged from the clinic or ad-
were seeking outpatient urgent care services.
mitted to the observation unit. He or she typically starts the
The team developed the following goals for the ideal
day assisting the observation unit with re-assessments and
process:
discharges, then focuses on the clinic when walk-in and SP
1. Treat patients in the least restrictive setting that can
patients begin arriving later in the day.
safely meet their needs.
2. Move the highest-risk patients to the safest location Sustainability Interventions: Building a Lean
(observation unit). Culture
3. Begin treatment as quickly as possible.
Some staff were initially skeptical of improvement efforts
4. Reduce unnecessary or redundant tasks.
because of a belief that things would not change. Others were
5. Use space more efficiently.
fearful of being punished for breaking the rules or criticiz-
6. Create the experience we would want for our fami-
ing their superiors. Leadership made considerable efforts to
lies or ourselves.
change the culture to one of staff feeling supported in values-
The team performed a gap analysis comparing the current
based decision making and problem solving in the moment.
process to the ideal, then developed new policies and pro-
This was primarily achieved via frequent contact and mod-
cedures for clinical assessments, patient flow, and space
eling (for example, regular rounding, open-door accessibility,
utilization.
working clinical shifts on the floor, and inclusion of front-
The new process was implemented July 1, 2014, and is
line staff in improvement efforts). In addition, we hardwired
as follows: WR arrivals are checked in and receive a brief
Lean concepts into our organizational structure in order to
medical screen as before. In addition, vital signs are per-
sustain improvements and continue developing the culture.
formed in a designated area in the WR to facilitate early
identification of acute medical issues. The WR is moni- Daily Huddles. We implemented daily huddles17 with key
tored by a behavioral health technician stationed there at all operational leaders, in which we ask “What do we need to
times, and peer supports have an increased presence. Pa- do to support the frontline staff today?” The huddles were
tients are brought to one of two triage assessment rooms (TR) implemented shortly after assuming management of the fa-
where the crisis worker performs a single streamlined as- cility in April 2014 and refined throughout the study period.
sessment using a newly developed tool to triage patients into The huddle centers around the shift report (Appendix 4, avail-
low/moderate/high risk categories based on dangerousness able in online article), which is prepared by frontline staff
to self/others and symptom acuity (Appendix 1, available in twice daily and contains key pieces of actionable data. For
online article). High-risk patients are automatically admit- example, multiple patients waiting for their initial psychi-
ted to the observation unit via a standing order protocol, atric evaluation may indicate a need to call in additional
eliminating the need to wait for BHMP orders. A nurse and BHMPs, whereas high numbers of patients waiting for trans-
behavioral health technician are called to the assessment room fer to external inpatient facilities may indicate a need for the
to begin the admission process and move patients to the ob- medical director to review the cases to ensure that they all
servation unit. Low- and moderate-risk patients are redirected do in fact meet medical necessity criteria for inpatient ad-
to the WR and called back into the TR assessment room mission and a need for leadership to work with external
to meet with the BHMP or crisis worker as needed. They stakeholders to address the backup. The huddle also high-
are not searched or separated from their belongings unless lights individual patients who may need specialized
278 Margaret E. Balfour, MD, PhD, et al Using Lean to Rapidly Transform a Behavioral Health Crisis Program

intervention. For example, data indicated that patients with normally distributed data (door-to-door dwell times and door-
developmental disabilities were more likely to be restrained; to-doctor times); t-tests were used to compare all other
now these individuals are flagged on the shift report, which measures. In addition, statistical process control charts were
triggers a review of the behavioral plan by the director of used to illustrate changes in throughput measures over time.
nursing. Some activities that once required a separate meeting
(for example, review of safety events) have been incorpo-
RESULTS
rated into the daily huddle, which both ensures that these
events are addressed as they occur and reduces time wasted Population and Encounters
on unnecessary meetings. We analyzed 10,546 encounters from January 1, 2014,
through December 31, 2014, and 10,353 encounters in the
Shift Leads. We made the transition to a structure of
postimplementation year of January 1, 2015, through De-
discipline-specific leads for each 12-hour shift, analogous to
cember 31, 2015. Demographic descriptors and monthly
Lean line managers,18 which includes a charge nurse, lead
volumes did not significantly differ across the study periods.
crisis worker, lead behavioral health technician, and lead unit
Some 61% of the population was male, and 20% were 18–
clerk. This transition occurred in the fourth quarter (Q4)
24 years of age; 35%, 26–40; 42%, 41–64; 3%, 65–84; and
of 2014 through Q1 2015 and replaced the previous struc-
0.1%, 85 years of age or older. Some 53.2% were white-
ture in which a single house supervisor was responsible for
non-Hispanic, 26.8% Hispanic, 4.8% African American,
the day-to-day operations on each shift, and management
4.5% Native American, 1.4% biracial, 0.6% Asian, and the
staff, often via retrospective chart review or monthly groups,
remainder classified as “other” or declined to answer.
performed clinical supervision. The shift leads are now em-
powered to solve problems in the moment affecting their
specific discipline’s responsibilities. More complex prob- Interventions
lems are addressed in the next daily huddle (or directed to Phase I Interventions (July–December 2014). Phase
the administrator on call if urgent). Clinical supervision is I outcomes are summarized in Table 1. There was a de-
performed by the shift leads, as they can identify opportu- crease of 225 minutes in the median door-to-door dwell time
nities for improvement and provide correction and feedback in the clinic (95% confidence interval [CI]: −224– −208;
in real time. This frees managers to use their time more ef- p < 0.0001). The change over time is depicted as a control
ficiently, focusing only on individual staff needing more chart in Figure 1a. There was a decrease of 2 hours in the
intensive intervention, so that they are free to engage in more median door-to-door dwell time in the observation unit (CI:
strategic planning and complex improvement activities. A −3.7– −2.0; p < 0.0001) despite the fact that 232 more pa-
modified Lean curriculum was designed for the shift leads tients per month were identified as high-risk and triaged to
to give them the tools to lead future improvement projects. that unit (CI: 163–299; p < 0.0001). The percentage of pa-
tients evaluated by a BHMP (as opposed by being seen by
Outcome Analysis social services staff only) increased by 21 percentage points
Data Extraction. Existing data reports had been de- (CI: 19–23; p < 0.0001). The mean number of emergent se-
stroyed immediately prior to our assuming management of curity calls per month decreased by 9.2 (CI: −16.3– −2.0;
the facility. Thus, baseline data were reconstructed concur- p = 0.017), and staff injuries decreased by 2.1 (CI: −4.1–
rently with the process improvement activities described in −0.02; p = 0.034). Injuries sustained in the clinic were elimi-
this study. The EHR was used to extract patient demo- nated (Figure 2). These improvements were sustained during
graphics, arrival and discharge times, and assessment times the following postimplementation year, as shown in Table 1.
for all adults presenting for services. Calls to security were
Phase II Interventions (October–December
compiled from daily security logs. Nonemergent calls such
2014). Phase II outcomes are summarized in Table 2. (For
as routine escorts were excluded. Staff injuries were com-
the statistical analyses, the preimplementation comparison
piled from incident reports. We did not have access to reliable
period for Phase II is July–September because the condi-
data on staff injuries or door-to-doctor times that occurred
tions that existed prior to Phase I do not provide a meaningful
prior to ConnectionsAZ assuming management of the CRC;
comparison condition because of the low percentage of pa-
thus pre-April 2014 data are not included. The percentage
tients receiving psychiatric evaluations by a BHMP and the
of hours on diversion (not accepting transfers from EDs
lack of standardized criteria for diversion.) Observation unit
because of overcapacity) was calculated from daily logs. Stan-
median door-to-doctor time decreased by 6.6 hours (CI:
dardized criteria for diversion were developed in July 2014;
−6.1– −5.1; p < 0.0001). The change over time is de-
thus pre-July data are not included.
picted, as a control chart in Figure 1b. There was an increase
Statistical Analysis and Data Presentation. Outcome following Phase I, as more patients were required to be evalu-
data were analyzed using Minitab 17 (Minitab Inc., State ated by the BHMP, then a reduction after the implementation
College, Pennsylvania) and XLSTAT (Addinsoft, New York of Phase II interventions targeted at BHMP staffing. Hours
City). Wilcoxon rank-sum tests were used to compare non- on diversion decreased by 73 percentage points (CI: −125–
Volume 43, No. 6, June 2017 279

−20; p < 0.0001). These improvements were sustained over

< 0.0001
0.0086

< 0.0001
0.004

0.005
0.019
the following postimplementation year, as shown in Table 2.

p
Redistribution of Space. More efficient use of space
One Year Postimplementation (Figure 3) resulted in 1,046 square feet of unused space in

(−241, −229)

(−14.3, −3.1)
(−1.7 − 0.3)

(−3.9, −0.4)
(107, 380)
the clinic (47% of the total clinic space). The clinic was re-
95% CI

(25, 28)
(Jan–Dec 2015)

modeled to create an overflow observation unit in February

2015, increasing the observation unit capacity from 25 to
34 patients. The seclusion/restraint room in the clinic is no
longer in use.
Difference
(vs. pre)

−0.7

−8.7
−2.1
−244

243
27
DISCUSSION
Table 1. Comparison of Phase I Outcome Variables: Pre- vs. Postimplementation and Pre- vs. One Year Postimplementation

This study describes the application of Lean methods to

84%
23.5
648

achieve a rapid and sustainable transformation of clinical op-

4.8
1.2
99

erations in a behavioral health crisis program. This initiative

began with an engagement and information-gathering phase,
< 0.0001
< 0.0001
< 0.0001
< 0.0001
0.017
0.034

during which we developed value-based outcome mea-

p

sures. Then a multidisciplinary team composed of leaders

Post Phase I Implementation

and frontline staff designed and implemented two consec-

(−224, −208)

(−16.3, −2.0)

Wilcoxon rank-sum tests were used to compare door-to-door dwell times; t-tests were used to compare all other measures.
(−3.7, −2.0)

(−4.1, −0.2)

utive sets of interventions. Phase I was implemented within

(163, 299)
95% CI

(19, 23)
(Jul–Dec 2014)

three months of our assuming management of the facility

and involved a redesign of flow, space utilization, and clin-
ical assessment and care protocols, resulting in improvements
to clinic throughput and safety measures without addition-
Difference
(vs. pre)

al staffing. Phase II involved improvements to the BHMP

−225
−2.0

−9.2
−2.1
232
21

staffing model, resulting in a dramatic decrease in observa-

tion unit door-to-doctor time.
Most published studies of Lean applications in the emer-
78%
22.2
118

637

4.3
1.2

gency setting focus on process outcomes and do not include

outcomes reflecting patient safety, quality, or effects on
(Jan–Jun 2014)

employees.19 Our study demonstrates improvements in such

measures, including staff injuries and calls to security to
57%
24.2

13.5
343

405
Pre

3.3

*Reliable data on staff injuries from before April 2014 data were not available.

manage violent patients. Other measures, such as incidents

of self-directed violence or restraint-related injuries to pa-
tients, were low at baseline and remained unchanged during
CI, confidence interval; BHMP, behavioral health medical professional.

the study period.

Number of Patients Triaged to Observation Unit per Month (mean)

The improvements described in this study were sus-

Observation Unit Door-to-Door Dwell Time (median in hours)

tained through the following year by incorporating Lean

concepts into our organizational structure and culture. Rapid
communication of information to people with the ability
Clinic Door-to-Door Dwell Time (median in minutes)

to problem-solve—a key component of sustainable Lean

implementations20—is accomplished via the new system of
Emergent Calls to Security per Month (mean)

shift leads, shift reports, and daily huddles. Shift leads solve
problems in the moment, while the shift report and huddles
ensure that more complex problems are communicated
quickly to higher-level managers. In addition, shift leads re-
Staff Injuries per Month (mean)*

ceived training in Lean methods to engage more of the

Patients Seen by BHMP (%)

workforce in future improvement activities.21

Unique Considerations for Behavioral Health

Services
Early assessment is associated with positive outcomes in the
ED setting,22–25 but to our knowledge this is the first study
demonstrating this approach in a behavioral health crisis
setting. Several key considerations unique to behavioral health
280 Margaret E. Balfour, MD, PhD, et al Using Lean to Rapidly Transform a Behavioral Health Crisis Program

Improvement in Throughput, January–December 2014

A. Clinic Door-to-Door Dwell Time

Phase I Phase II

2014

B. Observation Unit Door-to-Doctor Time

Phase I Phase II

2014

Figure 1: These X-bar charts depict improvement in throughput measures. Each data point represents the mean of a random
sample of up to 100. The center line (X-bar) represents the process mean. Upper control limits (UCL) and lower control
limits (LCL) are set at three standard deviations above and below the mean, respectively. S-charts plotting the standard
deviation verified that the processes were in control and are not displayed. Clinic door-to-door dwell time decreased fol-
lowing Phase I interventions, and this improvement was sustained during Phase II (Figure 1a). There was an increase in
observation unit door-to-doctor time following Phase I, as more patients were required to be evaluated by the behavioral
health medical professional (BHMP), then a reduction after the implementation of Phase II interventions targeted at BHMP
staffing (Figure 1b). Figure 1a is reprinted with permission of Springer SBM US, from Balfour ME, et al. Crisis Reliability
Indicators Supporting Emergency Services (CRISES): a framework for developing performance measures for behavioral health
crisis and psychiatric emergency programs. Community Ment Health J. 2016;52:1–9.

were aligned with Lean concepts to improve both safety and escalate quickly. It is equally important to engage with pa-
experience for this specialized population (Figure 4). Con- tients in crisis and treat them with respect in the least-
tinuous observation and proactive intervention are critical restrictive environment possible.26 The more efficient use of
to address symptoms and diffuse behaviors that may space resulted in the consolidation of staff onto the locked
observation unit, where the highest staff-to-patient ratio is
needed, and our new assessment process reduced the delay
Number of Staff Injuries for the Observation Unit in moving high-risk patients to this unit. To ensure con-
(Obs Unit) and Clinic, April–December 2014 stant observation, peer supports stay with these patients during
12
the time it takes to move them to the observation unit; thus,
Obs Unit what was once non-value-added waiting time gained value
10 Clinic via peer engagement. Similarly, the assignment of a peer and
technician to the waiting room creates a more therapeutic
8
# Staff Injuries

milieu for patients and families waiting for the clinic process.
6
Early segmentation of low- and moderate-risk patients allowed
us to dispense with the one-size-fits-all approach of treat-
4 ing everyone as dangerous and subjecting them to searches
and instead concentrate our highest level of precautions and
2 safety procedures on the high-risk subpopulation. The re-
allocation of unused clinic space into a smaller overflow
0
Baseline (Apr-Jun) Phase I (Jul-Sep) Phase II (Oct-Dec) observation unit allowed us to further individualize care for
high-risk patients. For example, patients with severe anxiety
Figure 2: The number of staff injuries steadily decreased and
were eliminated from the clinic setting altogether (see Table 1 or psychological trauma often feel more comfortable on this
for statistical analysis). smaller, quieter unit.
Volume 43, No. 6, June 2017 281

Table 2. Comparison of Phase II Outcome Variables: Pre- vs. Postimplementation and Pre- vs. One Year
Postimplementation
Post Phase II Implementation One Year Postimplementation
(Oct–Dec 2014) (Jan–Dec 2015)
Pre* Difference Difference
(Jul–Sep 2014) (vs. pre) 95% CI p (vs. pre) 95% CI p
Observation Unit 8.2 1.6 −6.6 (−6.1, −5.1) < 0.0001 1.4 −6.8 (−6.3, −5.5) < 0.0001
Door-to-Doctor Time
(median in hours)
Hours on Diversion (%) 90% 17% −73 (−125, −20) < 0.0001 34% −56 (−100, −4) < 0.0001
Wilcoxon rank-sum tests were used to compare door-to-doctor time; t-tests were used to compare hours on diversion.
*The preimplementation comparison period for Phase II is July–September because the conditions that existed prior to Phase I do not
provide a meaningful comparison condition due to the low percentage of patients receiving psychiatric evaluations by a behavioral health
medical professional and the lack of standardized criteria for diversion.
CI, confidence interval.

Challenges and Limitations could not be compared, as they were either lost or not con-
As stated, baseline data had been destroyed immediately prior sistently measured prior to our interventions. The need to
to our assuming management of the facility. The severity of intervene quickly also affected the ability to conduct more ex-
the safety concerns necessitated quick action; therefore, we were tensive preintervention measurements. For example, although
building reports, reconstructing baseline data, and creating a we currently administer the Hospital Survey on Patient Safety
quality program concurrently with the work described in this Culture27 and analyze for trends in staff responses, we did not
study. As a result, some measures such as patient satisfaction have the time or bandwidth to collect baseline data prior to

Redistribution of Space

Patient flow based on risk:

unknown risk
Police Entrance
low/mod risk
Sally Port Rooms (SP)
high risk

Only the obs storage

storage

unit Seclusion SR
Room is used SR
offline
Observation Unit

Unused clinic space

remodeled to create new
overflow observation unit TR

Before: TR

Waiting
Room
Walk-in
Entrance

Clinic area
2 Triage Rooms (TR)
Seclusion Room (SR) taken offline

Figure 3: Before the Phase I interventions, staff were spread out over a large area, with mixing of patients with unclear
risk profiles (blue arrows), who often slept overnight in unmonitored, non-ligature safe assessment rooms. After the new
process, risk level is determined early. Green arrows show the flow of low- and moderate-risk patients, and red arrows show
the flow of high-risk patients, who may arrive via the waiting room (walk-ins) or the gated sally port (law enforcement drops).
Staff are consolidated with the high-risk patients on the observation unit. More efficient flow resulted in unused space that
was converted to an overflow observation unit. The process improvement team developed the new flow using an en-
larged laminated floor plan and dry erase markers; their final product looked much like this electronic version.
282 Margaret E. Balfour, MD, PhD, et al Using Lean to Rapidly Transform a Behavioral Health Crisis Program

Alignment of Lean Concepts with Behavioral Health Clinical Goals

Reduce Decreased Treatment Least restrictive care:

• Stabilize & discharge
waste wait time starts earlier • Avoid restraints

Consolidate Safer care:

Efficient use Higher staff to
staff on obs Better monitoring of
of resources patient ratio
unit high risk patients

Assign peer to
Maximize
Focus on sit w/pt while Better patient
value-added
customer waiting for engagement
time
admission
Figure 4: Alignment of Lean concepts with behavioral health clinical goals improves both safety and experience for this
specialized population. Obs unit, observation unit; pt, patient.

our interventions. Such a formal assessment of organizational REFERENCES

culture change would have been a worthwhile endeavor. 1. Womack JP, Jones DT. Lean Thinking: Banish Waste and
Create Wealth in Your Corporation. New York City: Simon
& Schuster, 1996.
2. Toussaint JS, Berry LL. The promise of Lean in health care.
CONCLUSION Mayo Clin Proc. 2013;88:74–82.
This work demonstrates how Lean methods can be applied 3. Womack JP, et al. Going Lean in Health Care. IHI Innovation
to achieve rapid and sustained improvements in safety and Series white paper. Cambridge, MA: Institute for Healthcare
Improvement, 2005. Accessed Mar 27, 2017. https://
throughput in a behavioral health crisis setting. Lean is www.entnet.org/sites/default/files/GoingLeaninHealth
complementary to behavioral health clinical goals, as the CareWhitePaper-3.pdf.
focus on customer experience and eliminating waste can 4. Joosten T, Bongers I, Janssen R. Application of lean thinking
result in processes that deliver care quickly and safely while to health care: issues and observations. Int J Qual Health
also promoting engagement, individualized care, and Care. 2009;21:341–347.
5. D’Andreamatteo A, et al. Lean in healthcare: a
respect. comprehensive review. Health Policy. 2015;119:1197–
1209.
Acknowledgments. The authors acknowledge the contributions of all of 6. Kaplan GS, et al. Why Lean doesn’t work for everyone. BMJ
the staff at the Crisis Response Center. Qual Saf. 2014;23:970–973.
Conflicts of Interest. All authors report no conflicts of interest. 7. Ohno T. Toyota Production System: Beyond Large-Scale
Production. Portland, OR: Productivity Press, 1988.
8. van Rossum L, et al. Lean healthcare from a change
Margaret E. Balfour, MD, PhD, is Vice President for Clinical Innovation management perspective. J Health Organ Manag. 2016 May
and Quality, ConnectionsAZ, and Assistant Professor of Psychiatry, Uni-
16;30:475–493.
versity of Arizona. Kathleen Tanner, MA, LSSBB, is Director of Performance
Excellence, Connections SouthernAZ. Paul J. Jurica, PhD, formerly Quality
9. Imai M. Gemba Kaizen: A Commonsense Approach to a
Analyst, Connections SouthernAZ, is deceased. Dawn Llewellyn, BA, for- Continuous Improvement Strategy, 2nd ed. New York City:
merly Data Support Specialist, is Quality Analyst, Connections SouthernAZ. McGraw-Hill, 2012.
Robert G. Williamson, MD, is Chief Medical Officer, ConnectionsAZ. Chris 10. Moses J. What’s the Difference Between Research and QI?
A. Carson, MD, MBA, is Chief Executive Officer, ConnectionsAZ. Please IHI Open School. 2015. Accessed Mar 27, 2017. http://www
address correspondence to Margaret E. Balfour, margie.balfour@
.ihi.org/education/IHIOpenSchool/resources/Pages/
connectionshs.com.
Activities/Moses-ResearchVsQI.aspx.
11. Davidson L, et al. Peer support among persons with severe
mental illnesses: a review of evidence and experience. World
Psychiatry. 2012;11:123–128.
ONLINE-ONLY CONTENT
12. Wiler JL, et al. Implementation of a front-end split-flow model
to promote performance in an urban academic emergency
See the online version of this article for Appendix 1. Crisis
department. Jt Comm J Qual Patient Saf. 2016;42:271–
Triage Guide. Appendix 2. Visual Management System for 280.
Patient Tracking. Appendix 3. Value Stream Map of the 13. Balfour ME, et al. Crisis Reliability Indicators Supporting
Triage Process. Appendix 4. Sample Shift Report. Emergency Services (CRISES): a framework for developing
Volume 43, No. 6, June 2017 283

performance measures for behavioral health crisis and 21. Berwick DM. Improvement, trust, and the healthcare
psychiatric emergency programs. Community Ment Health workforce. Qual Saf Health Care. 2003;12 Suppl 1:i2–6.
J. 2016;52:1–9. 22. Arya R, et al. Decreasing length of stay in the emergency
14. Lighter DE. Basics of Health Care Performance Improvement: department with a split Emergency Severity Index 3 patient
A Lean Six Sigma Approach. Burlington, MA: Jones & Bartlett flow model. Acad Emerg Med. 2013;20:1171–1179.
Learning, 2013. 23. Imperato J, et al. Physician in triage improves emergency
15. Institute of Medicine. Crossing the Quality Chasm: A New department patient throughput. Intern Emerg Med.
Health System for the 21st Century. Washington, DC: 2012;7:457–462.
National Academy Press, 2001. 24. Rutman LE, et al. Creating a leaner pediatric emergency
16. Rother M, Shook J. Learning to See: Value Stream Mapping department: how rapid design and testing of a front-end
to Add Value and Eliminate Muda, ver. 1.3. Cambridge, MA: model led to decreased wait time. Pediatr Emerg Care.
Lean Enterprise Institute, 2003. 2015;31:395–398.
17. Sikka RK, Kovich K, Sacks L. How every hospital should start 25. Traub SJ, et al. Emergency department rapid medical
the day. Harv Bus Rev. 2014 Dec 5. Accessed Mar 27, 2017. assessment: overall effect and mechanistic considerations.
https://hbr.org/2014/12/how-every-hospital-should-start-the- J Emerg Med. 2015;48:620–627.
day. 26. Richmond JS, et al. Verbal de-escalation of the agitated
18. Huntzinger J. The roots of Lean: training within industry: the patient: consensus ztatement of the American Association
origin of Kaizen. Target. 2002;18(2):9–22. for Emergency Psychiatry Project BETA De-escalation
19. Holden RJ. Lean thinking in emergency departments: a Workgroup. West J Emerg Med. 2012;13:17–25.
critical review. Ann Emerg Med. 2011;57:265–278. 27. Sorra J, et al. AHRQ Hospital Survey on Patient Safety
20. Mazzocato P, et al. Lean thinking in healthcare: a realist review Culture: User’s Guide. Rockville, MD: Agency for Healthcare
of the literature. Qual Saf Health Care. 2010;19:376–382. Research and Quality, 2016.