PRACTICE MANAGEMENT

A Case Study of Quality Improvement

Methods for Complex Adaptive
Systems Applied to an Academic
Hepatology Program
John Fontanesi, PhD,* Anthony Martinez, MD, Toritsesan O. Boyo, MPH,
and Robert Gish, MD
Although demands for greater access to hepatology services that are less
costly and achieve better outcomes have led to numerous quality improvement
initiatives, traditional quality management methods may be inappropriate for
hepatology. We empirically tested a model for conducting quality improvement
in an academic hepatology program using methods developed to analyze and
improve complex adaptive systems. We achieved a 25% increase in volume
using 15% more clinical sessions with no change in staff or faculty FTEs, generating a positive margin of 50%. Wait times for next available appointments
were reduced from five months to two weeks; unscheduled appointment slots
dropped from 7% to less than 1%; no-show rates dropped to less than 10%;
Press-Ganey scores increased to the 100th percentile. We conclude that framing
hepatology as a complex adaptive system may improve our understanding of
the complex, interdependent actions required to improve quality of care, patient
satisfaction, and cost-effectiveness.
KEY WORDS: Hepatology; operations; time-motion; adaptive; clinic.

ealthcare delivery in America is fragmented,1,2

resulting in substantial waste3,4 and unacceptable
patient outcomes.5 Despite numerous national
and regional initiatives and the allocation of
substantial resources, much of the extant data and research
reflect substantial shortcomings in the provision of effective and reliable care.6,7 The introduction of quality improvement methods has done little to mitigate emergency
department overcrowding and ambulance diversion,8,9
improve surgical suite underutilization and decrease excessive staff overtime,10 or ensure that electronic medical
records improve care,11 reduce errors,12-14 increase patientprovider communication,15,16 or improve patient access to
care. Spending more than twice as much on healthcare
as other industrialized countries has been ineffective at
reducing disparities or mortality related to medical error.17
A number of authors have suggested that a reliance on
traditional industrial-quality improvement techniques may

*Director, Center for Management

Science in Health, and Department
of Medicine, Division of General
Internal Medicine, and Departments
of Family and Preventive Medicine
and Pediatrics, University of California,
San Diego School of Medicine, 9500
Gilman Drive, #8415, La Jolla, CA
92093-0821; phone: 619-543-3886;
e-mail: jfontanesi@ucsd.edu; Division
of Gastroenterology, Hepatologyand
Nutrition, University at Buffalo, Buffalo
GeneralMedical Center, Erie County
MedicalCenter, Buffalo, New York.
Director of Ambulatory Services, San
Mateo Medical Center, San Mateo,
California, and Division of Gastroenterology and Hepatology, Stanford University, Stanford, California; Division
of G
astroenterology and Hepatology,
Stanford University, Stanford, California, and Robert G. Gish Consultants
LLC, San Diego, California.
Copyright 2015 by
Greenbranch Publishing LLC.

not be appropriate in light of the complexity of healthcare

delivery systems.18-20 The basic argument is that methods
such as benchmarking, failure mode analysis, Lean, and
Six Sigma are inherently reductionistic as they decompose work into subtasks and subroutines with the goal of
identifying and then reducing variation and waste. There
is an explicit assumption that the processes under review
are deterministic and linear, with minimally complex
multivariate conditions for which there will be one best
way independent of context. Information used to identify
opportunities for improvement tends to use averages of aggregated data, collected over extended time horizons that
are difficult to contextualize at the point of service.
Medical providers typically work in a world of probabilities composed of dynamic treatment options, system
resources, and patient circumstances that require judgment about the various trade-offs in possible combinations
of treatment and lifestyle recommendations unique for a

www.greenbranch.com |800-933-3711

323

324Medical Practice Management| March/April 2015

given patient who can then elect to accept or reject any or
all of those recommendations. This is the antithesis of a
deterministic, well-behaved linear system. Providing care
requires the coordination of multiple entities, operationally managed across administratively distinct groups, each
of which functions semi-autonomously with its own set of
regulations, culture, and logistical requirements united primarily by the desire to provide quality care to an individual
patient. Indeed, a healthcare delivery system possesses all
the attributes of a complex adaptive system, as defined by
Nobel Laureate Murray Gell-Mann of the Santa Fe Institute21 and further refined for healthcare by Rouse19: that of
a dynamic, evolving, nonlinear set of interactions without
centralized control.
Numerous researchers in industries such as computer
sciences, 22, 23 economics, 24 social sciences, 25 and aerospace26 have explored the requirements for effective quality
management of complex adaptive systems. These generally
include the following:
77 Noncentralized control requires the locus of leadership
to switch from role-based actions of an individual to
team-level relationships in which teams:
Organize around a few simple principles focused on
customer goals. Patient-centered care is an example
of organizing around customer needs.
Are encouraged to experiment and innovate in achieving customer goals. This includes the opportunity to
fail as long as the failing is recognized quickly.
Receive timely information on individual customer
needs rather than aggregated data. An example is
the difference between receiving a report on the percentage of hypercholesterolemic patients prescribed
statins or a report showing that the current patient
is a 90-year-old hypercholesterolemic patient with
Alzheimer disease and prostate cancer.
77 Complex adaptive systems are sensitive to their starting
conditions. Small disturbances can have a surprisingly
profound impact on overall system behavior as the
initial disturbance amplifies and propagates downstream. A common example in healthcare occurs
when a patient (or key staff member) is late for the first
appointment of the day. Therefore, it is critical to find
methods to control starting conditions.
77 Complex adaptive systems are context sensitive. For
example, a recent meta-analysis of adoption of evidenced-based medicine found that small hospitals did
best when focused on specific practices, whereas larger
hospital systems did better when efforts were directed to
enhancing the process of culture change.
77 The meaning of data changes at different levels of organization. Using the example cited earlier, the percent of
hypercholesterolemic patients prescribed statins has
different values and meanings at the patientprovider
interface, providerhealth system interface, health

systemregulatory/payer interface, and regulator/

payersocietal interface.

METHODS
To explore how these concepts might be operationalized, we empirically tested a model for conducting quality improvement in a complex adaptive system using an
academic hepatology program. Hepatology was chosen
in part because of the willingness of the leadership and in
part because of the rapidly changing best practices and
multiple specialties involved in treating patients with liver
disease. Chronic liver disease (CLD) is on the increase,
currently affecting an estimated 15% of the U.S. population
and expected to continue to rise.27 It is the eighth leading
cause of death in the United States,28 with an estimated
direct cost of $9.1 billion dollars annually.29 New treatment
breakthroughs that can reverse or at least flatten the trajectory of CLD progression can occur only if a potential patient
can receive timely and appropriate services. The University
of California, San Diego Hepatology Program, as part of the
Division of Gastroenterology, is a nationally recognized program that trains new physicians, conducts a broad array of
research and clinical trials, and serves as the largest regional
provider of care for patients suffering acute and chronic
liver diseases. The mission statement of the hepatology
program is to be the complex liver disease center for the
Imperial, Kern, Orange, Riverside, San Bernardino and San
Diego counties and Southwest Arizona Region.
At the time we began our assessment, the average
wait for a first appointment was five months, but 7% of
all appointment slots went unfilled. The no show rates
exceeded 10%, and patient satisfaction was low. Staff members, organized along a number of different administrative
structures, were highly professional but focused on their
individual organizational imperatives rather than on the
purpose of the program. The resulting silos hindered
continuity of action and information flow, resulting in
burdensome and redundant documentation requirements
where patients were asked the same questions multiple
times, leaving the impression of fragmented and uncoordinated care. Quarterly quality reports often provided
information about processes over which no individual
type of staff member had control (e.g., patients satisfaction
with facility cleanliness) and served to increase isolation
between types of staff with the secondary consequence of
suppressing teamwork and innovation.
The scheduling schema was static in the sense that the
same number and types of appointments were available
on the same days of each month regardless of varying
demand. Furthermore, the schema used traditional proportional appointment allocation methodology (fixed appointment durations based on clinical characteristics such
as a new patient or a patient post-transplant) and assumed
that tardiness followed a normal Gaussian distribution.

www.greenbranch.com
|800-933-3711

Fontanesi, Martinez, Boyo, and Gish|Quality Improvement Methods 325

Table 1. Effects of Queueing on Key Clinic Functions

Activity
Registration*
Receptionist verifies contact and insurance information*
Time in waiting room*
Medications checked by staff*
Direct patient:provider time
Percent of provider recommendations recalled at end of visit*
Average time in clinic*
Patient receives after-visit summary*

Early in Clinic Session

Later in Clinic Session

5 minutes

2 minutes

93%

70%

2 minutes

19 minutes

48%

22%

37 minutes

22 minutes

75%

25%

62 minutes

58 minutes

100%

32%

* p = .01

Table 2. Analysis of Referral Patterns

p Value*
Normal Distribution
Shapiro-Wilk

.006

Kolmogorov-Smirnov test

.0097

Anderson-Darling

.005

Shapiro-Wilk

.023

Kolmogorov-Smirnov test

.049

Anderson-Darling

.045

Shapiro-Wilk

.512

Kolmogorov-Smirnov test

.831

Anderson-Darling

.678

Johnson Distribution

Weibull Distribution

*If p .05, then the distribution family does not fit the observed data.

The result of these assumptions produced queues that increased as the daily clinic progressed.
Receptionists responded to the queues by skipping
verification items such as a patients current contact and
insurance information, which, in turn, affected schedulers
ability to contact patients and increased the rate of insurance denial of benefits. Similarly, medical assistants were
less likely to initiate medication reconciliation, resulting in
physicians spending more time during the exam collecting basic information and less time explaining treatment
options and patients, as objectively measured at the end
of each visit, being less knowledgeable about their disease
and treatment options compared to the times when queues
did not exist. These complex interactions underscored the
need for quality management appropriate to a complex
adaptive system (Table 1).

INITIATION
Accordingly, under the Medical Directors direction, the first
step was to review and clarify the mission statement.This

statement became the fundamental organizing principle

to harmonize the various staffs activities and realign roles
and responsibilities in creating treatment teams. Activities
related to direct patient care for activities for each team
were directed by the physician leader, while purely administrative tasks such as human resource documentation
requirements remained with the original supervisors for
specific types of personnel. The hepatologists took existing
generic treatment protocols and translated them to the specific activities required for individual patients. For example,
initial consults for patients with decompensated cirrhosis
were conducted by the hepatologist who, with the team,
constructed a conditional treatment course for a specific
time horizon that included diagnostic tests and procedures,
and follow-up monitoring visits with the nurse practitioner.
As part of organizing around patient needs, team huddles
were initiated to prepare for upcoming patients and to better coordinate care.
These fundamental changes were supported by changes
in the scheduling schema. Shapiro-Wilk, KolmogorovSmirnov, and Anderson-Darling tests for normalcy found
service demand nonconforming to normal Gaussian assumptions (Table 2). This meant scheduling schemas could
not control service demand. However, Hurst exponents
yielded an index of 0.37, a nonrandom pattern to service
demand suggesting that a flexible scheduling schema
could anticipate and adjust to demand efficiently. Congestion in clinic due to queue development was analyzed with
the primary driver being that very few patients actually
arrived on time. Most arrived early, but a substantial
number arrived late, resulting in waves or pulses of patients all arriving at the registration desk at the same time.
Almost 20% of patients depended on family, friends, or
public transportation to get to their appointments and thus
did not have full control over their arrival time.
Although analysis indicated that length of time with
the provider in the exam room was driven, in part, by the
type, stage, and degree of disease control, it also showed
that two patients with clinically similar disease states could
require significantly different amounts of time based on

www.greenbranch.com
|800-933-3711

326Medical Practice Management| March/April 2015

Table 3. Change in Key Metrics
PreIntervention

PostIntervention

5 months

2 weeks

Unfilled appointment slots*

7%

1%

No-shows

17%

9%

Next appointment scheduled

prior to leaving clinic*

50%

93%

Press-Ganey scores*

50%

100%

Activity
Wait until next available
appointment*

*p = .01

their psychosocial and cognitive characteristics. Most

importantly, the hepatologists and their team could easily
predict which patients would require longer visit times.
This foreknowledge was used in a Bin-Packing scheduling
paradigm such that easy patients were scheduled in the
beginning of each clinic session with more needy patients
scheduled for longer appointments later in the session.

RESULTS
This initiative resulted in a 25% increase in volume, representing a total of 9500 visits. There were 2800 consults and
1400 procedures utilizing 15% more clinical sessions with
no change in either staff or faculty FTEs and generating a
positive margin of 50%. Additionally, wait times for next
available appointments were reduced from five months to
two weeks; unscheduled appointment slots dropped from
7% to less than 1%; no-show rates dropped to less than
10%; and Press-Ganey scores increased from the low 50th
percentile to the 90th to 100th percentile. Additionally, a
40% reduction in patient wait and staff idle times was noted
along with a four-fold increase in new referrals (Table 3).

CONCLUSION
The healthcare system is buffeted by often competing demands for greater access to more personalized care that
both is less costly and achieves better health outcomes.
Hepatology services for patients with acute and chronic
liver disease are no exception. The tendency has been to
try to meet the demands by applying quality improvement
management methods effective in other industries. However, patients with acute and chronic liver disease are cared
for within a classically complex adaptive system for which
traditional centralized command and control management techniques are inappropriate methodologies that risk
improvements in one care domain at a cost to the total
system of care.
The improvements noted in this case study required
considerable time, team effort, and, most importantly,
cultural change. There is certainly risk associated with
such change, but the results speak for themselves. It is

also important to note that in complex adaptive systems,

context matters. The issues affecting other hepatology programs will be similar to but not the same as those identified
at the University of California, San Diego. Solutions also
will need to be localized. However, the concept of team,
examining for patterns instead of averages, and focusing
on patient needs to organize the panoply of professionals
involved in caring for patients with chronic liver disease
may well prove to be universal. Y
REFERENCES
1. Cebul RD, Rebitzer JB, Taylor LJ, Votruba M. Organizational Fragmentation and Care Quality in the U.S. Health Care System. Working Paper
No. 14212. Cambridge, MA: National Bureau of Economic Research,
August 2008.
2. Eye on Patients. A Report by the Picker Institute for the American Hospital Association. Boston: Picker Institute, 2000.
3. Berwick DM, Hackbarth AD. Eliminating waste in US health care.
JAMA. 2012;307:1513-1516.
4. Building a Better Delivery System: A New Engineering/Health Care
Partnership. Washington, DC: The National Academies Press; 2005.
5. Weingart SN, Wilson RM, Gibberd RW, Harrison B. Epidemiology of
medical error. West J Med. 2000;172:390-393.
6. Radley DC, McCarthy D, Lippa JA, Hayes SL, Schoen C. Aiming
Higher: Results from a Scorecard on State Health System Performance, 2014. New York: The Commonwealth Fund, May 2014; www.
commonwealthfund.org/publications/fund-reports/2014/apr/2014state-scorecard. Accessed February 3, 2015.
7. Leape LL, Berwick DM, Bates DW. What practices will most improve safety? Evidence-based medicine meets patient safety. JAMA.
2002;288:501-507.
8. Burt CW, McCaig LF, Valverde RH. Analysis of ambulance transports
and diversions among US emergency departments. Ann Emerg Med.
2006;47:317-326.
9. Felton BM, Reisdorff EJ, Krone CN, etal. Emergency department
overcrowding and inpatient boarding: a statewide glimpse in time.
Acad Emerg Med. 2011;18:1386-1391.
10. Saleh KJ, Novicoff WM, Rion D, etal. Operating-room throughput:
strategies for improvement. J Bone Joint Surg Am. 2009;91:2028-2039.
11. Garg AX, Adhikari NK, McDonald H, etal. Effects of computerized
clinical decision support systems on practitioner performance and
patient outcomes: a systematic review. JAMA. 2005;293:1223-1238.
12. Friedberg MW, Chen PG, Van Busum KR, etal. Factors Affecting
Physician Professional Satisfaction and Their Implications for Patient
Care, Health Systems, and Health Policy. Santa Monica, California:
Rand Corporation; 2013.
13. Weingart SN, Toth M, Sands DZ, etal. Physicians decisions to override computerized drug alerts in primary care. Arch Intern Med.
2003;163(21):2625-2631.
14. Battles JB, Keyes MA. Technology and patient safety: a two-edged
sword. Biomed Instrum Technol. 2002;36:84-88.
15. McGrath JM, Arar NH, Pugh JA. The influence of electronic medical
record usage on nonverbal communication in the medical interview.
Health Informatics J. 2007;13:105-118.
16. Makoul G, Curry RH, Tang PC. The use of electronic medical records:
communication patterns in outpatient encounters. J Am Med Inform
Assoc. 2001;8:610-615.
17. Davis K, Stremikis K, Squires D, Schoen C. Mirror, Mirror on the Wall,
2014 Update: How the U.S. Health Care System Compares Internationally. New York: The Commonwealth Fund; June 2014.
18. Basole R, Rouse WB. Complexity of service value networks: conceptualization and empirical investigation. IBM Syst J. 2008;47(1):53-70.
19. Rouse WB. Health care as a complex adaptive system: implications for
design and management The Bridge. 2008;38(1):17-25.
20. Plesk P. Appendix B: Redesigning health care with insights from
the science of complex adaptive systems. In: Committee on Quality
of Health Care in America, ed. Crossing the Quality Chasm: A New
Health System for the 21st Century. Washington, DC: National Academy Press; 2001.

www.greenbranch.com
|800-933-3711

Fontanesi, Martinez, Boyo, and Gish|Quality Improvement Methods 327

21. Gell-Mann M. The Quark and the Jaguar: Adventures in the Simple
and the Complex. New York: Holt; 1994.
22. Simons E, Dana S, Griesser A. Automation tools supporting the
development of information standards for complex systems. In:
Frey DD, Fukuda S, Rock G, eds. Improving Complex Systems Today:
Proceedings of the 18th ISPE International Conference on Concurrent
Engineering. London: Springer; 2011:11-21.
23. Chen PP-S. The entity-relationship model: toward a unified view of
data. ACM T Database Syst. 1976;1(1):9-36.
24. Day RH. Complex Economic Dynamics, Volume I: An Introduction to
Dynamical Systems and Market Mechanisms. Cambridge, MA: The
MIT Press; 1994.
25. Jackson MO. A survey of network formation models: stability and
efficiency. In: Demange G, Wooders M, eds. Group Formation in

26.

27.
28.
29.

Economics: Networks, Clubs, and Coalitions. Cambridge, UK: Cambridge University Press; 2005:11-49.
Uckun S, Kurtoglu T, Bunus P, Tumer I, etal. Model-Based Systems
Engineering for the Design and Development of Complex Aerospace
Systems. SAE Technical Paper 2011-01-2664, October 18, 2011,
doi:10.4271/2011-01-2664.
Younossi ZM, Stepanova M, Afendy M, etal. Changes in the prevalence
of the most common causes of chronic liver diseases in the United
States from 1988 to 2008. Clin Gastroenterol Hepatol. 2011;9:524-530.
Hoyert DL, Xu J. Deaths: preliminary data for 2011. Natl Vital Stat Rep.
2012;61(6):1-51.
Pyenson B, Fitch K, Iwasaki K. Consequences of hepatitis C (HCV):
costs of a baby boomer epidemic of liver disease. New York: Milliman,
May 2009.

www.greenbranch.com
|800-933-3711