Applying the System Dynamics A pproach in Evaluating Clinical
Risk Management Policies in Three Healthcare C ompanies
Francesco Ceresia Giovan Battista Montemaggiore
Department of European Studies and International Integration Department of Business Management
University of Palermo (Italy) University of Palermo (Italy)
fceresia@ alice.it montemaggiore@ economia.unipa.it
Abstract
This paper explores and extends research on the role of system dynamics methodology as a
powerful approach to clinical risk management (CRM). We report our preliminary findings
on CRM in three healthcare organizations. We use system dynamics methodology for
exploring the multi-dimensional facets of hospitals’ complex operations management
systems. We address theoretical scholarly matters focusing on the depiction of managerial
insights to gain more understanding of CRM. We investigate the impacts of CRM
implementation on the hospital financial performance along with other indicators. We
provide a summary of our findings and their empirical and theoretical implications and
contributions.
Key words clinical risk management, healthcare systems, system dynamics, hospitals’
financial performance indicators, human resource management
1. Introduction
The need for exploring and understanding Clinical Risk Management (CRM) is important
in that hospitals are high-risk organizations with multi-faceted structural dynamics,
elaborate internal operations, varied external environments, fluid organizational cultures,
and multiple stakeholders with numerous interests and expectations. An increased level of
patient safety awareness came to light due to wide media coverage of clinical errors. The
high number of exorbitant compensation claims and the steep rise of insurance costs have
recently forced healthcare companies to seriously consider CRM.
We focus our efforts on drawing managerial insights for better understanding of human
behavior on the complex processes of CRM implementation. Our findings are based on
data collected from a sample of 204 clinical and managerial staff members of doctors,
nurses, and clinical risk managers in three healthcare companies in Italy.
The main objective of this paper is to present our research preliminary results along with
their implications. This will be accomplished by describing a research project aimed at
building a management flight simulator, based on the system dynamics (SD) methodology.
Our goal is to, ultimately, support healthcare companies’ management in experimenting
with different CRM policies by monitoring their potential effect on the financial and non-
financial performance indicators. The management flight simulator is intended to help
healthcare managers in designing CRM policies to guarantee both a satisfying level of
patient safety and a sustainable growth.
Risks associated with patient care cannot be totally eliminated; therefore, CRM plays a vital
role in enabling hospitals to enhance patient safety (Vincent, 2006). It is important to note
that Risk Management (RM) generally encompasses political, legal, and business
environment risks (Young et al., 2002). Additionally, CRM is a specific form of RM
focusing on clinical processes directly and indirectly related to the patient.
It follows that CRM can be defined as all structures, processes, instruments, and activities
which enable hospital staff to identify, analyze, contain, and manage risks while providing
clinical treatments and patient care (Walshe, 2001). Naturally, there are other aspects of
hospital governance that influence patient safety. They include financial or infrastructural
risk management, or health policy issues such as hospitals accreditation matters. The core
point is that systematic CRM integrates both a proactive and reactive approaches, and
frames the hospital as a system, instead of focusing on individuals and their potential for
committing errors (Corrigan et al.; 2001, Misson, 2001, Reason, 2000). This is why
hospitals are an ideal setting for understanding and implementing the system dynamics
paradigm.
It is important to note that CRM practices, such as the introduction of guidelines and
protocols, patient involvement, etc., do not take into account expenditures and their
influence on the employees’ behavior. Indeed, such practices, if not properly managed, can
give rise to medics and paramedics’ work overload bumout, which would inevitably
increase the probability of errors.
In other words, improving hospitals’ risk profile, although ethically incumbent, often
requires significant investments. Therefore, some healthcare companies opt not to invest in
CRM due to its high cost, and the added complexity of its operating procedures. Factored in
their decisions are also the difficulties experienced in appraising the outcome benefits from
investments aimed at reducing clinical risks.
The relevance of human factors in the occurrence of errors and effects of CRM policies on
the management of personnel also dictate the adoption of a human resource management
perspective, that may help healthcare managers in evaluating the role of staff behaviors in
the success of CRM policies and, therefore, in designing CRM interventions that take into
consideration the workforce attitudes, motivations, biases, etc.
Indeed, improvements of clinical risk profile often allow hospitals to realize important
savings on insurance costs. It can also boost institutions’ image and increase their
competitive advantage. For this reason, it is essential to adopt a systemic and multi-
dimensional (Berg, 2010) approach that allows healthcare companies to properly evaluate
CRM policies effects on organizations’ performance, in the short, and medium long term.
2. Literature Review and Theoretical Framework
This section provides literature review on two interrelated elements: the CRM process and
the system dynamics perspective. Recent work (Briner, 2010) suggests that despite the
multitude of initiatives, programs, systems, and tools that can be viewed as elements of
CRM there is a lack of knowledge concerning their implementation in hospitals.
Hospitals have always been concerned with enhancing patient safety. However, it is
noteworthy that this issue became a core consideration since the publication of the Institute
of Medicine reports “To Err Is Human, 1999”, and “Crossing The Quality Chasm, 2001”.
Afterwards, a widespread application of systematic CRM has taken place (Vincent, 2006;
Misson, 2001; Chiozza et al., 2006). At the organizational level, many RM tools have been
adapted from other high-risk industries such as aviation. Incident reporting is also gaining
increased acceptance among hospitals and is viewed as a possible method for promoting
learning from errors (Barach et al., 2000; Leape, 2002; Secker et al., 2001). In addition,
several patient safety initiatives have been launched at both the national and intemal levels
(Joint Commission, 2008).
2.1. Clinical Risk Management
Research by Young et al. 2002 and Misson (2001) shows that the following dimensions
represent three major variables of risk management:
e Risks to Patients: Following medical ethical standards is key to minimizing risks and
maintaining patient safety. This is in addition to compliance with statutory regulation;
learning from complaints; and also ensuring regular systems reviews and questioning -
by critical event audit.
e Risks to Practitioners: Ensuring that clinicians are immunized against infectious
diseases, work in a safe environment, and are helped to stay current as essential parts of
quality assurance.
e Risks to the Organization: Poor quality is a threat to any organization. In addition to
reducing risks to patients and staff, organizations need to ensure high quality
employment practices, by introducing measures to review individual and team
performance, and introducing well-designed policies on public involvement.
CRM is an approach for improving the quality and safe delivery of health care. This can be
accomplished by placing special emphasis on identifying conditions that put patients at
risk, and by establishing mechanisms to minimize or prevent these risks.
This point to the fact that CRM systems are essentially dedicated to delivering risk
reduction strategies. It is also important to emphasize that CRM goals include:
identification of risks, prevention of harm, injury and loss, and controlling systems and
processes with the deliberate goal of eliminating or reducing severity of damage.
It is important to note that healthcare companies adopt strategies to identify potential causes
of active or latent errors. They also implement organizational procedures aimed at
eliminating the causes of the identified errors. Therefore, the adoption of the CRM
approach in this research project is aimed at initiating a cultural change oriented toward
increasing patient safety in the companies participating in the project (McFadden et al.,
2009).
Vincent et al. (1998) proposed a general framework of factors influencing clinical practice
and contributing to medical adverse events (Table. 1).
Table 1. Framework of Factors Influencing Clinical Practice and Contributing to Adverse Events
(Vincent et al., 1998)
Framework Contributory Factors Examples of Problems That
Contribute to Errors
Institutional | Regulatory context Insufficient priority given by regulators
Medico-legal environment to safety issues;
National Health Service Legal pressures against open discussion,
Executive preventing the opportunity to learn from
adverse events
Organization | Financial resources and Lack of awareness of safety issues on
and constraints the part of senior management;
management | Policy standards and goals Policies leading to inadequate staffing
Safety culture and priorities levels
Work Staffing levels and mix of skills | Heavy workloads, leading to fatigue;
environment | Patterns in workload and shift Limited access to essential equipment;
Design, availability, and Inadequate administrative support,
maintenance of equipment leading to reduced time with patients
Administrative and managerial
support
Team Verbal communication Poor supervision of junior staff;
Written communication Poor communication among different
Supervision and willingness to | professions;
seek help Unwillingness of junior staff to seek
Team leadership assistance
Individual Knowledge and skills Lack of knowledge or experience;
staff Motivation and attitude Long-term fatigue and stress
member Physical and mental health
Task Availability and use of protocols | Unavailability of test results or delay in
Availability and accuracy of test | obtaining them;
results Lack of clear protocols and guidelines
Patient Complexity and seriousness of | Distress;
condition
Language and communication
Personality and social factors
Language barriers between patients and
caregivers
Although Vincent’s general framework depicts the main factors contributing to clinical
errors, the underlying approach is far from a root-cause analysis perspective for two reasons:
First, the root-cause analysis hypothesizes that there is a single or at least a small number of
root-causes, while clinical evidences demonstrate that errors are often a consequence of a
wide array of factors. Second, despite the primary aim of the root-cause analysis is to find
the real cause for errors, the main goal of a deeper analysis should be the identification of
gaps in the system, where the approach is much more proactive and forward-looking. For
these reasons, Vincent (2003) calls this deeper approach “systems analysis”. The research
team found it requisite to model the hospital’s systems utilizing the system dynamics
methodology to capture the complexity characterizing the environment where CRM
policies and behavioral operations techniques have to be implemented.
It is important to note here that the three dimensions of risk management identified above
are not intended to be captured by the study questionnaire (to be discussed below). Instead,
they were intended to serve as part of the theoretical framework to simply help readers
understand the landscape and the broader context of the research undertaken. This
reinforces professional research conventions requiring that the literature review/theoretical
framework to be much broader than the scope of the study questionnaire. In short, these
three factors will not be further explored or captured by the questionnaire.
2.2 The System Dynamics Perspective
This is the second element of our theoretical framework. System dynamics is a
methodology for understanding the behavior of complex systems over time. It deals with
internal feedback loops, time delays, stocks, and flows that affect the behavior of the entire
system. These elements help describe how even seemingly simple systems display baffling
nonlinearity (Sterman, 2001; Repenning, 2001). System dynamics uses tools like causal
mapping and simulation modeling (Bendoly et al., 2010). Traditional system dynamics
models incorporate boundedly rational individuals’ decisions as well as heuristics and
biases, and examine their impact in complex dynamic settings, where the results of
individuals’ decisions change the future state of the system which, in turn, influences future
decisions.
Research by Bendoly et al. 2010 demonstrates that there are two types of misperceptions of
feedback: structure and dynamics. Misperceptions of feedback structure are caused by
mental maps that have a poor representation of the complexity of the real system; for
instance, a mental model that ignores important feedback processes in the system.
Misperceptions of feedback dynamics are caused by inaccurate mental models of how the
system behaves. In this case, a mental model that fails to capture the impact caused by
accumulations will poorly infer their dynamics. Sterman’s work (1989) suggests that the
misperception of feedback arises from people’s adoption of deficient dynamic mental
models that guide decisions. These deficiencies include an event-based perspective,
focusing on specific events instead of the system structure that generates them; an open
loop view of causality where previous decisions lead to outcomes and do not change the
current state; failure to understand the impact of delays and of accumulations by not
separating cause and effect; and insensitivity to nonlinearities, which alter the structure and
behavior of the system. The dangers of these misperceptions were very well articulated by
(Bendoly et al. 2010). They suggest that these misconceptions cause decision making
errors.
Despite the relative newness of the adoption of the system dynamics methodology in the
CRM field, different examples of applications of the system dynamics approach to the
healthcare sector have been reported in the literature (Dangerfield, 1999; Wolstenholme,
1999; Homer & Hirsch, 2006). These important contributions highlighted the numerous
advantages of using system dynamics models to manage the complexity characterizing the
healthcare sector.
3. Clinical risk management in the healthcare context
CRM has not often found a real application in three healthcare companies. Healthcare
organizations limited their engagements to a formal implementation of the prescribed
procedures without any substantial improvement in the patient safety culture. In fact, many
initiatives were prompted by media campaigns about serious adverse events resulting from
clinical errors. These initiatives were concluded immediately after the initial euphoria.
National and regional institutions have managed the problem of medical malpractice by
enacting mandatory rules and regulations that have required healthcare companies to
participate into a data collection activity aimed at feeding a central error — monitoring
system. However, the central error monitoring system did not allow for collecting reliable
information about the so called “near miss” and “no harm” events. In addition, the internal
clinical risk committees limited their activities to suggesting procedures that have not been
implemented. Furthermore, in Italy, no relevant data are available about the occurrence of
clinical errors, their main causes, the definition of performance indicators aimed at
measuring the improvement in the management of the clinical risk (Trucco and Cavallin,
2006.
In short, it appears that incentives for healthcare companies to adopt CRM policies is
simply lacking. Therefore, it is important for such companies to realize that improvements
of their risk profile would not only allow them to obtain considerable savings on insurance
costs, but would also enable them to enhance their image, reputation, and increase their
competitive advantage. So far, very few companies have applied for or attained the
accreditation from the Joint Commission Intemational, the most prominent non-
governmental and non- profit organization that certifies healthcare organizations if they
meet a set of standard requirements designed to improve quality of care. From the above
analysis, we can conclude that despite the numerous attempts of the national and regional
governments to spread CRM practices, they remain quite limited. A real change in the
patient safety culture can be realized only if the required investments to improve healthcare
organizations’ risk profile are economically feasible. However, in order to properly
implement cost-benefit analyses, the healthcare companies’ management should quantify
short and medium-long term effects of CRM policies. Such policies may include a financial
aspect like compensation costs, insurance premiums, revenues, and the non-financial
variables including company image, customer satisfaction, personnel motivation.
4. An Assessment of the System Dynamics methodology application to CRM
Presently, in order to detect errors and assess their potential effects, clinical risk managers
adopt monitoring tools, such as incident reporting, clinical audit, and methods of process
analysis, such as the root-cause analysis and the hospital failure mode and effect criticality
analysis. However, these methods are based on a linear analysis of the causal relationships
characterizing the business processes. In particular, they do not take into account feedback
structure underlying the net of causality connecting the variables of the different company
sub-systems (Lee et al., 2009). Furthermore, these analyses are static (Cavallin et al., 2006),
namely they ignore delays normally existing between the triggering of the cause and the
occurrence of the related error and, consequently, they are not suitable to simulate future
trends (Treek, 2008).
Also, the present time clinical risk assessment methods are inadequate in helping healthcare
organizations in setting safety targets and evaluating safety performance improvement on a
quantitative basis (Trucco and Cavallin, 2006). Moreover, the root - cause analysis can also
be misleading because it focuses only on identifying the root cause, but an adverse event
usually does not have a single root cause (Trucco and Cavallin, 2006).
The limitations of the system dynamics methods discussed above may undermine the
identification of the real company processes’ criticalities. Similarly, the organizational
practices implemented to reduce the clinical risk, such as the “only therapy sheet”, the
introduction of guidelines and protocols, the patient involvement, etc., often increase
workload bumout, which inevitably augment the probability of errors. Therefore, it is
necessary to adopt a multi-dimensional and systemic (Cook and Rasmussen, 2005)
approach that allows hospitals to assess, according to a holistic perspective, the effects of
CRM policies on the company performance.
5. Research Methodology
The study was carried out in three healthcare companies: (1) a private hospital (identified
here as Hospital A), placed in a little town near to a big city (the capital of the Region),
which serves a population of 30.000 people; (2) a private hospital (identified here as
Hospital B), placed in the big city, which serves a population of about 700.000 people; (3) a
public hospital (identified here as Hospital C), placed in a medium town 200 kilometers
away from the big city, which serves a population of 120.000 people. Table 2 shows some
macro-variables of these hospitals.
Table 2. Some macro-variables of the hospitals involved in the research.
Hospital (A) (B) (C)
Location Small Town (29.000) Big Town (655.000) Medium Town
(77.000)
Type Private Private Public
Beds 60 (Normal = 54; Day | 94 (Normal =85; Day | 226 (Normal=187;
H. =6) H. =9) Day H.=39)
Employees 62 163 517
Annual Budget (2010)
€ 4.122.422,00
€ 12.922.323,00
€ 19.980.582,00
Annual Budget (2009)
€ 4.340.370,00
€ 12.650.595,00
€ 21.127.943,00
Average Income per € 4.330,28 (2010) € 2.127,83 € 2.510,75
Patient (2010)
Average Income per € 5.136,53 (2009) € 2.184,90 € 2.511,34
Patient (2009)
In-patients (2010) 952 (M = 665; S =| 6073 (M =1115; S | 7958 (M =6276;S =
287) =4958) 1682)
In-patients (2009) 845 (M = 613; S =| 5790 (M =991; S | 8413 (M =6906;S =
232) =4799) 1507)
ER NO NO YES
Clinical Risk Committee YES YES YES
Clinical Risk Manager YES YES YES
Hospital Surgery Orthopedics (surgery) | Midwifery (surgery), All Specialties
Specialties and Cardiology Urology (surgery) and
(pacemaker) Cardiology
(pacemaker)
As shown in Table 2, all hospitals have a clinical risk committee and a clinical risk
manager. According to the National Healthcare System, the Hospital “C” is not an
autonomous hospital from a managerial perspective, but it is part of a regional healthcare
District placed in a medium town in the middle of the region. The General Manager of the
Hospital “C” is the head of the regional Healthcare District, which groups two main
hospitals and other medical and surgical services.
5.1 Subjects
To collect data about the adopted CRM procedures, and the dynamics of clinical errors and
formal complaints at each of the three hospitals we interviewed: The General Manager, the
Medical Director and the Clinical Risk Manager. Furthermore, a questionnaire was
administered to the hospital personnel. To qualify for inclusion, staff members had to have
worked in the hospital for a minimum of one month prior to administering the
questionnaire. As a rule-of-thumb, we invited all personnel within a clinical area to
participate. - that influence or are influenced by the “working environment", e.g.,
Attending/Staff Physicians, Resident Physicians, Registered Nurses, Charge Nurses,
Pharmacists, Respiratory Therapists, and Technicians; responses were voluntary. Table 3
shows the main sample demographic characteristics.
Table 3. The sample demographics characteristics.
HOSPITALS
Hospital A Hospital B Hospital C
N % N % N %
Response 35 56% 41 25 128 25
Rate
Job Profile:
Doctor 11 31,43% 13 31,71% 42 32,81%
Nurse 10 28,57% 18 43,90% 51 39,84%
Staff 11 31,43% 10 24,39% 10 7,81%
missing data 3 857% 0 0,00% 25 19,53%
Sex:
Male 23 65,71% 14 34,15% 39 30,47%
Female Z 20,00% 26 63,41% 31 24,22%
missing data 5 14,29% 1 244% 58 45,31%
Data were collected in agreement with units’ leaders. To keep track, the questionnaires
were numbered, respondents’ names were not recorded, and there were no name-and-
number lists. The subjects were asked about their role -doctor, nurse, staff- and gender. No
personal information was collected to avoid fear of respondents’ identification. As shown
in Table 3, the total sample size was 204; and the response rate was: 56% in Hospital A;
25% in Hospital B; and 25% in Hospital C respectively. The response rates per job profiles
and gender are presented in the table.
5.2 Methods, data collection, and research instruments
5.2.1 Designing the causal loop diagram (CLD) by group model building (GMB)
sessions
Following Vennix et al. (1992), three main tasks were performed by modelers before the
intervention: elicitation of information, exploring courses of action or convergent tasks, and
evaluation. Once group members agreed about the procedures, the first phase started and
adaptation of the model was performed. During this phase, interviews, cognitive maps,
nominal group techniques, and workbooks were the main instruments used. During the
convergent tasks phase, the subjects were called to choose between alternative problems
elaboration, structural model and different policies. This phase was characterized by
intensive of face- to- face discussion techniques. During the evaluation phase, the group
discussed and agreed on the different issues. As a result of the intervention, not only
choices were assumed, but also changes involving the mental models were pursued.
Research by (Vennix et al., 1997) indicated that GMB has been viewed as a method to
facilitate a stimulating learning process. The main output of the GMB sessions was the
CLD, a document that describes the causal relationship between the key-variables of the
three healthcare companies in this study.
5.2.2 Exploring the CRM procedures in the hospital
To explore the CRM procedures adopted by the three hospitals, we asked the clinical risk
managers to refer to the official document describing the CRM company plan (Audit Plan).
The document defines the responsibilities, activities, and records to be made to ensure an
effective prevention of clinical risk caused by medical activities and the management of
adverse events. An Audit Plan had to be developed in accordance with the national laws.
Furthermore, the flow-charts for each hospital were analyzed, the overall clinical processes
were activated by the personnel to figure out each process phase. This was followed by the
evaluation of the clinical risks, and an estimation of the potential injuries to the patients and
the mechanisms adopted to avoid potential injuries.
10
5.2.3 Exploring the professional staff’s perceptions of the quality of the CRM in the
hospital
The questionnaire referred to above is based on the Vincent’s framework of factors
influencing clinical practice and contributing to adverse events, (Vincent et al., 1998). The
main aim of the questionnaire is to provide measurements of the seven main frameworks
depicted: Institutional, Organization and management, Work environment, Team,
Individual staff member, Task, Patient. A copy of the questionnaire appears in the
Appendix.
Since it was not possible to use the Vincent et al. (1998) questionnaire to measure the
clinical risk factors depicted as it contains no items to this effect, we utilized a
questionnaire developed by Sexton et al. (2006a) that explores the personnel attitude
conceming the safety culture in hospitals.
The Safety Attitude Questionnaire (SAQ) was initially developed to assess the quality of
safety and teamwork related norms and behaviors of individual workers, in a particular
setting (Sexton et al., 2006b-c). The safety culture has been defined as "the product of
individual and group values, attitudes, perceptions, competencies, and patterns of behavior
that determine the commitment to, and the style and proficiency of an organization's health
and safety management"(Sorra, 2004). The SAQ Short Form version adopted in this
research is a single page questionnaire with 36 items and demographics information (role
and gender). Each of the 36 items is answered using a five-point Likert scale (Disagree
Strongly, Disagree Slightly, Neutral, Agree Slightly, Agree Strongly). The questionnaire
comprises six categories: Teamwork Climate (6 items), Safety Climate (7 items),
Perceptions of Management (10 items divided in two sections: 5 items for Hospital
Management section and 5 items for Unit Management section), Job Satisfaction (5 items),
Working Conditions (4 items), and Stress Recognition (4 items). Table 4 shows the SAQ
factors definitions including example items for each of the six categories.
Table 4. SAQ factor definitions and example items.
Scale: Definition Example items Scale
Teamwork climate: Perceived | Disagreements are appropriately resolved (i-e., not
quality of collaboration between | who is right, but what is best for the patient)
personnel The physicians and nurses here work together as a
well-coordinated team
Job satisfaction: Positivity I like my job
about the work experience This hospital is a good place to work
Perceptions of management: Hospital Management section:
Approval of managerial action | Hospital management supports my daily efforts;
Hospital management is doing a good job
Unit Management section:
Unit management supports my daily efforts; Unit
management is doing a good job
Safety climate: Perceptions of a | I would feel safe being treated here as a patient
strong and proactive lam encouraged by my colleagues to report any
11
organizational commitment to
safety
patient safety concems I may have
Working conditions: Perceived
quality of the work environment
and logistical support (staffing,
equipment etc.)
This hospital constructively deals with problem
physicians and employees
All the necessary information for diagnostic and
therapeutic decisions is routinely available to me
Stress recognition:
Acknowledgement of how
performance is influenced by
stressors
situations
When my workload becomes excessive, my
performance is impaired
Iam more likely to make errors in tense or hostile
To develop a questionnaire that would fit with the Vincent’s assumption about the seven
frameworks, we added several new items to represent the clinical risk contributory factors
that are not represented in the SAQ. Table 5 shows the SAQ items added to each of the
Vincent’s factor frameworks.
Table 5. The item of the new ire to measure the Vincent's frameworks.
Framework ITEMS Example items added by authors
Institutional 4 item developed by the authors | The lawmaker does not sufficiently
4 ITEM protect the patient’s right to be treated
in compliance with high safety
standards
Organization 2 item developed by the authors | The organizational models adopted by
and 7 item from Safety climate [SC] | the company reveal a deep culture of
management | scale (SAQ) CRM
9 ITEM
Work 2 item developed by the authors | In my Unit there is a good balance
environment | 5 item from Perceptions of between the number of doctors and
11 ITEM Hospital Management [PHM] nurses
scale (SAQ)
4 item from Working Conditions
[WC] scale (SAQ)
Team 5 item from Perceptions of Unit
9 ITEM Management [PUM] scale
(SAQ)
4 item from Teamwork Climate
[TC] scale (SAQ)
Individual 3 item developed by the authors | The doctors have an expertise and
staff 5 item from Job Satisfaction [JS] | experience appropriate for the
Member scale (SAQ) complexity of clinical cases treated
12 ITEM 4 item Stress Recognition [SR]
scale (SAQ)
12
Task 3 item developed by the authors | It is expected to observe a strict
3 ITEM medical protocol for the most of the
clinical activities carried out in this
Hospital
Patient 3 item developed by the authors | Patients have difficulties in speaking
3 ITEM correctly
5.2.3.1 The internal consi: e of the new questi ire scales.
To test the internal consistency of the new scales, an exploratory and confirmatory factor
internal consistence analysis was conducted. We launched a reliability analysis using the
Cronbach’s Alpha Coefficient. In accordance with Nunnally (1978), we consider a
Cronbach’s Alpha value equal or greater than 0.70 as an acceptable reliability coefficient,
although lower thresholds are sometimes used by others. The confirmatory items reliability
analysis was conducted and a confirmatory factor analysis (CFA) was performed using the
statistical program AMOS 4.0 (Arbuckle, 1999).
The hypothesized factor structure defined in Table 5 was compared with the empirical data,
allowing each item to saturate on a single factor, and by setting to zero all other factor
loadings. Covariances between the factors were free parameters. To fix the measurement
scale of each factor, their variance was set at 1.0. The goodness of fit of the model was
verified by the following indices: ”; the ratio between y and the degrees of freedom of the
model (C/gl); the comparative fit index CFI (Bentler, 1990); the Tuker-Lewis index TLI
(Bentler & Bonett, 1980); the root mean square error of approximation (RMSEA).
The first framework depicted by Vincent et al. (1998) is the Institutional one. The
Cronbach's Alpha value is 0.544 (F = 6.584; p = .000), and the corrected item-total
correlation range between .20 and .42.
Table 6. Indices of goodness of fit of the model for the new Questionnaire Frameworks (N = 204).
Framework Model 2 g = x2/g TLI CFI RMSEA p
Institutional A 0,22 2 0,107 1 1 0 0,898
100,52 27 3,723 .758 .819 1 0.000
Organization and A
Management B 61,60 19 3,242 .841 .892 10 0.000
Work A 14510 44 3,298 .884 907 10 0.000
Environment
Team A 73,03 44 166 959 967 .05 0,004
Individual Stat? ~~ (248,01 544,593 619-6883 0.000
Member
B 98,30 48 2,048 .889 .919 .07 0.000
13
As Table 6 shows, the model fit to the data in a satisfactory way. The analysis of the
standardized estimates of factor loadings reveals that the estimated parameters are
substantial (range between .28 and .47) and the standard errors are acceptable (range
between .09 and .19).
The second framework depicted by Vincent et al. (1998) is the Organization and
Management one. The Cronbach's Alpha value is 0.755 (F = 15.992; p = .000), and the
corrected item-total correlation range between .26 and .63. One item shows a corrected
item-total correlation equal to .07. As shown in Table 6, the indices of goodness of fit of
the hypothesized factor structure (model A) show a fit that is not fully satisfactory. The
model was therefore modified (Model B) by taking steps, based on indications from the
post-hoc diagnostic procedure (Modification Indices - MI). We deleted the item that
showed the low corrected item-total correlation and added the covariance between the
errors of the item ORG02 and ORG03-SC. These modifications improved the fit between
the model and the data in a satisfactory way. The analysis of the standardized estimates of
factor loadings revealed that the estimated parameters are substantial (range between .55
and .79) and the standard errors are acceptable (range between .06 and .16).
The third framework depicted by Vincent et al. (1998) is the Work Environment one. The
Cronbach's Alpha value is 0.905 (F = 22.611; p = .000), and the corrected item-total
correlation range between .39 and .82. As shown in Table 6, the model fit to the data in a
satisfactory way. The analysis of the standardized estimates of factor loadings reveals that
the estimated parameters are substantial (range between .35 and .89) and the standard errors
are acceptable (range between .06 and .08).
The fourth framework depicted by Vincent et al. (1998) is the Team one. The Cronbach's
Alpha value is 0.899 (F = 13.064; p = .000), and the corrected item-total correlation range
between .43 and .75. As shown in Table 6, the model fit to the data in a satisfactory way.
The analysis of the standardized estimates of factor loadings reveals that the estimated
parameters are substantial (range between .36 and .77) and the standard errors are
acceptable (range between .05 and .11).
The fifth framework depicted by Vincent et al. (1998) is the Individual one. The Cronbach's
Alpha value is 0.791 (F = 86.998; p = .000), and the corrected item-total correlation range
between .25 and .61. As shown in Table 6, the indices of goodness of fit of the
hypothesized factor structure (model A) present a fit that is not fully satisfactory. The
model was therefore modified (Model B) by steps, based on indications from the post-hoc
diagnostic procedure (Modification Indices - MI). In particular, we then added the
covariance between the errors of the items IND09-SR, IND10-SR, IND11-SR, IND 12-SR.
This covariance between the errors of the variables referred to the stress recognition
dimension probably means that this factor structure reveals an autonomous sub-factor
called “stress recognition”. These modifications improved the fit between the model and
the data in a satisfactory way. The analysis of the standardized estimates of factor loadings
reveals that the estimated parameters are substantial (range between .33 and .70) and the
standard errors are acceptable (range between .03 and .16).
The sixth framework depicted by Vincent et al. (1998) is the Task one. The Cronbach's
Alpha value is 0.699 (F = 34.441; p = .000), and the corrected item-total correlation range
between .46 and .58. Since this factor has only three items, the factor structure gained from
14
the preliminary exploratory factor analysis was confirmed by the reliability analysis, and it
was not verified by a CFA.
The seventh, and final, framework depicted by Vincent et al. (1998) is the patient one. The
Cronbach's Alpha value is 0.317 (F = 1.740; p = .177), and the corrected item-total
correlation range between .00 and .30. So, the Cronbach's Alpha value for this scale is very
low. Besides, the value of Cronbach's Alpha if we delete the second item is negative (-.048),
due to a negative average covariance among items. This violates the general reliability
model assumptions. A fter checking that the item coding reveals no mistakes, we decided to
eliminate this factor for future analyses.
5.3 Building the Stock and Flow Model
Based on the CLD designed during the GBM sessions with the hospital management, a
stock and flow structure was built, with the main aim to observe the impact of the adopted
CRM policies on the hospital performance, both from a financial and non-financial
perspective. From a patient safety point of view, the system dynamics model estimates the
degree of physical impairment or disability at discharge.
According to Baker et al. (2004), to evaluate the degree of physical impairment or disability
at discharge, the physician reviewers were asked to determine, on the basis of evidence in
the medical record and their professional judgment, the degree of physical impairment
attributable to the adverse event over and above the patient’s disability from the underlying
disease on the day of discharge. Conceptually, a patient’s physical impairment or disability
at discharge, hereinafter called “patient injury rate”, can be viewed as the sum of two
different rates:
e a patient injury normal rate, which expresses the consciousness that every medical
intervention can produce a patient injury,
e a patient injury rate due to clinical error, express negative consequence for the patient
of a medical intervention, where the injuries produced could have been avoided by the
medical staff by strictly following medical procedures and protocols.
Obviously, hospitals can invest in CRM policies that could produce improved results. And,
as a direct consequence, hospitals can effectively reduce patient’s injury disability resulting
from clinical error. From a financial perspective, the system dynamics model estimates the
costs sustained by hospitals for managing the patients’ complaints; for covering legal fees,
and the increase of insurance premiums resulting from punitive damages. Following
Brennan et al. (1996), the system dynamics model contemplates an average cost for a
specific malpractice claim in accordance with the degree of the patient’s physical
impairment or disability. The following rules have been adopted about the average value of
a payment for a specific malpractice claims, and the indirect cost of legal services,
presented in Table 7.
15
Table 7. The Insurance cost dynamic with respect to the degree of physical impairment or disability
due to clinical errors.
Degree of physical impairment or disability Insurance cost | Legal consultant
increase (%) cost (€)
None 0
Minimal impairment, or recovery in 1 mo, or both 0 1500
Moderate impairment, recovery in 1-6 mo 0 5000
Moderate impairment, recovery in 6-12 mo 5 8000
Permanent impairment, degree of disability < 50% 20 15000
Permanent impairment, degree of disability > 50%; 80 20000
Death 200 50000
The rise of insurance cost, resulting from payment to patients as ordered by the National
Court of Justice, is affected by the patient’s age and social status. The actual payment for
the degree of patient’s physical disability may range between divergent minimum and
maximum values. Table 7 shows the average value as estimated by insurance professionals
and lawyers. The increase in insurance premiums, following the first compensation
sentence episode, is reported in Table 7. However, the percentage increase in insurance
premiums rates become lower with each successive episode.
6. Results
6.1 The causal loop diagram
The GMB sessions involving management of the healthcare companies participating in the
study allowed us to identify some of the main cause-effect relationships characterizing the
organizations’ system. As depicted in the Figure 1, the higher the number of people who
require a hospital treatment, all other things being equal, the greater the number of patients
requiring hospitalization to a certain healthcare company. As a consequence, the number of
treatments provided by a hospital increase as well as the number of patients cured reducing
the population to be cured (loop B;). An increase of treatments, due to a higher number of
patients, determines a rise in the number of adverse events due to clinical errors, if the
percentage of clinical errors does not change. This would worsen hospital reputation and,
hence, the number of potential patients (loop Bz).
However, when the number of treatments is augmented, hospitals would acquire more
earings which, in turn, would improve its financial standing allowing for investments in
CRM policies. These investments would reduce the liability (loop B3), and improve CRM
quality. This would lead to decreasing the number of adverse events due to clinical errors,
and would also have positive effects on the number of patients and, hence, treatments (loop
Rj). The effect of investments in CRM policies on CRM quality is not immediate, because
it is necessary for the new operating procedures to leave sediment in workers’ behavioral
16
pattems before real improvements can occur. As depicted in the loop R3;, a reduction of the
number of adverse events due to clinical errors, stemming from an improvement of CRM
quality obtained through investments in CRM policies, decreases the number of
compensation claims. This would lower the insurance premium with a positive impact on
financial revenues that can be re-invested in CRM policies. In the loop R: is described a
potential pathological phenomenon that can be caused by the national healthcare system.
Figure 1. The emerged causal loop diagram
+ Treatments
eamings
Patients RI
+ Insurance
7" ; ros \\,
Populationto gy ».. .
becured Treatments Financial
availability \
x Compensation
Patients * \, claims a
Hospital cured 53
reputation B2 4, .
\ Adverse events Investm ents
due to clinical inCRM
errors NN policies
Patients to be Re CRM om
treated again ¢ quality “+
Indeed, when adverse events due to clinical errors occur, it is possible that the same
patients have to be treated again for the same disease or different illness caused by the prior
treatment received. It is likely that these patients return to the same hospital to get a new
cure, as they may not be aware of the clinical errors and would continue trusting the same
doctors. This determines and augments the number of patients, increasing the number of
treatments and, all other things being equal, the number of adverse events due to clinical
errors (loop R2). These re-treatments bring to the hospital new earnings- in most cases the
government is paying for the medical treatments, triggering the previously described loop
Ry. Therefore, it can be concluded to a certain degree that it may be economically
advantageous for healthcare companies to commit clinical errors, if such errors do not
result in any significant negative consequence. In order to prevent such phenomenon, the
government should implement a stringent control system to verify as to why certain
patients are treated by the same hospital in a brief duration for the same or consequential
pathologies.
17
6.2 The personnel’s perceptions about the quality of the CRM in the hospital
Scale means, standard deviations, the proportion of positive scores (> 75 out of 100) and
alpha values are been estimates through an ANOVA post-hoc analysis (Tukey HSD)
presented in Table 8. The psychometric validation displayed that the coefficient alpha
ranged among the scales from .55 to .90, where the mean value range from .62 to .73 and
the standard deviation ranges from .12 to .17. The percent of positive scores (> .75 out of 1)
range from 22.5% to 53,4%. The inter-correlation between the questionnaire frameworks
has been calculated. The data demonstrate that overall frameworks are highly correlated
with one another. In fact, the factor inter-correlation ranged between .39 and .80 (overall
statistically significant at the 0.001 level, 2-tailed).
The one-way ANOVA with post-hoc test was employed to explore the difference of the
perception about the seven frameworks between the personnel of the three hospitals. The
results showed a significant difference between the personnel perceptions with regard to the
Institutional framework (F(2,203) = 36,922, p= .000), the Organization and Management
framework (F (2,203) = 41,776, p= .000), the Work Environment framework (F(2,203) =
65,179 , p= .000), the Team framework (F(2,203) = 10,072, p=.000), the Individual and
Staff Member framework (F(2,203) = 23,114, p= .000), and finally the Task framework
(F(2,203) = 43,402, p=.000).
Table 8. The ANOVA post-hoc analysis (Tukey HSD) and the percent of positive scores (=
0,75 out of 1)
95% Confidence Positive
Framework | Hospitals| N M | SD a Interval for Mean MIN | MAX | Score
‘rror| Lower Upper (%)
Bound Bound
0,80
A 128 | 0,56*| 0,14] 0,01 0,53 0,58 0,25
IST B 35| 0,74] 0,13] 0,02 0,69 0,78 0,41 0,95
€ 41| 0,72] 0,14} 0,02 0,67 0,76 0,40} 1,00
Total 204| O@2| 016) O01 0,60 064 025} 100) 225
A 128 | 0,63* | 0,12] 0,01 0,61 0,65 0,33] 0,85
ORG B 35} 0,77] 0,13} 0,02 0,73 0,82 0,50} 0,95
Cc 41| 0,79] 0,11} 0,02 0,76 0,83 0,50} 0,98
Total 204; 0,69} 014; 001 0,67 0,70 0,33) 098] 37,7
A 128] 0,55* | 0,12] 0,01 0,53 0,57 0,22} 0,82
WOR B 35| 0,77] 0,17] 0,03 0,71 0,83 0,38] 0,98
Cc 41| 0,77] 0,13} 0,02 0,73 0,81 0,36} 0,98
Total 204; 0,63} 017} 001 061 0,66 022) 098) 275
A 128 | 0,70* | 0,15} 0,01 0,67 0,72 0,24} 1,00
TEAM B 35} 0,80] 0,16} 0,03 0,74 0,85 0,36} 1,00
Cc 41| 0,79] 0,13} 0,02 0,75 0,83 0,31 1,00
Total 204| 073| 015| O01 O71 0,75 024) 1,00) 534
IND A 128 | 0,65*| 0,10] 0,01 0,63 0,67 0,38] 0,92
18
B 35 0,78| 0,14] 0,02 0,73 0,83| 0,50] 0,98
Cc 41| 0,73] 0,11| 0,02 0,70 0,77] 0,47] 0,98
Total 204| 0.69| 0,12| 0,01 0,67 0,70| 038) 098) 299
A 128 | 0,61*| 0,15] 0,01 0,58 0,63] 0,20] 0,93
TSK B 35[ 0,78] 0,15] 0,03 0,73 0,83] 0,47[ 1,00
Cc 41] 0,82| 0,13| 0,02 0,78 0,87] 0,40] 1,00
Total 204| 0,68| 0,18| 0,01 0,66 070] 020; 100) 338
The post-hoc analysis (Tukey HSD) shows that all the means referred to the Hospital C the
personnel’s perception about the CRM quality are statistically different with both the
Hospital A and Hospital B personnel’s perception (p = .000). No difference, instead, was
found between the Hospital A and B personnel’s perception (see Table 8).
6.3 The clinical adverse events of the three Hospitals
The data shown in the Table 9, present evidence on the number of compensation claims for
each hospital. Further, it shows that the highest percentage of compensation claim occurs in
Hospital C, while the lowest percentage happens in Hospital A. These data can be
explained by two main factors: first, the number of interventions made by a hospital per
year, and second, the technical difficulty of these interventions. In fact, it is well known
that a chirurgical intervention is riskier than the medical ones.
Table 9. Number of compensation claims (2008-2010) for hospital and incidence of compensation
claim per number of treatments.
Number of Number of overall treatments (with the incidence of
Hospital | Compensation Claims compensation claims for treatments)
2008 | 2009 | 2010 2008 2009 2010
A 0 0 0 |1016 (0%) | 845 (0%) | 952 (0%)
B 1 2 4 | 5298 (0,019%)| 5790 (0,035%) | 6073 (0,066%)
Cc 8 14 4 |9170 (0,087%)| 8413 (0,166%) | 7958 (0,050%)
It is important to note that a compensation claim does not necessarily mean that the related
adverse event is due to a clinical error made by hospital personnel. However, the data give
us a good indication about the dimension of the clinical risk that occurs in these hospitals.
19
6.4 The Stock and Flow Model
Based on the CLD described earlier, a stock and flow structure has been developed, with
the aim of observing the potential impact of some CRM policies on the performance of the
hospitals studied in this research. Figure 2 shows a section of the stock and flow structure
describing the hospitalization processes. The first stock (on the figure’s left), represents the
number of people of a specific population- a town- affected by some relevant clinical
events that require hospital treatments. As a consequence, these people could potentially
become hospital’s in-patients. For this reason, from the stock called “Population Affected
by Clinical Event” four different out-flows representing the above- mentioned alternatives.
As Figure 2 shows, from each of the three patients’ stock (patients waiting for ER, patients
treated by ER, Hospital Inpatients) departs two main different flows: the first one represents
the progression of the hospitalization process, while the second one depicts the negative
consequence of every medical activity, that is, the patient injury rate. Conceptually, the
patient injury rate can be viewed as the sum of two different rates: a patient injury normal
rate, a patient injury rate due to clinical error.
Figure 2. Stock and Flow Structure related to the Hospitalization Processes and the Effects of Clinical Risk Management on patients’ safety.
Patients Injuried
During Waiting Time
for Treatment in ER
Patients Injury Rate during
Waiting Time for
Treatment in ER due to
Clinical Error
Population Dead before
ER Treatment
Effect of cR Patients Injury Rate during
‘Treatment in H due to
Clinical Error
Population Went to
Other Hospitals
Patients Injuried during
Treatment in ER
Patients Injuried during
‘Treatment in Hospital
Paes Injury Rate during
Population Patients Injury ent in ER due to
Dying before Rate During Clinical Enor
ER Treatment Waiting Time
oe aie Bereiney Pellentsidduy os Patients Injury Patients Injury
al in El Rate During Rate During (> Normal Rate
| Population +) during
CXeooing to other reatment in Troatmertin Treatment in during
Hospitals Population ER Hospital Treatment in H
Affected by
Clinical Event Hales Distietged
pingiokt from Hospital
JI [1
LJ cy LS 2) 5 i = =
Population Affected\~ Patients in ER se Patients Treated in ER Hospital Inpatients
by Clinical Event waiting for Treatment Patients Patients Patients Effectively
Treatment in Treated in ER Treated in- Treated
and Admitted Hospital Patients Rate
as Inpatients
O
— Effect of CR
Rateot I) rete inER Cy Fameworks
Population and 7, on Patient
Affected
Discharged Injury
Effect of CRM
on Hospital
Effectiveness % Patients
Effectively
Treated in
Hospital
Ineffecvely
reated
pais Rate
20
The “patient injury rate” variable is affected by the clinical risk factors as defined by
Vincent et. al (1998). Nevertheless, a deeper analysis shows that just five of them can
directly affect the clinical practice as managed by the medical staff of a hospital. In fact,
both the “institutional” and “patient” frameworks (that were deleted as described in the
previous section) seem to refer to macro and micro scenarios respectively, while the other
five frameworks (Organization and Management; Work environment; Team; Individual
staff member, Task) refer to factors directly related to medical practice- which can be
improved through CRM policies.
It is important to note here that one of the main issues in building SD models is defining the
model boundaries. Such boundaries are set according to the main aims of the research.
Since, our research is a first attempt of applying SD to CRM from an organizational level
of analysis; we decided to build a simple but sensible SD model. Furthermore, this was also
acknowledged below in the section on “Conclusions, Implications, Limitations, and Future
Directions”. The core point is that the data presented here are based on the preliminary
results of a multi-phase research project being undertaken, hence, reflecting the research
exploratory nature. Despite the fact that there might be additional processes that could be
included in the SD model, these cannot be treated in this paper without renouncing the
simplicity of the model which, at this stage of the analysis, is fundamental to describing our
research results. In short, these additional processes will be addressed in a future follow up
paper.
7. Scenario Analysis
Based on the qualitative and quantitative analysis of the system structure outlined above,
four alternative policies have been compared in order to evaluate their potential effects on
the company’s performance.
As depicted in Table 5, these policies differ by the degree of improvement of the quality of
clinical risk contributory factors, which ranges from 0 (very low CRM quality) to 1 (very
his CRM quality):
in the base-run scenario, the hospitals’ policy is aimed at maintaining the current level
of CRM quality;
e inscenario 1, it is hypothesized that the hospitals decide to cut the actual investment in
CRM;
e in scenario 2, the hospitals plan to increase the CRM quality by 5% with respect to the
current level;
e in scenario 3, the hospitals invest in CRM interventions in order to augment its quality
by 10% with respect to the current level.
For the scenario analysis, a six- year time horizon is considered. The first two years (2009
and 2010) of the simulation runs are aimed at replicating the past performance of the
participating hospitals. The remaining four years, from 2011 to 2014, are intended to
forecast the potential impacts of the examined CRM policies on hospital performance,
financial and non-financial. The system dynamics model contains a cost function related to
21
CRM investments showing that a higher degree of CRM quality improvement indicates a
higher amount of money invested in CRM policies.
Insurance costs are also included as described in the CLD in section 6.1. However, because
of the complexity of the organizations studied, not all the costs and revenues are included in
the system dynamics model. Included are only revenues directly connected to CRM
policies and general costs - which are figured out as a percentage of the revenue volume.
As a consequence, the reference behavior reproduction was not possible for the net earning
variable. The following figures (figures 4, 5, 6) show the simulation results of the four
different scenarios. For brevity, the graphs report the results of just one of the three
hospitals. However, the simulation results of the three hospitals studied present very similar
behavioral patterns and, hence, the following scenario analysis can be representative of the
three healthcare companies.
In the base-run scenario, it is assumed that the hospitals maintained the initial level of CRM
quality during the simulation time. This policy could imply some minor incidents due to
clinical errors that may determine both a worsening of hospital reputation and an increase
of insurance costs, which negatively affect economic results. As the reader will note from
the behavior of “treatment earnings” compared to the dynamics of annual in-patients,
during 2010, the three hospitals experienced a reduction of the average revenue per patient.
This reduction was due to a decrease in the amount of funds reimbursed by the regional
government to the hospitals for the patient treatments and to a different treatments mix
required by patients. This reduction was also the main cause responsible for the worsening
of the economic results represented in the simulation results.
In scenario 1 (see Figure 3), the hospitals decide to abandon the CRM practices starting in
2011. The analysis of this scenario is aimed at evaluating the response of the system
dynamic model to such an “extreme” policy. This decision would determine a progressive
deterioration of CRM quality (indicated as “average value CR contributory factors”), due to
obsolescence process of medical tools and practices, which would lead to an increase of
clinical errors, a worsening of hospital reputation and, hence, a reduction of the number of
people going to the hospitals (indicated as “annual inpatients”).
Because of the higher number of clinical errors, hospitals would experience an increase of
the “% of treated patients affected by new clinical events” and of the “% of overall
complaints”. The reduction of treatments eamings, is due to a lesser number of patients,
and the increase of insurance costs, due to the higher number of compensation claims,
which would lead to a reduction of company “net earnings”.
In scenario 2 (see Figure 4), the hospitals decide, starting in 2010, to improve the CRM
quality by 5% with respect to their current level. As shown in the simulation results, this
decision could produce a positive effect on all the previously examined performance
indicators. The comparison between the base run and scenario 2 shows that an investment
in CRM policies would bring, in the medium term, higher net earnings. In fact, after an
initial reduction of the net earnings due to the investment costs, the economic results would
improve because of a higher number of patients and a lower number of compensation
claims, and hence, lower insurance costs.
22
Figure 3. Base Run (Reference) & Scenario 1 (Current) for Hospital B
‘Annual Inpatients Hospital Reputation
people/ve i
. [eee
\ oy
‘ \
Treatment Earnings ‘Average Value GR Frameworks
civ
Fearent | |, |Feurrent
11.000. , |
L | fi }
10.000. t °
Tnaurence costs Net Earnings
ce civ
waa Fez] t Ex]
‘Number of patients compensated by Insurance for degree of injury
ime 01/01/2010 | 04/01/2031 | 01/01/2012 | 01/01/2013 | 01/01/2014 | 01/01/2035
[femporary Disability forless than Imo 00 00} 00} 00} 00) 00)
fremporary Disability for i to 6 mo 00) oy oy 00) 0) 007
[Temporary Disability for more than 6 mo 00) oy 00) 0) 0) ‘007
[Permanent Disebility for less than 50% 00} oy oy oy oy 007
[permanent Disability for more than 50% 00 a0y 00 00 00y 00
[be ‘007 oy 007 oy oy 207
[eTemporary Dissbilty for less than = mo 007 007 007 00). 00). 0,
é 00) 00) 00 00). 00). 0)
me 00) 00) 00) 00. 00). 0)
0% 00) 00) 00 00) 00), 07
50% 007 007 00) 00). 00). 007
Froeath 00) 007 00 00). 00) a0y
i
In scenario 3 (see Figure 5), from 2010, the hospitals increase the investments in CRM in
order to improve the CRM quality by 10% compared to their current level. The comparison
between the scenarios 2 and 3 shows that the higher investment in CRM policies
represented in scenario 3 would determine a better performance, with respect to scenario 2,
in terms of patient safety. However, the higher investment costs required by this policy
would not be counterbalanced by higher revenues and insurance costs savings. As a result,
this policy would produce a worsening of the net earnings. Consequently, the management
could prefer the scenario 2, even though this would imply a lower level of CRM quality.
The adoption of the policy described in scenario 3 could be incentivized by the government
through offering healthcare companies’ tax exemptions or other financial aids.
23
Figure 4. Base Run (Reference) & Scenario 2 (Current) for Hospital B
inpatients Hospital Reputation
people/yr
2
i
“Treatment Earnings:
pon Caeser = =
‘Number of patients compensated by Insurance for degree of Injury
3 (01/01/2010 | 01/01/2033 | 01/01/2012 | 01/01/2013 | 03/03/2034 | 01/01/2015
007 3,00) 00}
00) 3,00) 00}
00} 3,00) 00}
00} 3,00) ‘00y
00} 3.00) 00}
00) 3,00) 00}
00} 00} 00}
00} 00} 3,00
00} ‘00y ‘00>
00) 00} 00}
00} 00} 00}
00} 00} 00}
rr
At this stage, we would like to comment on why we had to collapse the different factors
measured in the questionnaire into a single variable that represents CRM quality.
Coherently with the choice about the model boundaries definition, we decided to represent
the different CRM factors by a proxy variable that takes into account their average values.
This decision emanated from the lack, in the literature, of data about the specific values
describing the contribution of each factor to the overall variance of CRM quality. This
decision is also based on three premises: First, there is an explicit acknowledgement in the
literature about the ultra-complexity of the subject matter. As noted earlier, recent work
suggests that despite the multitude of initiatives, programs, systems, and tools that can be
viewed as elements of CRM; there is a lack of knowledge concerning their implementation
in hospitals (Briner, 2010). Second, we could have tried to estimate the value of these
parameters; but this is so complex that will require several researchers focused on this
specific topic. Third, this decision cannot be judged as a signal of disjunction between
questionnaire and SD pieces, but merely as the inevitable consequence of the lack of
empirical data about the value of these parameters.
24
Figure 5. Scenario 2 (Reference) & Scenario 3 (Current) for Hospital B
“Annual tnpationts Hospital Reputation
people/ve 1
° Feurene || |2, Feurrene
” i
on |
"Treatment Earnings "Average Value CR Frameworks
ene x
13.000. — | |
Insurance Costs Net Earnings
ciyr cyye
at || |sse. !
of injury
[ Time (03/03/2010 | 047 01/03/2012 | 03/ [ox7ox72015
[Temporary Disability for tess than 1 mo 5,00 00 00
[Temporary Disebility for 4 to 6 mo 5,007 00 ‘007
[Temporary Disability for more than 6 mo 5,00 5,007 07
iP for less than 50% 5,00 00 5,00
[perme for more than 50% 5,00 5,00) 00}
[pesth 5,007 00 005
[stempar ime 5,007 00 007
[Temporary Disability for 4 to 6 mo 5,00 3,00) 00}
[-Temporary Disability for more than 6 mo 5,007 3,00) 007
[SPermanent Disability for ess than 50% 5,007 5,00) 007
[pre ‘5096 5,00 3,00) ‘007
[Fbeath 5,007 9,00) "00
A summary of our analytical findings can be discussed as follows: In the scenario analysis
section, we explained that the performance indicators monitored to measure the potential
impact of CRM policies on financial and non- financial results present similar behavioral
pattems in the three hospitals. However, we need to take some considerations into account-
as noted below. The questionnaire results show that the two private hospitals present a
higher level of CRM quality compared to the third public hospital. This is confirmed both
from the personnel perception questionnaire results and by the number of compensation
claims received by the hospitals. This difference in the CRM quality can be explained by
the higher consciousness in the private hospitals’ personnel about the relevant impact of a
serious clinical adverse event on the financial stability of the healthcare organization they
work for.
In other words, the adverse event is perceived not only as a professional failure but also as a
potential threat to their workplace. Moreover, the financial support from the regional
government to the public hospital reduces the perception of the economic impact of an
adverse event and, hence, personnel’s attitude towards CRM practices. For this reason, the
public hospital presents greater margins of CRM quality improvement that require more
investments in terms of both medical equipment and personnel training and motivation
25
programs. Therefore, it would be useful to develop CRM laboratories linking both private
and public hospitals aimed at creating communities of practice- where sharing best
practices and fostering organizational learning.
8. Conclusions, Implications, Limitations, and Future Directions
We believe that our study has a broad appeal for researchers. Globally, while patient’s
safety and the quality of care declined, the aging population has substantially increased, so
did the skyrocketing cost of healthcare. The complexity of the profit maximization
phenomenon at the expense of patient’s safety has become a pressing issue requiring
remedial action. Greater awareness and concerns over the future of health care, in part,
steered us to study this thorny universal problem.
Our research findings suggest that it would be feasible for these companies, to invest
financial resources to achieve a certain level of CRM quality. However, according to the
simulation results, increasing the level of CRM quality could be at the expense of financial
bottom-line. Also, the financial costs related to additional investments may be higher than
the marginal benefits the companies could gain. As a consequence, if the national
healthcare system aims to accomplish higher level of CRM quality than the healthcare
companies’ “breakeven” threshold, it should make these investments more feasible and
sustainable through tax exemption policies, or other financial aid measures.
Considering the difficulty of having access to any hospital for the purposes of academic
research and data collection, we believe that gaining access to the three hospitals represents
a significant progress toward further exploration and understanding of CRM. Also, the
behavioral similarity among the three hospitals supports the assumption that our research
results can be extended to other hospitals.
The development of this research project was aimed at comparing different combinations of
CRM investments, and an investigation of their impacts on healthcare companies’ cash
flow. Similar to other studies, our research has a number of limitations, that can be
addressed in future research. One limitation is that the data presented here are based on the
preliminary results of a multi-phase research project being undertaken, hence, reflecting the
research exploratory nature.
A second limitation is represented by the model boundaries. Indeed, some system variables
that may influence the performance of the hospitals have not been considered, such as the
role of the Regional Healthcare Administration, the role of Unions, patients associations,
etc. However, these external influences were not considered essential for the goal of the
analysis, since they are beyond the hospitals’ management control. Nevertheless, despite
these limitations, this paper delivers results with implications for the application of system
dynamics methodology to CRM.
A third limitation relates to the fact that we had to represent the different CRM factors by a
proxy variable that takes into account their average values. As noted earlier, while this
decision emanated from the lack of empirical data in the literature about the value of these
parameters, it, nevertheless, represents a research limitation. Finally, the ultra-complexity
of the subject matter; and the paucity of quantitative research-based studies in the literature,
was an issue. However, the scarcity of material on the subject matter tumed out to be a two-
26
edge sword: on the one hand; it is a constraint in dictating the “exploratory” nature of the
study; and yet; on the other hand; it enhances the study’s contribution to the field via
charting an “unexplored” path to learn more about the elements of a paradigm for
navigating an “underdeveloped” research stream.
In sum, the academic and practical implications of our research can be summarized as
follows. CRM practices do not take into account the cost elements and their influence on
personnel management. Such conditions may foster wrong evaluations leading to the
postponement of the introduction of procedures aimed at improving the healthcare
companies’ risk profile. Furthermore, healthcare companies’ managements experience
serious difficulty in quantifying the benefits gained from investments aimed at reducing the
clinical risk. Therefore, it is necessary to provide healthcare companies’ managements with
a systemic and multi-dimensional approach that supports cost-benefit analysis of CRM
policies.
The international nature of the study may open the door and stimulate other researchers to
undertake comparative research studies in cross-cultural settings. We are calling on future
researchers to investigate and carry-on experimentation with system dynamics as a
significant paradigm for uncovering real performance indicators of operations management
effectiveness - not only from the perspective of organizational internal resources, but also
the contextual drivers and constraints imposed by the organization’s environments.
References
Arbuckle, J.L. ; Wothke, W. 1999. Amos 4.0 user's guide. Chicago, IL : SmallWaters.
Baker, G. R., Norton, P. G., Flintoft, V., Blais, R., Brown, A., Cox, J., et al., 2004. The
Canadian adverse events study: The incidence of adverse events among hospital patients
in Canada. Journal of Canadian Medical Association 170 (11), 1678-1686.
Barach,P.,Small,S.D. 2000.Reporting and preventing medical mishaps:lessons from non-
medical near miss reporting systems. British Medical Journal. 320(7237):759-763.
Bendoly, E., Croson, R., Goncalves, P., Schultz, K. 2010. Bodies of knowledge for research
in behavioral operations. Journal of Operations Management. 19(4):434-452.
Bentler, P., and Bonett. D. (1980). Significance tests and goodness of fit in the analysis of
covariances structures. Psychological Bulletin, 88, 588-606.
Bentler, 1990. Comparative fit indexes in structural model. Psychological Bulletin, 107,
238-246.
Berg, H.P., 2010. Risk management: procedures, methods and experiences. RT&A 1, 79-95.
Brennan et al. (1996)
Briner, M., Kessler,O. Pfeiffer, Y., Wehner,T., Manser, T. 2010.Assessing hospital clinical
risk management: Development of a monitoring instrument. BMC Health Services
Research. 10:337: 1-13.
Cavallin, M., Chiarelli, P., Genduso, G., Trucco P., 2006. Beyond Error Reporting:
Applying Control Charts in Clinical Risk Monitoring. Journal of Medicine and the
Person 4 (2), 62-69.
27
Chiozza, M.L., Plebani,M. 2006. Clinical governance: from clinical risk management to
continuous quality improvement. Clinical Chemistry and Laboratory
Medicine.44(6):694-698.
Cook, R., Rasmussen, J., 2005. “Going Solid”: a model of systems dynamics and
consequences of patient safety. Quality and Safety in Healthcare 14, 130-34.
Corrigan, J.M., Donaldson, M.S., Kohn, L.T., Maguire, S.K., Pike, K.C. 2001.Crossing the
quality chasm: a new health system for the 21% century. National Academy Press,
Washington D.C.
Dangerfield, B.C., 1999. System dynamics applications to European healthcare issues.
Journal of the Operational Research Society 50, 345-353.
Homer, J.B., Hirsch, G., 2006. System Dynamics Modeling for Public Health: Background
and Opportunities. A merican Journal of Public Health 96 (3), 452-458.
Joint Commission, 2008. National patient safety goals: accreditation program: hospital. In
The Joint Commission on Accreditation of Healthcare Organizations. 30.
Leape, L.L., 2002. Reporting of adverse events. The New England journal of Medicine. 347
(20): 1633-1638.
Lee, R.C., Cooke, D.L., Richards, M., 2009. A system analysis of a suboptimal surgical
experience. Patient Safety in Surgery 3 (1), 1-9.
Misson, J.C., 2001. A review of clinical risk management. 2001. Journal of Quality in
Clinical Practice. 21(4): 131-134.
Nunnally, J., 1978. Psychometric theory. McGraw-Hill, New Y ork.
Reason, J. 2000. Human error: models and management. British Medical journal. 320:768-
770.
Repenning, N.P., 2001. Understanding fire fighting in new product development. Journal of
Product Innovation Management. 18(5): 285-300.
Secker-Walker, J., Taylor-Adams, S. 2001.Clinical incident reporting. In Clinical Risk
Management: Enhancing Patient Safety .2" edition. Edited by Vincent, C. BMJ Books:
419-438. London.
Sexton, J.B., Helmreich, R.L., Neilands, T.B., Rowan, K., Vella, K., Boyden, J., Roberts,
P.R., Thomas, EJ., 2006 (a). The Safety Attitudes Questionnaire: psychometric
properties, benchmarking data, and emerging research. BMC Health Services Research 6,
44,
Sexton, B., Makary, M., Tersigni, A., Pryor, D., Hendrich, A., Thomas, E., Holzmuellerm,
C.G., Knight, A.P., Wu, Y., Pronovost, PJ., 2006 (b). Teamwork in the operating room.
Anaesthesiology 105(5), 877-884.
Sexton, J., Holzmueller, C., Pronovost, P., Thomas, E., McFerran, S., Nunes, J., Thompson,
D.A., Knight, A.P., Penning, D.H., Fox, H.E., 2006 (c). Variation in caregiver
perceptions of teamwork climate in labor and delivery units. Journal of perinatology 26,
463-470.
Sorra, J.S., Nieva, V.F., 2004. Hospital Survey on Patient Safety Culture. AHRQ
Publication No. 04-0041. Rockville, MD: Agency for Healthcare Research and Quality.
Sterman, J.D., 1989. Modeling managerial behavior: misperceptions of feedback in a
dynamic decision experiment. Management Science, 35:321-339.
Sterman, J.D., 2001. System dynamics modeling: tools for learning in a complex world.
California Management Review.43(4): 8-25.
28
Tréek, D., 2008. Using System Dynamics for Managing Risks in Information Systems.
WSEAS Transactions on Information Science and Applications 5 (2), 175-180.
Trucco, P., Cavallin, M., 2006. A quantitative approach to clinical risk assessment: The
CREA method. Safety Science 44, 491-513.
Vennix, J. A. M., Andersen D. F., Richardson G. P., Rohrbaugh J. 1992. Model-Building
for Group Decision Support - Issues and Altematives in Knowledge Elicitation.
European Joumal of Operational Research, Vol. 59, No. 1., 28-41.
Vennix, J. A. M., Andersen D. F., Richardson G. P.,. 1997. Group model building, art, and
science — Foreword. System Dynamics Review, Vol. 13, No. 2, 103-106.
Vincent, C., 2006. Patient Safety. Elsevier Churchill Livingstone, Edinburgh etc.
Vincent, C., Taylor-Adams, S., Stanhope, N., 1998. A framework for analyzing risk and
safety in clinical medicine. British Medical Journal 316, 1154- 1157.
Vincent, C.A., 2003. Understanding and responding to adverse events. New England
Joumal of Medicine, 348 (11), 1051-1056.
Walshe, K. 2001. The development of clinical risk management. In Clinical risk
management: enhancing patient safety, 2" edition: 45-60. edited by Vincent, C. BMJ
Books, London.
Wolstenholme, E.F., 1999. A patient flow perspective of UK Health Services: exploring the
case for new “intermediate care” initiatives. System Dynamics Review 15, 253-271.
Young, PC., Tomski, M. 2002. An introduction to risk management. Physical medicine and
rehabilitation Clinics of North America.13 (2):225-246.
