System Dynamics and Tobacco Treatment 1
USING DYNAMICS MODELING TO PROMOTE EFFECTIVE TOBACCO TREATMENT
PRACTICES IN COMMUNITY-BASED PRIMARY CARE SETTINGS
David Lounsbury, PhD?, Ralph Levine, PhD? and J amie Ostroff, PhD?
1Department of Epidemiology and Population Health, Albert Einstein College of Medicine,
Yeshiva University, Bronx, NY USA
Department of Psychology & Department of Community, Agriculture, Recreation, and Resource
Studies, Michigan State University, East Lansing, MI USA
3Department of Psychiatry and Behavioral Science, Memorial Sloan-Kettering Cancer Center,
New York, NY USA
Funded by the National Institutes of Health/National Institute on Drug Abuse (R03 DA022278-
01A1)
ABSTRACT
This paper describes formative field research to develop and test the utility of a system
dynamics modeling intervention intended to promote evidence-based tobacco treatment
practices in community-based primary care settings. Brief counseling interventions by primary
care providers have been shown to effectively promote tobacco cessation among patients who
smoke, yet many physicians are inconsistent in the way they intervene with their patients. Too
little time, poor training, lack of third-party reimbursement, competing clinical problems, and the
belief that their patients are not able to change explain, in part, why some physicians do not
adhere to evidence-based guidelines for treating tobacco use and dependence. Via a protocol
for conducting on-site office visits to small primary care practices located in medically
underserved urban communities, we tested the hypothesis that providers exposed to the
simulation tool would demonstrate better understanding and progress towards full
implementation of the US Public Health Service Guideline for Treating Tobacco Use and
Dependence. Results indicate that simulated output that reflects the dynamics of providers’
unique practice environment is associated with stronger behavioral intent than other forms of
feedback information, such as patient chart reviews.
INTRODUCTION
In this study, we examined the utility of system dynamics modeling as a means to develop a
simulation tool to foster understanding about how to improve implementation of the PHS
Guideline for Treating Tobacco Use and Dependence in primary care practices. We examine
the dynamics of tobacco treatment in small primary care practices (1-5 physicians per practice).
System dynamics models have been developed to study community and population impacts of
varied public health problems and policies (Homer and Hirsch 2006), including tobacco policies
(Feenstra, Hamberg-van Reenen et al. 2005; Levy, Bales et al. 2005; Cavana and Clifford
2006). However, these works have not modeled the dynamics of individual practice settings, nor
have they used system dynamics models to directly educate and influence physician practices.
In the first stage of this research, we worked collaboratively with an expert advisory group to
construct a working system dynamics model of the simulation tool. This version of the simulation
tool is now being subjected to a formative assessment in an academic detailing intervention with
a small sample of community-based primary care practices. Our formative assessment
examines: (1) feasibility and acceptability of using the simulation tool in an academic detailing
intervention, (2) changes in individual provider attitudes about and practices in tobacco
System Dynamics and Tobacco Treatment 2
treatment, (3) and implementation of new or improved office systems to improve tobacco
treatment at the practice level. We hypothesized that system dynamics modeling of the practice
environment will promote deeper understanding of and greater impetus to implement the PHS
Guideline.
The specific aims of this project were as follows:
Aim 1. To develop a system dynamics simulation tool as a decision aid for promoting in-depth
understanding about how best to implement currently recommended the clinical guidelines for
the treatment of tobacco use and dependence in community-based primary care settings;
Aim 2. To conduct a formative assessment of the simulation tool, delivered as part of an
academic detailing intervention to a cohort of small primary care practices in racially and
ethnically diverse, urban communities with high rates of smoking and poverty.
BACKGROUND AND SIGNIFICANCE
The Efficacy of Academic Detailing to Change Provider Practices. Brief counseling
intervention by primary care providers has been shown to effectively promote tobacco use
cessation, yet many physicians do not consistently adhere to this practice for all patients at each
appointment (Greco and Eisenberg 1993; Davis and Taylor-Vaisey 1995; Goldstein, DePue et al.
1998; Goldstein, Niaura et al. 2003). Significant barriers exist that can interfere with clinicians’
assessment and treatment of smokers. Many clinicians lack knowledge about how to identify
smokers quickly and easily, which treatments are efficacious, how treatments can be delivered,
and the relative efficacies of different treatments (Orleans 1993). Even if clinical knowledge is
strong, many physicians do not consistently use this intervention. Primary care physicians are
more likely to report counseling patients about smoking cessation than other medical
professionals, but are not more likely to refer them for counseling (Meredith, Yano et al. 2005).
Too little time, poor training, lack of third-party reimbursement, competing clinical problems, and
the belief that their patients are not able to change also explain why some physicians do not
adhere to the guideline (Glynn and Manley 1989; Cabana, Rand etal. 1999; Adsit, Fraser et al.
2005).
Academic detailing interventions typically involved multiple components, including provision of
written materials and sample supplies, didactic training, auditing (with feedback), ‘reminder’
systems, and one or more office-based consultations (Soumerai and Avorn 1990; Goldstein,
Niaura et al. 2003; Gandjour and Lauterbach 2005). A recent Cochrane review by O’Brien and
colleagues (O'Brien, Oxman et al. 2005) examined the effectiveness of educational outreach
visits, or academic detailing, to promote changes in medical and health care provider practices.
In 13 of 18 randomized trials examined, the targeted provider behavior was prescribing
practices. Three studies addressed preventive practices, including brief counseling for smoking
cessation (Avorn, Soumerai et al. 1992; Berings, Blondeel et al. 1994). Collectively, these
efforts help detailers establish a rapport with providers that, in turn, can generate effective
change in practices.
Although positive outcomes were observed in all studies in the review, interventions that
provided one or more of the following, including individual instruction, used audit and feedback
strategies, incorporated review by peers, and that successfully integrated ‘reminder’ systems,
were among the most effective for medical professionals (Steele, Fors et al. 1989; Dietrich,
O'Connor et al. 1992) (Wensing and Grol 1994; Yano, Fink et al. 1995; Weissman, Allison et al.
1999; Andrews, Tingen et al. 2001; Kiefe, Allison et al. 2001). Results did not reveal a clear
relationship between the number of office visits by detailers and impact on the provider,
although it was noted that interventions with as few as one or two visits had positive effects.
System Dynamics and Tobacco Treatment 3
Overall, academic detailing appears to be a promising way to change provider behaviors,
especially when the behavior was prescribing medications. However, additional research on
interventions intended to change preventive practices, including tobacco treatment practices
(Goldstein, Niaura et al. 2003), is needed. Although dissemination-only strategies (e.g.,
conferences and mailings) always demonstrated smaller effects than interventions involving
outreach visits or peer review, such interventions had varying levels of effective impact (Oxman,
Thomson et al. 1995).
We believe that the system dynamics modeling approach has the potential to transform how
clinical guidelines and scientific reviews are disseminated to busy professionals. A well-
designed simulation tool could greatly accelerate the rapport-building process between detailers
and providers. We hypothesize that the capability to automatically simulate the dynamics of
implementing practice changes during the course of either a didactic training session and/or an
office-based consultation would help an academic detailer quickly learn about a provider's
practice environment and help providers make practice-specific, cost-effective decisions about
how to most efficiently and rapidly attain (and/or sustain) evidence-based standards of tobacco
treatment for their patients. A tool with this capability would allow for quick comparison of
alternative ways of changing office procedures by generating scenarios that simulate different
combinations of role-sharing or resource exchange.
The system dynamics simulation tool we envision would be able to generate customized output,
on the spot, in the form of easy-to-read behavior-over-time charts and data tables. Results
would give a dynamic picture of demand on providers as well as patient outcomes over a
specified period of time. It could show how, for example, adding tobacco treatment time during
office visits will impact wait times over the course of a single day, or how combination NRT
impacts relapse rates for heavy smokers over a three year period. More generally, our
completed simulation tool would help providers answer critical questions such as: Which staff
members should (and can) be involved in the practice’s tobacco treatment strategies? How
effective are minimal interventions, such as clinician advice to stop smoking, for our patients, or
are more intensive interventions required? How does the duration of an intervention in number
of treatment sessions or in total face-to-face contact time substantially influence efficacy for our
patients? How much counseling time can we allocate during an office visit? What are the short-
term and long-term costs of not effectively treating tobacco use, to the practice and to our
patients? Which pharmacologic interventions will be easiest for our patients to adhere to and
may lead to greater patient contact? How many times do patients relapse before they quit for
good?
We expect that the capacity to address these types of questions with the simulation tool will help
primary care providers visualize the implementation of various features of the tobacco treatment
guidelines. In turn, we expect that providers will more quickly identify the mechanisms that will
drive effective tobacco treatment in their own practices.
System Dynamics Modeling to Foster Practice Change. Experts in change management and
health care quality improvement recommend that “rapid change” can be achieved by: (1)
employing strategies that break large goals into a series of smaller goals (i.e., ‘small wins’); (2)
fostering interdisciplinary workgroups to intervene at multiple points in the process of care; and
(3) conducting a series of pilot studies or projects to test and establish new practices. Through
our planned intervention, we attempt to use our system dynamics model to bring together two
types of knowledge to the problem of tobacco treatment in primary care: Professional
knowledge (i.e., evidence-based medicine) and knowledge for improvement (i.e., a systematic
approach to achieving change for improved care; informative feedback) (Headrick 2000).
System Dynamics and Tobacco Treatment 4
In public health and health services research, system dynamics modeling and other simulation
techniques have typically studied population-level problems (Lounsbury 2002; Feenstra,
Hamberg-van Reenen et al. 2005; Levy, Bales et al. 2005; Bar-Yam 2006; Homer and Hirsch
2006). This work has examined cost-effectiveness of new treatment modalities (i.e.,
pharmacotherapies) (Halpern, Khan et al. 2000), new public resources (e.g., state administered
tobacco ‘quit lines’) (Bentz, Bayley et al. 2006), as well as implementation of the PHS Guideline
itself (Cromwell, Bartosch et al. 1997; Torrijos and Glantz 2006). These studies often use cost-
benefit analytic techniques to assess cost of life-year, or quality-adjusted-life-year (QALY),
saved, which have provided useful, though highly variable results across studies. A recent
meta-analytic review of economic evaluations of smoking cessation determined that cost-
effectiveness ratios in such policy studies were huge (ranging from 120% to 5600%) (Ronckers,
Groot et al. 2005).
System dynamics modeling is a method that can help researchers and policy makers better
understand why such variability exists. A recent policy study that used system dynamics to
study the long term impact of implementing an excise tax on tobacco is a case in point. The
authors used system dynamics to answer questions related to price, tobacco sales, government
revenues, and smoking prevention (Cavana and Clifford 2006). The researchers reported that
policy analysts engaged in the study found the model useful and exciting, and well-suited to
society-level policy questions. Collaborating in the model-building process gave them insight
into the structure behind the processes, and helped them understand how a specific policy aim
could be achieved, or not. To our knowledge, system dynamics has not yet been applied to
dynamics modeling of tobacco treatment at the level of the primary care practice (Homer and
Hirsch 2006), though one simulation study, conducted in The Netherlands, examined the cost-
effectiveness of face-to-face smoking cessation interventions by general practitioners (Feenstra,
Hamberg-van Reenen et al. 2005; Homer and Hirsch 2006).
METHODS
Participating Practices. Twenty-five community based practices were recruited to the study.
On average, these practices are staffed by two full-time primary care providers. The largest
practice included five providers. The mean number of patient visits per week was 125 (minimum
30 per week; maximum 300 per week). Smoking prevalence, based on initial chart review data
collected from each practice, was 18% (minimum 5%; maximum 33%). The average co-pay was
$18.84 per visit (range $5 to $50), with an estimated average patient visit bill being $144 (range
$50 to $500). A high proportion of patients were covered by Medicare and/or Medicaid in most
practices. A substantial amount of time and effort was expended to recruit these practices (see
Table 1). A total of 196 Queens-based practices were approached in order to secure the
participation of 25 sites over recruitment period of 14 months. Recruitment efforts were focused
on practices affiliated with MetroP lus, a low or no-cost health insurance provider to eligible
people living in Manhattan, Brooklyn, the Bronx and Queens, through a variety of F ederally-
backed Medicaid and Medicare programs.
Participating providers’ awareness and use of patients support services, including the New York
State Quit Line, the New York State Fax-to-Quit Service, or of local smoking cessation support
groups, was limited (see Table 2). None of the participating practices had formal tobacco
treatment policies in place; similarly, none had a designated ‘tobacco cessation champion’
available to patients who smoked.
System Dynamics and Tobacco Treatment 5
a] Table Primary Care
Practice Recruitment Results
Targeted practices (Queens, NYC) 276 100%
Eligible MetroPlus Contacts 161 58%
Ineligible MetroPlus Contacts 80 29%
Cold calls 35 13%
Practice approached 196 100%
Visited; pending recruitment 107 55%
Recruited’ 25 13%
Refused” 38 19%
Excluded’ 13 7%
Unreachable* 13 7%
* includes 4 cold call practices.
2 not interested, too busy, few patients who smoke.
3 already a partner to Queens Quits.
4 could not locate, closed for business, bad address.
Table 2 - Patient Support Services and Pharmacotherapy Prescription Practices
NYS Quit Line 74% 13% 17%
NYS Fax-to-Quit 44% 4% 17%
Cessation support group - QHC 44% 0% 11%
Cessation support group - EHC 44% 0% 17%
Nicotine patch 61% 57% 44% 30%
Other NRT (Gums/Lozenges/Inhaler) 48% 48% 35% 22%
Zyban/Bupropion 52% 52% 44% 26%
Chantix/Varenicline 52% 39% 44% 30%
Model Development. The system dynamics method we have applied to develop the simulation
tool is consistent with standard procedures and techniques described in texts by seminal system
dynamicists such as Randers, Richardson, and Sterman (Randers 1980; Richardson and Pugh
Ill 1981; Sterman 2000). All model development has been conducted using Vensim (Ventana
Systems, Harvard, MA). Our finalized simulation tool is a theoretical representation of the
dynamics of a single primary care office visit with a patient who uses tobacco.
System Dynamics and Tobacco Treatment 6
Informed by the PHS Guideline for Treating Tobacco Use and Dependence, other published
literature on managing primary practices, as well as input from participating primary care
physicians, our model includes examines the dynamics of smokers visiting a given primary care
practice over a period of two to four years. The problem to be modeled has multiple
components, namely understanding: (1) How to facilitate a ‘quit attempt’; (2 ) How tobacco
treatment can stimulate growth of the practice; and (3) How practice treatment efforts impact
patient outcomes and practice outcomes (e.g., reimbursement), over time.
The conceptual framework for the general model links three broad domains: (1) Provider
Practices, (2) Patient Tobacco Use, and (3) Patient Health (see Figure 1). Tobacco use can be
viewed as a mediator of patients’ health and their use of primary care, in that everyone requires
some level of primary care at some point (whether for an acute, chronic or preventive health
matter)(Fetter, Averill et al. 1984; Ritzwoller, Goodman et al. 2005). Moreover, we know that
tobacco users are more likely to have respiratory and other health problems and, therefore, are
more likely to demand primary care services (Rigotti 2000; Rigotti 2002).
‘smoxng = Figure 1 -
eisai me Sui Conceptual
Framework for
&&) EE) _
To represent the interdependent nature of these domains, we have chosen to adopt
Hornbrook’s fundamental concept of health care episodes (Hornbrook, Hurtado et al. 1985).
Hornbrook and his colleagues are economists and health services researchers whose work
takes a theoretical approach to unitizing health care services and costs (Hornbrook and al.
2005). The concept of a health care episode is useful here because it “enables more
appropriate assessment of costs of care and, in addition, lends itself to analysis of the
processes as well as the outcomes of medical care” (Hornbrook, Hurtado et al. 1985)(p. 164).
A health care episode is defined as a series of health-related events with a beginning, an end,
and a course, all related to a given health problem that exists over a specific time period. For
our study, there are four types of episodes, namely: (1) smoking episodes, (2) quitting episodes,
(3) illness episodes, and (4) treatment episodes. Although an illness episode (e.g., the period of
time someone is sick with the flu) may be unrelated to a patient's current smoking behavior, it
nonetheless has the potential to bring the patient to the doctor's office, offering the opportunity
to address their tobacco use (Thompson, Michnich et al. 1988; McBride, Plane et al. 1997;
Sippel, Osborne et al. 1999; Easton, Husten et al. 2001; Katz, Muehlenbruch et al. 2002; Smith,
Sheffer et al. 2003). In other words, the visit is an opportunity to initiate a quitting episode, and
the system dynamics model is used to assess the dynamics of a quitting episode a function of
Provider Practices, Patient Tobacco Use, and Patient Health.
System Dynamics and Tobacco Treatment 7
In model runs, Provider Practices are presented in four modes, of ‘intervention strategies’ as
follows:
1. Counseling only (which and be moderated by three levels of provider's counseling skill -
low - medium - high);
2. Counseling with referral to Quit Line (an external, unlimited telephone counseling and
resource service);
3. Counseling and Quit Line with prescription of a single pharmacotherapy (e.g., nicotine
replacement therapies (NRTs), including patches, lozenges, gum, and inhalers - or non-
NRTs, including Bupropion [Zyban ®/Wellbutrin ®] and Varenicline [Chantix®)).
4. Counseling and Quit Line with prescription of a combination pharmacotherapy (¢.g.,
NRT patches and gum).
A stock-and-flow diagram of patients who smoke entering a practice, being exposed to the
provider's intervention strategy is shown in Figure 2. These patients are funneled into three
stocks according to their response to the intervention: Patients who continue to smoke, Patients
who have quit smoking, and Patients who quit, but then relapsed (i.e. resumed smoking).
quitter, quitting and relapsing, or simply not quitting (see Figure 2).
Figure 2 - Stock-and-Flow of Patients’ Response to an Intervention Strategy
Fraction who receive the Time for nonquitters
intervention but DO NOT to leave practice
quit
\ éDN/
Se a eae to AFD) porcentage
smoke Nonquitters parnesk ey
Fraction who receive iehistiot leaving practice
quit Nonquiters quitting i
Time for
quitters to
fe practice
ge | Pis who have
Effect of
PROVIDER Influx of iia Quitters leaving
COUNSELING ‘quitters 2
SKILL LEVEL practice
or Intervention
strategies
Percentage Pte who have
per week of ‘capes -——~<—K——e>
relapsers who
Baseline visits aniWagat Rese leaving
per week practice
Ralpsers quitting ag
Time for
relapsers to leave
amikon Styl fon) ractice
the intervention S— mee who .
Smokers entering the receive the SS
practice intervention
Time to recieve
( intervention
New patient Note: Structure is replicated for ‘Heavy
(feral factor Smokers’and ‘LightSmokers’
System Dynamics and Tobacco Treatment 8
Patients’ relapsing is simulated via a construct we defined as ‘quitting ambivalence,’ the
discomfort of quitting due to either nicotine withdrawal, side effects of cessation medication use,
or other psychological or behavior symptoms that can accompany a patient's quit attempt. A
stock-and-flow diagram of this structure is shown in Figure 3. Quitters’ Ambivalence drives
patient relapse for patients who are successful at making a quit attempt.
Figure 3 - Stock-and-Flow of Quitters’ Ambivalence about Cessation
COUNSELING PLUS QUIT
Relapse COUNSELING PLUS QUIT LINE and mono pharma
Rae LINE Effect of cessation Elfest of ceceation
ambivalence on relapse ambivalence on relapse
HS PROVIDER COUNSELING
; COUNSELING PLUS QUIT
ONLY Effect of cessation LINE !and combed phar
ambivalence on relapse Effect of cessation
“aa ambivalence on relapse
Quitters‘
Mounting ambivalence after ambivalence | weteg e
treatment due to withdraw! and about cessation aaa al
office visi
med side effects ge
Ambivalence decrease
fraction
Ambivalence increase
fraction after office
visit
Baseline
ambivalence
Average time to relapse
Note: Structure is replicated for
‘Heavy Smokers’and ‘LightS mokers’
The dynamics of Quitters’ Ambivalence are a function of the treatment mode that is employed
by the provider. Simulated output shows how a more intensive intervention strategy fosters
more patients quitting and extends the quitting episode (see Appendix: Ratio of Quitters to
Relapsers by Strategy; Proportion of Quitter to Relapsers by Strategy).
Other features of the model include structures to simulate patients’ change in ‘health severity’ as
a function of their tobacco use status and the tendency for successful patients (i.e., those who
have quit) to bring in new patients to the practice, via word of mouth. Finally, the model also
simulates reimbursement via new Medicaid and Medicare provisions, for time spent counseling
patients (see Figures 4 and 5).
These structures represent important feedback loops in the model that impressed participating
physicians. The lesson learned was that helping patients quit was good for business, as patients
who quit were more likely to remain under their care for a longer period of time, and, more
importantly, ushered in new patients (see Appendix: Effect of New Patient Referrals by
Quitters). More new patient visits translated into more stable practice environments, over time.
System Dynamics and Tobacco Treatment 9
Figure 4 - Stock-and-Flow of Average Health Severity of Patients, Visits per Year, and
Reimbursement from State-sponsored Medicaid and Medicare Programs
Reimbursment Frac of patients on
rate up to 10 Medicaid and Medicare
min
<Influx of
nonquitters> Nonquitters
<Pts who continue to cumulative Medicaid
Reimbursements and Medicare
submitted reimburse
nonquitters
Time to change
health severity of
nonquitters
Nonquitters average
visits per year
Average health
severity of
Change in nonquitters
health severity
of nonquitters
Initial health
severity
nonquitters
Health severity
goal for Note: Structure is replicated for
nonquitters ‘Heavy Smokers’and ‘LightS mokers’
Simulation analyses of expected reimbursement from State-sponsored Medicaid and Medicare
services in New York State were also modeled. Simulated output called attention to the fact that
non-quitters and relapsers (see Appendix: Status of Patients who are Receiving Treatment;
Reimbursement Rate Comparison), who comprised the largest numbers of tobacco patients in a
practice, accounted for most of the reimbursement revenue generated for the practice. In
addition, the model indicated that the health severity of patients who did not successfully quit
was worse than those who did quit successfully. Hence, physicians inferred that health
problems would likely translate into more office visits per year. The take home message here
was that providers can benefit financially from working consistently with all tobacco patients.
Formative Assessment of the Simulation Tool. Testing of the model is on-going, although
the current version has passed basic Verification tests, which are concerned with verifying that
the structure and the parameters of the system have been correctly incorporated into the model;
Legitimation tests, which affirm that the model follows commonly accepted principles or rules of
system structure and dimensional checks (i.e., units of measurement or quantification of the
variables on each side of equations are the same); and (3) Validation tests, which address the
extent to which the simulated behavior of the model is like the actual ‘real world’ problem
behavior it is intended to represent (Forrester and Senge 1980; Sterman 2000).
System Dynamics and Tobacco Treatment 10
Figure 5 - Stock-and-Flow of ‘Word-of-Mouth’ Effect on Generation of New Patients
Referrals by Patients who were Successful Quitters
<Nonquitters
quitting>
<Influx of
quitters> <Ralpsers
quitting again>
<Ralpse rate>
<Quitters leaving
practice>
Total quit rate
after intervention Discussion
SWITCH
Total loss of
{ quitters
Total discussion
about tx experience
by quitters Quitters OUT
Quitters IN adjust adjust on
on discussion discussion
smal J =>
_ level of Average discussion about tx patient
disoussion by experience by quitters referral
quitters
rate
<Pts who have
Note: Structure is replicated for
quit: smoking> ‘Heavy Smokers’and ‘LightS mokers’
We used a mixed (qualitative and quantitative) one-arm, pretest-posttest design to assess the
utility of the tool for participating providers. Note that involvement by any given practice lasted
approximately four months. To assess baseline level of implementation of office systems, we
completed a practice profile with assistance from the practice administer during our initial
planning meeting (M zero). In addition, at M1 Baseline and M3 Follow-up we asked the practice
Mo M1 M2 M3 M4 administrator or other staff
a @ a a a member to conduct a rapid
Recuimerts; Gassing fetfaciceFolowup Zrdfrecie patient chart review to obtain
overview ($50 incentive) session ($50 incentive) session objective evidence of tobacco
meeting (e100 i (ea aj screening and treatment
practices among providers.
Figure 6 - Intervention Design
System Dynamics and Tobacco Treatment 11
To determine the feasibility and acceptability of using the simulation tool, we analyzed discourse
from each practice’s academic detailing session (collected at M2 and M4) and from our project
logbook entries throughout the study. An evaluation of the detailing sessions was completed by
providers after each one was completed (see Figure 6).
Figure 7 - Comparison of Participating Providers’ Preference
for Feedback Format during Academic Detailing Sessions
Results indicate that
1008 providers found tailored
60% simulation output to be a
greater source of
50% Motivation than feedback
40% about their self-reported
practices or independent
30% chart reviews (see Figure
20% 7). The large majority of
participating providers
10% (93%) indicated that our
office-based feedback
0% .
sessions fostered a strong
Provider Practice Chart Review SD Modeling behavioral intent to
Survey Simulations enhance the way they
addressed the needs of
patients who used tobacco in their practice (see Table 4).
Our field experience indicates that simple behavior-over-time graphs easily capture the interest
of participating providers. Moreover, results support that these graphs, when accompanied with
brief explanatory text, appear to foster effective communication. We found that presentation of
the stock-and-flow diagrams that comprise the system dynamics model were difficult to explain
to participants, and time dedicated to this - which was on average limited to less than 20
minutes - seemed to dampen enthusiasm about the simulated output. However, some found
that the graphs shown with a tiny, partial stock-and-flow figure overlay would be the most
effective way to share results and build an understanding of the model structures that produce
specific simulated data.
Table 3 - Preliminary Assessment of Intent to Enhance Tobacco Treatment Practices
Provider motivated to make changes in the practice 93% 13
Increased motivation to provide follow-up and counseling 57% 8
Increased motivation to encourage relapsers 50% 7
Encouraged to ensure that patients have easy access to tobacco treament | 41% 6
Increased motivation to document use at every visit 36% 5
Encouraged to prescribe pharmacotherapy 36% 5
Motivated planning to educate staff 36% 5
Informed about ways to obtain reimbursement for counseling 29% 4
Promoted use of dedicated staff time for patient counseling/follow-up 29% 4
Encouraged to spend more time counseling patients 21% 3
Encouraged to generate own practices performance data 21% 3
System Dynamics and Tobacco Treatment 12
Future Directions. Additional model development is under consideration, per input from our
participating providers, such as: (1) Further disaggregation of patients (e.g., by ‘chronic’ vs.
‘non-chronic’ medical conditions or by adherence to meds); (2) Examination of the effect of
subsequent ‘booster’ office visits (as the effect of only one visit is featured in the current model);
(3) Examination of the effect of longer-term treatment sequelae (e.g., weight gain); (4)
Examination of the effect of integrating other behavioral treatments (e.g., diet, exercise, alcohol
consumption). The key question regarding whether or not to embellish on the existing version of
the model is the following: How much complexity is sufficient to motivate practice change?
Based on the results of this formative fieldwork with urban, primary care providers, we conclude
that system dynamics models are effective tools for communicating complexity to busy health
care providers. There is a need for further research that assesses actual practice change, as
the current study design was not able to detect more than an effect on behavioral intent to
change. Also, time and effort required to facilitate office visits suggests the need to explore
alternative ways to expose providers to the model, such as web-based platforms and more
structured user interfaces.
System Dynamics and Tobacco Treatment 13
REFERENCES
Adsit, R., D. Fraser, et al. (2005). "Changing clinical practice, helping people quit: The
Wisconsin cessation outreach model." Wisconsin Medical | ournal 104(4): 32-36.
Andrews, J., M.S. Tingen, et al. (2001). "Provider feedback improves adherence with AHCPR
smoking cessation guideline." Preventive Medicine 33: 415-421.
Avorn, ] .,S.B. Soumerai, et al. (1992). "A randomized trial of a program to reduce the use of
psychactive drugs in nursing homes." New England | ournal of Medicine 327(168-173).
Bar-Yam, Y. (2006). "Improving the effectiveness of health care and public health: A multiscale
complex systems analysis." American J ournal of Public Health 96(3): 459-466.
Bentz, C.J., B. Bayley, et al. (2006). "The feasibility of connecting physician offices to a state-
level tobacco quit line." American | ournal of Preventive Medicine 30(1): 31-37.
Berings, D., L. Blondeel, et al. (1994). "The effect of industry-independent drug information on
the prescribing of benzodiazepine in general practice." European J ournal of Clinical
Pharmacology 46(501-505).
Cabana, M. D.,C.S. Rand, etal. (1999). "Why don't physicians follow clinical practice
guidelines?" The | ournal of American Medical Association 282(15): 1458-1465.
Cavana, R. Y. and L. V. Clifford (2006). "Demonstrating the utility of system dynamics for public
policy analysis in New Zealand: The case of excise tax policy on tobacco." System
Dynamics Review 22(4): 321-348.
Cromwell, J., W.J. Bartosch, et al. (1997). "Cost-effectiveness of the clinical practice
recommendations in the AHCPR guideline for smoking cessation." | ournal of the
American Medical Association 278(21): 1759-1766.
Davis, D. A. and A. Taylor-Vaisey (1995). "Translating guidelines into practice: A systematic
review of theoretical concepts and research evidence in the adoption of clinical practice
guidelines." CMAJ 157: 408-416.
Dietrich, A.J} ., G. T. O'Connor, et al. (1992). "Cancer: Improving early detection and prevention.
A community practice randomized trial." British Medical | ournal 304(687-691).
Easton, A., C. Husten, et al. (2001). "Smoking cessation counseling by primary care women
physicians: Women Physicians' Health Study." Women & Health 32(4): 77-91.
Feenstra, T. L., H. H. Hamberg-van Reenen, et al. (2005). "Cost-effectiveness of face-to-face
smoking cessation interventions: A dynamic modeling study." Value in Health 8(3): 178-
190.
Fetter, R.B., R. F. Averill, et al. (1984). "Ambulatory visit groups: A framework for measuring
productivity in ambulatory care." Health Services Research 19(4): 415-437.
Forrester, ). W. and P. M. Senge (1980). "Tests for building confidence in system dynamics
models." TIMS Studies in the Management Science 14: 201-228.
Gandjour, A. and K. W. Lauterbach (2005). "How much does it cost to change the behavior of
health professionals? A mathematical model and an application to academic detailing."
Medical Decision Making 25(May-J une): 341-347.
Glynn, T.J. and M. W. Manley (1989). How to Help Your Patients Stop Smoking. A National
Cancer Institute Manual for Physicians. Bethesda, MD, Smoking, Tobacco and Cancer
Program, Division of Cancer Prevention and Control, National Cancer Institute, NIH.
System Dynamics and Tobacco Treatment 14
Goldstein, M.,) . DePue, et al. (1998). "A population-based survey of physician smoking
cessation counseling practices." Preventive Medicine 27: 720-729.
Goldstein, M. G., R. Niaura, et al. (2003). "An academic detailing intervention to disseminate
physician-delivered smoking cessation counseling: Smoking cessation outcomes of the
physicians Counseling Smokers Project." Preventive Medicine 36: 185-196.
Greco, P.J. and). M. Eisenberg (1993). "Changing physicians’ practices." New England
ournal of Medicine 329: 1271-1274.
Halpern, M. T., Z. M. Khan, et al. (2000). "Economic model of sustained-release bupropion
hydrocloride in health plan and work site smoking-cessation programs." American
Journal of Health-S ystems and Pharm 57: 1421-1429.
Headrick, L. A. (2000). "Two kinds of knowledge to achieve better care." | ournal of General
Internal Medicine 15(9): 647-655.
Homer, J. B. and G. B. Hirsch (2006). "System dynamics modeling for public health:
Background and opportunities." American | ournal of Community Psychology 96(3): 452-
458.
Hornbrook, M. C. and e. al. (2005). "Building a virtual cancer research organization." | ournal of
the National Cancer Institute Monographs 35: 12-25.
Hornbrook, M. C., A. V. Hurtado, et al. (1985). "Health are episodes: Definition, measurment
and use." Medical Care Review 42(2): 163-218.
Katz, D. A., D. R. Muehlenbruch, et al. (2002). "Effectiveness of a clinic-based strategy for
implementing the AHRQ smoking cessation guideline in primary care." Preventative
Medicine 35: 293-302.
Kiefe, C.1.,}.J. Allison, et al. (2001). "Improving quality improvement using achievable
benchmarks for physician feedback: A randomized controlled trial." | ournal of the
American Medical Association 285(22): 2871-2879.
Levy, D., S. Bales, et al. (2005). "The role of public policies in reducing smoking and deaths
caused by smoking in Vietnam: Results from the Vietnam tobacco policy simulation
model." Social Science & Medicine 62: 1819-1830.
Lounsbury, D. W. (2002). Understanding the dynamics of prevention, care, and empowerment:
A systems approach to HIV/AIDS policy innovation. East Lansing, MI, Michigan State
University.
McBride, P., M. Plane, etal. (1997). "Smoking screening and management in primary care
practices." Archives of Family Medicine 6: 165-172.
Meredith, L., E. Yano, etal. (2005). "Primary care provider attitudes are associated with
smoking cessation counseling and referral." Medical Care 43(9): 929-934.
O'Brien, T., A. D. Oxman, etal. (2005). "Educational outreach visits: Effects on professional
practice and health care outcomes (Review)." The Cochrane Collaboration Retrieved
January 15, 2006.
Orleans, C. T. (1993). Treating nicotine dependence in medical settings: A stepped-care model.
New York, Oxford University Press.
Oxman, A. D., M.A. Thomson, etal. (1995). "No magic bullets: A systematic review of 102 trials
of interventions to improve professional practice." CMAJ 153: 1423-1431.
System Dynamics and Tobacco Treatment 15
Randers, J . (1980). Elements of the System Dynamics Method. Cambridge, MA, Productivity
Press.
Richardson, G. P. and A. L. Pugh Ill (1981). Introduction to System Dynamics Modeling.
Portland, Oregon, Productivity Press.
Rigotti, N. (2000). "A 36-year-old woman who smokes cigarettes." ] AMA 284(6): 741-749.
Rigotti, N. (2002). "Treatment of tobacco use and dependence." The New England | ournal of
Medicine 346(7): 506-512.
Ritzwoller, D. P., M.J. Goodman, etal. (2005). "Creating standard cost measures across
integrated health care delivery systems." | ournal of the National Cancer Institute
Monographs 35: 80-87.
Ronckers, E. T., W. Groot, etal. (2005). "Systemic review of economic evaluations of smoking
cessation: Standardizing the cost-effectiveness." Medical Decision Making 25: 437-448.
Sippel, }. M., M. L. Osborne, et al. (1999). "Smoking cessation in primary care clinics." | ournal
of General Internal Medicine 14: 670-675.
Smith, P. O., C. E. Sheffer, et al. (2003). "Smoking cessation research in primary care treatment
centers: The SCRIPT-MS project." American | ournal of The Medical Scinece 326(4).
Soumerai, S. B. andJ. Avorn (1990). "Principles of educational outreach (‘Academic Detailing’)
to improve clinical decision making." | ournal of the American Medical Association
263(4): 549-556.
Steele, M., M. Fors, etal. (1989). "Cost effectiveness of two interventions for reducing outpatient
prescribing costs." DICP 23(497-500).
Sterman, J . (2000). Business Dynamics: Systems Thinking and Modeling for a Complex World.
Boston, McGraw-Hill.
Thompson, R.S., M. E. Michnich, et al. (1988). "Effectiveness of smoking cessation
interventions integrated into primary care practice." Medical Care 26(1): 62-76.
Torrijos, R. M. and S. A. Glantz (2006). "The US Public Health Service "treating tobacco use
and dependence clinical practice guidelines" as a legal standard of care." Tobacco
Control 15: 447-451.
Weissman, N. W.,J.J. Allison, et al. (1999). "Achievable benchmarks of care: The ABCs of
benchmarking." | ournal of Evaluation in Clincial Practice 5: 203-215.
Wensing, M. and R. Grol (1994). "Single and combined strategies for implementing changes in
primary care: A literature review." International | ournal of Quality Health Care 6: 115-
132.
Yano, E.M., A. Fink, etal. (1995). "Helping practices reach primary care goals: Lessons from
the literature." Archives of Internal Medicine 155: 1146-1156.
System Dynamics and Tobacco Treatment 16
APPENDIX - Sample simulation output - Typical participating primary care practice
Status of Patients who are Receiving Intervention
1,200
900 This figure helps providers
understand how many tobacco
2 patients are in their care. Note that
FI Non-quitters comprise the majority
q 600 of such patients.
a
300
0
1 24 47 70 93 116 139 162 185 208
Time (Week)
Nonquitters : Counseling PLUS QL & MONO-
Quitters : Counseling PLUS QL & MONO:
Relapsers : Counseling PLUS QL & MONO:
Sample Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
System Dynamics and Tobacco Treatment 17
Heavy Smokers Head-to-Head Comparison of Treatment Strategies
40
For heavy smokers, there is
an additive, beneficial effect
30 associated with more
intensive strategies.
20
10
0
1 24 47 70 93 116 139 162 185 208
Time (Week)
HS quitters : Counseling ONLY
HS quitters : Counseling PLUS Quit Line
HS quitters : Counseling PLUS QL & MONO
HS quitters : Counseling PLUS QL & COMBO
Sample Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
System Dynamics and Tobacco Treatment 18
Light Smokers Head-to-Head Comparison of Treatment Strategies
40
For light smokers, the most intensive
intervention has a large effect. Note that
30 Counseling only and Counseling with
Quit Line are equivalent in terms of
impact on light smokers.
20
Patients
10
1 24 47 70 93 116 139 162 185 208
Time (Week)
LS quitters : Counseling ONLY
LS quitters : Counseling PLUS Quit Line
LS quitters : Counseling PLUS QL & MONO
LS quitters : Counseling PLUS QL & COMBO
Sample Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
System Dynamics and Tobacco Treatment 19
Ratio of Quitters to Relapsers by Strategy
The ratio of Quitters to Relapsers is
greatest due to practices that employ the
—_— most intensive strategy (Counseling
15 PLUS Quit Line and Combination
Pharmacotherapy). Most patients
relapse within two to three months.
1 24 47 70 93 116 139 162 185 208
Time (Week)
Ratio of Quitters to Relapsers : Counseling ONLY
Ratio of Quitters to Relapsers : Counseling PLUS Quit Line
Ratio of Quitters to Relapsers : Counseling PLUS QL & MONO
Ratio of Quitters to Relapsers : Counseling PLUS QL & COMBO
Sample Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
System Dynamics and Tobacco Treatment 20
Proportion of Quitters among those who Make a Quit Attempt
0.8
0.6 The proportion of Quitters to Relapsers
is greatest due to practices that employ
2 —_—__ the most intensive strategy (Counseling
z 0.4 PLUS Quit Line and Combination
ie Pharmacotherapy). Most patients
~ relapse within two to three months.
0.2
0
1 24 47 70 93 116 139 8 162 185
Time (Week)
Proportion of quitters among quitters and relapsers in practice
Proportion of quitters among quitters and relapsers in practice
Proportion of quitters among quitters and relapsers in practice
Proportion of quitters among quitters and relapsers in practice
Counseling ONLY
208
Counseling PLUS Quit Line
: Counseling PLUS QL & MONO
Counseling PLUS QL & COMBO
Sample Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
System Dynamics and Tobacco Treatment 21
Proportion of Quitters among All Tobacco Uses in Practice
0.4
By examining the proportion of all
Quitters among all tobacco patients in
the practice, we observe that only the
0.3 most intensive intervention appears to
be clinically significantly better. Also, we
seen that the proportion of successful
quitters as a result of a single office visit
Percentage
0.2
is small, less than 10% after about 15
weeks’ time.
0.1
0
1 24 47 70 93 116 139 162 185 208
Time (Week)
Proportion of quitters among all tobacco users in practice : Counseling ONLY
Proportion of quitters among all tobacco users in practice : Counseling PLUS Quit Line
Proportion of quitters among all tobacco users in practice : Counseling PLUS QL & MONO
Proportion of quitters among all tobacco users in practice : Counseling PLUS QL & COMBO
Sample Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
System Dynamics and Tobacco Treatment 22
Effect of New Patient Referral by Quitters
40
The effect of new patient referrals by
quitters occurs rapidly, within two
months. Moreover, Heavy Smokers
30 || appear more likely to make more
referrals than Light Smokers.
20
+>
10
0
0 4 8 12 16 20 24 28 32 36 40 44 48 52
Time (Week)
Heavy Smokers entering the practice : New patients - no referral effe:
Light Smokers entering the practice : New patients - no referral effe
Heavy Smokers entering the practice : New patient referral effee
Light Smokers entering the practice : New patient referral effe
Sample Simulation Output: A Typical Primary Care Practice (18% smeking prevalence among new patients)
System Dynamics and Tobacco Treatment 23
State Reimbursement Rate Comparison
Two rates of reimbursement are offered
by the State for Medicaid and Medicare
patients. Simulated output shows that
Heavy Smokers account for higher
reimbursement, due to their more
frequent visits to the provider's office.
Over time, the reimbursements show a
substantial accumulation!
0 4 8 12 16 20 24 28 32 36 40 44 48 52
Time (Week)
HS cumulative state reimburse : State Reimbursement up to 10 min
LS Cumulative State Reimbursement : State Reimbursement up to 10 min
HS cumulative state reimburse : State Reimbursement more than 10 min
LS Cumulative State Reimbursement : State Reimbursement more than 10 min
Semple Simulation Output: A Typical Primary Care Practice (18% smoking prevalence among new patients)
