Improving medical decision making - Stroke prevention in AF

Improving
medical decision making

Stroke prevention in atrial fibrillation

D.L. Arts

Layout and printing: Optima Grafische Communicatie, Rotterdam, The Netherlands
Cover design: Magda Jurewicz
ISBN: 978-94-6169-963-3

Copyright © 2016 D.L. Arts
No parts of this thesis may be reproduced or transmitted in any form or by any means,
without the prior permission of the author.

Financial support by the Dutch Heart Foundation for the printing of this thesis is
gratefully acknowledged.

Improving medical decision making
Stroke prevention in atrial fibrillation

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad van doctor
aan de Universiteit van Amsterdam
op gezag van de Rector Magnificus
prof. dr. ir. K.I.J. Maex

ten overstaan van een door het College voor Promoties ingestelde commissie
in het openbaar te verdedigen in de Aula der Universiteit
op vrijdag 4 november 2016, te 11:00 uur
door Diederick Laurentius Arts
geboren te Nijmegen

P romotiecommissie :

Promotores: Prof. dr. H.C.P.M. van Weert
Overige leden: Universiteit van Amsterdam
Prof. dr. A. Abu-Hanna
Universiteit van Amsterdam
Prof. dr. H.R. Büller
Universiteit van Amsterdam
Prof. dr. R.J.G. Peters
Universiteit van Amsterdam
Prof. dr. N.F. de Keizer
Universiteit van Amsterdam
Dr. F.H. Rutten
Universitair Medisch Centrum Utrecht
Prof. dr. J. van der Lei
Erasmus Universiteit Rotterdam
Prof. dr. M.E. Numans
Universiteit Leiden

Faculteit der Geneeskunde
This research was supported by an unrestricted grant from the ZORRO foundation.

Table of contents

Chapter 1. General introduction and thesis outline 7
19
Chapter 2. Frequency and risk factors for under- and over-treatment in stroke
prevention for patients with non-valvular atrial fibrillation in general practice 31

Chapter 3. Reasons for intentional guideline non-adherence: a systematic 47
review
61
Chapter 4. Guideline-related barriers to optimal prescription of oral 73
anticoagulants in primary care
89
Chapter 5. Improving stroke prevention in patients with atrial fibrillation
107
Chapter 6. Effectiveness and usage of a decision support system to improve
stroke prevention in general practice: a cluster randomized controlled trial 125
137
Chapter 7. Acceptance and barriers pertaining to a general practice decision 145
support system for multiple clinical conditions: a mixed methods evaluation 153
157
Chapter 8. Development of a regression model and neural network to predict 161
stroke and comparison to the CHA2DS2-VAsc in a large dataset 165

Chapter 9. General discussion
Chapter 10. English summary

Nederlandse samenvatting
Curriculum vitae
Portfolio
List of publications
Dankwoord



Chapter 1

General introduction and thesis outline



General introduction and thesis outline

I ntrod u ction Chapter 1

In this thesis we address the problem of low guideline adherence in general, and specifi-
cally for atrial fibrillation (part A), as well as the potential of decision support systems for
improving it (part B). Furthermore, we investigate the use of complex prediction models
for improving stroke prevention in patients with atrial fibrillation (part C). Figure 1 dis-
plays an outline of the thesis and the topic of each chapter.

A. Current practice: guideline adherence

2 Current adherence 3 Intentional reasons 4 Appraisal of the
to Dutch for guideline Dutch AF guideline
AF guideline non-adherence

B. Improving guideline adherence with CDSS C. Improving disease
prediction
5 Randomized 6 Impact on guideline
controlled trial on adherence and 8 Predicting stroke
GP CDSS usage using computer-
based models
7 Evaluation of
user experience

Figure 1. Topics of the chapters in this thesis.
Chapters are indicated by the numbers next to the topic. AF: atrial fibrillation, GP: general practitioner,
CDSS: clinical decision support systems

P reliminaries

In the late 1900s, the medical community started to doubt their reliance on accumulated
personal experiences for medical decision making. Archibald Cochrane (1909-1988)
first criticized our practice of medicine in his 1972 book Effectiveness and Efficiency:
Random Reflections on Health Services [1]. He noted that many medical practices that
were presumed to be effective lacked evidence from controlled trials. Shortly thereafter,
researchers began publishing about the variation in treatment practices and document-
ing errors in clinical reasoning.
In 1990, the term “evidence-based medicine” was formally introduced and by 2000
it was well known throughout the medical community [2]. Currently, evidence-based
medicine (EBM) is defined as “an approach to medical practice intended to optimize
decision-making by emphasizing the use of evidence from well designed and conducted
research” with the intention to provide the best possible care for individual patients [3].

9

Chapter 1

EBM is now considered the gold standard for clinical practice, as treatment strategies
can be based on results from large groups of patients, as opposed to one physician’s
opinion. The most common implementation of EBM involves evidence-based clinical
practice guidelines (CPGs): “systematically developed statements to assist practitioner
and patient decisions about appropriate health care for specific clinical circumstances”
[4].

Young doctors are educated on the importance of applying EBM and the role of evidence
therein during internships [5]. But, are physicians, young and old, really using evidence-
based medicine in daily practice? Are we applying the best available evidence to make
decisions in relation to patients’ circumstances and preferences?

Guideline overload
Today, we have no shortage of CPGs. Medical societies are constantly creating and
updating guidelines for specific diseases, with more than 1000 guidelines being cre-
ated annually [6]. However, the mere provision of guidelines does not ensure doctors
adhere to them. It has consistently been shown that guideline adherence varies greatly,
but rarely exceeds 80% and is on average 55% [7, 8]. Why would physicians not follow
guidelines that often? Cabana et al. [9] describe possible reasons for non-adherence in
their systematic review (e.g., Lack of Awareness, Lack of Outcome Expectancy, and Guide-
lines Factors) and divide these into three different categories: Knowledge, Attitudes, and
Behavior. Other studies found barriers in similar categories [10, 11]. Knowledge of these
barriers to guideline adherence allows us to look for solutions that target specific issues
in guideline implementation and adherence.

Decision support
Making good use of guidelines requires a high-quality guideline, a willing doctor and
patient, the right tools, circumstances, and time. Part of the medical community has long
known that there are other methods to support clinicians besides (paper) guidelines.
Some turned to computers to help implement clinical practice guidelines even before
the term “evidence-based medicine” was introduced. These systems, which supported
healthcare workers in making clinical decisions, were aptly named “Clinician Decision
Support Systems” (CDSSs). One of the first of these systems was “MYCIN,” a 1976 rule-
based system that was designed to help doctors diagnose blood infections and select
the right treatment [12]. It used IF-THEN rules (e.g., IF patient has fever THEN recommend
antibiotics) to guide physicians to the right endpoint. Many, if not most, CDSSs still use
these basic IF-THEN rules to make recommendations today.
Parts A and B of this thesis relate to guideline adherence and decision support. What
specific reasons do physicians have for guideline non-adherence? Can we improve

10

General introduction and thesis outline Chapter 1

guideline adherence by supporting physicians with clinical decision support systems?
We attempted to answer these questions in the field of atrial fibrillation (AF), specifically
stroke prevention in AF, as guideline non-adherence is prevalent in this domain and the
potential for improving patient health is large (Box 1).

Box 1. Atrial Fibrillation
Atrial fibrillation (AF) is the most common form of cardiac arrhythmia (abnormal heart rhythm),
characterized by rapid and irregular beating. Its prevalence has increased over the last 30 years
[13]. In 2006 it ranged from 0.7% (age 55-59) to 17.8% (age > 85). On average, 2% to 3% of the
population is affected in Europe and North America [14].
Patients with AF are at high risk for stroke, with up to a five-fold risk compared to patients with-
out AF [15]. Apart from the increased risk, stroke severity also is increased [16]. Therefore, most
patients with AF need to be treated with antithrombotic medication to reduce stroke risk. Oral
anticoagulation (OAC), such as warfarin and the new oral anticoagulants (NOACs), can reduce risk
of stroke by 60% [17, 18].
Most recent guidelines recommend prescribing OAC for all but the lowest risk categories to allow
for easy implementation and optimal stroke risk reduction. To determine which patients should
receive OAC, stroke risk scores are used to predict individual risk of stroke for each patient. In
doing so, net benefit is balanced against possible side effects, mainly increased bleeding risk.
Several such risk-prediction schemes have been developed in the past and are used in interna-
tional guidelines [19, 20]. Currently, the CHA2DS2-VASc is the most widely used stroke risk score:
it consists of 7 variables that can add up to a total of 9 points [19]. Due to the high net benefit of
treatment with OAC, the stroke risk cut off for this treatment has been lowered over the past few
years [21]. The most recent European Society for Cardiology (ESC) guidelines recommend treat-
ment for patients scoring 1 point or more. This means that most patients over 65 years old with AF
should be receiving antithrombotic medication. For this thesis we focused on the Revised Dutch
GP guideline for atrial fibrillation, authored by the Dutch College of General Practitioners (NHG).
It introduced the CHA2DS2-VASc, but did not yet recommend treating all patients with 1 point.
Instead, it recommended weighing risk and benefit with the patient [22].

Part A: Current practice: guideline adherence

Adherence to stroke prevention guidelines in AF is poor [23-25]. This non-adherence
is mainly related to underuse of OAC in patients with medium to high stroke risk. Two
important reasons for OAC underuse are: 1) the complexity of the decision rules used,
and 2) physicians’ concerns with the bleeding risk associated with OAC. However, the
benefit of stroke prevention greatly outweighs the risk of bleeding due to OAC, and OAC
should therefore not be withheld when indicated [26].
To determine if guideline non-adherence in AF stroke prevention was as prevalent in
the Netherlands as it was in other countries [27, 28], we evaluate guideline adherence
amongst general practitioners (GPs) using a national database in Chapter 2. We inves-
tigate to what extent Dutch GPs appropriately prescribe anticoagulants to protect their
patients from stroke, and balance medication-related adverse event risk against poten-

11

Chapter 1

tial benefit. Next, we investigate potential barriers to guideline adherence. Some barriers
relate to the guideline itself: recommendations may not be clearly defined, making them
hard to implement. Likewise, the level of the evidence might be missing from recom-
mendations, allowing for uncertainty regarding their importance [9, 29]. Lugtenberg et
al. reported that suboptimal guideline features may have acted as a barrier for adherence
to the Dutch AF guideline [30]. To establish what specific features might limit guideline
adherence, we perform a thorough appraisal of the guideline in Chapter 3. Using estab-
lished methods of evaluating guidelines, we assess how optimization might improve its
implementation in daily practice.
Another guideline-related barrier to adherence is often mentioned by critics of guide-
lines: they argue that Randomized Controlled Trials are not representative of real-world
populations, and thus guideline recommendations are not applicable to their individual
patients [31]. To determine if physicians indeed have valid reasons for guideline non-
adherence, we set out to discover what intentional reasons they have for not acting
according to guideline recommendations in Chapter 4. We perform a systematic review
of studies on guideline adherence, and categorize reasons for non-adherence.

Part B: Improving guideline adherence with clinical
decision support systems

Can clinical decision support systems help improve guideline adherence? Over the past
years, many trials have tried to answer this question. Studies often focus on indirect
measures for quality of care and results indicate that usage of CDSSs can lead to improve-
ments in clinical practice and guideline implementation [32-34]. However, evidence for
direct improvement on patient health is lacking: the overall effectiveness of CDSS on
mortality has not been established, but a recent review did find a moderate improvement
in morbidity outcomes [35]. Therefore, CDSSs hold promise for the future of healthcare,
but can they help us apply guidelines in the best way possible? Current evidence is not
universally in favor of decision support, as negative side effects can occur when imple-
menting these systems, such as reducing the perceived need for human verification of
medication, providing inaccurate recommendations, or interruption of workflow [36].
Furthermore, there is no clear consensus on what makes decision support successful.
Studies have identified many possible factors for lack of effectiveness. These include:
lack of usability, lack of integration with host systems, lack of time to effectuate advice,
inapplicability to the patient, lack of integration with current workflow, and alert fatigue
[33, 37]. However, due to a lack of homogeneity in methodology and implementations
across a wide range of medical specialties, we have yet to definitively determine features
of effective decision support [38].

12

General introduction and thesis outline Chapter 1

To help answer questions regarding CDSS effectiveness and factors pertaining to effec-
tiveness, we began the Expert-AF project. Our goal was to investigate whether a CDSS
can improve adherence to stroke prevention guidelines in the Netherlands. Chapter 5
contains the trial protocol for the Expert-AF project, a cluster randomized controlled trial
that was run amongst Dutch GPs. We created our own clinical decision support system for
a widely used GP electronic health record system. Chapter 6 describes the effectiveness
of the system we implemented as measured by the change in the proportion of patients
treated according to the guideline. Furthermore, we collected reasons for non-adherence
to gain insight into why GPs would withhold a treatment with such high net benefit from
their patients. In Chapter 7 we use focus groups and surveys to evaluate the system we
created. We investigate how GPs experienced our system and what factors promoted or
limited use. By doing so, it provides a clear overview of the challenges one faces when
trying to implement a new system in an existing, busy workflow.

Part C: Improving disease prediction

While the topic of how to implement decision support is intriguing, we have thus far only
attempted to use computers to help with implementing pre-existing guidelines. These
guidelines are usually developed by multidisciplinary groups of experts who summa-
rize current literature and define an easy-to-use overview of best practices for specific
health conditions [39]. After a guideline has been completed, it is up to researchers or
electronic health record (EHR) vendors to “translate” a guideline into a decision rule.
This usually results in a simple IF-THEN rule, just like those used in MYCIN in 1976. The
average smartphone can run these rules in a split second and its resources will not be
used to their full capacity.
Can we use more complex methods that may improve disease prediction, regardless of
implementation? Can we do more with the computational power modern computers
offer? The answer is yes. There are more complex methods for making predictions that
might be more effective than traditional methods. It is important to note that this does
not necessarily involve computers, but in practice these methods are so complex that
only the computational power of a modern computer can solve the math involved. An
example is Artificial Neural Networks (ANNs). These networks, which mimic the mammal
brain by using connected neurons in order to learn, are often used by large web compa-
nies such as Google and Facebook. ANNs have been around in medicine for a long time.
For example, ANNs are currently used for the prediction of breast cancer and analysis of
MRI images [40, 41]. They are trained using a process called “supervised learning,” and
can automatically model complex non-linear relationships between risk factors. We cur-
rently break these relationships down into formulas to make them usable in daily clinical

13

Chapter 1

practice. This results in a loss of information about the subtleties of the relationships
between risk factors. Can we use computers to leverage these subtleties and improve
predictions? In Chapter 8 we try to improve stroke prediction in atrial fibrillation using
logistic regression and neural networks. We investigate whether these techniques are
better at predicting stroke than the established stroke risk stratification schemes and
determine how patients could benefit from these predictions.
The future of medical decision making
The medical community will continue to strive for the effective use of evidence in health-
care. In this thesis we investigate whether guidelines and computers can facilitate the
ongoing transition to evidence-based medicine in years to come. Are they the future of
medical decision making?

14

General introduction and thesis outline

R eferences Chapter 1

1. Cochrane A. Effectiveness and Efficiency: Random Reflections on Health Services: The Nuff-
2. ield Provincial Hospital Trust; 1972.
3. Gray JAM. Evidence-based Health Care & Public Health: Churchill Livingstone; 2009.
4. Wikipedia: Evidence-based_medicine 2016 [Available from: https:/​ /en.wikipedia.org/wiki/
5. Evidence-based_medicine.
In: Field MJ, Lohr KN, editors. Clinical Practice Guidelines: Directions for a New Program.
6. Washington (DC)1990.
7. Maggio LA, Tannery NH, Chen HC, Cate Ot, O’Brien B. Evidence-Based Medicine Training
in Undergraduate Medical Education: A Review and Critique of the Literature Published
8. 2006–2011. Acad Med. 2013;8​ 8(7):1​ 022‑8.
9. Jackson R, Feder G. Guidelines for clinical guidelines. A simple, pragmatic strategy for guide-
10. line development. 1998;​317(7156):​427‑8.
Grol R, Dalhuijsen J, Thomas S, in’t Veld C, Rutten G, Mokkink H. Attributes of clinical guide-
11. lines that influence use of guidelines in general practice: observational study. Br Med J.
12. 1998;3​ 17(7162):8​ 58‑61.
13. McGlynn EA, Asch SM, Adams J, Keesey J, Hicks J, DeCristofaro A, et al. The quality of health
care delivered to adults in the United States. N Engl J Med. 2003;3​ 48(26):​2635‑45.
14. Cabana, Rand CS, Powe NR, Wu AW, Wilson MH. Why Don ’ t Physicians Follow guidlines, A
15. Framework for Improvement. October. 1999;​0718.
16. Gurses AP, Marsteller JA, Ozok AA, Xiao Y, Owens S, Pronovost PJ. Using an interdisciplinary
17. approach to identify factors that affect clinicians’ compliance with evidence-based guide-
18. lines. Crit Care Med. 2010;3​ 8(8):S​ 282-S91.
19. IOM (Institute of Medicine). Graham R e. Clinical practice guidelines we can trust. National
Academies Press. 2011.
Musen MA. Stanford Medical Informatics: uncommon research, common goals. MD Comput.
1999;​16(1):​47-8, 50.
Camm AJ, Lip GYH, De Caterina R, Savelieva I, Atar D, Hohnloser SH, et al. 2012 focused
update of the ESC Guidelines for the management of atrial fibrillation. Eur Heart J. 2012;​
33(21):​2719‑47.
Zoni-Berisso M, Lercari F, Carazza T, Domenicucci S. Epidemiology of atrial fibrillation: Euro-
pean perspective. Clin Epidemiol. 2014;6​ :2​ 13‑20.
Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the
Framingham Study. Stroke. 1991;2​ 2(8):9​ 83‑8.
Lin HJ, Wolf PA, Kelly-Hayes M, Beiser AS, Kase CS, Benjamin EJ, et al. Stroke severity in atrial
fibrillation. The Framingham Study. Stroke. 1996;2​ 7(10):​1760‑4.
Hart RG. Meta-analysis Antithrombotic Therapy to Prevent Stroke in Patients who have noval-
vular atrial fibrillation. Ann Intern Med. 2007.
Ogilvie IM, Welner SA, Cowell W, Lip GY. Ischaemic stroke and bleeding rates in ‘real-world’
atrial fibrillation patients. Thromb Haemost. 2011;​106(1):​34‑44.
Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. Refining clinical risk stratification for
predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based
approach: the euro heart survey on atrial fibrillation. Chest. 2010;1​ 37(2):2​ 63‑72.

15

Chapter 1

20. Singer DE, Chang YC, Borowsky LH, Fang MC, Pomernacki NK, Udaltsova N, et al. A New Risk
Scheme to Predict Ischemic Stroke and Other Thromboembolism in Atrial Fibrillation: The
21. ATRIA Study Stroke Risk Score. Journal of the American Heart Association. 2013;2​ (3).
22. Lip GYH. Stroke and bleeding risk assessment in atrial fibrillation: when, how, and why? Eur
56. Heart J. 2013;3​ 4(14):1​ 041‑9.
23. fibrillation DCoGPGDGfA. Guideline Atrial fibrillation (second partial revision). Huisarts &
Wetenschap. 2013;
24. (8):3​ 92‑401.
25. Dinh T, Nieuwlaat R, Tieleman RG, Buller HR, van Charante NA, Prins MH, et al. Antithrombotic
drug prescription in atrial fibrillation and its rationale among general practitioners, internists
26. and cardiologists in The Netherlands--The EXAMINE-AF study. A questionnaire survey. Int J
Clin Pract. 2007;6​ 1(1):​24‑31.
27. Lip GYH, Lane DA. Stroke Prevention in Atrial Fibrillation A Systematic Review. Jama-Journal
of the American Medical Association. 2015;​313(19):1​ 950‑62.
28. Arts DL, Visscher S, Opstelten W, Korevaar JC, Abu-Hanna A, van Weert HCPM. Frequency
and Risk Factors for Under- and Over-Treatment in Stroke Prevention for Patients with Non-
29. Valvular Atrial Fibrillation in General Practice. PLoS One. 2013;8​ (7).
Ruff CT, Giugliano RP, Braunwald E, Hoffman EB, Deenadayalu N, Ezekowitz MD, et al. Com-
30. parison of the efficacy and safety of new oral anticoagulants with warfarin in patients with
atrial fibrillation: a meta-analysis of randomised trials. Lancet. 2014;3​ 83(9921):9​ 55‑62.
31. Ogilvie IM, Welner SA, Cowell W, Lip GY. Characterization of the proportion of untreated
32. and antiplatelet therapy treated patients with atrial fibrillation. Am J Cardiol. 2011;​108(1):​
33. 151‑61.
Nieuwlaat R, Capucci A, Camm AJ, Olsson SB, Andresen D, Davies DW, et al. Atrial fibrillation
34. management: a prospective survey in ESC Member Countries - The Euro Heart Survey on
Atrial Fibrillation. Eur Heart J. 2005;2​ 6(22):​2422‑34.
35. Shekelle PG, Kravitz RL, Beart J, Marger M, Wang M, Lee M. Are nonspecific practice guidelines
potentially harmful? A randomized comparison of the effect of nonspecific versus specific
36. guidelines on physician decision making. Health Serv Res. 2000;3​ 4(7):1​ 429.
Woolf S, Schünemann HJ, Eccles MP, Grimshaw JM, Shekelle P. Developing clinical practice
guidelines: types of evidence and outcomes; values and economics, synthesis, grading, and
presentation and deriving recommendations. Implementation Science. 2012;7​ (1):​1‑12.
Tonelli MR. In defense of expert opinion. Acad Med. 1999;​74(11):1​ 187‑92.
Bright TJ, Wong A, Dhurjati R, Bristow E, Bastian L, Coeytaux RR, et al. Effect of clinical
decision-support systems: a systematic review. Ann Intern Med. 2012;​157(1):2​ 9‑43.
Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical practice using clinical
decision support systems: a systematic review of trials to identify features critical to success.
BMJ. 2005;​330(7494):7​ 65.
Goud R, van Engen-Verheul M, de Keizer NF, Bal R, Hasman A, Hellemans IM, et al. The effect of
computerized decision support on barriers to guideline implementation: A qualitative study
in outpatient cardiac rehabilitation. Int J Med Inform. 2010;​79(6):4​ 30‑7.
Moja L, Kwag KH, Lytras T, Bertizzolo L, Brandt L, Pecoraro V, et al. Effectiveness of Computer-
ized Decision Support Systems Linked to Electronic Health Records: A Systematic Review and
Meta-Analysis. Am J Public Health. 2014;​104(12):​E12-E22.
Ash JS, Sittig DF, Campbell EM, Guappone KP, Dykstra RH. Some unintended consequences of
clinical decision support systems. AMIA Annu Symp Proc. 2007:​26‑30.

16

General introduction and thesis outline

37. Roshanov PS, Fernandes N, Wilczynski JM, Hemens BJ, You JJ, Handler SM, et al. Features Chapter 1
38. of effective computerised clinical decision support systems: meta-regression of 162 ran-
39. domised trials. BMJ. 2013;​346:f​ 657.
40. Medlock S, Wyatt JC, Patel VL, Shortliffe EH, Abu-Hanna A. Modeling information flows in
41. clinical decision support: key insights for enhancing system effectiveness. J Am Med Inform
Assoc. 2016.
Ellrodt G, Cook DJ, Lee J, Cho M, Hunt D, Weingarten S. Evidence-based disease management.
JAMA. 1997;​278(20):1​ 687‑92.
Meinel LA, Stolpen AH, Berbaum KS. Breast MRI lesion classification: Improved performance
of human readers with a backpropagation neural network computer-aided diagnosis (CAD)
system. Journal of magnetic …. 2007(Query date: 2014-01-02).
Priddy KL, Keller PE. Artificial neural networks: An introduction: books.google.com; 2005.

17



Chapter 2

Frequency and risk factors for under-
and over-treatment in stroke prevention

for patients with non-valvular atrial
fibrillation in general practice

D.L. Arts, S. Visscher, W. Opstelten, J.C. Korevaar,
A. Abu-Hanna, H.C.P.M. van Weert

Chapter 2

A bstract

Aims
To determine adequacy of antithrombotic treatment in patients with non-valvular atrial
fibrillation. To determine risk factors for under- and over-treatment.

Design
Retrospective, cross-sectional study of electronic health records from 36 Dutch general
practitioners in 2008.

Subjects
Primary care physicians (n = 36) and patients (n = 981) aged 65 years and over.

Main outcome measures
Rates of adequate, under and over-treatment, risk factors for under and over-treatment.

Results
Of the 981 included patients with a mean of age 78, 18% received no antithrombotic
treatment (under-treatment), 13% received antiplatelet drugs and 69% received oral
anticoagulation (OAC). Further, 43% of the included patients were treated adequately,
26% were under-treated, and 31% were over-treated. Patients with a previous ischaemic
stroke were at high risk for under-treatment (OR 2.4, CI: 1.6-3.5), whereas those with
contraindications for OAC were at high risk for over-treatment (OR 37.0, CI: 18.1-79.9).
Age over 75 (OR 0.2, CI: 0.1-0.3]), diabetes (OR 0.1, CI: 0.1-0.3), heart failure (OR 0.2, CI:
0.1-0.3), hypertension (OR 0.1, CI: 0.1–0.2) and previous ischaemic stroke (OR 0.04, CI:
0.02-0.11) protected against over-treatment.

Conclusions
In general practice, CHADS2-criteria are being used, but the antithrombotic treatment of
patients with atrial fibrillation frequently deviates from guidelines on this topic. Patients
with previous stroke are at high risk of not being prescribed OAC. Contraindications for
OAC, however, seem to be frequently overlooked.

20

Frequency and risk factors for under- and over-treatment in stroke prevention for patients Chapter 2

I ntrod u ction

Every patient over the age of 65 with atrial fibrillation (AF) existing over 48 hours, needs
anti-thrombotic treatment, as AF increases the risk of stroke fivefold in untreated patients
[1]. Moreover, stroke caused by AF is also associated with increased severity [2], thus
leading to increased mortality and disability. Stroke risk reduction is currently achieved
by administering antithrombotic therapy to patients with AF. Antithrombotic therapy with
oral anticoagulation (OAC) decreases the risk of stroke by approximately 60% and by
approximately 20% with antiplatelet drugs in patients with AF [3, 4].
Because OAC increases the risk of major bleeding, current national and international
guidelines on AF incorporate stroke risk stratification schemes to calculate stroke risk,
balance benefits and risks of OAC-treatment, and recommend appropriate therapy [5,
6]. The CHADS2 score is a commonly used stroke risk stratification scheme, which is
calculated by assigning 1 point each for the presence of congestive heart failure, hyper-
tension, age 75 years or older, and diabetes mellitus; and by assigning 2 points each for
history of stroke or transient ischaemic attack (TIA) [7, 8]. Thromboprophylaxis with OAC
is recommended in patients with moderate to high stroke risk (CHADS2 score ≥ 2), while
antiplatelet therapy is recommended for patients with low risk [3, 5, 7].
Former hospital based studies showed that OAC prescription rates for high-risk patients
vary from 52% to 67% [9, 10]. A study among Dutch general practitioners (GPs), inter-
nists, and cardiologists confirmed these findings of secondary care studies, albeit with
somewhat higher compliance rates (67-72%)[11]. Because general practitioners in the
Dutch healthcare system provide longitudinal and integral care to their patients, they are
in a good position to provide a patient-tailored treatment.
We report on adequacy of antithrombotic treatment for patients over 65 years old, using
routinely collected data from the GP’s electronic healthcare records (EHR). Furthermore,
we aim to identify risk factors for inadequate treatment.

M ethods

Data collection
Data were extracted from EHR of general practices that participated in the Nether-
lands Information Network of General Practice (LINH) in 2008 [12]. The LINH forms a
geographically well-distributed network of GPs in The Netherlands. The LINH patient
population has a stable size and its age distribution and gender ratios are similar to
those of the general Dutch population. The database holds International Classification
of Primary Care (ICPC) coded longitudinal data on morbidity, prescribing, and referrals
of about 340,000 individuals. For this study, practices were only eligible for extraction

21

Chapter 2

when their data met the necessary quality requirements, which pertained to the number
of records present per individual patient, diagnoses present per individual patient, and
number of registered ICPC codes (60% of all patient contacts). The number of ICPC codes
per patient was compared to similar practices and counts of previous years to ensure
completeness of the registration. These data had to be available for a period of at least
3 years.

Patients
Characteristics were extracted for all patients aged 65 years and older who suffered from
AF at the end of 2008. Medication prescriptions, including OAC and antiplatelet drugs,
relevant comorbidities and contraindications for OAC were selected from the database.
Patients with valvular abnormalities were excluded. Of all patients with atrial fibrillation,
those without any antithrombotic medication were identified using R [13]. For these
patients, correctness of the diagnosis of atrial fibrillation was checked with the treating
GP and patients with wrongly recorded atrial fibrillation were excluded.

Ethics Statement
In the Netherlands, there is no need to obtain consent when only registry data obtained
from routine care and without patient identifying information are used, as is stated in
the selection criteria for the Medical Research Involving Human Subjects Act (WMO)[14].

Treatment evaluation
In order to evaluate treatment adequacy and under and over-treatment, we first deter-
mined the recommended treatment for every patient according to the CHADS2 stroke
risk stratification scheme [8, 15]. If contraindications for OAC were present, the recom-
mended treatment was antiplatelet drugs. If no contraindications were present, the
CHADS2 score was calculated using ICPC-coded diagnoses. The recommended treatment
was antiplatelet drugs for scores < 2, and OAC for scores ≥ 2. We then compared the
recommended treatment with the actual treatment. We classified adequacy of treatment
as follows: patients who received no antithrombotic treatment and patients with CHADS2
scores ≥ 2 who received only antiplatelet drugs were classified as under-treated. Patients
with contraindications for OAC and/or with CHADS2 scores < 2 who were treated with
OAC were included in the over-treatment group. The remaining patients were classified
as adequately treated.

Statistical analysis
Variables were initially selected based on the HAS-BLED score [16]and our clinical
judgement of their relevancy to adequacy of treatment: conditions incorporated in the
CHADS2-score, contraindications for OAC and conditions that might be responsible for

22

Frequency and risk factors for under- and over-treatment in stroke prevention for patients Chapter 2

withholding indicated treatment or other conditions than atrial fibrillation that warranted
OAC treatment. These variables were: age, gender, diabetes, hypertension, congestive
heart failure, ischaemic heart disease, previous stroke, epilepsy, Parkinson’s disease,
cognitive impairment, risk of falling (history of falling or use of sedatives), established
contraindications for OAC (haemorrhagic stroke, history of large bleeds, kidney or liver
failure (including alcoholism), coagulation diseases), deep venous thrombosis or pulmo-
nary embolism and the use of heparins. We were not able to accurately assess labile
INR’s and actual hypertension > 160 mm Hg. Parkinson’s disease, epilepsy and risk of
falling were clustered into one variable, as were contraindications for OAC. We compared
these variables within the three treatment outcome groups and calculated chi-squares
to test for association.
To identify independent associations with adequate treatment we applied multivariate
logistic regression in which each treatment outcome group was compared against the
remaining groups. We used backward stepwise variable elimination based on the Akaike
Information Criterion, and checked for presence of collinearity using the variance infla-
tion factor [17]. The model’s accuracy was measured using the standard bootstrap proce-
dure with 100 bootstrap samples and the percentile bootstrap confidence intervals. All
analyses were performed with the R language [13] using the RMS library for R [18]. The
Area under the curve (AUC) was determined to assess the model’s discriminating power.
The AUC ranges from 0.5 (no discrimination) to a theoretical maximum of 1. Perfect dis-
crimination corresponds to an AUC of 1 and is achieved if the scores for all the cases are
higher than those for all the non-cases, with no overlap.

R es u lts

Patient characteristics
After selecting practices with the required availability of three years of complete data, 36
general practices with 148,528 patients remained. This resulted in a total number of 981
patients with AF. Mean age was 78 years and 46% were men. CHADS2 score was 1.9 on
average, with only 2 patients scoring 6 points. The majority (59%) had a CHADS2 score ≥
2 and 69% were treated with OAC (Figure 1). More than 16% of all included patients had
one or more contraindications for OAC use. One or more comorbidities were present in
81% of patients, with hypertension being the most prevalent (57%) (Table 1).

Treatment
A total of 420 patients (43%) were treated adequately; 251 patients (26%) were under-
treated, 172 out of these 251 (69%) patients did not receive any antithrombotic treat-
ment and 79 (31%) patients had a CHADS2 scores ≥ 2 but were on antiplatelet drugs.

23

Chapter 2

CHADS 6

CHADS 5

CHADS 4

CHADS 3

CHADS 2

CHADS 1

CHADS 0

0 50 100 150 200 250 300 350

Under-treatment Over-treatment Treatment according to guideline

Figure 1. Number of patients and proportions of under- and over-treatment per CHADS2 score in
patients with atrial fibrillation.

A total of 310 patients (32%) were over-treated, among whom 267 (86%) patients had
CHADS2 scores < 2, but were on OAC (table 1). Of all patients treated with OAC, almost
8% had contraindications.

Risk factors
Table 2 shows factors associated with treatment adequacy. No evidence for collinearity
between variables was discovered. Presence of contraindications for OAC was the only
risk factor that significantly increased the chance of over-treatment. Independent protec-
tive factors against over-treatment were age > 75, presence of diabetes, heart failure,
hypertension, and previous ischaemic stroke or TIA (all CHADS2 criteria). Previous stroke
or TIA increased the risk for under-treatment, as did heart failure to a lesser degree. Fac-
tors that significantly decreased the chance of under-treatment were not found.

D isc u ssion

In this analysis of GPs’ electronic healthcare records, we have shown antithrombotic
treatment in patients with non-valvular AF to be inadequate. Over half of patients did not
receive the recommended treatment. All CHADS2 criteria (age > 75, diabetes, heart failure,
hypertension, previous stroke) independently reduced the risk of over-treatment, as was
to be expected. Contraindications for OAC however increased this risk. Previous stroke or

24

Frequency and risk factors for under- and over-treatment in stroke prevention for patients

Table 1. Population and proportions per treatment outcome.

n Perc. n Perc. n Perc. n Perc.
Adequate
Under- treatment*
420 42.81%
All treatment Over-treatment 114 27.14% Significance
306 72.86%
Overall 981 251 25.59% 310 31.60% 167 39.76% <0.001
405 96.43% <0.001
Age 65 - 75 383 39.04% 88 35.06% 181 58.39% 142 33.81% <0.001 Chapter 2
158 37.62% <0.001
Age > 75 598 60.96% 163 64.94% 129 41.61% 92 21.90% <0.001
306 72.86% <0.001
Men 448 45.67% 110 43.82% 171 55.16% 60 14.29% 0.09
63 15.00% <0.001
Co-morbidity present 792 80.73% 211 84.06% 175 56.45% 101 24.05% <0.001
4 0.95% 0.75
Diabetes 217 22.12% 52 20.72% 23 7.42% 0 0.00% 0.65
0 0.00% <0.001
Heart failure 303 30.89% 85 33.86% 33 10.65% 1 0.24% 0.13
0 0.00% <0.001
Ischaemic heart disease 184 18.76% 40 15.94% 52 16.77% 1 0.24% 0.93
2 0.48% 0.23
Hypertension 756 77.06% 145 57.77% 105 33.87% 0.38
<0.001
Ischaemic stroke 120 12.23% 50 19.92% 10 3.23%

Falling, epilepsy, Parkinson’s 150 15.29% 42 16.73% 45 14.52%

Sedative use 228 23.24% 53 21.12% 74 23.87%

Contraindications for OAC *,** 82 8.36% 26 10.36% 52 16.77%

Haemorrhagic stroke 2 0.20% 1 0.40% 1 0.32%

Large bleeding 32 3.26% 5 1.99% 27 8.71%

Liver failure 3 0.31% 1 0.40% 1 0.32%

Kidney failure 1 0.10% 1 0.40% 0 0.00%

Coagulation disease 5 0.51% 1 0.40% 3 0.97%

Cognitive impairment 41 4.18% 17 6.77% 22 7.10%

CHADS2< 2 407 41.49% 86 34.26% 267 86.13% 54 12.86% <0.001
CHADS2 ≥ 2 574 58.51% 165 65.74% 43 13.87% 366 87.14% <0.001

*According to the Dutch GP guideline.
** Values displayed in this row reflect one or more contraindications.

Table 2. Models for over- and under-treatment.

1. Model for over-treatment

Variable OR 95% CI AUC††
0.13 - 0.89
Age >75 0.19† 0.08 - 0.28
0.09 - 0.25
Diabetes 0.14† 0.10 - 0.25
0.02 - 0.20
Heart failure 0.15† 0.11
18.11 - 79.93
Hypertension 0.14†

Previous stroke/TIA 0.04†

Contraindications for OAC* 36.98†

Removed stepwise**

Female sex 0.71 0.50 - 1.02
0.53 - 1.33
Ischaemic heart disease 0.84 0.64 - 1.84
0.78 - 1.83
Neurological disease*** 1.09

Sedative use 1.20

25

Chapter 2

Table 2. Models for over- and under-treatment. (continued)

2. Model for under-treatment OR 95% CI 3.48 AUC††
Variable 2.35† 1.57 - 0.54
Previous stroke/TIA
Removed stepwise** 1.10 0.81 - 1.51
Age >75 1.11 0.82 - 1.50
Female sex 0.91 0.62 - 1.31
Diabetes 1.45 1.03 - 2.03
Heart failure 0.76 0.50 - 1.14
Ischaemic heart disease 1.19 0.86 - 1.65
Hypertension 1.07 0.70 - 1.60
Neurological disease*** 1.22 0.72 - 2.03
Contraindications for OAC* 0.76 0.52 - 1.09
Sedative use

* Haemorrhagic stroke, large bleed in history, liver failure, kidney failure, coagulation disease,
cognitive impairment (dementia, psychosis).

** As selected by AIC. Odds ratios are reported for the complete model.
*** Parkinson’s disease, epilepsy, history of falling.
† P < 0.01
†† Area under the curve or concordance-index: a measure of discriminating power of the model.

TIA increased the chance for under-treatment and reduced the change of over treatment.
Female sex is associated with lower OAC rates in other studies [15, 18], but our data do
not support these findings.
Although our study was performed in primary care, demographic characteristics of our
study population were comparable with other studies on this topic [3, 19, 20]. Mean age
was slightly higher, which can be attributed to the fact that we only included patients of
65 years or older. Comorbidity rates were comparable, with the exception of ischaemic
heart disease, which showed markedly lower prevalence (19% vs. 28%) [11, 20, 21]. This
can be explained by the fact that other studies were mainly performed in secondary care.
We found antithrombotic treatment adequacy was low, but comparable to other studies
on antithrombotic treatment in patients with AF [3, 11, 19, 20]. These studies compared
stroke risk scores to antithrombotic prescriptions and did not take contraindications
in individual patients into account. This led to patients being classified as adequately
treated when they were actually over-treated due to contraindications for OAC, thus
leading to higher rates of adequate treatment.
Our data suggest that contraindications for OAC are not always evaluated, which we
ascribe to the fact that contraindications can be easily overlooked in busy everyday
practice. We found no significant independent associations between treatment adequacy
and sedative use, history of epilepsy/falling, or ischaemic heart disease. Our results sug-
gest previous stroke/TIA increases the risk of under-treatment, but this could be due to

26

Frequency and risk factors for under- and over-treatment in stroke prevention for patients Chapter 2

problems in the registration of contraindications (mainly stroke and renal failure). It is
possible that some patients actually suffered from a haemorrhagic stroke, in which case
the treatment would have been adequate.

Strengths and limitations
We studied antithrombotic treatment of atrial fibrillation in real life, by using routinely
registered data. These data were subjected to several rigorous quality checks, but our
evaluation was slightly hampered, by a lack of detail and incompleteness of diagnostic
coding for two diagnoses: Diagnoses of cerebral haemorrhage and ischaemic stroke
are both classified as sub-codes of the ICPC code K90, but in the majority of cases,
no sub-codes were specified. Given the much greater prevalence of ischaemic stroke,
we classified all unknown cases of stroke as ischaemic. Also, renal failure appeared
to be underreported, as we only found 2 cases of renal failure, which is lower than
expected in this population. This probably caused some misclassification that might
have led to inflation of reported under-treatment. On the other hand, the risk of being
over-treated due to the presence of contraindications for OAC is probably larger than
reported. Hypertension is both an indication, as well as a contra-indication for OAC, as it
is incorporated in both the CHADS2 and the HASBLED score. Because we were unable to
assess the exact blood pressure, we considered treated hypertension as an indication.
Instable INR’s are not coded in the GP’s EHR as treatment with OAC is monitored by
a separate service in the Netherlands. This might have led to slight over reporting of
overtreatment.

Implications for clinical practice and research
By including not only comorbidities and medication, but also contraindications for indi-
vidual patients, we have come close to a real-life treatment scenario. GPs should be more
aware of contraindications for OAC, as stroke risk does not justify the added risk of major
bleeding. They should also be aware of the presence of a previous ischaemic stroke, as
this is always a reason to prescribe OAC. Lastly, we would like to stress the importance of
accurate diagnostic coding and electronic record keeping. This will allow for more accu-
rate research and will enable the use of decision support systems to support caretakers
in applying complex guidelines. Decision support systems might play a vital role in the
improvement of stroke prevention, as has been stated and validated in other areas of
medicine [21, 22]. Further research should be done to assess the use of these systems in
medical practice, as clues from the medical history seem to be easily overlooked.

27

Chapter 2

Conclusions
In Dutch general practice, the antithrombotic treatment of patients with atrial fibrilla-
tion frequently deviates from guidelines on this topic. Patients with previous stroke are
at high risk of not being prescribed OAC. Contraindications for OAC, however, seem to
be frequently overlooked, resulting in over-treatment, which shows tailoring treatment
remains important.

28

Frequency and risk factors for under- and over-treatment in stroke prevention for patients

R eferences

1. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Chapter 2
2. Framingham Study. Stroke. 1991;​22(8):9​ 83‑8.
3. Lin HJ, Wolf PA, Kelly-Hayes M, Beiser AS, Kase CS, Benjamin EJ, et al. Stroke severity in atrial
4. fibrillation. The Framingham Study. Stroke. 1996;​27(10):​1760‑4.
5. Hart RG. Meta-analysis Antithrombotic Therapy to Prevent Stroke in Patients who have noval-
6. vular atrial fibrillation. Ann Intern Med. 2007.
7. Ogilvie IM, Welner SA, Cowell W, Lip GY. Ischaemic stroke and bleeding rates in ‘real-world’
8. atrial fibrillation patients. Thromb Haemost. 2011;1​ 06(1):​34‑44.
9. Camm AJ, Kirchhof P, Lip GY, Schotten U, Savelieva I, Ernst S, et al. Guidelines for the man-
10. agement of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the
11. European Society of Cardiology (ESC). Europace. 2010;1​ 2(10):1​ 360‑420.
Fuster V, Ryden LE, Cannom DS, Crijns HJ, Curbs AB, Ellenbogen KA, et al. ACC/AHA/ESC 2006
12. guidelines for the management of patients with atrial fibrillation - Executive summary. J Am
Coll Cardiol. 2006;4​ 8(4):​854‑906.
13. Nieuwlaat R, Capucci A, Lip GY, Olsson SB, Prins MH, Nieman FH, et al. Antithrombotic treat-
14. ment in real-life atrial fibrillation patients: a report from the Euro Heart Survey on Atrial
15. Fibrillation. Eur Heart J. 2006;​27(24):3​ 018‑26.
Gage BF, Waterman AD, Shannon W, Boechler M, Rich MW, Radford MJ. Validation of Clinical
Classification Schemes for Predicting Stroke. JAMA: The Journal of the American Medical As-
sociation. 2001;​285(22):​2864‑70.
Ogilvie IM, Welner SA, Cowell W, Lip GY. Characterization of the proportion of untreated
and antiplatelet therapy treated patients with atrial fibrillation. Am J Cardiol. 2011;​108(1):​
151‑61.
Nieuwlaat R, Capucci A, Camm AJ, Olsson SB, Andresen D, Davies DW, et al. Atrial fibrillation
management: a prospective survey in ESC Member Countries - The Euro Heart Survey on
Atrial Fibrillation. Eur Heart J. 2005;2​ 6(22):​2422‑34.
Dinh T, Nieuwlaat R, Tieleman RG, Buller HR, van Charante NA, Prins MH, et al. Antithrombotic
drug prescription in atrial fibrillation and its rationale among general practitioners, internists
and cardiologists in The Netherlands--The EXAMINE-AF study. A questionnaire survey. Int J
Clin Pract. 2007;​61(1):2​ 4‑31.
Verheij RA VDC, Abrahamse H, Davids R, Van den Hoogen H, Braspenning J, Van Althuis T.
Landelijk Informatienetwerk Huisartsenzorg. Feiten en cijfers over huisartsenzorg in Ned-
erland. Utrecht/Nijmegen: NIVEL/IQ; 2009 [Available from: http://www.nivel.nl/oc2/page.
asp?pageid=12673.
Team RDC. R: A Language and Environment for Statistical Computing. 2011.
(CCMO) CCoRIHS. WMO: Does your study have to be reviewed? 2013 [cited 2013 05-03-
2013]. Available from: http://www.ccmo-online.nl/main.asp?pid=10&sid=30&ssid=51.
Fuster V, Ryden LE, Cannom DS, Crijns HJ, Curtis AB, Ellenbogen KA, et al. ACC/AHA/ESC 2006
Guidelines for the Management of Patients with Atrial Fibrillation: a report of the American
College of Cardiology/American Heart Association Task Force on Practice Guidelines and the
European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Re-
vise the 2001 Guidelines for the Management of Patients With Atrial Fibrillation): developed
in collaboration with the European Heart Rhythm Association and the Heart Rhythm Society.
Circulation. 2006;1​ 14(7):​e257‑354.

29

Chapter 2

16. Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. A novel user-friendly score
(HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro
17. Heart Survey. Chest. 2010;1​ 38(5):1​ 093‑100.
18. Heiberger RM. HH: Statistical Analysis and Data Display: Heiberger and Holland. 2011.
19. Jr FEH. rms: Regression Modeling Strategies. 2012.
Gorin L, Fauchier L, Nonin E, Charbonnier B, Babuty D, Lip GY. Prognosis and guideline-adher-
20. ent antithrombotic treatment in patients with atrial fibrillation and atrial flutter: implications
of undertreatment and overtreatment in real-life clinical practice; the Loire Valley Atrial
21. Fibrillation Project. Chest. 2011;1​ 40(4):​911‑7.
22. DeWilde S, Carey IM, Emmas C, Richards N, Cook DG. Trends in the prevalence of diagnosed
atrial fibrillation, its treatment with anticoagulation and predictors of such treatment in UK
primary care. Heart. 2006;​92(8):​1064‑70.
de Lusignan S, van Vlymen J, Hague N, Thana L, Dzregah B, Chan T. Preventing stroke in people
with atrial fibrillation: a cross-sectional study. J Public Health (Oxf). 2005;2​ 7(1):8​ 5‑92.
Wess ML, Schauer DP, Johnston JA, Moomaw CJ, Brewer DE, Cook EF, et al. Application of a
decision support tool for anticoagulation in patients with non-valvular atrial fibrillation. J
Gen Intern Med. 2008;2​ 3(4):4​ 11‑7.

30

Chapter 3

Reasons for intentional guideline non-
adherence: a systematic review

D.L. Arts, A.G. Voncken, S.K. Medlock, A. Abu-Hanna, H.C.P.M. van Weert

Chapter 3

A bstract

Background
Reasons for intentional non-adherence to guidelines are largely unknown. The objective
of this systematic review was to gain insight into and categorize reasons for intentional
non-adherence and their validity. Non-adherence might be a conscious choice by either
the clinician or the patient, and is not influenced by external factors (e.g. lack of knowl-
edge or resources). We use the term intentional non-adherence to describe this class of
reasons for not following guideline recommendations.

Methods
Two independent reviewers examined MEDLINE citations for studies that investigated
reasons for guideline non-adherence. The obtained articles were assessed for relevance
and quality. Our search yielded 2912 articles, of which 16 matched our inclusion criteria
and quality requirements. We planned to determine an overall ranking of categories of
non-adherence.

Results
Seven studies investigated clinical reasons and performed adjudication, while nine
studies did not perform adjudication. Non-adherence varied between 8.2% and 65.3%.
Meta-analysis proved unfeasible due to heterogeneity of study methodologies. The per-
centage of reasons deemed valid by adjudication ranged from 6.6% to 93.6%. Guideline
non-adherence was predominantly valid; contra-indications and patient preference were
most often reported as reasons for intentional non-adherence.

Conclusion
We found a wide range of rates of non-adherence to clinical guidelines. This non-
adherence is often supported by valid reasons, mainly related to contra-indications and
patient preference. Therefore, we submit that many guideline deviations are intentional
and these deviations do not necessarily impact quality of care.

32

Reasons for intentional guideline non-adherence: a systematic review

I ntrod u ction

In recent years, the scientific community has shown an increased interest in clinical prac- Chapter 3
tice guidelines, and practitioners adherence to such guidelines. Guidelines can reduce
inappropriate variation in medical practice, thereby improving quality of care [1, 2] and
reducing costs[3]. Guidelines are increasingly used for quality management and health
care policy. Another use of guidelines is in remuneration of physicians by healthcare
insurers, where reaching a certain level of guideline adherence qualifies for additional
remuneration. However, adherence to guidelines varies greatly: several studies report
adherence rates of 10% to 80%[4, 5]. Most research into reasons for non-adherence
has been performed in the behavioral field. Cabana et al[6] describe multiple reasons
for non-adherence in their systematic review (e.g., Lack of Awareness, Lack of Outcome
Expectancy and Guidelines Factors) and divided these into three different categories:
Knowledge, Attitudes, and Behavior. Other studies found that guideline adherence is
related to characteristics of the clinician, guideline, system, and implementation[7, 8].
Many attempts[7, 9, 10] have been made to improve these circumstances linked to
guideline non-adherence, but even with support from leaders in the medical field,
availability on demand, clinical decision support systems (CDSS), and financial rewards,
non-adherence remains substantial[9, 10]. Some of this residual non-adherence is attrib-
utable to a conscious decision by the clinician or patient to not follow the guideline. In
this study we investigate these reasons for non-adherence. We use the term intentional
non-adherence to describe this class of reasons for not following guideline recommen-
dations. Unintentional non-adherence can occur due to external factors (such as lack of
knowledge about the guideline’s recommendations) or error on the part of the clinician
or patient (such as forgetting to prescribe or take a medication). In this paper we will not
investigate this type of non-adherence.
To study intentional reasons for non-adherence we consider documentation of an explicit
reason for not following the guideline to be evidence that the decision was intentional.
These documented reasons are the focus of this review. In this study we aim to categorize
and quantify reasons for intentional non-adherence and report on their appropriateness
(as defined by peers), if applicable. We expect our results to contribute not only to future
guideline development, but also to aid in assessing the validity of modern-day qual-
ity indicators. Finally, clinical decision support systems (CDSS) require that guidelines
explicitly mention exceptions, in order to be able to adequately apply a digital guideline
to every patient. Our study could make guideline developers aware of different types of
exceptions, and thus enable them to more effectively document them, developing more
differentiated guidelines and better CDSS.

33

Chapter 3

M ethods

We reviewed the existing literature with the objective of assessing reasons for intentional
non-adherence to clinical practice guidelines.

Search strategy and study selection
We searched MEDLINE using the following query: (guideline adherence [MeSH Major
Topic] OR practice guidelines as topic [MeSH Major Topic]) AND (reason OR reasons OR
perception OR perceptions OR attitude OR attitudes OR view OR views OR barrier OR
barriers OR facilitator OR facilitators). We applied the limits: “Humans”, “English”, and
“has abstract” and searched until October 1, 2014. Two independent reviewers (AV, DA)
individually assessed the resulting titles and abstracts and selected papers that fit the in-
and exclusion criteria described below. In cases where the reviewers disagreed, a third
reviewer (HW) was consulted. Selected full-text articles were assessed for relevance.
References and “related articles” of the selected articles were explored for potential
inclusion.
Inclusion criteria were:
• Reasons for intentional non-adherence to clinical guidelines were described.
Exclusion criteria were:
• Reasons for non-adherence were not collected within three months.
• Study did not assess actual clinical performance (i.e., vignette studies).
• A clear reference to the studied guideline was not provided.
• Data-collection was not explicitly described
• Study was of insufficient methodological quality (according to the methodological

criteria described below).
Articles concerning non-adherence to quality indicators, decision rules, clinical decision
reminders, or triage decisions were eligible if these were a derivative of a guideline.
There was no restriction regarding specialty or case-mix.

Exclusion of articles based on methodology
Two reviewers (AV, DA) assessed the methodological quality of the selected articles using
the “Dutch Cochrane checklists for assessing Cohort studies” [11]. This tool allows the
user to make an assessment of methodological quality; it assesses several key method-
ological aspects of a study, including, “population definition”, “risk of selection bias”
and “follow-up duration”. Both reviewers judged the articles to be of either sufficient
(all criteria of the checklist were met) or insufficient (one or more criteria of the checklist
were not met) methodological quality. These assessments were compared and disagree-
ments were resolved during a consensus meeting. Articles deemed to be of insufficient
methodological quality by both reviewers were excluded.

34

Reasons for intentional guideline non-adherence: a systematic review Chapter 3

Data extraction and category creation
We collected the following characteristics of the included studies: study design, year,
site, setting, country, target disorder, information technology used (if any), intervention,
type of guideline and its distribution, rates of non-adherence, and reasons for non-
adherence. To rank categories of reasons for non-adherence for each study, both review-
ers noted categories of reasons for non-adherence and their reported importance (rank)
in a structured form. Results were compared and merged into one result set. Different
but often overlapping categories were used in the included studies to classify reasons
for non-adherence. To be able to compare reported reasons, we created categories by
way of induction, i.e., “the process by which themes and categories emerge from the
data through the researcher’s careful examination and constant comparison.” [12]. Since
most articles did not report proportions of reasons for non-adherence, we performed
classification by ranking relative occurrence of a reason in an article by two reviewers
(DA, AV). Finally, we collected percentages of valid reasons for articles with adjudication
(judgment of validity by peers). The PRISMA statement was used as a guideline for this
systematic review [13].

Statistical analysis
We used an independent samples Krukis-Wallis test to test for association between
adherence rates and study/guideline characteristics: setting, type of intervention, ap-
plication of adjudication, and patient selection by use of exclusion criteria. To test for
association with the use of exclusion criteria we categorized studies into three groups:
with none, some, or extensive exclusion criteria. The “some” group was defined as one to
five exclusion criteria, the “extensive” group as more than five.

R es u lts

Out of the 2912 titles and abstracts screened, we selected 147 full-text articles based
on title and abstract (Figure 1). Of these 147 full articles, 119 did not meet our inclusion
criteria, leaving 28 articles. Reference checking did not provide any additional articles.
Of the remaining 28 articles, twelve articles were excluded because the methodology
was inadequate. Sixteen studies were retained for analysis; seven studies[14-20] applied
adjudication and nine studies[21-29] did not. The included articles were heterogeneous
in almost every aspect: setting, site, country (health care system), and target disorder
(table 1), as well as method of reporting; categories of reasons for non-adherence were
defined differently and frequencies of reasons were often not explicitly listed. None of
the include articles explicitly described using a form of information technology. Due
to this heterogeneity, meta-analysis was not feasible. All studies had a cohort design:

35

Chapter 3

Identification 2912 records identified
through MEDLINE searching.

2912 records after duplicates were removed

Screening 2912 screened on title 2765 records excluded on title
and abstract and abstract

Eligibility 0 additional records identified 147 full-text articles 131 full-text articles excluded:
through other sources. assessed for eligibility. • Did not asses reasons for non-

adherence: 119
• Inadequate methodology: 12

Included 16 studies included in
qualitative synthesis.

Figure 1. PRISMA flow diagram.

nine retrospective chart analyses, six prospective chart analyses, and one mixed retro-
spective and prospective chart analysis (Table 1).

The induction process resulted in five categories for intentional non-adherence: “pa-
tient decision”, contra-indications”, “patient demographics”, “physician decision”, and
“other” (table 2). Further granularization was not possible due to the lack of detail in the
categories described in the studies. The ranks of each category per article are shown
in table 3. The category “contra-indications” was mentioned most, followed by “patient
decision,” “other,” “patient demographics,” and finally “physician decision.” See box 1 for
hypothetical examples of the induced categories.

36

Table 1. Overview of studies

Author Adj Year Study design Setting Site Country Target Exclusion Intervention Guideline or Guideline
disorder criteria otherwise distribution

Owen [18] Yes 2002 Retrospective Secondary care Psychiatry inpatient USA Schizophrenia Some Antipsychotic drugs Clinical National
cohort ward performance
measure based
on guidelines

Irwin [16] Yes 2003 Prospective Secondary care Cardiology department Canada Bradycardia None Pacemaker Guideline National
Cohort implantation

Ardery [14] Yes 2007 Retrospective Primary care General Practice USA Hypertension Multiple Treatment according to National
cohort comorbid guidelines (therapeutic Guideline
Multiple illnesses and diagnostic)
Reasons for intentional guideline non-adherence: a systematic review(treatment,
Persell [19] Yes 2010 Prospective cohort Secondary care37Internal medicine USA primary and None Primary and secondary Quality NS
outpatient clinic secondary
prevention) prevention measures

Kmetik [17] Yes 2011 Retrospective Secondary care General internal USA Coronary artery None Drug therapy Quality NS
cohort medicine and disease measures
cardiology outpatient
clinic

Pediatric cardiology Three pediatric Outpatient follow-up Standardized
outpatient clinic and assessment Clinical
Farias [15] Yes 2012 Prospective cohort Secondary care USA cardiac None Assessment NS
Referral for diagnostic and
disorders work-up for surgery Management National
Vancomycine Plan National
Uijl [20] Yes 2012 Retrospective Secondary and Neurology outpatient Netherlands Epilepsy None treatment
cohort tertiary care clinic None Guideline

Evans [24] No 1996 Prospective Tertiary care All inpatient wards USA Infections Guideline
cohort

Chapter 3

Chapter 3
38
Table 1. Overview of studies (continued)

Author Adj Year Study design Setting Site Country Target Exclusion Intervention Guideline or Guideline
disorder criteria otherwise distribution
Community National
Marras [27] No 1998 Prospective and Tertiary Care All inpatient wards Canada Acquired Multiple Treatment with
retrospective Pneumonia comorbid antibiotic according to Guideline Local/ Internal
illnesses guideline
Community Regional
Retrospective Acquired Non- Decision of admittance
chart review and Pneumonia candidates National
Halm [26] No 2000 physician survey Emergency Emergency Department USA for outpatient according to a decision Guideline
at admittance Department Breast cancer treatment rule (Pneumonia Local/ regional

Colorectal Severity Index) Internal
cancer National
Balasubra- No 2003 Prospective cohort Secondary care Oncology department UK Some Treatment with (neo-) Guideline In House
manian [22] Community adjuvant chemo- and
Acquired radiotherapy
Pneumonia
Oliveria [28] No 2004 Retrospective Secondary care Oncology (inpatient USA Not in HMO Referral to oncologist Guideline
cohort and outpatient) Diabetes for 1yr and subsequent
treatment with
Aortic valve chemotherapy
stenosis
Aujesky [21] No 2009 Prospective cohort Secondary care Oncology (inpatient USA Multiple Choice of treatment Guideline
and outpatient) comorbid site using the
illnesses Pneumonia Severity
Index

Cornali [23] No 2009 Retrospective Secondary care Post-acute geriatric Italy None Treatment according to
cohort Tertiary care setting (inpatient) USA guidelines (therapeutic Guideline
Secondary care Norway and diagnostic)
Cardiology (inpatient
Freed [25] No 2010 Retrospective and outpatient) None Aortic valve Guideline
cohort replacement
Oncology & urology
Stensvold No 2011 Retrospective (inpatient and Prostate cancer Multiple Surgery and/or Guideline
[29] cohort outpatient) radiotherapy

Reasons for intentional guideline non-adherence: a systematic review

Table 2. Induction of the categories of reasons for non-adherence

Induced category Source categories

Patient decision Family decision
Infrequent clinic visits
Low compliance
Refusal of treatment

Demographics Age
Gender
Ethnicity
Limited level of physical activity

Physician decision Physician decision (unspecified) Chapter 3

Contra-indications Abnormal findings (laboratory / physical)
High operative risk
Limited life expectancy
Intolerance for recommended treatment
‘Did not fit the guideline’ (as reported in manuscript)
Extensive comorbidities

Other Insurance constraints
Other (as reported in manuscript)

Table 3. Non-adherence, proportion of appropriate reasons and ranked categories for non-adher-
ence per study.

Author Non-adherence Appropriate Ranking categories**

Owen [18] 39.4% 80% 1) Contra-indications, Other

Irwin [16] 50.6% Large 1) Contra-indications 2) Demographics 3) Physician decision
proportion*
Ardery [14] 31.6% 7% 1) Contra-indications 2) Patient decision3) Other
Persell [19] 22.2% 94% Ranking was not possible.
Kmetik [17] 24.3% 93% 1) Contra-indications
Farias [15] 8.2% 50% 1) Other 2) Physician decision 3) Contra-indications

Uijl [20] 18.6% 26% 1) Contra-indications 2) Patient decision

Evans [24] 65.3% NA 1) Physician decision, Other

Marras [27] 19.5% NA 1) Contra-indications 2) Other

Halm [26] 56.4% NA 1) Patient decision 2) Contra-indications 3) Physician decision

Balasubra-manian [22] 18% NA 1) Other 2) Contra-indications, Demographics

Oliveria [28] 30% NA 1) Physician decision 2) Contra-indications 3) Patient decision,
Other

Aujesky [21] 20.3% NA 1) Contra-indications 2) Patient decision 3) Physician decision

Cornali [23] 17.7% NA 1) Contra-indications 2) Demographics 3) Patient decision

Freed [25] 42% NA 1) Contra-indications2) Physician decision 3) Patient decision,
Demographics 4) Other

Stensvold [29] 16.7% NA 1) Demographics 2) Contra-indications, Patient decision

*Appropriateness was evaluated, but no explicit percentage was mentioned.
** Categories were ranked by occurrence. If a category was not mentioned, it was not ranked.
Categories with equal numbers of occurrences were ranked together.

39

Chapter 3

Box 1. Examples of induced categories.
A 72-year-old male visits the clinician because of newly detected atrial fibrillation. He is known to
have diabetes and hypertension; both conditions are well regulated by medication. According to
the Dutch guideline for atrial fibrillation this patient should receive warfarin.
The patient explicitly asks for the new oral anticoagulants (NOACs), as he is in Spain during the
winter months. He does not want to have his INR monitored there, nor does he want to manage it
himself. The reason for prescribing NOACs would be classified as patient decision. If this patient
has a high risk of falling, due to for instance polyneuropathy and vertigo, the reason not to pre-
scribe any anticoagulants would have been classified as a contra-indication. Refraining from any
anticoagulant treatment because of old age would have been classified as a demographic reason
and not prescribing anticoagulants because of a terminal illness would be classified as physicians’
decision.

Non-adherence varied between 8.2% and 65.3%. Validity of reasons for adjudicated
studies varied between 6.6% and 93.6%. Of studies with adjudication, all but Ardery
[14] and Uijl [20] judged the majority of reasons for non-adherence as valid (Table 2).

D isc u ssion

The objective of this study was to categorize and quantify reported reasons for inten-
tional non-adherence and their validity. We found a wide range of non-adherence rates.
When reasons for guideline non-adherence were adjudicated, they were mostly judged
as valid. The main categories of reasons for non-adherence were “contra- indications”
and “patient decision.”

Strengths
To our knowledge, this is the first review that assesses reasons for intentional guideline
non-adherence. We reduced selection bias by reviewing every title, abstract, and article
with two independent reviewers. Checking relevant references and “related articles”
in MEDLINE did not result in any new articles, confirming the sensitivity of our search
strategy. Due to limited resources, our search was limited to MEDLINE which could have
resulted in the omission of relevant papers.

Limitations
An important limitation of this study is lack of matching methodologies used the included
articles. Due to the heterogeneity of the reviewed articles, comparing reasons for non-
adherence was difficult, and some overlap in our defined categories could exist. Many
articles did not specify what contra-indications were already mentioned in the guideline,
so we assumed that when “contra-indications” were given as a reason for non-adherence,

40

Reasons for intentional guideline non-adherence: a systematic review Chapter 3

these were not mentioned in the guideline. Furthermore, the process of induction is, by
definition, a possible source of bias, as the interpretation of categories can be subjective.

Limitations of included papers
Despite matching our inclusion criteria and passing our methodological checklist, there
are several limitations to the included studies that stand out. First and foremost, with one
exception none of the studies attempted to use a standardized approach for evaluating
guideline non-adherence. Furthermore, many studies were retrospective chart analyses,
and while studies with risk for severe recall bias were excluded, this design still leaves
room for bias, as opposed to the prospective study design.
The significant heterogeneity of the studies included in this paper is the most likely cause
of the wide range of non-adherence rates we found. Studies varied in setting, population,
and design. Other studies on this topic confirm these findings [30, 31]. Differences in
study design, particularly retrospective vs. prospective, are reported as important factors
for the large spread in adherence rates, but we did not find this association. Additionally,
no associations were observed for setting, guideline characteristics, and target disease.
We expected guidelines with extensive exclusion criteria to be associated with higher
adherence rates, but our statistical analysis could not confirm this. It should be noted
that the limited number of studies included in the analysis would have required large
differences to amount to a statistically significant association.

Reasons for non-adherence
Most studies on reasons for non-adherence either come from the behavioral field or were
written in opinion articles and narrative reviews. In the behavioral field, a meta-synthesis
of qualitative research by Cabana [6] describes categories such as “inapplicability to the
patient”, “inability to reconcile patient preferences with guideline recommendations”
and “lack of outcome expectancy”. This corresponds with our categories of “contra-
indications”, “patient decision” and “physician decision”. However, the category “patient
demographics” is not mentioned in Cabana’s framework, and could be classified as an
additional reason in the category Attitudes. Lugtenberg et al. [32] found that “lack of
agreement with guideline recommendations” was the most prominent barrier in applying
guidelines in general practice, a barrier that also corresponds with “physician decision “.
A second important barrier they found was “applicability to the patient”, which relates to
out categories “contra-indications” and “patient demographics”. Gurses describes, “ex-
ception ambiguity,” defined as “the ambiguity on whether benefits of applying a particular
guideline to a specific patient outweigh the potential risks and patient discomfort.”[33].
This corresponds with our categories “contra-indications” and “physician decision”. The
large proportion of the category “patient decision” raises the question of whether there

41

Chapter 3

should be more patient involvement in guideline development, as patients have shown
an increased preference to be involved in their own care process [34].

Validity of reasons
For all but two of the studies [14, 20] with adjudication, the majority of reasons were
adjudicated as valid. The large proportion of reasons that were adjudicated as valid was
expected, since intentional guideline non-adherence, by definition, indicates that a phy-
sician has considered the guideline in the context of the patient, and found a reason to
deviate from the advice. This underscores the value and validity of professional clinical
judgment when applying guidelines in daily practice. Thus, ideal guideline adherence
might not be 100% adherence, but may be much lower, 69% to 98% in the included
studies.
The high frequency of valid reasons indicates that guidelines can be improved. A study
on applicability of clinical practice guidelines on elderly patients with comorbidities
showed that only one-third of the guidelines adequately discussed issues related to pa-
tients with comorbidities[35, 36]. This also corresponds with Gurses’ “Exception ambigu-
ity” and Cabana’s “Applicability to Patient” [33, 6] and with our “Contra-indications” and
“Demographics” categories. It should be noted that situations where a guideline simply
isn’t applicable to a patient, although the patient in question matches the described
target population of the guideline, deviating from the guideline might be the only right
thing to do. A possible solution for this issue could be to list every possible exception
for a guideline. Another, more viable solution is application of clinical decision support
systems, as those can hide all but the relevant exceptions for the current patient. Further
research should be done to investigate the feasibility of these concepts.

Recommendations for future researchers and guideline developers
Meta-analysis proved unfeasible due to heterogeneity of study methodologies. We
therefore propose that future researchers take a more structured approach to evaluating
guideline adherence. Farias [15] described Standardized Clinical Assessment and Man-
agement Plan (SCAMP) as a quality-improvement initiative that guides clinical decision-
making to standardize the assessment and management of patients with a specific
disorder. This system collects data on clinical deviations (DEVs) and reasons provided by
caregivers. Evaluation of these reasons leads to improvement of the SCAMP, thus creating
a dynamic guideline. Moreover, this results in a structured dataset with guideline devia-
tions that can easily be analyzed. This method could be part of a guideline evaluation
framework that has predefined categories of reasons for non-adherence, and formal
methods for adjudication of these reasons. Such a framework could greatly increase
comparativeness of studies on guideline non-adherence and thus result in generalizable
concepts for guideline improvement.

42

Reasons for intentional guideline non-adherence: a systematic review Chapter 3

Implications for decision support systems
The structural documentation of exceptions can also improve decision support systems,
making these more capable to provide recommendations for patients with multiple
comorbidities. Currently, conflicts often arise between overlapping guidelines, which
cannot be resolved by a computer. Using methods to formalize guidelines, and more
specifically intentions of guidelines, could prove invaluable for the future of decision
support [37]. Our study underwrites the need for flexible guidelines and decision sup-
port systems as described by Latoszek-Berendsen et. al., that allow a physician to “take
a road that may not be completely according to the guideline, but within its spirit” [37,
38]. As we have shown that physicians have many justifiable reasons for straying from
a guideline recommendation. Lastly, we highly recommend future decision support sys-
tems implement functionality that allows for user friendly, though mandatory reporting
on reasons for guideline deviations by physician users.
Conclusion
In this study we found wide ranges of non-adherence rates to clinical guidelines. This
non-adherence is often intentional and supported by valid reasons, mainly related to
contra-indications and patient preference. Therefore, we submit that many guideline
deviations are intentional and justifiable, and these deviations do not necessarily impact
quality of care.

43

Chapter 3

R eferences

1. Andrews EJ, Redmond HP. A review of clinical guidelines. Brit J Surg. 2004;9​ 1(8):9​ 56‑64.
2. Worrall G, Chaulk P, Freake D. The effects of clinical practice guidelines on patient outcomes
3. in primary care: A systematic review. Can Med Assoc J. 1997;​156(12):1​ 705‑12.
4. Brown PD. Adherence to guidelines for community-acquired pneumonia - Does it decrease
cost of care? Pharmacoeconomics. 2004;2​ 2(7):​413‑20.
5. Grol R, Dalhuijsen J, Thomas S, in’t Veld C, Rutten G, Mokkink H. Attributes of clinical guide-
lines that influence use of guidelines in general practice: observational study. Brit Med J.
6. 1998;​317(7162):​858‑61.
7. McGlynn EA, Asch SM, Adams J, Keesey J, Hicks J, DeCristofaro A et al. The quality of health
care delivered to adults in the United States. The New England Journal of Medicine. 2003;​
8. 348(26):2​ 635‑45.
9. Cabana, Rand CS, Powe NR, Wu AW, Wilson MH. Why Don ’ t Physicians Follow guidlines, A
Framework for Improvement. October. 1999;0​ 718.
10. Gurses AP, Marsteller JA, Ozok AA, Xiao Y, Owens S, Pronovost PJ. Using an interdisciplinary
11. approach to identify factors that affect clinicians’ compliance with evidence-based guide-
lines. Crit Care Med. 2010;3​ 8(8):​S282-S91.
12. IOM (Institute of Medicine). Graham R e. Clinical practice guidelines we can trust. National
13. Academies Press. 2011.
Jamtvedt G, Young JM, Kristoffersen DT, O’Brien MA, Oxman AD. Does telling people what
14. they have been doing change what they do? A systematic review of the effects of audit and
feedback. Qual Saf Health Care. 2006;​15(6):​433‑6.
15. Rosenthal MB, Frank RG, Li ZH, Epstein AM. Early experience with pay-for-performance - From
concept to practice. Jama-J Am Med Assoc. 2005;2​ 94(14):​1788-9311.
16. The Dutch Cochrane Centre. FORMULIER III voor het beoordelen van een COHORTONDER-
ZOEK. http://dcc.cochrane.org/sites/dcc.cochrane.org/files/uploads/cohort.pdf. 2015. Ac-
17. cessed on 15-03-2015
Zhang, Y, and Barbara MW. Qualitative analysis of content. Applications of social research
methods to questions in information and library science. 2009: 308‑319.
Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009) Preferred Reporting
Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(7):
e1000097.
Ardery G, Carter BL, Milchak JL, Bergus GR, Dawson JD, James Pa et al. Explicit and implicit
evaluation of physician adherence to hypertension guidelines. Journal of Clinical Hyperten-
sion (Greenwich, Conn). 2007;​9(2):​113‑9.
Farias M, Friedman KG, Powell AJ, de Ferranti SD, Marshall AC, Brown DW et al. Dynamic evolu-
tion of practice guidelines: Analysis of deviations from assessment and management plans.
Pediatrics. 2012;​130(1):​93‑8.
Irwin ME, Bainey KR, Senaratne MPJ. Evaluation of the appropriateness of pacemaker mode
selection in bradycardia pacing: how closely are the ACC/AHA guidelines followed? Pacing
and Clinical Electrophysiology : PACE. 2003;​26(12):​2301‑7.
Kmetik KS, O’Toole MF, Bossley H, Brutico CA, Fischer G, Grund SL et al. Exceptions to out-
patient quality measures for coronary artery disease in electronic health records. Annals of
Internal Medicine. 2011;1​ 54(4):​227-W77.

44

Reasons for intentional guideline non-adherence: a systematic review

18. Owen RR, Thrush CR, Hudson TJ, Mallory SR, Fischer EP, Clardy Ja et al. Using an explicit Chapter 3
guideline-based criterion and implicit review to assess antipsychotic dosing performance for
19. schizophrenia. -International journal for quality in health care : Journal of the International
20. Society for Quality in Health Care / ISQua. 2002;1​ 4(3):1​ 99‑206.
21. Persell SD, Dolan NC, Friesema EM, Thompson JA, Kaiser D. Frequency of inappropriate medi-
cal exceptions to quality measures. Annals of Internal Medicine. 2010.
22. Uijl SG, Leijten FSS, Moons KGM, Veltman EPHM, Ferrier CH, van Donselaar CA et al. Epilepsy
23. surgery can help many more adult patients with intractable seizures. Epilepsy Res. 2012;​
24. 101(3):​210‑6.
Aujesky D, McCausland JB, Whittle J, Obrosky DS, Yealy DM, Fine MJ. Reasons why emergency
25. department providers do not rely on the pneumonia severity index to determine the initial
26. site of treatment for patients with pneumonia. Clinical infectious diseases : an official publi-
27. cation of the Infectious Diseases Society of America. 2009;​49(10):​e100‑8.
28. Balasubramanian S. Audit of compliance to adjuvant chemotherapy and radiotherapy guide-
29. lines in breast cancer in a cancer network. The Breast. 2003;1​ 2(2):​136-41..
30. Cornali C. Implementation of guidlines for type 2 diabetes mellitus in a post acute geriatric
31. setting. Aging Clinical and Experimental Research. 2009;​21:​338‑43.
32. Evans M, Kortas K. Vancomycin use in a university medical center: comparison with hospital
33. infection control practices advisory committee guidelines. Infection control and hospital
epidemiology : the official journal of the Society of Hospital Epidemiologists of America.
1996;​17(12):7​ 73‑4.
Freed BH, Sugeng L, Furlong K, Mor-Avi V, Raman J, Jeevanandam V et al. Reasons for nonad-
herence to guidelines for aortic valve replacement in patients with severe aortic stenosis and
potential solutions. The American Journal of Cardiology. 2010;1​ 05(9):​1339‑42.
Halm Ea, Atlas SJ, Borowsky LH, Benzer TI, Metlay JP, Chang YC et al. Understanding physician
adherence with a pneumonia practice guideline: effects of patient, system, and physician
factors. Archives of Internal Medicine. 2000;​160(1):​98‑104.
Marras TK, Chan CK. Use of Guidelines in treating community-acquired pneumonia. Chest.
1998;1​ 13(6):1​ 689‑94.
Oliveria Sa, Yood MU, Campbell UB, Yood SM, Stang P. Treatment and referral patterns for
colorectal cancer. Medical Care. 2004;​42(9):​901‑6.
Stensvold A, Dahl Aa, Fosså SD, Axcrona K, Lilleby W, Brennhovd B et al. Clinicians’ use of
guidelines as illustrated by curative treatment of prostate cancer at a comprehensive cancer
center. Acta Oncologica (Stockholm, Sweden). 2011;​50(3):​408‑14.
Bauer MS. A review of quantitative studies of adherence to mental health clinical practice
guidelines. Harvard Review of Psychiatry. 2002;​10(3):​138‑53.
Subramanian S, Klosterman M, Amonkar MM, Hunt TL. Adherence with colorectal cancer
screening guidelines: a review. Prev Med. 2004;​38(5):​536‑50.
Lugtenberg M, Zegers-van Schaick J, Westert G, Burgers J. Why don’t physicians adhere to
guideline recommendations in practice? An analysis of barriers among Dutch general practi-
tioners. Implementation Science. 2009;4​ (1):5​ 4.
Gurses AP, Seidl KL, Vaidya V, Bochicchio G, Harris AD, Hebden J et al. Systems ambiguity and
guideline compliance: a qualitative study of how intensive care units follow evidence-based
guidelines to reduce healthcare-associated infections. Qual Saf Health Care. 2008;1​ 7(5):​
351‑9.

45

Chapter 3

34. Chewning B, Bylund CL, Shah B, Arora NK, Gueguen JA, Makoul G. Patient preferences for
35. shared decisions: A systematic review. Patient Educ Couns. 2012;​86(1):9​ ‑18.
36. Mutasingwa DR, Ge H, Upshur REG. How applicable are clinical practice guidelines to elderly
37. patients with comorbidities? Can Fam Physician. 2011;​57(7):​E253-E62.
Tinetti ME, Bogardus ST, Agostini JV. Potential pitfalls of disease-specific guidelines for
38. patients with multiple conditions. New Engl J Med. 2004;3​ 51(27):​2870‑4.
Medlock S, Opondo D, Eslami S, Askari M, Wierenga P, de Rooij SE et al. LERM (Logical Ele-
ments Rule Method): a method for assessing and formalizing clinical rules for decision sup-
port. International journal of medical informatics. 2011;​80(4):2​ 86‑95.
Latoszek-Berendsen A, Talmon J, de Clercq P, Hasman A. With good intentions. International
journal of medical informatics. 2007;7​ 6, Supplement 3:S​ 440-S6.

46

Chapter 4

Guideline-related barriers to optimal
prescription of oral anticoagulants in

primary care

A.L. Beukenhorst, D.L. Arts, W. Lucassen, K.J. Jager, S.N. van der Veer

Chapter 4

A bstract

Guidelines provide recommendations for antithrombotic treatment to prevent stroke in
people with atrial fibrillation, but oral anticoagulant prescriptions in Dutch primary care
are often discordant with these recommendations. Suboptimal guideline features (i.e.,
format and content) have been suggested as a potentially explanatory factor for this type
of discordance.
Therefore, we systematically appraised features of the Dutch general practitioners’
(NHG) atrial fibrillation guideline to identify guideline-related barriers that may hamper
its use in practice. We appraised the guideline’s methodological rigour and transparency
using the Appraisal of Guidelines, Research and Evaluation (AGREE) II tool. Additionally,
we used the Guideline Implementability Appraisal (GLIA) tool to assess the key recom-
mendations on oral anticoagulant prescription.
The editorial independence of the guideline group scored high (88%); scores for other
aspects of the guideline’s methodological quality were acceptable, ranging from 53%
for stakeholder involvement to 67% for clarity of presentation. At the recommendation
level, main implementation obstacles were lack of explicit statements on the quality of
underlying evidence, unclarity around recommendation strength, and use of ambiguous
terms that may hamper operationalization in electronic systems.
Based on our findings we suggest extending stakeholder involvement in the guideline
development process, standardising layout and language of key recommendations,
providing monitoring criteria, and preparing electronic implementation parallel with
guideline development. We expect this to contribute to optimising the NHG atrial fibril-
lation guideline, facilitating its implementation in practice, and ultimately to improving
antithrombotic treatment and stroke prevention in people with atrial fibrillation.

48