Working Paper n° 09-04
The effect of non-medical
factors on variations in the
performance of colonoscopy
among different health care
Karine Lamiraud, Alberto Holly,
Bernard Burnand, Pascal Juillerat,
Vincent Wietlisbach, Florian
Froehlich, Jean-Jacques Gonvers,
The effect of non-medical
factors on variations in the
performance of colonoscopy
among different health care
Karine Lamiraud1, Alberto Holly1,
Bernard Burnand2 Pascal Juillerat2,
Vincent Wietlisbach2, Florian
Froehlich3, Jean-Jacques Gonvers4,
1 Institute of Health Economics and Management (IEMS),
University of Lausanne
2 Healthcare Evaluation Unit, Institute of Social and
Preventive Medicine (IUMSP), Centre Hospitalier Universitaire
Vaudois and University of Lausanne, Lausanne, Switzerland
3 Department of Gastroenterology and Hepatology, Centre
Hospitalier Universitaire Vaudois and University of Lausanne;
and Department of Gastroenterology, University of Basle,
4 Department of Gastroenterology and Hepatology, Centre
Hospitalier Universitaire Vaudois and University of Lausanne,
This text is not to be cited without the permission
of the authors.
Working Paper n° 09-04
- 1 -
The effect of non-medical factors on variations in the performance of colonoscopy among
different health care settings
Karine Lamiraud, Institute of Health Economics and Management (IEMS), Faculty of
Business and Economics (HEC), University of Lausanne, Lausanne, Switzerland
Alberto Holly, Institute of Health Economics and Management (IEMS), Faculty of
Business and Economics (HEC), University of Lausanne, Lausanne, Switzerland
Bernard Burnand, Healthcare Evaluation Unit, Institute of Social and Preventive
Medicine (IUMSP), Centre Hospitalier Universitaire Vaudois, and University of
Lausanne, Lausanne, Switzerland
Pascal Juillerat, Healthcare Evaluation Unit, Institute of Social and Preventive Medicine
(IUMSP), Centre Hospitalier Universitaire Vaudois and University of Lausanne,
Vincent Wietlisbach1, Healthcare Evaluation Unit, Institute of Social and Preventive
Medicine (IUMSP), Centre Hospitalier Universitaire Vaudois and University of
Lausanne, Lausanne, Switzerland
- 2 -
Florian Froehlich, Department of Gastroenterology and Hepatology, Centre Hospitalier
Universitaire Vaudois and University of Lausanne, Lausanne; and Department of
Gastroenterology, University of Basle, Basle, Switzerland.
Jean-Jacques Gonvers, Department of Gastroenterology and Hepatology, Centre
Hospitalier Universitaire Vaudois and University of Lausanne, Lausanne, Switzerland
John-Paul Vader, Healthcare Evaluation Unit, Institute of Social and Preventive Medicine
(IUMSP), Centre Hospitalier Universitaire Vaudois and University of Lausanne,
Extranef – Dorigny
University of Lausanne
Phone: +41 21 692 34 36
Fax: + 41 21 692 36 55
This study was funded in part by the Swiss National Science Foundation (grant No 32-
- 3 -
Previous studies in the literature have shown significant variations in colonoscopy
performance, even when medical factors are taken into account. This study aimed to
examine the role of non-medical factors (i.e. embodied in health care system design) as
possible contributors to variations in colonoscopy performance.
We used patient data from a multicenter observational study conducted between 2000 and
2002 in 21 centers across 11 western countries. Variability was captured through two
performance outcomes (diagnostic yield and colonoscopy withdrawal time), jointly
studied as dependent variables using a multilevel two-equation system.
Results showed that open-access systems and high-volume colonoscopy centers were
independently associated with a higher likelihood of detecting significant lesions and
higher withdrawal durations. Fee for service (FFS) payment was associated with shorter
withdrawal durations, and had an indirect negative impact on the diagnostic yield.
Teaching centers exhibited lower detection rates and higher withdrawal times.
Our results suggest that gate-keeping colonoscopy is likely to miss patients with
significant lesions and that developing specialized colonoscopy units is important to
improve performance. Results also suggest that FFS may result in a lower quality of care
in colonoscopy practice and highlight that longer withdrawal times do not necessarily
mean higher quality in teaching-centers.
- 4 -
Key Words: Medical Practice Variation (MPV), performance, non-medical factors, panel
two-equation linear-probit model, colonoscopy
- 5 -
Colonoscopy is widely used for the diagnosis and treatment of colonic disorders. In most
adult patients with bowel problems and iron deficiency anemia, it is the preferred method
of colon evaluation and is recognized as one of the most effective methods when
screening for colorectal cancer and adenomatous polyps in asymptomatic patients over
the age of 50 as well as in younger patients at increased risk1,2.
However significant unexplained variations in colonoscopy performance have been
reported3-6 among endoscopists, remaining even when medical factors such as patient
characteristics, physician characteristics and the procedure’s technical aspects3 are taken
into account. To date no studies have provided evidence on how non-medical factors (i.e.
embodied in health care system design) might account for the remaining variance. This
study aimed to do exactly that.
To monitor colonoscopy performance, we used previous research3 and employed two
performance indicators: diagnostic yield (i.e. detection of a significant lesion) and
colonoscopy withdrawal time (i.e. duration during which inspection is primarily
performed). We examined the impact of four non-medical factors on these two measures
(i) Colonoscopy volume at the center: It is known that colonoscopy technique improves
with practice7,8. Hence one might expect high volume centers to be associated with higher
detection rates and longer withdrawal times.
- 6 -
(ii) Physician payment mechanism (fee for service -FFS- versus salaried): FFS systems
pay physicians a fee for each item or unit of care they provide. A salaried physician
receives a lump sum salary for a specified number of hours per week per year. One would
expect health care agents to respond to financial incentives inherent in payment systems9.
Theoretical and empirical research in other therapeutic areas suggests that shorter
procedure durations could be expected in FFS systems10 and these results might also be
expected to hold for withdrawal times. Most importantly, the existing literature provides
inconsistent results on whether the quality of care is different in FFS and salary
systems11. Our results may help further the understanding of the relationship between
mode of payment and quality of care, as expressed by the procedures’ performance.
(iii) Mode of access to the procedure (“open access” versus “gatekeeping”): open access
colonoscopy allows physicians to directly schedule patients for the procedure without
prior consultation. In gatekeeper systems, patients are scheduled for colonoscopy only
after prior consultation with a gatekeeper physician (see appendix for a precise
description of access to colonoscopy in open access and gatekeeping systems). Little is
known about whether gatekeeping has any impact on quality of care whatever the
medical condition12 and one might expect that that gatekeeping limits access. In the case
of colonoscopy, this restriction may be seen as positive, with only higher risk patients
being referred for colonoscopy. In this situation one could expect higher detection rates to
be associated with longer withdrawal times in gatekeeper systems. However, strict
selection criteria prevailing in gatekeeping systems might instead be seen negatively, as
patients5 with a significant and potentially treatable colonic pathology could be missed,
- 7 -
thus reducing detection rates and withdrawal times. Our study investigates whether
gatekeeping has a positive or negative impact on the performance of the procedure.
(iv) Center teaching status (teaching versus non-teaching). Evidence exists in other
therapeutic areas that teaching centers achieve better care than their non-teaching
counterparts13-15 but it is unclear how this benefit is achieved. If we consider
colonoscopy, the center’s teaching status might reflect the amount of care provided. If
junior colonoscopists are being taught during the procedure, the overall quantity of care
provided may be less, thereby resulting in a possible negative impact on diagnostic yield.
Also, teaching activities might lengthen the procedure duration including the withdrawal
Another major contribution of our study is the use of econometric methods able to jointly
handle the two performance indicators, thereby taking possible interactions between both
outcomes into account. In particular, we controlled for the possibility that unobserved
factors might have a positive impact on both the probability of having a significant lesion
and on withdrawal duration. We employed patient data from the EPAGE multicenter
observational study, conducted in 21 centers across 11 western countries.
The EPAGE study
The EPAGE study was carried out between December 2000 and February 2002 in 21
centers across 10 European countries (Czech Republic, Denmark, France, Germany,
Great Britain, Italy, Poland, Spain, Sweden and Switzerland) and Canada. It was
- 8 -
observational and prospectively included patients scheduled for colonoscopy. Each center
was requested to include 300 consecutive patients thereby preventing patient selection.
Patient and procedure data were collected using patient-specific questionnaires,
completed by the endoscopists. Center related data were collected using center-specific
questionnaires, completed by the head of each participating endoscopy center. The
required ethics committee approval, according to country-specific regulations, was
obtained before study commencement. Further information on the design and results of
the EPAGE study can be found elsewhere16,17.
Patient characteristics examined included demographics, health status, type of care
(inpatient or outpatient), the date of any previous colonoscopy, the specialty of the
referring physician (if any) and examination indications. Colonoscopy characteristics
examined included extensive information concerning the procedure execution in terms of
techniques used, appropriateness, as defined by the EPAGE panel using the Rand
appropriateness method18, and performance. We used two performance measures. The
first was the detection of cancer or adenomatous polyps during the examination. We
constructed a dichotomous dependent variable which indicated whether the procedure
detected at least one significant lesion. The second measure was withdrawal time. This
latter was defined as the time duration from cecal identification to withdrawal of
colonoscope from the anus. It included the time taken for maneuvers such as
polypectomy, performed during the withdrawal phase of the examination. Recent
research supports the belief that a minimum adequate colonoscopy withdrawal time can
- 9 -
be equated with the performance of the colonoscopy, in that longer withdrawal times may
indicate careful inspection of the colorectal mucosa, leading to higher detection rates3.
Center characteristics examined included the yearly volume of colonoscopies,
endoscopist method of payment (FFS or salary), mode of access to the procedure
(gatekeeper or open access) and the center’s teaching status (teaching versus non-
Furthermore, national health expenditure data were extracted for each country from the
2007 OECD Health Data Base and expressed as a percentage of the GDP. This variable
was used to represent country-specific characteristics.
As diagnostic yield was used as one of the two performance indicators, we excluded the
procedures whose primary goals were therapeutic from the study population.
Furthermore, only completed procedures (both caecum and terminal ileum reached) were
included as withdrawal time was used as a performance indicator. This has already been
implemented in previous studies5,19.
The econometric analysis aimed at identifying the impact of non-medical factors on both
the diagnostic yield and withdrawal time. The unit of analysis was the patient (i.e. the
colonoscopy). Non-medical factors were embodied in center-specific institutional
- 10 -
characteristics. The empirical analysis comprised two main features. First, the data had a
multilevel structure as patients were nested within centers. Hence, it was possible that
medical procedures conducted in the same center might be correlated. Second, we
handled two performance indices simultaneously, the first (withdrawal time) being
expected to have a direct impact on the second (diagnostic yield)3.
Although multilevel models are common in the field of medical practice variation20,
studies on hierarchical models using multiple outcome measures are scarce. In the
context of profiling the quality of care provided by different groups of physicians, most
efforts have been targeted at improving performance indices based on composite
In our study, both one linear and one non-linear multilevel equation were investigated.
ϑ represent the withdrawal time for patient i in center j.
ij Y denotes the diagnostic
yield for patient i in center j:
ij Y = when a significant lesion was found during the exam,
ij Y = otherwise. We proposed to capture unobserved heterogeneity among centers
through a random-effects specification and to estimate the following two-equation
- 11 -
) , 0 ( ∑
) , 0 (∑
We assumed that
θ was determined by a set of exogenous variables
ijY was simultaneously determined by
θ and a set of exogenous variables
Z was a vector of the non-medical variables.
X2were vectors of
variables incorporating patient characteristics (listed in Table 1), technical aspects of
colonoscopy (listed in Table 3) and physician dummies. The identification of the model
relied on valid instruments23, i.e. exclusion restrictions in the diagnostic yield equation.
Note that the recursive system (1) (only one endogenous variable appearing on the right-
hand side of the other equation) implies that the withdrawal duration has a direct impact
on diagnostic yield, while this latter has an indirect impact on the former, passing through
the correlation between the residuals of both equations (third factor endogeneity). Indeed,
the diagnostic yield itself is only known after the biopsied tissue is analyzed and thus
cannot directly impact on withdrawal duration. The two-equation system controls for the
fact that unobserved variables may exist which are correlated with one another and so
might simultaneously increase both the probability of finding a significant lesion and the
withdrawal time. As the random-effects structure implies that error terms are decomposed
into two parts20, the correlation is induced either by the correlation between individual-
v (perhaps capturing the presence of colorectal cancer etc.)
or by that between center-specific disturbances
u (perhaps capturing working
conditions, managerial style etc.).
- 12 -
The joint evaluation of the entire system of equations was based on the full-information
maximum likelihood estimator. The random-effects linear-probit equation system
required specific programming and was executed by Gauss software24.
Joint estimations were run on the total study population and on the subsample who
underwent first-time screening colonoscopies.
Population and procedure characteristics
A total of 6,004 patients were enrolled during the EPAGE study. The mean number
enrolled per center was 291 (range: 185 - 343). Overall 4,943 patients had a complete
colonoscopy performed for non-therapeutic purposes. Of these latter, 49.2% were men,
the mean age was 57.0 years and 21.9% had major health problems (ASA class III-V).
Major indications included surveillance after polypectomy (19.0%), hematochezia
(15.9%), uncomplicated lower abdominal pain (12.7%) and screening (11.6%). Four
hundred and twelve patients underwent first time screening procedures. Patient
characteristics are summarized in Table 1.
Table 2 shows that significant cancer or adenomatous polyp lesions were found in 18.4%
[5.8%-30.2%] of examinations over all centers. The diagnostic yield was only slightly
smaller (16.7%) in the population who had first-time screening colonoscopy. There were
also large disparities with respect to withdrawal times in the study population. The mean
- 13 -
withdrawal time was 9.7 minutes (±7.4). The overall mean withdrawal time for
examinations in which no significant lesion was found was 8.6 minutes (±6.1) compared
with 14.5 minutes (±10.5) in procedures where a significant lesion was detected. There
were also large variations among centers in terms of total colonoscopy duration. Mean
duration of the procedure was 22.8 minutes (±13.3).
Table 3 provides information concerning the execution of the procedures.
Center characteristics are summarized in Table 4. They express observed institutional
heterogeneity among centers. Centers differed greatly in terms of the number of
procedures carried out, with a yearly mean of 1913 (range: 400 – 5800). Physicians
received a salary in all but six centers, FFS payment being prevalent in these latter. Open
access colonoscopy was practiced in 13 centers while gatekeeping was prevalent in 8.
Sixteen centers were involved in teaching junior colonoscopists while the others were
non-teaching. Table 4 also reports the percentage of patients with major health problems
(ASA III-V) by type of center. This percentage was significantly higher in teaching
centers. However, patient health status was not significantly different between open-
access and gatekeeping centers or between FFS and salary payment systems. Neither did
patient health status significantly differ between high volume and low volume centers.
Regarding the type of physician referring patients for colonoscopy in the EPAGE study,
in gatekeeping centers, 74.2% were gastroenterologists, while only 10.7% were general
practitioners or internists and 15.1% other types of physicians. This suggests that a pre-
- 14 -
assessment consultation took place in most of the cases. On the contrary, in open access
systems, the majority of patients (51%) were referred by a general practitioner or an
internist, 31% were referred by a gastroenterologist and the remaining 18% were
admitted either without referral or after being referred by another type of physician. This
is in line with the access pattern described in the appendix.
The impact of non-medical factors
Regression results are displayed in Table 5. The first two columns display the results on
the study population. We reported the coefficients of the estimation in the withdrawal
equation and the partial effects in the yield equation. The conditions for identification of
this model were met. The type of care (inpatient versus outpatient), the use of
fluoroscopy and the difficulty of the examination were significant in the withdrawal
equation but not in the diagnostic yield equation and were consequently excluded from
the latter. Moreover, a Sargan test performed on the corresponding linear probability
model led to a non-rejection of the null hypothesis that these instruments were valid.
With regard to the diagnostic yield equation, adjusting for patient and technical-related
variables, gatekeeping and teaching activity were significantly negatively correlated with
the diagnostic yield. On average the probability of detecting a significant lesion was
respectively 8.1% and 4.7% less for gatekeeping systems and teaching centers. The
probability of detecting significant lesions was significantly higher in high volume
centers (on average +3.1% per 1000 colonoscopy increase). Physician payment type was
not significantly associated with the diagnostic yield. After adjusting for demographic
- 15 -
and technical-related variables, colonoscopy withdrawal times were significantly longer
in teaching centers (+1.79 minutes per procedure) and high volume centers (+2.14
minutes per 1000 colonoscopy increase). Gatekeeping and FFS approaches were
significantly associated with shorter withdrawal times, -2.37 and -3.1 minutes
respectively. Furthermore, longer withdrawal time was shown to be significantly
associated with a higher probability of detecting a significant lesion, thereby confirming
The models were run without center characteristics to show the effects of including non-
medical variables. For the withdrawal time equation, the Fisher test concluded that center
characteristics accounted for a significant amount (32%) of the total variation measured
in the colonoscopy withdrawal time. With respect to the yield equation, the likelihood
ratio test was used to compare the restricted and unrestricted models. The difference was
significant, suggesting that center variables significantly improved the model.
Center-specific error terms
u were significant for both the withdrawal time and
diagnostic yield equations. Furthermore, the correlations between center-specific and
individual error terms turned out to be positive and significant, suggesting that
unobserved center-specific and patient or procedure-related factors positively influence
both withdrawal time and the diagnostic yield.
- 16 -
The estimations run on the subsample of those who underwent first time screening
colonoscopy (columns three and four of Table 5) provide results that are slightly less
precise but not qualitatively different from those obtained for the whole study population.
The impact of other covariates
As far as other explanatory dimensions are concerned, the impact observed in our study
of most of the colonoscopy’s technical aspects was to be expected. In particular, the use
of more complex procedures and interventions carried out during the investigation
significantly increased the withdrawal time. Colonoscopy with biopsy and colonoscopy
with polypectomy were significant predictors of longer withdrawal times and of
diagnostic yield. Procedures judged non-difficult by the endoscopist had shorter
withdrawal times. Not surprisingly, the diagnostic yield was higher when the
colonoscopy was performed for an appropriate indication. Furthermore, detection rates
were significantly higher in men than in women, as previously described in other
studies25. However, neither sedation nor quality of colon cleansing was significantly
associated with withdrawal time or diagnostic yield, something which does not reflect
results from recent research26. Finally, longer withdrawal times were observed in
countries where health care expenditures as a share of GDP were higher. However, there
was no direct impact of health care expenditures on diagnostic yield.
DISCUSSION AND CONLUSION
The major contribution of our study is to highlight the role of non-medical factors as
contributors to variation in colonoscopy performance. First, our results highlighted that
- 17 -
high volume centers were associated with more significant diagnosis and with longer
withdrawal times. These concomitant effects can be accounted for by the high practice
levels which only high volume centers are able to reach. This training effect (“practice
makes perfect”) assumption has been emphasized in many studies to account for the
association between volume and outcome27. Patients, insurers and policy makers look
increasingly to provider volume as a surrogate indicator for quality of care28. Our results
confirm that this approach may also be important for colonoscopy, supporting the
argument for large specialized centers instead of small gastroenterology units. Note that
our results are based on adjustment for case-mix variables and therefore are not affected
by a continuing major objection against the validity of existing literature examining the
relationship between volume and outcome29. Furthermore, we considered the possible
reverse causality between colonoscopy volume and quality, that is to say the fact that
higher quality colonoscopy centers may attract more clients, which in turn increases their
procedure volume. We estimated an instrumental variable model for the withdrawal time
and diagnostic yield equations, using the size of the population treated at the center as an
instrument. The assumption of exogeneity of the colonoscopy volume variable was not
rejected showing that our results were not biased by any possible reverse causality. It
would be interesting to quantify the gains in efficiency which might be expected from
larger colonoscopy centers. One cannot exclude the possibility that some well-informed
patients may prefer a lower quality of care locally instead of traveling to a distant high-
volume center30. One may also wonder whether there is an optimal size for colonoscopy
- 18 -
Second, colonoscopy withdrawal times were significantly shorter for endoscopists
working in a FFS payment system ceteris paribus and the mode of payment was not
associated with the diagnostic yield. Results on withdrawal time can be interpreted as an
incentive effect associated with the mode of payment: In FFS systems, physicians may
perform more colonoscopies per day, thereby making each colonoscopy shorter. This
confirms the findings of other studies which show that FFS may result in shorter
procedures9. Although the lesion detection rate was not directly significantly influenced
by the type of payment, there may have been an indirect negative impact of FFS on
diagnostic yield arising from the withdrawal time. Altogether these results suggest that
FFS may induce a lower quality of care in the context of colonoscopy.
Third, gate-keeping systems were independently associated with a lower probability of
detecting significant lesions and shorter withdrawal times. This trend is attributable to a
negative selection effect. The correlation between the mode of access and the ratio
between the number of colonoscopies performed at a center and the patient population
treated there per year, confirms that gatekeeping systems limit access to the procedure. In
the EPAGE study, 0.76% and 0.58% of the patients treated underwent a colonoscopy in
open access and gatekeeping centers respectively. The difference is significant at the 5%
level. Our results suggest that gatekeeping systems, through the application of strict
selection criteria, may miss patients with significant and potentially treatable colonic
pathologies. In this respect, some authors6 are in favor of open-access colonoscopy for
screening in healthy patients. Our results strongly reinforce this approach. Another point
for further investigation is whether gatekeeping, which reduces the likelihood of
- 19 -
detecting significant lesions, saves on resources. Our data suggest that the percentage of
referrals is smaller in gatekeeping systems. Hence it could be interesting to compute the
costs of open-access and gatekeeping systems respectively, identifying to what extent
patient selection gives a cost effective intervention level. In this way one could evaluate
areas where open access may not be cost effective. The current literature in other
therapeutic areas suggests that restricting access to specialists through gatekeeping does
not significantly reduce specialist referrals and is not necessarily cost effective 31,32.
Fourth, we showed that colonoscopy withdrawal times were significantly longer in
teaching centers. However, the likelihood of detecting a lesion was significantly lower
there ceteris paribus. These results confirm that the teaching of junior colonoscopists
lengthens the withdrawal time for teaching purposes. Furthermore, the results show that,
for equal withdrawal times, the amount of care provided might be less in teaching centers
than in their non-teaching counterparts. Additional regressions run on the subsamples of
teaching and non-teaching centers suggest that the marginal effect of increased
withdrawal time on the probability of detecting a significant lesion was on average lower
in teaching centers (0.3% per additional minute versus 0.8%). Our results have major
policy implications. Although the efficiency of teaching centers is lower, one cannot
evaluate teaching and non-teaching centers in the same way, as the former not only carry-
out procedures but also invest in training.
It is important to stress that our results were maintained for the subsample of first time
screening colonoscopies. This subsample analysis was important for comparative
- 20 -
purposes because previous studies have mostly examined the relationship between
withdrawal time and adenomas or advanced adenomas in patients undergoing first time
screening examinations5,19. Indications in our study not only included “Screening for
Colorectal Cancer” but also “Hematochezia”, “Uncomplicated Abdominal Pain” and
“Surveillance After Polypectomy”. This mixture could make results’ interpretation more
difficult. Note that, when the study was carried out, screening colonoscopy was not
reimbursed by basic insurance in some of the countries and consequently the study most
probably under-reported this indication. Consequently, the distribution of indications
should be regarded with caution. However, this possible effect does not jeopardize the
results of the study on non-medical factors.
Additional data would have been helpful in understanding the residual variance in the
estimations. In particular, patient data such as weight, height or waist circumference were
not collected33. Other information not collected but which could have been useful
included the patient’s bowel habits and certain psychological factors such as coping style
and anxiety, which are known to have an impact on the ease of the procedure. Little
information was available concerning endoscopists, the regressions only including
physician dummies referring to the endoscopist performing the procedure. Neither
individual income data nor information as to whether endoscopists worked privately in
addition to their hospital duties, was collected.
In this work, we studied two performance indicators jointly, thereby controlling for the
biases induced by the interactions between both. The results confirmed previous research
- 21 -
suggesting that withdrawal time has a positive effect on diagnostic yield and the
correlations between residuals of both equations were significant, thus defending our
two-equation approach. We also considered using center average withdrawal time as a
predictor of diagnostic yield rather than the actual time in each specific case as an
alternative way of resolving the endogeneity of withdrawal time. However, the average
withdrawal time was not significant in this single-equation approach, thus giving
additional support to our two-equation approach. Furthermore, it should be acknowledged
that our results were not qualitatively different when taking into account potential
selection biases due to the fact that we restricted our analysis to completed colonoscopies.
This was performed using a three equation system (adding a third equation to account the
factors associated with completed colonoscopy-Data available on request).
Studying the factors associated with the diagnostic yield and the withdrawal time of the
procedure, we raised the major issues of colonoscopy performance increasingly being
discussed in western countries because of the rapidly rising demand for colonoscopy as a
screening procedure. This is especially pertinent in the context of North America where
colorectal cancer is now the second largest cause of cancer death34. The demand for
colonoscopy is often perceived as potentially overwhelming there. For example, current
guidelines (enactment of federal law in the United States, effective July 1st 2001, entitling
Medicare beneficiaries to a screening colonoscopy every 10 years) for colonoscopy
screening could require as many as 7.7 million American people to undergo a
colonoscopy each year35. Only 4.4 million procedures were performed in 1999. In this
context, managing quality related36 problems is crucial. There has been increasing
- 22 -
recognition that colonoscopy may not always succeed in detecting important colon
pathology37. This may result in the future overuse of colonoscopy because of growing
concern about finding serious pathological features in subsequent examinations. This
overuse would consume valuable colonoscopy resources, drive up health care costs and
expose patients to unnecessary risk.
In conclusion, our study adds to the extensive literature about non-medical factors
influencing medical performance by focusing on the specific case of colonoscopy. Open-
access is more likely to detect significant lesions and is associated with longer
withdrawal times. High volume centers exhibit higher detection rates together with longer
withdrawal times, suggesting that they should be favored. FFS systems may be
problematic as they tend to be associated with shorter withdrawal durations and so may
have an indirect negative impact on the diagnostic yield. Our results, relying on a two-
equation estimation, also confirm that detection rates and withdrawal time are
significantly associated ceteris paribus, thus confirming previous preliminary findings3
based on single-equation approaches. However our results also highlight that longer
procedures do not necessarily mean higher quality, as in the specific setting of teaching
centers which are involved in the training of junior endoscopists. Hence, the results of the
present study highlight the non-medical sources for improved performance. A dedicated
colonoscopy unit may balance out the advantages and costs of each solution. Further
research is crucially needed, both in clinical and economic literature, in order to reach
firm policy conclusions.
- 23 -
In practice, in open-access systems patients can access colonoscopy either directly
without physician referral or can be referred for colonoscopy by any physician (general
practitioner, internist or specialist), the latter being the most common situation. Usually,
patients go to their primary care practitioner (PCP) because of either abdominal pain or
rectal bleeding. At this point, the PCP prescribes the colonoscopy. The endoscopic center
schedules the appointment and a pre-procedure bowel preparation (4 liters of liquids the
day before), so that the patient will go to the center for the first time on the day of the
examination after already having had the bowel preparation. At this point, immediately
before the procedure, the patient may be visited by the endoscopist. However, since the
patient has already had the bowel preparation, the examination will be performed
irrespectively of whether the endoscopist deems the indication appropriate or not.
GATEKEEPING SYSTEM: The patient showing symptoms is usually referred by the
gatekeeper PCP to a gastroenterologist. In this case, the gastroenterologist decides, during
the visit (pre-assessment consultation), whether or not a colonoscopy is required. If it is,
he/she will make the appointment for the patient who will in turn have the bowel
preparation after having talked to the specialist. This results in patients being mostly
referred to colonoscopy by gastroenterologists in gatekeeping systems.
- 24 -
1. U.S. Preventive Services Task Force. Screening for Colorectal Cancer: U.S. Preventive
Services Task Force Recommendation Statement. Annals for Internal Medicine.
2. Levin B, Lieberman DA, McFarland B, et al. Screening and Surveillance for the Early
Detection of Colorectal Cancer and Adenomatous Polyps 2008: A Joint Guideline From
the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer,
and the American College of Radiology. Gastroenterology. 2008; 134 (5), 1570-1595.
3. Barclay RL, Vicari JJ, Doughty AS, Johanson JF, Greenlaw RL. Colonoscopic
Withdrawal Times and Adenoma Detection during Screening Colonoscopy. New England
Journal of Medicine. 2006;355 (24), 2533 – 2541.
4. Rex DK, Petrini JL, Baron TH, et al. Quality indicators for colonoscopy.
Gastrointestinal Endoscopy. 2006;63(4),S16-S28.
5. Sanchez W, Harewood GC, Petersen BT. Evaluation of polyp detection in relation to
procedure time of screening or surveillance colonoscopy. American Journal of
6. Rex D., Lieberman, D. Feasibility of screening colonoscopy: Discussion of issues and
recommendations regarding implementation.
6. Arblaster MJ, Collopy BT, Elliott PR. Colonoscopy in a private hospital: continuous
quality improvements in practice. Australian Clinical Review. 1992;12:71-6.
- 25 -
7. Church JM. Complete Colonoscopy: how often? and if not, why not? American
Journal of Gastroenteroly. 1994;89:556-60.
9. Lindenauer PK, Remus D, Roman S, et al. Public Reporting and Pay for Performance
in Hospital Quality Improvement. New England Journal of Medicine. 2007;356:486-96.
10. Gosden T, Forland F, Kristiansen IS, et al. Impact of payment method on behaviour
of primary care physicians: a systematic review. Journal of Health Services Research and
11. Armour BS, Pitts MM, Maclean R, et al. The effect of explicit financial incentives on
physician behavior. Arch Intern Med. 2001;161:1261-1266.
12. Philipps KA, Haas JS, Liang SY, et al. Are Gatekeeper Requirements Associated with
Cancer Screening Utilization? Health Services Research. 2004;39:153-178.
13. Ayanian JZ, Weissman JS. Teaching hospitals and quality of care: a review of the
literature. Milbank Q. 2002;80(3):569-593.
14. Taylor DH, Whellan DJ, Sloan FA. Effects of admission to a teaching hospital on the
cost and quality of care for Medicare beneficiaries. New England Journal of Medicine.
- 26 -
15. Silber JH, Rosenbaum PR, Romano PS, et al. Hospital Teaching Intensity, Patient
Race, and Surgical Outcomes. Arch Surg. 2009;144(2):113-120.
16. Vader JP, Wietlisbach V, Harris JK, et al. Gastroenterologists overestimate the
appropriateness of colonoscopies they perform: an international observational study.
17. Burnand B, Harris JK, Wietlisbach V, et al. Use, appropriateness and diagnostic yield
of screening colonoscopy: an international observational study (EPAGE).
Gastrointestinal Endoscopy. 2006;63(7):1018-26.
18. Vader JP, Burnand B, Froehlich F, et al. The European Panel on Appropriateness of
Gastrointestinal Endoscopy (EPAGE): project and methods. Endoscopy. 1999;31:572-8.
19. Simmons DT, Harewood GC, Baron TH, et al. Impact of endoscopist withdrawal
speed on polyp yield: implications for optimal colonoscopy withdrawal time. Aliment
Pharmacol Ther. 2006. 24:965-971.
20. Rice N, Jones A. Multilevel models and health economics. Health Economics.
- 27 -
21. Landrum MB, Bronskill SE, Normand SL. Analytic Methods for constructing cross-
sectional profiles of health care providers. Health Services Research and Outcomes
22. Landrum MB, Normand SL, Rosenheck RA. Selection of related multivariate means:
Monitoring psychiatric care in the Department of Veterans Affairs. Jounal of the
American Statistical Association. 2003;98(461):7-16.
23. Maddala GS. Limited-dependent and qualitative variables in econometrics.
Cambridge University Press 1983.
24. Huguenin J. The Mutivariate Normal Distribution and Multivariate Probit Analysis.
PhD Thesis 2004, University of Lausanne.
25. Regula J, Rupinski M, Kraszewska E, et al. Colonoscopy in Colorectal-Cancer
Screening for Detection of Advanced Neoplasia. New England Journal of Medicine.
26. Radaelli F, Meucci G, Sgroi G, et al. Technical Performance of Colonoscopy: The
Key Role of Sedation/Analgesia and Other Quality Indicators. Am J Gastroenterol. 2008.
- 28 -
27. Halm EA, Lee C, Chassin MR. Is volume related to outcome in health care? A
systematic review and methodologic critique of the literature. Annals of Internal
28. Daly JM. Society of Surgical Oncology Presidential Address: Volume, Outcome, and
Surgical Specialization. Annals of Surgical Oncology. 2004;11(2):107-114.
29. Dimick JB, Finlayson SRG, Birkmeyer JD. Regional Availability Of High-Volume
Hospitals For Major Surgery. Health Affairs.2004,
30. Hospital volume and health care outcomes, costs and patient access. Effective Health
31. Ferris TG, Chang Y, Blumenthal D, Pearson SD. Leaving gatekeeping behind. Effects
of opening access to specialists for adults in a health maintenance organization. New
England Journal of Medicine. 2001;345:1312-7.
32. Pati S, Shea S, Rabinowitz D, et al. Does Gatekeeping Control Costs for Privately
Insured Children? Findings From the 1996 Medical Expenditure Panel Survey.
- 29 -
33. Anderson JC, Gonzales JD, Messina CR, Pollack BJ. Factors that predict incomplete
colonoscopy: thinner is not always better. The American Journal of Gastroenterology.
34. American Cancer Society. Cancer Facts and Figures 2008. Atlanta: American Cancer
35. Zamir S, Rex D. An Initial Investigation of Efficiency in Endoscopy Delivery. The
American Journal of Gastroenterology. 2002;97:1968-1972.
36. Rizek R, Paszat L, Stukel T, et al. Rates of Complete Colonic Evaluation After
Incomplete Colonoscopy and Their Associated Factors: A Population-Based Study.
Medical Care. 2009;47(1):48-52.
37. Baxter NN, Goldwasser MA, Paszat LF, Saskin R, Urbach DR, Rabeneck L.
Association of colonoscopy and death from colorectal cancer: a population-based, case–
control study. Ann Intern Med. 2009;150(1).
- 30 -
Table 1: Patient characteristics (n = 4943)
AGE (mean, std) 61.7
MALE (%) 49.2% 42.1% 58.5%
HEALTH STATUS (%)
Major health problems (ASA III – V*) (%)
PREVIOUS COLONOSCOPY (%) 37.4 10.6 53.4
MAIN INDICATIONS FOR COLONOSCOPY
Colorectal cancer screening (%)
Uncomplicated Lower Abdominal Pain (%)
Surveillance after polypectomy (%)
*American Society of Anesthesiologists (ASA)
Definition of ASA Status:
Class I Normal, Healthy
Class II Patient with mild systemic disease
Class III Patient with severe systemic disease
Class IV Patient with severe systemic disease that is a threat to life
Class V Moribund patient who is not expected to survive without the operation
Class VI A declared brain-dead patient whose organs are being removed for donor purposes.
- 31 -
Table 2: Colonoscopy Performance (n = 4943)
SIGNIFICANT LESION (DIAGNOSTIC YIELD) (%)
SIGNIFICANT LESION FOR PATIENTS UNDERGOING
FIRST TIME SCREENING COLONOSCOPY (%)
WITHDRAWAL TIME (MEANS)
WITHDRAWAL TIME FOR PATIENTS UNDERGOING
FIRST TIME SCREENING COLONOSCOPY
- 32 -
Table 3: Technical aspects of colonoscopy (n = 4943)
Quality of colon cleansing: completely clean (%)
Deep sedation (%)
Use of fluoroscopy (%)
Biopsies taken (%)
Polypectomy performed (%)
Difficulty of colonoscopy (1 – 6 scale, by
ascending degree of difficulty) (according to the
Appropriate (according to EPAGE panel)
- 33 -
Table 4: Center characteristics (n = 21) and patient health status by center characteristics
Number of centers and
colonoscopies by center
(n = 21)
Patient health status by center
(n = 4943)
% patients with
(NUM. COLONOSCOPIES/ YR)
ACCESS TO COLONOSCOPY
CENTER TEACHING STATUS
- 34 -
Table 5: Nested data two-equation system
COLONOSCOPY WITHDRAWAL EQUATION
First time screening
Coef P>| z|
dF/dx P>| z|
Major health problems (ASA III-IV)
Indication for Colonoscopy: Screening for Colorectal Cancer
Indication for Colonoscopy : Hematochezia
Indication for colonoscopy : Uncomplicated Abdominal Pain
Indication for colonoscopy : Surveillance after Polypectomy
Quality of colon cleansing : completely clean or clear liquid aspired
Use of fluoroscopy
Presence of a 2nd endoscopist
Difficulty (1- 6 scale by ascending degree of difficulty)
Center colonoscopy volume / 1000
FFS payment system
Access to endoscopy : gatekeeping system
Health care expenditure (%GDP) in the country/10
DIAGNOSTIC YIELD EQUATION
Major health problems (ASA III-IV) *
Indication for Colonoscopy: Screening for Colorectal Cancer*
Indication for Colonoscopy : Hematochezia*
Indication for colonoscopy : Uncomplicated Abdominal Pain*
Indication for colonoscopy : Surveillance after Polypectomy*
Quality of colon cleansing : completely clean or clear liquid aspired* 0.002
Center colonoscopy volume / 1000
FFS payment system*
Access to endoscopy : gatekeeping system*
Health care expenditure (%GDP) in the country/10
Covariance between unobserved center effects
Covariance between error terms
asignificant , bPhysician dummies were included
(*)dF/dx is for discrete change of dummy variable from 0 to 1; P>|z| is the test of the underlying coefficient being 0
Page 38 Download full-text