Professional Documents
Culture Documents
S0169-2607(15)00093-0
http://dx.doi.org/doi:10.1016/j.cmpb.2015.04.007
COMM 3915
To appear in:
Received date:
Revised date:
Accepted date:
26-10-2014
12-4-2015
13-4-2015
Please cite this article as: K.-H. Chen, J.-M. Wu, T.-W. Ho, H.-J. Yu, F. Lai, A CrossHospital Cost and Quality Assessment System by Extracting Frequent Physician Order
Set from a Nationwide Health Insurance Research Database, Computer Methods and
Programs in Biomedicine (2015), http://dx.doi.org/10.1016/j.cmpb.2015.04.007
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
ip
t
cr
Kuo-Hsin Chen5,#, MD; Jin-Ming Wu1,4,#, MD; Te-Wei Ho1, MS; Hwan-Jeu Yu2,*,
PhD; and Feipei Lai1,2,3, PhD
us
University, Taiwan
2
an
te
*Corresponding author
Ac
ce
p
Highlights
A system that provides a convenient way for physicians to retrieve and compare
clinical pathways among health care providers about herniorrhaphy.
The frequent physician order sets were derived from the National Health
Database.
1
Page 1 of 34
ip
t
Abstract
cr
Purpose
Clinical pathways fall under the process perspective of health care quality. For care
us
providers, clinical pathways can be compared to improve health care quality. The
objective of this study was to design a convenient physician order set comparison
an
system based on claim records from the National Health Insurance Research Database
(NHIRD) of Taiwan.
Methods
Data were retrieved from the NHIRD for the period of 20032007 for frequent
te
physician order sets found in hospital surgical hernia repair inpatient claim records.
Ac
ce
p
The derived frequent physician order sets were divided into five frequency thresholds:
80%, 85%, 90%, 95%, and 100%. A consistency index was defined and calculated to
understand each care providers adherence to clinical pathways. In addition, the
average count of physician orders, average amount of cost, Charlson comorbidity
index, and recurrence rate were calculated; these variables were considered in
frequent physician order sets comparison.
Results
Records for 3,262 patients from 257 hospitals were retrieved. The frequent physician
order sets of various frequency thresholds, Charlson comorbidities, and recurrence
rates were extracted and computed for comparison among hospitals. A recurrence rate
2
Page 2 of 34
ip
t
each hospital (3.504.41 vs. 7.096.57; P<.001), and non-medical center facility type
(P=.042). A multivariable Cox regression analysis indicated an association of low
cr
recurrence rates with consistency index only (one percentage increased: OR=0.973;
us
CI:0.9570.990; P=.002).
an
Conclusions
The proposed system leveraged the claim records to generate frequent physician order
sets at hospitals, thus solving the difficulty in obtaining clinical pathway data. This
allows medical professionals and management to conveniently and effectively
te
Keywords
compare and query similarities and differences in clinical pathways among hospitals.
Ac
ce
p
Page 3 of 34
Introduction
ip
t
cr
1980s, Donabedian proposed a conceptual model for studying and evaluating health
us
care service quality [1]. This model classified the quality of care into three
perspectives: structure, process, and outcome. The structure perspective focused on
an
"the material, facilities, equipment, human resources, and organization of the health
care providers". The process perspective focused on "what is actually done in giving
and receiving care", including the activities involved in making a diagnosis and
recommending or implementing treatment. The outcome perspective focused on the
"effects of care based on the patients' health conditions". There are many issues
te
associated with the three health service delivery perspectives noted above. There are
many concerns associated with implementing these three health service delivery
Ac
ce
p
Page 4 of 34
different care, particularly regarding surgical procedures. This situation can result in
differences in elements such as hospital stays, laboratory tests, the use of medication
and blood products, outpatient treatment, complications, readmissions, total cost,
ip
t
outcomes, and patient satisfaction. All of these issues are of concern to patients, health
professionals, and health care management [5]. The application of clinical pathways is
cr
one way to address the quality-of-care challenge, as they provide guidance for
us
an
hernia surgery, pulmonary complications and hospital stays for head and neck
reconstruction [9] or minimize hospital costs and hospital stays for both hepatectomy
[10] and kidney surgery [11]. Thus, clinical pathways have been developed and
applied in many hospitals.
Clinical pathways have taken many forms. A typical clinical pathway contains a set of
te
physician order activities to be executed during a defined time period for a specific
clinical care objective (e.g., exam, anesthesia, drug prescription). The pathway can be
Ac
ce
p
used to coordinate the clinical care team, standardize practices, and reduce the
variance in health care delivery. The benefits of clinical pathways include achieving
expected outcomes, promoting effectiveness and efficiency of clinical care, and
optimizing resource utilization. Taken together, these benefits lead to a higher quality
of care [12].
The Problems
Best practice clinical pathways come from two sources: creation of clinical pathways
and adherence to clinical pathways. The creation of clinical pathways is usually based
on existing evidence and medical rules discussed in a professional committee. This
process can ensure that the clinical pathways are as objective as possible. However, a
5
Page 5 of 34
group of professionals under the same health care institution may still overlook or
exclude an important factor or activity. Day-to-day usage of the clinical pathways
may also identify areas of improvement. In addition, hospitals could possibly find
ip
t
ways to improve their own clinical pathways by reviewing those of other hospitals.
Though comparing clinical pathways in use at other institutions could be helpful, it
cr
can be difficult to obtain this information. Even when details of the clinical pathways
us
of another hospital are obtained, the resulting outcome of these clinical pathways can
be difficult to determine. Moreover, it is never known whether the other hospital
an
precisely follows the clinical pathways. Thus, it is difficult to make a comparison with
another hospital. In summary, we need a practical and convenient method for
te
Objectives
pathways.
Among the diverse components of the clinical pathway, sets of physician orders are a
Ac
ce
p
Page 6 of 34
ip
t
cr
computation algorithm, and overall system architecture of our study. A sample hernia
us
surgical operation procedure and its results are shown in Section 3. In Section 4, we
discuss the results of the sample comparison and address the contribution and
an
Data Source
te
The data used in our analysis were drawn from the NHIRD. The dataset used a
representative sample (N=1,000,000) selected from the entire database of beneficiary
Ac
ce
p
information [14]. Data extracted included registration files and original data for
reimbursement claims. The registration files included information on patients,
physicians, and care providers. The identification data for each file was scrambled to
ensure confidentiality. Based on the hospital registration file, we categorized the
hospitals into three types: medical center, regional hospital, and district hospital. The
inpatient expenditures file contained patient admission claim data, including
admission date, discharge date, and up to five ICD9 operation procedure codes for
this hospitalization as well as the patient's gender, date of birth, hospital ID, and
physician ID. The inpatient orders file contained the physician orders related to the
above inpatient expenditures admission record. The physician orders were divided
into different categories including medication description and amount, special
7
Page 7 of 34
ip
t
patients.
cr
Subjects Selection
us
To demonstrate the usability of the system, we selected and compared results from a
hernia repair operation. "Hernias are associated with reduced quality of life and high
an
unilateral inguinal hernia repair for the first time between 2003 and 2007 were
retrieved from the NHIRD, and included ICD9 procedure codes 53.00, 53.01, 53.02,
te
Ac
ce
p
The original NHIRD data comprise detailed records, but our study required us to
aggregate information. Therefore, we prepared the data in order to meet the aggregate
query requirements. This operation was similar to the transition from a transaction
database to a data warehouse. Transaction databases are used for daily operations, and
data warehouses for business decision making. First, ETL (extract, transform, and
load) must be performed on the transaction database to construct the warehouse
database. Then, the normalized transaction database must usually be de-normalized to
accelerate and fulfill the aggregate query requirements of the data warehouse. The
tasks to be performed in advance are as follows: extract the target detailed medical
records that meet the desired procedure codes, apply exclusion criteria, calculate the
number of cases grouped by hospital ID and then sort them by the number of cases,
8
Page 8 of 34
calculate the average number of physician orders per hospital, and calculate the
average amount of cost per hospital. Once complete, this process yields a population
and number overview regarding the target surgical operation. Subsequently, we can
ip
t
cr
us
The clinical pathway for a surgical operation at a specific hospital can be derived
from the physician orders claimed in the insurance records. Our approach was to
an
physician orders that reached the respective thresholds. For example, a physician
order that reaches 80% frequency indicate that this physician order occurred in more
than 80% of the claim cases. Similarly, if a physician order reaches 100% frequency,
te
it means that it occurred in all claim cases. In theory, higher frequency of physician
order sets will result in fewer physician orders that qualify. Physician order sets that
Ac
ce
p
reached one of the thresholds were considered frequent physician order sets. For
example, the frequent physician order sets for some surgical procedure of a care
provider could look like Table 1. The meaning of most of the order codes are found in
Table 6. The concept of the extraction and comparison process is shown in Figure 2.
Consistency Index
Table 1 demonstrates that orders 118 occurred in more than 80% of the hospital's
total cases. If the hospital's average number of physician orders applied per case is 20,
then we can say that 18/20 (90%) of physician orders per case reach the 80%
frequency threshold. On the other hand, if the hospital's average number of physician
orders per case is 40, we can say that only 18/40 (45%) of physician orders per case
9
Page 9 of 34
reach the 80% frequency threshold; the former situation is more consistent than the
latter. To understand the consistency of each hospital, we defined a consistency index
as follows:
ip
t
Consistency Index =
Where
cr
us
an
Let us use Table 1 as an example. For a surgical procedure, if the hospital's average
number of physician orders per case is 20, the consistency index of this hospital can
be calculated as follows:
te
Ac
ce
p
In order to assess the relationship between a patient's past health status and the current
health outcome, we computed the comorbidity index of every patient and the average
comorbidity index of every hospital. We used Charlson's model [17] to compute
patient comorbidity, and examined the correlation between morbidity and ICD9
disease and procedure codes, as adapted by Deyo [18]. We designed a software
10
Page 10 of 34
module to compute the Charlson score for each patient by using patient ID and index
date. After computing each patient's Charlson score, we computed the average
Charlson score for each hospital. We approximated each hospitals comorbidity status
ip
t
cr
Recurrence
us
One indicator of health care quality is patient recurrence rate. The system retrieved
patients recurrence records by identifying the ICD9 code 550.91 (hernia recurrence)
an
from the index date until the end of 2008. As this database is nationwide, we also
determined whether the patient returned to the same or a different hospital. The
system displayed all recurrence records after the index date. The system also showed
the physician ID associated with each surgical procedure. We then determined if
different physicians had varying favorite physician orders, assessing if they differed
Ac
ce
p
order sets.
te
from the frequent physician orders derived from the hospital's frequent physician
Comparison Algorithm
Page 11 of 34
set. Then, the system displays the marked union order set list to the user. The pseudo
code of the comparison procedure is listed as follows:
Algorithm Comparison
Input: FrequentOrderSetA, FrequentOrderSetB
te
System Architecture
an
us
cr
ip
t
Output:UnionOrderSet
Set UnionOrderSet = Union of FrequentOrderSetA and FrequentOrderSetB
For each PhysicianOrder t in FrequentOrderSetA
Begin
If t exists in UnionOrderSet
Set a mark in the A field of matched PhysicianOrder of UnionOrderSet
End
For each PhysicianOrder t in FrequentOrderSetB
Begin
If t exists in UnionOrderSet
Set a mark in the B field of matched PhysicianOrder of UnionOrderSet
End
Return UnionOrderSet
Ac
ce
p
application based on SQL Server 2008 and running on Microsoft Internet Information
Services. Figure 4 depicts the system architecture, and Figure 5 the hierarchy of
functionalities.
Results
Page 12 of 34
claim cases in that year or that hospital. The listed fields include patient gender, age,
Charlson score, admission date, discharge date, length of stay, amount of cost, and
average number of physician orders applied. When users are interested in a specific
ip
t
case, they can further click to view the detailed physician orders applied on the case.
The aggregate record distributions from 2003 to 2007 were 520, 688, 717, 680, and
cr
657, respectively. The ratio of patients receiving hernia repair surgical procedure for
us
the first time was about 6.520.77 per ten thousand (652/1,000,000).
an
hospitals, and 159 district hospitals. The aggregated count of cases for the top twelve
hospitals is shown in Table 2. Among these twelve hospitals were 9 medical centers, 2
regional hospitals, and 1 district hospital. The average Charlson score was at a normal
level, 1.720.39, and the average number of physician orders applied to patients was
te
Ac
ce
p
amount of cost for each physician order category. An example is shown in Table 3.
Frequent Physician Order Sets and Consistency Index
We viewed the extracted frequent physician orders for hospitals according to different
frequency thresholds: 80%, 85%, 90%, 95%, and 100%. The results were similar to
Table 1. The consistency indices for these twelve hospitals (Table 4) varied from
32.68% to 82.09%.
The ratio of frequent physician order sets to average number of physician orders
applied per case for each hospital at each frequency threshold are depicted in Figure 6.
Higher consistency indices are indicated by higher ratio scales and flatter curves.
13
Page 13 of 34
ip
t
threshold of frequent physician order sets. For convenience, the system displayed the
results of the comparison (including Charlson score, recurrence rate, categorized
cr
amount of cost, and frequent physician order sets) in one table. We separated the table
us
into two smaller tables here to aid explanation. Table 5 shows the comparison of the
Charlson score, recurrence rate, and categorized amount of cost. Hospital J had a
an
group of medium Charlson score patients, medium amount of cost, and an extremely
low recurrence rate.
The comparison results of frequent physician orders are shown in Table 6. The three
medical centers presented common orders such as CBC-I, and general biochemistry
Ac
ce
p
examinations.
te
pulse or ear oximetry and so on. Hospital J performed more general biochemistry
7.096.57; P<.001), and non-medical center hospital type (P=.042) were associated
with low recurrence rates. Multivariable Cox regression analysis (Table 8) revealed
that only a high proportion of consistency index was associated with low recurrence
14
Page 14 of 34
ip
t
Discussion
cr
The present analysis of the proposed system indicates that a higher consistency index
us
leads to a lower recurrence rate. This suggests that efforts to increase consistency
index may be a good solution for hospitals with low consistency and high recurrence
an
rates; a change in consistency rate may yield improved health care delivery. Based on
the results of our analyses, hospitals should aim to achieve a benchmark consistency
index of 70%.
2.
te
frequency threshold.
1.
Compute the consistency index based on the derived frequent physician order
Ac
ce
p
sets.
3.
Review the variation between the universal physician order set and the frequent
physician order sets.
4.
Compare the frequent order sets with those of other care providers.
5.
6.
Our proposed system can easily support the implementation of these improvement
steps. With the annual addition of new NHIRD data, the system can provide insights
on innovative practices and allow continuous updating of the sets of physician orders.
Clinical pathways are germane to the process perspective of health care quality.
Comparing the sets of physician orders across care providers is a good approach to
15
Page 15 of 34
identifying areas for clinical improvement and improving health care service delivery
overall. Our system provides a convenient point of access to this part of the clinical
pathway, because data is often challenging to obtain. Physicians may use the
ip
t
comparative data to better understand the hospital eco-system, their current practices,
and ways to improve current practices.
cr
We chose hernia repair as a starting point for this system because it can be examined
us
as a single pure surgical operation more easily than other conditions with a higher
proportion of mixed operational procedures[19]. Furthermore, hernia repair is less
an
complicated than other conditions, and the operation required is relatively well
established, compared with operations for other conditions. In previous reports,
recurrence after herniorrhaphy may be associated with skills of surgical mesh fixation
[20], characteristics of mesh [21, 22], and types of hernia [23]. We expect that this
system can provide even more benefits for physicians when increasingly complex
te
Ac
ce
p
patient. In this study, there is no statistically significant difference between low and
high recurrence rate in terms of the Charlson score (Table 7). It means that the
Charlson score has no association with recurrence after herniorrhaphy. Our study has
similar findings compared to one meta-analysis study, which reports that female
gender, direct inguinal hernias at the primary procedure, operation for a recurrent
inguinal hernia, and smoking history are significant risk factors for recurrence after
inguinal hernia surgery [24].
16
Page 16 of 34
Limitations
ip
t
cr
only assessed quantity and did not consider the physical influence of the physician
us
orders. For example, a surgical method may be considered more influential than a lab
test method. This circumstance may be improved by controlling for the weight of
an
te
Variance of Physicians
Ac
ce
p
Page 17 of 34
In contrast, analyses computed from the individual patient base might yield another
order set, a frequent association physician order set. If we further considered the
sequence of these physician orders during a hospital stay, it will lead to a frequent
ip
t
sequential physician order set. These results may be obtained from data models by
cr
us
purpose. In the future, the system can be improved to provide general-purpose queries
for any procedure code operation.
an
and are enrolled on the variance record. Most important of all, some deviations may
te
Ac
ce
p
function. In the future, our system may do regular mining for individual hospital to
find the serial change of uncommon physician orders, which may become standard
afterward.
Conclusions
Page 18 of 34
displayed for comparison to evaluate the related potential quality. Results suggest that
a higher consistency index leads to lower recurrence rates. Care-provider
administrators can use a consistency index of 70% as a reference benchmark when
ip
t
cr
Acknowledgement: nil
us
Conflicts of Interest
Ac
ce
p
te
an
None declared.
19
Page 19 of 34
References
Ac
ce
p
te
an
us
cr
ip
t
[1] A. Donabedian, The quality of care. How can it be assessed?, JAMA : the journal
of the American Medical Association, 260 (1988) 1743-1748.
[2] B.W. Ellis, S. Johnson, The care pathway: a tool to enhance clinical governance,
Clinical performance and quality health care, 7 (1999) 134-144.
[3] H. Campbell, R. Hotchkiss, N. Bradshaw, M. Porteous, Integrated care pathways,
Bmj, 316 (1998) 133-137.
[4] B.W. Ellis, S. Johnson, A clinical view of pathways of care in disease management,
International journal of health care quality assurance incorporating Leadership in
health services, 10 (1997) 61-66.
[5] M. Romero Simo, V. Soria Aledo, P. Ruiz Lopez, E. Rodriguez Cuellar, J.L.
Aguayo Albasini, [Guidelines and clinical pathways. Is there really a difference?],
Cirugia espanola, 88 (2010) 81-84.
[6] R.W. Miller, A.G. Lee, J.S. Schiffman, T.C. Prager, R. Garza, P.F. Jenkins, P.
Sforza, A. Verm, D. Kaufman, W. Robinson, E. Eggenberger, R.A. Tang, A practice
pathway for the initial diagnostic evaluation of isolated sixth cranial nerve palsies,
Medical decision making : an international journal of the Society for Medical
Decision Making, 19 (1999) 42-48.
[7] B. Willis, L.T. Kim, T. Anthony, P.C. Bergen, F. Nwariaku, R.H. Turnage, A
clinical pathway for inguinal hernia repair reduces hospital admissions, The Journal of
surgical research, 88 (2000) 13-17.
[8] R.E. Kelly, Jr., A. Wenger, C. Horton, Jr., D. Nuss, D.P. Croitoru, J.P. Pestian, The
effects of a pediatric unilateral inguinal hernia clinical pathway on quality and cost,
Journal of pediatric surgery, 35 (2000) 1045-1048.
[9] J.K. Yeung, J.F. Dautremont, A.R. Harrop, T. Asante, N. Hirani, S.C. Nakoneshny,
V. de Haas, D. McKenzie, T.W. Matthews, S.P. Chandarana, C. Schrag, J.C. Dort,
Reduction of pulmonary complications and hospital length of stay with a clinical care
pathway after head and neck reconstruction, Plastic and reconstructive surgery, 133
(2014) 1477-1484.
[10] L. Zhu, J. Li, X.K. Li, J.Q. Feng, J.M. Gao, Impact of a clinical pathway on
hospital costs, length of stay and early outcomes after hepatectomy for hepatocellular
carcinoma, Asian Pacific journal of cancer prevention : APJCP, 15 (2014) 5389-5393.
[11] T. Tarin, A. Feifer, S. Kimm, L. Chen, D. Sjoberg, J. Coleman, P. Russo, Impact
of a common clinical pathway on length of hospital stay in patients undergoing open
and minimally invasive kidney surgery, The Journal of urology, 191 (2014)
1225-1230.
[12] L. De Bleser, R. Depreitere, K. De Waele, K. Vanhaecht, J. Vlayen, W. Sermeus,
20
Page 20 of 34
Ac
ce
p
te
an
us
cr
ip
t
Page 21 of 34
Tables
05216K
02006K
03001K
57017B
96025B
96026B
32001C
08036B
08011C
an
12 18001C
14 A013718277
15 39004C
16 48011C
17 96007C
18 03026K
te
13 75607C
Ac
ce
p
11 09005C
10 08013C
us
cr
Order Code
ip
t
Cases Recurrence
Avg.
Charlson Avg.
Score
Amount Orders
A District Hospital
170
3(1.76%)
1.32
45,535
26
B Medical Center
158
8(5.06%)
1.61
42,422
34
C Medical Center
120
1(0.83%)
2.40
38,021
32
D Medical Center
100
0(0.00%)
1.28
32,433
33
E Medical Center
79
12(15.19%)
1.85
46,171
31
F Regional Hospital
68
0(0.00%)
2.21
51,266
40
G Medical Center
66
2(3.03%)
2.20
29,791
33
22
Page 22 of 34
66
2(3.03%)
1.29
40,982
34
Medical Center
66
3(4.55%)
1.59
43,067
37
Medical Center
55
0(0.00%)
1.62
38,867
38
K Regional Hospital
55
1(1.82%)
1.87
41,220
40
L Medical Center
49
1(2.04%)
1.43
31,041
38
87.67 2.75(3.13%)
1.72
40,068
34.33
Average
ip
t
H Medical Center
Category Name
cr
Amount
3,722
us
Basic Medical
Laboratory Examination
1,852
203
Injection
Therapeutic Treatment
Anesthesia
13,517
4,263
Special Material
55
1,656
38,864
Ac
ce
p
te
Total
13,346
Other
66
184
Operation
an
X-RAY
26
22
22
22
21
20
82.09%
34
16
15
11
32.68%
32
18
17
17
12
40.66%
33
16
14
13
11
34.92%
31
16
16
16
12
41.00%
40
22
21
18
13
37.78%
33
24
24
22
19
13
59.12%
34
22
20
17
13
45.96%
37
20
19
17
17
13
50.07%
38
23
23
20
17
46.31%
40
19
19
18
14
37.14%
38
22
20
17
14
38.71%
23
Page 23 of 34
Table 5. The comparison of Charlson score, recurrence rate, and categorized amount
G
1.61
2.20
1.62
5.06%
3.03%
0.00%
Basic Medical
3,787
3,826
Laboratory Examination
1,519
1,344
Charlson Score
Recurrence Rate
257
Injection
72
Therapeutic Treatment
Blood Transfusion and Bone Marrow
Transplantation
Anesthesia
Special Material
77
66
234
177
184
13,615
13,816
13,517
3,369
3,224
4,263
22
67
55
18,390
5,828
13,346
1,019
1,230
1,656
42,417
29,787
38,864
133
te
Total
Other
203
an
Operation
1,852
198
us
X-RAY
3,722
cr
ip
t
Care Provider
Ac
ce
p
Order Name
B G J
1 02006K
v v v
2 03001K
v v v
3 03026K
v v v
4 05216K
5 08011C
CBC-I (WBC,RBC,Hb,Hc,platelet
count,MCV,MCH,MCHC )
6 08036B
7 09002C
8 09005C
v v
9 09015C
v v
1 09021C
v v
v v
v v v
24
Page 24 of 34
0
General Biochemistry Examination - K(Potassium)
v v
1
09023C
2
1
09025C
3
1
09029C
4
1
18001C
5
E.K.G. (Electrocardiography)
1
32001C
6
Chest Examination
1
39004C
7
Parenteral Fluid
1
48011C
8
1
57017B
9
2
75607C
0
2
96007C
1
Spinal Anesthesia
ip
t
1
09022C
1
v v v
cr
te
an
us
v v v
v v v
v v v
v v
v v v
v v v
Ac
ce
p
v v
2
96025B
2
v v v
2
96026B
3
Pre-Anesthesia Evaluation
v v v
2
A013382100
4
v v
2
A013718277
5
2
A020444100
6
2
A025485209
7
2
A033698277
8
Antibiotics Cephalosporins
v
v
v
v
25
Page 25 of 34
9 W
3 NDN041622N
Special Material - IV Catheter(Teflon)
0 BD
3 NDN041622N
Special Material -IV Catheter(Teflon)
1 TM
ip
t
cr
P value
us
Recurrence rate 2%
70.99 23.88
52.6020.30
Charlson score
1.871.33
1.880.87
.953
2.214.28
2.892.82
.133
Number of surgeons
at each hospital
3.504.41
7.096.57
<.001
Regional hospital
12
46
33
128
31
Ac
ce
p
District hospital
te
Medical center
<.001
.042
Hospital type
an
Consistency index(%)
Consistency index
Hospital type
Number of surgeons
at each hospital
a
P value
OR
95% CI (lowerupper)
.002
0.973
0.9570.990
.475
1.758
0.3738.279
.421
1.036
0.9501.129
Every 1% increased
Considered as categorical variable
26
Page 26 of 34
Figure legends
Figure 1. The selection of study subjects.
cr
ip
t
us
an
Figure 6. The ratio of frequent physician order sets to average number of orders
te
Abbreviations
Ac
ce
p
Page 27 of 34
ip
t
cr
Ac
ce
p
te
an
us
28
Page 28 of 34
Ac
ce
p
te
an
us
cr
ip
t
Figure 1
Page 29 of 34
Ac
ce
pt
ed
an
us
cr
Figure 2
Page 30 of 34
Ac
ce
pt
ed
an
us
cr
Figure 3
Page 31 of 34
Ac
ce
pt
ed
an
us
cr
Figure 4
Page 32 of 34
Ac
ce
p
te
an
us
cr
ip
t
Figure 5
Page 33 of 34
Ac
ce
pt
ed
an
us
cr
Figure 6
Page 34 of 34