Running head: ANALYSIS OF SURGEON AND ANESTHESIA 1

Analysis of Surgeon and Anesthesia Predict Variability in Surgical Procedure Times

Cejamil S. Calderon

Hahn School of Nursing and Health Sciences, University of San Diego

HCIN 549: Biostatistics

Professor Alexandrea Cronin

20 August 2021
ANALYSIS OF SURGEON AND ANESTHESIA 2

Analysis of Surgeon and Anesthesia Predict Variability in Surgical Procedure Times

The authors for the article are Dr. David Strum associate professor of the Anesthesiology

Department of Queens University Kingston general Hospital, Dr. Allan Sampson professor and

chairman for the Department of statistics at the University of Pittsburgh, Dr. Jerrold May

professor of business administration and intelligence systems for the Joseph M. Katz Graduate

School of Business at the University of Pittsburgh, and Dr. Luis Vargas professor of operations,

decision sciences, and artificial intelligence for the Joseph M. Katz Graduate School of Business

at the University of Pittsburgh.

The cost of health care is steadily rising more and more each year. There is an obvious

economic impact amongst medical institutions around the world. To better manage fiscal

budgetary deficits amongst medical institutions the reduction of costs within the operation suits

must be addressed and thoroughly analyzed. The journal aims to develop statistical models that

identify and estimate the effects of major sources of variability amongst surgical procedure

times.

The value of the retrospective study would ultimately present and describe the variability

and identify the associated factors that may help predict, and control by improving scheduling

and reducing the cost of surgical services (Strum et al., 2000). Due to the personal accounts of

being an anesthesia technologist; this article was reviewed simply because of the particular

interest in the business aspect of surgical care and the cost of healthcare delivery when it comes

to effective surgical procedure times. The article presents the overutilization or under-utilization

pertaining to hospital resources and anesthesia care.

Methodology of Surgeon and Anesthesia Variability Research

ANALYSIS OF SURGEON AND ANESTHESIA 3

The methodology of predicting variability of surgical times amongst the surgeon and type

of anesthesia was compiling electronic health records of all surgical cases performed in an

anonymous hospital over a 7-year period from 1989 through 1995. The hospital's institutional

review board (IRB) had granted authorization to use anonymous patient information. The study

utilizes data sets consisting of 46,322 surgical cases. The independent variables included were

the surgical time (surgeon incision open to close), the total procedure time (patient in and out of

operating room), patient age, gender, surgeon, anesthesiologist, American Society of

Anesthesiologist’s risk class (ASA), type of anesthesia, ICD-9 codes, and Current Procedural

Terminology codes (CPT) (Strum et al., 2000).

The problem's that had arisen during the data collection process was that 60,643 patient

records were originally utilized for the study, 779 were omitted from the study because of

incomplete data pertaining to the surgical times and the total procedure time, thus consisting of

59,864 patient records with one and three CPT codes (Strum et al., 2000). Furthermore, 10,740

patient records had two different CPT codes. The research team had decided to utilize records

that only had one CPT code to reduce confounding factors, thus resulting in a total of 46,322

patient records (Strum et al., 2000). Overall, the methodology of extracting specific data of one

CPT code clearly depicts the study's main objective of proving surgical and anesthesia variability

and reducing the cost of surgical services.

Data Structures, Statistical Analysis, and Results

For the researchers to minimize any potential biased regarding their results, the team had

attempted to data-mine the patient records. As represented in Appendix A Data mining would

make analyzing the data simple and intellectually feasible to process. The data structures

represented four categories from original data, Initial data, Representative data, and Analytic
ANALYSIS OF SURGEON AND ANESTHESIA 4

data. The research team had utilized the Initial data that consists of 46,322 surgical cases, 3096

single CPT codes, 268 surgeons, and 151 anesthesiologists (Strum et al., 2000). The research

team had utilized the Initial database by developing a table that depicted 20 surgical categories

based on 20 primary headings of the CPT with the three most used CPT codes, as shown in

appendix B (Strum et al., 2000). The research team had realized that there were imbalances in

the 20 categories between the CPT codes and the total surgical times and total procedure time.

The imbalances were driven by anatomy, physiology, and technology which dictated the number

of surgical procedures encompassed by the primary CPT code categories (Strum et al., 2000).

The research team had utilized descriptive statistics to summarize the ST and TT for the 20

categories, as shown in appendix C and D. The team would later utilize the data subset of the

representative database containing 13,196 cases and the remaining 40 CPT codes as shown in the

appendix F (Strum et al., 2000). At the time of the study, the software and hardware technology

were not adequate to achieve a standard of statistical modeling. The research team had further

reduced the sample size of 12,125 cases amongst the 40 CPT codes: classifying it as the

analytical database (Strum et al., 2000). The reduction in cases was required due to analysis of

variance modeling (ANOVA) because one factor, ASA risk class was not recorded whenever

surgery was performed with local anesthesia. Due to the missing ASA values eliminating all

missing values would have eliminated local anesthesia from the studies analysis (Strum et al.,

2000).

As for the statistical analysis the research team had utilized descriptive statistics to

display the representative databases and utilized a six-factor analysis of variance (ANOVA) to

examine anesthesiologist factors as illustrated in Appendix B through H. Furthermore, the

research team utilized a multivariate linear regression model that included five factors to indicate
ANALYSIS OF SURGEON AND ANESTHESIA 5

the natural logarithm (Ln) as presented in Appendix I through N (Strum et al., 2000). Lastly, the

researcher team had determined that within each CPT code a significance level was used to test

each factor was P= 0.05 if the P-value represented was less than 0.05 then the CPT code would

be statistically significant for the ST and TT (Strum et al., 2000).

The results of the study were categorized by each database. The Initial database presented

a total of 8,834,789 mins of total operating room usage (TT) and 6,109,704 minutes of surgical

procedures (ST) (Strum et al., 2000). Furthermore, 1,424,197 minutes if initial anesthesia time

accounted for. 639,240 minutes of preparation and surgical positioning, and 661,648 minutes of

anesthesia emergence and transport time (Strum et al., 2000). Among each procedural category,

cardiovascular surgery represented the most surgical and operating room time. The

representative database presented a total of 2,316,472 minutes of TT and 1,633,817 minutes of

ST (Strum et al., 2000). Anesthesia time was indicated as 368,220 minutes, total preparation, and

positioning time 142,967 minutes, and 171,468 of anesthesia emergence and transport time

(Strum et al., 2000). The representative database reflects the initial database, and the

characteristics of the surgeries are within the parameters of contemporary surgery and anesthetic

academic practice (Strum et al., 2000). The analytical database and the representative database

drew strong comparisons towards each other. The analytic database was presented by developing

an ANOVA table as presented in Appendix H. To evaluate the correlation between the

independent variables, with the ST and TT the research team had then conducted a multivariate

linear regression model to present the data as presented in Appendix I through N (Strum et al.,

2000).

Analysis on Conclusions of Surgeon and Anesthesia Variability Study

ANALYSIS OF SURGEON AND ANESTHESIA 6

The research team had utilized the data mining approach effectively and was able to

synthesize the large and diverse data set by cataloging the data from the primary forty CPT code

headers (Strum et al., 2000). The research team had determined that the most influential source

of variability in procedural times was the surgeon itself (Strum et al., 2000). The surgeon

competency, experience, and workflow were the key indicators of increased or decreased in the

duration of the procedural time (Strum et al., 2000). The surgeon work rate has an important

implication because it can solely determine the cost over utilization or under-utilization of

hospital resources (Strum et al., 2000). Additional influential factors would also include the type

of anesthesia, gender, age, and ASA class (Strum et al., 2000). Furthermore, case-specific

relativity would be a factor due to the turnover (proper cleaning) of each room in-between cases

(Strum et al., 2000). The complexity of the surgical case would equal the more personnel being

staffed, equipment, and resources used in each complex case thus cleaning and prepping at the

end of each case takes a longer amount of time.

The research team had stated two strategies that may control hospital costs. One strategy

would be taking account of the specific surgeon for a specific scheduled case (Strum et al.,

2000). For example, having an experienced and tenured cardiac surgeon for triple bypass

surgery. The last strategy that the research team indicated was to schedule the more complex and

lengthy surgical cases first (Strum et al., 2000). This would give the surgical schedule more

versatility throughout the workday thus minimizing cost amongst the hospital by not

underutilizing or over-utilizing resources (Strum et al., 2000). For example, limiting registered

nurses or surgical technicians to logging in overtime. Lastly, the research team had indicated that

having a competent, well established administrative and scheduling team are key factors to drive

effective scheduling and allocation of resources amongst the hospital.

ANALYSIS OF SURGEON AND ANESTHESIA 7

Strengths and Weakness of the Surgeon and Anesthesia Variability Study

Overall, the journal article was well organized but the statistical methods that were

utilized in the variability study were overly complex. The study utilized multiple linear

regression modules that seem to be repetitive. There were numerous limitations to this study. The

classification and selection of CPT codes and subdivision of those CPT codes was a limiting

factor to the study (Strum et al., 2000). CPT codes are the single most important reporting tool to

a modern-day physician. The CPT Code and its nomenclatures is the universal language that

helps the administration of the hospital with claims processing, serves as the foundation for local,

regional, and national utilization analysis (Strum et al., 2000). Furthermore, the research team

reducing the diversity of patient population and physician population within their study was a

limiting factor (Strum et al., 2000). This would include certain procedures not being accounted

for and minimizing the experience and inexperienced physicians (residents, fellows.) (Strum et

al., 2000). Lastly, the research team only utilized a single academic health care institution;

therefore, creating bias within the hospital's subspecialties (Strum et al., 2000). According to the

research team, this compromise was a necessity to obtain sufficient data from the specific cases

(Strum et al., 2000).

The research team thoroughly illustrated synthesizing the data by adapting the data

mining approach especially at the time of study. The illustrations, graphs, and tables that were

presented had helped understand the statistical analysis. The study can be further improved by

analyzing multiple institutions and having a surgeon-specific databases that would describe

factors affecting the variability of surgeries are needed to improve scheduling (Strum et al.,

2000). This study was conducted over 20 years ago, if this study was conducted today, it would

have been easier and more precise to synthesize the large data set. The research community has
ANALYSIS OF SURGEON AND ANESTHESIA 8

seen dramatic changes in statistical analysis software with the introduction of innovating

programs such as Tableau, SPSS, and Python just to name a few. Utilizing these programs can

diversify and prevent bias for future studies concerning surgeon and anesthesia variability of

surgical times and developing future strategies to save cost towards a health care institution.
ANALYSIS OF SURGEON AND ANESTHESIA 9

References

Strum, D. P., Sampson, A. R., May, J. H., & Vargas, L. G. (2000, May 1). Surgeon and Type of

Anesthesia Predict Variability in Surgical Procedure Times. Anesthesiology.

https://doi.org/10.1097/00000542-200005000-00036.
ANALYSIS OF SURGEON AND ANESTHESIA 10

Appendix A

Schematic illustration of the hierarchy of databases produced by data mining.

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000
ANALYSIS OF SURGEON AND ANESTHESIA 11

Appendix B

Initial Database: CPT Codes and Brief Definitions for the Three Most Numerous
Procedures in Each Category, and the Range of Codes in Each Category

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000
ANALYSIS OF SURGEON AND ANESTHESIA 12

Appendix C

Initial Database: Descriptive Statistics Summarizing ST for Each of the 20 Categories

Cat N P Sum % Mean SD Median Min Max
1 3,702 254 313,234 5.13 85 81 60 5 805
2 9,348 912 1,105,155 18.1 118 84 102 3 960
3 2,195 219 240,327 3.93 109 91 85 2 650
4 8,290 296 1,510,173 24.7 182 132 175 2 1,055
5 647 27 81,544 1.33 126 94 108 8 595
6 80 12 8,079 0.13 101 84 73 23 555
7 4,135 103 663,491 10.9 160 158 105 4 1,130
8 4,638 169 352,742 5.77 76 114 33 1 980
9 1,131 82 149,810 2.45 132 102 115 2 745
10 0 0 0 0 0 0 0 0 0
11 464 27 50,715 0.83 109 69 96 5 497
12 2,748 126 316,468 5.18 115 93 89 5 755
13 483 26 21,130 0.35 44 32 35 5 215
14 232 20 40,350 0.66 174 90 154 40 698
15 3,448 243 781,436 12.8 227 165 195 1 1,035
16 3,174 182 326,381 5.34 103 63 90 2 465
17 639 65 77,109 1.26 121 74 107 5 734
18 292 14 17,632 0.29 60 29 55 5 300
19 2 1 96 0 48 31 48 26 70
20 674 15 53,832 0.88 80 60 70 1 485
Cat 5 category number; N 5 number of cases in the category; P 5 number of procedures in the
category; Sum 5 total ST (min) for cases in the category; % 5 percent of ST in each category
relative to the total; SD 5 standard deviation (min); Min 5 minimum ST (min); Max 5 maximum
ST (min)(46,322 cases and3,096 different procedures in the initial database).
Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000
ANALYSIS OF SURGEON AND ANESTHESIA 13

Appendix D

Initial Database: Descriptive Statistics Summarizing TT for Each of the 20 Categories

Cat N P Sum %Total Mean SD Median Min Max
1 3,702 254 471,169 5.33 127 93 100 12 850
2 9,348 912 1,689,938 19.1 181 97 165 14 1055
3 2,195 219 375,480 4.25 171 106 145 5 725
4 8,290 296 2,087,762 23.6 252 158 255 5 1185
5 647 27 117,346 1.33 181 110 160 30 675
6 80 12 13,144 0.15 164 95 135 85 685
7 4,135 103 934,943 10.6 226 181 165 20 1320
8 4,638 169 545,869 6.18 118 135 70 10 1065
9 1,131 82 200,066 2.26 177 121 150 18 895
10 0 0 0 0 0 0 0 0 0
11 464 27 79,241 0.9 171 79 160 60 572
12 2,748 126 472,573 5.35 172 105 145 25 884
13 483 26 41,558 0.47 86 50 70 30 625
14 232 20 57,168 0.65 246 104 225 75 874
15 3,448 243 1,035,385 11.7 300 177 275 25 1150
16 3,174 182 480,357 5.44 151 70 135 15 555
17 639 65 109,206 1.24 171 85 160 5 870
18 292 14 32,977 0.37 113 37 109 30 355
19 2 1 185 0 93 25 92.5 75 110
20 674 15 90,422 1.02 134 71 125 7 585
Cat 5 category number; N 5 number of cases in the category; P 5 number of procedures in the category;
Sum 5 total ST (min) for cases in the category; % 5 percent of ST in each category relative to the total; SD
5 standard deviation (min); Min 5 minimum ST (min); Max 5 maximum ST (min)(46,322 cases and
3,096 different procedures in the initial database).
Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000
ANALYSIS OF SURGEON AND ANESTHESIA 14

Appendix F

Representative Database: Definitions of the 40 Procedures and Summary Statistics for Age
(in years)

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000
ANALYSIS OF SURGEON AND ANESTHESIA 15

Appendix G

Representative Database: Statistics Summarizing ST and TT

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000
ANALYSIS OF SURGEON AND ANESTHESIA 16

Appendix H

Analytic Database: Tabular Comparisons of Main Effect ANOVAs by Level of Significance

for ST and TT (12,125 Cases and 40 Procedures in the Analytic Database C)
Table 6. Analytic Database: Tabular Comparisons of Main Effect ANOVAs by Level of
Significance for ST and TT (12,125 Cases and 40 Procedures in the Analytic Database C)
P value Age Anesthesia ASA Gender Surgeon
Surgical time
(ST)

P<.05 7(17.5) 15(44.1) 4(14.3) 6(19.4) 30(75)

p>.05 33(82.5) 19(55.9) 24(85.7) 25(80.6) 10(25)

Analyzed 40(100) 34(100) 28(100) 31(100) 40(100)

Omitted 0 6 12 9 0
Total 40 40 40 40 40

Total Time (TT)

P<.05 5(12.5) 18(52.9) 5(17.9) 6(19.4) 30(75)

P>.05 35(87.5) 16(47.1) 23(82.1) 25(80.6) 10(25)

Analyzed 40(100) 34(100) 28(100) 31(100) 40(100)

Omitted 0 6 12 9 0

Total 40 40 40 40 40

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000

Appendix I
ANALYSIS OF SURGEON AND ANESTHESIA 17

Boxplot comparisons of ST ordered by increasing median surgical time (minutes).

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000

Appendix J
ANALYSIS OF SURGEON AND ANESTHESIA 18

Plot of the surgeon-adjusted means of the natural log of ST against the means of the
natural log of ST (LnST) (12,125 cases, 40 procedures, and 96 surgeons in the analytic
database).

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000

Appendix K
ANALYSIS OF SURGEON AND ANESTHESIA 19

Plot of the unlogged, adjusted, surgeon median ST plotted against the weighted median ST
for each procedure, illustrating proportionately increased variability with longer
procedure times (n 5 12,125 cases, 40 procedures, and 96 surgeons in the analytic
database).

N
ote. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000

Appendix L
ANALYSIS OF SURGEON AND ANESTHESIA 20

Plot of the square root of the mean square error of the five-factor ANOVA model versus
the mean LnST for each procedure (12,125 cases, 40 procedures, and 96 surgeons in the
analytic database).

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000

Appendix M
ANALYSIS OF SURGEON AND ANESTHESIA 21

Analytic database: plot of the F ratio (mean square error of the main-effect factor surgeon
divided by mean square error of the total regression using the five-factor ANOVA model)
with respect to weighted mean LnST (12,125 cases, 38 procedures, with 2 outliers omitted
from the analytic database).

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 2000

Appendix N
ANALYSIS OF SURGEON AND ANESTHESIA 22

Plot of the adjusted anesthesia median ST (minutes) versus median ST for each procedure.
Closed circle 5 local anesthesia; open circle 5 general anesthesia; closed triangle 5 regional
anesthesia: open triangle 5 monitored anesthesia (12,125 cases and 40 procedures in the
analytic database).

Note. Surgeon and Type of Anesthesia Predict Variability in Surgical Procedure Times, 200