Content uploaded by Htun Teza
Author content
All content in this area was uploaded by Htun Teza on Jan 23, 2024
Content may be subject to copyright.
External validation and revision of Penn incisional
hernia prediction model: A large-scale
retrospective cohort of abdominal operations
Amarit Tansawet
a,b
, Pawin Numthavaj
a,*
, Htun Teza
a
,
Anuchate Pattanateepapon
a
, Pongsathorn Piebpien
c
, Napaphat Poprom
d
,
Suphakarn Techapongsatorn
b
, Gareth McKay
e
, John Attia
f
,
Preeda Sumritpradit
d
, Ammarin Thakkinstian
a
a
Department of Clinical Epidemiology and Biostatistics, Faculty of Medicine Ramathibodi Hospital, Mahidol
University, Bangkok, Thailand
b
Department of Surgery, Faculty of Medicine Vajira Hospital, Navamindradhiraj University, Bangkok, Thailand
c
Information Technology Department, Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok,
Thailand
d
Department of Surgery, Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok, Thailand
e
Centre for Public Health, School of Medicine, Dentistry, and Biomedical Sciences, Queen's University Belfast,
Northern Ireland, UK
f
Centre for Clinical Epidemiology and Biostatistics, Hunter Medical Research Institute, School of Medicine and Public
Health, The University of Newcastle, New Lambton, New South Wales, Australia
article info
Article history:
Received 20 May 2023
Received in revised form
20 July 2023
Accepted 24 July 2023
Available online 7 August 2023
Keywords:
Abdominal surgery
External validation
Incisional hernia
Prediction score
abstract
Background: Incisional hernia (IH) manifests in 10%e15% of abdominal surgeries and pa-
tients at elevated risk of this complication should be identified for prophylactic interven-
tion. This study aimed to externally validate the Penn hernia risk calculator.
Methods: The Ramathibodi abdominal surgery cohort was constructed by linking relevant
hospital databases from 2010 to 2021. Penn hernia risk scores were calculated according to
the original model which was externally validated using a seven-step approach. An
updated model which included four additional predictor variables (i.e., age, immunosup-
pressive medication, ostomy reversal, and transfusion) added to those of the three original
predictors (i.e., body mass index, chronic liver disease, and open surgery) was also eval-
uated. The area under the receiver operating characteristic curve (AUC) was estimated, and
calibration performance was compared using the HosmereLemeshow goodness-of-fit
method for the observed/expected (O/E) ratio.
Results: A total of 12,155 abdominal operations were assessed. The original Penn model
yielded fair discrimination with an AUC (95% confidence interval (CI)) of 0.645 (0.607, 0.683).
The updated model that included the additional predictor variables achieved an acceptable
AUC (95% CI) of 0.733 (0.698, 0.768) with the O/E ratio of 0.968 (0.848, 1.088).
Conclusion: The updated model achieved improved discrimination and calibration perfor-
mance, and should be considered for the identification of high-risk patients for further
hernia prevention strategy.
©2023 Published by Elsevier Ltd on behalf of Royal College of Surgeons of Edinburgh
(Scottish charity number SC005317) and Royal College of Surgeons in Ireland.
*Corresponding author. Department of Clinical Epidemiology and Biostatistics, Faculty of Medicine, Ramathibodi Hospital, Mahidol
University, Rama VI Road, Ratchathewi, Bangkok 10400, Thailand. Fax: þ66 22011284.
E-mail address: pawin.num@mahidol.ac.th (P. Numthavaj).
the surgeon 22 (2024) e34ee40
https://doi.org/10.1016/j.surge.2023.07.008
1479-666X/©2023 Published by Elsevier Ltd on behalf of Royal College of Surgeons of Edinburgh (Scottish charity number SC005317) and
Royal College of Surgeons in Ireland.
Introduction
Incisional hernia (IH), a protrusion of visceral tissue at the
area of an incision due to incomplete surgical wound healing,
occurs in approximately 10e15% of post-surgical proced-
ures.
1
Control group evidence from randomized controlled
trials (RCTs) investigating prophylactic mesh placement
during fascia closure suggested an 11.4e52.3% risk of IH in
high-risk patients.
2e4
Data from cost analysis of IH repair
indicated that reduced IH incidence could represent signifi-
cant cost savings.
5
As such, accurate identification of high
risk IH patients in need of prevention interventions is
essential.
Many risk factors associated with IH have been identified,
including type of surgery, high body mass index (BMI), and
surgical site infection (SSI).
6,7
However, accurate identification
of high risk IH patients is important in order for them to
receive prophylaxis intervention. We completed a systematic
review and reported several prediction models developed for
this purpose following general abdominal surgery.
8e11
Of
these models,
8e11
the number of predictor variables ranged
from 3 to 17, and their discriminatory performance ranged
from 0.77 to 0.92 in terms of concordance statistics; the Penn
Hernia Risk Calculator
11
was the most recent and available as a
mobile phone application. This model can be applied to all
types of abdominal surgery and is considered to offer clinical
utility, although it remains to be externally validated.
Prediction models generally perform well in the discovery
test cohort but are less specific and sensitive when validated
externally. As such, every prediction model should be exter-
nally validated and revised before their application in
different populations.
12,13
This study aims to validate the Penn
Hernia Risk Calculator in the Thai Ramathibodi abdominal
surgical dataset and improve model performance, as appro-
priate, to better identify high risk IH patients for targeted
prophylactic intervention.
Materials and methods
Study design
We constructed a retrospective cohort of adult abdominal
surgical patients in Ramathibodi Hospital from January 2010 to
August 2021. This cohort was compared with the original Penn
cohort, which included 29,739 patients undergoing intra-
abdominal operations from January 2005 to June 2016.
11
The
data were retrieved from different sources using International
Statistical Classification of Diseases and Related Health Prob-
lems (ICD) codes for operation and diagnosis (ICD-10), labora-
tory, medication, and billing data using linked encrypted
patient identification (i.e., hospital and admission number).
Eligible patients were identified using ICD-9-CM for intra-
abdominal operations if they were 18 years or older, not
pregnant or in the postpartum period, and underwent intra-
abdominal surgery not related to abdominal wall hernia, see
Supplementary Fig. 1. Patients whose IH was diagnosed before
their operation were also excluded. This study adhered to the
Transparent Reporting of a Multivariable Prediction Model for
Individual Prognosis or Diagnosis (TRIPOD) reporting
guidelines.
14
Only records with complete data for the Penn IH model's
predictors were included for external validation. Sixteen pre-
dictor variables were included in the original Penn IH model,
two of which were not considered as only Thai nationals were
included and the Elixhauser comorbidity score was not per-
formed as part of routine patient assessment. The 14
remaining predictor variables were available and used for
validation, including 8 preoperative factors (i.e., BMI, smoking
status, chronic obstructive pulmonary disease (COPD),
chronic liver disease, cancer, history of chemotherapy/radia-
tion therapy, antiplatelet/anticoagulant use, previous
abdominal surgery) and 6 intra-operative factors (i.e., open
approach, emergency surgery, emergency vascular surgery,
laparoscopic hysterectomy, concurrent ostomy procedure,
and small bowel obstruction). The outcome of interest was
any IH post-surgery, which was identified by incisional hernia
diagnosis (ICD-10) or incisional hernia repair (ICD-9-CM).
The following data were retrieved: patients’baseline
characteristics (age, sex, BMI), American Society of Anesthe-
siologists (ASA) physical status classification, smoking status,
underlying diseases (i.e., COPD, chronic liver disease, and
diabetes), cancer, chemotherapy and radiation therapy, con-
current medication (i.e., antiplatelet/anticoagulant and
immunosuppressive medication), history of abdominal sur-
gery, history of incisional hernia repair, surgical factors (i.e.,
wound classification, open approach, emergent laparotomy,
emergent vascular procedure, concurrent ostomy procedure,
ostomy reversal, colorectal procedure, laparoscopic hyster-
ectomy, small intestinal obstruction, and inflammation pa-
thology), transfusion, intensive care unit admission, post-
operative complications (i.e., SSI, wound complication,
pneumonia), and IH occurrence.
List of abbreviations
ASA American Society of Anesthesiologists
AUC Area under the receiver operating
characteristic curve
BMI Body mass index
CI Confidence interval
COPD Chronic obstructive pulmonary disease
IH Incisional hernia
IQR Interquartile range
NPV Negative predictive value
O/E ratio The observed/expected ratio
PPV Positive predictive value
RCT Randomized controlled trial
SD Standard error
SSI Surgical site infection
the surgeon 22 (2024) e34ee40 e35
Statistical analysis
Data were described by frequency and percentage for cate-
gorical variables, mean and standard error (SD) or median and
interquartile range (IQR) for continuous data. Summary
characteristics and risk factors were compared between the
Ramathibodi and Penn cohorts
11
using Chi-square tests. Pre-
dictor variables were regressed on IH occurrence using uni-
variate and multivariate logistic equations, and the
coefficients and 95% confidence intervals (95% CIs) estimated.
The Penn model was validated as follows (additional details
are provided in Appendix 1)
13
: 1) Composite risk scores were
calculated based on the original model,
11
then against the IH
outcome by logistic regression to assess the original model
performance in the Ramathibodi dataset. 2) Model coefficient
revision was performed by adding each original model pre-
dictor variable individually to the model containing only the
risk score, and only significant predictors were retained.
3) Potential predictors, not considered in the original model
but that were significantly associated with IH occurrence were
added to the original model. 4) All original predictors were re-
fitted on IH outcome in multivariate logistic regression to re-
estimate b-coefficients based on the Ramathibodi cohort
data. 5) Only significant predictors from step 4 were retained
in the model. 6) Only predictors identified in step 3 and 5 were
simultaneously considered and only significant predictors
were retained. 7) As per step 6, with only pre-operative and
intra-operative predictors considered.
Discrimination performance was assessed by estimating
concordance statistics (i.e., the area under the receiver oper-
ating characteristic curve (AUC)). HosmereLemeshow
goodness-of-fit chi-square tests of the observed/expected
outcome (O/E) ratio, and the O/E plot, were used to assess
calibration performance.
The best model was selected on the basis of both
discrimination and calibration performance. A composite risk
score was constructed based on coefficients for the selected
model, and was further categorized based on the distribution
frequencies at 25th, 50th, and 75th percentiles as a cut-off.
Then, sensitivity, specificity, positive and negative predictive
values (PPV and NPV), and likelihood ratios were estimated for
each cut-off. Significance was considered for p-value <0.05.
Stata version 17 (StataCorp, Texas, USA) was used for all sta-
tistical analyses.
Results
Characteristics of patients
A total of 423,704 operations were recorded in the Ram-
athibodi Surgery databases for the period January 2010 to
August 2021, see Supplementary Fig. 1. Of these, 18,358 were
identified as abdominal surgeries using ICD-9-CM codes for
various kinds of intra-abdominal procedures. Of the 18,358
abdominal surgeries, 16,731 records met our inclusion criteria,
although only 12,155 records (11,617 patients) had complete
data and were included for external validation of the Penn IH
model. The median follow-up time (IQR) was 23.4 (6.3e52)
months. The mean age (SD) was 57 (16.1) years, and 38.4% of
patients were male. Biliary surgery was the most frequently
performed procedure (41.5%), followed by gastrointestinal
(24%), colorectal (19.5%), and gynecologic procedures (10.2%).
A total of 178 out of 12,155 patients had IH occurrence with an
incidence (95% CI) of 1.5% (1.3%, 1.7%).
Predictive factors
Significant differences between the Ramathibodi patients
and the Penn cohort were observed, see Table 1. Among 14
predictor variables included in the Penn model, two pre-
dictors (i.e., emergency vascular surgery and laparoscopic
hysterectomy) had no IH occurrence, and therefore their
coefficients could not be estimated leaving the 12 remaining
predictors to calculate a Penn risk score. Of these, six pre-
dictors (i.e., BMI, chronic liver disease, antiplatelet/antico-
agulant use, open surgery, concurrent ostomy, and previous
abdominal surgery) were significantly associated with IH in
the Ramathibodi data, see Supplementary Table 1. All sig-
nificant predictors had the same directions of association as
in the Penn cohort. Two of these 6 predictor variables (i.e.,
open surgery and previous abdominal surgery) had similar
coefficients in both the Ramathibodi and Penn datasets, 0.36
versus 0.35 and 0.82 versus 0.85, respectively, in contrast to
the remaining coefficients which were substantially
different.
Performance of Penn model
External validation of the Penn model was based on the orig-
inal weighted score and predictor variable coefficients as
previously reported,
11
see Table 2. The original model pro-
vided fair discrimination for both coefficient and weighted
score approaches (step 1) with AUCs (95% CI) of 0.634 (0.595,
0.674) and 0.645 (0.607, 0.683), respectively.
Model revision
Model revision (step 2, 4, and 5) focused on the original pre-
dictors and showed little improved performance with AUCs
(95% CI) of 0.679 (0.641, 0.717), 0.692 (0.655, 0.729), and 0.689
(0.652, 0.726), respectively. Additional predictors significantly
associated with IH identified from univariate regression
(Supplementary Table 2) were included in the model (step 3),
with improved discrimination performance of 0.729 (0.693,
0.765). Step 6, which simultaneously considered the original
significant predictors from step 5 and additional predictors
from step 3, improved the AUC to 0.743 (0.707, 0.778). Finally,
step 7, which considered only pre- and intra-operative pre-
dictors and excluded SSI from the model, resulted in an AUC
of 0.733 (0.698, 0.768). All models demonstrated good calibra-
tion performance, where the O/E ratio ranged from 0.968 to
1.031. More details from each validation step are described in
Appendix 2.
the surgeon 22 (2024) e34ee40e36
The final model (step 7) included only pre- and intra-
operative data and may prove more clinically applicable,
given its acceptable discrimination and calibration perfor-
mance (Table 2 and Fig. 1). The following equation was con-
structed based on the predictor variable coefficients derived
from step 7 (Supplementary Table 3).
ln"Pþ
IH
1Pþ
IH#¼5:71 þ1:11xðAge 45 65Þþ1:63xðAge >65Þ
0:39xðBMI <18Þ0:57xðBMI 18 25Þ
þ0:64xðBMI >30Þþ0:92xðCirrhosisÞ
þ0:74xðImmunosuppressive drugÞ
þ0:50xðOpen surgeryÞþ2:06xðOstomy reversalÞ
þ0:60xðTransfusionÞ
The risk scores calculated based on predictor variable
coefficients ranged from 6.28 to 1.38, which were strati-
fied into very low, low, moderate, and moderate-high based
on thresholds of 5.17, 4.60, and 4.07 representing the
25th, 50th, and 75th percentiles, see Table 3. Sensitivity,
specificity, PPV, and likelihood ratios are presented in
Table 3.
Discussion
The original Penn
11
score provided reasonable discriminatory
performance in their original datasetwith an AUC of 0.84 but its
Table 2 ePenn model performance validation in the
Ramathibodi cohort data.
Step Model AUC (95% CI) O/E (95% CI)
1 Coefficient 0.634 (0.595, 0.674) 1.031 (0.930, 1.132)
Weighted score 0.645 (0.607, 0.683) 1.021 (0.897, 1.145)
2 Coefficient 0.646 (0.607, 0.684) 1.026 (0.919, 1.134)
Weighted score 0.679 (0.641, 0.717) 1.006 (0.906, 1.106)
3 Coefficient 0.727 (0.691, 0.763) 0.984 (0.847, 1.120)
Weighted score 0.729 (0.693, 0.765) 0.984 (0.894, 1.074)
4 0.692 (0.655, 0.729) 0.978 (0.875, 1.081)
5 0.689 (0.652, 0.726) 0.995 (0.891, 1.100)
6 0.743 (0.707, 0.778) 0.967 (0.861, 1.072)
7 0.733 (0.698, 0.768) 0.968 (0.848, 1.088)
AUC the area under the receiver operating characteristic curve, CI
confidence interval, O/E the observed/expected outcome ratio.
Table 1 eSummary characteristics for Ramathibodi and Penn cohorts.
Predictors, n (%) Penn cohort (N ¼29,739) Ramathibodi cohort (N ¼12,155) P-value
Incisional hernia 1127 (3.8) 178 (1.5) <0.001
Race, Caucasian 18,702 (62.8) NA
Age, years
<45 8837 (29.7) 2887 (23.8) <0.001
45e65 13,895 (46.7) 5168 (42.5)
>65 7007 (23.5) 4100 (33.7)
Sex, male 10,894 (36.6) 4667 (38.4) 0.001
BMI, kg/m
2
<18 1103 (3.7) 662 (5.5) <0.001
18e25 8021 (26.9) 6811 (56.0)
>25e30 9928 (33.4) 3451 (28.4)
>30 10,687 (35.9) 1231 (10.1)
Smoker 8102 (27.2) 27 (0.2) <0.001
COPD 8632 (29.0) 207 (1.7) <0.001
Hypertension 14,776 (49.6) 3798 (31.3) <0.001
Diabetes 5720 (19.2) 1463 (12.0) <0.001
Cirrhosis NA 206 (1.7) NA
2þElixhauser comorbidity score 18,711 (62.9) NA NA
Cancer 6654 (22.3) 3853 (31.7) <0.001
Chemotherapy/Radiotherapy 1306 (4.4) 1954 (16.1) <0.001
Antiplatelet/Anticoagulant 3016 (10.1) 1572 (12.9) <0.001
Emergency surgery 3523 (11.8) 3434 (28.3) <0.001
Open surgery 11,628 (39.1) 5431 (44.7) <0.001
Concurrent Ostomy NA 753 (6.2) NA
Ostomy reversal NA 56 (0.5) NA
Small bowel resection NA 416 (3.4) NA
Large bowel surgery
Partial colectomy NA 1902 (15.7) NA
Proctectomy NA 288 (2.4) NA
Emergency vascular procedure 354 (1.2) 2 (0.02) <0.001
Laparoscopic hysterectomy 2446 (8.2) 92 (0.8) <0.001
History of abdominal surgery 3781 (12.7) 652 (5.4) <0.001
Small bowel obstruction 3561 (11.9) 508 (4.2) <0.001
Wound complication NA 660 (5.4) NA
BMI body mass index, COPD chronic obstructive pulmonary disease, NA not available.
the surgeon 22 (2024) e34ee40 e37
performance decreased when evaluated in the Ramathibodi
data (AUC ¼0.645). There may be several reasons to explain the
difference observed. First, IH incidence in the Ramathibodidata
was approximately 2-fold lower than in the Penn data,
11
i.e.,
1.5% vs 3.8%. Second, there were significant differences in the
characteristicsand risk factors between both cohorts, see Table
1. As such, only six out of the 14 original predictor variables
were significant in the Ramathibodi dataset; all had the same
direction of association in both cohorts. However, only three of
the 16 original predictors were retained in the revised/updated
models. In addition, the significant original predictor variables
(i.e., emergency surgery which was the most significant),
emergent vascular procedure, and laparoscopic hysterectomy
were not significantly associated with IH in the Ramathibodi
data, which were likely significant contributors to the variation
in model performance observed across both datasets. These
findings support the need for model revision and validation in
external independent datasets.
Additional predictor variables were considered in revision
steps if they were identified from other IH prediction
models
8,10,15
or fascial dehiscence
16e18
or were significantly
associated with IH in univariate logistic regression
(Supplementary Table 2). Even though the Elixhauser comor-
bidity score was not available, ASA classification which
captures patient's status was considered in this step. However,
it was removed from the model during stepwise selection.
Integration of the new predictor variables, including age,
immunosuppressive medication, ostomy reversal, SSI, and
transfusion significantly improved the Ramathibodi model
performance with an AUC of 0.743 and the O/E ratio of 0.967
(step 6). Given the reported 178 IH cases in the Ramathibodi
dataset and the rule of thumb that requires ten events per
predictor variable, the model derived in step 6 was less likely to
suffer from model overfitting.
Surgical techniques incorporated during abdominal fascia
closure such as small-bite fascial suturing
19
and mesh rein-
forcement can minimize IH incidence.
2e4
Unfortunately, in-
formation for neither small-bite fascial closure nor
prophylactic mesh placement was available in our electronic
databases. Recent meta-analyses
20e22
have provided evidence
of the benefits associated with mesh on hernia prevention,
especially with regard to onlay and retromuscular place-
ment.
20,22,23
Therefore, identification of patients at higher IH
risk based on information available before or during surgery
(step 7) would be clinically helpful for fascia-enhanced pro-
phylactic intervention allocation. As such, incorporation of
post-operative predictors such as SSI may offer limited value
to enable IH prophylactic intervention. Nevertheless,
Fig. 1 eRevised incisional hernia prediction model performance for abdominal surgery a) Receiver operating characteristic
curve b) Calibration plot.
Table 3 eRevised Ramathibodi incisional hernia risk classification score using only pre- and intra-operative predictor
variables.
Thresholds Sensitivity (%) Specificity (%) PPV (%) LR (þ)
6.28 100 (97.9, 100) 5.9 (5.5, 6.3) 1.6 (1.3, 1.8) 1.06 (1.06, 1.07)
5.17 97.2 (93.6, 99.1) 19.9 (19.2, 20.6) 1.8 (1.5, 2.1) 1.21 (1.18, 1.25)
4.60 77.5 (70.7, 83.4) 55.6 (54.7, 56.5) 2.5 (2.1, 3.0) 1.75 (1.61, 1.89)
4.07 58.4 (50.8, 65.8) 74.1 (73.3, 74.9) 3.3 (2.7, 3.9) 2.26 (1.99, 2.57)
LR likelihood, PPV positive predictive value, 95% confidence intervals are shown in parentheses.
the surgeon 22 (2024) e34ee40e38
although SSI was removed from the risk prediction model, its
importance should not be overlooked as opportunities to
reduce post-operative SSI would likely result in lower IH risk.
Our model performance was less than that reported for the
HERNIA score,
9
and Fischer et al.‘s models,
10
which yielded
AUCs of 0.77,
9,10
and much lower than that of Veljkovic et al.
8
(AUC ¼0.92). The Veljkovic model was based on data from 603
patients and included only 4 predictor variables (BMI, suture
length to incision length ratio, time to suture removal or
complete epithelialization, and SSI) with relatively short
follow-up time (6.9 ±2.1 months),
8
representing one pre-
operative, one intra-operative, and two post-operative factors.
The HERNIA score is a well-known IH prediction model
derived from data from 428 patients using only 3 predictor
variables (BMI, COPD, and surgical approach [laparotomy or
hand-assisted laparoscopy]).
9
Given that current procedures
tend to be limited to minimally invasive laparoscopic tech-
niques, hybrid procedures such as hand-assisted laparoscopy
is relatively unpopular, which perhaps makes the HERNIA
score model less applicable.
The Penn hernia risk calculator was derived from the original
Fischer et al. model by the same group
10
using Cox regression
on data from 12,373 patients. Seventeen predictors were orig-
inally included in the model; six related to surgical procedures
(bariatric surgery, small bowel resection, proctectomy, partial
colectomy, ostomy creation, and ostomy reversal). Unlike the
Penn hernia risk calculator,
11
Fischer's model focused solely on
patients undergoing elective open abdominal surgery. Thus,
generalizability of this model to laparoscopic surgery or acute
procedures is questionable. Of the four IH models, only the
HERNIA score has been externally validated and revised,
although model performance measures such as concordance
statistics or calibration plots were not reported.
24
Given the
HERNIA score
9
may be less clinically applicable, we did not
validate it. Lack of data for suture length to incision length ratio
and time to suture removal or complete epithelialization also
precluded us from evaluating the Veljkovic model.
8
Our study had several limitations. First, not all predictor
variables were considered in the external validation for the
following reasons: Elixhauser comorbidity scores were not
available and race/ethnicity was not applicable as our data
were based solely in a Thai setting. Second, BMI was missing
in 27.4% of all subjects and therefore external validation was
undertaken in only those cases with complete data (12,155
records), which may have resulted in some degree of bias.
Third, some clinically insignificant IHs might not be detected
because imaging was not routinely used for hernia detection
in actual clinical practice at our settings. Finally, this exter-
nally validated updated model focused solely on abdominal
surgery as other surgery-specific models could not be evalu-
ated given the restrictions of the data collected.
Conclusion
Although the original Penn hernia risk calculator did not perform
well in the Ramathibodi IH surgical cohort, a revised model
achieved improved discrimination and calibration perfor-
mance. This revised model included age, BMI, chronic liver
disease, immunosuppressive medication, open surgery,
ostomy reversal, and transfusion, which helped identify those
patients at increased risk of IH and those most in need of tar-
geted intervention thus guiding and improving clinical care.
Ethical approval
This study was approved by the Ramathibodi Human Research
Ethics Committee (MURA2022/224) before data retrieval.
Author contributions
This study was conceptualized by A.Ta. and S.T. under super-
vision of P.N., P.S., and A.T. Data was linked to construct study
cohort by H.T., A.P., and P.P. Data cleaning was performed by
H.T., A.Ta., A.P., and N.P. A.Ta. performed data analysis under
supervision of P.N. and A.T. Manuscript was drafted by A.Ta.
and revised by G.MK., J.A., and A.T. All authors have read and
approved this manuscript before submission.
Funding
This study was funded by the National Research Council of
Thailand (NRCT #N42A640323). The sponsor had no role in the
design or conduct of the study.
Disclosure information
All authors declared no conflict of interest.
Study registration
Thai Clinical Trials Registry (TCTR20220704001), 4 July 2022,
retrospectively registered.
Consent to participate
Not applicable.
Consent for publication
Not Applicable.
Patient and public involvement
No patient and public involved.
Data statement
Access to the dataset from the current study is available from
the corresponding author upon reasonable request and
approval.
the surgeon 22 (2024) e34ee40 e39
Acknowledgement
This manuscript was a component of Amarit Tansawet's
training and dissertation in an international Ph.D. program
(Clinical Epidemiology), at the Department of Clinical Epide-
miology and Biostatistics, Faculty of Medicine, Ramathibodi
Hospital, Mahidol University, Bangkok, Thailand. The authors
gratefully acknowledge the assistance of Ms. Wipada Purawat.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.surge.2023.07.008.
references
1. Kingsnorth A. The management of incisional hernia. Ann R
Coll Surg Engl 2006;88(3):252e60.
2. Guti
errez de la Pe~
na C, Medina Achirica C, Domı´nguez-
Adame E, Medina Dı´ez J. Primary closure of laparotomies with
high risk of incisional hernia using prosthetic material:
analysis of usefulness. Hernia 2003;7(3):134e6.
3. Jairam AP, Timmermans L, Eker HH, Pierik R, van Klaveren D,
Steyerberg EW, et al. Prevention of incisional hernia with
prophylactic onlay and sublay mesh reinforcement versus
primary suture only in midline laparotomies (PRIMA): 2-year
follow-up of a multicentre, double-blind, randomised
controlled trial. Lancet 2017;390(10094):567e76.
4. Glauser PM, Brosi P, Speich B, K€
aser SA, Heigl A, Rosenberg R,
et al. Prophylactic intraperitoneal onlay mesh following
midline laparotomy elong-term results of a randomized
controlled trial. World J Surg 2019;43(7):1669e75.
5. Gillion JF, Sanders D, Miserez M, Muysoms F. The economic
burden of incisional ventral hernia repair: a multicentric cost
analysis. Hernia 2016;20(6):819e30.
6. Bosanquet DC, Ansell J, Abdelrahman T, Cornish J, Harries R,
Stimpson A, et al. Systematic review and meta-regression of
factors affecting midline incisional hernia rates: analysis of
14 618 patients. PLoS One 2015;10(9):e0138745.
7. Walming S, Angenete E, Block M, Bock D, Gessler B, Haglind E.
Retrospective review of risk factors for surgical wound
dehiscence and incisional hernia. BMC Surg 2017;17(1):19.
8. Veljkovic R, Protic M, Gluhovic A, Potic Z, Milosevic Z,
Stojadinovic A. Prospective clinical trial of factors predicting
the early development of incisional hernia after midline
laparotomy. J Am Coll Surg 2010;210(2):210e9.
9. Goodenough CJ, Ko TC, Kao LS, Nguyen MT, Holihan JL,
Alawadi Z, et al. Development and validation of a risk
stratification score for ventral incisional hernia after
abdominal surgery: hernia expectation rates in intra-
abdominal surgery (the HERNIA Project). J Am Coll Surg
2015;220(4):405e13.
10. Fischer JP, Basta MN, Mirzabeigi MN, Bauder AR, Fox JP,
Drebin JA, et al. A risk model and cost analysis of incisional
hernia after elective abdominal surgery based on 12,373
cases. The case for targeted prophylactic intervention. Ann
Surg 2016;263(5):1010e7.
11. Basta MN, Kozak GM, Broach RB, Messa CAt, Rhemtulla I,
DeMatteo RP, et al. Can we predict incisional hernia?:
development of a surgery-specific decision-support interface.
Ann Surg 2019;270(3):544e53.
12. Altman DG, Vergouwe Y, Royston P, Moons KGM. Prognosis
and prognostic research: validating a prognostic model. BMJ
2009;338:b605.
13. Moons KGM, Kengne AP, Grobbee DE, Royston P, Vergouwe Y,
Altman DG, et al. Risk prediction models: II. External
validation, model updating, and impact assessment. Heart
2012;98(9):691.
14. Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent
reporting of a multivariable prediction model for individual
prognosis or diagnosis (TRIPOD): the TRIPOD Statement. BMC
Med 2015;13(1):1.
15. Basta MN, Mirzabeigi MN, Shubinets V, Kelz RR, Williams NN,
Fischer JP. Predicting incisional hernia after bariatric surgery:
a risk stratification model based upon 2161 operations. Surg
Obes Relat Dis 2016;12(8):1466e73.
16. Webster C, Neumayer L, Smout R, Horn S, Daley J,
Henderson W, et al. Prognostic models of abdominal
wound dehiscence after laparotomy. J Surg Res
2003;109(2):130e7.
17. van Ramshorst GH, Nieuwenhuizen J, Hop WC, Arends P,
Boom J, Jeekel J, et al. Abdominal wound dehiscence in adults:
development and validation of a risk model. World J Surg
2010;34(1):20e7.
18. Cole J, Hughey S, Metzger A, Geiger P, Fluke L, Booth GJ.
Machine learning to predict fascial dehiscence after
exploratory laparotomy surgery. J Surg Res 2021;268:514e20.
19. Deerenberg EB, Harlaar JJ, Steyerberg EW, Lont HE, van
Doorn HC, Heisterkamp J, et al. Small bites versus large bites
for closure of abdominal midline incisions (STITCH): a
double-blind, multicentre, randomised controlled trial. Lancet
2015;386(10000):1254e60.
20. Jairam AP, L
opez-Cano M, Garcia-Alamino JM, Pereira JA,
Timmermans L, Jeekel J, et al. Prevention of incisional hernia
after midline laparotomy with prophylactic mesh
reinforcement: a meta-analysis and trial sequential analysis.
BJS Open 2020;4(3):357e68.
21. Depuydt M, Allaeys M, de Carvalho LA, Vanlander A,
Berrevoet F. Prophylactic mesh after midline laparotomy:
evidence is out there, but why do surgeons hesitate? World J
Surg 2021;45(5):1349e61.
22. Tansawet A, Numthavaj P, Techapongsatorn S,
Wilasrusmee C, Attia J, Thakkinstian A. Mesh position for
hernia prophylaxis after midline laparotomy: a systematic
review and network meta-analysis of randomized clinical
trials. Int J Surg 2020;83:144e51.
23. Tansawet A, Numthavaj P, Techapongsatorn S, McKay G,
Attia J, Pattanaprateep O, et al. Risk-benefit assessment of
onlay and retrorectus mesh augmentation for incisional
hernia prophylaxis: a secondary analysis from network meta-
analysis. Int J Surg 2021;92:106053.
24. Cherla DV, Moses ML, Mueck KM, Hannon C, Ko TC, Kao LS,
et al. External validation of the HERNIAscore: an
observational study. J Am Coll Surg 2017;225(3):428e34.
the surgeon 22 (2024) e34ee40e40