Surgery xxx (2019) 1e7

Contents lists available at ScienceDirect

Surgery
journal homepage: www.elsevier.com/locate/surg

Discharge disposition to skilled nursing facility after emergent general

surgery predicts a poor prognosis
Anghela Z. Paredes, MD, MSa,*, Azeem T. Malik, MBBSb, Marcus Cluse, BSa,
Scott A. Strassels, PhD, PharmDa, Heena P. Santry, MD, MSa, Daniel Eiferman, MDa,
Christian Jones, MD, MSc, Daniel Vazquez, MDa
a
Department of Surgery, Division of Critical Care, Trauma, and Burn, The Ohio State University Wexner Medical Center, Columbus
b
Department of Orthopaedics, The Ohio State University Wexner Medical Center, Columbus
c
Department of Surgery, Division of Acute Care Surgery, Johns Hopkins Medicine, Baltimore, MD

a r t i c l e i n f o a b s t r a c t

Article history: Background: Emergency general surgery can have a profound impact on the functional status of even
Accepted 8 April 2019 previously independent patients. The role and influence of discharging a patient to a skilled nursing
Available online xxx facility, however, remains largely unknown.
Methods: We queried the American College of Surgeons National Surgical Quality Improvement Program
for community-dwelling adults who underwent 1 of 7 emergency general surgery procedures and were
discharged home or to a skilled nursing facility from 2012 to 2016. Propensity score matching and
multivariable regression analyses were performed to determine the relationship between discharge
disposition and outcomes.
Results: Overall, 140,922 patients met the inclusion criteria. The majority were discharged home (95.9%).
After applying 1:1 propensity score matching, in comparison to patients discharged home, individuals
discharged to a skilled nursing facility had a greater odds of respiratory (odds ratio 2.32; 95% confidence
interval, 1.59e3.38) and septic complications (odds ratio 1.63, 95% confidence interval 1.12e2.36) after
discharge. Furthermore, following surgery, individuals discharged to a skilled nursing facility had a
greater odds of 30-day readmission (odds ratio 1.14; 95% confidence interval, 1.01e1.29), and death
within 30 days of the procedure (odds ratio 2.07; 95% confidence interval, 1.65e2.61).
Conclusion: After accounting for patient severity and perioperative course, discharge to a skilled nursing
facility is an independent risk factor for death, readmission, and postdischarge complications.
© 2019 Elsevier Inc. All rights reserved.

Introduction current smoking, have been associated with considerable

morbidity after EGS.6 Furthermore, postoperative complications,
More than 3 million people are hospitalized annually owing to a such as stroke and major bleeding, have been associated with death
condition requiring emergency general surgery (EGS), exceeding after EGS.7 Unlike other populations, patient optimization before
the yearly incidence of both new-onset diabetes and cancer.1e3 operative intervention is not feasible. Despite the known morbidity
These patients, plagued by unforeseen emergencies, represent a and mortality issues after EGS, the effect of discharge disposition on
complex population posing a substantial public health burden.3,4 As postdischarge outcomes remain poorly understood.
such, EGS has been associated with a high risk of death and post- Previous studies have estimated that 10% to 40% of EGS pa-
operative complications.5 Specifically, older age, male sex, and tients will require assistance once discharged, including either
comorbidities, such as chronic obstructive pulmonary disease and the use of home health care or a postacute care facility, such as a
skilled nursing facility (SNF) or intermediate care facility.8,9
Furthermore, recent studies have demonstrated that the quality
Presented at the Central Surgical Association Annual Meeting, Palm Harbor, FL, of a SNF has a marked impact on occurrence of complications and
March 2019. Selected best paper by a new member. readmission rates after pancreatectomy.10 The effect of discharge
* Reprint requests: Anghela Z. Paredes, MD, MS, Department of Surgery, Division
of Critical Care, Trauma, and Burn, The Ohio State University, Wexner Medical
to a SNF in the setting of EGS has not been explored fully. In this
Center, 395 W. 12th Ave, Suite 670, Columbus, OH 43210-1267. context, the objective of the current study was to characterize
E-mail address: Anghela.Paredes@osumc.edu (A.Z. Paredes). disposition after EGS using the database of the American College

https://doi.org/10.1016/j.surg.2019.04.034
0039-6060/© 2019 Elsevier Inc. All rights reserved.
2 A.Z. Paredes et al. / Surgery xxx (2019) 1e7

of Surgeons National Surgical Quality Improvement Program thrombosis or pulmonary embolism), renal failure (progressive
(ACS-NSQIP). Specifically, we sought to explore and compare renal insufficiency or acute renal failure), and cardiac abnormalities
the impact of discharge to a SNF versus home on 30-day read- (myocardial infarction or cardiac arrest) were characterized into a
missions, the incidence of postdischarge complications, and pre- or postdischarge complication based on duration of stay and
postdischarge 30-day mortality among a group of community- days from operation until date of complication. If the complication
dwelling individuals facing an unexpected emergency operation. occurred after the discharge date, the complication was considered
We hypothesize that a patient’s postdischarge course will vary a postdischarge complication.
based on discharge destination.
Statistical analysis
Methods
The cohort was subdivided based on discharge destination
Acquisition of data and study population (home versus SNF). Demographics, patient characteristics, and
clinical characteristics were compared between discharge desti-
Using the ACS-NSQIP database, we identified patients
18 years nations. To account for possible confounding between the 2 groups,
old who between 2012 to 2016 underwent 1 of the 7 previously a 1:1 propensity score-matching was performed based on age, sex,
defined EGS procedures (appendectomy, cholecystectomy, lapa- race, ASA class, body mass index, functional health status, comor-
rotomy, colectomy, small bowel resection, peptic ulcer repair, or bidities, wound class, type of operation (high-risk versus low-risk),
lysis of adhesions) that account for 80% of the operative volume, and predischarge complications. Continuous variables were
mortality, complications, and cost of EGS nationally (Supplemental described using measures of central tendency and the Wilcoxon
Table I).11,12 This database is a collection of operative data on over rank-sum test. Categorical variables were compared using pro-
150 variables, including patient demographics, intraoperative portions and the c2 or Fisher exact test as appropriate. Multivari-
variables, and 30-day postoperative mortality and morbidity for able logistic regression models were constructed adjusting for
patients undergoing operative intervention. Data were gathered by demographics, comorbidities, predischarge clinical characteristics,
surgical clinical reviewers who have undergone extensive training and type of operation (high-risk versus low-risk) to evaluate the
to ensure accurate data extraction.13,14 association between discharge disposition and postdischarge
To identify individuals who were community-dwellers before morbidity, 30-day readmission, and 30-day mortality. All analyses
presentation, patients who were transferred from an acute care were performed using SPSS version 24 (IBM, Armonk, NY) other
hospital, nursing home, chronic care facility, intermediate care fa- than propensity matching which was performed using SAS version
cility, outside emergency department, or unknown or other facility 9.4 (SAS Institute Inc., Cary, NC).
were excluded. Patients who underwent an elective operation, had
operative intervention more than 48 hours after presentation, or Results
who died during the index admission were also excluded. Dispo-
sition to either home or a SNF was necessary for inclusion in the Patient characteristics
final cohort.
A total of 140,922 patients who underwent EGS from 2012 to
Study variables and outcomes 2016 and met inclusion criteria were identified in the ACS-NSQIP
database. Median patient age was 43 years (IQR 29e59); approxi-
Clinical and demographic data reflect the patient’s status on mately half of the patients were female (n ¼ 72,536; 51.5%). Most
admission to the hospital. Abstracted data included age, sex, race or patients were white (n ¼ 100,420; 71.3%), and an overwhelming
ethnicity, classification according to the American Society of majority had an independent functional health status before EGS
Anesthesiology (ASA), and preoperative functional status (inde- (n ¼ 137,923; 97.9%). One in 3 patients was diagnosed with a septic
pendent, partially dependent, totally dependent, unknown). condition (n ¼ 46,885; 33.3%) preoperatively. Most patients
Functional status was determined by the surgical clinical reviewers underwent a low-risk operation (n ¼ 110,445; 78.4%). The most
based on the functional status 30 days before the operation. Inde- common in-hospital complications included sepsis (n ¼ 9,507,
pendent was defined as “does not require assistance from another 6.8%) and respiratory (n ¼ 5,127, 3.6%). Within 30 days of operation,
person for any activities of daily living, including one who functions 5.5% were readmitted and 0.4% (n ¼ 592) died after discharge.
independently with the use of prosthetics, equipment, and/or de- Among the 5712 patients (4.1%) discharged to SNF, most were
vices.”15 Partially dependent patients “require some assistance older (median age: 75, [IQR 66e82] vs 41 [IQR 28e57]) and more
from another person for activities of daily living,” whereas someone commonly totally dependent (n ¼ 102, 1.8% vs n ¼ 194, 0.1%) before
who is totally dependent “requires total assistance for all activities admission. Additional comparison of demographics and clinical
of daily living.”15 The comorbid conditions of interest included characteristics among patients who were discharged home or to a
>10% body weight loss in last 6 months, diabetes mellitus, dyspnea, SNF is shown in Table I. Patients who were discharged to a SNF were
current smoker within 1 year, preoperative ventilator use, history more likely to have been admitted with sepsis (59.4% vs 32.1%,
of severe chronic obstructive pulmonary disease, presence of as- P < .001). A greater proportion of patients discharged to an SNF
cites, congestive heart failure, hypertension, preoperative renal underwent high-risk operations in comparison to patients dis-
failure, dialysis dependence, history of disseminated cancer, pre- charged home (87.1% vs 18.9%, P < .001; Table II). In addition,
operative septic conditions (sepsis, septic shock, and septic in- patients discharged to a SNF were more than 6 times as likely to
flammations), preoperative wound infection, chronic steroid use, have experienced a predischarge complication (60.5% vs 9.3%, P <
and the history of a bleeding disorder. Similar to other studies, .001).
operative interventions were categorized as high-risk (laparotomy,
colectomy, small bowel resection, peptic ulcer repair, lysis of ad- Comparison of outcomes based on discharge destination
hesions) or low-risk (appendectomy, cholecystectomy).12,16 Com-
plications including wound, sepsis (sepsis or septic shock), Given the drastic differences in demographic and preoperative
respiratory (unplanned intubation, pneumonia, or postoperative characteristics between patients discharged to an SNF versus home
ventilator use >48 hours), thromboembolic disorders (deep venous and to minimize the potential for confounding, a 1:1 analysis
 A.Z. Paredes et al. / Surgery xxx (2019) 1e7 3

Table I
Baseline demographic and preoperative characteristics of community-dwelling patients undergoing
emergency general surgery based on discharge destination after operation

Home (n ¼ 135,210) SNF (n ¼ 5,712) P value

n % n %

Age (y) <.001

18e40 65,276 (48.3%) 146 (2.6%)
41e65 51,030 (37.7%) 1,103 (19.3%)
66e80 14,968 (11.1%) 2,240 (39.2%)
>80 3,936 (2.9%) 2,223 (38.9%)
Sex <.001
Male 66,187 (49%) 2,199 (38.5%)
Female 69,023 (51%) 3,513 (61.5%)
Race <.001
White 95,855 (70.9%) 4,565 (79.9%)
Black 11,560 (8.4%) 744 (13.0%)
Other 8,150 (6.0%) 165 (2.9%)
Unknown 19,645 (14.5%) 238 (4.2%)
ASA grade <.001
IeII 103,866 (76.8%) 572 (10%)
>II 31,444 (23.2%) 5,140 (90%)
Functional status <.001
Independent 133,023 (98.4%) 4,900 (85.8%)
Partially dependent 883 (0.7%) 641 (11.2%)
Totally dependent 194 (0.1%) 102 (1.8%)
Unknown 1,110 (0.8%) 69 (1.2%)
BMI (kg/m2) <.001
<25.0 45,804 (33.9%) 2,381 (41.7%)
25.0e29.9 45,251 (33.5%) 1,596 (27.9%)
30.0e34.9 25,238 (18.7%) 906 (15.9%)

35.0 18,917 (14%) 829 (14.5%)
>10% body weight loss in last 6 m 965 (0.7%) 238 (4.2%) <.001
Diabetes mellitus <.001
Insulin-dependent 3,391 (2.5%) 483 (8.5%)
Not insulin-dependent 5,695 (4.2%) 676 (11.8%)
Dyspnea <.001
At rest 233 (0.2%) 120 (2.1%)
With moderate exertion 2,137 (1.6%) 429 (7.5%)
Current smoker 25,582 (18.9%) 923 (16.2%) <.001
Preoperative ventilator use 1,039 (0.8%) 382 (6.7%) <.001
COPD 2,541 (1.9%) 781 (13.7%) <.001
Ascites 713 (0.5%) 206 (3.6%) <.001
CHF 372 (0.3%) 230 (4.0%) <.001
Hypertension 29,647 (21.9%) 3,943 (69.0%) <.001
ARF 179 (0.1%) 176 (3.1%) <.001
Dialysis-dependent 577 (0.4%) 157 (2.2%) <.001
Malignancy 1,407 (1.0%) 382 (6.7%) <.001
Preoperative septic condition 43,492 (32.1%) 3,392 (59.4%) <.001
Preoperative wound infection 532 (0.4%) 213 (3.7%) <.001
Corticosteroid use 2,986 (2.2%) 594 (10.4%) <.001
Bleeding disorder 3,690 (2.7%) 1,034 (18.1%) <.001

ARF, acute renal failure; BMI, body mass index; CHF, congestive heart failure; COPD, chronic
obstructive pulmonary disease; SIRS, systemic inflammatory response syndrome.

involving propensity score matching was completed. Patients were 1.65e2.61). Rate of death after the 30 days discharge time point
matched based on age, sex, race, ASA class, body mass index, could not be determined because of the restriction of the ACS-
functional health status, comorbidities, wound class, type of oper- NSQIP database only recording 30-day post discharge data.
ation, and specific in-hospital complications. After matching, 4,872
(50%) patients were discharged home, and 4,872 (50%) individuals Discussion
were discharged to a SNF. Between these new cohorts, only ASA
grade and preoperative ventilator use remained different (P < .05, Annually, millions of people in the United States face unforeseen
Supplemental Table II). A multivariable regression analysis was EGS.3 For some, the operative intervention can catalyze a tragic
performed adjusting for baseline clinical and demographic char- course of events. The postdischarge period including occurrence of
acteristics, type of operation, operative time, and postoperative complications, readmission, and mortality for EGS patients dis-
morbidity (Table III). Individuals discharged to a SNF had 63% charged to a SNF remains unknown. In the present study among a
greater odds having a septic complication (OR 1.63; 95% CI, group of community-dwelling individuals who survived the index
1.12e2.36) and more than 2 times the odds of developing a respi- hospitalization, 4.1% were discharged to a SNF after EGS. Not sur-
ratory complication after discharge (OR 2.32; 95% CI, 1.59e3.38). prisingly, patients discharged to a SNF were more likely to have
Discharge to a SNF was predictive of readmission within 30 days multiple comorbidities, undergo high-risk surgery and have at least
(OR 1.14, 1.01e1.29). Lastly, patients discharged to a SNF had one perioperative complication. To enable the analysis of the as-
considerably greater odds of dying within 30 days of operation sociation of discharge to a SNF and postdischarge outcomes of in-
compared with patients discharged home (OR 2.07; 95% CI, terest while controlling for confounding as fully as possible, we
4 A.Z. Paredes et al. / Surgery xxx (2019) 1e7

Table II
Perioperative and postoperative characteristics of community-dwelling patients EGS based on
discharge destination after an operation

Home (n ¼ 135,210) SNF (n ¼ 5,712) P value

n % n %

Type of operation <.001

Low risk 109,707 (81.1%) 738 (12.9%)
High risk 25,503 (18.9%) 4,974 (87.1%)
Wound classification <.001
Clean 5,616 (4.2%) 310 (5.4%)
Clean/contaminated 38,266 (28.3%) 1,417 (24.8%)
Contaminated 59,607 (44.1%) 1,080 (18.9%)
Dirty 31,721 (23.5%) 2,905 (5.9%)
Total operative time (min) <.001
0e60 85,913 (63.5%) 1,351 (23.7%)
61e120 38,120 (28.2%) 2,530 (44.3%)
>120 11,177 (8.3%) 1,831 (32.1%)
Duration of stay (d) <.001
0e2 99,347 (73.5%) 230 (4%)

3 35,863 (25.5%) 5,482 (96.0%)
Any predischarge complication 12,599 (9.3%) 3,455 (6.5%) <.001
Readmission 6,998 (5.2%) 784 (13.7%) <.001
Postdischarge mortality 251 (0.2%) 341 (6.0%) <.001

Table III present study additionally highlights the substantial mortality that
Results of multivariable logistic regression analysis for evalu- is independently associated with discharge to a SNF. Collectively,
ating the association of SNF discharge versus home for out-
comes of interest after EGS
these data have important implications in the discharge planning
after EGS and the close monitoring that may be needed especially
Postdischarge outcomes Adjusted OR (95% CI) for patients not discharged home.
Wound complications 0.87 (0.71e1.06) The effect of discharge to a SNF on outcomes is debated, but as
Respiratory complications 2.32 (1.59e3.38) may be suspected, the outcomes depend on the underlying popu-
Thromboembolic complications 1.32 (0.89e1.98)
lation and the postacute care needs of the patient. Patient optimi-
Renal complications 1.17 (0.66e2.05)
Cardiac complications 1.42 (0.8e2.51) zation, including discharge timing and increasing patient complexity,
Septic complications 1.63 (1.12e2.36) has been attributed by SNF providers as a major challenge in meeting
Readmissions 1.14 (1.01e1.29) the needs of their residents.17 Furthermore, optimization of resources
Mortality 2.07 (1.65e2.61)
available at SNFs may also be a contributing factor in the results
noted in the present study. For example, among patients undergoing
a pancreatectomy, patients discharged to a SNF rated as below-
included a propensity adjustment for key demographic and clinical average quality as determined by The Centers for Medicare and
characteristics. Holding clinical and perioperative characteristics Medicaid Services ratings, were more likely to be readmitted
equal, the present study found that the odds of death, readmission, compared to patients discharged to an above-average SNF.10 In a
and complications after discharge are increased in patients dis- different study, Ogunneye et al did not find an association between
charged to a SNF compared wtih those discharged home. To our the quality of the SNF and patient outcomes among patients with
knowledge, this is the first study to consider the differential ex- decompensated heart failure.18 Given that 2019 will mark the start of
periences and outcomes of EGS patients by discharge destination. the SNF value-based purchasing program that will penalize or
These findings expand important prior work characterizing the reward SNFs based on 30-day all cause readmission, however, our
field of EGS. Havens et al reported that EGS itself is an independent data support the need for improvements in patient selection and the
risk factor for death and postoperative complications.5 Specifically transition of care from acute hospitalization to a SNF. To this end,
among EGS patients, they found that one-third experienced a Schoenfeld et al found that patients who were discharged to a SNF
complication, and 1 in 8 died.5 In our current study, the morbidity that cared for a larger proportion of surgical patients had a decreased
rate was 11%. One possible reason for the disparate results may be likelihood of rehospitalization.19 Perhaps future directions by The
secondary to differences in the underlying population. The present Centers for Medicare and Medicaid Services should also focus on
study sought to measure differences between community dwellers designing and designating SNF centers with expertise in various
presenting from home who underwent 1 of 7 procedures shown to conditions in order to optimize the quality of care delivered to pa-
account for the most admissions, deaths, complications, and inpa- tients too healthy for a long-term acute care hospital but too complex
tient costs. In contrast, Havens et al included patients from both to return to home directly.
EGS and emergency vascular surgery populations and patients Our findings should be considered in the context of several
transferred from outside facilities whose community-dwelling limitations. First, similar to other retrospective cohort studies, the
status was unknown.5,11 Although the present study noted that present study is subject to information bias. To limit information
patients discharged to an SNF were older and more likely to have bias, the ACS-NSQIP database was selected owing to the rigorous
endured a complicated hospital course, we found that after utiliz- and validated methods of data abstraction.13,14 In addition,
ing 1:1 propensity matching and maintaining clinical and operative although ACS-NSQIP contains data that have been found to be more
characteristics similar between both cohorts, patients discharged to accurate than data obtained from administrative claims data
a SNF still had >2 times the odds of dying within the first month sources, data is limited to 30 days after operation, which may be
postoperatively once discharged in comparison to patients dis- insufficient to capture all complications especially of patients dis-
charged home. In addition, they were more likely to experience charged to SNF.20 Also, the ACS-NSQIP database does not contain
septic and respiratory complications. Unlike other studies, the information on patient’s insurance, socioeconomic status,
 A.Z. Paredes et al. / Surgery xxx (2019) 1e7 5

functional status at time of discharge, or the quality of SNF. Another References

important limitation to this study is the effect of confounding. We
attempted to address this potential limitation by using a 1:1 pro- 1. National Center for Chronic Disease Prevention and Health Promotion. National
pensity score matching, including variables predictive of the out- Diabetes Statistics Report: Estimates of Diabetes and Its Burden in the United
States. Atlanta, GA: National Center for Chronic Disease Prevention and Health
comes of interest, which has been shown to limit the amount of Promotion; 2017.
residual imbalance while maximizing the power to detect a dif- 2. Noone A, Howlader N, Krapcho M, et al. SEER Cancer Statistics Review,
ference between the groups.21 1975e2015. Bethesda, MD: National Cancer Institute; 2017.
3. Gale SC, Shafi S, Dombrovskiy VY, Arumugam D, Crystal JS. The public health
Despite these limitations, this study has important findings of burden of emergency general surgery in the United States: A 10-year analysis
interest to patients, families, providers, and policymakers. of the Nationwide Inpatient Sampled2001 to 2010. J Trauma Acute Care Surg.
Although the cause of a complex, postdischarge course requiring 2014;77:202e208.
4. Ogola GO, Gale SC, Haider A, Shafi S. The financial burden of emergency general
SNF discharge is multifactorial, the challenges may stem from a lack
surgery: National estimates 2010 to 2060. J Trauma Acute Care Surg. 2015;79:
of specialized resources needed to care for these critical patients 444e448.
that have survived an unexpected insult. In addition, future studies 5. Havens JM, Peetz AB, Do WS, et al. The excess morbidity and mortality of
emergency general surgery. J Trauma Acute Care Surg. 2015;78:306e311.
using institutional data are needed to capture the social factors and
6. Akinbami F, Askari R, Steinberg J, Panizales M, Rogers Jr SO. Factors affecting
motivators that contribute to the decision to be discharged to a SNF. morbidity in emergency general surgery. Am J Surg. 2011;201:456e462.
Future initiatives that seek to improve the transition from hospital 7. McCoy CC, Englum BR, Keenan JE, Vaslef SN, Shapiro ML, Scarborough JE.
to SNF while incentivizing quality of care are necessary to give the Impact of specific postoperative complications on the outcomes of emergency
general surgery patients. J Trauma Acute Care Surg. 2015;78:912e918; discus-
opportunity for previous community-dwelling individuals to re- sion 918e919.
turn home. In the absence of such approaches, a worsening of 8. Joseph B, Zangbar B, Pandit V, et al. Emergency general surgery in the elderly:
present findings could be possible in the future. Too old or too frail? J Am Coll Surg. 2016;222:805e813.
9. Havens JM, Olufajo OA, Cooper ZR, Haider AH, Shah AA, Salim A. Defining rates
In conclusion, although only a minority of community dwelling and risk factors for readmissions following emergency general surgery. JAMA
patients were discharged to a SNF after EGS, this discharge is in itself Surg. 2016;151:330e336.
an independent risk factor for death and postdischarge complica- 10. Paredes AZ, Hyer JM, Beal EW, et al. Impact of skilled nursing facility quality on
postoperative outcomes after pancreatic surgery. Surgery. [Epub ahead of print]
tions. Future studies and care remodeling techniques should focus on 11. Scott JW, Olufajo OA, Brat GA, et al. Use of national burden to define operative
attenuating this disparity in outcomes. Such techniques might focus emergency general surgery. JAMA Surg. 2016;151:e160480.
on discussing the increased morbidity and mortality with the patient 12. Feeney T, Castillo-Angeles M, Scott JW, et al. The independent effect of emer-
gency general surgery on outcomes varies depending on case type: A NSQIP
early in the planning phase, restructuring discharge planning, and
outcomes study. Am J Surg. 2018;216:856e862.
increasing provider follow-up after discharge. 13. Cohen ME, Ko CY, Bilimoria KY, et al. Optimizing ACS NSQIP modeling for
evaluation of surgical quality and risk: Patient risk adjustment, procedure mix
adjustment, shrinkage adjustment, and surgical focus. J Am Coll Surg. 2013;217:
Funding/Support 336e346, e331.
14. Raval MV, Pawlik TM. Practical guide to surgical data sets: National Surgical
Quality Improvement Program (NSQIP) and Pediatric NSQIP. JAMA Surg.
The authors have indicated that they have no conflicts of
2018;153:764e765.
interest (or funding) regarding the content of this article. 15. Skube SJ, Lindemann EA, Arsoniadis EG, Akre M, Wick EC, Melton GB. Char-
acterizing functional health status of surgical patients in clinical notes. AMIA Jt
Summits Transl Sci Proc. 2018;2017:379e388.
Conflict of interest/Disclosures 16. Castillo-Angeles M, Uribe-Leitz T, Jarman M, et al. Transferred emergency
general surgery patients are at increased risk of death: A NSQIP propensity
score matched analysis. J Am Coll Surg. 2019;228:871e877.
The ACS NSQIP and hospitals participating in the ACS NSQIP are
17. Britton MC, Ouellet GM, Minges KE, Gawel M, Hodshon B, Chaudhry SI. Care
the source of the data used. They have not verified and are not transitions between hospitals and skilled nursing facilities: Perspectives of
responsible for the statistical validity of the data analysis or the sending and receiving providers. Jt Comm J Qual Patient Saf. 2017;43:565e572.
18. Ogunneye O, Rothberg MB, Friderici J, Slawsky MT, Gadiraju VT, Stefan MS. The
conclusions derived by the authors. We acknowledge H.P.S has
association between skilled nursing facility care quality and 30-day read-
received grant support from the Agency for Healthcare Research mission rates after hospitalization for heart failure. Am J Med Qual. 2015;30:
and Quality (R01HS022694). She is a paid consultant by the John- 205e213.
son & Johnson Company on a fragility fracture advisory board. The 19. Schoenfeld AJ, Zhang X, Grabowski DC, Mor V, Weissman JS, Rahman M.
Hospital-skilled nursing facility referral linkage reduces readmission rates
submitted work is not related to this topic. No other disclosures. among Medicare patients receiving major surgery. Surgery. 2016;159:
1461e1468.
20. Steinberg SM, Popa MR, Michalek JA, Bethel MJ, Ellison EC. Comparison of risk
Supplementary materials adjustment methodologies in surgical quality improvement. Surgery.
2008;144:662e667.
Supplementary material associated with this article can be 21. Austin PC, Grootendorst P, Anderson GM. A comparison of the ability of
different propensity score models to balance measured variables between
found, in the online version, at https://doi.org/10.1016/j.surg.2019. treated and untreated subjects: A Monte Carlo study. Stat Med. 2007;26:
04.034. 734e753.

Discussion

Dr Juan Asensio (Omaha, NE): This is an interesting retro- specifically to explore the impact of discharge to a Skilled
spective large database 5-year study utilizing the National Nursing Facilities (SNF) on 30-day readmissions, as well as to
Quality Improvement Program (NSQIP) of the American College examine the incidence of post discharge complications and post
of Surgeons (ACS). The authors have clearly defined their discharge 30-day mortality. Their hypothesis being that the pa-
objective, which is to characterize disposition of patients tients’ post discharge course and outcomes will vary based on
following emergency general surgery (EGS) utilizing NSQIP data, discharge destination.
6 A.Z. Paredes et al. / Surgery xxx (2019) 1e7

The authors also focused on comorbidities present in this pa- being in the home group versus the other equivalent person with
tient population, as well as postoperative complications stratified the equivalent probability to be assigned to the SNF discharge
to 2 different patient groups. Those that underwent what was group.
deemed high risk surgical procedures such as: laparotomy, colec- We matched for the multiple differences that were seen, but
tomy, small bowel resection, peptic ulcer repair and lysis of adhe- additionally what we did is we performed multi-variable regres-
sions versus those that underwent what was defined as low risk sion analysis in order to be able to determine the relationship
procedures such as appendectomy and cholecystectomy. taking into account all of the matched variables to determine
Their results are clearly presented with great detail. Sur- whether the association was still present.
prising, was the median age of 43, not surprising was that Despite matching, there was still sufficient power to be able to
the majority of patients underwent low risk procedures (78.4%) detect the differences between the groups in regards to ASA class
and most were discharged homed95.9%. Of the 5,127 (4.1%) and in regards to the preoperative ventilator use, but in order to
patients transferred to a skilled nursing facility (SNF) their account for those differences, multivariable regression analysis was
readmission, rate postoperative complications and mortality performed.
were higher. Despite having a similar cohort, patients that were discharged
A review of their outcomes reveals that for all parameters to a SNF had higher odds of these complications, had higher odds
tested there was statistical significance for both univariate and of readmission, and had higher odds of mortality. What's impor-
stepwise logistic regression, so that the data does support the tant for us to take home is that when we decide that a patient
conclusions. Thus, the authors have proven what is intuitively should be discharged to a SNF, this represents a population that is
predicted and in concordance with other data available in the at high risk. There's something that happens during this post-
literature. operative period, this post discharge period, that we need to
However, there are a number of questions for the authors: watch over. Maybe this population is one that we need to have a
1. Can you describe the entire list of high risk versus low risk greater coordination of care so that we improve the quality of care
procedures? Missing from the list, for instance, are soft tissue that these individuals are receiving.
infections requiring debridement and drainage and possibly Like I mentioned, we don't have all of the information in regards
others. to what drives the decision-making process. This is a complex de-
2. The authors chose propensity score analysis, which since its cision, and although we are looking at the association of discharge
description in 1983 appears to be used more and more to a SNF and the outcomes of interest, it highlights the importance
frequently, but in my opinion, not well understood in terms of it of future studies to examine what is happening and what resources
applicability. What are the drastic differences you allude to are needed for practitioners to equip individuals caring for these
between patients discharged home versus those discharged to a patients in order to optimize their care.
SNF? Dr Steven De Jong (Maywood, IL): Congratulations on a very
3. What are the confounding variables that you needed to control well analyzed study, database, and a very well-presented study.
for, by choosing propensity score matching. In my review of the I may have misinterpreted one of your slides, but it seemed that
data provided in supplemental Table I, only 2 of the predictors of the group discharged home had a longer operative time compared
outcome achieved statistical significancedthe dichotomous to patients discharged to a SNF was significantly longer. Could this
ASA risks I to II versus greater than ASA II and preoperative be another significant conclusion from your study? Secondly, did
ventilator usage. Can you explain this? the time interval the patients stayed in the SNF or the long-term
4. Beyond what is obvious, can you describe the real and tangible care facility affect the expected outcome?
applications to the conclusions of your study. Dr Anghela Paredes: Thank you, Dr DeJong. In regards to the
In summary, the data supports conclusions of this study, the operative time, we did not match for operative time, more so
statistics are sound, and the study is solid. All that remains for the because we matched for the complexity of the procedure. So we
authors is to explain from the results of this study how can this be would hope that the complexity of the procedure would capture
applied to the population at large? the operative time, and that the high-risk operations were more
I would like to congratulate Dr Paredes for her excellent pre- likely to be longer operations that would be equally distributed
sentation, and the colleagues from Ohio State University. between the 2 groups.
I thank the Central Surgical Association for the privilege of being In regards to how long these individuals were at a SNF, unfor-
able to discuss this paper. tunately, the information regarding SNF length of stay or even SNF
Dr Anghela Paredes: Thank you, Dr Asensio. The benefit of quality is available on a Medicare database which is not able to be
propensity analysis is that it allows researchers to control for merged with the ACS NSQIP database. So in regards to how long
confounding on several variables. What ultimately we want to do these individuals were at a SNF and whether that affected these
with propensity matching, or multivariable regression analysis is outcomes, it's not able to be performed with the current data set,
limit the effect that confounders can have on the outcomes. What but further follow-up would be important and needed to be done in
we specifically want to study is whether SNF as a variable affects order to examine this relationship.
the odds of having a post-discharge complication, a readmission, or Dr Robert C.G. Martin (Louisville, KY): One question about your
a mortality. matching. Did you match these patients who had their surgical
In order to account for the differences that we saw between procedures already inpatient, or were these all elective surgeries
the patients that were discharged home and discharged to a SNF, that then had this problem?
as we have found that they were older, they were more likely to The second part is that the immediacy of readmission is really
have complications, and have multiple comorbidities we kind of bi-modal effect. There's usually within the first 4, 5 days
matched for those variables, thus creating as similar as possible of discharge, which ultimately really is not a SNF issue, it's
group of individuals that were in this data set. Additionally, what probably a discharge management issuedthey should not have
propensity score matching does is it assigns the probability of a been dischargeddand then it is after that 7 to 10 days, then out to
person to be assigned to 1 of the 2 groups, and then matches that 30 days, which is obviously what NSQIP gives you. Did you look at
probability to another person, and so one gets randomized to any of those types of variables for readmissions?
 A.Z. Paredes et al. / Surgery xxx (2019) 1e7 7

Dr Anghela Paredes: In regards to looking at the variables for Studies have shown up to 33% of readmissions are preventable. So
readmission in regards to the timing of readmission, we did not further studies would be needed to examine exactly the date of
specifically look at whether these were readmissions that were close readmission and whether that at all had an influence.
to the discharge date, and further studies are needed to see whether Sorry if I didn't make myself clear, for patients to be included in
patients are being readmitted from these facilities because maybe the cohort, they had to have undergone an emergent procedure,
they weren't ready to be discharged, and maybe an extra day or 2 meaning that the operation had to be done within 48 hours of
days within the hospital would have prevented that readmission. admission.