155

C OPYRIGHT
2018 BY T HE J OURNAL OF B ONE AND J OINT S URGERY, I NCORPORATED

Resident Participation in Fixation of

Intertrochanteric Hip Fractures
Analysis of the NSQIP Database
Alexander L. Neuwirth, MD, Russell N. Stitzlein, MD, Madalyn G. Neuwirth, MD, Rachel K. Kelz, MD, MPH, and
Samir Mehta, MD

Investigation performed at the Hospital of the University of Pennsylvania, Philadelphia, Pennsylvania

Background: Future generations of orthopaedic surgeons must continue to be trained in the surgical management of hip
fractures. This study assesses the effect of resident participation on outcomes for the treatment of intertrochanteric hip
fractures.
Methods: The National Surgical Quality Improvement Program (NSQIP) database (2010 to 2013) was queried for in-
tertrochanteric hip fractures (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] code
820.21) treated with either extramedullary (Current Procedural Terminology [CPT] code 27244) or intramedullary (CPT
code 27245) fixation. Demographic variables, including resident participation, as well as primary (death and serious
morbidity) and secondary outcome variables were extracted for analysis. Univariate, propensity score-matched, and
multivariate logistic regression analyses were performed to evaluate outcome variables.
Results: Data on resident participation were available for 1,764 cases (21.0%). Univariate analyses for all intertro-
chanteric hip fractures demonstrated no significant difference in 30-day mortality (6.3% versus 7.8%; p = 0.264) or
serious morbidity (44.9% versus 43.2%; p = 0.506) between the groups with and without resident participation. Multi-
variate and propensity score-matched analyses gave similar results. Resident involvement was associated with prolonged
operating-room time, length of stay, and time to discharge when a prolonged case was defined as one above the 90th
percentile for time parameters.
Conclusions: Resident participation was not associated with an increase in morbidity or mortality but was associated
with an increase in time-related secondary outcome measures. While attending surgeon supervision is necessary, resi-
dents can and should be involved in the care of these patients without concern that resident involvement negatively
impacts perioperative morbidity and mortality.
Level of Evidence: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

A
lthough efforts to reduce the rate of hip fractures in the screw continues to be the gold standard, the use of cepha-
United States have been successful, the prevalence of these lomedullary nails has risen over the past decade13,14.
injuries has increased secondary to an aging population1,2. In the United States, orthopaedic surgery residency pro-
Nearly half of hip fractures occur in the intertrochanteric region3. grams employ a mentorship model to transition medical gradu-
Treatment for this fracture subtype includes fracture reduction ates into independent surgeons. The impact of resident
and stabilization with either an extramedullary implant (e.g., a involvement on intraoperative and postoperative complications
sliding hip screw) or an intramedullary implant (e.g., a cepha- has been under close investigation recently. Schoenfeld et al.
lomedullary nail). Intramedullary implants are generally deemed showed a mild to moderate increase in the risk of complications
superior for fixation of unstable fracture patterns4; however, de- following hip and knee arthroplasty when trainees participated in
bate exists as to which method is superior for the treatment of the surgical care of patients15. Cvetanovich et al. recently reviewed
stable intertrochanteric fracture patterns5-12. While the sliding hip the cases of 1,382 patients following total shoulder arthroplasty

Disclosure: There was no external source of funding. On the Disclosure of Potential Conflicts of Interest forms, which are provided with the online version
of the article, one or more of the authors checked “yes” to indicate that the author had a relevant financial relationship in the biomedical arena outside the
submitted work (http://links.lww.com/JBJS/E537).

J Bone Joint Surg Am. 2018;100:155-64 d http://dx.doi.org/10.2106/JBJS.16.01611

 156
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

and showed that resident involvement was not a risk factor for 30- ‡100 billion cells/L; international normalized ratio (INR) of
day postoperative complications16. Additional studies utilizing the £1.5 and >1.5; and bilirubin of <2 and ‡2 mg/dL.
American College of Surgeons National Surgical Quality Im- The primary outcomes for this study were death or se-
provement Program (NSQIP) database found that resident in- rious morbidity within 30 days of the index procedure. Addi-
volvement was not associated with increased complications tional secondary outcomes including the incidence of specific
following total knee arthroplasty (24,529 cases)17 or with in- complications, total operative time (time from incision to
creased complications following total hip arthroplasty (13,109 closure), total length of hospital stay, time from admission to
cases)18. Furthermore, Bohl et al. assessed the theoretical “July the operating room, and time from the operation to hospital
effect” and found no association between the time of year when discharge were assessed. Perioperative time parameters above
residents graduate to new levels of responsibility and the risk of the 90th percentile were designated as prolonged.
complications for patients having hip and knee arthroplasty19. Primary outcomes were analyzed for the entire cohort as
NSQIP is a multi-institution national database that collects well as for each subgroup (intramedullary and extramedullary
over 150 variables, including preoperative, intraoperative, and fixation). Serious morbidity was defined as a complication car-
30-day postoperative data, to identify risk factors and perioper- rying an Accordion severity grading of ‡3 (any complication
ative complications. A trained clinical nurse reviewer at each site requiring reoperation or interventional radiographic procedures,
collects and inputs data for all major surgical procedures20,21. or resulting in failure of ‡1 organ systems)22. Specific complica-
To our knowledge, there are currently no studies evalu- tions were further grouped into common organ systems.
ating the impact of trainee involvement on complication rates
in the treatment of intertrochanteric hip fractures. Our pri- Statistical Methods
mary objective was to identify whether resident participation is Univariate analysis of resident participation was performed
associated with 30-day postoperative mortality and/or serious using the Pearson chi-square test for all categorical variables in
morbidity using the NSQIP database. Our null hypothesis the full cohort and for each of the surgical (intramedullary or
proposed that resident involvement was not associated with an extramedullary fixation) subgroups to identify associations
increased rate of death or serious morbidity following surgical with mortality, morbidity, and death or serious morbidity as
fixation of intertrochanteric hip fractures with either extra- well as intraoperative parameters and postoperative compli-
medullary or intramedullary implants. cations. Association with primary and secondary outcomes was
deemed significant for p values of £0.05.
Materials and Methods Subsequently, independent logistic forward stepwise re-

T he NSQIP Participant Use Data File (PUF) for 2010 to 2013

was accessed, and patients were identified using the Inter-
national Classification of Diseases, 9th Revision, Clinical Mod-
gression models were developed using variables identified as
significantly associated with each primary and secondary out-
come on univariate analysis to determine the association be-
ification (ICD-9-CM) coding for intertrochanteric hip fracture tween resident involvement and the primary and secondary
(820.21). Subsequently, Current Procedural Terminology (CPT) outcomes, with adjustment for potential confounders.
codes were utilized to identify patients treated with extramed- Propensity score matching is utilized to minimize selec-
ullary (27244) or intramedullary (27245) implants. Only cases tion bias of a treatment group by estimating treatment effects in
that contained complete data for surgical trainee participation observational, nonrandomized studies23. The “case” (resident
were included in the final analysis. The 2013 NSQIP PUF con- involvement) and “control” (attending surgeon alone) groups
tained data from 435 participating institutions21. were matched on the basis of logistic regression modeling that
Resident involvement during surgical treatment of in- identified variables significantly associated with death or serious
tertrochanteric hip fractures was examined as a binary variable: morbidity in the overall cohort24. Cases were matched 1:1 using
resident involved or attending surgeon alone. Additional cat- the nearest neighbor matching algorithm without replace-
egorical variables included advanced age (‡70 years); sex; race; ment25, with a caliper size of 0.1 · log(propensity score standard
functional status; body mass index (BMI) classified as under- deviation)26. Unmatched control subjects were discarded.
weight (<18.5 kg/m2), normal (18.5 to 24.9 kg/m2), overweight Within the matched cohort, the treatment effect (resident in-
(25.0 to 29.9 kg/m2), and obese (‡30 kg/m2); American Society volvement) was measured by directly comparing the case and
of Anesthesiologists (ASA) classification; diabetes mellitus re- control groups.
quiring insulin; history of hypertension requiring medication; Data were transferred into STATA format using the Stat/
history of cerebrovascular accident (CVA) or stroke; smoking; Transfer statistical program (version 11.0; Circle Systems), and
chronic obstructive pulmonary disease (COPD); alcohol abuse; analysis was performed using Stata 12.0/IC statistical software
ascites; steroid use; renal failure requiring dialysis; dissemi- (StataCorp). This study was reviewed and was deemed exempt
nated malignancy; chemotherapy administration within 30 by our institutional review board.
days of surgery; and radiation therapy administration within 90
days of surgery. Preoperative laboratory values were analyzed as Results
binary variables as follows: creatinine level of <2 and ‡2 mg/dL; Patient Demographics
albumin of <3 and ‡3 g/dL; white blood-cell (WBC) count of
<4,000 and ‡12,000 cells/mL; platelet count of <100 billion and E xtraction from the NSQIP database identified 8,384
patients who underwent intertrochanteric hip fracture
 157
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

TABLE I Comparative Demographics for All Patients Undergoing Surgical Fixation of Intertrochanteric Hip Fracture Grouped by Resident
Involvement*

Characteristic* Overall (no. [%]) Resident Involved (no. [%]) Attending Surgeon Alone (no. [%]) P Value

Total 1,764 (100.0) 552 (31.3) 1,212 (68.7)

Sex† 0.036
Male 501 (28.4) 178 (32.2) 323 (26.7)
Female 1,261 (71.5) 374 (67.8) 887 (73.2)
Age 0.005
<70 yr 258 (14.6) 100 (18.1) 158 (13.0)
‡70 yr 1,506 (85.4) 452 (81.9) 1,054 (87.0)
Race <0.001
Caucasian 1,304 (73.9) 352 (63.8) 952 (78.5)
African-American 38 (2.2) 23 (4.2) 15 (1.2)
Asian 17 (1.0) 10 (1.8) 7 (0.6)
Unknown 405 (23.0) 167 (30.3) 238 (19.6)
BMI 0.020
Underweight 489 (27.7) 175 (31.7) 314 (25.9)
Normal 712 (40.4) 220 (39.9) 492 (40.6)
Overweight 382 (21.7) 99 (17.9) 283 (23.3)
Obese 181 (10.3) 58 (10.5) 123 (10.1)
Functional status <0.001
Independent 1,122 (63.6) 330 (59.8) 792 (65.3)
Dependent 623 (35.3) 221 (40.0) 402 (33.2)
Unknown 19 (1.1) 1 (0.2) 18 (1.5)
ASA class† 0.160
I 17 (1.0) 6 (1.1) 11 (0.9)
II 330 (18.7) 92 (16.7) 238 (19.6)
III 1,108 (62.8) 341 (61.8) 767 (63.3)
IV 306 (17.3) 113 (20.5) 193 (15.9)
V 1 (0.1) 0 (0) 1 (0.1)
Comorbidities
Diabetes 314 (17.8) 105 (19.0) 209 (17.2) 0.365
Cardiac disease 298 (16.9) 96 (17.4) 202 (16.7) 0.706
Hypertension 1,231 (69.8) 371 (67.2) 860 (71.0) 0.112
CVA or stroke 100 (5.7) 43 (7.8) 57 (4.7) <0.001
Smoking 237 (13.4) 87 (15.8) 150 (12.4) 0.053
COPD 232 (13.2) 73 (13.2) 159 (13.1) 0.951
Alcohol abuse 68 (3.9) 21 (3.8) 47 (3.9) 0.998
Chronic steroid use 86 (4.9) 29 (5.3) 57 (4.7) 0.619
ESRD on dialysis 44 (2.5) 21 (3.8) 23 (1.9) 0.017
Disseminated malignancy 23 (1.3) 9 (1.6) 14 (1.2) 0.414
Hematocrit of £30% 353 (20.0) 112 (20.3) 241 (19.9) 0.640
Creatinine level of ‡2 mg/dL 90 (5.1) 32 (5.8) 58 (4.8) 0.331
Albumin of <3 g/dL 164 (9.3) 60 (10.9) 104 (8.6) 0.003
White blood-cell count 0.175
<4,000 cells/mL 17 (1.0) 3 (0.5) 14 (1.2)
‡12,000 cells/mL 431 (24.4) 147 (26.6) 284 (23.4)
Platelets <100 billion cells/L 73 (4.1) 18 (3.3) 55 (4.5) 0.458
INR >1.5 71 (4.0) 23 (4.2) 48 (4.0) 0.730
Bilirubin ‡2 mg/dL 20 (1.1) 10 (1.8) 10 (0.8) 0.021

*ASA = American Society of Anesthesiologists, CVA = cerebrovascular accident, COPD = chronic obstructive pulmonary disease, ESRD = end-stage
renal disease, and INR = international normalized ratio. †Data were missing for 2 patients.
 158
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

fixation from 2010 to 2013. Data on resident participation time was 53 minutes (median, 45 minutes; interquartile
were available for 1,764 patients (21.0%); residents partici- range, 32.5 to 64 minutes). Mean total hospital length of stay
pated in the cases of 552 patients (31.3%), and 540 (97.8%) was 7.7 days (median, 5 days; interquartile range, 4 to 8 days)
of 552 residents were categorized as “Attending & Resident and mean length of stay from the time of the operation to
in OR [operating room].” Extramedullary implants were discharge was 6.4 days (median, 4 days; interquartile range, 3
used in 2,982 (35.6%) of 8,384 patients, while 5,402 (64.4%) to 6 days).
were managed with an intramedullary implant. General Unadjusted analysis (Table II) of the overall cohort
anesthesia rates were not significantly different for proce- demonstrated that resident participation was not associated
dures in which residents participated (75.2%) compared with a significant difference with respect to death or serious
with those in which attending surgeons performed the case morbidity (49.1% versus 48.1%; p = 0.699), mortality (6.3%
alone (71.1%). versus 7.8%; p = 0.264), or serious morbidity (44.9% versus
Analysis of the full cohorts (resident involved and at- 43.2%; p = 0.506). Similarly, there were no significant differ-
tending surgeon alone) demonstrated a significant difference ences for resident participation with respect to death or serious
with respect to the following variables: sex, age, race, BMI, morbidity, mortality, or serious morbidity when patients
functional status, prior cerebrovascular accident or stroke, re- treated with extramedullary fixation (Table III) and those
nal disease, albumin level, and bilirubin level (Table I). treated with intramedullary implants (Table IV) were analyzed
separately.
Univariate Models The majority of all postoperative complications classified
Of the 8,384 patients in the NSQIP, 4,179 (49.8%) died or as serious morbidity were related to blood loss requiring
had a serious morbidity; there were 503 deaths (6.0%) and transfusion, with 719 (40.8%) of 1,764 of all patients meeting
4,025 serious complications (48.0%). Mean total operative this criterion. There were no significant associations for any

TABLE II Univariate Analysis for All Surgically Treated Intertrochanteric Hip Fractures Grouped by Resident Participation

Resident Involved Attending Surgeon Alone

(N = 552) (no. [%]) (N = 1,212) (no. [%]) P Value

Complications
Death or serious morbidity 271 (49.1) 583 (48.1) 0.699
Death 35 (6.3) 95 (7.8) 0.264
Serious morbidity 248 (44.9) 524 (43.2) 0.506
Organ space infection 1 (0.2) 0 (0) 0.138
Wound complication 7 (1.3) 10 (0.8) 0.377
Venous thromboembolism 9 (1.6) 23 (1.9) 0.697
Respiratory 26 (4.7) 53 (4.4) 0.751
Renal 3 (0.5) 9 (0.7) 0.637
Neurologic 4 (0.7) 10 (0.8) 0.826
Cardiac 19 (3.4) 30 (2.5) 0.252
Bleeding 226 (40.9) 493 (40.7) 0.916
Sepsis 19 (3.4) 25 (2.1) 0.085
Time variables
Total operative time <0.001
0-90 min 436 (79.0) 1,151 (95.0)
>90 min 116 (21.0) 61 (5.0)
Total length of stay <0.001
0-14 days 466 (84.4) 1,109 (91.5)
>14 days 83 (15.0) 102 (8.4)
Time to operating room 0.418
0-1 days 418 (75.7) 939 (77.5)
‡2 days 134 (24.3) 273 (22.5)
Time to discharge <0.001
0-12 days 471 (85.3) 1,111 (91.7)
>12 days 78 (14.1) 100 (8.3)
 159
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

TABLE III Univariate Analysis for Intertrochanteric Hip Fractures Treated with Extramedullary Fixation Grouped by Resident Participation

Resident Involved Attending Surgeon Alone

(N = 168) (no. [%]) (N = 435) (no. [%]) P Value

Complications
Death or serious morbidity 72 (42.9) 207 (47.6) 0.296
Death 6 (3.6) 27 (6.2) 0.202
Serious morbidity 63 (37.5) 192 (44.1) 0.139
Time variables
Total operative time <0.001
0-90 min 132 (78.6) 419 (96.3)
>90 min 36 (21.4) 16 (3.7)
Total length of stay 0.052
0-14 days 135 (80.4) 378 (86.9)
>14 days 32 (19.0) 56 (12.9)
Time to operating room 0.770
0-1 days 122 (72.6) 321 (73.8)
‡2 days 46 (27.4) 114 (26.2)
Time to discharge 0.167
0-12 days 139 (82.7) 380 (87.4)
>12 days 28 (16.7) 54 (12.4)

postoperative complication categories with resident involve- 7.2 days; p = 0.071) or time from the operation to discharge (6.6
ment (Table II). versus 5.9 days; p = 0.181). Further analysis of the proportion of
Analysis of perioperative time variables demonstrated that cases falling in the 90th percentile or higher (defined as pro-
resident participation was associated with a significant increase in longed) for time parameters demonstrated that resident partici-
mean operative time (68.9 versus 45.7 minutes; p < 0.001); there pation was associated with prolonged operative time (p < 0.0001),
was not a significant difference in overall length of stay (8.1 versus prolonged total length of stay (p < 0.001), and prolonged time to

TABLE IV Univariate Analysis for Intertrochanteric Hip Fractures Treated with Intramedullary Fixation Grouped by Resident Participation

Resident Involved Attending Surgeon Alone

(N = 384) (no. [%]) (N = 777) (no. [%]) P Value

Complications
Death or serious morbidity 199 (51.8) 376 (48.4) 0.079
Death 29 (7.6) 68 (8.8) 0.487
Serious morbidity 185 (48.2) 332 (42.7) 0.079
Time variables
Total operative time <0.001
0-90 minutes 304 (79.2) 732 (94.2)
>90 minutes 80 (20.8) 45 (5.8)
Total length of stay <0.001
0-14 days 331 (86.2) 731 (94.1)
>14 days 51 (13.3) 46 (5.9)
Time to operating room 0.337
0-1 days 296 (77.1) 618 (79.5)
‡2 days 88 (22.9) 159 (20.5)
Time to discharge <0.001
0-12 days 332 (86.5) 731 (94.1)
>12 days 50 (13.0) 46 (5.9)
 160
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

TABLE V Patient Characteristics for Propensity Score-Matched Cohorts*

Characteristics* Overall (no. [%]) Resident Involved (no. [%]) Attending Alone (no. [%]) P Value

Total 1,058 (100.0) 529 (50.0) 529 (50.0)

Sex 1.000
Male 338 (31.9) 169 (31.9) 169 (31.9)
Female 720 (68.1) 360 (68.1) 360 (68.1)
Age 0.738
<70 yr 170 (16.1) 87 (16.5) 83 (15.7)
‡70 yr 888 (83.9) 442 (83.6) 446 (84.3)
ASA class 0.512
I/II 174 (16.5) 92 (17.4) 82 (15.5)
III 659 (62.3) 331 (62.6) 328 (62.0)
IV/V 225 (21.3) 106 (20.0) 119 (22.5)
Functional status 0.659
Independent 645 (61.0) 319 (60.3) 326 (61.6)
Dependent 413 (39.0) 210 (39.7) 203 (38.4)
Comorbidities
Diabetes 197 (18.6) 98 (18.5) 99 (18.7) 0.937
Cardiac disease 195 (18.4) 96 (18.2) 99 (18.7) 0.812
Hypertension 730 (69.0) 359 (67.9) 371 (70.1) 0.425
CVA or stroke 62 (5.9) 41 (7.8) 21 (4.0) 0.019
Smoking 154 (14.6) 86 (16.3) 68 (12.9) 0.117
COPD 142 (13.4) 71 (13.4) 71 (13.4) 1.000
Alcohol abuse 43 (4.1) 21 (4.0) 22 (4.2) 0.694
Chronic steroid use 55 (5.2) 27 (5.1) 28 (5.3) 0.890
ESRD on dialysis 34 (3.2) 19 (3.6) 15 (2.8) 0.486
Disseminated malignancy 15 (1.4) 9 (1.7) 6 (1.1) 0.435
Hematocrit of £30% 220 (20.8) 109 (20.6) 111 (20.9) 0.912
Creatinine level of ‡2mg/dL 57 (5.4) 29 (5.5) 28 (5.3) 0.886
Albumin level of <3 g/dL 107 (10.1) 56 (10.6) 51 (9.6) 0.085
White blood-cell count <.001
<4,000 cells/mL 13 (1.2) 3 (0.6) 10 (1.9)
‡12,000 cells/mL 154 (14.6) 140 (26.5) 14 (2.6)
Platelets <100 billion cells/L 32 (3.0) 17 (3.2) 15 (2.8) 0.822
INR >1.5 40 (3.8) 22 (4.2) 18 (3.4) 0.843
Bilirubin of ‡2 mg/dL 11 (1.0) 5 (0.9) 6 (1.1) 0.904

*ASA = American Society of Anesthesiologists, CVA = cerebrovascular accident, COPD = chronic obstructive pulmonary disease, ESRD = end-stage
renal disease, and INR = international normalized ratio.

discharge (p < 0.001) in the overall cohort. When cases were 529 cases performed by an attending surgeon alone. Significant
analyzed by fixation type, resident participation was associated differences between groups with regard to demographic char-
with prolonged operative time for both the extramedullary (p < acteristics, clinical factors, and preoperative laboratory values
0.001) and intramedullary (p < 0.001) cohorts (Tables III and IV, initially observed in the unmatched cohort were successfully
respectively). In the intramedullary group, resident participation balanced in the propensity score-matched cohort (Table V).
was also associated with prolonged total length of stay (p < 0.001) Univariate analysis of the matched cohort demonstrated that
and prolonged time to discharge (p < 0.001). resident involvement was not significantly associated with 30-
day postoperative death or serious morbidity, mortality, or
Propensity Score-Matched Cohort Analysis overall rate of a serious complication (Table VI). Sepsis was the
Propensity score matching resulted in a cohort of 1,058 cases only complication category demonstrating a significant dif-
evenly split between 529 cases with resident involvement and ference between the groups (with an increased rate in the
 161
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

TABLE VI Impact of Resident Involvement on Postoperative Outcomes After Propensity Score Matching

Resident Involved Attending Surgeon Alone

(N = 529) (N = 529) P Value

Complications (no. [%])

Death or serious 258 (48.8) 239 (45.2) 0.242
morbidity
Death 34 (6.4) 38 (7.2) 0.625
Serious morbidity 235 (44.4) 211 (39.9) 0.135
Organ space infection 1 (0.2) 0 (0) 0.317
Wound complication 7 (1.3) 5 (0.9) 0.561
Venous 9 (1.7) 8 (1.5) 0.807
thromboembolism
Respiratory 24 (4.5) 24 (4.5) 1.000
Renal 3 (0.6) 2 (0.4) 0.654
Neurologic 4 (0.8) 2 (0.4) 0.413
Cardiac 18 (3.4) 13 (2.5) 0.362
Bleeding 214 (40.5) 196 (37.1) 0.256
Sepsis 18 (3.4) 7 (1.3) 0.026
Time variables (no. [%])
Total operative time <0.001
0-90 min 418 (79.0) 505 (95.5)
>90 min 111 (21.0) 24 (4.5)
Total length of stay <0.001
0-14 days 448 (85.0) 498 (94.1)
>14 days 79 (15.0) 31 (5.9)
Time to operating room 0.661
0-1 days 404 (76.4) 410 (77.5)
‡2 days 125 (23.6) 119 (22.5)
Time to discharge <0.001
0-12 days 453 (86.0) 501 (94.7)
>12 days 74 (14.0) 28 (5.3)

resident cohort; p = 0.026). Similar to the unmatched analysis, Discussion

resident involvement was associated with prolonged operative
time (p < 0.001), prolonged total length of stay (p < 0.001), and
prolonged time from the operation to discharge (p < 0.001).
T his study is an observational analysis of a large national
cohort evaluating the impact of resident involvement on
30-day postoperative complications including death and seri-
ous morbidity as well as perioperative time parameters. In a
Multivariate Model comparison of surgical procedures done with resident in-
In a logistic, forward stepwise regression model, there was no volvement and those with an attending surgeon operating
significant association between resident involvement and the alone, resident participation was not associated with death or
primary clinical outcomes of interest (death, serious morbidity, serious morbidity in univariate unmatched analysis (49.1%
and death or serious morbidity) in the overall cohort. These versus 48.1%; p = 0.699), univariate propensity score-matched
findings were also true for both the extramedullary and in- analysis (48.8% versus 45.2%; p = 0.242), and multivariate
tramedullary fixation subsets (Table VII). The multivariate logistic regression analysis (odds ratio, 0.73; p = 0.082). Fur-
model, however, did show that even after adjustment for po- thermore, subset analysis of the latter demonstrated resident
tential confounders, resident involvement was associated with participation was not associated with death or serious mor-
prolonged operative time (p < 0.001) and prolonged length of bidity for either extramedullary fixation (p = 0.052) or intra-
stay (p = 0.027) in the overall cohort, with prolonged operative medullary fixation (p = 0.425). These findings held true when
time in both the extramedullary (p < 0.001) and intramedul- mortality and serious morbidity were analyzed separately.
lary (p < 0.001) groups, and with prolonged length of stay (p = Resident participation was associated with prolonged
0.008) and prolonged time to discharge (p = 0.020) in the perioperative time variables. Prolonged operative time, total
intramedullary group (Table VII). length of stay, and time from the operation to discharge were
 162
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

TABLE VII Effect of Resident Involvement on Postoperative Outcomes After Multivariate Analysis*

Overall EM IM
Outcomes OR P Value OR P Value OR P Value

Complications
Death 0.73 0.271 0.49 0.355 0.85 0.180
Serious morbidity 0.86 0.383 0.46 0.053 1.08 0.719
Death or serious 0.73 0.082 0.47 0.052 0.85 0.425
morbidity
Time variables
Operative time of >90 4.13 <0.001 7.47 <0.001 3.36 <0.001
minutes
Length of stay of >14 days 1.92 0.027 1.41 0.548 2.67 0.008
Time to operating room of 0.80 0.259 0.70 0.361 0.89 0.621
‡2 days
Time to discharge of >12 1.70 0.082 0.88 0.834 2.35 0.020
days

*EM = extramedullary fixation, IM = intramedullary implant, and OR = odds ratio.

more common in the resident participation group in the surgical trainees do not adversely affect outcomes across several
overall cohort. Subset analysis demonstrated that prolonged different common orthopaedic surgical procedures. Prior
operative time was increased in both extramedullary and studies involving spine surgery and total joint arthroplasty have
intramedullary groups, while prolonged length of stay and failed to demonstrate an association between resident partici-
prolonged time to discharge were increased only in the intra- pation and the rate of complications with respect to the time of
medullary group. Intramedullary nailing is performed for the academic year17,27. Further, NSQIP analyses of both primary
more complex hip fractures, which, combined with the in- total hip and primary total knee arthroplasties showed no as-
creased frailty of this group as highlighted in the descriptive sociation between resident involvement and the rate of com-
analysis, may account for the increased length of stay. plications and demonstrated no difference in complication
Hip fractures are an increasingly prevalent challenge rates as a function of house staff seniority17,18. Additionally,
facing orthopaedic surgeons. On completion of residency Cvetanovich et al. found that resident involvement was not a
training, graduates are expected to safely manage intertro- risk factor for complications within 30 days after total shoulder
chanteric hip fractures in both community and academic set- arthroplasty16. Pugely et al. performed an NSQIP analysis of all
tings. In an era when health care must be delivered effectively orthopaedic procedures across 6 domains (primary total joint
and efficiently, surgeons and hospitals have intensified efforts arthroplasties, revision total joint arthroplasties, basic ar-
to improve patient outcomes and reduce complications. Mor- throscopies, advanced arthroscopies, lower extremity traumas,
bidity and mortality remain the most important and readily and spine fusions)28, concluding that resident involvement is
available outcome metrics. Questions remain as to the effects of associated with equivalent morbidity and mortality despite
surgical trainee participation on patient outcomes. Our in- increases in operative time. Additionally, they noted that the
vestigation focused on identifying whether resident participa- postgraduate year of the residents failed to show an association
tion was associated with complications in the surgical with surgical outcomes. More recently, in a study of patients
management of intertrochanteric hip fractures. The NSQIP undergoing elective posterior cervical fusion, Lee et al. reported
database allows evaluation of short-term complications asso- a minimal association between resident involvement and
ciated with surgical procedures in a large cohort, comprising morbidity, with resident involvement found to be a predictor
cases from a variety of practice settings, with >400 participating for increased blood transfusions, increased operative time, and
institutions representing several categories of health-care increased length of stay29. These findings, together with our
delivery systems (e.g., academic centers, community hospi- study, suggest that resident involvement in orthopaedic pro-
tals, and urban centers). cedures is not associated with increased risk to patients. Our
Data are currently limited with respect to the impact of study did reveal that resident involvement is associated with
resident involvement in the operative management of inter- increased time-related variables, including prolonged operative
trochanteric hip fractures. The present study expands on the time, prolonged total length of stay, and prolonged time to
subject and is unique in the literature, with the subset analysis discharge (using 90th percentile cutoffs). Despite the associa-
of extramedullary compared with intramedullary fixation. Our tion with prolonged time variables, resident participation did
findings are consistent with other series that have found that not show an increase in 30-day mortality or serious morbidity.
 163
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

The finding of an increased operative time with surgical projecting eventual outcomes. Moreover, the NSQIP data-
trainees is not unexpected, as the educational component base does not identify procedural steps performed by the
certainly adds to the length of the procedure. Increased length resident in the operating room, precluding the ability to
of stay may be attributable to less efficient disposition of pa- clearly define the level of participation and autonomy of the
tients by trainees than by advanced practice providers with involved resident in each case. Granular details of the op-
specific training in the care of hospitalized patients. Further- erations are also not available for analysis, e.g., the specific
more, patients receiving resident-involved care were more fracture pattern, quality of reduction, implant(s) used, or
likely to be malnourished and with dependent functional sta- intraoperative challenges, all of which could impact peri-
tus—factors that perhaps necessitated prolonged postoperative operative time parameters as well as postoperative compli-
inpatient optimization. Overall, our findings suggest that res- cations and ultimate outcomes of fracture union and patient
ident and fellow training incurs increased financial costs to function. Additionally, it is not currently possible to eval-
health systems, but not at the expense of short-term patient uate the data by type of institution, so while the NSQIP
outcomes. As the United States moves forward with health-care database contains data from all practice settings, the true
financing initiatives, such as bundled care, it is imperative that scalability of our findings is not known. Despite its limita-
the Centers for Medicare & Medicaid Services recognize the tions and the inability to project long-term outcomes, ac-
increased cost in health care associated with trainee education knowledging the multifactorial nature of outcomes and
and not penalize the academic health centers or the learners. complications following surgical fixation of intertrochan-
Increased pressure on surgeon-educators via participation in teric hip fractures, the NSQIP database, when analyzed
cost-cutting measures should not impact resident education. critically with appropriate statistical methodology, does
Interestingly, we found several significant demographic provide a robust opportunity to analyze the association of
variables that indicate that residents were involved in the care resident participation on 30-day outcomes. Similar to the work
of sicker patients. Despite evidence that patients cared for by that has been done in general surgery, our study opens the door
residents were sicker, clinical outcomes were not different for future, prospective evaluations of the impact of the trainee on
than those of patients treated by attending surgeons alone; outcomes in hip fracture surgery and the granular evaluation of
perhaps, this suggests that academic and/or tertiary care fa- specific factors. n
cilities are well equipped to care for sicker patients and can
help to mitigate increased risk associated with advanced pa-
tient comorbidities. Efficiency and value in hip fracture sur-
gery must be balanced with the obligation to educate the Alexander L. Neuwirth, MD1
trainee. Russell N. Stitzlein, MD1
The merits of this study, as with other studies that Madalyn G. Neuwirth, MD1
Rachel K. Kelz, MD, MPH1,2
utilize the NSQIP database, are that it utilizes a robust da- Samir Mehta, MD1,2
tabase that has been shown to be reliable for capturing
short-term complications30,31. The variety of geographic 1Departments of Orthopaedic Surgery (A.L.N., R.N.S., and S.M.) and

and demographic representation helps with generalizability General Surgery (M.G.N and R.K.K.), University of Pennsylvania Health
of findings. Despite these strengths, studies of this nature System, Philadelphia, Pennsylvania
have known limitations. Our study addressed resident par- 2Perelman
ticipation, so we were limited to 21.0% of the 8,384 cases School of Medicine, University of Pennsylvania,
Philadelphia, Pennsylvania
included for 2010 to 2013 because of missing data with re-
gard to whether residents participated in the procedures. E-mail address for S. Mehta: Samir.Mehta@uphs.upenn.edu
Additionally, the NSQIP database captures only the first 30
days after surgery, limiting its utility at demonstrating or ORCID iD for S. Mehta: 0000-0001-5191-2403

References
1. Brauer CA, Coca-Perraillon M, Cutler DM, Rosen AB. Incidence and mortality of hip 6. Ahrengart L, Törnkvist H, Fornander P, Thorngren KG, Pasanen L, Wahlström P,
fractures in the United States. JAMA. 2009 Oct 14;302(14):1573-9. Honkonen S, Lindgren U. A randomized study of the compression hip screw and
2. Cummings SR, Rubin SM, Black D. The future of hip fractures in the United Gamma nail in 426 fractures. Clin Orthop Relat Res. 2002 Aug;401:209-22.
States. Numbers, costs, and potential effects of postmenopausal estrogen. Clin 7. Chen F, Wang Z, Bhattacharyya T. Convergence of outcomes for hip fracture fixation by
Orthop Relat Res. 1990 Mar;252:163-6. nails and plates. Clin Orthop Relat Res. 2013 Apr;471(4):1349-55. Epub 2012 Nov 27.
3. Burge R, Dawson-Hughes B, Solomon DH, Wong JB, King A, Tosteson A. Inci- 8. Parker MJ, Handoll HH. Gamma and other cephalocondylic intramedullary nails
dence and economic burden of osteoporosis-related fractures in the United States, versus extramedullary implants for extracapsular hip fractures in adults. Cochrane
2005-2025. J Bone Miner Res. 2007 Mar;22(3):465-75. Database Syst Rev. 2010 Sep 8;9:CD000093.
4. Kokoroghiannis C, Aktselis I, Deligeorgis A, Fragkomichalos E, Papadimas D, 9. Utrilla AL, Reig JS, Muñoz FM, Tufanisco CB. Trochanteric gamma nail and
Pappadas I. Evolving concepts of stability and intramedullary fixation of intertro- compression hip screw for trochanteric fractures: a randomized, prospective, com-
chanteric fractures—a review. Injury. 2012 Jun;43(6):686-93. Epub 2011 Jul 14. parative study in 210 elderly patients with a new design of the Gamma nail. J Orthop
5. Kaplan K, Miyamoto R, Levine BR, Egol KA, Zuckerman JD. Surgical management Trauma. 2005 Apr;19(4):229-33.
of hip fractures: an evidence-based review of the literature. II: intertrochanteric 10. Pajarinen J, Lindahl J, Michelsson O, Savolainen V, Hirvensalo E. Pertrochan-
fractures. J Am Acad Orthop Surg. 2008 Nov;16(11):665-73. teric femoral fractures treated with a dynamic hip screw or a proximal femoral nail. A
 164
TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
d
R E S I D E N T P A R T I C I PAT I O N IN F I X AT I O N O F IN T E RT RO CH A N T E R I C HIP
V O LU M E 1 00-A N U M B E R 2 J A N UA R Y 17, 2 018
d d
FRACTURE S

randomised study comparing post-operative rehabilitation. J Bone Joint Surg Br. a zero pneumonia rate in general surgery patients. Perm J.2012
2005 Jan;87(1):76-81. Winter;16(1):39-45.
11. Hardy DC, Descamps PY, Krallis P, Fabeck L, Smets P, Bertens CL, Delince PE. 21. American College of Surgeons National Surgical Quality Improvement Program.
Use of an intramedullary hip-screw compared with a compression hip-screw with a User Guide for the 2012 ACS NSQIP Participant Use Data File (PUF). 2014 Nov.
plate for intertrochanteric femoral fractures. A prospective, randomized study of one https://www.facs.org/;/media/files/quality%20programs/nsqip/2013_acs_
hundred patients. J Bone Joint Surg Am. 1998 May;80(5):618-30. nsqip_puf_user_guide.ashx. Accessed 2017 Aug 21.
12. Bohl DD, Basques BA, Golinvaux NS, Miller CP, Baumgaertner MR, Grauer JN. 22. Strasberg SM, Linehan DC, Hawkins WG. The Accordion severity grading system
Extramedullary compared with intramedullary implants for intertrochanteric hip of surgical complications. Ann Surg. 2009 Aug;250(2):177-86.
fractures: thirty-day outcomes of 4432 procedures from the ACS NSQIP database. J 23. Austin PC. An introduction to propensity score methods for reducing the effects
Bone Joint Surg Am. 2014 Nov 19;96(22):1871-7. of confounding in observational studies. Multivariate Behav Res. 2011 May;46
13. Anglen JO, Weinstein JN; American Board of Orthopaedic Surgery Research (3):399-424. Epub 2011 Jun 8.
Committee. Nail or plate fixation of intertrochanteric hip fractures: changing pattern 24. Dugoff EH, Schuler M, Stuart EA. Generalizing observational study results: ap-
of practice. A review of the American Board of Orthopaedic Surgery database. J Bone plying propensity score methods to complex surveys. Health Serv Res. 2014 Feb;49
Joint Surg Am. 2008 Apr;90(4):700-7. (1):284-303. Epub 2013 Jul 16.
14. Barton TM, Gleeson R, Topliss C, Greenwood R, Harries WJ, Chesser TJS. A 25. Austin PC. The performance of different propensity score methods for esti-
comparison of the long Gamma nail with the sliding hip screw for the treatment of mating marginal hazard ratios. Stat Med. 2013 Jul 20;32(16):2837-49. Epub 2012
AO/OTA 31-A2 fractures of the proximal part of the femur: a prospective randomized Dec 12.
trial. J Bone Joint Surg Am. 2010 Apr;92(4):792-8. 26. Austin PC. Optimal caliper widths for propensity-score matching when esti-
15. Schoenfeld AJ, Serrano JA, Waterman BR, Bader JO, Belmont PJ Jr. The impact mating differences in means and differences in proportions in observational studies.
of resident involvement on post-operative morbidity and mortality following ortho- Pharm Stat. 2011 Mar-Apr;10(2):150-61.
paedic procedures: a study of 43,343 cases. Arch Orthop Trauma Surg. 2013 27. Auerbach JD, Lonner BS, Antonacci MD, Kean KE. Perioperative outcomes and
Nov;133(11):1483-91. Epub 2013 Sep 1. complications related to teaching residents and fellows in scoliosis surgery. Spine
16. Cvetanovich GL, Schairer WW, Haughom BD, Nicholson GP, Romeo AA. Does (Phila Pa 1976). 2008 May 1;33(10):1113-8.
resident involvement have an impact on postoperative complications after total 28. Pugely AJ, Gao Y, Martin CT, Callagh JJ, Weinstein SL, Marsh JL. The effect of
shoulder arthroplasty? An analysis of 1382 cases. J Shoulder Elbow Surg. 2015 resident participation on short-term outcomes after orthopaedic surgery. Clin Orthop
Oct;24(10):1567-73. Epub 2015 May 5. Relat Res. 2014 Jul;472(7):2290-300.
17. Haughom BD, Schairer WW, Hellman MD, Yi PH, Levine BR. Does resident 29. Lee NJ, Kothari P, Kim C, Leven DM, Skovrlj B, Guzman JZ, Steinberger J, Cho
involvement impact post-operative complications following primary total knee ar- SK. The impact of resident involvement in elective posterior cervical fusion. Spine
throplasty? An analysis of 24,529 cases. J Arthroplasty. 2014 Jul;29(7):1468- (Phila Pa 1976). 2016 Feb 1;(1). Epub 2016 Feb 1.
1472.e2. Epub 2014 Mar 6. 30. Cima RR, Lackore KA, Nehring SA, Cassivi SD, Donohue JH, Deschamps C,
18. Haughom BD, Schairer WW, Hellman MD, Yi PH, Levine BR. Resident involve- Vansuch M, Naessens JM. How best to measure surgical quality? Comparison of the
ment does not influence complication after total hip arthroplasty: an analysis of Agency for Healthcare Research and Quality Patient Safety Indicators (AHRQ-PSI)
13,109 cases. J Arthroplasty. 2014 Oct;29(10):1919-24. Epub 2014 Jun 6. and the American College of Surgeons National Surgical Quality Improvement Pro-
19. Bohl DD, Fu MC, Golinvaux NS, Basques BA, Gruskay JA, Grauer JN. The gram (ACS-NSQIP) postoperative adverse events at a single institution. Surgery.
“July effect” in primary total hip and knee arthroplasty: analysis of 21,434 2011 Nov;150(5):943-9. Epub 2011 Aug 27.
cases from the ACS-NSQIP database. J Arthroplasty. 2014 Jul;29(7):1332-8. 31. Davenport DL, Holsapple CW, Conigliaro J. Assessing surgical quality using ad-
Epub 2014 Feb 10. ministrative and clinical data sets: a direct comparison of the University Health System
20. Fuchshuber PR, Greif W, Tidwell CR, Klemm MS, Frydel C, Wali A, Rosas E, Clopp Consortium Clinical Database and the National Surgical Quality Improvement Program
MP. The power of the National Surgical Quality Improvement Program—achieving data set. Am J Med Qual. 2009 Sep-Oct;24(5):395-402. Epub 2009 Jul 7.