As health care costs mount higher, invested stakeholders are scrutinizing surgical practices with an eye toward improving value—defined as the patient's outcome relative to the cost of care.16 Insurers, patients, and government agencies are all beginning to make decisions about health care based on perceived value. Surgical complications are often broadly defined as “any deviation from the normal postoperative course.”4 This definition is clearly subjective and prone to interobserver variability.11 Within the past several years, nationwide databases, such as those hosted by the Cleveland Clinic (Cardiovascular Information Registry),11 the Agency for Healthcare Research and Quality (Patient Safety Indicators),9,17 and the American College of Surgeons (ACS) (National Surgical Quality Improvement Program [NSQIP]), have begun prospectively collecting data on surgical complications using objective criteria.8,10
In particular, the NSQIP database has strict criteria for defining a specified set of frequent postsurgical complications. Data are derived from hundreds of US hospitals in a wide range of settings, from small rural community hospitals to major academic medical centers. In 2006, there were 121 participating sites, and in 2013, there are more than 400 sites. All major surgical procedures performed are identified using Current Procedural Terminology (CPT) codes, and the standardized set of variables is collected from either all cases for low-volume hospitals or a sample group representing 20% of all cases performed for large-volume sites. The NSQIP utilizes trained raters to identify these complications directly from the medical records. The raters are frequently audited using test cases, and their data are ignored if they fail these auditing checks.19
With the NSQIP data set, it is possible to determine the most frequent complications in neurological surgery. Identifying these complications will allow us to better distribute resources to prevent and mitigate the most costly and common adverse events in our patients. Moreover, understanding the type and frequency of complications in neurosurgery will allow us to better coordinate care with stakeholders—insurers, governmental institutions, and patients. In this study, we used the NSQIP database to examine trends and predictors of complications in neurosurgical patients.
Methods
Data for the years 2006 to 2011 were obtained from the ACS NSQIP database. Files were acquired in delimited text format and parsed using both SPSS version 20 (IBM Corp.) and Matlab R2012a (MathWorks Inc.). Neurosurgical cases were extracted by querying the data for which the primary surgeon's specialty was listed as “neurological surgery,” and then each of the CPT codes was individually reviewed to determine whether the case was cranial or spinal (Table 1).
CPT codes included in an analysis of neurosurgical complications
Spine Codes | Cranial Codes |
---|---|
0090T | 21137 |
0092T | 21139 |
0093T | 21180 |
0095T | 21299 |
0096T | 35001–5 |
0163T | 35201 |
0195T | 35301 |
0196T | 35800 |
0202T | 37605 |
10121 | 37799 |
10140 | 49422 |
10180 | 60600 |
11000 | 61150–888 |
11042–4 | 62000–362 |
20005 | 69511 |
20100 | |
20696 | |
20902 | |
20922 | |
20930 | |
20931 | |
20936 | |
20938 | |
20999 | |
21026 | |
21501 | |
21510 | |
21610 | |
21615 | |
22010 | |
22015 | |
22100–14 | |
22206–26 | |
22318–27 | |
22520–95 | |
22600–32 | |
22800–99 | |
27080 | |
27202 | |
27218 | |
27280 | |
42725 | |
49215 | |
63001–308 | |
63685 | |
63700–41 | |
64573–85 |
Descriptive statistics are represented as the means ± standard deviation. Univariate statistics were calculated using the chi-square test, and 95% confidence intervals are presented for the resultant risk ratios. A multivariate model was constructed with the significant variables from univariate analysis (p < 0.05) using binary logistic regression with a maximum number of iterations of 20. Significant predictors from this model were identified, and their risk ratios were presented, along with the 95% confidence intervals. Data were plotted using Matlab.
Results
Records from 1,777,035 surgical cases were compiled. Neurosurgeons performed 38,396 of these cases. Characteristics of the neurosurgical patients are presented in Table 2 and are compared with those of general surgical patients to reveal the unique features of neurosurgical patients while highlighting the landscape of complications encountered. Of note, compared with nonneurosurgical patients, the neurosurgical patients were more likely to be male and white. African American patients had a RR of 0.75 (95% CI 0.72–0.78) of undergoing neurosurgery as opposed to any other surgery, and Asians/Pacific Islanders had an RR of 0.75 (95% CI 0.70–0.81).
Summary of characteristics in neurosurgical compared with nonneurosurgical patients
Characteristic | Neurosurgical Patients (%) | Nonneurosurgical Patients (%) | p Value | RR (95% CI) |
---|---|---|---|---|
total no. of patients | 38,396 | 1,738,639 | ||
mean age (yrs) ± SD | 56.0 ± 15.0 | 55.7 ± 17.0 | <0.0001 | |
mean height (in) ± SD | 66.9 ± 4.2 | 66.2 ± 4.1 | <0.0001 | |
mean weight (lb) ± SD | 186.7 ± 46.9 | 186.2 ± 54.4 | <0.0001 | |
mean anesthesia time (mins) ± SD | 238.4 ± 129.2 | 161.5 ± 106.3 | <0.0001 | |
mean op time (mins) ± SD | 162.6 ± 110.1 | 110.0 ± 91.3 | <0.0001 | |
mean length of stay (days) ± SD | 4.8 ± 10.2 | 4.3 ± 9.9 | <0.0001 | |
sex* | ||||
male | 19,521 (50.8) | 737,714 (42.4) | 1† | |
female | 18,800 (49.0) | 999,673 (57.5) | 0.72 (0.70–0.73)‡ | |
race* | ||||
white | 30,360 (79.1) | 1,319,127 (75.9) | 1† | |
black or African American | 2,943 (7.7) | 171,462 (9.9) | 0.75 (0.72–0.78)‡ | |
Asian/Pacific Islander | 731 (1.9) | 42,327 (2.4) | 0.75 (0.70–0.81)‡ | |
American Indian or native Alaskan | 225 (0.6) | 11,992 (0.7) | 0.82 (0.72–0.93)‡ | |
unknown | 4,137 (10.8) | 193,731 (11.1) | ||
inpatient vs outpatient* | ||||
inpatient | 32,871 (85.6) | 1,121,524 (64.5) | 1† | |
outpatient | 5,525 (14.4) | 617,112 (35.5) | 0.31 (0.30–0.32)‡ | |
diabetes* | ||||
no | 32,918 (85.7) | 1,477,932 (85.0) | 1† | |
yes | 5,478 (14.3) | 260,704 (15.0) | 0.94 (0.92–0.97)‡ | |
insulin-dependent | 1,925 (5.0) | 102,225 (5.9) | 0.85 (0.81–0.89)‡ | |
not insulin-dependent | 3,553 (9.3) | 158,479 (9.1) | 1.00 (0.97–1.04) | |
tobacco use* | ||||
no | 28,693 (74.7) | 1,391,997 (80.1) | 1† | |
yes | 9,702 (25.3) | 346,621 (19.9) | 1.35 (1.32–1.38)‡ | |
hypertensive patient on antihypertensive(s)* | ||||
no | 20,392 (53.1) | 936,936 (53.9) | 1† | |
yes | 18,004 (46.9) | 801,689 (46.1) | 1.03 (1.01–1.05)‡ | |
emergency case* | ||||
no | 36,108 (94.0) | 1,525,718 (87.8) | 1† | |
yes | 2,288 (6.0) | 212,916 (12.2) | 0.46 (0.44–0.48)‡ |
Cases with unknown values, when encountered, were omitted from analysis.
Reference.
Denotes a statistically significant difference.
As might be expected, neurosurgical cases were far more likely to involve inpatient procedures rather than outpatient (outpatient RR 0.31, 95% CI 0.30–0.32). However, neurosurgery was less likely to be done emergently compared with nonneurosurgical procedures (RR 0.46, 95% CI 0.44–0.48). Moreover, neurosurgical procedures were significantly longer than nonneurosurgical ones, both in terms of anesthesia time (76.9 minutes, p < 0.0001) and operative time (52.6 minutes, p < 0.0001). Neurosurgical patients also stayed on average 0.5 day longer than the other surgical patients (p < 0.0001).
Other demographic features were also compared between neurosurgical and nonneurosurgical patients. Neurosurgery patients were less likely to have diabetes (RR 0.94, 95% CI 0.92–0.97) but more likely to be on medication for hypertension (RR 1.03, 95% CI 1.01–1.05). Neurosurgery patients were also more likely to smoke (RR 1.35, 95% CI 1.32–1.38).
In neurosurgical patients alone, we evaluated the frequency of complications documented in the NSQIP database. Cases were split into spinal (28,017 cases) and cranial (10,041 cases) surgery groups, and complications were analyzed independently (Table 3); the other 338 cases represented peripheral nerve and nonneurosurgical cases performed by neurosurgeons. Complications occurred in 14.3% (5507 of 38,396) of neurosurgical cases. Cranial procedures were significantly more likely to produce complications than spinal procedures (RR 2.575, 95% CI 2.386–2.779), with 23.6% of cranial procedures leading to complications compared with 11.2% of spinal procedures. The most frequent complication in the neurosurgical procedures overall or in either spinal or cranial procedures alone was bleeding requiring a transfusion, which occurred in 4.5% of patients, and was more likely to occur in cranial than in spinal patients (RR 1.255, 95% CI 1.136–1.385). The second most common complication was reoperation within 30 days of the initial operation, and this occurred in 4.3% of neurosurgical patients overall and was more likely in cranial than in spinal patients (RR 2.11, 95% CI 2.01–2.22). The third most common complication was failure to wean from mechanical ventilation for more than 48 hours postoperatively, which occurred in 2.5% of neurosurgical patients. In fact, all complications except 5 were more likely in cranial procedures than in spinal ones. Of these 5, there was no significant difference in 4 and only 1 in which spinal procedures were significantly more likely to generate a complication: superficial surgical site infection (cranial RR 0.67, 95% CI 0.52–0.86). The 5 nonsignificantly different complications were wound dehiscence, deep surgical site infection, peripheral nerve injury, and graft, prosthesis, or flap failure.
Frequency of complications in 38,058 neurosurgical cases*
Parameter | Cranial Op (%) | Spinal Op (%) | Cranial & Spinal Op Combined (%) | RR, Cranial vs Spinal (95% CI) |
---|---|---|---|---|
no. of cases | 10,041 | 28,017 | ||
complication | ||||
superficial incisional infection | 72 (0.7) | 301 (1.1) | 373 (1.0) | 0.667 (0.517–0.862)† |
deep incisional infection | 45 (0.4) | 166 (0.6) | 211 (0.6) | 0.756 (0.545–1.051) |
organ/space infection | 105 (1.0) | 91 (0.3) | 196 (0.5) | 3.220 (2.434–4.259)† |
wound dehiscence | 24 (0.2) | 82 (0.3) | 106 (0.3) | 0.817 (0.519–1.286) |
postop pneumonia | 367 (3.6) | 212 (0.8) | 579 (1.5) | 4.830 (4.085–5.711)† |
unplanned reintubation | 300 (3.0) | 182 (0.6) | 482 (1.3) | 4.599 (3.831–5.522)† |
pulmonary embolism | 137 (1.4) | 107 (0.4) | 244 (0.6) | 3.573 (2.777–4.596)† |
failure to wean from ventilator for >48 hrs postop | 759 (7.6) | 191 (0.7) | 950 (2.5) | 11.088 (9.477–12.973)† |
progressive renal insufficiency | 24 (0.2) | 36 (0.1) | 60 (0.2) | 1.860 (1.110–3.116)† |
acute renal failure requiring dialysis | 21 (0.2) | 20 (0.1) | 41 (0.1) | 2.930 (1.589–5.403)† |
urinary tract infection | 402 (4.0) | 458 (1.6) | 860 (2.3) | 2.449 (2.146–2.795)† |
stroke/CVA | 219 (2.2) | 42 (0.1) | 261 (0.7) | 14.549 (10.466–20.225)† |
coma (>24 hrs) | 113 (1.1) | 5 (0.0) | 118 (0.3) | 63.060 (25.756–154.395)† |
peripheral nerve injury | 10 (0.1) | 32 (0.1) | 42 (0.1) | 0.872 (0.429–1.773) |
cardiac arrest requiring CPR | 65 (0.6) | 49 (0.2) | 114 (0.3) | 3.701 (2.556–5.359)† |
myocardial infarction | 34 (0.3) | 54 (0.2) | 88 (0.2) | 1.757 (1.145–2.697)† |
bleeding requiring transfusion | 540 (5.4) | 1201 (4.3) | 1741 (4.6) | 1.255 (1.136–1.385)† |
graft/prosthesis/flap failure | 2 (0.0) | 7 (0.0) | 9 (0.0) | 0.797 (0.166–3.837) |
deep venous thrombosis requiring therapy | 259 (2.6) | 225 (0.8) | 484 (1.3) | 3.212 (2.690–3.834)† |
sepsis | 363 (3.6) | 251 (0.9) | 614 (1.6) | 4.035 (3.441–4.732)† |
septic shock | 128 (1.3) | 80 (0.3) | 208 (0.5) | 4.464 (3.380–5.898)† |
reop w/in 30 days | 738 (7.3) | 899 (3.2) | 1637 (4.3) | 2.113 (2.013–2.217)† |
any complication | 2373 (23.6) | 3134 (11.2) | 5507 (14.5) | 2.575 (2.386–2.779)† |
CPR = cardiopulmonary resuscitation; CVA = cerebrovascular accident.
Denotes a statistically significant difference.
Using univariate analysis, we examined potential patient predictors that might be associated with the occurrence of a complication (Table 4). Many of these factors showed a significant association with complications. For example, African Americans were 1.51 times more likely to experience a complication than whites, and Asians/Pacific Islanders were 1.58 times more likely. Complications were also more likely to occur in inpatients (RR 4.95, 95% CI 4.28–5.73), patients with diabetes (RR 1.39, 95% CI 1.31–1.48), emergent cases (RR 3.83, 95% CI 3.64–4.04), and frequent users of alcohol (RR 1.36, 95% CI 1.20–1.54). Altogether, 34 characteristics were significantly associated with complications on univariate analysis.
Univariate predictors of complications among 5495 neurosurgical patients*
Factor | No. of Complications (%) | RR (95% CI) |
---|---|---|
sex | ||
male | 2685 (48.9) | 1† |
female | 2810 (51.1) | 1.09 (1.03–1.14) |
race‡ | ||
white | 4083 (74.1) | 1† |
black or African American | 596 (10.8) | 1.51 (1.39–1.62)§ |
Asian or Pacific Islander | 147 (2.7) | 1.58 (1.36–1.82)§ |
American Indian or native Alaskan | 21 (0.4) | 0.70 (0.47–1.05) |
spinal vs cranial | ||
spinal | 3134 (56.9) | 1† |
cranial | 2373 (43.1) | 2.11 (2.01–2.22)§ |
inpatient vs outpatient | ||
outpatient | 179 (3.3) | 1† |
inpatient | 5328 (96.7) | 4.95 (4.28–5.73)§ |
diabetes | ||
no | 4470 (81.2) | 1† |
yes | 1037 (18.8) | 1.39 (1.31–1.48)§ |
insulin-dependent | 439 (8.0) | 1.68 (1.54–1.83)§ |
not insulin-dependent | 598 (10.9) | 1.24 (1.15–1.34)§ |
tobacco use | ||
no | 4265 (77.4) | 1† |
yes | 1242 (22.6) | 0.86 (0.81–0.91)§ |
alcohol use‡ | ||
no | 3827 (94.8) | 1† |
yes | 208 (5.2) | 1.36 (1.20–1.54)§ |
dyspnea‡ | ||
no | 4985 (90.5) | 1† |
moderate exertion | 405 (7.4) | 1.32 (1.20–1.44)§ |
at rest | 116 (2.1) | 2.92 (2.53–3.36)§ |
ventilator dependent prior to op | ||
no | 4977 (90.4) | 1† |
yes | 530 (9.6) | 6.06 (5.79–6.34)§ |
severe COPD | ||
no | 5146 (93.4) | 1† |
yes | 361 (6.6) | 1.65 (1.50–1.81)§ |
current pneumonia‡ | ||
no | 3988 (99.0) | 1† |
yes | 41 (1.0) | 4.10 (3.37–4.99)§ |
congestive heart failure‡ | ||
no | 5538 (98.7) | 1† |
yes | 69 (1.3) | 3.79 (3.23–4.45)§ |
MI w/in 6 mos‡ | ||
no | 3992 (99.1) | 1† |
yes | 37 (0.9) | 3.31 (2.62–4.18)§ |
angina in past mo‡ | ||
no | 3998 (99.3) | 1† |
yes | 30 (0.7) | 1.44 (1.05–1.98)§ |
hypertensive patient on antihypertensive(s) | ||
no | 2390 (43.4) | 1† |
yes | 3117 (56.6) | 1.48 (1.40–1.55)§ |
prior angioplasty/amputation for PVD‡ | ||
no | 3963 (98.4) | 1† |
yes | 66 (1.6) | 1.73 (1.40–2.14)§ |
renal failure on dialysis | ||
no | 5428 (98.6) | 1† |
yes | 79 (1.4) | 3.49 (2.98–4.09)§ |
altered mental status‡ | ||
no | 3461 (85.9) | 1† |
yes | 568 (14.1) | 4.13 (3.87–4.40)§ |
coma prior to op‡ | ||
no | 3845 (95.4) | 1† |
yes | 184 (4.6) | 5.84 (5.46–.25)§ |
hemiplegia/hemiparesis prior to op‡ | ||
no | 3592 (89.2) | 1† |
yes | 437 (10.8) | 2.73 (2.52–2.95)§ |
history of transient ischemic attack(s)‡ | ||
no | 3888 (96.5) | 1† |
yes | 141 (3.5) | 1.49 (1.28–1.73)§ |
prior stroke w/ persistent neurological deficit‡ | ||
no | 3504 (87.0) | 1† |
yes | 525 (13.0) | 3.50 (3.26–3.75)§ |
prior stroke w/ no deficit‡ | ||
no | 3875 (96.2) | 1† |
yes | 154 (3.8) | 2.04 (1.78–2.34)§ |
CNS tumor‡ | ||
no | 3349 (83.1) | 1† |
yes | 680 (16.9) | 1.72 (1.60–1.85)§ |
paraplegia/paraparesis prior to op‡ | ||
no | 3731 (92.6) | 1† |
yes | 298 (7.4) | 1.90 (1.72–2.11)§ |
quadriplegia/quadriparesis prior to op‡ | ||
no | 3944 (97.9) | 1† |
yes | 85 (2.1) | 2.51 (2.11–2.98)§ |
disseminated/metastatic cancer‡ | ||
no | 5048 (92.3) | 1† |
yes | 423 (7.7) | 2.00 (1.83–2.17)§ |
open wound | ||
no | 5279 (95.9) | 1† |
yes | 228 (4.1) | 2.56 (2.30–2.85)§ |
chronic steroid use | ||
no | 4994 (90.7) | 1† |
yes | 513 (9.3) | 1.58 (1.46–1.71)§ |
>10% weight loss in last 6 mos | ||
no | 5389 (97.9) | 1† |
yes | 118 (2.1) | 1.94 (1.66–2.26)§ |
bleeding disorder | ||
no | 5124 (93.0) | 1† |
yes | 383 (7.0) | 2.53 (2.33–2.76)§ |
transfused >4 units RBCs prior to op | ||
no | 5416 (98.3) | 1† |
yes | 91 (1.7) | 4.93 (3.87–4.99)§ |
chemotherapy in past mo‡ | ||
no | 3941 (97.8) | 1† |
yes | 88 (2.2) | 2.22 (1.87–2.65)§ |
radiotherapy in past 90 days‡ | ||
no | 3943 (98.6) | 1† |
yes | 56 (1.4) | 2.23 (1.80–2.78)§ |
sepsis prior to op‡ | ||
no | 4870 (89.0) | 1† |
SIRS | 426 (7.8) | 3.57 (3.32–3.84)§ |
sepsis | 143 (2.6) | 3.82 (3.40–4.30)§ |
septic shock | 32 (0.6) | 4.07 (3.22–5.16)§ |
emergency case | ||
no | 4437 (80.6) | 1† |
yes | 1070 (19.4) | 3.83 (3.64–4.04)§ |
admission quarter | ||
1st | 1201 (21.8) | 1† |
2nd | 1127 (20.5) | 1.00 (0.93–1.08) |
3rd | 1507 (27.4) | 1.06 (0.99–1.13) |
4th | 1672 (30.4) | 1.08 (1.00–1.15)§ |
COPD = chronic obstructive pulmonary disease; MI = myocardial infarction; PVD = perivascular disease; RBC = red blood cell; SIRS = systemic inflammatory response syndrome.
Reference.
Cases with unknown values, when encountered, were omitted from analysis.
Denotes a statistically significant difference.
Using the predictors identified with univariate analysis, we then constructed a model using binary logistic regression to control for confounding. This analysis identified 15 significant predictors (Table 5). African American race continued to be a significant risk factor for developing complications (RR 1.24, 95% CI 1.10–1.41), whereas American Indian or Native Alaskan race continued to be protective (RR 0.53, 95% CI 0.28–0.98). However, Asian or Pacific Islander race was no longer a significant risk factor.
Significant multivariate predictors of complications
Factor | OR (95% CI) |
---|---|
sex | |
female | 1* |
male | 0.90 (0.83–0.97) |
race | |
white | 1* |
black or African American | 1.24 (1.10–1.41)† |
Asian or Pacific Islander | 1.28 (0.98–1.65)† |
American Indian or native Alaskan | 0.53 (0.28–0.98)† |
spinal vs cranial | |
spinal | 1* |
cranial | 1.41 (1.27–1.56)† |
inpatient vs outpatient | |
inpatient | 1* |
outpatient | 0.31 (0.26–0.37)† |
diabetes | |
no | 1* |
insulin-dependent | 1.30 (1.11–1.53)† |
not insulin-dependent | 1.08 (0.95–1.23) |
alcohol use | |
no | 1* |
yes | 1.25 (1.04–1.52)† |
dyspnea | |
no | 1* |
moderate exertion | 1.05 (0.71–1.56) |
at rest | 1.17 (1.00–1.37)† |
ventilator dependent prior to op | |
no | 1* |
yes | 5.16 (3.79–7.03)† |
severe COPD | |
no | 1* |
yes | 1.34 (1.12–1.60)† |
congestive heart failure | |
no | 1* |
yes | 1.73 (1.01–2.96)† |
hypertensive patient on antihypertensive(s) | |
no | 1* |
yes | 1.16 (1.06–1.27)† |
renal failure on dialysis | |
no | 1* |
yes | 2.03 (1.29–3.21)† |
altered mental status | |
no | 1* |
yes | 1.59 (1.33–1.91)† |
prior stroke w/ persistent neurological deficit | |
no | 1* |
yes | 1.60 (1.34–1.91)† |
prior stroke w/ no deficit | |
no | 1* |
yes | 1.81 (1.44–2.28)† |
paraplegia/paraparesis prior to op | |
no | 1* |
yes | 1.92 (1.63–2.26)† |
quadriplegia/quadriparesis prior to op | |
no | 1* |
yes | 2.63 (1.93–3.59)† |
open wound | |
no | 1* |
yes | 1.60 (1.25–2.04)† |
chronic steroid use | |
no | 1* |
yes | 1.28 (1.10–1.48)† |
preop transfusion w/ >4 units RBCs | |
no | 1* |
yes | 5.02 (3.07–8.22)† |
sepsis prior to op | |
no | 1* |
SIRS | 2.07 (1.69–2.54)† |
sepsis | 2.37 (1.70–3.31)† |
septic shock | 0.82 (0.41–1.65) |
brain or spine tumor | |
no | 1* |
yes | 1.42 (1.24–1.61)† |
emergency case | |
no | 1* |
yes | 2.27 (1.96–2.64)† |
Reference.
Denotes a statistically significant difference.
Many risk factors continued to be significantly associated with complications, such as diabetes, prior stroke, open wound, chronic steroid use, presurgical sepsis, and the need for a preoperative transfusion with > 4 units of red blood cells (Table 5). Cranial procedures still produced more complications than spinal procedures (RR 1.41, 95% 1.27–1.56), despite controlling for patient comorbidities.
Discussion
Minimizing surgical complications has been a goal in neurosurgery ever since Harvey Cushing began rigorously documenting his own surgical errors more than a century ago.12 However, understanding and mitigating complications has taken on renewed importance in the era of “pay for performance” and governmental and insurance reimbursement metrics.5,6,15,20,22 Therefore, we sought to describe the current state of complications in neurological surgery, to provide data to help understand how new regulations should be applied, and, of course, to better target interventions to improve patient well-being.
Complications occurred in 14.3% of neurosurgical cases. The most frequent complications, in either cranial or spinal operations, were blood loss requiring transfusion and reoperation within 30 days, followed, somewhat surprisingly, by failure to wean from mechanical ventilation for > 48 hours postoperatively, which occurred in 7.6% of cranial cases and 0.7% of spinal cases. This latter complication deserves note because, in our experience, it is seldom explicitly discussed with patients preoperatively, even though empirically it is one of the most frequent complications. The reasons for the high prevalence of this complication deserve more study.
Overall, complications were far more likely to occur in cranial procedures than in spinal ones (23.6% cranial vs 11.2% spinal). Even when accounting for other patient risk factors, cranial procedures were still 1.41 times more likely to involve complications than spinal procedures (95% CI 1.27–1.56). The source of this discrepancy is unclear from the data captured within the NSQIP data set. It should be stressed, however, that the NSQIP is only an incomplete reflection of patient characteristics. The nature of diseases bringing one to cranial neurosurgery (for example, glioma, hydrocephalus) are far different from those for spine surgery (for example, herniated intervertebral discs, spinal instability), and additional data on patient characteristics and risk factors not present in the NSQIP database could contribute to the observed difference in complication frequencies.
Many of the predictors of complications were expected and are traditionally considered as high-risk comorbidities from a pathophysiological standpoint (for example, presurgical sepsis, prior stroke, diabetes). Others were somewhat harder to explain. For example, African Americans are significantly more likely to experience complications than whites, with an RR of 1.24 (95% CI 1.10–1.41). This racial/ethnic disparity in surgical complication rates has been described across multiple surgical disciplines; however, the exact source of this bias is unknown.3,7,14,18,21 Although many comorbid conditions were controlled for in our multivariate model, it is possible that some untracked characteristics in African Americans make them more vulnerable to complications (for example, disease type or severity) or that socioeconomic factors (for example, access to care) are leading to more complications. In either case, we must seek the cause of this disparity to improve the health of our patients.
Several limitations to our study are worth noting. The NSQIP database is a well-curated set of records in which trained analysts enter every complication using strict criteria. Nonetheless, it has several inherent limitations, especially as regards neurosurgical patients. First, the 30-day analysis of complications that is standardized across all sites from which data are gathered is a major strength of the NSQIP database; however, a number of complications unique to the neurosurgical population may not be accurately represented. For example, delayed hemorrhage from an incompletely treated vascular malformation would not be recorded, and there would be no distinction between a postoperative neurological deficit that is permanent versus one that resolves over time. Tracked complications are, by necessity, general and cover a wide range of categories: incision related (both infection and dehiscence), pulmonary (pneumonia, failure to wean from ventilator, and so forth), urinary tract infections, strokes, cardiac arrest or myocardial infarction, renal failure, bleeding, peripheral nerve injury, deep venous thrombosis or pulmonary embolism, sepsis, and reoperation. The NSQIP data set is used by many surgical disciplines, and complications unique to neurosurgery and related disciplines are irrelevant to the majority of the database's users (for example, CSF leaks, pseudomeningoceles). Similarly, counting complications by using broad measures may not reflect the nuances encountered in neurosurgical care. For example, rates of failure to wean from a ventilator within 48 hours probably included patients with planned postoperative mechanical ventilation to manage seizures or to control intracranial pressure. In addition, preoperative morbidity variables are similarly generalized and broad. The categories include race, sex, history of tobacco or alcohol use, history of bleeding disorders, transfusions, chemo- or radiotherapy, sepsis, diabetes, disseminated malignancy, and a host of common pulmonary, cardiac, and renal conditions. Variables more relevant to neurosurgery included an altered mental status, history of stroke or transient ischemic attack, paralysis (hemiplegia, paraplegia, quadriplegia), or a CNS tumor. However, the presence of a CNS tumor is a “yes/no” field in the NSQIP database, which ignores distinctions between benign tumors such as meningiomas versus aggressive tumors such as glioblastomas. The database could have even greater value for the neurosurgical community if it included complications that are unique to, but represent a significant burden, to neurosurgical patients, such as postoperative seizures or meningitis. In the same regard, to facilitate more refined risk adjustment in the neurosurgical population, we must collect risk adjustment variables long suspected to be predictors of complications in neurosurgery such as history of prior craniotomies or cranial/spinal radiation or radiosurgery.
These issues highlight the need for neurosurgery-specific databases, which will track relevant complications and comorbidities. Some exciting progress has been made on such databases, particularly with the development of the National Neurosurgery Quality and Outcomes Database (N2QOD), sponsored by the NeuroPoint Alliance and managed by Vanderbilt's Institute for Medicine and Public Health.1,2,13 Neurosurgery-specific databases will go a long way in addressing these concerns, and current efforts to prospectively collect neurosurgery-specific data are important steps toward optimizing the care of patients.
Conclusions
Complications endanger patients and deplete scarce health care resources. To defend against complications, we attempted, first, to understand their prevalence and, second, to identify factors predictive of their occurrence. Analyzing data from more than 38,000 neurosurgical patients in the NSQIP database, we found that the most frequent complications were blood loss requiring transfusion and reoperation within 30 days, followed by failure to wean from mechanical ventilation postoperatively. We also found that complications were more frequent in cranial than in spinal procedures and more likely to occur in African Americans than in whites. The sources of these disparities remain to be determined, and further research with particular attention to complications unique to neurosurgery is clearly necessary. Once we define the most costly and harmful complications, we can begin to rationally direct our quality improvement efforts.
Acknowledgments
The ACS NSQIP and the hospitals participating in the ACS NSQIP are the sources of the data used herein; however, they have not verified and are not responsible for the statistical validity of our data analysis or derived conclusions.
We thank Rita Mistry, M.P.H., for her input on the manuscript and her expertise.
Disclosure
Dr. Parsa was supported in part by the Reza and Georgianna Khatib Endowed Chair in Skull Base Tumor Surgery. This work was supported in part by a Socioeconomic Fellowship from the Congress of Neurological Surgeons (J.R.). The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.
Author contributions to the study and manuscript preparation include the following. Conception and design: all authors. Acquisition of data: all authors. Analysis and interpretation of data: all authors. Drafting the article: all authors. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Parsa. Study supervision: Parsa.
References
- 1↑
Asher AL, , McCormick PC, , Selden NR, , Ghogawala Z, & McGirt MJ: The National Neurosurgery Quality and Outcomes Database and NeuroPoint Alliance: rationale, development, and implementation. Neurosurg Focus 34:1 E2, 2013
- 2↑
Asher AL, , McGirt MJ, , Glassman SD, , Groman R, , Resnick DK, & Mehrlich M, et al.: Regulatory considerations for prospective patient care registries: lessons learned from the National Neurosurgery Quality and Outcomes Database. Neurosurg Focus 34:1 E5, 2013
- 3↑
Blum MA, , Singh JA, , Lee GC, , Richardson D, , Chen W, & Ibrahim SA: Patient race and surgical outcomes after total knee arthroplasty: an analysis of a large regional database. Arthritis Care Res (Hoboken) 65:414–420, 2013
- 4↑
Dindo D, , Demartines N, & Clavien PA: Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 240:205–213, 2004
- 5↑
Emanuel E, , Tanden N, , Altman S, , Armstrong S, , Berwick D, & de Brantes F, et al.: A systemic approach to containing health care spending. N Engl J Med 367:949–954, 2012
- 6↑
Glickman SW, , Schulman KA, , Peterson ED, , Hocker MB, & Cairns CB: Evidence-based perspectives on pay for performance and quality of patient care and outcomes in emergency medicine. Ann Emerg Med 51:622–631, 2008
- 7↑
Ibrahim SA, , Stone RA, , Han X, , Cohen P, , Fine MJ, & Henderson WG, et al.: Racial/ethnic differences in surgical outcomes in veterans following knee or hip arthroplasty. Arthritis Rheum 52:3143–3151, 2005
- 8↑
Ingraham AM, , Richards KE, , Hall BL, & Ko CY: Quality improvement in surgery: the American College of Surgeons National Surgical Quality Improvement Program approach. Adv Surg 44:251–267, 2010
- 9↑
Kaafarani HM, , Borzecki AM, , Itani KM, , Loveland S, , Mull HJ, & Hickson K, et al.: Validity of selected patient safety indicators: opportunities and concerns. J Am Coll Surg 212:924–934, 2011
- 10↑
Khuri SF, , Henderson WG, , Daley J, , Jonasson O, , Jones RS, & Campbell DA Jr, et al.: Successful implementation of the Department of Veterans Affairs' National Surgical Quality Improvement Program in the private sector: the Patient Safety in Surgery Study. Ann Surg 248:329–336, 2008
- 11↑
Koch CG, , Li L, , Hixson E, , Tang A, , Phillips S, & Henderson JM: What are the real rates of postoperative complications: elucidating inconsistencies between administrative and clinical data sources. J Am Coll Surg 214:798–805, 2012
- 12↑
Latimer K, , Pendleton C, , Olivi A, , Cohen-Gadol AA, , Brem H, & Quiñones-Hinojosa A: Harvey Cushing's open and thorough documentation of surgical mishaps at the dawn of neurologic surgery. Arch Surg 146:226–232, 2011
- 13↑
McGirt MJ, , Speroff T, , Dittus RS, , Harrell FE Jr, & Asher AL: The National Neurosurgery Quality and Outcomes Database (N2QOD): general overview and pilot-year project description. Neurosurg Focus 34:1 E6, 2013
- 14↑
Mukherjee D, , Patil CG, , Todnem N, , Ugiliweneza B, , Nuño M, & Kinsman M, et al.: Racial disparities in Medicaid patients after brain tumor surgery. J Clin Neurosci 20:57–61, 2013
- 15↑
Nelson AR: Pay-for-performance programs: ethical questions and unintended consequences. J Fam Pract 56:10 Suppl A 16A–18A, 2007
- 17↑
Romano PS, , Geppert JJ, , Davies S, , Miller MR, , Elixhauser A, & McDonald KM: A national profile of patient safety in U.S. hospitals. Health Aff (Millwood) 22:154–166, 2003
- 18↑
Scarborough JE, , Bennett KM, & Pappas TN: Racial disparities in outcomes after appendectomy for acute appendicitis. Am J Surg 204:11–17, 2012
- 19↑
Shiloach M, , Frencher SK Jr, , Steeger JE, , Rowell KS, , Bartzokis K, & Tomeh MG, et al.: Toward robust information: data quality and inter-rater reliability in the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg 210:6–16, 2010
- 20↑
Song Z, , Cutler DM, & Chernew ME: Potential consequences of reforming Medicare into a competitive bidding system. JAMA 308:459–460, 2012
- 21↑
Vogel TR: Update and review of racial disparities in sepsis. Surg Infect (Larchmt) 13:203–208, 2012
- 22↑
Woolhandler S, , Ariely D, & Himmelstein DU: Why pay for performance may be incompatible with quality improvement. BMJ 345:e5015, 2012