Research Letter
Systematic Identification of NSQIP-Defined Postoperative Complications Using the ICD-10
Kenneth Williams, Masami Tabata-Kelly, Yu-Jen Chen, et al
JAMA Surg Published Online: August 12, 2026
doi: 10.1001/jamasurg.2026.3431
The creation of the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) inspired coordinated efforts to improve surgical quality, in part, through the systematic tracking of 18 postoperative complications, and it remains the gold standard.1 Prior studies mapping International Classification of Diseases, Ninth Revision (ICD-9), codes to NSQIP-defined complications facilitated the use of administrative data to assess surgical quality.2,3 Several subsequent studies validated International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), codes to identify individual surgical complications; however, none have validated a comprehensive ICD-10 complication list aligned with NSQIP.4 We conducted a study to systematically develop a comprehensive surgical complication list aligning ICD-9, ICD-10, and NSQIP, to support assessment of surgical complications in administrative data.
Methods
We adapted our study framework from an algorithm published by Columbo et al.5 This study was approved by the Mass General Brigham institutional review board. Given the use of deidentified, population-level data, the requirement for informed consent was waived.
First, we compiled a comprehensive list of ICD-9 codes aligned with NSQIP complications based on prior validation studies (eTable 1 in Supplement 1). We used the Centers for Medicare & Medicaid Services General Equivalence Mapping crosswalk to convert to ICD-10 codes (eTable 2 in Supplement 1). In parallel, 2 reviewers (K.W. and W.S.) manually generated a list of ICD-10 codes aligned with NSQIP complications through a review of the ICD-10 coding schema and literature review. We then compared the crosswalk-derived and manually generated ICD-10 lists. Concordant codes between lists were included, and discrepancies were adjudicated by a reviewer (K.W.).
The finalized ICD-10 list was assessed for reliability using the Healthcare Cost and Utilization Project National Inpatient Samples (2011-2019). We compared distributions of patient characteristics and procedure groups year-to-year to determine if patient populations were consistent over time and compared complication rates across ICD-9 (2011-2014) and ICD-10 (2016-2019) eras to assess for alignment. Mann-Kendall tests were performed to assess the general direction and significance of trends over time.
Results
We identified 262 ICD-9 NSQIP-aligned complication codes, which mapped to 294 candidate ICD-10 codes. Our comparison between the ICD-9/ICD-10 crosswalk-derived list and the manually generated ICD-10 list yielded 181 overlapping codes and 333 discrepancies. After adjudication, the final ICD-10 NSQIP-aligned list included a total of 513 codes, representing 251 more codes than the original ICD-9 NSQIP-aligned list.
Baseline characteristics (age, race and ethnicity, sex, length of stay, admission type, and Charlson Comorbidity Index) across the study period were comparable (Table 1). Surgical procedure rates remained stable across the transition period from ICD-9 to ICD-10 with general surgery/vascular surgery (29.6% and 29.2%, respectively) and orthopedic surgery (28.3% and 29.3%, respectively) comprising the majority of cases. We observed an increase in overall complications across the entire study period from 2011 to 2019 (Table 2). Deep space surgical site infection (SSI), organ SSI, pulmonary embolism, pneumonia, progressive renal insufficiency, and urinary tract infections saw statistically significant decreases while unplanned intubation, acute renal failure, bleeding, cardiac arrest, myocardial infarction, sepsis, and septic shock saw statistically significant increases. The remainder of the complications held stable.
Table 1. Comparison of Patient Characteristics and Surgical Specialty Procedure Rates Using ICD-9 (2011-2014) and ICD-10 (2016-2019) Codes

| Characteristic | No. (%) | |
|---|---|---|
| ICD-9 codes | ICD-10 codes | |
| Unweighted cohort, No. | 7 316 225 | 7 548 291 |
| Weighted cohort, No. | 36 174 301 | 37 741 443 |
| Age, median (IQR), yb | 57.4 (39.3-70.0) | 57.3 (35.8-69.9) |
| Length of stay, median (IQR), db | 2.7 (1.5-5.0) | 2.5 (1.3-4.8) |
| Sexb | ||
| Male | 14 480 758 (40.0) | 14 509 557 (38.4) |
| Female | 21 673 978 (59.9) | 23 222 531 (61.5) |
| Unknown | 19 565 (0.1) | 9355 (0.02) |
| Race and ethnicityb,c | ||
| Asian or Pacific Islander | 826 511 (2.3) | 1 217 594 (3.2) |
| Black | 4 035 216 (11.2) | 4 370 380 (11.6) |
| Hispanic | 3 636 492 (10.1) | 4 228 269 (11.2) |
| Native American | 189 024 (0.5) | 216 770 (0.6) |
| White | 23 967 200 (66.3) | 25 198 172 (66.8) |
| Other | 1 095 404 (3.0) | 1 171 524 (3.1) |
| Unknown | 2 424 454 (6.7) | 1 338 734 (3.5) |
| Admission typeb | ||
| Nonelective admission | 16 775 181 (46.4) | 18 510 705 (49.0) |
| Elective admission | 19 267 125 (53.3) | 19 139 013 (50.7) |
| Unknown | 131 995 (0.4) | 91 725 (0.2) |
| CCIb | ||
| 0 | 19 334 868 (53.4) | 19 521 636 (51.7) |
| 1 | 8 105 043 (22.4) | 7 620 498 (20.2) |
| 2 | 4 089 396 (11.3) | 4 266 979 (11.3) |
| ≥3 | 4 644 994 (12.8) | 6 332 330 (16.8) |
| Surgical specialty | ||
| Cardiothoracicb | 5 323 721 (14.7) | 5 162 480 (13.7) |
| General surgery/vascular surgerya | 10 702 846 (29.6) | 11 007 277 (29.2) |
| Orthopedic surgeryb | 10 245 275 (28.3) | 11 065 598 (29.3) |
| Neurosurgeryb | 3 536 749 (9.8) | 3 132 075 (8.3) |
| Otherd | 10 046 911 (27.8) | 10 586 774 (28.1) |
Table 2. Comparison of Complication Rates Using NSQIP-Aligned ICD-9 (2011-2014) and ICD-10 (2016-2019) Codes

| Complication | No. (%) | P value | |||||||
|---|---|---|---|---|---|---|---|---|---|
| 2011 | 2012 | 2013 | 2014 | 2016 | 2017 | 2018 | 2019 | ||
| Unweighted cohort, No. | 2 014 011 | 1 798 431 | 1 762 921 | 1 740 862 | 1 902 639 | 1 899 360 | 1 875 094 | 1 871 198 | NA |
| Weighted cohort, No. | 9 663 229 | 8 992 155 | 8 814 602 | 8 704 315 | 9 513 191 | 9 496 794 | 9 375 467 | 9 355 991 | NA |
| Weighted complications | |||||||||
| All | 3 034 603 (31.40) | 2 911 810 (32.38) | 2 934 629 (33.29) | 2 981 496 (34.25) | 3 050 163 (32.06) | 3 093 309 (32.57) | 3 156 630 (33.67) | 3 198 540 (34.19) | .009 |
| Acute renal failure | 703 480 (7.28) | 672 930 (7.48) | 707 190 (8.02) | 763 535 (8.77) | 853 094 (8.97) | 883 340 (9.30) | 937 310 (10.00) | 992 510 (10.61) | .002 |
| Bleeding | 1 129 549 (11.69) | 1 107 440 (12.32) | 1 144 110 (12.98) | 1 170 606 (13.45) | 1 230 054 (12.93) | 1 234 679 (13.00) | 1 251 515 (13.35) | 1 265 900 (13.53) | .002 |
| Cardiac arrest – CPR | 49 517 (0.51) | 50 210 (0.56) | 54 820 (0.62) | 61 110 (0.70) | 58 985 (0.62) | 62 350 (0.66) | 62 010 (0.66) | 60 980 (0.65) | .04 |
| Deep space SSI | 125 493 (1.30) | 107 570 (1.20) | 103 325 (1.17) | 100 055 (1.15) | 1730 (0.02) | 1610 (0.02) | 1570 (0.02) | 1540 (0.02) | .001 |
| DVT | 172 572 (1.79) | 159 210 (1.77) | 153 880 (1.75) | 154 990 (1.78) | 120 705 (1.27) | 119 820 (1.26) | 121 750 (1.30) | 122 405 (1.31) | .06 |
| Myocardial infarction | 411 822 (4.26) | 417 865 (4.65) | 414 690 (4.70) | 421 795 (4.85) | 472 440 (4.97) | 485 155 (5.11) | 498 020 (5.31) | 516 985 (5.53) | .002 |
| Mechanical ventilation | 199 004 (2.06) | 299 800 (3.33) | 294 250 (3.34) | 285 820 (3.28) | 305 420 (3.21) | 277 020 (2.92) | 254 885 (2.72) | 240 435 (2.57) | .39 |
| Organ SSI | 125 493 (1.30) | 107 570 (1.20) | 103 325 (1.17) | 100 055 (1.15) | 15 060 (0.16) | 14 015 (0.15) | 13 045 (0.14) | 11 940 (0.13) | .001 |
| Pulmonary embolism | 81 800 (0.85) | 75 285 (0.84) | 72 105 (0.82) | 73 085 (0.84) | 68 245 (0.72) | 67 515 (0.71) | 70 680 (0.75) | 68 650 (0.73) | .04 |
| Pneumonia | 422 343 (4.37) | 374 175 (4.16) | 366 930 (4.16) | 367 015 (4.22) | 267 030 (2.81) | 258 125 (2.72) | 257 155 (2.74) | 251 460 (2.69) | .002 |
| Progressive renal insufficiency | 136 695 (1.41) | 119 745 (1.33) | 109 710 (1.24) | 103 885 (1.19) | 96 200 (1.01) | 88 820 (0.94) | 82 260 (0.88) | 81 550 (0.87) | .001 |
| Sepsis | 392 652 (4.06) | 363 430 (4.04) | 381 595 (4.33) | 415 000 (4.77) | 414 890 (4.36) | 442 305 (4.66) | 478 315 (5.10) | 498 305 (5.33) | .009 |
| Septic shock | 169 125 (1.75) | 175 420 (1.95) | 189 305 (2.15) | 209 190 (2.40) | 187 150 (1.97) | 196 110 (2.07) | 215 130 (2.29) | 231 060 (2.47) | .009 |
| Superficial SSI | 208 226 (2.15) | 184 845 (2.06) | 179 755 (2.04) | 179 530 (2.06) | 343 390 (3.61) | 365 375 (3.85) | 370 460 (3.95) | 325 955 (3.48) | .27 |
| Stroke | 296 508 (3.07) | 275 520 (3.06) | 274 810 (3.12) | 278 810 (3.20) | 130 950 (1.38) | 137 390 (1.45) | 145 235 (1.55) | 148 440 (1.59) | .17 |
| Unplanned intubation | 199 004 (2.06) | 299 800 (3.33) | 294 250 (3.34) | 285 820 (3.28) | 530 315 (5.57) | 538 240 (5.67) | 554 680 (5.92) | 566 695 (6.06) | .009 |
| Urinary tract infection | 582 439 (6.03) | 521 025 (5.79) | 507 155 (5.75) | 501 515 (5.76) | 457 535 (4.81) | 439 370 (4.63) | 416 815 (4.45) | 387 900 (4.15) | .001 |
| Venous thromboembolism | 201 444 (2.08) | 186 060 (2.07) | 179 555 (2.04) | 181 775 (2.09) | 165 515 (1.74) | 165 630 (1.74) | 170 985 (1.82) | 170 140 (1.82) | .06 |
| Wound disruption | 77 866 (0.81) | 69 845 (0.78) | 71 560 (0.81) | 70 875 (0.81) | 61 180 (0.64) | 67 045 (0.71) | 71 570 (0.76) | 73 170 (0.78) | >.99 |
Discussion
To the best of our knowledge, this is the first study to systematically develop an ICD-9 to ICD-10 crosswalk for NSQIP complications using a large administrative dataset. We observed a statistically significant increase in overall complication rates from 2011 to 2019, but this appears to be a secular trend that began before the transition from ICD-9 to ICD-10. Likewise, increasing and decreasing rates of specific complications appear to be secular trends unrelated to changes in ICD coding. This suggests the 2 coding systems identify postoperative complications at comparable rates. Previous work indicates variable sensitivity and specificity of both ICD-9 and ICD-10for identification of comorbid conditions in administrative data, compared with record review, which remains the gold standard.6 While our study did not assess sensitivity of ICD-10 for detection of surgical complications, our findings do indicate that ICD-10–based detection performs as well as ICD-9–based approaches that were previously validated through record review. In this context, this study serves to corroborate the reliability of an ICD-9/ICD-10 crosswalked list of NSQIP aligned codes to assist in the evaluation of large administrative datasets and as a framework for aligning future ICD iterations.