Key Points
Question
Is there an association between hospital perioperative quality (measured by failure to rescue or perioperative mortality) and long-term survival after noncardiac surgery?
Findings
In this national cohort study of 654 093 US veterans who underwent noncardiac surgery at 98 hospitals, there was a dose-dependent association between care at hospitals with higher failure to rescue or mortality rates and long-term risk of death irrespective of the occurrence of postoperative complications.
Meaning
Given this study’s findings of hospital-level variation in postoperative long-term mortality regardless of the occurrence of complications, future studies are needed to better understand the institutional differences in care processes that are associated with variable longer-term survival.
This cohort study uses data from the Veterans Affairs Surgical Quality Improvement database to assess whether an association exists between hospital perioperative quality and long-term survival after noncardiac surgery.
Abstract
Importance
There is known variation in perioperative mortality rates across hospitals. However, the extent to which this variation is associated with hospital-level differences in longer-term survival has not been characterized.
Objective
To evaluate the association between hospital perioperative quality and long-term survival after noncardiac surgery.
Design, Setting, and Participants
This national cohort study included 654 093 US veterans who underwent noncardiac surgery at 98 hospitals using data from the Veterans Affairs Surgical Quality Improvement Program from January 1, 2011, to December 31, 2016. Data were analyzed between January 1 and November 1, 2021.
Exposures
Hospitals were stratified separately into quintiles of reliability-adjusted failure to rescue (FTR) and mortality rates. Patients were further categorized as having a complicated or uncomplicated postoperative course.
Main Outcomes and Measures
The association between hospital FTR or mortality performance quintile (with quintile 1 representing low FTR or mortality and quintile 5 representing very high FTR or mortality) and overall risk of death was evaluated separately using multivariable shared frailty modeling among patients with a complicated and uncomplicated postoperative course.
Results
For the overall cohort of 654 093 patients, the mean (SD) age was 61.1 (13.2) years; 597 515 (91.4%) were men and 56 578 (8.7%) were women; 111 077 (17.0%) were Black, 5953 (0.9%) were Native American, 467 969 (71.5%) were White, 42 219 (6.5%) were missing a racial category, and 26 875 (4.1%) were of another race; and 37 538 (5.7%) were Hispanic. Hospital-level 5-year survival for patients with a complicated course ranged from 42.7% (95% CI, 38.1%-46.9%) to 82.4% (95% CI, 59.0%-93.2%) and from 76.2% (95% CI, 74.4%-78.0%) to 95.2% (95% CI, 92.5%-97.7%) for patients with an uncomplicated course. Overall, 47 (48.0%) and 83 (84.7%) of 98 hospitals were either in the same or within 1 performance quintile for FTR and mortality, respectively. Among patients who had a postoperative complication, there was a dose-dependent association between care at hospitals with higher FTR rates and risk of death (compared with quintile 1: quintile 2 hazard ratio [HR], 1.05 [95% CI, 0.99-1.12]; quintile 3 HR, 1.17 [95% CI, 1.10-1.26]; quintile 4 HR, 1.30 [95% CI, 1.22-1.38]; and quintile 5 HR, 1.34 [95% CI, 1.22-1.43]). Similarly, increasing hospital FTR rates were associated with increasing risk of death among patients without complications (compared with quintile 1: quintile 2 HR, 1.07 [95% CI, 1.01-1.14]; quintile 3 HR, 1.10 [95% CI, 1.04-1.16]; quintile 4 HR, 1.15 [95% CI, 1.09-1.21]; and quintile 5 HR, 1.10 [95% CI, 1.05-1.19]). These findings were similar across hospital mortality quintiles for patients with complicated and uncomplicated courses.
Conclusions and Relevance
The findings of this cohort study suggest that the structures, processes, and systems of care that underlie the association between FTR and worse short-term outcomes may also have an influence on long-term survival through a pathway other than rescue from complications. A better understanding of these differences could lead to strategies that address variation in both perioperative and longer-term outcomes.
Introduction
Nearly 20 million operations are performed in the US every year.1 However, the quality and safety of surgical care varies widely, with nearly 2-fold variation in risk-adjusted perioperative mortality across hospitals.2,3 Although many factors likely play a role in this outcome variability, short-term variation in hospital perioperative mortality has primarily been attributed to institutional differences in the management, rather than the prevention, of complications.2,4,5 Many structural and microsystem factors (eg, increased nurse-patient ratios, improved communication between care team members, and institutional safety culture) have been associated with more timely identification and treatment of complications.6,7,8,9 These factors are believed to directly influence more proximate postoperative outcomes, but it is possible that they could also affect patients’ longer-term survival.
Although there are numerous studies detailing hospital perioperative outcome variation and associated factors, much less is known about the extent to which this variation might be associated with differences in longer-term postoperative survival. At the patient level, recent data demonstrate that an increasing number of postoperative complications is associated with a dose-dependent association with worsening long-term survival.10 Although postoperative complications have a lasting, detrimental influence on the long-term survival of surgical patients, the potential effect of variation in the quality of care provided at the hospital where an operation is performed has not been characterized. In particular, it is unclear whether differences in hospital perioperative care systems are associated with worse long-term survival. Additionally, assessing whether this variation only affects patients who have postoperative complications (who are known to have worse perioperative and longer-term outcomes) or whether it affects the surgical care of all patients (regardless of whether postoperative complications occur) would represent important information.
A better understanding of whether surgical care at hospitals that deliver high- or low-quality perioperative care is associated with longer-term outcomes would be important for patients, hospital systems, and quality improvement initiatives alike. This type of variation could indicate hospital-specific care processes that influence not only patient perioperative outcomes but also longer-term survival. As such, the goals of this analysis are to (1) describe variation in long-term survival after noncardiac surgery among patients with and without postoperative complications and (2) evaluate the association between hospital perioperative quality (represented by adjusted failure to rescue [FTR] or perioperative mortality) and long-term postoperative survival.
Methods
Study Design
This was a national cohort study of patients undergoing non-cardiac surgery between January 1, 2011, and December 31, 2016, using the Veterans Affairs (VA) Surgical Quality Improvement (VASQIP) database. Participation in VASQIP is mandatory for all VA hospitals, and this program has provided continuous quality monitoring and oversight since 1991.11 Detailed information from a representative sample of noncardiac surgery cases performed at all VA hospitals is prospectively collected by trained nurse abstractors using well-defined and established protocols, and a full data dictionary is provided as supplemental material.12,13 This study was approved by the Baylor College of Medicine’s institutional review board and the Michael E. DeBakey VA Medical Center Research and Development Committee, which deemed the study to be exempt from review because deidentified VASQIP data were used. This study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Study Population
Patients undergoing noncardiac surgery during the study period were identified in VASQIP (n = 702 496). As is standard for analyses using VASQIP data, for patients who underwent more than 1 operation within a 30-day period, only information from the first procedure was used. Patients without a recorded date of operation or an unknown length of follow-up were excluded. Patients with procedures identified as transplant, cardiac, pain, or ophthalmologic were also excluded. Finally, we excluded patients treated at 29 ambulatory surgical centers and 12 facilities with a standard complexity designation. Standard complexity facilities were excluded because between 80% and approximately 100% of these patients underwent ambulatory procedures.
Participant race and ethnicity were self-reported. These variables were included in the analysis to provide a more complete understanding of baseline cultural and environmental influences, if present.
Variables
The primary exposures of interest were each treating hospital’s reliability-adjusted FTR or mortality rate with hospitals stratified into quintiles (both were calculated for each hospital). Thirty-day mortality and FTR, defined as 30-day death after any VASQIP-defined postoperative complication, served as indicators of hospital perioperative quality. Reliability adjustment is an empirical bayesian approach that shrinks each hospital’s outcome rate toward the population mean to a degree that is inversely proportional to its surgical volume.14 To calculate each hospital’s reliability-adjusted FTR or mortality rate, we first created a multivariable logistic regression model evaluating FTR or mortality as the outcome adjusting for the occurrence of multiple complications, patient frailty measured using the Risk Analysis Index (RAI), a diagnosis of malignancy as the operative indication, procedural complexity (assigned by the VA National Surgery Office to each Current Procedural Terminology code based on the complexity and level of risk associated with the operation), surgical specialty, and facility complexity (based on hospital resource availability and the breadth and complexity of operations performed). The RAI is a validated tool for measuring patient frailty that incorporates information about each patient’s age, sex, functional status, and comorbid conditions.15,16 Because the RAI is a composite variable containing information from covariates that would typically be included in multivariable modeling, as in prior work, it was used alone to avoid issues related to collinearity.16,17 The log(odds) of FTR or mortality were obtained using postestimation prediction. Hierarchical regression (patients clustered within hospitals) including a random effect for the hospital was then used to derive the shrinkage factor, which was applied to create the hospital-level reliability-adjusted estimates. Hospitals were then stratified into quintiles of reliability-adjusted FTR or mortality rates, respectively (very low FTR or mortality [ie, best performance in quintile 1] to very high FTR or mortality [ie, worst performance in quintile 5]).
Patients were categorized as having had a complicated or uncomplicated postoperative course. A patient with an uncomplicated course did not have any postoperative complications. VASQIP-defined 30-day complications include myocardial infarction, cardiac arrest, cerebrovascular event, coma, failure to wean from mechanical ventilation within 48 hours, reintubation, pneumonia, kidney insufficiency, kidney failure, urinary tract infection, superficial surgical site infection, deep wound infection, organ space infection, fascial dehiscence, pulmonary embolism, deep vein thrombosis, bleeding requiring a transfusion of 4 or more units of packed red blood cells, sepsis, vascular graft failure, and clostridium difficile colitis. The occurrence of any of these complications qualified the patient as having had a complication. Major complications were defined similar to complications but excluded patients who only had superficial surgical site infection, urinary tract infection, kidney insufficiency, or isolated deep vein thrombosis.5,17,18 Secondary complications were identified among patients who had more than one complication and constituted those occurring after the initial complication (eg, for a patient who had 3 complications, the second and third would constitute the secondary complications). A full list of the VASQIP variables used in this analysis are available in eTable 1 in the Supplement.
Statistical Analysis
The primary outcome was overall risk of death (ie, death from any cause). Long-term follow-up was available through the VA’s Vital Status file. The Kaplan-Meier method was used to estimate survival rates at each hospital. Multivariable Cox shared frailty modeling was used to evaluate the association between the treating hospital’s mortality and FTR performance quintile and overall risk of death. Shared frailty modeling was selected to account for the associated nature of the data; patients treated at the same hospital were likely to have been exposed to similar perioperative care and thus have potentially similar outcomes. Model covariates were similar to those used to perform reliability adjustment. Separate models were used to evaluate patients with a complicated and an uncomplicated postoperative course given notable differences in the long-term outcome of patients who have postoperative complications.10 A case-complete approach was used, as only 0.6% of patients had any missing covariate data. A P value of <.05 was considered significant, and all tests were 2-sided.
To evaluate the robustness of our findings, we conducted several sensitivity analyses. First, we stratified patients based on the type of surgical procedure they underwent: outpatient (n = 349 078), low-risk inpatient (procedure associated with an inpatient hospital stay and a perioperative mortality rate <1%; n = 185 025), or high-risk inpatient surgery (n = 119 990).10 Second, we conducted an analysis including only patients considered “robust” based on the RAI (RAI <20; n = 526 703).19,20 Third, we excluded patients with a principal diagnosis of malignancy as the indication for the procedure (n = 81 923 excluded). Fourth, we applied a 90-day landmark to focus only on patients who survived beyond the perioperative period (n = 12 980 excluded). Fifth, we restricted our analysis to include only patients treated at hospitals with the highest complexity designation to control for implicit heterogeneity in facility resources that may have allowed for a more robust response to postoperative complications (n = 526 703 veterans treated at 70 hospitals). Finally, because race and ethnicity information was missing or omitted for 8.4% (n = 63 582) of the primary cohort, we conducted a sensitivity analysis including it as a model covariate in our main models (ie, overall complicated and uncomplicated patients). The main study findings were unchanged by inclusion of race and ethnicity. All analyses were conducted between January 1 and November 1, 2021, using Stata, version 15.1 (StataCorp LLC).
Results
The final cohort (Figure 1) included 654 093 patients treated at 98 hospitals after excluding patients without an operative date or follow-up information (n = 281), those who underwent transplant, cardiac, pain, or ophthalmologic procedures (n = 3213), and patients treated at 29 ambulatory surgical centers (n = 20 114) or 12 standard complexity facilities (n = 24 795). Among the 98 hospitals included in the cohort, median hospital mortality and FTR rates were 1.1% (IQR, 0.9%-1.3%) and 8.3% (IQR, 6.8%-9.9%), respectively. Overall, 47 (48.0%) and 83 (84.7%) of 98 hospitals in the cohort were either in the same quintile or within 1 quintile for adjusted mortality and FTR, respectively.
Figure 1. Patient Flow Diagram for Inclusion in Final Analytic Cohort and Sensitivity Analyses.
RAI indicates Risk Analysis Index; VASQIP, Veterans Affairs Surgical Quality Improvement Program.
For the overall cohort of 654 093 patients, the mean (SD) age was 61.1 (13.2) years; 597 515 (91.4%) were men and 56 578 (8.7%) were women; 111 077 (17.0%) were Black; 5953 (0.9%) were Native American; 467 969 (71.5%) were White, 42 219 (6.5%) were missing a racial category, and 26 875 (4.1%) were of another race; and 37 538 (5.7%) were Hispanic. Baseline sociodemographic, clinical, and procedural information for the cohort are presented in the Table stratified by the treating hospital’s reliability-adjusted FTR quintile. Differences in demographic and clinical patient characteristics were no greater than 5% across hospitals, except for the proportion of Black and Hispanic patients treated at very low (14.6% [19 747 of 135 265 patients] and 4.3% [5773 of 135 265 patients], respectively) and very high (21.4% [28 584 of 133 768 patients] and 9.7% [13 020 of 133 768 patients], respectively) FTR hospitals. Intermediate or complex procedural complexity varied by less than 5% across hospital quintiles (41.5% [56 137 of 135 265 patients], 44.2% [56 781 of 128 704 patients], 43.7% [56 285 of 128 729 patients], 42.8% [54 654 of 127 627 patients], and 39.8% [53 202 of 133 768 patients] from very low FTR to very high FTR quintiles). eTable 2 in the Supplement provides sociodemographic, clinical, and procedural information across adjusted mortality quintiles. eTable 3 in the Supplement provides the most commonly performed procedures according to hospital FTR quintile.
Table. Sociodemographic, Clinical, and Procedural Characteristics by Reliability-Adjusted Hospital FTR Quintiles.
| Characteristics | Adjusted FTR quintiles, No. (%)a | ||||
|---|---|---|---|---|---|
| Very low (n = 135 265; H = 20) | Low (n = 128 704; H = 20) | Medium (n = 128 729; H = 18) | High (n = 127 627; H = 20) | Very high (n = 133 768; H = 20) | |
| Sociodemographic characteristics | |||||
| Age, y | |||||
| 18-44 | 15 351 (11.4) | 15 633 (12.2) | 14 585 (11.3) | 14 791 (11.6) | 15 650 (11.7) |
| 45-54 | 17 337 (12.8) | 17 452 (13.6) | 16 893 (13.1) | 17 364 (16.6) | 18 818 (14.1) |
| 55-64 | 72 465 (53.6) | 67 631 (52.6) | 68 454 (53.2) | 66 830 (52.4) | 70 695 (52.9) |
| 65-74 | 12 944 (9.6) | 11 890 (9.2) | 12 156 (9.8) | 12 156 (9.5) | 12 173 (9.1) |
| ≥75 | 17 168 (12.7) | 16 098 (12.5) | 16 153 (12.6) | 16 486 (12.9) | 16 432 (12.3) |
| Sex | |||||
| Female | 11 217 (8.3) | 11 895 (9.2) | 11 159 (8.7) | 10 772 (8.4) | 11 535 (8.6) |
| Male | 124 048 (91.7) | 116 809 (90.8) | 117 570 (91.3) | 116 855 (91.6) | 122 233 (91.4) |
| Race | |||||
| Black | 19 747 (14.6) | 19 212 (14.9) | 20 245 (15.7) | 23 289 (18.3) | 28 584 (21.4) |
| Native American | 1248 (0.9) | 919 (0.7) | 1373 (1.1) | 1145 (0.9) | 1268 (1.0) |
| White | 95 251 (70.4) | 99 059 (76.9) | 91 479 (71.1) | 88 746 (69.5) | 93 434 (69.9) |
| Otherb | 11 042 (9.1) | 4907 (3.9) | 7668 (5.9) | 7619 (5.9) | 5478 (4.0) |
| Missing | 7977 (5.9) | 4607 (3.6) | 7964 (6.2) | 6828 (5.4) | 5004 (3.7) |
| Hispanic ethnicity | 5773 (4.3) | 3277 (2.6) | 7952 (6.2) | 7516 (5.9) | 13 020 (9.7) |
| Missing | 3191 (2.4) | 1 528 (1.2) | 4206 (2.0) | 2346 (2.7) | 2125 (1.6) |
| ASA class | |||||
| 1 | 2948 (2.2) | 3422 (2.7) | 2380 (1.9) | 2677 (2.1) | 3044 (2.3) |
| 2 | 35 905 (26.5) | 31 792 (24.7) | 28 801 (22.4) | 29 918 (23.4) | 38 306 (28.6) |
| 3 | 84 834 (62.7) | 81 508 (63.3) | 85 792 (66.7) | 81 881 (64.2) | 79 947 (59.8) |
| 4 | 11 302 (8.4) | 11 514 (9.0) | 11 496 (8.9) | 12 806 (10.0) | 11 797 (8.8) |
| Missing | 158 (0.1) | 295 (0.2) | 103 (0.1) | 124 (0.1) | 390 (0.3) |
| Current smoker | 43 783 (32.4) | 87 269 (32.2) | 88 043 (31.6) | 86 825 (31.9) | 93 318 (30.2) |
| Robust RAI status | 56 812 (42.0) | 57 275 (44.5) | 55 447 (43.1) | 55 443 (43.4) | 61 409 (45.9) |
| Missing | 318 (0.2) | 70 (0.1) | 29 (0.0) | 150 (0.1) | 90 (0.1) |
| Preoperative functional status | |||||
| Independent | 124 452 (92.0) | 121 048 (94.1) | 118 556 (92.1) | 119 836 (93.9) | 127 717 (95.5) |
| Partially dependent | 8955 (6.6) | 6135 (4.8) | 8366 (6.5) | 6265 (4.9) | 4671 (3.5) |
| Totally dependent | 1832 (1.4) | 1489 (1.2) | 1792 (1.4) | 1495 (1.2) | 1327 (1.0) |
| Missing | 0 | 0 | 0 | 0 | (0.2) |
| Clinical characteristics | |||||
| Diabetes | |||||
| None or controlled by diet | 105 648 (78.1) | 99 902 (77.6) | 99 363 (77.2) | 97 878 (76.7) | 103 844 (77.6) |
| Oral medication | 16 280 (12.0) | 16 432 (12.8) | 16 462 (12.8) | 16 567 (13.0) | 16 921 (12.7) |
| Daily insulin | 13 336 (9.9) | 12 368 (9.6) | 12 903 (10.0) | 13 181 (10.3) | 13 002 (9.7) |
| Missing | 0 | 0 | 0 | 0 | 0 |
| CHF within 30 d | 119 980 (88.7) | 114 742 (89.2) | 113 721 (88.3) | 113 819 (89.2) | 122 412 (91.5) |
| Dyspnea | 13 852 (10.2) | 13 005 (10.0) | 13 699 (10.6) | 12 522 (9.8) | 10 395 (7.8) |
| None | 1433 (1.1) | 956 (0.7) | 1309 (1.0) | 1286 (1.0) | 961 (0.7) |
| Exertion | 4374 (3.2) | 2783 (2.2) | 2527 (2.0) | 3199 (2.5) | 2364 (1.8) |
| Rest | 1546 (1.1) | 1480 (1.2) | 1556 (1.2) | 1858 (1.5) | 1394 (1.0) |
| >10% Loss of body weight | 119 980 (88.7) | 114 742 (89.2) | 113 721 (88.3) | 113 819 (89.2) | 122 412 (91.5) |
| Dialysis within 2 wk | 13 852 (10.2) | 13 005 (10.0) | 13 699 (10.6) | 12 522 (9.8) | 10 395 (7.8) |
| Acute preoperative kidney failure | 603 (0.5) | 424 (0.3) | 386 (0.3) | 507 (0.4) | 399 (0.3) |
| Cancer diagnosis | 18 441 (13.6) | 15 569 (12.1) | 16 022 (12.5) | 15 461 (12.1) | 18 286 (13.7) |
| Procedural data | |||||
| Emergency operation | 5292 (3.9) | 6382 (5.0) | 5489 (4.3) | 5639 (4.4) | 5351 (4.0) |
| Procedural complexity | |||||
| Standard | 79 128 (58.5) | 71 923 (55.9) | 72 444 (56.3) | 72 973 (57.2) | 80 566 (60.2) |
| Intermediate | 53 439 (39.5) | 54 645 (42.5) | 53 341 (41.4) | 52 234 (40.9) | 50 843 (38.0) |
| Complex | 2698 (2.0) | 2136 (1.7) | 2944 (2.3) | 2420 (1.9) | 2359 (1.8) |
Abbreviations: ASA, American Society of Anesthesiologists; CHF, congestive heart failure; FTR, failure to rescue; H, hospitals; RAI, Risk Analysis Index.
Missing values are provided where present. If not provided, there were no missing values for the listed variable.
Includes Asian (n = 2657 [0.4%]), Native Hawaiian or Pacific Islander (n = 4796 [0.7%]), unknown by patient (n = 9839 [1.5%]), or declined to answer (n = 19 422 [3.0%]).
Overall, 44 191 patients (6.8%) had at least 1 postoperative complication. Figure 2 demonstrates rates of perioperative outcomes stratified by hospital FTR and mortality quintiles. Complication (6.9% and 6.1%), major complication (4.6% and 4.0%), and secondary complication rates (28.6% vs 26.5%, respectively) were similar when comparing very low- to very high-FTR hospitals. Complication, major complication, and secondary complication rates were highest in the third hospital FTR quintile. The FTR rates were significantly different across hospitals (5.8% for very low and 10.8% for very high FTR; trend test, P < .001). Complication (6.4% and 6.6%) and major complication (4.3% and 4.4%) rates were similar when comparing very low to very high mortality hospitals, but secondary complication rates (30.5% vs 26.8%, respectively) differed. Mortality rates were significantly different across hospitals (0.8% for very low and 1.4% for very high mortality hospitals; trend test, P < .001).
Figure 2. Unadjusted Complications, Major Complications, Secondary Complications, and Failure to Rescue (FTR) or Mortality Rates According to Hospital Reliability–Adjusted FTR and Mortality Quintiles.
Median follow-up for the cohort was 42 months (IQR, 24-60 months). Figure 3 presents 5-year survival estimates for each hospital among patients with a complicated or uncomplicated postoperative course. Overall, 5-year survival for patients with a complicated course ranged from 42.7% (95% CI, 38.1%-46.9%) to 82.4% (95% CI, 59.0%-93.2%). Five-year survival ranged from 76.2% (95% CI, 74.4%-78.0%) to 95.2% (95% CI, 92.5%-97.7%) for patients with an uncomplicated course. Among patients with a complicated course, there was a dose-dependent association between increasing hospital FTR quintile and risk of death (Figure 4A) (compared with quintile 1: quintile 2 hazard ratio [HR], 1.05 [95% CI, 0.99-1.12]; quintile 3 HR, 1.17 [95% CI, 1.10-1.26]; quintile 4 HR, 1.30 [95% CI, 1.22-1.38]; and quintile 5 HR, 1.34 [95% CI, 1.22-1.43]). Among patients without complications, treatment at high FTR hospitals was also associated with increased risk of death (compared with quintile 1: quintile 2 HR, 1.07 [95% CI, 1.01-1.14]; quintile 3 HR, 1.10 [95% CI, 1.04-1.16]; quintile 4 HR, 1.15 [95% CI, 1.09-1.21]; and quintile 5 HR, 1.10 [95% CI, 1.05-1.19]). There was a similar pattern among complicated and uncomplicated patients across hospital mortality quintiles (Figure 4B).
Figure 3. Unadjusted Hospital 5-Year Survival Among Patients With and Without Complications.
Each point represents 1 hospital’s 5-year survival estimate for patients with or without complications. Error bars represent 95% CIs for survival estimates.
Figure 4. Risk of Death According to Hospital Reliability–Adjusted Failure to Rescue (FTR) and Mortality Quintiles Among Patients With and Without Complications.
Cox shared frailty models were adjusted for occurrence of multiple complications (in complicated patient models), patient frailty measured using the Risk Analysis Index, a diagnosis of malignancy as the operative indication, procedural complexity, surgical specialty, and facility complexity.
In the sensitivity analyses for hospital mortality, all findings mirrored those seen in the primary analysis for patients with and without complications. In those for hospital FTR, similar findings were noted among patients with complications. The only exceptions were the outpatient surgery and 90-day landmark analyses. In these groups, the dose-response association was inconsistent for patients treated at high- and very high–FTR hospitals (outpatient: quintile 4 HR, 1.26 [95% CI, 1.12-1.42]; quintile 5 HR, 1.22 [95% CI, 1.06-1.42]; 90-day landmark: quintile 4 HR, 1.18 [95% CI, 1.11-1.25]; quintile 5 HR, 1.10 [95% CI, 1.02-1.19]). Among patients without complications, all sensitivity analyses for hospital FTR were similar to the primary analysis.
Discussion
Characterizing institutional variation in both perioperative and longer-term outcomes is a critical first step toward improving surgical quality. While prior work has described variation in perioperative outcomes, to our knowledge, this cohort study is the first assessment of both hospital-level variation in long-term survival after noncardiac surgery and the association between the quality of perioperative hospital care and long-term survival. In addition to strong concordance between hospital mortality and FTR, we believe this study provides 2 other important findings. First, there is substantial hospital-level variation in patient long-term survival—in particular, among patients who have postoperative complications. Second, the quality of perioperative care provided at each hospital appears to be associated with long-term risk of death, regardless of the presence of complications. In light of prior work suggesting that variation in hospital-level perioperative mortality is attributed to differences in the timely identification and management of complications, it is possible that there is simultaneous institution-level variability in predischarge or postdischarge care systems that affect longer-term survival.
Over the past decade, an increasing number of patients with postoperative complications are surviving beyond the immediate perioperative period.3 As such, rather than surgical quality improvement initiatives focusing exclusively on perioperative outcomes, it may be useful to consider a pivot toward indicators that better capture the quality of intermediate and longer-term care for patients who have postoperative complications. As an example, a previously described conceptual model for the long-term sequelae of a hospitalization for sepsis could serve as an excellent template for this sort of shift—specifically, the components of high-quality sepsis care span from hospitalization to long-term follow-up and include timely source control, postdischarge rehabilitation, screening for new chronic medical conditions, adequate medication reconciliation, and assurance of adequate support systems.21 Given that patients treated at hospitals with poor FTR performance demonstrate higher perioperative mortality rates as well as worse longer-term survival, this could indicate that some hospitals are more effective at instituting these types of practices for surgical patients than others.
An important finding from our study was that the association between treatment at hospitals with poor FTR performance (which specifically measures the institution’s ability to identify and treat patients with complications) and worse long-term survival was present among patients who did not have complications. A potential explanation is that care processes associated with successful complication rescue intersect with those for patients who do not have complications. As an example, perioperative management of diabetes requires efficient physician-system interactions, as either hyper- or hypoglycemia increases the risk and severity of postoperative complications.22 As such, the optimal system has built-in redundancies for routine surveillance and management of perioperative blood glucose that may also be beneficial in the care of all postoperative patients, regardless of their clinical course.23,24 When these redundancies are not present or not optimized, patients may be at risk for complications going unrecognized or medical needs that extend beyond the immediate hospitalization going unattended. Other such examples include opioid prescribing practices, preoperative risk assessment, and routine use of the intensive care unit.25,26,27 Poor management with respect to any of these areas could result in pre- or postdischarge morbidity and an increase in longer-term mortality. Variation in these sorts of everyday practices could contribute to the increased risk of death for patients without complications who receive surgical care at hospitals with low FTR performance.
Although this work was performed on veterans treated in VA hospitals, there are several reasons why these data are important and might generalize to non-VA care settings. First, an important barrier to studying this topic is the lack of available data sources containing reliable information about complications and long-term survival with hospital identifiers. Although administrative and electronic health record data typically provide information about long-term survival, they are less reliable for ascertaining information about perioperative complications.28,29,30 By comparison, data from the American College of Surgeons National Surgical Quality Improvement Program provide highly reliable information about perioperative complications but do not contain long-term follow-up or a hospital identifier. As such, VASQIP is a unique (if not the only) data source for evaluating this issue in a robust, reliable fashion. Second, there is existing precedent for quality initiatives developed in the VA system being implemented in the private sector—VASQIP being a prime example. Owing in large part to the success of VASQIP after it was implemented in the early 1990s, it was later used as a template for the design and implementation of the American College of Surgeons National Surgical Quality Improvement Program, which has been highly successful in the private sector.11,31 Finally, as the nation’s largest integrated health care system, the VA represents a unique entity in which novel approaches to longer-term care of patients with postoperative complications could be developed, implemented, and eventually disseminated to other non-VA hospitals and health systems.
The U-shaped pattern noted for complications, major complications, and secondary complications when hospitals were stratified based on FTR (with the highest morbidity rates noted among middle-quintile hospitals) merits additional consideration. This pattern was not observed across hospital mortality quintiles. This finding suggests a possible inherent limitation in using FTR to evaluate hospital perioperative quality. Specifically, it is possible that hospitals that perform well in preventing complications (ie, hospitals that have a low complication rate and thus a smaller denominator for calculating FTR) may demonstrate higher rates of FTR. In these cases, it may be useful to also consider performance in terms of perioperative mortality to create a more holistic approach to assessing hospital quality that factors in the impact of both morbidity and mortality.
Limitations
These study findings should be interpreted in light of several important limitations. First, VASQIP does not provide detailed information on facility-level characteristics. This information could have been used to assess specific hospital-level features that were associated with better long-term survival. This might also help explain the U-shaped nature of the complication rates observed across adjusted FTR quintiles. It is possible that specific aspects about the clinical environment in middle-tier hospitals placed patients at higher risk for complications. In our cohort, a greater proportion of procedures performed at middle-quintile facilities were more complex and therefore potentially more morbid. Additional facility-level data—such as VA hospital academic affiliation, which is not provided in VASQIP—would have been useful to explore the association between hospital teaching status, complications, and long-term mortality. Similarly, given the limitations of the data set, we could not fully address the confounding influence of socioeconomic or structural factors on patient outcomes. Although future studies are certainly needed to address this important question, our findings were robust to a number of different subgroup and sensitivity analyses (including controlling for race and ethnicity). Furthermore, our study cohort was comprised almost entirely of male veterans. As such, the generalizability of our findings to women and to patients treated outside the VA is unclear. VASQIP does not provide information about how complications are managed; nor does it allow ascertainment of how and when the care team recognized complications. Similarly, VASQIP only provides information on a standardized set of complications that do not include procedure-specific complications. As such, it is possible that the association between uncomplicated patients and hospitals with low FTR performance is explained by residual confounding from complications not coded in the VASQIP database (anastomotic leak, pneumothorax, etc). However, given the consistency in the findings between the primary analysis and our sensitivity analyses, our findings appear robust to varying assumptions about the data and the study cohort. Finally, information about cause-specific mortality is not included in VASQIP. This might have allowed an evaluation of whether there was variation in the causes of postoperative death across institutions.
Conclusions
This cohort study found wide hospital-level variation in long-term outcomes after surgery. Given strong concordance between the findings when stratifying hospitals based on FTR and mortality, we believe this suggests not only that a substantial portion of variation in perioperative mortality is accounted for by FTR, but that longer-term survival may also be affected by the quality of perioperative care. Owing to improvements in complication rescue in recent years, a growing population of patients will need longer-term management of the potential morbidity caused by complications beyond the perioperative period. Furthermore, the presence of hospital-level differences in postoperative long-term mortality regardless of the occurrence of complications suggests future studies are needed to better understand and highlight institutional differences in care processes that are associated with variable longer-term survival.
eTable 1. VASQIP Variables
eTable 2. Sociodemographic, Clinical, and Procedural Characteristics by reliability adjusted hospital mortality Quintiles
eTable 3. Ten most common procedures associated with patients treated at reliability adjusted hospital FTR Quintiles
References
- 1.Steiner CA, Karaca Z, Moore BJ, Imshaug MC, Pickens G. Surgeries in hospital-based ambulatory surgery and hospital inpatient settings, 2014: statistical brief #223. In: Healthcare Cost and Utilization Project (HCUP) Statistical Briefs. Agency for Healthcare Research and Quality: 2017. [PubMed] [Google Scholar]
- 2.Ghaferi AA, Birkmeyer JD, Dimick JB. Complications, failure to rescue, and mortality with major inpatient surgery in medicare patients. Ann Surg. 2009;250(6):1029-1034. doi: 10.1097/SLA.0b013e3181bef697 [DOI] [PubMed] [Google Scholar]
- 3.Fry BT, Smith ME, Thumma JR, Ghaferi AA, Dimick JB. Ten-year trends in surgical mortality, complications, and failure to rescue in Medicare beneficiaries. Ann Surg. 2020;271(5):855-861. doi: 10.1097/SLA.0000000000003193 [DOI] [PubMed] [Google Scholar]
- 4.Ghaferi AA, Dimick JB. Variation in mortality after high-risk cancer surgery: failure to rescue. Surg Oncol Clin N Am. 2012;21(3):389-395, vii. doi: 10.1016/j.soc.2012.03.006 [DOI] [PubMed] [Google Scholar]
- 5.Massarweh NN, Anaya DA, Kougias P, Bakaeen FG, Awad SS, Berger DH. Variation and impact of multiple complications on failure to rescue after inpatient surgery. Ann Surg. 2017;266(1):59-65. doi: 10.1097/SLA.0000000000001917 [DOI] [PubMed] [Google Scholar]
- 6.Sheetz KH, Dimick JB, Ghaferi AA. Impact of hospital characteristics on failure to rescue following major surgery. Ann Surg. 2016;263(4):692-697. doi: 10.1097/SLA.0000000000001414 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ghaferi AA, Birkmeyer JD, Dimick JB. Hospital volume and failure to rescue with high-risk surgery. Med Care. 2011;49(12):1076-1081. doi: 10.1097/MLR.0b013e3182329b97 [DOI] [PubMed] [Google Scholar]
- 8.Ghaferi AA, Osborne NH, Birkmeyer JD, Dimick JB. Hospital characteristics associated with failure to rescue from complications after pancreatectomy. J Am Coll Surg. 2010;211(3):325-330. doi: 10.1016/j.jamcollsurg.2010.04.025 [DOI] [PubMed] [Google Scholar]
- 9.Smith ME, Wells EE, Friese CR, Krein SL, Ghaferi AA. Interpersonal and organizational dynamics are key drivers of failure to rescue. Health Aff (Millwood). 2018;37(11):1870-1876. doi: 10.1377/hlthaff.2018.0704 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Portuondo JI, Itani KMF, Massarweh NN. Association between postoperative complications and long-term survival after non-cardiac surgery among veterans. Ann Surg. PUblished online January 15, 2021. doi: 10.1097/SLA.0000000000004749 [DOI] [PubMed] [Google Scholar]
- 11.Khuri SF, Daley J, Henderson W, et al. ; National VA Surgical Quality Improvement Program . The Department of Veterans Affairs’ NSQIP: the first national, validated, outcome-based, risk-adjusted, and peer-controlled program for the measurement and enhancement of the quality of surgical care. Ann Surg. 1998;228(4):491-507. doi: 10.1097/00000658-199810000-00006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Khuri SF, Daley J, Henderson WG. The comparative assessment and improvement of quality of surgical care in the Department of Veterans Affairs. Arch Surg. 2002;137(1):20-27. doi: 10.1001/archsurg.137.1.20 [DOI] [PubMed] [Google Scholar]
- 13.Henderson WG, Daley J. Design and statistical methodology of the National Surgical Quality Improvement Program: why is it what it is? Am J Surg. 2009;198(5)(suppl):S19-S27. doi: 10.1016/j.amjsurg.2009.07.025 [DOI] [PubMed] [Google Scholar]
- 14.Dimick JB, Ghaferi AA, Osborne NH, Ko CY, Hall BL. Reliability adjustment for reporting hospital outcomes with surgery. Ann Surg. 2012;255(4):703-707. doi: 10.1097/SLA.0b013e31824b46ff [DOI] [PubMed] [Google Scholar]
- 15.Arya S, Varley P, Youk A, et al. Recalibration and external validation of the risk analysis index: a surgical frailty assessment tool. Ann Surg. 2020;272(6):996-1005. doi: 10.1097/SLA.0000000000003276 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Varley PR, Borrebach JD, Arya S, et al. Clinical utility of the Risk Analysis Index as a prospective frailty screening tool within a multi-practice, multi-hospital integrated healthcare system. Ann Surg. 2021;274(6):e1230-e1237. doi: 10.1097/SLA.0000000000003808 [DOI] [PubMed] [Google Scholar]
- 17.Shah R, Attwood K, Arya S, et al. Association of frailty with failure to rescue after low-risk and high-risk inpatient surgery. JAMA Surg. 2018;153(5):e180214. doi: 10.1001/jamasurg.2018.0214 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in hospital mortality associated with inpatient surgery. N Engl J Med. 2009;361(14):1368-1375. doi: 10.1056/NEJMsa0903048 [DOI] [PubMed] [Google Scholar]
- 19.George EL, Hall DE, Youk A, et al. Association between patient frailty and postoperative mortality across multiple noncardiac surgical specialties. JAMA Surg. 2021;156(1):e205152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Shinall MC Jr, Arya S, Youk A, et al. Association of preoperative patient frailty and operative stress with postoperative mortality. JAMA Surg. 2020;155(1):e194620. doi: 10.1001/jamasurg.2019.4620 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Prescott HC, Angus DC. Enhancing recovery from sepsis: a review. JAMA. 2018;319(1):62-75. doi: 10.1001/jama.2017.17687 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Jones CE, Graham LA, Morris MS, et al. Association between preoperative hemoglobin A1c levels, postoperative hyperglycemia, and readmissions following gastrointestinal surgery. JAMA Surg. 2017;152(11):1031-1038. doi: 10.1001/jamasurg.2017.2350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rehman H, Akeroyd JM, Ramsey D, et al. Facility-level variation in diabetes and blood pressure control in patients with diabetes: findings from the Veterans Affairs national database. Clin Cardiol. 2017;40(11):1055-1060. doi: 10.1002/clc.22769 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Venkitachalam L, McGuire DK, Gosch K, et al. Temporal trends and hospital variation in the management of severe hyperglycemia among patients with acute myocardial infarction in the United States. Am Heart J. 2013;166(2):315-324 e311. doi: 10.1016/j.ahj.2013.05.012 [DOI] [PubMed] [Google Scholar]
- 25.Hill MV, McMahon ML, Stucke RS, Barth RJ Jr. Wide variation and excessive dosage of opioid prescriptions for common general surgical procedures. Ann Surg. 2017;265(4):709-714. doi: 10.1097/SLA.0000000000001993 [DOI] [PubMed] [Google Scholar]
- 26.Wijeysundera DN, Austin PC, Beattie WS, Hux JE, Laupacis A. Variation in the practice of preoperative medical consultation for major elective noncardiac surgery: a population-based study. Anesthesiology. 2012;116(1):25-34. doi: 10.1097/ALN.0b013e31823cfc03 [DOI] [PubMed] [Google Scholar]
- 27.Gillies MA, Power GS, Harrison DA, et al. Regional variation in critical care provision and outcome after high-risk surgery. Intensive Care Med. 2015;41(10):1809-1816. doi: 10.1007/s00134-015-3980-1 [DOI] [PubMed] [Google Scholar]
- 28.Lawson EH, Louie R, Zingmond DS, et al. A comparison of clinical registry versus administrative claims data for reporting of 30-day surgical complications. Ann Surg. 2012;256(6):973-981. doi: 10.1097/SLA.0b013e31826b4c4f [DOI] [PubMed] [Google Scholar]
- 29.Rosen AK, Itani KM, Cevasco M, et al. Validating the patient safety indicators in the Veterans Health Administration: do they accurately identify true safety events? Med Care. 2012;50(1):74-85. doi: 10.1097/MLR.0b013e3182293edf [DOI] [PubMed] [Google Scholar]
- 30.Mull HJ, Borzecki AM, Loveland S, et al. Detecting adverse events in surgery: comparing events detected by the Veterans Health Administration Surgical Quality Improvement Program and the Patient Safety Indicators. Am J Surg. 2014;207(4):584-595. doi: 10.1016/j.amjsurg.2013.08.031 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Hall BL, Hamilton BH, Richards K, Bilimoria KY, Cohen ME, Ko CY. Does surgical quality improve in the American College of Surgeons National Surgical Quality Improvement Program: an evaluation of all participating hospitals. Ann Surg. 2009;250(3):363-376. doi: 10.1097/SLA.0b013e3181b4148f [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable 1. VASQIP Variables
eTable 2. Sociodemographic, Clinical, and Procedural Characteristics by reliability adjusted hospital mortality Quintiles
eTable 3. Ten most common procedures associated with patients treated at reliability adjusted hospital FTR Quintiles
