Variations in Surgical Safety According to Affiliation Status With a Top-Ranked Cancer Hospital

Abstract

This database study analyzes the association between hospital affiliation with top-ranked cancer hospitals and surgical safety at affiliate hospitals.

Top-ranked cancer hospitals have increasingly shared their trusted brands with unranked hospitals in the community through hospital affiliations.¹ Public perception in the United States is that the safety of care at hospitals improves after affiliation with a top-ranked hospital, a belief that increases preference for these hospitals.^2,3 The extent to which affiliation with a top-ranked cancer hospital is associated with cancer surgery outcomes is unknown.

Methods

This database study was conducted with approval from the Yale University Institutional Review Board, which deemed that written informed consent was not required because the study does not directly involve human subjects. Hospitals with a brand-sharing affiliation with a top-50 cancer hospital (based on US News and World Report rankings) were evaluated from January 1, 2013, to October 1, 2016 (n = 338 affiliate hospitals). Data were analyzed from July 2018 to June 2019. These unranked affiliates were compared with 2729 unranked hospitals that were not affiliated with a top-ranked cancer hospital (nonaffiliates) using methods described in a previous study of the topic.⁴ Hospital characteristics were abstracted from the American Hospital Association Data & Insights database (www.ahadata.com), and patient attributes and 90-day mortality after complex cancer surgery (nonemergent pulmonary lobectomy, colectomy, gastrectomy, pancreaticoduodenectomy [Whipple], or esophagectomy) were collected from the Centers for Medicare & Medicaid database (MEDPAR).

Adjusted hierarchical logistic regression and difference-in-difference modeling were used to compare 90-day mortality over time at affiliates and nonaffiliates. To test whether similar patients undergoing similar procedures would have a difference in outcome at affiliate and nonaffiliate hospitals, overall and procedure-specific hierarchical logistic regression models were adjusted for patient and procedure covariates (age, sex, race, Elixhauser comorbidities, procedure type, partial vs total resection [colectomy and gastrectomy], type of admission [elective vs urgent]), including an indicator for affiliation status and a random hospital-level effect. To understand the relative independent influence of hospital characteristics that are associated with affiliation, hierarchical logistic regression models were adjusted for patient, procedure, and hospital covariates (age, sex, race, Elixhauser comorbidities, procedure type, partial vs total resection [colectomy and gastrectomy], type of admission, hospital bed size, surgical volume, Commission on Cancer accreditation status, nurse-to-bed ratio, teaching, and hospital ownership status) with a random hospital effect. Two-tailed P values < .05 were considered statistically significant. All analyses were performed using SAS statistical software version 9.4 (SAS Institute). The statistical analysis was performed between July 2018 and June 2019.

Results

In total, 73 680 patients underwent surgery at nonaffiliate hospitals, and 11 464 patients underwent surgery at affiliate hospitals. After adjusting for patient and procedural variables, 90-day mortality after complex cancer surgery was higher at nonaffiliate than affiliate hospitals (OR, 1.13; 95% CI, 1.03-1.23) (P = .01). However, further adjustment for hospital characteristics known to effect surgical safety mitigated the safety advantage of affiliate hospitals (OR, 1.09; 95% CI, 0.99-1.19) (P = .06) (Table).

Table. Risk-adjusted Odds Ratios of 90-day Mortality at Nonaffiliate vs Affiliate Hospitals^a.

Procedure	OR (95% CI)b	P Value
Odds of 90-Day Mortality at Nonaffiliate vs Affiliate Hospitals Adjusted for Patient and Procedure Characteristicsb
Lobectomy	1.17 (0.96-1.42)	.12
Colectomy	1.13 (1.01-1.26)	.03
Gastrectomy	1.06 (0.77-1.46)	.71
Esophagectomy	1.27 (0.86-1.87)	.23
Whipple	1.05 (0.76-1.47)	.80
All procedures	1.13 (1.03-1.23)	.01
Odds of 90-Day Mortality at Nonaffiliate vs Affiliate Hospitals Adjusted for Patient, Procedure, and Hospital Characteristicsc
Lobectomy	1.16 (0.95-1.42)	.14
Colectomy	1.08 (0.96-1.20)	.20
Gastrectomy	1.07 (0.80-1.55)	.52
Esophagectomy	1.24 (0.82-1.86)	.31
Whipple	1.04 (0.74-1.46)	.83
All procedures	1.09 (0.99-1.19)	.06

Abbreviation: OR, odds ratio.

^aAffiliate hospitals, those affiliated with top-ranked cancer hospitals, served as the reference, with an OR greater than 1 favoring affiliates. Ninety-day mortality was assessed in patients older than 65 years who underwent nonemergent pulmonary lobectomy, colectomy, gastrectomy, pancreaticoduodenectomy (Whipple), or esophagectomy for cancer.

^bAdjusted for patient age, sex, race, admission type, year of surgery, Elixhauser comorbidity score. The model includes random hospital effect. Although tumor characteristics are not available in Medicare data, previous studies have shown that the stage distribution of surgically managed cancers is similar across different types and sizes of hospitals.⁵

^cAdjusted for patient age, sex, race, admission type, year of surgery, Elixhauser comorbidity score, bed size, teaching status, Commission on Cancer accreditation, hospital ownership status, annual complex cancer surgical volume, and nurse-to-bed ratio. Model includes random hospital effect.

One hundred forty-four affiliations began during the study period, allowing the affiliates to be analyzed in the year before and after affiliation took place. Several observations were made relating to hospital characteristics associated with safety (Figure, A). Affiliates had a more favorable mix of hospital characteristics than nonaffiliates before the affiliation took place, that is, more beds (median, 234 vs 134), a higher nurse-to-bed ratio (median, 1.4 vs 1.3), and a higher proportion of affiliates were accredited by the Commission on Cancer (74% [n = 107 of 144] vs 35% [n = 978 of 2729]; P < .01). Hospital characteristics improved slightly at affiliates after affiliation (median nurse-to-bed ratio, 1.4 before affiliation vs 1.5 after affiliation; P < .001). Hospital characteristics improved to a similar extent at nonaffiliates during the study period (median nurse-to-bed ratio, 1.3 in 2013 vs 1.4 in 2015; P < .001).

Figure. Hospital Characteristics, 90-Day Mortality, and Surgical Volume .

Attributes in the year prior to affiliation with a top-ranked cancer hospital were compared with those of the year after affiliation (the year of affiliation was excluded; thus, the analysis spanned 3 years). The McNemar test was used for categorical and Wilcoxon signed rank test for continuous variables. Preaffiliation hospital characteristics were compared with nonaffiliate hospital characteristics using the χ² test for categorical and median 2-sample test for continuous variables. Ninety-day mortality and complex cancer surgery case volume was assessed in patients older than 65 years who underwent nonemergent pulmonary lobectomy, colectomy, gastrectomy, pancreaticoduodenectomy (Whipple), or esophagectomy for cancer. Mortality and volume data in the year prior to and the year after affiliation were only available for hospitals that affiliated in 2014 and 2015 (n = 88). CoC indicates Commission on Cancer.

^aDifference between affiliates and nonaffiliates is significant.

^bDifference between before and after group is significant.

Mortality and case volume data were available for 88 affiliates (Figure, B). Ninety-day mortality decreased from 9.8% to 6.3% (P < .001) and surgical volume increased from 8.0 to 9.5 (P < .001) at affiliates after affiliation. However, an adjusted difference-in-difference model demonstrated that although mortality was significantly lower at affiliated than nonaffiliated hospitals throughout the study period (regression coefficient = −0.176; P = .03), mortality in general decreased over time at all hospitals, and this downward slope was not significantly changed by affiliation (mean slope difference, 0.063; P = .71).

Discussion

Ninety-day mortality was lower when surgery was performed at hospitals affiliated with top-ranked cancer hospitals compared with nonaffiliated hospitals. A favorable mix of hospital characteristics associated with safety at affiliate hospitals appeared to contribute to this mortality advantage (the mortality difference between affiliates and nonaffiliates lost significance after adjustment for these hospital characteristics); thus, affiliate status appears to be a marker but not a robust, independent predictor of favorable outcomes. Additionally, these data suggest that favorable hospital attributes and outcomes predated these affiliations. Furthermore, positive changes occurring at hospitals following affiliation similarly occurred at nonaffiliates during this time period. Therefore, we failed to identify a clear effect of affiliation on surgical safety at affiliate hospitals, which is consistent with a previous study of network affiliations.⁶ Interestingly, the current findings specifically contrast the results of our previous survey of the US public, which observed the perception that significant favorable improvements would occur within the first year of affiliating with a prominent cancer hospital.^2,3 Ultimately, it appears that the modest safety advantage of complex cancer surgery at affiliate hospitals may result from top-ranked cancer hospitals selectively affiliating with safer hospitals.