Introduction

As healthcare across the US evolves, there is an increased emphasis on identifying factors that impact quality of care.1 In particular, surgeon- and hospital-related characteristics play a critical role in dictating access to high-quality care and subsequent patient outcomes. Specifically, hospital characteristics such as total and procedure-specific hospital volume, teaching designation, nurse-to-patient ratio, magnet status, occupancy rate and case-mix have been associated with patient outcomes.2,3,4,5,6,7 Similarly, at the level of the treating surgeon, total and procedure-specific volume, as well as fragmented practice have an impact on quality of care.8,9,10 Admission for a surgical procedure is often a high-risk and complex episode of care, and thus the utilization of such metrics can aid in quality improvement initiatives.11 These metrics gain even further significance in the setting of complex pancreatic surgery, which is often accompanied by high rates of postoperative morbidity, mortality and readmission.12,13,14

Multiple tools have been proposed to quantify the overall quality of care processes, as well as patient outcomes during and after a surgical episode. For example, textbook outcome (TO) is one such metric that serves as a composite metric consisting of no postoperative complications, no extended length of stay (LOS), no 90-day readmission, and no 90-day mortality.14 Failure to rescue (FTR) is another metric that signifies preventable postoperative mortality secondary to a complication.15 In fact, FTR has been adopted as a key patient safety indicator (PSI-04) by the Agency for Health Research and Quality (AHRQ).16 Furthermore, nonroutine discharge status has been associated with increased expenditures and worse quality of life, and, in turn, been incorporated into hospital rankings.17

From a patient-centered perspective, postoperative outcomes and overall patient satisfaction are dependent on a combination of the treating surgeon, as well as the treating hospital. However, the impact of receiving treatment by a care team that is subspecialized in a specific domain of surgery remains relatively unexplored. Specifically in hepatopancreatic (HP) surgery, surgeons and hospitals may disproportionately perform either hepatic or pancreatic procedures. As such, we assessed the impact of undergoing surgery by a care team that is subspecialized in pancreatic surgery on patient outcomes such as TO, FTR, discharge status and index admission expenditures.

Methods

Data Source

The 100% Medicare Standard Analytic Files (SAFs) from 2013–2017 were utilized to initially identify patients who underwent either hepatic or pancreatic surgical procedures. The SAFs are collected and maintained by the Centers for Medicare and Medicaid Services; the SAFs consist of individual patient-level data on demographics, diagnoses, interventions and short- and long-term outcomes. Patients were identified using relevant International Classification of Diseases, Ninth and Tenth Editions (ICD-9/10) procedure codes (Supplementary Table 1). Specifically, patients who underwent pancreatic surgery between 2013–2017, were aged 65 years or older, and enrolled in either Medicare parts A or B were included in the final analytic cohort. Furthermore, only patients who underwent treatment by a surgeon and hospital that performed both hepatic and pancreatic surgical procedures during the study period were included. In instances of multiple surgical procedures, only the first surgical procedure was considered in the analytic cohort. Treating surgeons and hospitals were identified based on the unique National Provider Identifier (NPI) number.18 Patients who received health maintenance organization (HMO) payments, had missing data on surgical procedure codes, surgeon or facility NPI, the components of textbook outcome (TO), FTR, discharge status and index admission expenditures were excluded. The Ohio State University institutional review board deemed this study exempt from approval.

Variables and Outcomes of Interest

The primary independent variable of interest was surgery subspecialization index (SSI), which was defined as a composite measure of the degree of surgeon and hospital subspecialization in pancreatic surgical procedures. Although only pancreatic surgery was assessed in the final analytic cohort, SSI was calculated based on both hepatic and pancreatic surgical volume. Specifically, SSI was calculated based on surgeon pancreatic volume, surgeon total hepatopancreatic volume, hospital pancreatic volume and hospital total hepatopancreatic volume:

$$\frac{Surgeon \ pancreatic \ surgical \ volume}{Surgeon \ total \ hepatopancreatic \ surgical \ volume} \times \frac{Hospital \ pancreatic \ surgical \ volume}{Hospital \ total \ hepatopancreatic \ surgical \ volume}$$

SSI was subsequently categorized into tertiles (i.e., low, intermediate, and high). In essence, a high SSI signified that the majority of surgeries performed by the treating surgeon and hospital over the study period were pancreatic, while a low SSI signified that the surgeon and hospital primarily performed hepatic procedures.

The primary dependent variables of interest included metrics of care processes and postoperative outcomes including TO and its individual component metrics, as well as FTR, discharge to home and index admission expenditures. TO was defined as the absence of any postoperative complications, extended length of stay (LOS) (> 75th percentile), 90-day readmission, and 90-day mortality.14 FTR was defined as mortality secondary to a postoperative complication.6 Discharge status was categorized as discharge to home versus discharge in any other capacity or to any other facility.17 Index admission expenditure in US dollar amounts was determined and treated as a continuous variable.19 Other patient-level sociodemographic and clinical characteristics such as age, gender, race/ethnicity (White versus minority), Charlson Comorbidity Index (CCI) and admission type (elective versus non-elective) were also recorded. Treating hospitals were categorized according to teaching status, as well as hospital volume for stratified analyses (high volume: > 50th percentile versus low volume: ≤ 50th percentile).

Statistical Analysis

Continuous variables were reported as median and interquartile range (IQR) and compared using Kruskal–Wallis H tests, while categorical variables were reported as frequencies and percentages and compared with chi-square tests. The distribution of SSI categories relative to surgeon- and hospital-specific pancreatic subspecialization were illustrated using kernel density plots. Mixed effects multivariable logistic regression with a random effect for treating hospital was utilized to assess the association of SSI with TO and its components, FTR and discharge to home. The association of SSI with index admission expenditure was estimated by utilizing gamma regression with a log link and interpreted as relative differences. Additional models were stratified by hospital volume status to assess the impact of SSI among both low- and high-volume hospitals. All models were adjusted for age, gender, race/ethnicity, CCI, hospital teaching status and type of admission, in line with prior literature.20 Predicted margins were used to estimate and illustrate adjusted probabilities of the events of interest. All tests were two-sided, and the level of significance was set at α = 0.05. Statistical analyses were conducted in STATA, v17 (College Station, TX; StataCorp LLC).

Results

Patient, Surgeon, and Hospital Characteristics

Among 19,625 patients who underwent pancreatic surgery, median age was 72.0 years (IQR 68.0–77.0). The majority of patients was male (n = 10,324, 52.6%) and identified as White (n = 17,537, 89.4%). Overall, patients underwent surgery by 1,416 unique surgeons at 677 unique hospitals. Most patients underwent surgery by a care team characterized by a high SSI (Low SSI: n = 5,428, 27.7% vs. Intermediate SSI: n = 6,801, 34.7% vs. High SSI: n = 7,396, 37.7%) (Fig. 1). Notably, patients who underwent surgery by a high SSI care team were older (Low SSI: 72.0 years, IQR 68.0–77.0 vs. Intermediate SSI: 72.0 years, IQR 68.0–77.0 vs. High SSI: 73.0, IQR 69.0–77.0), and more likely to be White (Low SSI: n = 4,798, 88.4% vs. Intermediate SSI: n = 6,081, 89.4% vs. High SSI: n = 6,658, 90.0%), as well as to receive treatment at a teaching hospital (Low SSI: n = 3,986, 73.1% vs. Intermediate SSI: n = 5,190, 76.3% vs. High SSI: n = 6,092, 82.4%) (all p < 0.05). In contrast, patients who received surgery by a low SSI care team were more likely to have a greater comorbidity burden (CCI) (Low SSI: 3.0, IQR 2.0–8.0 vs. Intermediate SSI: 2.0, IQR 2.0–8.0 vs. High SSI: 2.0, IQR 2.0–8.0) (p < 0.001) (Table 1).

Fig. 1
figure 1

Kernel density plot signifying the distribution of the surgery subspecialization index (SSI) categories relative to (a) surgeon subspecialization in pancreatic surgery and (b) hospital subspecialization in pancreatic surgery

Full size image
Table 1 Characteristics of patients who underwent pancreatic resection, stratified by SSI
Full size table

Outcomes

Overall, 41.9% (n = 8,232) of patients achieved a TO. Achievement of TO increased incrementally with higher SSI (Low SSI: n = 2,094, 38.6% vs. Intermediate SSI: n = 2,890, 42.5% vs. High SSI: n = 3,248, 43.9%) (p < 0.001). A similar trend was observed relative to the individual components of TO. For example, patients treated by a high SSI care team were less likely to experience postoperative complications (Low SSI: n = 1,704, 31.4% vs. Intermediate SSI: n = 1,829, 26.9% vs. High SSI: n = 1,898, 25,7%), extended LOS (Low SSI: n = 1,771, 32.6% vs. Intermediate SSI: n = 2,048, 30.1% vs. High SSI: n = 2,134, 28.9%), and 90-day mortality (Low SSI: n = 496, 9.1% vs. Intermediate SSI: n = 508, 7.5% vs. High SSI: n = 537, 7.3%) (all p < 0.001). No differences in 90-day readmission were noted, however, regardless of SSI (p = 0.104) (Fig. 2). Furthermore, home discharge increased incrementally with SSI (Low SSI: n = 2,120, 39.1% vs. Intermediate SSI: n = 2,963, 43.6% vs. High SSI: n = 3,288, 44.5%) (p < 0.001), while FTR was highest among patients treated by low SSI providers (Low SSI: n = 339, 6.2% vs. Intermediate SSI: n = 319, 4.7% vs. High SSI: n = 360, 4.9%) (p < 0.001). Of note, there were no differences in expenditures related to the episode of care relative to SSI (Low SSI: $22,096.7, IQR $16,951.2–$36,712.0 vs. Intermediate SSI: $21,688.0, IQR $17,066.9–$36,290.6 vs. High SSI: $22,160.4, IQR $16,914.3–$35,705.3).

Fig. 2
figure 2

Bar graph depicting the achievement of textbook outcome (TO) and its components, stratified by surgery subspecialization index (SSI) categories.

* denotes statistical significance (p < 0.05)

Full size image

After adjusting for relevant sociodemographic and clinical characteristics, as well as fixed and random hospital effects, higher SSI was associated with higher odds to achieve a TO [referent: Low SSI; Intermediate SSI: OR 1.16 (95%CI 1.06–1.27); High SSI: OR 1.23 (95%CI 1.11–1.35)]. Specifically, higher SSI was incrementally associated with lower odds of postoperative complications [referent: Low SSI; Intermediate SSI: OR 0.84 (95%CI 0.76–0.92); High SSI: OR 0.76 (95%CI 0.69–0.85)], extended LOS [referent: Low SSI; Intermediate SSI: OR 0.85 (95%CI 0.77–0.94); High SSI: OR 0.84 (95%CI 0.75–0.94)], 90-day mortality [referent: Low SSI; Intermediate SSI: OR 0.81 (95%CI 0.69–0.95); High SSI: OR 0.78 (95%CI 0.66–0.92)], and 90-day readmission [referent: Low SSI; Intermediate SSI: OR 0.94 (95%CI 0.87–1.02); High SSI: OR 0.92 (95%CI 0.84–0.99)]. Higher SSI was also associated with lower odds of experiencing FTR [referent: Low SSI; Intermediate SSI: OR 0.76 (95%CI 0.63–0.91); High SSI: OR 0.77 (95%CI 0.64–0.94)], and increased odds of being discharged home [referent: Low SSI; Intermediate SSI: OR 1.23 (95%CI 1.10–1.37); High SSI: OR 1.22 (95%CI 1.07–1.38)] (Fig. 3). Of note, adjusted expenditures were similar regardless of SSI [referent: Low SSI; Intermediate SSI: OR 0.98 (95%CI 0.95–1.00); High SSI: OR 0.98 (95%CI 0.95–1.00)] (Table 2).

Fig. 3
figure 3

Adjusted probability, relative to surgery subspecialization index (SSI) percentile, of (a) achieving a textbook outcome (TO), (b) developing failure to rescue (FTR), (c) being discharged home.

Dotted lines signify 95% confidence intervals

Full size image
Table 2 Multivariable mixed effects logistic regression analyzing the impact of SSI on postoperative outcomes
Full size table

Hospital Volume and SSI

The effect of SSI was subsequently assessed within low- and high-volume hospitals. Notably, patients treated by high SSI care teams had higher odds of TO in low-volume [referent: Low SSI; Intermediate SSI: OR 1.14 (95%CI 1.01–1.28); High SSI: OR 1.15 (95%CI 1.02–1.31)], and to a greater extent high-volume hospitals [referent: Low SSI; Intermediate SSI: OR 1.16 (95%CI 1.01–1.32); High SSI: OR 1.26 (95%CI 1.09–1.45)]. Moreover, within low-volume hospitals, SSI was incrementally associated with lower odds of 90-day mortality [referent: Low SSI; Intermediate SSI: OR 0.74 (95%CI 0.61–0.92); High SSI: OR 0.73 (95%CI 0.60–0.88)], as well as FTR [referent: Low SSI; Intermediate SSI: OR 0.75 (95%CI 0.60–0.92); High SSI: OR 0.76 (95%CI 0.60–0.95)]. Interestingly, higher SSI care in low-volume hospitals was associated with a reduction in index admission expenditures [referent: Low SSI; Intermediate SSI: OR 0.96 (95%CI 0.92–0.99); High SSI: OR 0.93 (95%CI 0.90–0.97)].

Similarly, among high-volume hospitals, higher SSI was associated with lower odds of postoperative complications [referent: Low SSI; Intermediate SSI: OR 0.78 (95%CI 0.68–0.91); High SSI: OR 0.70 (95%CI 0.60–0.82)] and extended LOS [referent: Low SSI; Intermediate SSI: OR 0.78 (95%CI 0.66–0.91); High SSI: OR 0.82 (95%CI 0.69–0.97)]. Furthermore, higher SSI was also associated with higher odds of being discharged home [referent: Low SSI; Intermediate SSI: OR 0.31 (95%CI 1.12–1.54); High SSI: OR 1.32 (95%CI 1.10–1.58)], yet expenditures per episode of index care were not impacted by SSI [referent: Low SSI; Intermediate SSI: OR 1.01 (95%CI 0.97–1.06); High SSI: OR 1.03 (95%CI 0.98–1.07)] (Table 3).

Table 3 Multivariable mixed effects logistic regression analyzing the impact of SSI on postoperative outcomes, stratified by hospital volume status
Full size table

Discussion

The “volume-outcome” relationship, relative to both the individual treating surgeon as well as the treating hospital, has been demonstrated across a wide range of complex surgeries.9,21,22,23,24 Although the underlying mechanism of the association between hospital/surgeon volume and patient outcomes is likely multifactorial, several investigators have argued that it may serve as a proxy for other systems-level factors.24 Other metrics of quality that have been proposed include nurse-to-patient ratio, hospital magnet status and occupancy rate, case-mix, as well as degree of surgeon fragmented practice.3,4,5,6,7,10 The effect of these factors is especially pronounced in complex surgeries such as pancreatic surgery, which are often accompanied by poor patient outcomes.12,13,14 As such, the identification of surgeon- and hospital-related factors that influence patient outcomes is critical for targeted quality improvement initiatives. The present study is unique because it demonstrated that even within HP surgery, subspecialization of the care team was associated with improved patient outcomes after pancreatic surgery. Notably, compared with a low degree of HP subspecialization, higher categories of SSI were associated with 16% and 23% higher odds of achieving an optimal TO following pancreatic resection – a trend noted across TO components. Furthermore, higher SSI was also associated with higher odds of being discharged home, while being associated with lower odds of FTR. The association of SSI with improved postoperative outcomes persisted in both low- and high-volume hospitals. Of note, the benefits of higher SSI were derived without a commensurate increase in expenditures, and in the case of low-volume hospitals, undergoing surgery by a highly subspecialized care team was actually associated with a decrease in costs.

Since William Steward Halsted established the first formal surgical residency program at the Johns Hopkins Hospital in 1889, surgical treatment of gastrointestinal pathologies largely fell under the domain of general surgeons.25 However, the past 30 years have been marked by a ‘paradigm drift’ towards subspecialization after general surgery board certification.26 Although this accelerating trend towards subspecialization is likely multifactorial, a survey demonstrated that surgical residents opted to pursue fellowship training to obtain further technical expertise in areas of interest, to increase financial compensation, to maintain a better work-life balance, as well as to pursue academic interests.27 A similar effect has been observed in various other medical specialties such as radiology, physiatry, urology and ophthalmology.28,29,30,31 However, the impact of further specialization within hepatopancreatic (HP) surgery on patient outcomes remains relatively unexplored. The present study proposed the SSI as a measure of the degree of care team subspecialization. The benefit of this formulation was that it incorporated pancreatic surgical volume of both the surgeon and hospital in a single composite metric. Furthermore, it inherently controlled for the total hepatopancreatic surgical volume of the surgeon and hospital, thus increasing its validity. In turn, the results of the present study demonstrated that – while overall volume is an important indicator of quality – specificity within the case-mix even among HP surgeons can play an integral role in dictating patient care outcomes. Furthermore, these results highlight that even after hepatopancreaticobiliary fellowship training, increasing disease site subspecialization, either organically as a surgical practice matures or through a concerted effort, may be associated with improved patient outcomes. However, this must be balanced with ensuring access to hepatopancreatic care as well as maintaining an adequate supply of ‘general’ hepatopancreatic surgeons across the workforce.28

The ability to measure quality of care in a holistic manner is essential to foster a patient-centered focus within healthcare.1 As such, efforts to develop and implement such indicators have accelerated in recent times. In particular, compared with individual patient safety indicators such as postoperative complications, extended LOS, mortality, and readmission, composite metrics such as TO have been increasingly utilized due to a variety of reasons. For example, individual outcomes such as 90-day mortality may be challenging to analyze due to low event rates.32,33,34 Due to the involvement of multidisciplinary teams and the need to coordinate care, composite outcomes such as TO can also more holistically evaluate quality, as opposed to focusing on specific departmental domains. In the present study, being treated by a care team characterized by intermediate and high SSI was associated with 16% and 23% higher odds of achieving a TO, respectively. These results indicated that even within a subspeciality such as HP surgery, a focused pancreatic surgical practice on the part of the surgeon and hospital resulted in improved short-term postoperative outcomes for patients undergoing pancreatic resection. A similarly beneficial effect was observed considering other quality metrics such as FTR and discharge to home. Of note, FTR, also referred to as PSI-04,15 continues to maintain widespread appeal and utilization in quality improvement initiatives. Although prior literature has demonstrated lower FTR rates among high-volume hospitals,35 the impact of subspecialized practice on indicators such as FTR has not been previously evaluated. Furthermore, although it has been hypothesized that increasing subspecialization in general surgery is related at least in part to seeking additional financial compensation on the part of surgeons,26 this did not translate to increased expenditures related to the episode of care. In fact, despite substantial improvement in various quality metrics, expenditures remained similar when undergoing treatment at highly subspecialized practices compared with patients treated by care teams with a low degree of subspecialization.

Although previous attempts have been made to quantify the interaction of surgeon and hospital volume, a single composite metric that incorporates both has not been proposed to date. For example, Paredes et al. reported increased odds of developing a postoperative complication after a pancreaticoduodenectomy performed by a low-volume surgeon, regardless of hospital volume status or the nurse-to-patient ratio.8 Furthermore, a systematic review by Van Den Broeck et al. noted that both surgeon and hospital volume were independently associated with improved perioperative outcomes after prostatectomy.36 In the present study, the independent effect of care team subspecialization on postoperative outcomes, regardless of hospital volume, was validated in multiple ways. First, the nature of the SSI formulation intrinsically controls for both total surgeon and hospital hepatopancreatic volume. To further validate these findings, the mixed effects multivariable regression models were stratified according to hospital volume status. Among both low- and high-volume hospitals, increasing subspecialization was associated with higher odds of achievement of TO. A particularly interesting finding was that among low-volume hospitals, greater subspecialization was associated with decreased expenditures. Although this finding is possibly due to multiple underlying reasons, a contributing factor may be the generally better inpatient postoperative outcomes observed among patients who underwent care by a highly subspecialized team. Specifically, higher expenditures in the low SSI group may in part be due to the accrual of costs such as through the need for additional interventions for postoperative complications or increasing costs due to an extended LOS. As such, these results highlight that the benefits of subspecialized practice relative to patient outcomes extend across hospitals of different volume status. Furthermore, within low-volume hospitals, receiving care by a highly subspecialized team may decrease the financial costs associated with complex surgical episodes of care. However, these results may have been confounded by surgeon experience. Although the “volume-outcome” relationship has been extensively validated in prior literature, the impact of surgeon experience on patient outcomes remains a subject of controversy. For example, Kelz et al. and Anderson et al. reported no differences in patient outcomes among early-career versus experienced surgeons.37,38 Conversely, other authors have noted a steep surgical learning-curve, with worse outcomes observed among patients treated by early-career surgeons.39,40,41 Although the present dataset did not allow for the analysis of years of surgical experience as a covariate, future studies should assess the impact of surgeon experience on patient outcomes.

The results of the present study must be interpreted in light of several limitations. As with any retrospective study utilizing data from an administrative billing database, selection and misclassification bias were possible. Furthermore, as the study population was limited to elderly Medicare beneficiaries who were largely White, these results may not be generalizable to younger, commercially insured non-Medicare patients. Furthermore, as only hepatopancreatic surgeons and hospitals were considered, further studies are required across other surgical subspecialties to validate these findings. Moreover, operating room team dynamics such as assistance by another attending surgeon or hepatopancreatic surgical fellow versus a junior resident could not be ascertained; in turn, the composition of the surgical team may have had an impact on the overall quality of care and subsequent patient outcomes. Finally, although the analyses controlled for total surgeon and hospital volume as well as hospital teaching status, other metrics of quality such as nurse-to-patient ratio, hospital occupancy rate and fragmentation of practice were not accounted for.

In conclusion, the present study demonstrated the beneficial impact of care team subspecialization on pancreatic surgical outcomes. Notably, higher SSI was independently associated with improved postoperative outcomes such as TO and its components, FTR and discharge to home after pancreatic surgery. Furthermore, these benefits were derived without a commensurate increase in expenditures. Amidst increasing efforts to improve the quality of care that patients receive, surgical “super” subspecialization plays an important role in driving patient outcomes, regardless of total surgeon or hospital volume. Future studies should consider the use of SSI and investigate its applicability relative to other surgical subspecialties.