Introduction

Over the last decades a significant upward trend has been registered in the total number of spinal surgical procedures worldwide, including spinal fusion for degenerative spine diseases and surgical treatments for spinal tumors. This increase is the result of new developments in surgical techniques, new instrumentation devices and bone grafting materials, and introduction of relevant technological innovations. Despite these advancements in spinal surgery, the incidence of adverse events (AEs) (unexpected or undesirable events occurring as result of the surgery) and surgical complications (SCs) (causing any deviation from the normal postoperative course) is still high and clinically relevant. Yadla and colleagues reported a 24.2% incidence of major complications following thoracolumbar surgery [1] and 18.2% in cervical surgery [2]. These data were confirmed by other studies [3,4,5,6]. AEs and SCs represent a critical issue, having a negative impact on the patient's outcomes and quality of life. AEs and SCs have also a significant economic impact on healthcare systems and on the patient's psychological approach to surgery.

In 2009, the World Health Organization (WHO) published a manual (Guidelines for Safe Surgery), to support 10 "highly recommended" practices aimed at improving surgical safety. In order to implement these recommendations WHO produced and tested a checklist focused on intraoperative surgical safety [7]. The effectiveness of the WHO Surgical Safety Checklist in reducing AEs has been analyzed in different surgical contexts, and we recently evaluated its impact in spinal surgical complications’ rate [5].

However, the introduction of checklist systems did not solve the problem of AEs and SCs in spinal surgery, and some poor outcomes are not yet identified and graded as complications by the classification systems validated in the literature.

The purpose of this study is to analyze in detail the AEs that occurred in our tertiary reference center for spinal surgery, using a reliable method to capture and classify intraoperative and postoperative AEs, namely SAVES V2 system (Spinal Adverse Events Severity system) [8]. Moreover, we propose to analyze the effect of AEs on the hospital length of stay, which is a measure of the impact on the healthcare system. Finally, we would highlight the risk factors associated to the occurrence of AEs and to the prolonged length of stay, in order to hypothesize possible strategies to reduce complications and improve patients’ safety in spinal surgery.

Methods

Study design

This study was performed at a tertiary referral center for spine surgery, following the approval of the local Ethics Committee on December 14th, 2016 (protocol number 0022814). The study population included patients affected by degenerative, traumatic, infectious (hematogenous spondylodiscitis) and oncological spinal diseases, who underwent surgical treatment from January 2017 to January 2018. Minor patients were included. Surgeries were performed by five experienced spine surgeons of the same team. The investigation was conducted in accordance with the ethical principles of the latest version of the Declaration of Helsinki. The signature of a study-specific informed consent was obtained before patients’ enrollment.

Patients were classified in two groups, Non-Oncological and Oncological patients, and the two groups were analyzed separately.

During the study period the intraoperative and postoperative adverse events were prospectively collected, without using any validated classification system. The ward staff (1 research nurse and 1 clinical research assistant) prospectively collected in a database the AEs detected during the surgery and the hospitalization or reported by surgeons who visited patients during the follow-up period, as previously described [5].

Later, an independent MD observer retrospectively captured intraoperative and postoperative AEs of the same patients surgically treated from January 2017 to January 2018. The data were extracted from patients’ electronic health records, without checking the database filled by the ward staff. All the AEs were retrospectively recorded and classified according to SAVES V2 system [8], which is reported in Tables 1, 2, 3 (including SAVES V2 severity grades, categories of intraoperative AEs, categories of postoperative AEs). Postoperative AEs were differentiated in “early postoperative”, occurring in the first 30 days after surgery, and “late postoperative”, occurring more than 30 days after surgery. All AEs were categorized according to SAVES V2 severity grades from grade 1 to grade 6 (Table 1). The same data collection was repeated separately by a spine surgeon in training and by a clinical researcher, in order to assess the inter-observer reliability. The three different professional figures encountered the same number and type of complications. It was only necessary to find an agreement on the definition of “dural tear” and “nerve root injury” as adverse events, since in some surgeries on cancer patients these events may be consciously caused during the removal of the spinal tumor and in these cases they cannot be considered an error.

Table 1 SAVES-V2 severity grades
Full size table
Table 2 SAVES-V2 intraoperative AE categories
Full size table
Table 3 SAVES-V2 postoperative AE categories
Full size table

AEs collected by both methods, prospectively and retrospectively, were compared. Then we focused only on AEs collected retrospectively using SAVES V2 system, which were considered for descriptive and statistical analyses.

Patients’ characteristics including sex, age, comorbidities (described according to Charlson Comorbidity Index), ASA score and prior exposure to spinal surgery were collected. Information on the type and setting (emergency or elective) of the surgical procedure performed were also reported. The type of surgical procedures is represented in Fig. 1.

Fig. 1
figure 1

Type of surgical procedures recorded from January 2017 to January 2018

Full size image

The length of stay (LOS) during the index admission was calculated for each patient as the time from the date of admission to the date of discharge. Prolonged LOS was defined as ≥ 75th percentile LOS (in days) for each procedure category.

Second unplanned surgeries occurring during the index admission and during the follow-up period within two years from the index surgery were also retrieved.

Statistical analysis

Data were summarized using mean ± standard deviation or median and range value for continuous variables, and by absolute and relative frequency for categorical variables. Chi-square or Fisher Exact test, as appropriate were used to compare AEs between groups of patients. Two-way analysis of variance (ANOVA) on ranks was used to assess the effect of AEs and the effect of the different surgical procedures on LOS. Univariate and multivariate logistic regression models were used to determined odds ratio (ORs) for selected potential risk factor of complication and prolonged length of stay. Given the limited number of events and observations, variable categories were grouped when appropriate and the procedures with less than 5 cases were excluded from logistic models. A p value of < 0.05 was considered statistically significant. All statistical analyses were carried out using SAS 9.4 (SAS Institute, Cary, NC) software.

Results

AEs collected prospectively and retrospectively occurring in patients treated between January 2017 and January 2018

In the 1-year period from January 2017 to January 2018 a total of 336 patients underwent spinal surgery. Of these patients 223 were treated for non-oncological conditions while 113 had an oncological disease requiring surgical intervention to the spine. The demographic data of the study population are shown in Table 4. The mean age was 64.2 years (range 13–91) for the non-oncological group and 52.4 years (range 6–83) for the oncological group. Comorbidities were collected: the mean value of Charlson Comorbidity Index [CCI] was 4.7 in the oncological group and 2.4 in the non-oncological group. Increasing CCI was associated with an increasing likelihood of occurrence of any complications, and particularly of a minor complication [9], and a CCI greater than 2 can be considered more at risk for complications.

Table 4 Patients’ characteristics
Full size table

Previous spinal surgery was performed in 37% of oncological patients and 30% of non-oncological patients.

As reported in Fig. 2, we prospectively collected, without a validated capture system, a total of 97 AEs (20 intraoperative AEs, 44 early postoperative AEs and 33 late postoperative AEs). Fifty-seven patients (17%) of the cohort had at least one complication.

Fig. 2
figure 2

Summary of the adverse events data for Prospective and Retrospective collection

Full size image

Two hundred ten AEs were recorded retrospectively using SAVESV2 system (30 intraoperative AEs, 138 early postoperative AEs and 42 late postoperative). Ninety-nine patients (29.5%) of the cohort had at least one complication.

Overall a higher number of AEs was retrospectively recorded using SAVES system compared to the prospective recording without the use of any capture system (97/336 vs 210/336, p < 0.001).

Intraoperative AEs

Intraoperative AEs are reported in Fig. 3. The most common intraoperative AE was dural tear: We recorded 5 cases in the non-oncological group and 9 cases in the oncological group (14 cases out of 336 patients, incidence 2.2%). Other intraoperative complications were massive blood loss, visceral injury, vascular injury, hardware malposition requiring revision, cardiac failure. Intraoperative AEs were more frequent in the oncological group (23 AEs/113 patients) than in the non-oncological group (7 AEs/223 patients).

Fig. 3
figure 3

Intraoperative adverse events in non-oncological and oncological patients

Full size image

Early postoperative AEs

Early postoperative AEs are reported in Fig. 4. They account for most of the recorded AEs (138/210, 66%). Also, early postoperative AEs were more frequent in the oncological group of patients than in the non-oncological group (46% vs 38%). The most common early postoperative AEs were included in the category “others” (25 cases out of 336 patients, incidence 25.3%): In this category we included postoperative fever, wound secretion, anemia, respiratory infections, drug-related complications, arrhythmia, postoperative neuropathic pain.

Fig. 4
figure 4

Early postoperative adverse events in non-oncological and oncological patients

Full size image

Late postoperative AEs

Late postoperative AEs are reported in Fig. 5. The most common late postoperative AEs were construct failure with or without loss of correction for the non-oncological group (68% of all the late postoperative AEs in this group; incidence 17/223 patients, 7.6%), while wound dehiscence was the most common complication in the oncological group (41% of the late postoperative AEs in this group; incidence 7/113 patients, 6.2%). The rate of late postoperative complications was similar in the two groups of patients (15% for oncological patients and 11% for non-oncological patients).

Fig. 5
figure 5

Late postoperative adverse events in non-oncological and oncological patients

Full size image

Elective versus emergency surgery

As reported in Table 5, we compared the total number of complicated patients and the total number of AEs in the group of patients undergoing elective surgery and in the group pf patients undergoing emergency surgery.

Table 5 Comparison of AEs between patients undergoing elective or emergency spine surgery
Full size table

We observed no significant differences in the number of complicated patients with respect to the total number of patients between the elective and the emergency group, and this result was the same both in the oncological patient group and in the non-oncological patient group.

However, we observed that the total number of AEs with respect to the total number of patients was significantly higher in the emergency group than in elective group (p = 0.04).

Revision surgeries

Only one unplanned surgery was performed during the index admission for a hematoma occurring in an oncological patient.

However, 15 patients in the non-oncological group (6.7%) and 10 patients in the oncological group (8.8%) were readmitted within two years for a complication requiring a second unplanned surgery (Table 6). These complications had a severity grade ranging from 3 to 5 and the readmissions involved a mean hospitalization period of 17 days (Table 6).

Table 6 Second unplanned surgeries
Full size table

Moreover, we reported the index procedures performed for non-oncological (Table 7) and oncological (Table 8) patients. From these tables we obtain that in the non-oncological group 29/223 (13%) patients underwent a surgical revision procedure of a previous surgery performed in our center or in another center (“hardware revision surgery”). In the oncological group, 9 patients out of 113 underwent a surgical revision procedure (8%) (5 "hardware revision surgery" and 4 "debridment" for infection). We registered a total of 38 surgical revisions out of 336 index procedures (11%) (Fig. 1).

Table 7 Summary of LOS for Index Procedure for the non-oncological group
Full size table
Table 8 Summary of LOS for Index Procedure for the Oncological group
Full size table

Length of stay

The length of stay (LOS) during the index admission was calculated for each type of surgical procedure, both for patients having intraoperative and early postoperative AEs (“complicated patients”) and for patients without AEs (“uncomplicated patients”) (Table 7 and 8). LOS was calculated as the time from the date of admission to the date of discharge. In the non-oncological group, LOS was significantly higher for complicated patients compared to uncomplicated patients (F = 44.11, p = 0.0000) and LOS was significantly different between type of surgical procedures (F = 5.63, p = 0.0000). Similarly, in oncological group of patients a significant effect both of complication (F = 21.01, p = 0.0000) and different type of surgical procedure (F = 4.24, p = 0.0004) on LOS was observed.

The mean increase of LOS due to AEs is 8 days. As reported in Tables 7 and 8, LOS increased in the group of complicated patients for all the procedures examined, and AEs had a significant impact on LOS particularly for some procedures: hardware revision surgery, laminectomy, posterior fusion (TLIF or PLIF) with instrumentation, posterior decompression with instrumentation, posterior deformity correction and fusion (for non-oncological patients); surgical decompression and stabilization, intralesional resection¸ vertebrectomy (for oncological patients).

Risk factors for prolonged length of stay

Prolonged LOS was defined as ≥ 75th percentile LOS (in days). In Tables 9 and 10, the 75th percentile LOS is reported for each procedure category, in the non-oncological and in the oncological group, respectively. The procedures with less than five cases were excluded.

Table 9 Length of stay by procedure category for non-oncological patients
Full size table
Table 10 Length of stay by procedure category for oncological patients
Full size table

In the non-oncological group, the surgical time longer than 3 h was a risk factors for prolonged LOS, as well as the presence of a postoperative AE (Table 11). In the multivariate regression model only the presence of a postoperative AE was confirmed as risk factor (OR = 3.3 (95% CI 1.5–7.1)).

Table 11 Univariate logistics regression model of potential risk factors for prolonged LOS in non-oncological patients
Full size table

In the oncological group, in addition to the surgical time longer than 3 h and the presence of a postoperative AE, age greater than 65 years and comorbidity index > 2 were also significant risk factors for prolonged LOS (Table 12). In the multivariate regression model surgical time and postoperative AEs were confirmed as risk factors for prolonged LOS (respectively: OR = 2.7 (95% CI 1.0–7.3), OR = 3.5 (95% CI 1.3–9.3)).

Table 12 Univariate logistics regression model of potential risk factors for prolonged LOS in oncological patients
Full size table

In a subset of patients with available data, we also observed that the number of instrumented levels and the degree of severity of complications were not associated with a prolonged length of stay.

Risk factors for AEs

As reported in Fig. 6, in the non-oncological group, there were 58 patients with surgical AEs (26%) and in the oncological group there were 38 patients with AEs (33.6%).

Fig. 6
figure 6

Presence of adverse events in non-oncological and oncological patients

Full size image

In the non-oncological group, there were different factors associated to AEs (Table 13): age > 75 years old, anatomical region, surgical time longer than 3 h, presence of instrumentation, presence of previous surgeries. In the multivariate logistic regression model only the surgical time longer than 3 h was confirmed as risk factor for AEs (OR = 5.0 (95% CI 1.8–13.9)).

Table 13 Univariate logistics regression model of potential risk factors for AEs occurrence in non-oncological patients
Full size table

In the oncological group we observed that the surgical time longer than 3 h, the presence of instrumentation and the presence of previous surgery were associated to an increased risk of AEs (Table 14). In the multivariate logistic regression model only the presence of previous surgeries was confirmed as risk factors for AEs (OR = 3.4 (95% CI 1.3–8.6)).

Table 14 Univariate logistics regression model of potential risk factors for AEs occurrence in oncological patients
Full size table

Discussion

To evaluate the quality of surgical care is increasingly relevant, as surgery has become technologically advanced and highly specialized and involves invasive procedures performed also on high-risk and complex patients. Surgical outcomes, and in particular complications, are indicators for surgical quality assessment, and patients’ safety is considered a primary goal to be achieved, considering that complications have a demonstrated association with increased burden on the healthcare system, resulting in increased resource utilization, longer hospital stay, and higher costs.

In order to achieve this goal, the first step is an objective capture and classification of adverse events and complications occurring in a specific surgical context. Thus, we focused our analysis on AEs and SCs occurring in spinal surgery, which has been reported to record a high rate of intra- and postoperative AEs [3, 10].

First of all, we tried to understand if the introduction of the WHO Surgical Safety Checklist (SSCS) had an impact on the complications’ rate in our tertiary referral center [5]. Even if the results were promising, we evidenced the lack of a reliable classification system for AEs and SCs and started to introduce the SAVES V2 system [8] to capture and classify AEs and SCs both in retrospective and prospective cohorts of patients surgically treated for different types of spinal diseases.

As recently reported in a systematic literature review [11], other classification systems were used before the introduction and validation of SAVES v2 system. The most commonly used were the classification “intraoperative and perioperative” and the Glassman classification [12] (major and minor, where significant complications requiring reoperation or leading to permanent deficit are major complications and other general medical adverse events or perioperative events with time limited effect are minor complications). Other papers use the McDonnell’s classification [13], the International Classification of Disease, Ninth Edition, Clinical Modification (ICD-9-CM), the Common Terminology Criteria of Adverse Events (CTCAE) or the Clavien-Dindo classification [14].

The prevalence of complication reporting in surgical spine articles was 64.2% [11]. A high variability was observed concerning definition of complications, assessors, timing to record complications, and the classification systems used. The quality of complication and adverse event reporting in spinal surgery literature has been recently analyzed [15].

We thought that to start focusing on how to reduce the high complication rates in spine surgery, a better reporting is mandatory and we tried to analyze a series of patients surgically treated in our tertiary referral center during one year, including patients affected by oncological (34%) and non-oncological (66%) diseases, in order to (i) capture, classify and grade all the AEs according to SAVES v2 system and (ii) evaluate which risk factors could be associated with the occurrence of AEs and with an indicator of poor surgical outcome, as the prolonged length of stay.

Comparing a prospective collection of AEs without a concerted methodology of capture and a validated classification system to a retrospective accurate collection and classification with SAVES system, we detected a significant difference in the complications’ rate: 17% of patients had at least one AE in the prospective collection, while 29.5% of patients had at least one AEs in the retrospective collection with SAVES system.

The capture and classification of AEs according to SAVES system allowed to detect a higher rate of AEs in the oncological group of patients (26% of patients with complications in the non-oncological group and 34% of patients with complications in the oncological group). In particular, we observed a significantly higher number of intraoperative AEs in oncological patients than in non-oncological patients. Also, early postoperative AEs were more frequent in oncological patients than in non-oncological ones. This evidence can be associated with more demanding surgical procedures necessary to treat patients affected by spinal tumors and with the frailty of this category of patients. We recently reported an analysis of AEs, classified with SAVES V2, in a large number of patients treated at our center by en bloc resection for primary spinal tumors [16].

Frailty has been associated with increasing disability, hospitalization, adverse health outcomes and death. A number of observational studies have shown that frailty worsened postoperative outcomes as morbidity, mortality and length of stay (LOS) [17, 18] and the severity of frailty syndrome has been reported to be directly correlated with post-surgical mortality rates and complications [19]. As frailty is correlated to general surgery outcome, it might also predict the outcome in patients undergoing high-demanding spinal surgery [20]. The most common intraoperative AE resulted to be the dural tear (14/336 patients, 4%). Accidental dural tears have been reported in the literature as a relevant AE in spinal surgery with an overall incidence that significantly varied in the studies and ranged from 0.4% to 15.8%. Centers with a high number of surgical procedures reported lower rates of dural tears and vice versa [21, 22]. Revision surgery, age, diabetes, obesity have been reported as risk factors for dural tears [21].

Early postoperative AEs account for the most frequent category of complications in the present study (138/336 patients, 41%). Using SAVES V2 the reported incidence of early postoperative complications varies in the literature from 12.4 to 77% [4, 22]. An interesting consideration is that the category of early postoperative AEs with higher frequency is “other”. In this category we classified a large variety of medical complications, and some of them had a significant frequency. Other authors have also reported a large number of complications collected in the “other” category. We wonder if all of the complications collected in the “other” category were clinically relevant. As recently suggested [23], there is the need of a stringent definition of what is considered or not an AE. We propose to revise the categories of AEs included in SAVES V2 in order to better classify these events.

Concerning late postoperative AEs, the rate was similar in the two groups of patients. However, the most common was construct failure with or without loss of correction for the non- oncological group (17/223 patients, 7.6%) and wound dehiscence for the oncological group (7/113 patients, 6.2%). These data agree with those previously reported [5].

Comparing elective and emergency patients, we observed that the number of complicated patients was not significantly different in the two categories, however, patients undergoing emergency surgery developed a significantly higher number of AEs with respect to elective patients. These results are different from those previously reported by Street et al [24] and Karstensen et al [4], who observed no significant increase of AEs in emergency versus elective surgery.

Moreover, late AEs were frequently associated to a revision surgery: In fact, we observed only one unplanned surgery due to an AE during the hospitalization period, while 15/223 patients (6.7%) in the non-oncological group and 10/113 patients in the oncological group (8.8%) were readmitted within two years for a second unplanned surgery due to AEs. These AEs had a significant impact for the patients and the healthcare system, as they were classified with severity grade ranging from 3 to 5 (according to SAVES v2 system) and required a mean hospitalization period of 17 days.

We analyzed the length of stay for the two groups of patients (oncological and non-oncological) and for different types of surgical procedures inside each group. We compared LOS between complicated and uncomplicated patients and observed that LOS was significantly increased in the presence of AEs both in the non-oncological and in the oncological groups of patients.

As described by Wallner et al. [25], we analyzed the effect of different risk factors on prolonged length of stay, considering prolonged LOS ≥ 75th percentile LOS (in days).

Following univariate and multivariate regression analyses of different risk factors for prolonged LOS, we observed that only the presence of postoperative AEs affected the LOS for non-oncological patients in the multivariate model, while for the oncological group of patients both postoperative AEs and surgical time longer than 3 h were significant risk factors for prolonged LOS in the multivariate model.

Thus, the occurrence of intraoperative and early postoperative AEs significantly increases the length of stay for both oncological and non-oncological patients with a consequent impact on the costs of the healthcare system. In our public Hospital the cost for the hospitalization of a patient submitted to spinal surgery is 1082 euros per day (mean value). This cost includes examinations and medications performed during the hospitalization period. As the mean increase of LOS due to AEs was 8 days, we hypothesize an average increase of costs for the hospitalization of complicated patients around 8000 euros. This estimation could be higher due to the necessity of additional exams, consultations, and medications necessary to resolve the adverse events.

Our results highlight that it is mandatory to reduce AEs in order to minimize the costs for the healthcare system.

Considering risk factors for AEs, we observed at multivariate analysis that surgical time longer than 3 h significantly increased the incidence AEs in the non-oncological group, while the presence of a previous surgery was a risk factor for AEs in the oncological group of patients.

These results indicate that the reduction of surgical time (which is associated to high demanding procedures) and of revision surgeries is a relevant goal to achieve in order to reduce complications.

A limitation of our analysis of AEs is the lack of the patients’ point of view: We captured and classified AEs from the clinicians’ point of view, however, failed back surgery syndrome [26] is known as a category of complications perceived by the patients following spinal surgery, while no adverse events are detected by the surgeon.

Thus, we propose to emphasize the patient’s point of view for the evaluation of AEs and SCs, and consequently for a shared surgical decision-making.

Conclusion

The results of this study indicate that a validated capture system as SAVES v2 can represent a first step to correctly face the relevant problem of complications in spinal surgery.

Considering that the rate of AEs and SCs in spinal surgery is still high despite the improvement of technical skills, we hypothesize that non-technical skills could be relevant for the occurrence of these events and we suggest that teamwork, interprofessional collaboration and communication should be implemented in order to improve patients’ safety. We think that Health organizations in western countries are committed to improving patient safety through education of staff and teamwork education programs, including the use of checklists and other tools elaborated to highlight risk factors for AEs and SCs, and address them [27].

Thus, we propose to introduce in surgical departments one or more professional figures dedicated to patients’ safety, who monitor the critical points where AEs and SCs could arise and support surgeons and other healthcare professionals in the application of preventive measures to reduce them, during surgery but also in the pre-operative and postoperative periods.