Incidence and risk factors of spinal epidural hemorrhage after spine surgery: a cross-sectional retrospective analysis of a national database

South Korea has a National Health Insurance System (NHIS) that covers approximately 98% of the overall Korean population, and 99% of National Health Insurance (NHI) claims data have been electronically reported since 2005. The claims data of Health Insurance Review and Assessment Service (HIRA) pertains to patients’ diagnosis, treatment, surgical history, procedures performed, and prescription drug information. However, the extensive volume and complicated structure of the claims data limit research approaches. To improve accessibility for researchers to use the claims data, HIRA has established the patient sample data sets that have been validated by five different validity tests [21].

These patient sample datasets are as follows: adult patient group, pediatric patient group, all patient group, and inpatient group. The adult patient sample represents approximately 1 million patients aged > 65 years annually (20% of the original data), whereas the pediatric patient sample takes into account 1.1 million patient aged < 20 years annually (10% of the original data). All patient samples represents overall 1.4 million patients annually, which represents 3% of the original population. Lastly, national inpatient sample pertains to 700,000 inpatients (13%), and approximately 400,000 outpatients (1%) annually. All sample dataset contain sampling weights allow obtaining estimated national population. With a 95% concordance, the estimated population and the actual population demonstrated a high level of representativeness [22].

Our study analyzed the customized NHIS dataset of HIRA National Inpatient Sample of 2014 (HIRA-NIS-2014), which contains claims data from January 1, 2014 to December 31, 2014 [23]. This period is defined as the research period in this study. Sampling weights were applied to give best representative estimation of the actual population. The analysis in this study was based on the number of pertinent cases, instead of the number of patients, as one patient might undergo multiple spinal surgeries during the research period. For example, if a patient underwent two different spine surgeries during their independent hospitalizations, the detected number of pertinent cases would be two.

The data that support the findings of this study are available from HIRA but restrictions apply to the availability of these data, which were used under license for the current study. Thus, these data are not publicly available. However, the data are available from HIRA upon reasonable request and with permission.

In Korea, currently there are no disease/treatment codes that exclusively represent SEH. Based on existing studies, this study converted the diagnostic code for postoperative SEH (998.12) indicated in the International Statistical Classification of Diseases and Related Health Problems, 9th revision (ICD-9) to the most appropriate disease code in the 6th Korea Informative Classification of Disease (KOICD-6) codes [24]. As KOICD codes does not differentiate postoperative hematoma from hemorrhage, SEH cases in this study included both spinal epidural hemorrhage and spinal epidural hematoma.

Based on code selection recommendations from the Centers for Disease Control and Prevention/National Center for Health Statistics and KOICD, we first extracted all cases of postoperative hemorrhage and hematoma (T81 and T810), and then narrowed down to severe SEH. Severe SEH cases were defined as the ones that required medical treatment codes for the removal or drainage of hematoma in the central nervous system [9, 25]. T81 and T810 are defined as hemorrhage and hematoma which excludes adverse effect of drug, hematoma of obstetric wound, or failure of transplanted grafts. If the SEH case also had hemorrhage controlling or hematoma drainage codes such as S4755, S4756 or S4754, it would be included in the group of severe SEH cases. Detailed information on the disease and treatment codes were listed in Supplementary Table 1.

Next, all selected cases were individually investigated by three different medical professionals for concurrent diseases, treatment provided, and medications to demonstrate detailed information on severe SEH after spine surgery during patient hospitalizations. Subsequently, all claims records of selected patients during the study period, the disease/treatment codes for the complications of surgical and medical care and postoperative disorders of the nervous system after spine surgery were examined. Detailed information on the disease and treatment codes is listed in Supplementary Table 1.

As an illustration, in a patient with spinal disease undergoing spine surgery, the incidence of postoperative hemorrhage and hematoma was first examined. Then, the intervention for postoperative SEH or spinal nerve compression was evaluated throughout the study period. If any of the codes indicated postoperative disorders of the nervous system or complication from surgical care, every detail of the case was flagged for further review. All variables for a pertinent case were relevant to the hospitalization period. Detailed information on the SEH codes is presented in Supplementary Table 1.

To construct the conceptual framework of analysis, the extant studies on incidence and risk factors of SEH were extensively reviewed. The terms “(spinal epidural hematoma) AND (spine surgery) AND (risk factor)” were searched in the MEDLINE database. Simultaneously, the codes for spinal disease were referenced by literature review. In case a patient had a spinal disease code and underwent surgery in relation to the main disease, the patient was considered to have undergone spine surgery. In addition, the risk factors considered in this study were based on the results of the literature review. For example, the definition of extended surgery hours (≥ 2 h) and loss of blood were based on the literature search. The use of nonsteroidal anti-inflammatory drugs (NSAIDs) was defined as the administration of NSAIDs at any time during the hospitalization. The total medical cost was the total expenditure on the hospitalization in a case that was claimed by the medical institution. Insurance types were categorized as the NHI and Medical Aid/Veterans Medical Benefits. Blood loss was calculated by the blood transfusion claims codes of the case. Type of hospitals was classified into two categories: general hospitals (e.g., tertiary academic hospitals and general hospitals) and small hospitals (e.g., hospitals and clinics). The location of hospitals was divided into three categories: metropolitan area (Seoul), other metropolitan areas(e.g., Busan, Daegu, Incheon, Gwangju, Daejon, Ulsan and Sejong), and rural area (all other locations). The duration of hospital stay was calculated by subtracting the admission date from the discharge date.

In case a patient record included any invasive procedures such as lumbar puncture, myelography, and epidural anesthesia, then it was considered that the patient underwent invasive procedures. If a patient had thrombocytopenia or coagulation factor deficiency, the patient was regarded to have a bleeding factor. Trauma case was determined if the patient was admitted through the emergency department.

After establishing the analytical dataset containing the variables identified as risk factors, we conducted a descriptive analysis. A pairwise correlation test was applied to determine the level of correlation among the set of variables included in the study model. A chi-square test and independent two-sample t-test were used accordingly to determine if there was any significant association between a set of independent variables and both categorical and continuous dependent variables. As the final step, a multivariate logistic regression was utilized prior to which multicollinearity was examined by referring to the variance inflation factor. Data with a variance inflation factor of > 5 were excluded from the model. In the multivariate logistic regression model, this study tries to identify factors that are associated with the incidence of SEH. Sampling weight was used to obtain better estimation of the original population. All statistical analyses were performed using Stata 14.0 [26].

Based on the inpatient records, out of a total of 31,964,373 cases, there were 17,549 cases of spinal surgery in 2014. As shown in Fig. 1, the total number of cases of hemorrhage and hematoma occurring after spine surgery was 201 (1.15%), and there were no severe SEH cases as assessed by the codes of hemorrhage evacuation, decompression, or postoperative complication caused by spine surgery.

Based on the extensive review of recent studies, Table 1 presents the incidence rate and risk factors of SEH. The incidence rate varied from 0 to 23.6% owing to the differences in diagnoses, detection methods, and sample sizes. The identified risk factors are as follows: obesity, hypertension, diabetes, bleeding factors, use of anticoagulant/antiplatelet medications, alcohol consumption, malignancy, previous surgeries, female sex, surgery type (lumbar) and approach, age of 65 and above years, use of NSAIDs, blood loss, trauma, prolonged surgical time, and invasive procedures.

Patient characteristics were obtained from the 17,549 spinal surgery cases (201 patients with SEH and others). The values for age, total medical cost, and duration of hospital stay were greater for SEH cases than for non-SEH cases. Lumbar spine surgery type and posterior approach showed significance in the univariate analysis. Of all SEH cases, 199 involved treatment with NSAIDs. Among all patients with SEH, 14 underwent surgery for > 2 h in contrast to 187 patients who underwent surgery for ≤2 h. Regarding blood loss, 135 patients were reported to have no blood loss in the claims dataset, and 41 had < 0.5 L of blood loss in the SEH group. Others experienced a blood loss of > 0.5 L (N = 25). Most spine surgery procedures wherein SEH occurred (N = 139) were performed at small hospitals, whereas only few such procedures (n = 63) were performed in general hospitals. The results of the detailed descriptive analyses are shown in Table 2.

A pairwise correlation analysis that was performed before the univariate analysis revealed that there was no significant correlation among the independent variables. In the univariate analysis, we found that most independent variables that have been included in the study model were associated with the incidence rate of SEH. The following variables were found to have a statistically significant relationship with SEH. Age (p = 0.03), total medical cost (p < 0.001), lumbar spine surgery (p < 0.001), anterior-approach surgery (p < 0.001), hypertension (p = 0.01), use of NSAIDs (p = 0.01), bleeding factors (p < 0.001), surgical time >  2 h (p < 0.001), blood loss > 0.5 L (p < 0.001), number of hospital beds > 200 (p = 0.03), and hospitals located in the rural areas. The variables with marginal significance (0.05 ≤ p < 0.20), such as sex, infections, and small hospital types, were also included in the multivariable logistic regression model. The use of anticoagulant was not found to be a significant risk factor as per the univariate analysis, however, we included the variable in the multivariate analysis since other extant studies demonstrated it as a strong risk factor of SEH. Table 3 presents the results of the univariate analysis. Lumbar spine surgery, blood loss > 0.5 L, surgery > 2 h, hypertension, use of NSAIDs, presence of bleeding factors, small hospitals, and location of the hospital were identified as risk factors for SEH. However, anterior approach and anticoagulant use showed a protective effect on SEH incidence. The analyses of the survey data included the calculation of sampling weights, which ensured about 95% of concordance of the patient sample dataset with the actual population [21]. The details on the results of the multivariate analysis are presented in Table 4. The analysis results from the non-weighted dataset are provided in Supplement Table 2.

This is the first study to utilize a national database of Korean population to investigate the incidence and risk factors of SEH. Of total 17,549 cases of spine surgery, 201 (1.15%) were found to be the case of symptomatic SEH, and there was no case of severe SEH detected leading to continuous nerve damage and reoperation.

So far, only few studies have assessed the topic of SEH in the Asian population [12, 40‐43], where the reported incidence of SEH with nerve deficit ranges from 0.1 to 0.9%. Imajo et al. utilized a series of nationwide survey results on the complications among Japanese patients who underwent spine surgery from 1994 to 2001 (N = 31,380) and reported SEH incidence rate of 0.9%, whereas Anno et al. reported an incidence rate of 0.4% based on data from a single-center [43]. They found that SEH had a positive association with prolonged surgical time, intraoperative blood loss, and old age. The lumbar procedure was also identified as a risk factor in a previous study, and the neurological outcome was found to be poor in two cases.

Yamada et al. reported that postoperative SEH evacuation was performed in 32 of 8250 patients at their institution, out of which 14 patients developed severe paralysis [44]. The patients who underwent SEH evacuation within 24 h of the onset of symptoms had better clinical outcomes. The duration of onset for SEH symptoms ranged from 3 h to 3 days, and the median duration was 24 h after the initial surgery. The duration of SEH evacuation ranged from 2 h to 5 days after the onset of the symptoms. Yi et al. analyzed 3720 spine surgery cases over a period of 7 years (1998–2005) at a single institution in Korea and reported nine cases (0.24%) of severe neurologic deterioration requiring surgical decompression [42]. Lumbar spine surgery, anticoagulation therapy, and highly vascularized tumor were also reported as the risk factors of SEH. Better clinical outcomes have been consistently noted in patients who undergo early intervention to evacuate SEH.

There were no cases of reoperation for hemorrhage removal or neurological deterioration due to SEH in our analytic dataset, which could be explained based on the following reasoning. In Korea, the period of hospitalization after spine surgery is more than three-fold longer than that in other countries [45]. The average duration of hospital stay for our patient sample was 13.48 days, which is more than three-fold longer than the international average of 4 days [46]. The SEH group showed a longer hospital stay at 19.35 days, and the medical care team might have had adequate time to detect the potential development of epidural hemorrhage. Therefore, postoperative hemorrhage and hemorrhage were estimated to develop in 1.15% of the total patients; this incidence is higher than that in similar ethnic groups based on the literature. However, the incidence of severe SEH requiring reoperation was minimal.

Our analysis of the risk factors of SEH confirmed previous findings that the anterior-approach, lumbar surgery, intraoperative blood loss, prolonged surgical time, high blood pressure, use of NSAIDs, and concurrent bleeding factors are associated with SEH. Moreover, the type and location of the hospital were also found to be associated with the risk of SEH, which could suggest that the differences in the medical service quality based on the location of the hospitals could lead to different surgical outcomes. Inconsistent quality of inter-institution and inter-nation medical care has been a major global concern [46].

In our univariate analysis, anticoagulant use was an insignificant risk factor of SEH (p = 0.51). However, the use of anticoagulant and/or antiplatelet agents has been considered as one of the most important triggers for SEH incidence as reported previously [8, 42, 47, 48]. Subsequently, the use of anticoagulant was incorporated in the multivariate analysis to demonstrate its effects on postoperative hemorrhage and hematoma, which showed no statistical (p = 0.08; OR = 0.66 [0.42–1.05]).

Postoperative thromboprophylaxis is applied only to 7.89% of patients undergoing spine surgery in Korea compared to 56.3% of the patients in North America. The apprehension against the use of anticoagulant because of suspected incidence of severe SEH might be the reason for the differences in the results in our analysis. However, thromboprophylaxis with optimal dose has been known to have no negative effect on the incidence of SEH [49]. As the preparation of thromboprophylaxis protocol has been incorporated into the hospital accreditation evaluation in Korea, standardized and optimized thromboprophylaxis might be considered in qualified patients without contraindication [19].

Our study attempts to fill the existing gap in the data on SEH. This is the first study examining the risk factors of SEH based on a national insurance database. Previous analyses and reports were based on single-center data and on an availability basis. Consequently, the findings of our study on the risk factors of SEH could help clinicians make suitable clinical decisions for patients undergoing spine surgery. Moreover, these data from this pilot study can be utilized for further studies on SEH and for the development of an exclusive disease code for SEH.

Although our study makes a significant contribution to the literature, the absence of an exclusive disease code for SEH in KOICD-6 is its limitation. The study attempted to overcome this constraint that exists in the national insurance database by utilizing disease/treatment codes and interventions for severe cases of SEH. Although the HIRA-NIS dataset is derived from the national claims data, it only contains information on treatments being used during hospitalization, and does not provide the exact administration time. Further studies focused at bridging these gaps are strongly encouraged.