A Review of Neuraxial Epidural Morbidity: Experience of More Than 8,000 Cases at a Single Teaching Hospital

This project was conducted as a Quality Improvement initiative and as such did not require specific approval from the St. Vincent’s Hospital Human Research and Ethics Committee (St. Vincent’s Hospital, Melbourne, Victoria, Australia). The St. Vincent’s Hospital APS database is a set of prospectively gathered data on all patients who have been under the care of the APS since it began in 1990. Each patient on the APS had a tracking sheet initiated by the anesthesiologist inserting the epidural catheter, immediately postinsertion, at the same time that the epidural drug infusion was prescribed. The tracking sheet contained baseline information regarding the procedure and details of the epidural placement. It also recorded daily information regarding pain scores, motor blockade, the presence of defined adverse events (nausea, pruritus, site infection), and free-text notes. On the rare occasions that the APS tracking sheets were not initiated or were misplaced, the pain service was notified, and a tracking sheet commenced when the APS registrar was called for epidural orders or with requests for epidural removal. After patient discharge from the hospital, these data were entered into a computerized database (including free-text comments). Any patient returning to the hospital with a problem related to APS care had appropriate additional comments added to their record by consultant anesthesiologists, registrars, or, from 2002, the APS nurse. A search of this database was performed for cases of serious or suspected neuraxial epidural morbidity, namely epidural abscesses and hematomas. Precoded fields included insertion site infection  (defined as two or more of purulence, erythema, or tender swelling at the site), pyrexia  (temperature of greater than 38.5°C for more than 6 h), excess motor blockade , and abnormal termination of infusion  (i.e. , earlier than planned). Free text comments fields were searched with key terms, including abscess , hematoma , scan , magnetic resonance imaging  (MRI ), computerised tomography  (CT ), back pain , weakness , paralysis , bladder , inflam , pus , purulent , antibiotics , and coagulation . Records of patients identified in this manner were retrieved and reviewed. In addition, the original APS paper tracking sheets were all accessed. Cases of nonneuraxial epidural morbidity, such as pruritus, hypotension, nausea, and vomiting, were not considered further, and an early subset of this data has been published elsewhere.8Our data were cross-checked against the hospital’s diagnostic group and costing database, which is completed by medical records staff after discharge. A detailed hospital medical records–based coding search was performed for epidural hemorrhage  and intracranial and intraspinal abscess , which are both specifically coded diagnoses, from January 1990 to December 2005.

Determination of the “denominator” (i.e. , the total number of epidural catheters inserted and run on the wards postoperatively) was also possible from the APS database, and thus a precise complication rate was determined. Note that the database did not include single-shot epidurals, epidurals where a catheter was inserted and removed in the operating suite (e.g. , failed block, epidural anesthesia only), or other neuraxial blockade (e.g. , spinals).

Data were retrieved from a computerized database (Microsoft Access; Microsoft Corporation, Redmond, WA). Bivariate categorical data were analyzed using chi-square analysis, Fisher exact test, or logistic regression as appropriate. Ranked ordinal data were reduced to two categories for logistic regression. A P  value of less than 0.2 was used to accept components in sequential multivariate analysis, and a P  value of less than 0.05 was considered statistically significant. Odds ratios (ORs) are expressed with 95% confidence intervals (95% CIs). Mean values are expressed as mean ± SD. Analysis was undertaken using STATA statistical software version 8.0 (Stata Corporation, College Station, TX).

Over the 16 yr from February 1990 to the end of December 2005, 8,210 epidural catheters were inserted for postsurgical analgesia at St. Vincent’s Hospital and followed up by the APS. The annual caseload is shown in figure 1. The mean duration of infusion was 2.8 ± 1.3 days (mean ± SD), ranging from less than 1 day to 13 days, with 32 patients having infusions for 7 days or more. The mean age was 59.3 ± 16.6 yr (mean ± SD), and the most common surgical procedure coded was general abdominal procedures including upper abdominal surgery (31% of all cases). Details of epidural catheter management and surgical categories are shown in table 1.

During this period, 30 MRIs and 2 CT scans were performed to investigate possible neuraxial complications. From the mid-1990s, MRI scans were used in preference to CT scanning because of the greater fidelity in defining intraspinal pathology.5The total number of MRI examinations performed annually in our institution since 1994 is shown in figure 2along with the numbers ordered by the APS. In 16 cases, epidural abscess was suspected; in 10 cases, the indication was for suspected hematoma; and in 4 cases, the possibility of either was considered. Therefore, imaging for possible abscess occurred in 1:410 epidurals (0.24%), and imaging for possible hematoma occurred in 1:586 epidurals (0.17%). Epidural abscess was confirmed in 6 scans, an epidural hematoma was confirmed in 1 scan, and a subarachnoid hematoma was confirmed in 1 scan (fig. 1and table 2).

Of the 8,210 patients, 6 were diagnosed with an epidural abscess (1:1,368 or 0.073%; 95% CI, 0.015–0.135%). Five patients were successfully treated with antibiotics alone. One patient required surgical decompression (1:8,210 or 0.01%; 95% CI, 0–0.07%). Clinical signs indicating an MRI/CT scan for possible abscess (n = 20 patients) were evidence of an epidural insertion site infection (12 cases), pyrexia (8 cases), epidural insertion site infection plus pyrexia (7 cases), back pain (4 cases), or neurologic signs (1 case) (table 2). The combination of pyrexia and epidural insertion site infection was present in 5 of the 6 epidural abscess patients (with site tenderness only in the remaining patient). Seven of the 20 MRI scans sought for suspected epidural abscess were from patients with pyrexia and epidural insertion site infection, whereas this combination was present in only 30 patients overall (0.36%).

To ensure that no diagnoses of epidural abscess were missed, we cross-checked our 8,210 APS patients with 171 epidural abscess patients who were identified by the hospital medical records–based search over the same period. These patients were cross-referenced by Unit Record Number and again (separately) by surname with our APS database. This process identified only one possible additional epidural abscess patient; however, on chart review, this coding had been erroneously generated by a note made by the Infectious Diseases registrar, “Impression—Epidural Abscess,” before a normal MRI scan and uneventful subsequent follow-up.

The vertebral level of catheter insertion was mid to low thoracic (T6–T11) in 87% of patients (median insertion levels are presented in table 1). The incidence of epidural insertion site infection was independently associated with the level of insertion being 2.8% at the thoracic levels (T6–T11) compared with 0.8% at the lumbar (L1–L4) levels (P < 0.01, n = 6,266). However, mean duration of catheter insertion was longer for thoracic catheters than lumbar catheters (thoracic, 3.0 ± 1.3 days; lumbar, 2.1 ± 1.2 days; P < 0.01), due in large part to the planned use of epidural analgesia for a maximum of 48 h in orthopedic surgical patients. Logistic regression analysis relating epidural insertion site infection (184 cases) to patient age, duration of catheter insertion, spinal level of insertion, and type of surgery (n = 5,115 for all included categories) was performed. A significant association was identified between epidural insertion site infection and duration of catheterization (in days: OR, 1.4; 95% CI, 1.2–1.6; P < 0.001), i.e. , for each postoperative day the epidural remained in situ , the risk of site infection increased by 40% per day. In addition, general abdominal or thoracic surgery was associated with a greater risk for insertion site infections than orthopedic or vascular surgical procedures (OR, 3.3; 95% CI, 1.7–6.4).

Bupivacaine (0.1%) with fentanyl (0, 2, or 4 μg/ml) was the predominant local anesthetic used for postoperative infusion before 1996. Ropivacaine (0.2%) with fentanyl (0, 2, or 4 μg/ml) has been used almost exclusively since 1996.

There have been two changes in type of skin preparation solution over the 16 yr: aqueous povidone iodine (1990–1993), alcoholic chlorhexidine (1994–2003), and alcoholic iodine (2004–2005). This information along with the annual rate of insertion site infection is shown in figure 3.

Two spinal hematomas (1:4,105 or 0.024%; 95% CI, 0–0.06%) were diagnosed, although only one of these was a true epidural hematoma (table 2). Neither of these required surgical intervention or resulted in adverse sequelae. The clinical indications for imaging on suspicion of spinal hematoma were unexpectedly prolonged (not resolving over 12 h) leg weakness in 13 cases, one of which also had back and leg pain, and individual cases of leg pain, neck pain, and catheter displacement during full anticoagulation. The number of investigations requested has increased over the past 6 yr of the report period. Despite this, no case of epidural hematoma was identified over this time.

Low-molecular-weight heparins were introduced to our institution for thromboprophylaxis in high-risk patients (e.g. , orthopedics) in 1992. Since this time, the standard thromboprophylaxis regimen in our institution has been 5,000 U dalteparin once daily, commenced 6–12 h postoperatively.

Of concern is the risk that epidural catheters may become displaced at a time of peak anticoagulant activity. The accidental displacement rate for catheters has remained relatively constant at 9.9% over the 16-yr period. Displacement is more common for thoracic level catheters (T6–T11, 10.9%) than lumbar (L1–L4, 9.8%; P < 0.01, Fisher exact test). One patient with subarachnoid bleeding (patient 3) was associated with a catheter removal approximately 1 h after 5,000 U dalteparin subcutaneously.

We cross-checked our 8,210 APS patients with 115 epidural hematoma patients who were identified by the hospital medical records–based search over the same period. These patients were cross-referenced by Unit Record Number and again (separately) by surname with our APS database. There were no additional matches.

The combined risk of serious neuraxial complications (abscess or hematoma) was therefore 1:1,026 or 0.10% (95% CI, 0.04–0.19%). The need for surgical intervention was 1:8,210 (0.01%; 95% CI, 0–0.07%). The decision to manage patients conservatively or surgically was based on the presence and progression of neurologic signs and was made in consultation with specialist neurologic, neurosurgical, and infectious disease clinicians. No patient sustained any permanent neurologic deficit.

We found the incidence of major adverse neuraxial events related to epidural catheter use for postoperative analgesia to be approximately 1:1,000. This is consistent with data reported from other large series.4,6The distinguishing aspect of this data is that it was collected entirely prospectively and from a single institution; therefore, data consistency, clinical practices, and true denominator information are known. Epidural abscess was diagnosed in 6 cases (0.07%) after 20 imaging investigations for this possibility, and was associated with epidural insertion site infection and pyrexia. Spinal hematoma was found in 2 cases (0.02%).

The early diagnosis of epidural abscess is important because delay in therapy can lead to permanent neurologic complications. In this series, the combination of epidural insertion site infection and pyrexia (> 38.5°C) occurred in 30 cases, 5 of which had an epidural abscess. The remaining patient with a confirmed epidural abscess had an epidural insertion site infection without pyrexia. A rapid passage to imaging in these cases, aided by early involvement of neurologic, neurosurgical, and infectious disease clinicians, meant that conservative management (4–7 months of antibiotic therapy) was effective in five patients. Imaging for abscess in our series occurred from 5 to 11 days after catheter placement. In the patient requiring surgical management (case 8), clinical suspicion (pyrexia, epidural insertion site infection, back pain) led to an initial MRI, which was reported (and reviewed) as normal (fig. 4A), and the patient was sent home. Two days later (11 days after insertion), this patient represented with neurologic signs. A repeat MRI was performed (fig. 4B), revealing a T1–T9 epidural abscess, which was surgically decompressed. This highlights the importance of continued follow-up, because the progression of epidural abscess to the point of clinical diagnosis typically takes days rather than hours.

Epidural insertion site infection is clearly a factor in epidural abscess formation. We have demonstrated a significant association with duration of insertion (OR, 1.4), as has been observed by others,1and upper body procedures (abdominal or thoracic) (OR, 3.3). It should be noted that sterile insertion techniques were used for all epidural catheter insertions over the entire 16 yr. This involved sterile hand scrub, sterile gown, gloves, and mask with skin preparation as noted and sterile draping. The change to alcohol-based skin preparation solutions from 1994 was based on advice from our microbiology department that alcohol-based antiseptics were more effective than aqueous solutions. This has been confirmed in recent investigations.9 

The absence of diagnosed epidural abscesses, but not epidural hematoma, before the year 2000 may relate to a number of factors. Insertion site infection rates were not dramatically different over this period, and it is more likely that a decreased threshold for clinical concern coupled with the increased accessibility of MRI scans contributed to the increased number of scans performed, and this in turn may have led to an increase in diagnosis of epidural abscess. Although MRI scanning has been available at our institution since 1991, newer scanners since 1996 have brought increased efficiency, image resolution, and accessibility. Figure 2demonstrates the steady yearly increase in the number of MRI examinations performed, especially in the past 2 yr. MRI is currently the investigation of choice in cases of suspected epidural abscess or hematoma.5,10To minimize the possibility of having missed even one case of epidural abscess during this time because of the impact that this would have on the overall incidence in this series, we accessed our hospital medical records database. The search we performed retrieved disease-group coding for both epidural abscesses and hematomas from January 1990 to December 2005. We found no missed events from this search.

We consider the possibility that heightened vigilance in response to literature reports in the late 1990s of the relatively high frequency of epidural hematoma and abscess resulted in the detection of epidural space infections that may not have proceeded to neurologically threatening abscesses. The rarity of neurologic symptoms at the time of diagnosis and the response to conservative therapy in our series lends support to this theory. In the early to mid 1990s, epidural insertion site infections were often caused by flucloxacillin-sensitive Staphylococcus aureus  species, and it is possible that treatment of these site infections may also have inadvertently treated deeper infections that were not investigated by CT or MRI scans at that time. Finally, the infrequency of these events underlines the value of a large series such as is presented here, because a sample over even 5 yr may significantly underestimate or overestimate the overall incidence.

The clinical criteria for which an MRI scan is performed are important, as discussed above. Royakkers et al.  advocated MRI scans in any patient after epidural who develops back pain with evidence of local and systemic infection, regardless of the presence or absence of neurologic signs.5We would support this principle, but suggest that more selective criteria would be more clinically reasonable. That is, the combination of epidural insertion site infection plus pyrexia would seem to be a sound basis for investigation with an MRI, and the presence of any third factor (including, of course, neurologic changes or back pain) would mandate prompt investigation. Although pyrexia after surgery has a number of causes and these should be investigated, it remains an important clinical sign in conjunction with an insertion site infection. The health economic impact of spinal cord compression from epidural abscess is considerable, and therefore, the use of MRI scanning in cases where a high index of clinical suspicion exists is justified—this would have resulted in less than two additional scans per year in our series based on the above criteria.

The introduction of low-molecular-weight heparin since 1992 has been identified as a factor in increasing the incidence of postoperative epidural hematoma.11We identified only one intraspinal hematoma associated with low-molecular-weight heparin in this period, although there has clearly been an increase in diagnostic investigations for problems such as persistent motor blockade on suspicion of neuraxial complications.

In conclusion, over 16 yr and 8,210 cases, we have identified a 1:1,368 incidence of epidural abscess and a 1:4,105 incidence of spinal hematoma. These incidences are consistent with current literature. Epidural abscesses were associated with epidural insertion site infection and pyrexia in all but one case. Catheter site infection was in turn associated with abdominal or thoracic surgery and duration of epidural infusion. There were no long-term neurologic complications in any patient. The low rate of surgical intervention and lack of any long-term sequelae is the distinguishing feature of this series.

The authors thank Neil Liddell, Ph.D. (Department of Medical Imaging, St. Vincent’s Hospital Melbourne, Fitzroy, Victoria, Australia).