Improving Adherence to Practice Guidelines for Anticoagulation in Patients Receiving Neuraxial Anesthesia Using an Electronic Order Entry Alert System

Abstract

Objectives:

Epidural analgesia is frequently used to alleviate postoperative pain. Though rare, epidural hematoma continues to be a feared complication of neuraxial analgesia. The risk of epidural hematoma is likely increased when certain regimens are used for prophylaxis/treatment of venous thromboembolism. To help decrease the risk, we developed an alert in our electronic medical record to assist providers with adherence to published guidelines addressing neuraxial analgesia and anticoagulation.

Methods:

Patient data were collected retrospectively 3 months before and 3 months after the initiation of the computerized alert to assess the effectiveness of the alert. Patients were included if they had a procedure code associated with epidural analgesia. Pregnant patients and children were excluded. Type and frequency of antithrombotic medications were recorded for comparison to published practice guidelines.

Results:

Using Poisson regression to describe the data, patients with epidurals after the best practice alert observed a 61% decrease in the expected number of days of exposure to inappropriate doses of anticoagulation versus patients treated before implementation of the alert.

Conclusion:

Unapproved antithrombotic administration was significantly reduced after initiation of the alert system. This simple electronic alert was found to have a protective effect for patients receiving both anticoagulation and epidural analgesia.

The benefits of epidural analgesia are numerous. When considering thoracic trauma, certain types of surgery (i.e., abdominal, vascular, cardiac) and intractable angina pectoris, epidural analgesia results in superior pain control and patient satisfaction when compared with systemic medications.^1–3 Under some circumstances, epidural analgesia can reduce the incidence of postoperative pulmonary dysfunction, the need for postoperative mechanical ventilation, and the incidence of paralytic ileus.^1,2 Anesthesiologists often offer epidurals to hospitalized patients for posttraumatic and postoperative pain relief unless there are reasons that preclude this treatment. The contraindications to epidural analgesia include patient refusal, infection at the needle insertion site, coagulopathy or other bleeding disorder, intracranial hypertension, and severe aortic and mitral stenosis.¹ Epidural anesthesia is not without risk. Some complications from epidural analgesia depend on the medication used for pain relief. For example, local anesthetics administered into the epidural space can cause bradycardia and hypotension (from sympathetic blockade) and motor weakness. Opioids in the epidural space can lead to nausea, vomiting, pruritis, sedation, and respiratory depression.¹ Other possible complications include postdural puncture headache, infection, and a rare incidence of epidural abscess or epidural hematoma. Epidural hematoma is more likely to occur in patients with underlying coagulopathy.^1,4

Many of the same patients who receive neuraxial analgesia for postoperative pain control are also placed on chemical thromboprophylaxis to decrease the risk of developing deep venous thrombosis or pulmonary embolism. As Rosencher et al⁵ mentions, “roughly half of all patients not given thromboprophylaxis will develop venographically evident deep vein thrombosis after major orthopedic surgery, and 10% of those who do, will suffer a pulmonary embolism.” Also, “the risk [of hematoma] is increased 15-fold by concomitant use of anticoagulant therapy when appropriate precautions are not taken.⁵”

Specific risk factors exist for the development of an epidural hematoma related to neuraxial block. Medication-related factors include the concomitant use of anticoagulants, which are frequently administered to hospitalized patients. Procedure-related factors include both catheter insertion and removal, a traumatic procedure or multiple needle passes, the presence of blood in the catheter during insertion or removal, and the type of neuraxial block (indwelling epidural catheter > single-shot epidural block > single-shot spinal block). Patient-related factors include inherited coagulopathy and acquired coagulopathy (i.e., liver or renal failure).⁶

The commonly cited incidence in the literature of neuraxial hematoma is 1:150,000 for epidural technique and 1:220,000 for spinal technique.⁷ Despite the relatively small incidence, patients who develop a neuraxial hematoma often experience severe morbidity. The American Society of Regional Anesthesia (ASRA) has published evidence-based practice guidelines entitled “Regional Anesthesia In the Patient Receiving Antithrombotic or Thrombolytic Therapy.⁷” Owing to the rarity of spinal hematoma, randomized controlled trials would be difficult to conduct to establish guidelines, and thus their “consensus statements represent the collective experience of recognized experts in the field of neuraxial anesthesia and anticoagulation.⁷ These guidelines are based on case reports, clinical series, pharmacology, hematology, and risk factors for surgical bleeding.⁷” The most common agents given for chemical thromboprophylaxis against venous thomboembolism (VTE) are subcutaneous heparin and enoxaparin. The 2010 ASRA guidelines regarding postoperative dosing of these agents are summarized below:

For subcutaneous heparin⁷:

There is no contraindication to neuraxial techniques in patients receiving heparin, 5000 units subcutaneously twice daily, but the incidence of hematoma may be decreased by delaying administration onset until after the block and can be increased in debilitated patients after prolonged therapy.

The safety of neuraxial block in total daily doses greater than 10,000 units or more frequently than twice-daily dosing has not been established.

For subcutaneous enoxaparin⁷:

Twice-daily dosing is associated with an increased risk of spinal hematoma. The first dose of enoxaparin should not be given sooner than 24 hours postoperatively; and catheters should be removed before initiating dosing, with administration delayed for 2 hours after catheter removal.

With single-daily dosing, the first postoperative dose should be given no sooner than 6 to 8 hours after surgery, and the second postoperative dose should be given no sooner than 24 hours after the first dose. The catheter should be removed at least 10 to 12 hours after the last dose, with resumption of dosing occurring no sooner than 2 hours after catheter removal.

There are multiple risk factors for developing a neuraxial hematoma and many potential opportunities for prevention. Gupta⁸ outlined one such method when he described an electronic clinical decision support system that functioned to alert providers when an epidural solution was being ordered on a patient that was receiving an anticoagulant included on a master list of prohibited anticoagulants. This list had been established in conjunction with the pharmacy. In the 3 months before the warning system, Gupta noted that 213 epidurals were placed with 26 order conflicts. In a 3-month period after the alerts were established, 237 epidurals were placed with only 11 order conflicts. Of the 11 potential conflicts postwarning system establishment, most were situations when the epidural had already been removed, but the epidural solution order had not been discontinued yet.⁸

West Virginia University Hospital implemented a similar electronic Best Practice Alert (BPA) system to notify providers of anticoagulant dosing restrictions for patients with a neuraxial catheter. A chart review was conducted 3 months before and 3 months after implementation of the BPA to see if the frequency of prohibited dosing or prohibited anticoagulant administration could be reduced in patients with neuraxial catheters.

METHODS

A retrospective cohort study was undertaken to assess the effectiveness of the alert system. Patients were identified using procedural codes associated with epidural analgesia at West Virginia University Hospital for 3 months before and after the initiation of the BPA (April 20, 2013). Pregnant women and children were excluded. The BPA would appear when a physician initiated or modified an anticoagulant order for a patient with a documented epidural (Figs. 1–3). The BPA would require administrators to select an “Accept” or “Cancel” button before proceeding to the medication order screen. Both the Accept and Cancel buttons would make the BPA disappear and allow the ordering provider to order the anticoagulant at any dose without further barrier. For patients with epidural catheters, the BPA would be expected to fire anywhere between 1 and 5 times. This frequency was expected, as epidural catheters are usually in place for 1 to 5 days, and orders are most frequently initiated or adjusted on daily patient rounds and upon hospital admission. Documentation of an epidural catheter in the day of care flow sheet by the bedside nurse (usually entered at the time of placement or shortly after epidural placement) was used to trigger the BPA. The BPA would not become active if the epidural catheter was not documented in the flow sheet. The specifications for the BPA were designed by pharmacists at WVU, who specified the patient criteria for generating the alert (i.e., a patient with a documented epidural catheter) and defined which user tasks would fire the alert (i.e., ordering anticoagulant medication).

FIGURE 1.

Healthcare provider attempting to order anticoagulation for a patient with a neuraxial catheter in place.

FIGURE 3.

Healthcare provider attempting to order enoxaparin in a patient with a neuraxial catheter in place.

During the chart review, the type and number of doses of each anticoagulant were recorded. Other information collected during the chart review included the primary service of the patient with the catheter, the reason for neuraxial catheter placement, the number of days each catheter remained in place, and any bleeding complications. Demographic information including age, comorbid conditions, sex, home anticoagulants, and body mass index was also recorded (Table 1). A total of 85 patients met the inclusion criteria for the study, including 39 before the alert and 46 after the alert. Most patients had thoracic epidurals, with only 4 lumbar epidurals before the BPA and 12 after the BPA. The only reasons for having an epidural catheter in the study population were rib fractures and postoperative pain management. Postoperative pain management was the most common reason for epidural placement, with posttraumatic pain from rib fractures being the indication for epidural placement for 9 epidurals before the BPA and 8 epidurals after the BPA. Table 2 lists the multiple types of surgeries the patients included in the study underwent with the benefit of postoperative epidural analgesia. Nine patients were excluded from the study. For purposes of the statistical analysis, one patient was excluded owing to having an epidural for both indication categories (postoperative pain and rib fractures). The other 8 patients were excluded because they were not given any anticoagulation during their hospital stays; and thus, they had no risk of receiving an inappropriate dose of anticoagulant. No other anticoagulants besides heparin or enoxaparin were given to patients with neuraxial catheters during the study period. No bleeding complications occurred. The primary services included trauma surgery, general surgery, surgery-oncology, gynecology, urology, cardiothoracic surgery, and vascular surgery.

TABLE 1.

Patients' Demographics

	Before (n=37)	After (n=39)	P
Age	60.0 (±15.6)	60.2 (±12.6)	0.937*
Sex
Female	18(48.6%)	18(46.2%)	0.827†
Male	19(51.4%)	21 (53.8%)
Comorbidities
CAD§	9 (24.3%)	5 (12.8%)	0.196†
CHF‖	4 (10.8%)	4 (10.3%)	0.937‡
DM 2¶	9 (24.3%)	6(15.4%)	0.328†
COPD#/Emphysema	11 (29.7%)	5 (12.8%)	0.071†
Home anticoagulants
Aspirin and/or clopidogrel	13(35.1%)	12(30.8%)	0.686†
NSAID**	2 (5.40%)	7 (17.9%)	0.154‡
Warfarin	4 (10.8%)	2 (5.10%)	0.425‡

^*P values calculated with a t test for continuous data.

^†P values calculated with a t test for χ² for categorical data.

^‡P values calculated with a t test for Fisher exact test for categorical data.

^§Coronary artery disease.

^‖Congestive heart failure.

^¶Type 2 diabetes mellitus.

^#chronic obstructive pulmonary disease.

^**Nonsteroidal anti-inflammatory drug.

TABLE 2.

List of All Types of Surgeries Undergone By Patients With Epidurals for Postoperative Pain Management During the Entire Study Period

Surgeries with Epidural Pain Management*
Thoracic/Aortic aneurysm repair
Pancreatectomy
Nephrectomy
Exploratory laparotomy for abdominal trauma
Small/large bowel resection
Ventral hernia repair
Cystectomy
Lobectomy
Hysterectomy with bilateral salpingo-oophorectomy
Cholecystectomy
Splenectomy
Aortofemoral bypass
Adrenalectomy
Colocutaneous fistula repair
Vesicovaginal fistula repair
Hip fracture repair

^*Includes all types of surgeries performed on patients who received epidural catheters for postoperative pain management during the entire study period.

In the 3 months before the BPA, 1 patient required 2 epidural catheters for separate abdominal surgeries. The surgeries were performed for the same underlying disease process. All other patients with epidural catheters before the BPA were different patients. In the 3 months after the BPA, one patient required 2 epidural catheters for separate abdominal surgeries, which also addressed the same disease process. Besides these 2 epidural catheters, all other epidurals placed after the BPA were placed in separate patients. None of the patients included in the 3 months before the BPA required epidurals in the 3 months immediately after implementation of the BPA. For the purposes of the statistical evaluation, the patients with 2 epidurals during each time period were treated as though they had one epidural for the sum of days each epidural was in place. This data adjustment was considered statistically valid, as there was no reason to believe either patient was treated differently during the periods of time each of their separate epidurals were in place (i.e., same primary service and surgery for the same disease process). With regard to patient demographics, the average age of patients in the groups before and after the BPA was not significantly different. With regard to patient comorbid conditions, there were no significant differences between patients with epidurals included in the study before and after the BPA (Table 1).

The primary outcome of interest for the study was the number of days patients with epidural catheters were exposed to at least one inappropriate dose of anticoagulant. The statistical software used for data analysis was R version 3.1.2 (R Core Team 2014, R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org/). The effect of the BPA was investigated using Poisson regression to model the number of days patients with epidurals were exposed to 1 or more inappropriate doses of anticoagulant. A stepwise model selection was performed using AIC (Akaike Information Criteria) as the selection criteria. The candidate model was initially specified as alert status (before or after the BPA), age, body mass index, sex, anticoagulant type (enoxaparin or heparin), epidural type (thoracic or lumbar), and clinical indication (postoperative pain or rib fractures). After the selection procedure, only alert status (P = 0.018), age (P = 0.388), anticoagulant type (P < 0.001), and indication (P = 0.063) were included in the selected model. The coefficients included in the model are shown in Table 3.

TABLE 3.

Predictors for Inappropriate Anticoagulant Use in the Poisson Regression Model

		Coefficient	SE*	Mean Ratio	Lower	Upper
Alert	After versus before BPA†	−0.940	0.397	0.391	0.179	0.851
Age	One-year increase	−0.027	0.013	0.973	0.948	0.999
Anticoagulant	Enoxaparin versus heparin	−1.598	0.438	0.202	0.086	0.477
Indication	Rib Fx‡ versus postoperative pain	0.849	0.457	2.340	0.955	5.726

^*Standard error.

^†Best practice alert.

^‡Rib fractures.

RESULTS

Using the Poisson regression model, patients with epidural catheters placed after the BPA observed a 61% decrease in the expected number of days of exposure to inappropriate doses of anticoagulation versus patients treated before implementation of the alert. Regardless of the alert status, patients treated with enoxaparin observed a 79.8% decrease in the expected number of days of exposure to inappropriate doses of anticoagulation versus patients treated with heparin.

DISCUSSION

This study provides evidence that the BPA significantly reduced inappropriate anticoagulant use in patients with indwelling epidural catheters. In the study population, enoxaparin was administered more correctly more often than heparin to patients with epidural catheters. This observation may indicate that there is a greater knowledge of dosing enoxaparin to patients with epidural catheters in comparison to heparin. However, this finding may simply indicate that routinely administered doses of heparin for deep venous thrombosis prophylaxis are unapproved for patients with epidural catheters. Conversely, routinely administered doses of enoxaparin may be considered acceptable for patients with epidural catheters. Clinical decision support systems have shown to improve adherence to practice guidelines and to decrease ordering errors in multiple areas. As mentioned earlier, Gupta showed increased adherence to anticoagulation guidelines in patients with neuraxial anesthesia with a computerized alert system.⁸ With the exception of Gupta’s study, no similar studies were found with regard to use of a computerized physician order entry system to promote adherence to anticoagulant guidelines in patients with neuraxial catheters. As inappropriate anticoagulant use in patients with epidurals can increase the potentially devastating risk of epidural hematoma, the goal of increased adherence to practice guidelines regarding anticoagulants is a worthy goal.

Other arenas for effective alerts built into computerized physician order entry systems are described in the literature. Smith et al⁹ describes an alert built into a computerized physician order entry system at a multihospital center aimed at decreasing the incidence of inappropriate blood component transfusions including packed red blood cells and plasma. Alerts were triggered if packed red cells were ordered for patients with hemoglobin values greater than 8 g/dL or if plasma was ordered for patients with international normalized ratio (INR) values less than 1.6 (compatible with institutional guidelines of the involved hospitals). The alerts were not triggered for patients younger than 18 years or patients in the emergency department. Of the orders that triggered alerts, approximately 10% of packed red cell orders were canceled over a 15-month study period, and approximately 20% of plasma orders were canceled during a 10-month study period.⁹ Howell et al¹⁰ describes an alert built into a computerized physician order entry system to decrease the ordering of inappropriate Papanicolaou tests for cervical cancer screening. This alert was implemented in the primary care clinics and would display if a Papanicolaou test was ordered in a women younger than 21 years or older than age 70 years in hopes of improved adherence to the American College of Gynecology 2009 cervical cancer screening guidelines. Although providers were able to easily override these alerts, a significant decrease in the frequency of Papanicolaou tests ordered in these age groups was noted.¹⁰ Nachtigall et al¹¹ evaluated whether improved adherence to guidelines for antibiotic therapy in critically ill patients occurred after implementation of computer-assisted decision support system. This computer-assisted decision support system contained treatment guidelines for most infections a provider would encounter in the surgical intensive care unit. Based on information entered by a health care provider, diagnostic procedures were recommended and suggestions were made for empiric antibiotic therapy. This clinical decision support system also provided links to relevant reference materials. Implementation of this complex computer-assisted decision support system significantly improved adherence to guidelines from 61% before implementation (measured from January to April 2006) to 71% several months after implementation (measured from February to March 2010).¹¹

Medical systems with computerized order entry systems afford a unique opportunity to increase awareness and to prevent common errors. The “To Err is Human” report describes 4 tiers in the approach to addressing the culture of patient safety. As included in the report, 1 of the 4 tiers is “creating safety systems inside health care organizations through the implementation of safe practices at the delivery level. This level is the ultimate target of all the recommendations.¹²” This report also mentions that medication errors occur frequently in hospitals, and although many of these errors do not result in harm, the errors that result in harm are costly.¹²

In addition to having a direct protective effect, the BPA likely increased awareness to unapproved dosing regimens for heparin and enoxaparin in patients receiving neuraxial analgesia. Because WVUH is a teaching facility, residents frequently place orders for medications to provide prophylaxis against venous thromboembolism. Nonanesthesiology residents are not typically aware (at least initially) of the specific practice guidelines regarding anticoagulants and neuraxial catheter placement. Owing to the large number of resident providers at WVUH and to the frequency with which resident providers transition from one teaching service to another, directing structured educational programs toward residents responsible for ordering anticoagulants on patients with epidural catheters is a difficult task. In addition, patients with epidurals appear on a wide variety of surgical and medical services. However, education can significantly reduce the incidence of medical errors. A study performed by Campino et al¹³ demonstrated that the use of multiple informational sessions given by hospital pharmacists significantly reduced the incidence of medication errors in a neonatal intensive care unit. This neonatal care unit included 5 pediatric residents. These informational sessions were created after a pilot study identified the portions of the drug administration process that lead to the most errors.¹³ In addition, a study conducted by Alagha et al¹⁴ in the pediatric intensive care unit in Egypt showed that resident physician education that included the provision of a structured medication chart significantly reduced common medication errors detected during thorough chart reviews. These medication charts were designed by pharmacists.¹⁴ Therefore, a future project for the anesthesiology residency program at WVUH will be to conduct educational sessions to large groups of incoming residents and to provide laminated charts showing appropriate dosing regimens of heparin and enoxaparin in patients with epidurals. The charts will also include the minimum time that should elapse between the last dose of heparin and enoxaparin administered and epidural catheter placement and removal, and also the time at which heparin and enoxaparin can be restarted after epidural catheter placement and removal. As residents at WVUH from all departments are now required to participate in quality improvement projects to graduate from residency, which is one of many ways WVUH addresses improved patient safety, it will not be hard to find residents willing to undertake this project.

One limitation of the BPA is that it requires documentation of an epidural catheter in the day of care flow sheet to become an active alert. In this study, it is possible that some patients did not have an epidural documented appropriately. If so, an ordering provider would not have received an alert. It is not known what effect, if any, this had on the results. Different triggers for the BPA may exist that are less prone to provider documentation error. For example, the epidural medication infusion order could trigger the alert system, similar to the study performed by Gupta.⁸ It is rare for a patient to have an indwelling epidural catheter and no epidural infusion ordered. One possible exception is a patient whose epidural infusion is stopped owing to persistent hypotension. Creating a fail-safe trigger for the BPA represents an opportunity for improved safety. A further limitation of the BPA is that it can be overridden easily, i.e., the ordering physician may still order the inappropriate medication. Although restrictive to physician practice, it should be possible to develop a system that cannot be easily overridden.

Although this study was small, medication errors with regard to heparin and enoxaparin and patients with epidural catheters were detected at an unacceptable rate. Even after the alert, although with a reduced frequency, errors in dosing anticoagulants in patients with epidural catheters still occur. There are further opportunities to reduce the frequency of these medication errors. Unfortunately, health care systems that do not have computerized order entry systems in place will not be able to adopt the system used in this study to alert providers about patients with epidural catheters and anticoagulant guidelines. However, this study may provide education regarding potential errors that occur in all health care systems that offer patients epidural catheters.

CONCLUSIONS

Patients with neuraxial catheters require special precautions with respect to chemical thromboprophylaxis. Practice guidelines are available that address this topic, but they are not widely appreciated outside of the anesthesiology community. Computerized physician order entry systems afford a unique opportunity to improve compliance with evidence-based practice guidelines. This study shows that guideline adherence can be improved through the implementation of a simple electronic alert system. The alert system reduced, but did not eliminate, the risk of unapproved anticoagulant usage in patients with epidural catheters. Fail-safe alerts within that electronic medical record that are not easily overridden may improve compliance further.

FIGURE 2.

Healthcare provider attempting to order heparin for a patient with a neuraxial catheter in place.