Does Operative Management of Epidural Abscesses Increase Healthcare Expenditures up to 1 Year After Treatment?

Abstract

Background

The incidence of spinal epidural abscesses is increasing. What is more, they are associated with high rates of morbidity and mortality. Advances in diagnostic imaging and antibiotic therapies have made earlier diagnosis and nonoperative management feasible in appropriately selected patients. Nonoperative treatment also has the advantage of lower immediate healthcare charges; however, it is unknown whether initial nonoperative care leads to higher healthcare charges long term.

Questions/purposes

(1) Does operative intervention generate higher charges than nonoperative treatment over the course of 1 year after the initial treatment of spinal epidural abscesses? (2) Does the treatment of spinal epidural abscesses in people who actively use intravenous drugs generate higher charges than management in people who do not?

Methods

This retrospective comparative study at two tertiary academic centers compared adult patients with spinal epidural abscesses treated operatively and nonoperatively from January 2016 through December 2017. Ninety-five patients were identified, with four excluded for lack of billing data and one excluded for concomitant intracranial abscess. Indications for operative management included new or progressive motor deficit, lack of response to nonoperative treatment including persistent or progressive systemic illness, or initial sepsis requiring urgent source control. Of the included patients, 52% (47 of 90) received operative treatment with no differences in age, gender, BMI, and Charlson comorbidity index between groups, nor any difference in 30-day all-cause readmission rate, 1-year reoperation rate, or 2-year mortality. Furthermore, 29% (26 of 90) of patients actively used intravenous drugs and were younger, with a lower BMI and lower Charlson comorbidity index, with no differences in 30-day all-cause readmission rate, 1-year reoperation rate, or 2-year mortality. Cumulative charges at the index hospital discharge and 90 days and 1 year after discharge were compared based on operative or nonoperative management and secondarily by intravenous drug use status. Medical records, laboratory results, and hospital billing data were reviewed for data extraction. Demographic factors including age, gender, region of abscess, intravenous drug use, and comorbidities were extracted, along with clinical factors such as symptoms and ambulatory function at presentation, spinal instability, intensive care unit admission, and complications. The primary outcome was charges associated with care at the index hospital discharge and 90 days and 1 year after discharge. All covariates extracted were included in this analysis using negative binomial regression that accounted for confounders and the nonparametric nature of charge data. Results are presented as an incidence rate ratio with 95% confidence intervals.

Results

After adjusting for demographic and clinical variables such as age, gender, BMI, ambulatory status, presence of mechanical instability, and intensive care unit admission among others, we found higher charges for the group treated with surgery compared with those treated nonoperatively at the index admission (incidence rate ratio [IRR] 1.62 [95% CI 1.35 to 1.94]; p < 0.001) and at 1 year (IRR 1.36 [95% CI 1.10 to 1.68]; p = 0.004). Adjusted analysis also showed that active intravenous drug use was also associated with higher charges at the index admission (IRR 1.57 [95% CI 1.16 to 2.14]; p = 0.004) but no difference at 1 year (IRR 1.11 [95% CI 0.79 to 1.57]; p = 0.55).

Conclusion

Multidisciplinary teams caring for patients with spinal epidural abscesses should understand that the decreased charges associated with selecting nonoperative management during the index admission persist at 1 year with no difference in 30-day readmission rates, 1-year reoperation rates, or 2-year mortality. On the other hand, patients with active intravenous drug use have higher index admission charges that do not persist at 1 year, with no difference in 30-day readmission rates, 1-year reoperation rates, or 2-year mortality. These results suggest possible economic benefit to nonoperative management of epidural abscesses without increases in readmission or mortality rates, further tipping the scale in an evolving framework of clinical decision-making. Future studies should investigate if these economic implications are mirrored on the patient-facing side to determine whether any financial burden is shifted onto patients and their families in nonoperative management.

Level of Evidence

Level III, therapeutic study.

Introduction

Background

Spinal epidural abscesses are an increasingly common and devastating condition [4, 5, 7], with a rising prevalence because of the aging population and increased intravenous drug use (IVDU) [9]. They pose a particular challenge to the healthcare system because they require multidisciplinary care, can be refractory to multiple lines of treatment, and often occur among the underinsured population [4, 5, 7, 9, 26]. Treatment is also costly, with the mean charge for hospital admission for spinal epidural abscesses estimated at nearly USD 160,000 [26].

Rationale

As spinal epidural abscesses have become more common, there has been increasing debate regarding the pursuit of operative versus nonoperative treatment [10, 20, 27, 34]. For decades, urgent surgical decompression was viewed as paramount [2, 15]; however, advances in MRI, interventional procedures to identify pathogens, and improvements in antibiotic therapy have facilitated successful nonoperative care. Spinal epidural abscesses still result in substantial morbidity and mortality, with recent studies suggesting that 10% to 20% of patients with this diagnosis will die during treatment, and 30% to 50% will experience permanent neurologic deficits [1, 7, 18, 24, 26]. Even with this morbidity, however, nonoperative treatment has been successful [18] and is increasing in popularity [16, 18, 27]. In other areas of medicine where the paradigm shift has trended toward nonoperative management, such as with appendicitis, economic analyses have found increased expenditures associated with nonoperative management due to the increased monitoring and outpatient interventions required [25, 28]. It is unknown whether the rising prevalence of nonoperative management of spinal epidural abscesses may similarly be inadvertently increasing healthcare charges, and if these differences persist in long-term treatment.

In addition to surgical management, IVDU is another factor whose effect on long-term charges associated with the care of spinal epidural abscesses is unknown. Intravenous drug users tend to be underinsured, and care of IVDU-associated infections is associated with financial costs to healthcare systems [11, 29]. It is unknown if charges specific to spinal epidural abscesses are higher in patients with active IVDU and whether these differences persist in long-term management.

Research Questions

We therefore asked: (1) Does operative intervention generate higher charges than nonoperative treatment over the course of 1 year after the initial treatment of spinal epidural abscesses? (2) Does the treatment of spinal epidural abscesses in people who actively use intravenous drugs generate higher charges than management in people who do not?

Patients and Methods

Study Design and Setting

The clinical data for this retrospective comparative study were accessed from the research patient data registry of Mass General Brigham, a tertiary urban referral network. This data repository automatically captures sociodemographic and clinical details for all patients presenting for care in the healthcare system and has been used to study aspects of orthopaedic healthcare policy [6, 32]. Furthermore, the research patient data registry is linked to the National Death Index and therefore captures patient deaths, even if the patients’ terminal care was not delivered at our institution.

Study Participants

We identified patients who underwent treatment for spinal epidural abscesses from January 2016 to December 2017. The analysis was restricted to these years because comprehensive charge data are not present in our clinical registry before 2016. We included adult patients aged 18 years and older with an inpatient hospital admission in which a spinal epidural abscess was one of the primary diagnoses. Patients were excluded if no billing data were available, if the record review failed to confirm the diagnosis of a spinal epidural abscess, or if an intracranial abscess was present [10]. After confirming the patients’ eligibility, we reviewed the records to obtain data on demographic factors, clinical factors, and billing based on charges specific to that patient and encounter (Fig. 1). Indications for operative management at the two institutions included new or progressive motor deficit, lack of response to nonoperative treatment including persistent or progressive systemic illness, or initial sepsis requiring urgent source control [33].

Fig. 1.

Study inclusion flowchart.

Study Variables

Sociodemographic factors extracted included age at the initial presentation, gender, race, BMI, comorbidities as defined by the Deyo-modified Charlson criteria [8], active IVDU at the time of admission, and primary insurance type. Clinical factors extracted included the abscess location (cervical, thoracic, lumbar, or sacral), infectious pathogens, antibiotic regimen and duration of treatment, ambulatory status and symptoms on presentation, intensive care unit admission, discharge disposition, and operative versus nonoperative management, as well as any changes in treatment plan. A change in management was defined as a change from nonoperative to operative management after an initial 72-hour treatment period. We further observed patients for any readmission within 30 days of hospital discharge.

Symptoms on presentation were categorized as asymptomatic, axial pain only, radicular pain (for example, neck or back pain with additional radicular arm or leg pain, without sensorimotor impairment), or paresis or paralysis (defined as sensory or motor deficits for the bilateral upper extremities, bilateral lower extremities, or a hemiparetic distribution). Bowel or bladder dysfunction was categorized as paresis or paralysis. Neurologic decline as evidenced by progression from one category to a more severe category was also recorded.

Billing data were extracted based on actual charges specific to the patient and encounter. Charges for the total index admission were recorded, including surgical charges (operating room and anesthesia charges as well as surgical implant and disposable supplies). All-cause charges from the index discharge date to 90 days after discharge were recorded in a similar fashion, which also included any readmission charges and, if applicable, charges associated with revision or planned repeat surgical procedures. Related charges were recorded from 90 days to 1 year after the index discharge date. These were defined as any encounter related to the management of the spinal epidural abscesses, such as revision spine surgery, readmission related to the spinal epidural abscesses, spine clinic visit, infectious disease visit, or primary care physician visit in which primarily spinal complaints were discussed and/or spinal imaging was ordered. As a tertiary care facility, the parent institution often discharges patients to facilities outside the network without available billing data. Therefore, national averages for rehabilitation and skilled nursing facility stays after complex spinal surgery were used for patients, depending on their discharge disposition [23].

Cost estimates for antibiotics were calculated by multiplying the dosage and duration of treatment by the per-dose wholesale acquisition cost [31]. Per-dose costs were derived by averaging the minimum wholesale acquisition cost from each manufacturer of a given antibiotic as reported in Red Book (IBM Corp). Dosage amounts and treatment lengths were extracted from clinical documentation at the time of discharge. Antibiotic charges were included in the total estimates of 30-day, 90-day, and 1-year charges, regardless of discharge disposition. For patients who were discharged home with home services for parenteral antibiotic therapy, daily estimates from a previous study [19] were added for direct nonmedication charges, including medication supplies and home nursing visits.

All USD amounts were converted to the 2019 equivalent before analysis based on the United States Bureau of Labor Statistics Consumer Price Index [34]. Amounts extracted from a previous study [18] were adjusted based on the average of Consumer Price Index values during the study period.

Bias

There are several possible sources of bias in this study. First, surgeons may have selected criteria for operative management unrelated to the underlying spinal pathology that posed a selection bias for magnitude of charges. These might include bias for or against selecting operative management for certain individuals based on insurance type, race, or IVDU or other social situations. The operative and nonoperative groups showed no difference in any demographic factors, thus decreasing the likelihood of selection bias, although some element may still exist. Second, there was potential for transfer bias in that individuals who underwent nonoperative treatment may have been less likely to follow up within the health system, and that individuals with active IVDU with unstable social situations may have also been less likely to follow up with the health system. Nonetheless, analysis demonstrated no difference in the proportion of patients who completed follow-up at 1 year between patients managed operatively versus nonoperatively and between patients with and without active IVDU. Third, there was potential for assessment bias in the outcomes measured. This study used charges for several reasons, chiefly that it was most likely to be consistent across payors and hospitals and that it would be less likely to be affected by patients who were uninsured without financial support. In this study, charges represented the projected, but not actual, financial impact of providing care during the study. This does not, however, provide a proxy for cost and healthcare systems and providers should not interpret the results of this study as being representative of cost. The use of charges should also not be taken out of the context of clinical care. The outcomes assessed by this study, including 30-day all-cause readmission rate, 1-year reoperation rate, and 2-year mortality rate, were equivalent between patients treated operatively and nonoperatively as well as those with or without active IVDU. Nevertheless, we acknowledge that these are crude measures of clinical outcomes. An ideal analysis would be able to measure disability- or quality-adjusted life years in conjunction with charges; however, to our knowledge, such measurement metrics have not been developed with respect to spinal epidural abscesses. Validation of such tools is a goal for future study.

Primary and Secondary Study Outcomes

Our primary study goal was to assess differences in charges associated with care between patients managed operatively and nonoperatively for spinal epidural abscesses during the index hospital admission, at 90 days after discharge, and 1 year after discharge.

Our secondary study goal was to assess differences in charges associated with care between patients with and without active IVDU being treated for spinal epidural abscesses during the index hospital admission, at 90 days after discharge, and 1 year after discharge.

To answer both of these questions, we compared cumulative charges at the index admission, at 90 days, and at 1 year by group while adjusting for demographic and clinical covariates.

Sample Characteristics

Ninety-five patients underwent inpatient management for a spinal epidural abscess during the study period, four of whom were excluded because of a lack of billing data and one patient was excluded for simultaneous management of an intracranial abscess. Ninety patients met the inclusion criteria, of whom 52% (47) underwent operative management and 48% (43) underwent nonoperative management. Patients had a mean age of 57 ± 16 years overall; 60% (54 of 90) were men, and the mean BMI was 28.6 ± 8.0 kg/m².

Between patients managed operatively and nonoperatively, there were no differences in demographics such as age, gender, BMI, Charlson comorbidity index, race, smoking status, and insurance type (Table 1). There were also no differences in pretreatment ambulatory status, anatomic region, length of stay, 30-day readmission rate, 1-year reoperation rate, 2-year mortality rate, complication rate, and rate of completed 1-year follow-up between operatively and nonoperatively managed patients. Patients undergoing surgical management, however, were more likely to present with paresis or paralysis than those undergoing nonsurgical treatment (45% [21 of 47] versus 21% [9 of 43]; p = 0.04), pathologic fracture (17% [8 of 47] versus 2% [1 of 43]; p = 0.02), and require an ICU stay during admission (51% [24 of 47] versus 19% [8 of 43]; p = 0.001) (Table 1).

Table 1.

Patient characteristics based on operative or nonoperative management

Parameter	Operative (n = 47)	Nonoperative (n = 43)	p value
Age in years	58.0 ± 15.2	56.6 ± 16.0	0.51
Gender
Men	57 (27)	63 (27)
Women	43 (20)	37 (16)	0.61
BMI in kg/m2	28.5 ± 8.0	28.7 ± 8.0	0.85
Charlson comorbidity index score	3.0 ± 2.2	3.4 ± 2.5	0.42
Race or ethnicity			0.67
Asian	0 (0)	2 (1)
Black	9 (4)	9 (4)
White	83 (39)	72 (31)
Hispanic or Latino	6 (3)	9 (4)
Unknown	0 (0)	2 (1)
Other	2 (1)	5 (2)
Smoking status	49 (23)	53 (23)	0.67
Insurance type			0.28
Private	47 (22)	30 (13)
Medicare	19 (9)	30 (13)
Medicaid	34 (16)	37 (16)
Other	0 (0)	2 (1)
Intravenous drug use	26 (12)	33 (14)	0.46
Pretreatment ambulatory function			0.63
Independent ambulation	83 (39)	79 (34)
Dependent ambulation	13 (6)	12 (5)
Nonambulation	4 (2)	9 (4)
Patient complaints pretreatment			0.04
Axial pain	40 (19)	65 (28)
Radicular	15 (7)	14 (6)
Paresis or paralysis	45 (21)	21 (9)
Pathologic fracture or spinal instability	17 (8)	2 (1)	0.02
Region			0.07
Cervical	34 (16)	12 (5)
Thoracic	21 (10)	28 (12)
Lumbar	45 (21)	58 (25)
Sacrum	0 (0)	2 (1)
Pathogen (may add up to more than 100% due to multiple organisms)
Methicillin-sensitive Staphylococcus aureus	64 (30)	42 (18)	0.04
Methicillin-resistant S. aureus	11 (5)	16 (7)	0.43
Escherichia coli	2 (1)	7 (3)	0.27
Streptococcus species	11 (5)	9 (4)	0.83
Polymicrobial	11 (5)	5 (2)	0.29
Other	11 (5)	16 (7)	0.43
Intensive care unit stay during hospital stay	51 (24)	19 (8)	0.001
Length of stay in days	13 ± 16	8 ± 14	0.4
Two-year mortality	11 (5)	16 (7)	0.43
Complications	30 (14)	26 (11)	0.71
Change in management	23 (11)	0 (0)	< 0.001
Thirty-day all-cause readmission	17 (8)	21 (9)	0.64
Patients with 1-year follow-up	72 (34)	63 (27)	0.33
Surgery after index admission	15 (7)	9 (4)	0.42

Data presented as mean ± SD or % (n).

Charges for antibiotics, rehabilitation or facility stays, outpatient charges, and readmission charges were not different between the two groups at any timepoint (Table 2). The number of days of parenteral antibiotic therapy was greater in patients treated nonoperatively than in those treated operatively (44 days versus 33 days; p = 0.03).

Table 2.

Charges based on operative versus nonoperative management

Parameter	Operative (n = 47)	Nonoperative (n = 43)	p value
Length of stay in days	16 ± 13	14 ± 8	0.40
Inpatient cost per day	20,408 ± 7560	12,325 ± 5930	< 0.001
Initial admission charges	313,650 ± 299,823	161,988 ± 148,855	< 0.001
Thirty-day charges
Outpatient charges	22,584 ± 44,702	14,896 ± 9940	0.06
Readmission charges	20,823 ± 68,631	19,765 ± 47,569	0.98
Total antibiotic medication costs	3418 ± 2605	4054 ± 5326	0.54
Charges excluding admission	43,407 ± 78,714	34,661 ± 47,210	0.39
Total charges	357,057 ± 301,449	196,649 ± 152,277	< 0.001
Ninety-day charges
Outpatient charges	30,446 ± 45,094	23,368 ± 15,332	0.19
Readmission charges	31,573 ± 83,193	41,360 ± 78,234	0.84
Total antibiotic medication costs	5320 ± 5949	6281 ± 7812	0.57
Charges excluding admission	62,019 ± 90,478	64,728 ± 82,322	0.87
Total charges	375,669 ± 307,971	22,6716 ± 168,037	< 0.001
One-year charges
Outpatient charges	38,867 ± 56,906	26,937 ± 17,728	0.08
Readmission charges	67,923 ± 174,203	171,041 ± 704,079	0.39
Total antibiotic medication costs	5618 ± 7547	6805 ± 8138	0.51
Charges excluding admission	106,790 ± 179,979	197,979 ± 707,003	0.04
Total charges	420,440 ± 357,723	359,966 ± 731,814	0.37

All charges are presented in USD in 2019-equivalent dollars. All variables are expressed as mean ± SD.

Between patients with an active history of IVDU versus no history, there were no differences in gender or race. Patients with a history of IVDU were likely to be younger (44 ± 12 years versus 63 ± 13 years; p < 0.001), with a lower BMI (26.0 ± 6.4 kg/m² versus 29.6 ± 8.4 kg/m²; p = 0.02), lower Charlson comorbidity index (1.7 ± 2.3 versus 3.8 ± 2.1; p < 0.001), higher proportion of smokers (88% [23 of 26] versus 36% [23 of 64]; p < 0.001), and more likely to have Medicaid (88% [23 of 26] versus 14% [9 of 64]; p < 0.001).

There were no differences in clinical factors between the groups, including pretreatment ambulatory status, pretreatment sensorimotor deficits, presence of concomitant pathologic spinal fracture or instability, anatomic abscess region, ICU stay during admission, length of stay, 30-day all-cause readmission rate, 1-year reoperation rate, 2-year mortality, complication rate, rate of completed 1-year follow-up, and change from initial nonoperative to operative management (Supplemental Table 1; http://links.lww.com/CORR/A624).

Ethical Approval

This study was approved the institutional review board at Mass General Brigham, Somerville, MA, USA (number 2021P000050).

Statistical Analysis

We performed unadjusted analyses using the Wilcoxon rank-sum test for continuous data or chi-square tests for categorical data to evaluate nonparametric data for differences in group means according to management type. We conducted a multivariable analysis using charges as the continuous dependent variable with the type of management (operative versus nonoperative) as the primary predictor. We performed a second analysis with the presence or absence of IVDU as the primary predictor. Negative binomial regression was used to account for confounders and the nonparametric nature of charge data, with demographic and clinical factors considered as covariates. Results are presented using incidence rate ratios (IRR) and 95% confidence intervals. The IRR and 95% confidence intervals exclusive of 1.0, with a p value < 0.05, were defined a priori as statistically significant. Statistical testing was conducted using Stata version 16.1 (StataCorp).

Results

Comparing Charges: Operative Versus Nonoperative Management

After adjusting for demographic and clinical variables such as age, gender, BMI, ambulatory status, presence of mechanical instability, length of stay, intensive care unit admission, and other factors (Table 3), we found higher charges for the group treated with surgery compared with those treated nonoperatively at the index admission (IRR 1.62 [95% CI 1.35 to 1.94]; p < 0.001) (Table 3) and for total charges at 1 year (IRR 1.36 [95% CI 1.10 to 1.68]; p = 0.004) (Fig. 2). There were no differences in 30-day readmission rates or 2-year mortality rates.

Table 3.

Adjusted analysis of charges based on operative management versus nonoperative management

Parameter	IRR (95% CI)	p value
Adjusted index admission charges	1.62 (1.35-1.94)	< 0.001
Adjusted total 30-day charges	1.54 (1.29-1.82)	< 0.001
Adjusted total 90-day charges	1.41 (1.17-1.69)	< 0.001
Adjusted total 1-year charges	1.36 (1.10-1.68)	0.004
Adjusted 1-year charges exclusive of index admission	1.05 (0.55-2.00)	0.89

An IRR greater than 1 indicates surgical management had a higher value; an IRR less than 1 indicates nonoperative management had a higher value; all values have been adjusted for demographic factors, as well as preadmission ambulatory status, symptoms on presentation, mechanical instability, spinal anatomic region, intensive care unit stay during the index hospital stay, 1-year mortality, presence of complications, change from nonoperative to operative management, and 30-day readmission rate; IRR = incident rate ratio.

Fig. 2.

This graph shows 1-year cumulative charges, comparing operatively and nonoperatively treated patients. At 1 year, the higher charges associated with the index admission in patients undergoing operative management persisted to increased cumulative charges at 1 year, despite similar or even elevated charges after admission. A color image accompanies the online version of this article.

Comparing Charges: Patients Who Used Intravenous Drugs Versus Those Who Did Not

After adjusting for demographic and clinical variables such as age, gender, BMI, ambulatory status, presence of mechanical instability, length of stay, intensive care unit admission, and other factors, patients with active IVDU had higher charges at the index admission (IRR 1.57 [95% CI 1.16 to 2.14]; p = 0.004) but no difference at 1 year (IRR 1.11 [95% CI 0.79 to 1.57]; p = 0.55) (Supplemental Table 2; http://links.lww.com/CORR/A625). There were no differences in 30-day readmission rates or 2-year mortality rates.

Discussion

Spinal epidural abscesses are increasingly common and associated with high rates of acute mortality and long-term morbidity. They are also expensive to treat, with costs of care ranging from USD 100,000 to USD 460,000 per episode [12, 26]. Recently, nonoperative management in select patients has gained popularity to avoid the associated morbidity of surgery [17]. A prior study found that in properly indicated patients, the long-term outcomes of nonoperative treatment for spinal epidural abscesses are no different from those after surgical intervention [18]. Appropriately selected nonoperative care might amplify cost savings if surgery can be avoided successfully. To our knowledge, the long-term ramifications of the initial operative or nonoperative treatment for spinal epidural abscesses regarding healthcare charges have not been explored. The results of our analysis indicate that surgical intervention is associated with greater charges compared with nonoperative management over the course of 1 year after the index admission without differences in readmission rates, reoperation rates or conversion to the operative group, or 2-year mortality rates. Intravenous drug use, on the other hand, was associated with increased charges initially, but no difference in cumulative charges at 1 year. This suggests that nonoperative treatment can be undertaken without concern that this will lead to increased long-term charges and provides additional support for the paradigm shift toward nonoperative management in indicated patients. This also suggests that healthcare systems should not shy away from the care of intravenous drug users as their immediate increase in healthcare charges does not persist long term.

Limitations

We recognize several limitations to this study. First, this was a retrospective analysis at two tertiary referral centers, which may be subject to regional and institution-specific care delivery and financial practices. As previously mentioned, there was some potential for selection bias in treatment selection, which we investigated and found no baseline differences between patients managed operatively and nonoperatively. Nonetheless, there may be unaccounted factors that pose a selection bias, such as the time of patient presentation, which were not investigated in this study. The group with patients using intravenous drugs was younger and healthier than the group without IVDU, and although we adjusted for clinical and demographic factors in our definitive analysis, it is also possible that this group did not persist in having higher charges at 1 year because they had fewer comorbidities that accumulated charges.

There was also an element of assessment bias. We accounted for most direct charges, but there are many aspects of patient care that could have been overlooked in the analysis (such as institutional operational and facilities charges), in addition to the burdens patients encounter (such as time and financial burden associated with transportation to and from medical appointments). The charges associated with certain aspects of care such as nursing facility dispositions and home healthcare charges were extrapolated from a previous study [18]. We also used certain parameters such as intensive care admission as a proxy for severe systemic illnesses like septic shock because there may be many reasons for intensive care unit admission, and this was most likely to directly correspond to any increases in hospital charges. Furthermore, we did not account for the presence of concomitant spinal osteomyelitis or discitis. This was due to the fact that the duration of intravenous antibiotics is more likely to be dictated and extended by the presence of an epidural abscess rather than the presence of osteomyelitis and discitis. Although formal consensus guidelines from infectious disease societies exist regarding duration of treatment only for osteomyelitis and discitis [21] and not for epidural abscesses, informal recommendations in the literature note treatment of epidural abscesses to be at least as lengthy, if not greater, than those for osteomyelitis [33]. For the purposes of this study, then, we may assume that co-occurrence of osteomyelitis or discitis did not have a clinically meaningful effect on treatment duration. Similarly, we also did not assess for myelopathy since in this retrospective review, documentation of myelopathy was not always available in the acute setting, and typical acute neurologic deficits were used for clinical decision making. This presents one of the challenges of the retrospective design and could be rectified in a prospective study.

We also did not have data on patients who may have chosen to receive subsequent care at a different facility, thus raising the question of transfer bias. Nevertheless, the proportion of patients who completed follow-up at 1 year was not different between groups, and the nature of our integrated electronic medical record and the research patient data registry allows for a comprehensive oversight of healthcare encounters that occur in our healthcare system, even if these encounters are not in the specific facility where patients received their initial care. This minimizes the number of patients lost to surveillance and allows for a complete evaluation of mortality, which was low in this group of patients who survived the initial hospitalization (approximately 4% at 1 year). Nonetheless, we recognize that our modalities for surveillance are not exhaustive. Finally, this study only examined healthcare charges associated with the treatment of epidural abscesses. The nature of our registry does not allow for an assessment of function or residual deficits in patients treated operatively and nonoperatively. Although this work is an important step in understanding the contribution of patient and clinical factors to healthcare charges, future studies should correlate these with functional outcomes.

Comparing Charges: Operative Versus Nonoperative Management

The index admission charges were higher in the group treated operatively than in the group treated nonoperatively at all timepoints considered in this analysis. The duration of parenteral antibiotic administration was longer (by 11 days) in the group treated nonoperatively. Although the charges of the nonoperatively treated group were somewhat higher after discharge, this was not enough to offset the initial charges associated with operative intervention. Nonetheless, patients treated nonoperatively could have additional clinical care burdens. Unlike other conditions (for example, appendicitis), which have undergone a similar paradigm shift from surgical management to antibiotic therapy alone [13, 14], spinal epidural abscesses require long-term parenteral antibiotic therapy [3, 21]. However, parenteral therapy has a greater financial and logistical burden [19, 22], which may be prohibitive for patients who are underinsured and those who lack access to transportation and home social support. This may compound any long-term functional deficit sustained as a result of the spinal epidural abscesses. Due to the long-standing expectation that spinal epidural abscesses should be managed operatively, some hospital systems and providers may shy away from nonoperative management, despite increasing evidence of its efficacy. Our findings underscore that nonoperative management has lower immediate and long-term charges without increases in readmission, reoperation, or mortality rates, and hospitals should enact protocols on the management of spinal epidural abscesses so that their treatment is standardized across a wide range of multidisciplinary providers.

Future studies should build on this work by examining the patient-facing component of long-term care after management of an epidural abscess. It is unknown whether the decreased charges associated with nonoperative care in this study were compensated for by increased patient or caregiver burden.

Comparing Charges: Patients Who Used Intravenous Drugs Versus Those Who Did Not

We found that patients with active IVDU had higher charges than patients without IVDU initially, but that this difference did not persist for cumulative 1-year charges. It is unsurprising that the short-term charges are higher in patients with IVDU, as this is aligned with other work on healthcare charges associated with infections resulting from IVDU [17, 30]. These findings may stem from other medical and behavioral health comorbidities that are associated with IVDU alone, thus reaching beyond interventions specific to the treatment of spinal epidural abscesses. Nonetheless, Dwivedi et al. [11] reported that the treatment of spinal epidural abscesses in patients with IVDU resulted in financial losses of USD 5768 per patient for the institutions providing care. Viewed in this light, the use of nonoperative management when it is indicated for patients with IVDU who have spinal epidural abscesses may result in savings for health centers treating these individuals and more efficient allocation of resources to other aspects of care such as cessation of drug use and treatment of associated medical comorbidities and mental health concerns. Based on our findings, however, the lower charges do not persist out to 1 year, unlike with operative patients, even with similar follow-up rates and clinical outcomes between the group with and without IVDU. This suggests that healthcare systems can mitigate their long-term concerns around caring for patients with IVDU and epidural abscesses.

Conclusion

We found higher charges associated with operatively managed spinal epidural abscesses during the initial admission, which persists to 1 year after treatment. Patients with active IVDU have higher associated healthcare charges at the index admission, which diminish over the course of 1 year after the index admission. Neither group had any difference in length of follow-up, 30-day readmission rate, 1-year reoperation rate, or 2-year mortality. We believe that the findings will be of use to patients, clinicians, and healthcare systems. Although nonoperative treatment for spinal epidural abscesses has become increasingly accepted in certain contexts to avoid surgical morbidity, we showed that this could also be an effective means of reducing healthcare charges when treating patients with this condition, especially in those patients who may be predisposed because of IVDU. We caution that this study was not designed to be a direct comparison of operative and nonoperative treatment for spinal epidural abscesses, and there are situations (such as progressive neurologic deficit, paralysis, and systemic sepsis) in which surgical intervention would be the only indicated approach. However, in clinical situations in which there may be clinical equipoise, such as patients with nontoxic conditions who have normal neurologic and ambulatory function, avoiding surgical intervention may be attractive, and associated lower healthcare charges would be an additional benefit. Further work is necessary in these patients, including an evaluation of patient- and caregiver-facing impact of nonoperative treatment. Furthermore, validated evaluation of disability-adjusted life years is lacking in spinal epidural abscesses, which could provide additional data on functional status beyond the clinical outcomes described in this study. In the interim, our results provide reassurance to clinicians who are using nonoperative treatment for spinal epidural abscesses (in appropriate clinical scenarios) that this approach is an effective use of healthcare resources in the short term and long term without decreasing clinical outcomes, and that surgical management—much more than concomitant intravenous drug use—dictates increased long-term charges associated with care.