Trends and Healthcare Use Following Different Cholesteatoma Surgery Types in a National Cohort, 2003-2019.

Abstract

Objective:

To describe national trends in cholesteatoma management.

Study Design and Setting:

Retrospective analysis Optum Clinformatics® Database from 2003-2019.

Patients:

16,179 unique adult and pediatric patients who received cholesteatoma surgery.

Interventions and Main Outcome Measures:

Patients were categorized into three groups by initial surgical modality: canal wall down (CWD), canal wall up (CWU), and tympanoplasty without mastoidectomy (TnoM). Three major comparisons between groups were performed: 1) temporal trends, 2) clinical and sociodemographic determinants, and 3) healthcare use in terms of total costs and incidence of postoperative imaging and subsequent surgery.

Results:

Overall, 23.2% received initial CWD surgery, 44.3% CWU, and 32.5% TnoM. 1) The incidence of initial CWD surgery decreased (OR = 0.98, 95% CI [0.97,0.99]), while CWU increased (OR = 1.02, 95% CI [1.01,1.03]), and TnoM remained stable over the study period (OR = 0.99, 95% CI [0.98,1.00]). 2) Relative to CWU, TnoM surgery was less likely in adults, patients with prior complications, and non-White patients, while being more likely in patients with higher household income. CWD was more likely than CWU in adults, patients with prior complications, and non-White patients, while income had no effect. 3) Postoperative costs for CWU and CWD were similar. In two years following initial surgery, postoperative imaging and/or subsequent surgery was performed in 45.48% of CWD, 57.42% of CWU, and 41.62% of TnoM patients.

Conclusions:

Incidence of initial CWD surgery decreased and social disparities in cholesteatoma management were observed. Postoperative imaging or second look surgery were performed in less than 60% of patients with initial CWU surgery and over 40% of patients with initial CWD.

Introduction

Several surgical approaches and treatment paradigms for cholesteatoma currently exist, each having strong advocates.^1–8 In the early 1900s, radical mastoidectomy was the only operative means of cholesteatoma removal.⁹ By the 1950s, tympanoplasty had become widely established and the combined tympanomastoidectomy procedure continues to be used today. In this procedure, the posterior aspect of the bony external auditory canal is removed to access the middle ear, hence “canal-wall-down” (CWD). Further evolution in cholesteatoma management centered on preserving the posterior canal wall via simple mastoidectomy extended into the facial recess, hence “canal-wall-up” (CWU). More recently, advancements in imaging, microscopy, and endoscopy have allowed for a variety of even less invasive approaches without mastoidectomies; these approaches access the middle ear exclusively through the canal via postauricular or transcanal incisions, and are herein referred to as “tympanoplasty without mastoidectomy” (TnoM) for the purposes of this study. Traditionally, CWD is thought to result in lower rates of recurrence while requiring more intensive postoperative aural care due to potential loss of self-cleaning ability,¹⁰ while CWU and TnoM preserve these functions but result in higher rates of recurrence, thereby requiring postoperative surveillance imaging or staged second look surgeries as part of routine care.^2,11–13

While there are a number of indications that would strongly favor CWD surgery (such as significant canal wall erosion or otic capsule erosion in an only hearing ear), the decision as to which initial surgical approach relies largely on surgeon discretion and preferences. Given the variability of ideologies, characterization of national practice patterns for cholesteatoma management is important to identify areas where healthcare quality and value can be improved. In this study, we use a commercial and Medicare Advantage claims database from 2003 to 2019 to assess trends in cholesteatoma management on a national level. While the nuances of otologic surgery are not captured in administrative data, this approach affords a large and nationally representative sample with high-quality, objective data to study healthcare use. Here, we categorize patients by their initial surgical type based on billing data. We hypothesize that preferences of initial surgical type have shifted over time, social disparities in cholesteatoma management exist, and national postoperative healthcare use (specifically surveillance imaging and staged second-look or revision surgery) was comparable among CWU, CWD, and TnoM patients.

Methods

A retrospective analysis of insurance claims data using Optum’s Clinformatics® Data Mart Database (OptumInsight, Eden Prairie, MN) from 2003 to 2019 was conducted. The Clinformatics Data Mart Database is a de-identified database, which captures all healthcare services (including but not limited to inpatient, outpatient, and pharmaceutical claims) that were billed to a single, large health insurance provider and Medicare Advantage. The database had an annual enrollment of approximately 12 million unique individuals across all 50 states, and during the specified period, represented medical claims from approximately 9.5 million unique healthcare providers and 5,500 unique hospitals. Optum data is publicly available to researchers for a fee per year of data. This data source has been previously used in otolaryngology literature.^14–17 This study was exempt from review by the Institutional Review Board of Stanford University School of Medicine.

Detailed descriptions of all codes used are in Supplemental Table 1. Only patients who had cholesteatoma surgery as defined by a cholesteatoma diagnosis code (ICD-9: 380.xx; ICD-10: H60.xxx ,H71.xxx) associated with otologic surgery procedure code (CPT: 696xx) were included in the cohort. Initial surgery was defined as the first encounter of cholesteatoma surgery in the database. This assumption must be noted as it is not possible definitively determine if this was each patients’ true first-ever cholesteatoma surgery. Diagnoses of mastoiditis and petrositis (ICD-9: 383.xx; ICD-10: H70.xxx), bacterial meningitis (ICD-9: 320.xx; ICD-10: H70.xxx), intracranial abscess (ICD-9: 324.0; ICD-10: G06.0), subdural empyema (ICD-9: 324.9, ICD-10: G06.2), sigmoid sinus thrombosis (ICD-9: 325, ICD-10: I67.6), or facial nerve injury (ICD-9: 351.x, 951.4; ICD-10: G51.x, S04.xxxx) prior to initial surgical date were considered preoperative complications. To assess postoperative surveillance use, all temporal bone CTs (CPT: 704xx), brain MRIs (CPT: 705xx), and otologic surgeries (CPT: 696xx), subsequent to the initial surgical date were identified. Subsequent otologic surgery included all staged, second look, and revision surgeries, but excluded mastoid debridement. Demographic and socioeconomic data were obtained directly from the database. Total healthcare use costs were considered the sum of all unique charges billed to insurance in a specified period. The number of outpatient visits to otolaryngologists and emergency departments where otologic diagnoses were billed (ICD-9: 38x.xxx; ICD-10: H[6-9]x.xxx) were also counted in specified time periods.

Patients were categorized into one of three groups by initial surgical type: TnoM, CWU, or CWD. The TnoM group contained all patients who had tympanoplasty without mastoidectomy for cholesteatoma, which encompasses a variety of approaches including but not limited to transcanal and postauricular incisions and use of microscopic and endoscopic visualization. Three major comparisons across each initial surgical type were performed: 1) temporal trends, 2) clinical and sociodemographic determinants, 3) and healthcare use in terms of total costs and incidence of postoperative imaging and subsequent surgery.

All unique patients in the database were used to assess temporal trends and determinants of initial surgical type. To assess healthcare use, only patients with continuous enrollment within the database from −2 years relative to initial surgery date (to better select for true initial cholesteatoma surgery) to +2 years (to ensure capture of healthcare use 2 years postoperatively). In sensitivity analyses, models were rerun on patients with continuous enrollment from −2 to +4 years relative to the date of initial surgery. Costs were the sum of all unique charges paid by insurance in a specified time period and were not distinguished as costs directly related to cholesteatoma versus other medical comorbidities. To account for inflation, costs were adjusted to 2020 purchasing power by the year of initial surgery. The number of billed post-operative appointments after the 90-day global period following the day of otologic surgery were counted. Proportions of patients by initial surgical modality between continuously enrolled cohorts were compared to the overall cohort.

Statistical analyses were performed using Stata 16 (StataCorp LP, College Station, TX). Temporal trends were assessed using logistic regression and effect sizes were reported as odds ratios (OR) with 95% confidence intervals (CI). Clinical and sociodemographic determinants of initial surgical modality were assessed using multinomial logistic regression, where nominal outcomes variables (TnoM and CWD) were compared against a designated base variable (CWU). Here, effect sizes with 95% CIs were reported as linear coefficients representing log odds of the outcomes modeled as a linear combination of the following predictor variables: age, prior complication diagnoses, race/ethnicity, and annual household income. Within each categorial variable, categories were compared against a designated reference, where a coefficient >0 indicates a positive association and coefficient <0 indicates a negative association. To account for the right-skewed distribution of healthcare use data, results were reported as medians with interquartile ranges (IQR) and compared across groups with Kruskal-Wallis H tests followed by pairwise Mann-Whitney U tests with Bonferroni correction. Pearson χ² tests were used for comparing percentages of binary outcomes across groups followed by pairwise χ² tests with Bonferroni correction. An alpha of p<0.05 was used to define statistical significance.

Results

Patient characteristics are shown in Table 1. Overall, 16,179 unique patients who received cholesteatoma surgery were identified, 3,955 pediatric (<18 years), 12,224 adults (≥18 years). Proportions of initial cholesteatoma surgery type by age group are shown in Figure 1. Of these patients, 2,044 patients were continuously enrolled −2 to +2 years relative to the initial surgery date, while 1,113 patients were continuously enrolled −2 to +4 years relative to the initial surgery date.

Table 1:

Patient characteristics. Continuous enrollment is defined by number of years relative to the initial surgery date. TnoM = tympanoplasty without mastoidectomy, CWU = canal wall up, CWD = canal wall down.

Variable		All-Comers (n = 16,179)	Continuously Enrolled −2y to +2y (n = 2,044)	Continuously Enrolled −2y to +4y (n = 1,113)
Index surgery modality
	TnoM	5,256 (32.5)	680 (33.3)	366 (32.9)
	CWU	7,164 (44.3)	944 (46.2)	532 (47.8)
	CWD	3,759 (23.2)	420 (20.6)	215 (19.3)
Age at initial surgery date, n (%)
	<18 years	3,955 (24.5)	482 (23.6)	281 (25.3)
	≥18 years	12,224 (75.6)	1,562 (76.4)	832 (74.8)
Sex, n (%)
	Female	7,133 (44.1)	941 (46)	511 (45.9)
	Male	9,044 (55.9)	1,103 (54)	602 (54.1)
Prior complicated diagnoses, n (%)
	Present	1,323 (8.2)	224 (11)	140 (12.6)
	Absent	14,856 (91.8)	1,820 (89)	973 (87.4)
Race, n (%)
	Asian	606 (3.8)	90 (4.4)	51 (4.6)
	Black	898 (5.6)	112 (5.5)	63 (5.7)
	Hispanic	1,439 (8.9)	202 (9.9)	104 (9.3)
	Unknown	2,220 (13.7)	134 (6.6)	77 (6.9)
	White	11,016 (68.1)	1,506 (73.7)	818 (73.5)
Annual Household Income, n (%)
	<$100,000	6,378 (39.42)	945 (46.23)	502 (45.10)
	≥$100,000	4,528 (27.99)	784 (38.36)	448 (40.25)
	Unknown	5,273 (32.59)	315 (15.41)	163 (14.65)

Figure 1:

Proportion of initial cholesteatoma surgery in patients age ≥18 (“adult”, n = 12,224) and <18 years (“child”, n = 3,955) compared to overall cohort (“all”, n = 16,179).

Temporal Trends in Initial Cholesteatoma Surgery

In the overall cohort (n=16,179), the incidence of initial TnoM surgery remained stable throughout the study period (OR/year=0.99, 95% CI [0.98,1.00], p=0.077), while CWU incidence increased (OR/year=1.02, 95% CI [1.01,1.03], p<0.001) and CWD incidence decreased (OR/year=0.98, 95% CI [0.97,0.99], p<0.001). Therefore, over time, patients had slightly higher odds of receiving CWU, slightly lower odds of receiving CWD, and equivocal odds of receiving TnoM surgery.

Pediatric patients had a lower incidence of initial CWD surgery than adults (Figure 1) and the incidence decreased further over time (OR/year=0.96, 95% CI [0.94,0.98], p<0.001). The incidence of TnoM and CWU surgery among children did not change over the study period (OR/year=1.01, 95% CI [0.99,1.03], p<0.141; and OR/year=1.01, 95% CI [0.99,1.02], p=0.174; respectively).

In adults, incidence of initial CWU surgery was higher than TnoM and CWD (Figure 1) and increased over time (OR/year=1.02, 95% CI [1.01,1.03], p<0.001). The incidence of TnoM surgery among adults did not change (OR/year=1.00, 95% CI [0.99,1.00], p=0.772) while CWD decreased (OR/year=0.98, 95% CI [0.97,0.99], p<0.001). All considered, the incidence of initial CWU surgery increased for adult patients while incidence of initial CWD decreased for both adult and pediatric patients.

Social Determinants of Initial Cholesteatoma Surgery

Results of the multinomial logistic regression are shown in Table 2. After adjusting for covariables, adult patients, patients with prior complications, and Hispanic patients were independently less likely to have initial TnoM surgery than CWU in the overall cohort (n=16,179). Higher annual household income was strongly predictive of TnoM surgery with a notably large effect size (Table 2).

Table 2:

Results of the multinomial logistic regression (n=16,179). Nominal outcomes (TnoM = tympanoplasty without mastoidectomy, CWD = canal wall down) were compared to a designated base (CWU = canal wall up). Log odds of the outcomes are modeled as a linear combination of predictor variables. Effect size is reported as a linear coefficient where negative predictors are <0 and positive predictors are

		TnoM (compared to CWU base)			CWD (compared to CWU base)
		Coefficient	95% CI	P-value	Coefficient	95% CI	P-value
Per Calendar Year		−0.01	−0.02, 0.00	0.027*	−0.02	−0.03, −0.01	<0.001*
Age
	<18 years	Reference			Reference
	≥18 years	−0.67	−0.77, −0.57	<0.001*	0.30	0.17, 0.44	<0.001*
Pre-operative Complication
	Absent	Reference			Reference
	Present	−0.61	−0.80, −0.42	<0.001*	0.30	0.14, 0.46	<0.001*
Race/Ethnicity
	White	Reference			Reference
	Black	−0.19	−0.38, 0.01	0.057	0.22	0.03, 0.41	0.024*
	Hispanic	−0.21	−0.36, −0.05	<0.001*	0.22	0.07, 0.38	0.005*
	Asian	−0.08	−0.30, 0.15	0.501	0.26	0.03, 0.49	0.029*
Annual Household Income
	<$100,000	Reference			Reference
	>$100,000	22.90	2.74, 43.06	0.026*	−0.04	−0.14, 0.06	0.437

In contrast, adult patients, patients with prior complications, Asian, Black, and Hispanic patients were all independently more likely to receive initial CWD surgery than CWU. However, annual household income did not have an effect on CWD versus CWU (Table 2).

Healthcare Use Associated with Initial Cholesteatoma Surgery Type

The median total healthcare use cost (sum of all unique charges to insurance) 2 years prior to index surgery date was $13,971 (IQR $6,447 to $33,328) while the median cost 2 years after initial surgery was $28,053 (IQR $9,324 to $69,112). Healthcare use costs both before and after initial surgery differed across initial surgery modalities (Table 3). As expected, pairwise comparisons showed that TnoM patients had lower preoperative costs than CWU patients (p<0.001), and CWU patients had lower preoperative costs than CWD patients (p<0.001). Postoperatively, TnoM patients had lower costs than CWU and CWD patients (p<0.001 each). However, CWU and CWD patients had similar post-operative costs (p=0.8520).

Table 3:

Total healthcare use costs grouped by index surgery type (n = 2,044; TnoM = tympanoplasty without mastoidectomy, CWU = canal wall up, CWD = canal wall down). US dollar amounts ($) are adjusted to 2020 purchasing power by year of surgery. Equality of groups was tested using Kruskal-Wallis H tests. Results were reported as χ² with ties with 2 degrees of freedom. The null hypothesis of equality between population was rejected if p-value > 0.05.

	TnoM, $ (IQR)	CWU, $ (IQR)	CWD, $ (IQR)	χ2 with ties	p-value
2y before index	11,176 (5,009 to 26,894)	14,241 (6,701 to 35,712)	18,097 (8,772 to 39,002)	43.886	<0.001
2y after index	19,808 (6,111 to 46,801)	33,768 (11,882 to 75,041)	32,462 (10,863 to 85,487)	55.279	<0.001

In the 2 years post initial surgery, the median number of otolaryngology clinic return visits after the immediate postoperative 90-day global period was 3 (IQR 1 to 6). Notably, this number also does not include any appointments within the 90-day global period for any subsequent surgeries. There was no difference in the number of visits across initial surgical modalities (Kruskal-Wallis: χ² with ties=2.101, p=0.350). The percentage of patients who had at least one ED visit with a billed ear otologic diagnosis was 17.91%. There was no difference in ED use across groups (Pearson: χ²=1.849, p=0.462).

The percentage of patients who received post-operative imaging (CT or MRI) or subsequent surgery (including staged, second look, and revision surgeries) was 41.62% for TnoM, 57.42% for CWU, and 45.48% for CWD. These proportions were significantly different between groups (Pearson: χ² =13.515, p=0.001) with CWU having significantly higher use of subsequent imaging or surgery than TnoM or CWD (p<0.001 each). Breakdowns for groups by imaging modality and subsequent surgery are shown in Table 4.

Table 4:

Percentages of patients receiving surgery, CT, or MRI subsequent to initial surgery. Time period is indicated in number of years subsequent to initial surgery date (n = 2,044 for 2 years, n = 1,113 for 4 years).

	TnoM, %		CWU, %		CWD, %
	2 years	4 years	2 years	4 years	2 years	4 years
Surgery	27.65	34.43	41.21	45.68	29.29	33.95
CT	21.32	30.33	23.62	35.15	24.05	38.60
MRI	6.76	13.39	11.86	17.48	7.86	13.49
Any of the above	41.62	53.83	57.42	66.17	45.48	58.14

A sensitivity analysis was performed on patients with continuous follow up 4 years after the initial surgery date. As expected, the percentage of patients who received imaging or surgery subsequent to the initial surgery date increased to 53.83% for TnoM, 66.17% for CWU, and 58.14% for CWD. Again, these proportions significantly differed between groups (Pearson: χ² =13.515, p=0.001) with CWU having significantly higher use than TnoM or CWD (p<0.001 each, Table 3).

When temporal trends in imaging and surgery 2 years postoperatively were considered, incidence of subsequent surgery decreased (OR=0.97, 95% CI [0.93,0.99], p=0.015), CT use did not change (OR=1.00, 95% CI [0.97,1.04], p=0.961), and MRI use increased (OR=1.07, 95% CI [1.02,1.13], p=0.007). Postoperative imaging and surgery use broken down between children and adults is shown in Figure 2. In children, subsequent surgery was more common than imaging (Figure 2) and no changes were observed over time (surgery: OR=0.99, 95% CI [0.93,1.05], p=0.643; CT: OR=0.99, 95% CI [0.92,1.07], p=0.836; MRI: OR=1.07, 95% CI [0.93,1.24], p=0.334). In adults, incidence of subsequent surgery and CT were similar (Figure 2) and remained stable over time (surgery: OR=0.97, 95% CI [0.93, 1.00], p=0.054; CT: OR=1.00, 95% CI [0.96,1.04], p=0.886), however, the incidence of MRI increased during the study period (OR=1.06, 95% CI [1.01,1.12], p=0.030).

Figure 2:

Percentages of patients that received otologic surgery, CT, and MRI in 2 years following initial a) CWD, b) CWU, and c) TnoM by age ≥18 (“adult”, n = 1,592) and <18 years (“child”, n = 482).

Discussion

This study is the first to our knowledge to characterize US national trends in cholesteatoma management using administrative data. In our cohort, initial CWD became less common over time, socioeconomic disparities in cholesteatoma management exist, and that TnoM surgery was common (in children, the most common initial surgery type). Surveillance imaging and second-look surgery is not performed in over a third of all non-CWD patients. Additionally, we found that CWD was associated with higher-than-expected postoperative surveillance use, suggesting that it is not definitive surgery in many cases.

CWD became less popular over time.

In recent decades, a divergence of opinion exists regarding which initial surgery technique is best used to manage cholesteatoma; some surgeons advocate primarily for CWD,^18,19 others prefer CWU,^20–22 and yet others have described various techniques to fit individual situations.^2,23,24 We hypothesized that cholesteatoma management followed a general trend within surgery as a whole towards less invasive approaches. Indeed, we found that incidence of CWD surgery decreased in the previous two decades in our cohort. While this annual decrease in effect size may be small, it is statistically significant and moreover clinically significant when considering annual changes over the 16-year study period. Meanwhile, incidence of initial CWU surgery increased in adult patients. It is important to note that many strategies that could not be clearly categorized became more popular during the study period such as the “inside-out” technique,^25–28 and many strategies of canal wall reconstruction were described.^29–32 These popularized strategies may not have been consistently coded by different surgeons.

What is perhaps more interesting is the high incidence of TnoM surgery. While CWD versus CWU is extensively debated in literature,^5–8 TnoM approaches are seldom discussed in comparison. In our cohort, initial TnoM cholesteatoma surgery accounted for one third of our overall cohort and half of all children, making it more common than CWD. One of the limitations in our study is that there is no way to distinguish acquired from congenital cholesteatoma using the ICD diagnose codes. Initial TnoM cholesteatoma surgery may have been the preferred method to treat congenital cholesteatomas,³³ which may explain how it became the most common initial surgery type in our pediatric cohort. Considering that the TnoM group encompasses a variety of techniques (postauricular vs transcanal incisions, microscopic vs endoscopic visualization, etc.), future efforts are needed to identify indications and measure outcomes for various TnoM techniques. In recent years, there is rapidly increasing interest on endoscopic ear surgery (EES) for cholesteatoma, however a claims database analysis methodology will not be able to analyze EES outcomes until CPT codes specific for endoscopic ear surgery become standardly utilized.

Social disadvantage predicts more invasive initial surgery.

Given that much of cholesteatoma management relies on surgeon discretion and preferences of the surgeon, it is important to identify disparities in care. In our cohort, non-White patients tended to receive more invasive surgery. One possible explanation is that socially disadvantaged patients typically present with more advanced disease, necessitating more invasive approaches.^34–36 This explanation is consistent with prior work demonstrating racial/ethnic disparities in hearing healthcare in both children and adults.^37,38 However, it is also possible that surgeons elect to perform more “definitive” surgeries on patients with less reliable access to routine follow up and surveillance, which has previously been suggested.³⁹ This argument is supported by our income analyses, which showed that higher household income was strongly associated with TnoM over CWU surgery. While this effect may be explained in part by higher income patients presenting with less advanced disease, this disparity likely also reflects differences by patient social status in terms of counselling, priorities, and understanding on the parts of both surgeon and patient. Therefore, it is important for surgeons to be cognizant of social determinants of cholesteatoma management in order to reduce healthcare disparities in otology.

Healthcare use associated with CWU and CWD is more similar than expected.

While preserving the posterior canal wall may be beneficial for hearing and hygiene outcomes, it is thought to carry an increased the risk for residual or recurrent disease.¹³ Consequently, postoperative surveillance with second-look surgery or imaging has been widely advocated as part of routine care following non-CWD approaches.^7,40,41 With respect to this standard, surveillance after non-CWD surgery was less common than expected in our cohort. In the 2 years following initial CWU surgery, over 40% of patients with continuous insurance coverage did not receive any form of subsequent imaging or surgery. It should be noted that this figure likely overestimates true surveillance use given the inclusion of all otologic surgeries rather than second-look procedures specifically. While some patients who did not receive postoperative surveillance may have been lost to follow-up in this continually insured cohort, it is likely that surgeons elected not to pursue surveillance in a significant proportion of patients. The proportion without surveillance is even larger in TnoM patients, suggesting that surgeons may elect not to pursue surveillance when disease is less advanced (such as in well-encapsulated congenital cholesteatoma limited to the anterosuperior quadrant), or feel that surveillance via clinical history and exam is adequate. However, specific criteria for when this is appropriate is ill defined in literature and future work in this area are needed to increase healthcare quality and value.

CWD surgery, in contrast, is often thought of as a definitive surgery that requires no subsequent imaging or surgery. This notion is pervasive in literature, resident education, and patient counselling.^1,19,20 However, we demonstrate that approximately 45% of initial CWD patients received imaging or surgery 2 years postoperatively, suggesting that CWD procedures may not be definitive and many cases needed additional imaging studies and procedures. Reasons for reoperation on a CWD mastoidectomy may include staged or revision OCR, mastoid obliteration, or scar revision (mastoid debridement was excluded from this analysis). Reimaging after CWD mastoidectomy may be performed on an as-needed basis for issues such as unexpected bulges under the skin flap or unexplained vertigo. Given these findings, it is important for otologists to be collectively aware of these national healthcare use patterns when counselling patients about CWD procedures.

Analyses of total healthcare use costs prior to initial surgery suggests that disease complexity as proxied by cost correlated with the extent of the initial surgical approach, i.e., patients who underwent less invasive surgery had lower preoperative healthcare costs. Patients with higher preoperative costs may have had more complicated ear disease but is also likely that higher preoperative costs represent higher non-otologic comorbidity burden. It is possible that physicians tended to pursue more invasive surgeries in sicker patients to reduce operative time with wider exposures or that less invasive approaches were less important for these patients.

When comparing postoperative total costs, CWU and CWD were found to be similar. This may be because treatment paradigms are equally cost effective, but more likely it is because that postoperative imaging and surgery use is similar between CWU and CWD. Surgeon preference is likely a major driver of this finding; while some surgeons advocate for staging CWU but not CWD procedures, advocates for staging may routinely perform staged surgery on initial CWD as well, while advocates for single-stage surgery may not perform staged surgery on either. Similarly, the use of postoperative imaging varies widely, likely also due to differences in surgeon preference and disease processes. Future work comparing cost effectiveness of cholesteatoma treatment paradigms when closely followed would guide high value decision making.

Limitations.

This study has several limitations in addition to those previously discussed. As with all claims data, clinical information is inferred from the billing codes and is consequently susceptible to coding error and non-specific coding. Furthermore, relevant information that cannot be inferred from billing data such as degree of hearing loss could not be accounted for. Similarly, the level of detail inferred from codes is limited by specificity of available codes. For example, endoscopic surgeries could not be distinguished from microscopic surgeries within the TnoM group due to lack of specific billing codes. Surgeon intention cannot be inferred from coding, so while all subsequent otologic surgeries could be identified, staged second-look surgeries could not be distinguished from revision surgeries such as tympanoplasty for repair of perforation.

For analyses on the overall cohort (n = 16,179), initial cholesteatoma surgery was assumed to be the first cholesteatoma surgery that appeared in the database. However, it is likely that a subset of patients received their first true cholesteatoma surgery previously while on a different insurance plan. Analyses performed on the continuously enrolled cohort (n = 2,044) ensured that patients did not receive any otologic surgery for 2 years prior, however, there is no way to definitively confirm whether this was the patients true first surgery. The study cohort consisted primarily of commercially insured patients; while age, gender, and geographic distribution were comparable to national benchmarks, patients with high annual income were overrepresented, thereby limited the generalizability of results to the entire US population.

Costs reported on were for total healthcare costs in specified time periods rather than ear-specific costs. As such, differences in costs between cholesteatoma surgical modalities could be attenuated due to non-otologic comorbid healthcare costs. Future work where ear-specific healthcare costs could be extracted are needed.

Finally, this study investigates national healthcare use following cholesteatoma surgery and does not comment on rates of recurrent or recidivistic disease. Future studies with longer follow-up duration and greater clinical detail during follow-up are needed.

Conclusions.

Incidence of initial CWD surgery decreased over time. Social differences in cholesteatoma management were observed. Rates of postoperative imaging and further revision surgery were similar between initial CWU and CWD surgery. Cholesteatoma surgery without mastoidectomy was common, particularly in children, and warrants future study.