Racial and Ethnic Disparities in Utilization of Tonsillectomy among Medicaid-Insured Children

Abstract

Objective:

To examine racial differences in tonsillectomy ±adenoidectomy (T&A) for sleep-disordered breathing (SDB) among Medicaid-insured children.

Study design:

Retrospective analysis of the 2016 MarketScan Multistate Medicaid Database was performed for children ages 2 to <18 years with a diagnosis of SDB. Patients with medical complexity and infectious indications for surgery were excluded. Racial groups were categorized into non-Hispanic White, non-Hispanic Black, Hispanic, and Other. Adjusted multivariate logistic regression was used to determine if race/ethnicity was a significant predictor of obtaining T&A, PSG, and time to intervention.

Results:

There were 83,613 patients with a diagnosis of SDB that met inclusion criteria, of which 49.2% were female with a mean age of 7.9 ± 3.8 years. The cohort consisted of White (49.2%), Black (30.0%), Hispanic (8.0%), and Other (13.2%) groups. Overall, 15.4% underwent T&A. Black (82.2%) and Hispanic (82.3%) children had significantly higher rates of no intervention and White patients had the lowest rate of no intervention (76.9%; P < .0001) and the highest rate of T&A (18.7%; p<0.0001). Mean time to surgery was shortest in White compared with Black children (p<0.0001). Logistic regression adjusting for age and sex showed that Black children had 45% reduced odds of surgery (95% CI: 0.53–0.58), Hispanic 38% (95% CI: 0.58–0.68), and Other 35% (95% CI: 0.61–0.70) compared with White children with Medicaid insurance.

Conclusions:

Racial and ethnic disparities exist in the utilization of T&A for children with SDB enrolled in Medicaid. Future studies that investigate possible sources for these differences and more equitable care are warranted.

Pediatric obstructive sleep-disordered breathing (SDB) affects as many as 11–15% of children and represents a continuum of sleep derangements, from primary snoring to obstructive sleep apnea (OSA).(1–4) Periodic upper airway obstruction often leads to intermittent decreases in oxygen levels, hypercapnia, and sleep disturbance.(5) A sleep study, or polysomnography (PSG), is the gold-standard diagnostic study to characterize the severity of airway obstruction.(6) Consequences of untreated SDB in school-aged children include declines in neurocognitive performance, academic deficits, behavioral impairments, and adverse cardiovascular effects in cases of severe OSA.(7–12)

Tonsillectomy with or without adenoidectomy (T&A) is generally considered the first-line treatment for children aged 2–18 years with SDB and adenotonsillar hypertrophy.(13–15) T&A can lead to improvements in behavior, polysomnographic measures, and overall quality of life in even mild OSA.(19–25) Use of health services and overall health-related costs, which are higher in children with OSA, are reduced following surgical treatment with T&A.(26, 27) Obstructive SDB, rather than infectious causes such as tonsillitis, is now the leading indication for tonsillectomy.(15)

Studies have demonstrated significant demographic disparities in the prevalence of SDB by race/ethnicity and socioeconomic status (SES).(28) SDB disproportionately affects children of Black race and low-income in the United States.(29–37) This observation is profound given that vulnerable children are already at higher baseline risk for comorbid conditions associated with SDB, including obesity, smoke exposure, poor school performance, and behavioral impairments .(38–41) Children from lower SES backgrounds are at additional risk of shorter sleep durations and more sleep difficulties than those living in higher SES families.(42, 43)

It is estimated that in the United States, 2 in every 5 children are Medicaid beneficiaries.(44) Medicaid, with the exception of limited medical conditions, is only provided to patients below the federal poverty level, with a median national eligibility criteria of 138–195% beneath the federal poverty level.(45, 46)

The use of a mega-database, such as the MarketScan Multi-State Medicaid Database, provides population level claims information and integrates de-identified patient-level health data, hospital discharges and electronic medical records. To remove potential confounding secondary to access to care due to insurance, we sought to investigate racial/ethnic variations in SDB interventions in a cohort of children with the same public insurance type. We hypothesize that among otherwise healthy children with SDB and Medicaid insurance, there will be differential attainment of T&A for racial and ethnic minorities.

METHODS

Data Source and Population

We analyzed data from the 2016 Truven MarketScan Multi-State Medicaid Database, which includes the medical and surgical insurance claims of more than 44 million Medicaid enrollees from multiple states in the US. The database contains records of inpatient visits, outpatient services, prescription drug claims, and other medical care. All claims include a unique identification code that can link enrollee claims across services and admissions, allowing for longitudinal analyses. Institutional Review Board approval from the Medical University of South Carolina was not required for this study because analyses were conducted using de-identified data.

Eligibility criteria were chosen to assemble a relatively low-risk cohort of children ages 2–18 years old undergoing T&A for obstructive symptoms. A flow diagram (Figure ; available at www.jpeds.com) illustrates the patient selection process for the study. The database was queried for patients that had a primary diagnosis of SDB symptoms or documented OSA by the International Classification of Diseases, 10^th Revision (ICD-10) codes (G4730, G4733, J353, J351, R0683) in 2016. These specific codes were chosen in order to include only obstructive indications for T&A. We excluded children younger than 2 years of age because PSG is recommended prior to T&A for this age group by the American Academy of Otolaryngology-Head and Neck (AAO-HNS) guidelines.(6) Indication for T&A in this cohort was for obstructive SDB; therefore children with infectious or inflammatory processes of acute or chronic adenotonsillitis were excluded. In order to study children of homogeneous medical complexity, we excluded high-risk populations that would otherwise be recommended preoperative PSG, including genetic syndromes, craniofacial abnormalities, micrognathia, skeletal and neuromuscular disorders, cerebral palsy, sickle cell disease, and mucopolysaccharidoses (Table V; available at www.jpeds.com).(6)

Figure 1.

Cohort flow diagram of children with sleep-disordered breathing that were identified in the Medicaid database. Patients included in the final analyses are shown by intervention pathway.

Abbreviation: PSG, polysomnography

Table 5.

Exclusion Codes for the Study Cohort.

Diagnoses of adenotonsillitis	ICD-10
Chronic tonsillitis	J35.01
Chronic tonsillitis and adenoiditis	J35.0, J35.03, J35.9
Acute recurrent streptococcal tonsillitis	J03.01
Acute tonsillitis, unspecified	J03.90
Acute recurrent tonsillitis, unspecified	J03.91
Diagnoses warranting PSG	ICD-10
Down syndrome	Q90.9
Anomalies of skull and face	Q75.9
Micrognathia	Q87.0
Achondroplasia	Q77.4
Neuromuscular disorders	G70.9
Cerebral palsy	G80.9
Sickle cell disease	D57.2
Mucopolysaccharidoses	E76.3

Abbreviations: ICD-10, International Classification of Diseases, Tenth Revision; PSG, polysomnography

Index dates for SDB diagnosis ranged from January 1 to September 30, 2016 and patients were followed for an additional 3 months from their index diagnosis date to establish patient clinical pathways. The final cohort underwent four possible intervention pathways: children that underwent T&A alone without PSG, PSG before T&A, PSG without subsequent surgery, and no intervention at all. Surgeries were identified using Current Procedural Terminology (CPT) codes for tonsillectomy (CPT 42825, 42826) and adenotonsillectomy (CPT 42820, 42821). T&A was defined as tonsillectomy with or without adenoidectomy. We excluded children that underwent adenoidectomy alone as this is often performed for reasons other than SDB, such as nasal obstruction or chronic otitis media. Additionally, we identified children that had pre-operative attended PSG (CPT 95810, 95782, 95807, 95808). Baseline patient demographics and clinical variables including age, sex, race/ethnicity, PSG, and surgical intervention were assessed for all patients. Race/ethnicity in the database were separated into four categories: non-Hispanic White, non-Hispanic Black, Hispanic, and Other. The primary outcome in this study was proportion of patients undergoing T&A for SDB, compared across racial/ethnic groups in an otherwise healthy cohort of Medicaid insured children. The secondary outcome was time to surgery stratified by race/ethnicity. Additional analyses were performed for likelihood of PSG and undergoing any intervention.

Statistical Analyses

Categorical variables were summarized by frequency and percentage. All continuous variables were assessed for normality using the Shapiro-Wilk test or Kolmogorov-Smirnov test depending on sub-group sample size. Continuous variables were summarized by mean ± standard deviation or interquartile range (IQR). Comparisons of baseline characteristics and outcomes (categorical variables) were performed using a Chi Square test or the Fisher exact test where appropriate. For continuous variables, a Kruskal-Wallis test was used to compare more than two groups. A multivariable logistic regression was used to examine the association between race/ethnicity and the primary outcome variable of PSG, T&A attainment, and any intervention. Those covariates that were significantly associated in the univariate analysis were included for analysis in our final regression model that adjusted for age and sex and were reported as adjusted odds ratios (OR). All analyses were performed with SAS 9.4 (SAS Institute, Cary, NC). The p value level of significance was set at an α-level of 0.05 for all statistical tests.

RESULTS

There were 83,613 patients from the 2016 MarketScan Medicaid Database with a diagnosis of SDB (Table I). The mean age was 7.9 (SD, 3.8) years. Most patients were under the age of 13 years (85.0%), mostly White (48.8%), and male (50.8%). Only 6.5% of the entire cohort underwent a sleep study. As for surgical intervention, 15.4% of patients underwent surgery, of which the majority (92.8%) underwent adenotonsillectomy and 7.2% underwent tonsillectomy alone without adenoidectomy. Of the 12,884 surgery patients with a listed discharge location, almost all (99.5%) children had outpatient surgery.

Table 1.

Demographics and Surgery Status for Pediatric Patients with Diagnosis of Sleep-Disordered Breathing from the Truven MarketScan Medicaid Database (n=83,613).

Characteristic	Patients in Cohort n=83,613 (%)
Age
Age, Mean ± SD	7.9 ± 3.8
Sex
Male	42,509 (50.8)
Female	41,104 (49.2)
Race
White	40,781 (48.8)
Black	25,077 (30.0)
Hispanic	6,711 (8.0)
Other	11,044 (13.2)
PSG	5,445 (6.5)
Surgery	12,884 (15.4)
Adenotonsillectomy	11,956 (14.3)
Tonsillectomy alone	928 (1.1)

Abbreviation: PSG, polysomnography

Table 2 demonstrates the four possible treatment pathways taken by patients that met inclusion criteria by racial groups. After receiving an ICD-10 diagnosis of SDB, 79.4% of all patients had no intervention of PSG or T&A. Black and Hispanic patients had significantly higher rates of no intervention after diagnosis and White patients had the lowest rate of no intervention (76.9%; p<0.0001). Black patients had the highest rate of undergoing PSG followed by surgery (1.6%; p=0.0044). Black and Other groups also had the highest rate of PSG without having surgery afterwards (6.0%; p<0.0001). White patients had the highest rate of surgery after receiving a diagnosis of SDB without a diagnostic PSG (17.4%; p<0.0001).

Table 2.

Treatment Pathway Distribution for Pediatric Patients with Sleep-Disordered Breathing.

	Total	White	Black	Hispanic	Other	P-valuea
Treatment Pathway	n=83,613	n=40,781	n=25,076	n=6,711	n=11,044
No Intervention	66,413 (79.4)	31,357 (76.9)	20,607 (82.2)	5,521 (82.3)	8,928 (80.8)	<0.0001
Surgery alone	11,755 (14.1)	7,099 (17.4)	2,567 (10.2)	775 (11.5)	1,314 (11.9)	<0.0001
PSG before surgery	1,129 (1.4)	511 (1.3)	394 (1.6)	84 (1.3)	140 (1.3)	0.0044
PSG and no surgery	4,316 (5.2)	1,814 (4.5)	1,509 (6.0)	331 (4.9)	662 (6.0)	<0.0001

^aSignificance determined by Chi square test of subgroup analysis (right four columns) by race.

Abbreviations: PSG, polysomnography

Table 3 demonstrates the mean time from diagnosis to intervention by race/ethnicity. The overall mean days to surgery for the cohort was 49.1 days (IQR, 4.0–72.0). There were significant differences in the mean days to surgery when stratified by race/ethnicity with a lowest mean of 43.7 (IQR, 0.0–65.0) days to surgery for White patients (p<0.0001) and mean days to PSG of 44.8 (IQR, 0.0–71.0) (p=0.0005).

Table 3.

Time to Surgery and Polysomnography for Pediatric Patients with Sleep-Disordered Breathing.

	Total	White	Black	Hispanic	Other	P-valuea
Mean days to surgery [IQR]	49.1 [4.0–72.0]	43.7 [0.0–65.0]	59.0 [19.0–87.0]	54.8 [15.0–78.0]	53.5 [8.5–81.0]	<0.0001
Mean days to PSG [IQR]	48.0 [0.0–76.0]	44.8 [0.0–71.0]	49.9 [0.0–77.0]	51.7 [0.0–80.0]	51.0 [0.0–81.5]	0.0005

^aSignificance determined by Kruskal-Wallis test of subgroup analysis (right four columns) by race

Abbreviations: PSG, polysomnography

Logistic regression was performed to determine the race/ethnicity-stratified odds of obtaining a sleep study, undergoing surgery, or receiving any intervention while adjusting for age and sex (Table 4). With each year increase in age, patients were 7% (95% CI: 1.06–1.08) more likely to undergo a sleep study and 5% (95% CI: 0.94–0.95) less likely to undergo surgery. Female patients were less likely to undergo PSG (OR=0.88; 95% CI: 0.83–0.93) and significantly more likely to undergo surgery (OR=1.18; 95% CI: 1.14–1.23) when compared with males. Black and Other groups had higher odds of obtaining a PSG (OR=1.35; 95% CI: 1.27–1.43 and OR=1.29; 95% CI: 1.19–1.40, respectively) compared with White children. For all children in the cohort that had surgery performed regardless of the treatment pathway (n=12,883), minority race/ethnicity children were significantly less likely than White patients to undergo surgery (Black OR=0.59; 95% CI: 0.56–0.61, Hispanic OR=0.65; 95% CI: 0.60–0.70), and Other OR=0.66; 95% CI: 0.62–0.70). A sub-analysis was performed on patients that had surgery without PSG (n=11,754) to remove the potential confounder of mild sleep apnea results on PSG and thus recommendations against T&A in favor of watchful waiting. The logistic regression demonstrated consistent differences in rates of surgery among racial groups that had T&A without PSG when controlling for age and sex, with Black children having 45% reduced odds of surgery (95% CI: 0.53–0.58), Hispanic patients 38% (95% CI: 0.58–0.68), and Other 35% (95% CI: 0.61–0.70) compared with White children. As far as undergoing PSG and/or T&A, racial and ethnic minority children with SDB were less likely to receive any intervention compared with White children (Black OR=0.72; 95% CI: 0.69–0.75, Hispanic OR=0.72; 95% CI: 0.67–0.77, and Other OR=0.79; 95% CI: 0.75–0.83).

Table 4.

Logistic Regression Models Predicting Likelihood of Obtaining PSG, Surgery or Any Intervention.

	PSG Obtained n=5,445		Surgery Performed n=12,883a		Interventionb n=17,199a
Race	Patients, n (%)	Odds Ratio [95% CI]	Patients, n (%)	Odds Ratio [95% CI]	Patients, n (%)	Odds Ratio [95% CI]
White	2,325 (42.7)	Reference	7,610 (59.0)	Reference	9,424 (54.8)	Reference
Black	1,903 (35.0)	1.35 [1.27–1.43]	2,961 (23.0)	0.59 [0.56, 0.61]	4,470 (26.0)	0.72 [0.69–0.75]
Hispanic	415 (7.6)	1.09 [0.98–1.21]	859 (6.7)	0.65 [0.60, 0.70]	1,190 (6.9)	0.72 [0.67–0.77]
Other	802 (14.7)	1.29 [1.19–1.40]	1,453 (11.3)	0.66 [0.62, 0.70]	2,115 (12.3)	0.79 [0.75–0.83]
Female	2,535 (46.6)	0.88 [0.83–0.93]	6,746 (52.4)	1.18 [1.14–1.23]	8,728 (50.8)	1.09 [1.05–1.12]
Age		1.07 [1.06–1.08]		0.95 [0.94–0.95]		0.99 [0.98–0.99]

^aExcluded 1 value for missing variables

^bIntervention defined as receipt of PSG and/or surgery

Abbreviations: CI, confidence interval; PSG, polysomnography

DISCUSSION

The American Academy of Pediatrics stated that research on eliminating health and healthcare disparities related to race, ethnicity, and SES be a priority.(47) our study of a Medicaid-only population decreases the potential interaction of race/ethnicity with SES given the criteria of meeting federal poverty levels in this group of otherwise healthy children. We demonstrated that children of Black, Hispanic, and Other race/ethnicity were 35–45% less likely to undergo T&A after being diagnosed with SDB compared with their White peers. In addition, racial and ethnic minority groups had significantly longer wait times from diagnosis of SDB to surgical intervention and were significantly less likely to receive any intervention after diagnosis compared with White children. These findings are especially striking given that SDB is two- to four-times more prevalent among Black than White children.(31, 32)

Our results support previous studies that show underuse of T&A among racial and ethnic minority children in the United States.(18, 28, 48, 49) A longitudinal investigation from the Cleveland Family Study that was comprised of a community-based urban sample, showed that Black children with SDB were 50% less likely to have undergone self-reported T&A than other racial groups.(49) In a Mississippi impoverished rural community medical center, health services used for tonsillectomy and tympanostomy tubes revealed that White children were almost 3 times more likely to undergo T&A.(48) Our findings support a prior cost analysis by Chang et al that suggested underuse of T&A among Black children in Missouri insured with Medicaid.(50) However, the prior study included adenotonsillar infections as an indication for T&A whereas the present study analyzed a larger sample size across different regions and included only children that underwent surgery for obstruction. Moreover, we were able to incorporate information regarding additional racial and ethnic minority groups and identified disparate attainment of T&A among Hispanic and Other minority races and ethnicities.

Heller et al investigated the impact of the AAO-HNS 2011 tonsillectomy and PSG clinical practice guidelines and found that average annual rates of tonsillectomy use during 2004–2017 were lower among Black and Hispanic children than among White children overall for both infectious and non-infectious indications in the states of South Carolina and Florida.(51) However, the authors also found that tonsillectomy use remained greater among publicly insured than privately insured children. This is a distinction from previous reports in which public insurance was correlated with longer wait times to PSG and surgical evaluation by an otolaryngologist compared with children with private insurance.(52, 53)

Utilization of PSG varies not only by society guidelines, (16, 54, 55) but also by institutional practice patterns, availability of PSG centers, and specialist type.(55, 56) Despite various guideline recommendations, most children who undergo T&A for SDB do so without polysomnographic evidence that documents mild, moderate or severe OSA.(57, 58) The Childhood Adenotonsillectomy Trial (CHAT) included children with mild to moderate polysomnographically documented OSA and demonstrated that tonsillectomy improved behavior, quality of life, sleep study measures, and OSA symptoms but did not lead to significant changes in objective measures of attention or executive function compared with watchful waiting over a 7-month period.(19) Of note, the study did not include children younger than 5 years of age, a population in whom T&A is often performed and who may be most sensitive to the effects of disturbed sleep.(59) mild SDB may be associated with more severe neurobehavioral impairment that is more easily reversed with tonsillectomy.(21) Although there is likely both overuse and underuse of tonsillectomy in our healthcare system, effective and timely treatment for the right candidates could mean reducing substantial short and long-term health burdens associated with symptoms and comorbidities exacerbated by SDB, especially among vulnerable populations in which SDB is most prevalent.(60)

The results of the present study show that the overall likelihood of undergoing any intervention, PSG and/or surgery, was still lower in racial and ethnic minority groups compared with White children. A sub-analysis did show differential use of PSG by race, particularly with Black children obtaining the highest frequency of sleep studies, followed by the Other group. One explanation for this finding may be a higher prevalence of obesity among Black children, which has been shown to be more common in this population along with severe OSA.(29) Our study design purposely excluded high-risk children that would warrant PSG based on perioperative comorbidities and age, however, obesity and discordant clinical examination were factors not reliably captured in the database. It has been suggested that children with SDB that undergo T&A are more likely to be obese, and Black children who are obese are more likely to have SDB.(61) Conversely, Redline et al showed that the increased risk of SDB in Black patients is independent of the effects of obesity or respiratory problems.(29) In our study, PSG utilization was highest in Black children, suggesting that race may not be a limiting factor in access to care for this diagnostic procedure or that this observation may be due to other factors not readily captured in the database, such as patient or provider preference. For example, obtaining PSG before consideration of tonsillectomy may be a reflection of racial variation in preferences towards nonsurgical and less invasive treatment options. Moreover, individual social and relational elements may influence how physicians refer patients and recommend treatment options for children with SDB.(62) This may account for the finding that White children had the highest rate of direct surgical intervention without PSG after receiving a diagnosis of SDB. One study found that a greater disposition for tonsillectomy was observed in non-Hispanic White parents and parents of patients with SDB who had been previously evaluated by a pediatrician.(63)

A number of patient, provider, and system level factors could account for our observations. It has been shown that communities with more Black and Hispanic residents have fewer ambulatory surgery centers and lower overall outpatient surgery use.(64) Further considerations include parental educational level, primary language, cultural beliefs, transportation and logistical challenges, and systemic racism.(65) Racism is a social determinant of health that negatively impacts the well-being of children and their families through implicit and explicit biases, institutional structures, and interpersonal relationships; failure to identify and address racism at its institutional, personally mediated and internalized levels will continue to threaten health equity for all children.(66) Although structural racism may lead to differences in the manner in which families are counseled about treatment options, a lack of trust in healthcare providers among the Black community has important implications in medical decision-making.(67) Individuals that experience health inequities also experience sleep deficiencies that are associated with a wide range of suboptimal health outcomes and poorer overall functioning.(68) Therefore, it is crucial to provide equitable diagnostic and surgical care to vulnerable groups with SDB, because better sleep quality for individuals exposed to childhood adversity is associated with improved social functioning, emotional regulation, and academic performance.

The strengths of the present study include use of a large, comprehensive national claims database encompassing multistate data rather than single state or institutional data. Such aggregate population-based data are able to inform us of practice patterns on differential utilization of T&A as well as PSG. Limitations of the present study are related to the nature of claims data, including variability in provider coding. In the absence of reported patient symptoms, clinical findings, and PSG results, we strategically selected ICD-10 codes that were representative of obstructive SDB. However, accurately capturing children with SDB using codes alone is a challenge given that SDB represents a continuum of disease. For this reason and differences in coding practices by specialists, we chose to include diagnoses ranging from primary snoring to OSA. We recognize the limitations of including snoring as a code as this may comprise children without hypopneas and apneas, however, the frequency of parentally reported snoring, rather than apnea-hypopnea index, has been shown to predict cognitive and behavioral problems in young children.(69) Moreover, several studies (70–72) including a systemic review of the literature,(73) have shown that clinical evaluation is inaccurate at distinguishing primary snoring from obstructive sleep apnea syndrome. Parental reports using even validated surveys such as the sleep questionnaire also do not robustly correlate with OSA severity.(74)

These data were based on a large convenience sample and do not represent a random sample; therefore these results are not generalizable to all populations. Although the Marketscan database covers broad geographic regions, it does not include patient-level identifiers regarding geographic location and we were therefore not able account for regional differences in tonsillectomy rates, which has been cited as a significant factor in differing rates of pediatric tonsillectomy across the country.(75) The Marketscan database also lacks information including household income, parental education level, and census tracts, therefore we could not confirm whether Medicaid enrollment acted as a surrogate for low SES despite the qualifying criteria to meet federal poverty levels among non-medically complex children. Our analysis did not address uninsured or underinsured children for whom barriers in access to surgical care highly impact. Lastly, patients and families that completed PSG and had a surgical option recommended but subsequently declined cannot be captured in claims databases.

Our finding of disproportionately low T&A rates among nonwhite children insured with Medicaid constitutes a disparity in healthcare access and utilization. Such differential use may represent inequitable access to care, particularly because racial minority groups are at an increased risk of SDB. Mitigating health inequities requires an understanding of the interrelated social determinants of health with the goal of developing targeted and culturally appropriate interventions. Although the reasons for the observed findings are not discernable using solely claims data, these results provide impetus to further investigate the multilevel factors behind racial/ethnic disparities in the diagnosis and treatment of SDB.

List of Abbreviations:

AAO-HNS

American Academy of Otolaryngology-Head and Neck Surgery

ICD-10

International Classification of Diseases, 10^th Revision

PSG

polysomnography

SDB

sleep-disordered breathing

T&A

tonsillectomy with or without adenoidectomy

US

United States