Disparities in Access to Deep Brain Stimulation Surgery for Parkinson Disease: Interaction Between African American Race and Medicaid Use

Importance  African American individuals experience barriers to accessing many types of health care in the United States, resulting in substantial health care disparities. To improve health care in this patient population, it is important to recognize and study the potential factors limiting access to care.

Objective  To examine deep brain stimulation (DBS) use in Parkinson disease (PD) to determine which factors, among a variety of demographic, clinical, and socioeconomic variables, drive DBS use in the United States.

Design, Setting, and Participants  We queried the Nationwide Inpatient Sample in combination with neurologist and neurological surgeon countywide density data from the Area Resource File. We used International Classification of Diseases, Ninth Revision codes to identify discharges of patients at multicenter, all-payer, nonfederal hospitals in the United States diagnosed with PD (code 332.0) who were admitted for implantation of intracranial neurostimulator lead(s) (code 02.39), DBS.

Main Outcomes and Measures  We analyzed factors predicting DBS use in PD using a hierarchical logistic regression analysis including patient and hospital characteristics. Patient characteristics included age, sex, comorbidity score, race, income quartile of zip code, and insurance type. Hospital characteristics included teaching status, size, regional location, urban vs rural setting, experience with DBS discharges, year, and countywide density of neurologists and neurological surgeons.

Results  Query of the Nationwide Inpatient Sample yielded 2 408 302 PD discharges from 2002 to 2009; 18 312 of these discharges were for DBS. Notably, 4.7% of all PD discharges were African American, while only 0.1% of DBS for PD discharges were African American. A number of factors in the hierarchical multivariate analysis predicted DBS use including younger age, male sex, increasing income quartile of patient zip code, large hospitals, teaching hospitals, urban setting, hospitals with higher number of annual discharges for PD, and increased countywide density of neurologists (P < .05). Predictors of nonuse included African American race (P < .001), Medicaid use (P < .001), and increasing comorbidity score (P < .001). Countywide density of neurological surgeons and Hispanic ethnicity were not significant predictors.

Conclusions and Relevance  Despite the fact that African American patients are more often discharged from hospitals with characteristics predicting DBS use (ie, urban teaching hospitals in areas with a higher than average density of neurologists), these patients received disproportionately fewer DBS procedures compared with their non–African American counterparts. Increased reliance on Medicaid in the African American population may predispose to the DBS use disparity. Various other factors may be responsible, including disparities in access to care, cultural biases or beliefs, and/or socioeconomic status.

Deep brain stimulation (DBS) is an important treatment option in the long-term management of Parkinson disease (PD). Although DBS remains an elective procedure, 89% of patients with PD ultimately develop dyskinesias and motor fluctuations within 10 years of initiating medical treatment with levodopa.¹ A proportion of these patients with severe dyskinesias, medically intractable tremors, or motor fluctuations will be carefully selected as candidates for DBS. Indeed, multiple randomized clinical trials have shown DBS to be more effective than continued medical management in well-selected patients.^2-5 Since Food and Drug Administration approval in 2002, the age of patients undergoing DBS is increasing steadily without a corresponding increase in complication rates.⁶ These studies strongly suggest that DBS is both an efficacious and safe procedure for patients with PD.

Despite the abundance of evidence demonstrating the effectiveness of DBS for PD, this promising therapy may be underused in certain patient populations, particularly minorities. In a study performed during the emergence of DBS for PD, Eskandar and colleagues⁷ found that 85.9% of patients undergoing surgery for PD were white, with African American patients comprising only 0.6% of total cases. This stands in stark contrast to the proportion of African American patients diagnosed with PD (4.8%-28%).^8-13 Studies in other surgical fields have observed similar discrepancies, demonstrating that African American patients received disproportionately fewer surgical procedures compared with other racial groups.^14-22 To the best of our knowledge, no studies exist that examine potential barriers to receiving DBS that could account for DBS underuse in the African American population. We examined the Nationwide Inpatient Sample (NIS) in conjunction with the Area Resource File (ARF) to identify factors that may affect access to this important therapy.

Institutional review board approval was not required for this study using a publicly available, national database. Data were collected from the NIS (Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality),²³ a stratified sample of all patient discharges from approximately 20% of nonfederal hospitals in the United States from 2002 to 2009. Each discharge in the data set is weighted by the Healthcare Cost and Utilization Project to extrapolate the total annual patient discharge information. Parkinson disease discharges were identified using International Classification of Diseases, Ninth Revision (ICD-9) codes for PD (332.0). Methods from previous studies selected discharges within the NIS using only the primary diagnosis field, extracting only a subset of the entire PD population. Our method allowed for any of 15 diagnosis fields to contain PD with the expectation that this would be more representative of the true population of patients with PD while also increasing our sample size. The rationale for this decision is supported by the finding that patients with PD are most often hospitalized for comorbid diseases, with only 15.0% admitted expressly for PD.²⁴

We combined data from the NIS with the ARF,²⁵ a basic county-specific database containing more than 6000 socioeconomic and environmental variables for each of the nation’s counties. The ARF provides data detailing the number of neurologists and neurosurgeons by county, among other variables. We linked the most recent ARF file (2011-2012) by county Federal Information Processing Standard code to the NIS discharge data in accordance with a previously described method.²⁰ This allowed the most recent and extensive information regarding neurologist and neurosurgeon density (2010) to be matched to the NIS data set (2002-2009). Relative neurologist and neurosurgeon countywide densities did not significantly fluctuate (range of average density, 4.29-4.74 neurologists/100 000 individuals and 1.66-1.76 neurosurgeons/100 000 individuals; P = .69 and P = .84, respectively) in different years within the ARF.

Age, sex, modified comorbidity score, race, income quartile of the patient’s zip code, form of payment for hospital admission, and presence of ICD-9 code for implantation/replacement of intracranial neurostimulator lead(s) (02.39) were extracted from the NIS database. The NIS provides data for primary and secondary payers for each patient discharge. We combined both categories of payers into 1 of 4 mutually exclusive categories: “private insurance,” “Medicaid without private insurance,” “Medicare with neither private insurance nor Medicaid,” and “other.” The NIS provides 6 categories for race/ethnicity: “white,” “black,” “Hispanic,” “Asian/Pacific Islander,” “Native American,” and “other.” We will subsequently refer to the NIS “black” categorization as “African American.” Medical comorbidities were defined using a modified version of the Elixhauser comorbidity score.²⁶ The score is an assessment of the general comorbidity associated with a given patient, which includes a set of 30 comorbidity markers. We used a previously described⁷ modification of the score, which excluded the 2 neurological comorbidity variables, “other neurological deficit” and “paralysis,” such that the highest possible comorbidity score was 28.

Hospital size (small, medium, large), region (Northeast, Midwest, South, West), setting (urban or rural), teaching hospital status, year of discharge, experience (number of discharges by hospital with ICD-9 code of PD), and density of neurologists and neurological surgeons by hospital county were identified in the ARF and NIS databases.

Univariate analyses used Mann-Whitney U tests and χ² tests using built-in and custom Matlab scripts (Matlab; MathWorks). A hierarchical logistic regression model (SAS procedure GLIMMIX; SAS Institute) was used to analyze variables that predicted DBS use. Predictor variables included age, sex, modified comorbidity score, race, income quartile of patient zip code, payer status, hospital experience, size (number of patient beds), region, setting (urban or rural), teaching status, year of discharge, and the density of neurologists and neurological surgeons. The unique hospital identification code served as the nesting variable. We accounted for missing data with single imputation based on deterministic regression modeling in an R environment (R Development Core Team). Multinomial logistic, binomial logistic, and linear regression models were used as appropriate. P < .05 was considered statistically significant. Statistics are reported by sample mean (SD).

From 2002 through 2009, there were 2 408 302 PD discharges from nonfederal hospitals in the United States; 114 168 (4.7%) of these were African American, and 18 312 (0.8%) of the total PD discharges represented patients who had been admitted for DBS. Among this population of patients with PD admitted for DBS, only 120 (0.1%) represented DBS for African American PD discharges. Unadjusted analysis revealed African American PD discharges to be associated with significantly fewer DBS procedures relative to non–African American discharges (unadjusted odds ratio, 0.13; 95% CI, 0.11-0.16; P < .001). Table 1 shows the patient and hospital characteristics of discharges with the diagnosis of PD and discharges of DBS for PD from 2002 through 2009.

Figure 1 shows the total number of discharges for both PD and DBS for PD between 2002 and 2009. The racial composition of discharges and DBS for PD discharges as represented by Figure 2 highlights the isolated discrepancy between the proportion of African American PD discharges and the proportion of African American PD discharges for DBS.

Controlling for patient and hospital factors that impact DBS use in patients with PD, African American status (P < .001) and use of Medicaid relative to Medicare and private insurance (P < .001) predicted nonuse of DBS. Interestingly, Hispanic ethnicity did not have a significant effect on DBS use (P = .74).

For the hospital-related variables, teaching status, urban setting, and large size were positive predictors of DBS use (all P < .001), as well as greater number of annual discharges for PD (P = .008). Notably, the average countywide density of neurologists predicted DBS use (P = .006) while the density of neurological surgeons had no significant effect on DBS use (P = .97). Table 2 shows the odds for use of DBS for PD for the variables included in the hierarchical model.

Table 3 shows the patient and hospital characteristics of African American discharges for PD from 2002 through 2009. To assess the factors predicting DBS use in the African American population alone, we repeated the hierarchical multivariate model for African American discharges (urban vs rural setting was removed for model convergence). Table 3 shows the odds ratios for DBS use in African American patients with PD for the variables included in the hierarchical model. Notably, factors that predicted DBS use in the general PD population (sex, discharge zip code income, hospital size, year, hospital experience, and countywide density of neurologists) did not predict use in African American patients with PD (P > .05). However, Medicaid use, relative to private insurance or Medicare use, continued to predict lack of DBS use in this population (P < .001). This finding suggests an important role for Medicaid status in determining DBS use in the African American PD population.

Given the negative effect of Medicaid use within both the general and African American PD population, we repeated the hierarchical multivariate model stratifying each race/ethnicity into Medicaid and non-Medicaid groups with an additional predictor denoting each combination. Surprisingly, white patients using Medicaid received significantly more DBS surgeries than African American patients using private insurance and Medicare (odds ratio, 2.23; 95% CI, 1.47-3.37; P < .001), suggesting a unique combination of Medicaid and race/ethnicity as the cause of the present access disparity. Indeed, African American patients with Medicaid underwent the fewest DBS procedures of any other group combination of race/ethnicity and insurance status while African American patients with non-Medicaid insurance underwent the second fewest procedures.

To our knowledge, our investigation is the first to study the variables affecting DBS use in patients with PD. When controlling for patient- and hospital-related factors, African American PD discharges are nearly 8 times less likely to undergo DBS surgery relative to white patients. This is consistent with prior investigations that demonstrate reduced access to surgical care for African American patients,^16-19,22 including neurosurgical procedures.^14,15,20,21 Other studies also demonstrate access to care and treatment disparities for African American patients with PD,^27,28 including underrepresentation in clinical trials for PD.^29-31 Given this finding, we sought to examine the potential barriers to DBS in this population.

It has been suggested that African American patients with PD are less likely to receive specialized care (eg, DBS surgery) because of medical comorbidities.^32,33 Indeed, our study reveals that African American PD discharges were associated with higher mean comorbidity scores. Yet, African American race still predicted decreased DBS use after adjusting for comorbidity score. This finding is consistent with data from the National Institutes of Health Exploratory Trials in Parkinson’s Disease demonstrating that comorbidities were less likely to be the reason minorities were underrepresented in study participation.³¹ In addition, recent evidence has shown that more DBS procedures are being performed on patients with PD with higher comorbidity scores over the same period.³⁴ Taken together, these observations suggest that the apparent disparity in the African American patients receiving DBS stems from underrepresentation in the initial screening population.³¹

Geographic concerns, especially access to practitioners, have also been suggested as a major barrier to surgical care.¹⁸ Yet our analysis directly contradicts this theory. We show that despite being more frequently discharged from urban teaching hospitals in areas with a higher than average density of neurologists and neurosurgeons, African American PD discharges underwent disproportionately fewer DBS procedures.

Several studies have shown that reduced income impacts access to neuro-oncologic care,²⁰ resection of pituitary tumors,²¹ and DBS use.⁷ Still, our analysis showed that income was not a significant predictor of DBS use in African American patients with PD. Alternatively, type of insurance is an important predictor of DBS use in the African American PD population.

African American patients with PD were significantly more likely to use Medicaid. The proportion of patients with PD using Medicare was similar between African American patients and the overall discharge population, with the increase in African American Medicaid discharges paralleled by a decreased percentage of African American discharges using private insurance. Medicaid use remained the only significant predictor of not receiving DBS, after controlling for other predictors of DBS use. It is reasonable to surmise that African American patients with PD on Medicaid have a systematic disadvantage in accessing DBS surgery.

In our experience, neurologists tend to refer potential patients with PD to neurosurgery if they are judged to have adequate social support. The rationale is that patients with deep brain stimulators need well-coordinated neurological and neurosurgical follow-up. Lin and Popp³⁵ have suggested that those with private insurance have better access to primary care, a better skill set to negotiate among various treatment options, and more preoperative and postoperative social support. Our analysis shows that African American patients receiving DBS rely more heavily on Medicaid, which, by definition, serves those in a disadvantaged background. Future studies should attempt to address the gaps in care experienced by these groups to reduce the barriers to DBS and increase referrals for surgical treatment.

Other neurosurgical studies have shown that those with government insurance are less likely to undergo surgery than those with private insurance.^7,15,36 This phenomenon may reflect practitioner avoidance of increased administrative burdens and low monetary reimbursement from government-funded insurance³⁷ or differential denial of reimbursement by insurance type.³⁸ Still, although individual insurance reimbursement varies between individual plans and across states, most Medicare, Medicaid, and private insurance plans cover DBS for PD. Indeed, under the National Coverage Determination for Deep Brain Stimulation,³⁹ government-funded insurance will cover DBS for PD if proper inclusion criteria are met. In our experience, almost all patients with PD use (1) only Medicare or private insurance or (2) 80% Medicare with a 20% supplement by a secondary insurance carrier for DBS. Thus, it appears the decreased DBS use among those relying on Medicaid alone (of which there is a higher proportion in the African American PD population) may be indicative of more proximate access issues (eg, disadvantaged background, less knowledge of available treatment options) rather than procedural reimbursement. Medicaid is likely one of the myriad factors influencing DBS use in the African American population and should be explored further in subsequent studies.

In our stratified regression analysis, separating racial/ethnic identifiers into Medicaid and non-Medicaid groups, we found that white patients with PD using Medicaid received significantly more DBS surgeries compared with African American patients with PD not using Medicaid. This supports a multifactorial etiology of DBS underuse that includes both socioeconomic factors and cultural factors.

Cultural differences between racial and ethnic groups represent an important variable for which our present study cannot account. Cultural differences in the perceptions, attitudes, spirituality, and symptom expressivity of African American patients^13,40-42 play a prominent role in the access to care disparity. Future studies are necessary to investigate the extent to which these differences generalize to the PD population.

The NIS is a retrospective data set with inherent limitations. The NIS records account for unique patient discharges, which do not necessarily correspond to distinct patients with PD. Accordingly, patients with similar attitudes or biases for or against DBS, or other factors that may covary with receiving DBS (eg, poorly controlled PD), may be overrepresented in the database in the form of repeated admissions and subsequent discharges. The inclusion of discharges with any diagnosis of PD may be susceptible to selection bias. However, a supplemental analysis (eTables 1, 2, and 3 in Supplement) using solely discharges with a primary diagnosis of PD yielded consistent results. Our study is also limited by variability and errors in hospital reporting of race/ethnicity because the NIS does not enforce a standard method of racial/ethnic reporting.

We sought discharge records using the ICD-9 code for PD and provided a comorbidity score associated with each discharge. Still, this is unable to capture the full clinical context of patients with PD. For example, minorities tend to be seen with more advanced disease⁴³ and may have less access to medication,⁴⁴ less frequent visits to practitioners,⁴⁵ less secure social support networks,⁴⁶ and a higher prevalence of exclusionary criteria to DBS such as dementia.^47,48 These factors likely contribute to the disparity in DBS use. Additionally, the ICD-9 code 02.93 (implantation or replacement of intracranial neurostimulator lead[s]) leaves the possibility that a small number of discharges reflect revisions rather than initial DBS implantation. Still, there remains no a priori rationale for the revision proportion of discharges with ICD-9 code 02.93 to vary by race/ethnicity. Furthermore, implantation records from Medtronic, the sole manufacturer of deep brain stimulators, reveals that there were 19 957 new deep brain stimulators for PD placed from 2002 to 2009 (R. DiTota, MBA, Medtronic, written communication, December 7, 2012), suggesting our data (n = 18 312) comprise primarily new DBS electrode implantations.

Last, we corrected for access to neurologists yet we cannot account for access to movement disorder specialists, specifically. Movement disorder specialists may provide higher-quality referrals for DBS than general neurologists.⁴⁹ It is possible that African American patients have limited access to movement disorder specialists, resulting in a lack of DBS use. Altogether, these shortcomings should be understood in the context that this is the largest database reporting the racial compositions of the DBS and PD populations in the United States.

Our study is limited to a period (2002-2009) where DBS for PD was used primarily for those whose medical control failed and who developed motor fluctuations, expressed overwhelming adverse effects, and had generally moderate or advanced PD. In 2009, the first of a series of highly influential studies reported superiority of DBS over “best medical management” and advocated for earlier intervention with DBS therapy.³ Thus, the impact of the disparity in the access to DBS for PD may be even greater in recent years.

Ultimately, DBS remains an elective procedure for carefully selected patients with PD whose disease is suboptimally controlled by medical management. This represents a minority of patients with PD. Nevertheless, access to DBS should not vary solely on racial or ethnic background. Our study shows that African American patients are singularly less likely to receive DBS for PD, even despite the presence of predictors that positively predict receiving DBS in the general PD population. Future studies should address this disparity in access to DBS in the context of rising health care costs and elective, quality of life–related surgery.

To our knowledge, this is the first study that attempts to identify the potential barriers to receiving DBS for PD. When controlling for patient and hospital characteristics as well as the geographic availability of clinical specialists, African American status alone predicted nonuse of DBS. Access to neurologists, the gatekeepers of DBS, predicted DBS use while access to neurological surgeons did not have a significant effect on DBS use. This study suggests that type of insurance (ie, Medicaid vs non-Medicaid) is an important barrier to DBS use in the African American PD patient population. Future prospective studies must investigate the gaps in care experienced by African American patients with PD, particularly those using Medicaid, to improve access to DBS.

Corresponding Author: Andrew K. Chan, BS, Department of Neurological Surgery, Neurological Institute, Columbia University Medical Center, 710 W 168th St, New York, NY 10032 (akc2136@columbia.edu).

Accepted for Publication: November 7, 2013.

Published Online: January 6, 2014. doi:10.1001/jamaneurol.2013.5798.

Author Contributions: Mr Chan and Dr McGovern had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Mr Chan and Dr McGovern share first authorship and responsibility for the integrity of the work as a whole, from inception to published article.

Study concept and design: Chan, McGovern, Sheehy, Zacharia, Mikell, McKhann.

Acquisition of data: Chan, McGovern, Sheehy.

Analysis and interpretation of data: Chan, McGovern, Brown, Sheehy, Zacharia, Bruce, Ford.

Drafting of the manuscript: Chan, McGovern.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Chan, McGovern, Brown, Sheehy, Zacharia, Bruce.

Administrative, technical, and material support: Ford.

Study supervision: McGovern, Sheehy, Zacharia, Mikell, McKhann.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by a grant from the Doris Duke Charitable Foundation to Columbia University to fund Clinical Research Fellows Mr Chan and Dr Sheehy.

Role of the Sponsors: The Doris Duke Charitable Foundation had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.