Abstract
Abstract
BACKGROUND
For those suffering from Parkinson’s disease, deep brain stimulation (DBS) offers significant relief from the cardinal symptoms of the disease as well as the dyskinesias and motor fluctuations that emerge after long-term medical management with levodopa [1]. Both the subthalamic nucleus (STN) and the globus pallidus internus (GPi) have been shown to be effective stimulation targets to relieve the motor symptoms related to Parkinson’s disease and to improve quality of life [2, 3]. Consequently, other outcomes of DBS, such as the impact on mental health, can now be examined to help determine which target may be preferred for individual patients. Several studies have begun to investigate outcomes of DBS on mood and mental health with varied results. Some studies have shown increased symptoms of depression and psychosis following DBS [2, 4–6]. Studies using mood-rating scales, such as the Beck Depression Inventory, have produced mixed results. Some studies report improved mood ratings following surgery [7–9] while others found no difference [10–13]. In studies reporting differences in mood following DBS, most note changes in the first year following surgery, while studies looking at long-term outcomes generally report no difference [7, 15]. More specifically, in the larger study from which our Veteran population was drawn, there was a worsening in mood (assessed by the Beck Depression Inventory) after STN stimulation, and an improvement in mood after GPi stimulation at two years compared to baseline [2].
Despite a growing literature on occurrence of mental health-related adverse events including depression and psychosis following DBS for Parkinson’s disease [3, 17], studies have not characterized the impact of DBS on the use of mental health services and medications. This study sought to understand how DBS target location for Parkinson’s disease may affect use of these services. In addition to understanding potential differences in mood following DBS, it is beneficial to know if these changes affect healthcare use and costs. This study is the first to examine mental health healthcare use and costs following STN and GPi DBS for individuals with Parkinson’s disease. Mental health outpatient visits, medication use, inpatient admissions, and costs were compared by target site for a 36-month time period following DBS.
METHODS
This study retrospectively examined a subset of Veterans who participated in a large multi-site randomized two phase clinical trial which compared best medical treatment to DBS (phase 1) and STN to GPi DBS (phase 2) for Parkinson’s disease [2, 19]. The Human Studies Subcommittee of the Hines VA Research and Development Service approved the study, and consent was obtained from all patients. The enrollment criteria and study procedures for the clinical trial have been previously described in detail [2, 18]. Briefly, patients with idiopathic Parkinson’s disease were eligible for the clinic trial if they were Hoehn and Yahr stage ≥2 (off medication), age ≥21, and levodopa responsive with disabling motor symptoms. Exclusion criteria included prior surgery for Parkinson’s disease, active alcohol or drug abuse problems, and dementia (MMSE, <25 or Mattis DRS >2 SD below the mean of age-matched controls).
Data source
We examined 161 Veterans who were randomized to receive GPi (n = 85) or STN (n = 76) DBS and consented to have their medical records reviewed. Data gathered from the clinical trial and national VA Medical SAS data sets were used to compare demographics, healthcare encounters (outpatient and inpatient), and medication use. Data available from the clinical trial also included patient functional status, quantified from the Unified Parkinson’s Disease Rating Scale (UPDRS) [20], and quality of life using the Parkinson’s Disease Questionnaire (PDQ39) [21]. We also used clinical trial data to assess the quantity of Parkinson’s disease medication use at all time points by converting Parkinson’s disease related medications to levodopa equivalent daily doses (LEDD) [22].
We included healthcare encounters for which the primary diagnosis code was mental health-related (ICD-9 codes: 290–319). We excluded 3% (23 of 751) of the outpatient encounters where the primary diagnosis code was mental health-related but the codes identifying the location of the appointment were unlikely to be mental health-related (e.g. urology). The majority of visits with mental health-related primary diagnosis codes were for appointments in individual or group mental health clinics.
Medication prescription data were acquired from the VA Decision Support System (DSS) pharmacy files. We selected mental health-related medications based on relevant VA Drug Classes (i.e., antidepressants, benzodiazepine derivatives, anticonvulsants, antipsychotics, sedative/hypnotics) for analysis.
Cost data were available from the VA DSS National Data Extracts (NDEs) for outpatient visits, medications, and hospital admissions. DSS extracts information from the VA’s accounting and payroll system and combines this with workload information from patient care and administrative departments to produce cost estimates. Costs of prescriptions from VA pharmacies were based on the VA’s acquisition and dispensing costs in the VA DSS NDE Pharmacy datasets [23].
Statistical analysis
Veterans’ age and Unified Parkinson’s Disease Rating Scale (UPDRS) part III (clinician-scored motor evaluation) scores were compared by DBS target group using t-test; race, sex, and marital status were compared by DBS target group using chi-square analysis for the proportion of patients in each group. We performed a repeated measures analysis of variance (ANOVA) to examine the effects of treatment group and time on mental health outpatient clinic visits and medication use. Mental health outpatient visits, medications, and inpatient utilization by DBS target were additionally examined using chi-square analysis. We performed a logistic regression analysis to assess the associations of mental health medication use, DBS target group, and changes in Parkinson’s medications (using levodopa equivalents [LEDD]). Finally, we estimated total costs per patient for mental health service utilization and medication use by GPi and STN. We calculated 95% confidence intervals for the differences in means between groups using non-parametric bias-corrected accelerated bootstrap procedures.
RESULTS
Patient demographics
At baseline, the two treatment groups, GPi (n = 85) and STN (n = 76), did not differ in demographic characteristics, functional status, or mental health care use (Table 1). Veterans had a mean age of 61 years, and the majority was white (99% ) and married (83% ) (Table 1). The GPi group had significantly more males than the STN group (100% vs 94% , p = 0.05). UPDRS III motor scores off and on medication did not differ at baseline. In the 6 months prior to the DBS surgery, mental health outpatient visits and mental health medication use did not differ by group; approximately 25% of patients utilized mental health outpatient care and 54% of the patients used mental health medications (Table 1).
Mental health outpatient clinic use
Approximately 25% of patients used mental health outpatient services at any of the study assessment points over 36 months. There were no significant differences in the proportion of patients with mental health outpatient services by treatment group or by time by ANOVA. These results were consistent with chi-square analysis of use at each 6-month time interval from baseline to 36-months (Fig. 1A).
For patients with mental health outpatient visits, the most frequent mental health ICD-9 code was depressive disorder NEC (not elsewhere classified) with 23% of patients overall receiving this ICD-9 code within the 36-month follow-up period (Table 2). Other prevalent mental health diagnostic codes included mentaldisorder NOS (not otherwise specified) (17% ), other mood disorder (10% ), and anxiety state NOS (5.6% ). The proportion of patients with these diagnostic codes did not differ by treatment group (Table 2).
Mental health medication use
Approximately 50% of patients received mental health prescriptions at any of the study assessment points over 36 months of follow up. An ANOVA showed significantly greater mental health medication use during both the baseline to 6 month (p = 0.02) and 6 to 12 month (p = 0.003) time intervals. In addition, the ANOVA showed an interaction effect of time period and DBS target site, with greater mental health medication use for STN patients during the 6 to 12 month interval (p = 0.02). Chi-square analyses demonstrated a significantly larger proportion of STN than GPi patients received mental health medications from VA pharmacies 6 to 12 months post-surgery (69.7% vs. 52.9% , chi-square = 4.75, p = 0.03) (Fig. 1B).
Antidepressants were the most frequently prescribed psychiatric medications, with 63% of patients receiving prescriptions for antidepressants over the 36-month period (Table 2). Other commonly prescribed mental health drug classes included anticonvulsants (35% ), benzodiazepine derivatives (32% ), and antipsychotics (26% ). There were no significant differences in prescribed drug classes by DBS target (Table 2).
Prior to DBS, 54% of the Veterans (87 patients) received mental health medications. Within the first six months following DBS, an additional 9 GPi patients (10.5% ) and 15 STN patients (19.7% ) received new mental health prescriptions. In the 6 to 12 month interval following surgery, an additional 3 GPi patients (3.5% ) and 6 STN patients (7.9% ) received new mental health prescriptions. However, of the 33 new patients receiving mental health prescriptions in the year following DBS, only 8 of these patients (5 GPI and 3 STN) continued to receive mental health prescriptions through the 36-month period. All but one of these new users received antidepressants.
Relationship between LEDD and mental health medication use
In the logistic regression analyses, we examined the relationship between mental health medication use, DBS target, and change in LEDD following surgery. There was a significant difference in mental health medication use by DBS target site during the 6 to 12 month post-operative period (p = 0.01). A trend towards difference in mental health medication use based on change in LEDD (p = 0.05) was noted.
Mental healthcare costs
We also examined healthcare costs associated with mental health use (i.e., outpatient clinic visits, medications, and inpatient admissions) following DBS. There were no statistically significant differences in costs by group for medication use or inpatient visits at any time interval (Table 3). There were statistically significant differences in cost for outpatient clinic visits for the six months prior to surgery and the six months following surgery with greater costs for the GPi group (p < 0.05) (Table 3). There were no significant differences in total cost for mental health outpatient clinic visits, medications, or inpatient visits over 36 months (Table 4).
DISCUSSION
This study sought to characterize mental health healthcare utilization, medication use and associated costs by STN and GPi target over time following DBS for Parkinson’s disease. The results of this study suggest that mental health healthcare utilization (outpatient and inpatient visits) did not change over time or by DBS target site. The proportion of patients with mental health outpatient visits did not differ by DBS target site and did not change as a result of DBS surgery. Mental health inpatient utilization did not differ significantly by DBS target site or time with only a small percentage of patients requiring inpatient admissions for mental health diagnoses (5.6% overall). Depression was the most prevalent mental health diagnosis overall (23% ) and antidepressants were the most commonly prescribed mental health medication (63% ).
Unlike outpatient and inpatient visits, mental health medication use did significantly differ between surgical target sites and over time, with a greater proportion of patients using mental health medications in the year following surgery. In the 6 to 12 month interval following surgery, the STN DBS group had a greater increase in mental health medication use than the GPi DBS group. However, of the 33 patients with new mental health prescriptions in the year following surgery, all but 8 had their prescriptions discontinued by 36 months, suggesting that these might be transitory changes related to DBS intervention and not chronic post-operative mental health conditions.
It is unclear why those in the STN DBS group show temporary increases in mental health medication use. Several studies have investigated changes in mood and demonstrated worsening mood following DBS with most of the changes occurring in the first year [2, 11], though others have shown improvement [7–9]. In the clinical trial from which our Veteran sample was drawn, there was a decrease in mood after STN and an improvement after GPi stimulation at two years compared to baseline using the Beck Depression Inventory [2]. However, the current study suggests these differences in mood do not translate into sustained differences in mental health healthcare use in Veterans beyond the one-year increase in mental health medication use. It is possible that STN electrode location alters mood more so than the GPi target. Okun et al. (2009) did find that stimulation ventral to the optimal DBS stimulation sites produced worse subjective reported mood scores, however they found no difference between unilateral STN versus GPi groups [16]. It is also possible that the greater reduction in LEDD following STN DBS compared to GPi DBS may have a temporary effect on patient mood [11, 24–27]. Consistent with this, our data showed a trend in the relationship between LEDD reduction and greater likelihood of mental health medication receipt. While the proposed effects of STN DBS on mood should not be minimized, they may be related to the significant LEDD reduction for STN DBS compared to GPi DBS, which would require further studies with greater power to investigate. Furthermore, this may lead to protocols that minimize large changes in LEDD following surgery to minimize mood disparities [11].
Additionally, the differences in mental health medication use in the first year following surgery do not align with outpatient mental health use as might be expected. This may suggest that the intensity of mood changes is not substantial enough to demand regular outpatient clinical evaluation or that the symptoms resolved quickly enough with medication that outpatient visits are not reflective of the mood findings. Furthermore, this discrepancy may reveal barriers such as difficulty obtaining access to outpatient services.
Overall costs associated with mental health use, while substantial, also did not differ between DBS target groups. Over time, only the costs related to outpatient clinic visits showed a difference between groups, with the GPi group having greater costs than the STN group in the 6-months before and after surgery. These differences are not reflected in the overall proportion of patients requiring mental health outpatient visits at these times and therefore may be reflective of a few individuals who were requiring more outpatient visits during this time. It should also be noted that despite just 9 patients requiring inpatient stays for mental health needs, these admissions comprised a substantial portion of mean total costs. The costs of mental health admissions emphasize the need to follow patients with mental health changes closely in attempt to minimize these hospitalizations.
Some limitations to the study should be considered. First, we examined mental health use only for patients enrolled at the VA sites from the original randomized trial, since we had comprehensive information about Veteran healthcare visits, prescriptions, and associated costs. Therefore, our sample size was smaller and not sufficiently powered for some of the outcomes of interest. Further, the findings may not be generalizable to the non-Veteran population. The groups did not differ demographically except for a greater proportion of males in the GPi group compared to the STN. It should be noted that this difference is by approximately 5% , with both having predominately males. As outlined above, the high percentage of males is reflective of the Veteran population used for this study. To ensure that the difference in females was not influencing the data, we reran the total cost and utilization analyses excluding the women and found no differences. Finally, as this is a post-hoc analysis, we cannot be certain about the reasons for medication prescription or consistent use of ICD-9 codes.
This is the first study to investigate mental health-related healthcare utilization and costs following DBS for Parkinson’s disease. The past literature has primarily assessed mental health outcomes following DBS for Parkinson’s by examining adverse events (including psychosis, depressive episodes) [2, 28] and by assessing mood changes using tools such as the Beck Depression Inventory [8–10, 16] with mixed results. Costs have been investigated for the target groups on a whole [29] and for specific Parkinson’s disease medication use [27] but have not directly analyzed for mental health treatment. The current study used mental health-related healthcare utilization and costs as a way to operationalize the effects of DBS on patients’ mental health. Despite suggestions for mood differences between targets in the main study, this did not translate into sustained differences in mental healthcare utilization or cost.
CONFLICT OF INTEREST
The authors have no conflict of interest to report.
Footnotes
ACKNOWLEDGMENTS
The authors wish to acknowledge the Department of Veterans Affairs (DVA) Health Services Research & Development program (IIR 08-124) for funding the current study. Additionally authors wish to acknowledge the Cooperative Studies Program, Office of Research and Development, and the National Institute of Neurological Diseases and Stroke, for their funding support of the original clinical trial.