Medicare patient experience with vagus nerve stimulation for treatment-resistant depression

Abstract

Background:

Major depressive disease (MDD) represents a cost burden to the US healthcare system: approximately one-third of MDD patients fail conventional treatment: multiple failures define treatment-resistant depression (TRD). Vagus nerve stimulation (VNS) therapy is an approved adjunctive treatment for TRD.

Objective:

To study the healthcare utilization experience of Medicare beneficiaries implanted with VNS (VNSBs) during Medicare coverage, compared with beneficiaries with TRD (TRDBs) and managed depression (Mdeps).

Methods:

A retrospective analysis of 100% standard analytic file (SAF) Medicare claims from 2006–2009 using specific criteria to identify a VNSB dataset, compared to TRDs and Mdeps datasets (extract of 5% sample SAF from 2001–2009) and 2009 general Medicare beneficiaries (GMBs). Comparative analysis included demographics, mortality, healthcare utilization, and costs.

Results:

Among patients meeting study criteria for VNSBs (n = 690), TRDBs (n = 4639), Mdeps (n = 7524), and GMBs (n > 36 million), VNSBs were on average: younger, more likely to be female, and white, with Medicare eligibility due to disability. Of the VNSBs in the 2-year post-implantation period: 5% died; 22% experienced no negative events (defined as hospitalizations for psychoses or poisoning, emergency room use, electroconvulsive therapy, or poisoning, suicidal ideation, or self-harm diagnoses); 29% experienced multiple negative events; and 41% had either a single hospitalization or only all-cause ER visits. VNSBs experiencing negative events had more complex co-occurring psychiatric diagnoses. The annual mortality rate for VNSBs post-implant was 19.9 deaths per 1000 patient years, compared with 46.2 (CI: 41.9–51.6) and 46.8 (CI: 43.4–50.4) deaths for TRDBs and Mdeps, respectively. The medical costs per patient-year post-VNS implantation for VNSBs ($8749) was similar to the Mdeps ($8960; CI $8555–$9381) and was substantially lower than TRDBs ($13,618; CI $12,937–$14,342).

Conclusions:

VNSBs achieving positive health outcomes (measured by lack of negative events post-implantation) tend to have fewer psychiatric co-occurring conditions. Lowered costs post-implantation with evidence of response to VNS suggest the therapy represents an option for carefully screened TRDBs who have failed other therapies.

Limitations:

Administrative data are missing pharmaceuticals and clinical measures. Data for the VNS population were not available pre-implantation for comparison to post-implantation experience. Cost comparisons are adjusted for missing costs in the VNS dataset.

Keywords::

• Health outcomes
• Medicare beneficiaries
• Medicare costs
• Healthcare service utilization
• Treatment-resistant depression
• Vagus nerve stimulation

Introduction

Treatment-resistant depression (TRD) is defined in the literature as generally requiring at least two failures of adequate antidepressant therapy, and is a recognized diagnostic grouping of patients suffering repeated and/or chronic major depressive episodes1–3. The National Institute of Mental Health-sponsored STAR*D (Sequenced Treatment Alternatives to Relieve Depression) study assessed the likelihood of patients who were being treated for depression achieving remission of their symptoms4. The study found that if a patient did not achieve remission after two acute treatment steps, the likelihood of achieving remission with further drug treatments was very low (13.7%) with a third treatment, and 13% with a fourth treatment. In addition to the persistence of symptoms, patients with TRD utilize more healthcare resources. Consequently their care is associated with higher healthcare expenditures5–9. In general, the TRD literature describes a patient population that continues to experience the most debilitating effects of major depressive illness despite multiple pharmacologic interventions and often including electroconvulsive therapy (ECT).

A report released in September 2011 by the Agency for Health Care Research and Quality (AHRQ) summarized available studies describing the effectiveness of non-pharmacologic interventions for TRD and concluded that the comparative effectiveness literature is extremely limited with low strength of evidence for most of the interventions studied: comparative clinical research on non-pharmacological interventions in a TRD population is in its infancy1,2. Several small clinical studies do suggest sustained remission of major depression symptoms for 1–2 years with vagus nerve stimulation (VNS) therapy, as well as a relatively benign adverse event profile10–13. The 2010 American Psychiatric Association treatment guidelines for major depressive disorder discuss VNS therapy as a non-pharmacologic intervention to be considered as an option for adjunctive therapy for treatment-resistant depression14.

VNS therapy involves intermittent stimulation of the left cervical vagus nerve via an implantable pulse generator and lead and an external programming system used for adjusting stimulation settings15,16. VNS is a non-pharmacologic treatment that was approved by the US Food and Drug Administration (FDA) for the treatment of epilepsy in 1997 and for the treatment of TRD in 200517. Specifically, VNS therapy for TRD is indicated as an adjunctive treatment for patients with chronic or recurrent depression who have not responded to four or more adequately delivered antidepressant treatments. VNS therapy was prescribed for a limited number of Medicare beneficiaries with TRD, but a subsequent non-coverage decision by the Centers for Medicare and Medicaid Services (CMS) in 2007 essentially halted further use in this population. Among the reasons cited for the non-coverage decision was inadequate data for use in the Medicare population. Following this decision, other insurers were hesitant to approve payment for the therapy, and its utilization for TRD patients decreased, making it increasingly difficult to produce the data needed to evaluate its effectiveness.

The purpose of this study was to conduct a retrospective analysis of Medicare claims for those patients who were implanted with VNS during the time period when the therapy was covered by Medicare and for which data were available (January 1, 2006 to June 30, 2007) and to describe patient characteristics, healthcare service utilization, mortality, health outcomes, and costs of care for these patients. To establish a context for these findings, a parallel analysis was performed to provide the basis for comparing VNS patient experience with Medicare beneficiaries who have TRD but were not treated with VNS, as well as Medicare beneficiaries with depression that is managed with non-VNS treatment options (e.g., managed depressed). A further benchmark for some parts of the study is a comparison to the general Medicare beneficiaries in 2009. As distinct from clinical studies, this study uses administrative claims data to examine the entire population implanted with the VNS system during this Medicare coverage period, relying upon proxy data for the clinical screening criteria normally applied in clinical studies. It is intended to complement existing studies and is fundamentally descriptive.

VNS patient experience is compared to patients with TRD and managed depression identified from 2001–2009, a longer period than the VNS Medicare coverage period. The longer time period permits following Medicare patients with depression over 4–8 years to establish the context for the costs of care for this population, since they cannot be systematically matched to the VNS population.

During this longer time period, changes in pharmacotherapy would not have had a major impact on hospitalization for depression or service utilization. A few antidepressants were introduced during this period, some of which were not widely adopted (e.g., the selegeline patch). The use of anti-psychotics for augmentation of antidepressant non-response, introduced in the last few years, did not widely impact practice, as a definitive pathway to alleviate TRD was not established4. One important intervention during this time was the FDA decision to put a warning regarding suicidal thoughts and behavior on antidepressant medications, which may have contributed to decreased use of antidepressants, and may have affected suicide rates. The focus of this change was, however, on younger populations that are not the target population for VNS therapy.

The evidence available for assessing the effectiveness of VNS, as delineated by the AHRQ study, is based on studies with 100–300 patients or fewer, including those with major depression and bipolar depression diagnoses1,2. While lacking the clinical tests typically used to measure depression symptoms and disease severity (e.g., Hamilton Depression Rating Scale, Montgomery-Asberg Depression Rating Scale), administrative claims data can be used for a population-based assessment of patient experience and service utilization following an intervention. The study reported here is the first to examine Medicare beneficiary experience with VNS used to treat TRD. The study examined the experience of 690 Medicare beneficiaries implanted with VNS and followed for a minimum of 2 years.

Methods

Data sources

Medicare Standard Analytic Files (SAFs) for 100% of fee-for-service beneficiaries containing all inpatient and outpatient hospital claims were used to select patients who received adjunctive VNS therapy. The SAFs for the study period do not include exact dates of service, but do include a calendar quarter in which services were delivered. The calendar quarter including the implantation procedure is referred to here as the ‘index quarter’, and all patients were followed for 8–10 quarters following the index quarter. SAF data for these patients for prior years were not available to the research team except in a 5% beneficiary sample that was deemed inadequate to study patient characteristics and experience prior to VNS implantation. Therefore, this study is limited to patient experience following VNS implantation. A comparative analysis of TRD and managed depressed patients was performed using the Medicare SAF claims for a nationally representative 5% sample of fee-for-service beneficiaries from 2001–2009. Note that the SAFs do not include any prescription drug data, and therefore no evaluation of patient history with pharmacologic treatment failures was available.

Selecting patients who received adjunctive VNS therapy

Patients were selected if they were 18 years or older and had the VNS system implanted between January 1, 2006 and June 30, 2007 for a diagnosis of depression. Depression diagnoses used were International Classification of Diseases (Ninth Revision; ICD-9) codes: 292.2x, 296.3x, 300.4, 309.0, 309.1, and 311, representing major depressive disorder single and recurrent episodes, excluding ‘mild’, reactive depressions, adjustment disorder with depressed mood, prolonged depressive reaction, and depressive disorder not elsewhere classified. The study population includes patients with bipolar disease diagnoses, but bipolar diagnosis codes were excluded from the definition criteria to ensure that the study population is primarily suffering from major depressive disorder as distinct from a primarily bipolar disorder population with multiple depression codes in claims, and bipolar disorder was treated as a co-occurring psychiatric condition in this study. Patients with any claims history of epilepsy were excluded (ICD-9 codes 345.xx, 780.39, 293.0, 333.2). Given the depression indication for VNS, this study assumes all patients who did not have an epilepsy diagnosis and received the therapy met the depression indication criteria for VNS specified by the FDA.

Selecting patients with TRD and patients with managed depression

Definition of the TRD study sample was based on the following assumptions: (a) depressed patients may initially receive care from a primary care practitioner or clinic, as well as from a psychiatrist or mental health clinic; (b) if initial treatment for depression through a primary care provider fails, the patient would likely be referred to a psychiatrist or mental health clinic; (c) initial diagnosis and treatment may be initiated during a psychiatric hospitalization; (d) patients undergoing multiple treatment regimens for depression will eventually be managed by psychiatrists and in mental health clinics; (e) patients with regular psychiatrist visits, only a single psychiatric hospitalization, and no ECT are likely to be patients whose depression is managed (e.g., not treatment-resistant); and (f) patients with frequent psychiatric visits, any evidence of ECT, and two or more psychiatric hospitalizations in a 2-year period are likely to represent the TRD population.

A medication management visit is defined as a Medicare Part B claim (e.g., visit to a physician, outpatient clinic, or emergency room) in which a psychiatric medication management Current Procedural Terminology (CPT®) code was billed by a psychiatrist or primary care physician (general practice, family practice, or internal medicine), and the visit was not billed in the hospital inpatient place of service. A psychiatric hospitalization is defined as a hospital inpatient discharge from a psychiatric hospital, a psychiatric unit within a short-term acute care hospital, or the discharge is billed under a psychiatric Diagnosis-related Group (DRG) in a medical bed in a short-term acute care hospital. Hospitalizations that do not meet these criteria are defined as ‘medical’.

The TRD study sample was defined as a depressed patient (meeting above criteria) who also satisfied one of the following three conditions: (a) in the 2-year identification period the patient had between eight and 17 medication management visits, and had two or more psychiatric hospitalizations; (b) in the 2-year identification period the patient had between eight and 17 medication management visits, and had at least one ECT treatment; or (c) in the 2-year identification period the patient had 18 or more medication management visits.

Definition of a patient with managed depression was as follows: (a) having had at least one inpatient, outpatient hospital, or other Part B Medicare claim with a depression diagnosis in at least every two calendar quarters throughout a 2-year (eight calendar quarters) period; and (b) during that 2-year period the patient had at least two non-inpatient Part B claims with a depression diagnosis made by a psychiatrist. If a patient had multiple such 2-year periods during 2001–2009, the first period was chosen as the ‘identification period’.

The managed depressed study sample was comprised of depressed patients meeting the above stated criteria who did not receive ECT, had no more than one psychiatric hospitalization, and had 8–11 medication management visits (one every 2–3 months) during a 2-year identification period. (See Table 1 for summary of study populations.)

Table 1.  Abbreviated definitions of populations and samples in the study.

VNS population (n = 690)

All Medicare FFS beneficiaries who received VNS for depression and who did not have epilepsy.

Treatment resistant depression (TRD) sample (n = 4639)

5% sample of all Medicare FFS beneficiaries who in the 2-year identification period had major depression and averaged two or more psychiatric hospitalizations per year, averaged at least one medication management visit each 6 weeks, or who ever received ECT.

Managed depression (MDep) sample (n = 7524)

5% sample of all Medicare FFS beneficiaries who in the 2-year identification period had major depression, who did receive ECT, had no more than one psychiatric hospitalization, and averaged one medication management visit every 2–3 months.

2001–2009 Medicare FFS sample (n = 1,777,382)

5% sample of all Medicare FFS beneficiaries, 2001–2009.

2009 Medicare FFS population (n = 36,751,501)

All Medicare FFS beneficiaries in 2009.

Patients in both the TRD and managed depressed study samples were followed for various lengths of time in the 9-year study period (2001–2009). About one-third of TRD patients and one-quarter of managed depressed patients were observed for 8–9 years. Both samples were analyzed based on sub-samples observed for 2-year increments of time following the 2-year identification period.

Analysis of healthcare utilization and utilization-related outcomes

Utilization rates for all study populations were calculated for hospitalization and for ER use. Patients were categorized on the basis of specific healthcare utilization referred to in this study as negative events (i.e., psychiatric hospitalization, emergency room (ER) use, hospitalization, ECT, or diagnoses for poisoning, risk for suicide or self-injury). These negative events represent tangible clinical outcomes, serving as a proxy for standard clinical efficacy measures, such as a Hamilton Depression Ratings Scale (HAMD), that are not available in administrative claims databases. Such services were described as negative events only to reflect the expectation that a patient with a positive response to VNS therapy post-implantation and a managed depressed patient would not experience such events and occurrence of such events would be evidence of ongoing experience of serious symptoms of depression. Note that coding for suicidal ideation and self-injury represent a provider concern for these behaviors and therefore may undercount actual experience. Patient healthcare utilization was categorized into four groups: no negative events, negative events, marginally negative events, and uncertain outcomes. The ‘no negative events’ category included patients who had no emergency room use, no psychiatric hospitalizations, no hospitalization for poisoning, no ECT, no diagnoses for poisoning, self-injury, self-harm, or suicidal ideation. The ‘negative events’ category included patients who had any amount of ECT post-implantation, two or more psychiatric hospitalizations (could include psychiatric as well as hospitalization for a poisoning Diagnosis Related Group (DRG)) or other self-harm/suicidal ideation diagnosis, or two or more diagnoses on claims of poisoning, suicidal ideation, self-harm, or self-injury. The ‘marginally negative events’ category included patients who had a single psychiatric hospitalization but no ECT or ER visits, and none of the other negative outcomes. The ‘uncertain events’ category included patients who had no other negative events other than ER visits, as such visits cannot be directly attributed to depression. Marginal and uncertain outcomes are combined in the tables.

Note that very few patients had revisions, replacements, or removals of the VNS system during the study period; therefore, these events were removed from the analysis of negative events for VNS patients. In addition, patients who had died were removed from the analysis of negative events because deaths cannot be attributed in claims data to any particular diagnosis or cause.

Data use requirements by CMS prohibit disclosure of cell sizes between ‘1’ and ‘10’, or any data from which such cell size could be deduced. To comply with these requirements, some data presented in the tables are blinded using an asterisk or expressed as a range of values.

Analysis of per patient-year cost

Medicare costs were extracted from claims for all patients in the sample for Medicare covered services, except prescription pharmaceuticals. Costs for physician services and durable medical equipment were not available for the VNS population. A per patient-year cost of care was developed based on all Medicare and required beneficiary co-pay/co-insurance payments for the first and second year post-implantation for VNS patients, and for the 2-year identification period and cohorts of TRD and managed depression patients for 2-year increments following the identification period. Per patient-year costs (including physician and Durable Medical Equipment [DME] costs) for each group were compared to the general Medicare Fee-for-service (FFS) population from 2009 (2009 Medicare FFS). Per patient-year costs for the VNS population were also compared to TRD and managed depression per patient-year costs excluding the physician service and DME costs to enable direct comparisons.

All costs reported for the VNS population, TRD, and managed depression samples are in 2010 dollars. We standardize costs to 2010 dollars by re-pricing services from the year in which each service is delivered, using 2010 standard Medicare prices where they are available (e.g., drugs), and, for some services (e.g., hospitalization, outpatient hospital, skilled nursing facility, and home health), by applying the rate update percentages published in annual final regulations. Physician rates are re-priced by using the inflation adjustments passed by Congress or embedded in applicable regulations that change the conversion factor that is multipled by the relative value units assigned to each code. An average inflation factor is applied to codes and services that changed from year-to-year. This approach to standardizing payments preserves the regional geographic variation that exists in Medicare payments for most services.

Statistical analysis

Descriptive analyses of demographic characteristics (age, gender, race, and reason for Medicare eligibility) were conducted separately for the VNS, TRD, MDep, and 2001–2009 Medicare FFS patient populations. Patients with various co-morbid medical conditions and co-occurring psychiatric conditions were studied using groupings of ICD-9 codes developed empirically based on codes observed on patient claims; and the proportion of patients with the various conditions was summarized using percentages. Mortality was calculated based on deaths per 1000-patient-years.

Where sample estimates are presented (e.g., referring to the TRD sample, managed depression sample, or the 2001–2009 Medicare FFS sample), 95% confidence bounds have been presented in the associated table, or if the statistic is not from a table, in the text. Where a sample is involved in a comparison statement, we have presented the p-value in the text. All sampling estimates were calculated by bootstrap on 10,000 repetitions.

Results

Demographic characteristics

A total of 690 fee-for-service Medicare patients met the criteria for the VNS study population (Table 2). The TRD sample included 4639 patients, and the managed depression sample included 7524 patients. VNS patients were 73% female vs 67% for TRD and 69% for managed depression patients. On average VNS patients were 51.9 years of age compared to 56.6 years for TRD and 58.7 years for managed depression patients. VNS patients were 97% white vs 88% of TRD patients and 87% of managed depression patients. Among VNS patients, 81% were eligible for Medicare due to disability compared with 62% for TRD and 50% for managed depression patients. The general Medicare population in 2009 was 55% female, had a mean age of 70.1 years, was 83% white, and included 17% eligible for Medicare due to disability.

Table 2.  Demographic characteristics of VNS population, TRD sample, and managed depression sample.

	VNS population			TRD sample			MDep sample			2009 Medicare FFS population
	All	Alive	Dead	All	Alive	Dead	All	Alive	Dead
Number of patients	690	653	37	4639	3693	946	7524	6074	1450	36,751,501
Sex [CI]
% male	27%	26%	43%	33% [31–34%]	32% [31–34%]	34% [31–37%]	31% [30–32%]	30% [29–32%]	36% [33–38%]	45%
% female	73%	74%	57%	67% [66–69%]	68% [66–69%]	66% [63–69%]	69% [68–70%]	70% [68–71%]	64% [62–67%]	55%
Race [CI]
% white	97%	97%	100%	88% [88–89%]	88% [87–89%]	92% [90–93%]	87% [87–88%]	86% [86–87%]	92% [90–93%]	83%
% all other	3%	3%	0%	12% [11–12%]	12% [11–13%]	8%  [7–10%]	13% [12–13%]	14% [13–14%]	8%  [7–10%]	17%
Mean age (in years) [CI]	51.9	51.5	57.5	56.6   [56–57]	53.8   [53–54]	67.3   [66–68]	58.7   [58–59]	56.1   [56–56]	69.7   [69–70]	70.1
% < 40 years	15%	*	*	17% [16–19%]	20% [19–22%]	6%   [4–7%]	12% [11–13%]	14% [13–15%]	4%   [3–5%]	3%
% 40–64 years	67%	*	*	46% [45–48%]	51% [50–53%]	28% [25–31%]	50% [49–51%]	55% [54–57%]	26% [23–28%]	20%
% 65+ years	18%	16%	46%	36% [35–38%]	29% [27–30%]	66% [63–69%]	38% [37–39%]	31% [29–31%]	71% [68–73%]	77%
Medicare status [CI]
% aged w/o ESRD	19%	17%	46%	37% [36–39%]	70% [67–72%]	44% [43–45%]	49% [48–50%]	43% [41–44%]	75% [73–77%]	82%
% disabled w/o ESRD	81%	83%	54%	62% [61–64%]	30% [27–33%]	56% [54–57%]	50% [49–52%]	57% [56–58%]	23% [21–26%]	17%

ESRD, end stage renal disease; FFS, fee-for-service; MDep, managed depressed; TRD, treatment-resistant depression; VNS, vagus nerve stimulation.

Note: Data use requirements by CMS prohibit disclosure of cell sizes between ‘1’ and ‘10’, or any data from which such cell size could be deduced. To comply with these requirements, some data presented here are blinded using an asterisk or expressed as a range of values. Race in U.S. Medicare claims data classifies each beneficiary as ‘White,’ ‘Black,’ ‘Hispanic,’ ‘North American Native,’ ‘Asian,’ ‘Other,’ or ‘Unknown.’

Mortality

The annual mortality rate for the VNS population post-implantation was 19.9 deaths per 1000 patient-years (age < 40: 25.4; 40–64: 10.3; ≥ 65: 52.0), compared to 46.2 in the first 2 years after the identification period for the TRD sample (age < 40: 12.5; 40–64: 25.8; ≥ 65: 81.4), and 46.8 for the same period for the managed depression sample (age < 40: 15.1; 40–64: 24.6; ≥ 65: 77.0) (Table 3). The mortality rate for the general Medicare fee-for-service population between 2001–2009 was 48.2 deaths per 1000 patient-years (age < 40: 9.9; 40–64: 24.0; ≥ 65: 55.6). In the VNS population, 37 patients (5%) died during the 2-year post-implantation study period.

Table 3.  Comparison of deaths per 1000 patient-years, hospitalizations per patient-year, and payments per patient-year: VNS population, TRD sample, managed depression sample, and 2001–2009 Medicare FFS sample.

	VNS population, years 1 and 2 post-implantation	TRD sample, years 1 and 2 post-identification		MDep sample, years 1 and 2 post-identification		2001–2009 Medicare FFS sample, calendar years 2001–2009
Deaths per 1000 patient-years [CI]
Overall	19.9	46.2	[41.9–50.6]	46.8	[43.4–50.4]	48.2	[48.1–48.3]
Under 40 years	25.4	12.5	[7.0–19.1]	15.1	[9.1–21.8]	9.9	[9.7–10.2]
40–64 years	10.3	25.8	[20.9–30.9]	24.6	[20.8–28.5]	24.0	[23.8–24.2]
65 years and older	52.0	81.4	[72.8–90.2]	77.0	[70.5–83.5]	55.6	[55.5–55.8]
Hospitalizations per patient-years [CI]
Overall	0.71	0.94	[0.89–0.99]	0.50	[0.47–0.52]	0.35	[0.35–0.36]
Under 40 years	0.93	1.50	[1.30–1.71]	0.39	[0.33–0.46]	0.36	[0.35–0.36]
40–64 years	0.68	0.89	[0.82–0.96]	0.48	[0.45–0.51]	0.32	[0.32–0.32]
65 years and older	0.63	0.79	[0.73–0.84]	0.54	[0.50–0.57]	0.36	[0.36–0.36]
Payments per patient-years [CI]
Excluding physician and DME costs	$8749	$13,618	[$12,937–$14,342]	$8,960	[$8,555–$9,381]	$6670	[$6641–$6702]
Including physician and DME costs	N/A	$19,492	[$18,550–$20,515]	$13,252	[$12,752–$13,779]	$9757	[$9722–$9794]

DME, durable medical equipment; FFS, fee-for-service; MDep, managed depressed; N/A, not applicable; TRD, treatment-resistant depression; VNS, Vagus nerve stimulation.

Psychiatric co-occurring and medical comorbid diagnoses

Inpatient and outpatient hospital claims were checked for ICD-9 diagnosis codes in any position (primary or secondary) on the claim during the index period and the first and second post-implantation years. If the patient had one or more diagnoses on any claims it was reported as either a psychiatric co-occurring condition or a medical co-morbid condition (Table 4). About one fifth of the VNS population studied had diagnoses of bipolar disorder, and a small proportion of patients had diagnoses of personality disorder, drug abuse related conditions, post-traumatic stress disorder, or schizoaffective disorder. One in four patients had diagnoses of anxiety, obsessive-compulsive disorder (OCD), or phobias during the index period. Hypertension is reported for more than half the patients during the index period and decreases in prevalence post-implantation. Frequency of reported joint, limb, back, and chronic pain appears to increase from the index period into the post-implantation period. The VNS population analysis is based only on hospital inpatient and outpatient claims, and likely undercounts diagnoses compared to the TRD and managed depressed samples that include physician services claims.

Table 4.  Prevalence of co-occurring psychiatric conditions and medical comorbidities.

	VNS population (n = 690)		TRD sample: Full study period (n = 4639)	MDep sample: Full study period (n = 7524)
	First year post–VNS	Full follow-up period^a
Co–occurring psychiatric conditions (% with > 2 co-occurring diagnoses (listed below): [CI]
Anxiety, OCT, phobias	11%	16%	59% [58–61%]	43% [42–45%]
Bipolar disorder	20%	24%	30% [29–32%]	10%  [9–10%]
Sleep disorder	3%	7%	16% [15–17%]	13% [12–14%]
Personality disorder	5%	7%	16% [15–17%]	4%  [4–5%]
Drug abuse related conditions	3%	4%	17% [16–18%]	6%  [5–7%]
Post-traumatic stress disorder	*	3%	10%  [9–11%]	4%  [4–5%]
Schizoaffective disorder	*	3%	12% [11–13%]	2%  [2–2%]
Delusions and other psychoses	*	*	17% [16–19%]	6%  [5–6%]
Schizophrenia	*	*	11% [10–12%]	3%  [3–3%]
Alcohol related conditions	*	*	8%  [7–9%]	3%  [2–3%]
Co-occurring psychiatric conditions (% with > 1 co-occurring diagnosis): [CI]
Poisoning by drugs	6%	8%	9%  [8–10%]	2%  [2–2%]
Suicide attempt or self-inflicted injury	10%	15%	7%  [6–8%]	1%  [1–1%]
Personal history of non-compliance with medical care	*	3%	7%  [6–7%]	2%  [1–2%]
Eating disorders	*	3%	6%  [6–7%]	3%  [2–3%]
Suicide ideation	8%	14%	6%  [6–7%]	1%  [1–1%]
Medical comorbidities (% with > 1 co-occurring diagnosis): [CI]
Join/limb/back pain	33%	39%	73% [71–74%]	70% [69–71%]
Hypertension	44%	41%	68% [67–69%]	68% [67–69%]
Disorders of lipoid metabolism	39%	39%	62% [61–64%]	64% [63–65%]
Heart disease	18%	22%	56% [54–57%]	50% [49–51%]
Malaise and fatigue	13%	22%	52% [51–54%]	44% [43–46%]
Long term current use of other high risk medications	23%	29%	42% [41–44%]	40% [39–41%]
Anemias	8%	13%	39% [38–41%]	35% [34–36%]
Hypothyroidism	24%	26%	35% [34–36%]	29% [28–30%]
COPD	11%	16%	34% [33–35%]	30% [29–31%]
Diabetes	17%	19%	32% [31–34%]	31% [30–32%]
Asthma	11%	14%	21% [20–23%]	18% [20–23%]
Chronic pain	11%	18%	20% [19–21%]	18% [17–19%]
Overweight and obesity	9%	14%	17% [16–18%]	14% [13–15%]
Migraine	8%	10%	12% [11–13%]	9%  [8–10%]

MDep, managed depressed; TRD, treatment-resistant depression; VNS, vagus nerve stimulation.

Note: Data use requirements by CMS prohibit disclosure of cell sizes between ‘1’ and ‘10’, or any data from which such cell size could be extrapolated. To comply with these requirements, some data presented here are blinded using an asterisk or expressed as a range of values.

^aVNS patients were followed for up to 10 calendar quarters following VNS implantation, and left the study early if they died. Note that prevalence statistics will generally be higher when identified over a longer period (10 quarters) than a shorter one (1 year).

Looking at the TRD and managed depressed samples: 59% of TRD patients had anxiety, OCD, or phobias compared to 43% of managed depressed patients; 30% of TRD patients had bipolar disorders compared to 10% of managed depressed patients; 17% of TRD patients had delusions, other psychoses, or drug abuse and related conditions compared to 6% of managed depressed patients. In all other co-occurring psychiatric diagnoses, the TRD sample had a higher frequency of these diagnoses compared with managed depressed patients (for each, p < 0.0001). A similar pattern can be observed for medical comorbid diagnoses, though the differences are not as large.

Negative events post-implantation for the VNS population and post-identification period for TRD and managed depressed samples

In the VNS population, patient deaths and patients with revisions/replacements and removals of the VNS system were removed from the study of negative events because these events cannot be attributed in claims data to any particular diagnosis or cause. Of the 629 VNS patients classified, 150 (24%) experienced no negative events through the 2-year post-implantation period, suggesting a positive health outcome for those patients; 282 (45%) experienced ER visits for unknown reasons or had a single hospitalization during the post-implantation period, which cannot be determined to be related to depression or the VNS implant; and 197 (31%) experienced negative events post-implantation suggesting persistence of major depression symptoms despite the VNS intervention (Table 5).

Table 5.  Demographic characteristics and Medicare enrollment status of VNS patient population categorized by negative events.

Demographic characteristics/Medicare enrollment status	VNS population, 2 years post-VNS implant^a
	All patients, full follow-up period (n = 690)	No negative events (n = 150, 24%)	Marginally negative or uncertain events (n = 282, 45%)	Negative events (n = 197, 31%)
Sex
% Male	27%	26%	23%	27%
% Female	73%	74%	77%	73%
Age groups^b
% under 40 years	15%	9%	18%	16%
% 40–64 years	67%	70%	66%	63%
% 65 years and older	18%	21%	16%	21%
Medicare status
% Aged w/wo ESRD^c	19%	21%	16%	21%
% Disabled w/wo ESRD	81%	79%	84%	79%

ESRD, end stage renal disease; VNS, vagus nerve stimulation; w/wo, with or without.

^aNote that 61 patients were excluded from the negative events analysis (not the overall comparison numbers in the first column) because they were not present for the full 2 years used to identify negative events.

^bIn the Standard Analytic files, age is that in the previous year, and thus Medicare enrollment appears to begin at 64 years of age. To correct for this, we have grouped patients by ages of 40–63 years and 64+ years, but have presented it as 40–64 years and 65+ years.

^cNote ESRD excluded from VNS.

In the TRD sample, the appropriate comparison period is the 2-year increments after the identification period, since the identification period requires the presence of hospitalizations and/or ECT. For negative events to be measurable, patients needed to be alive and have data for the full 2 years. In the first and second year post-identification period, 1429 of 3797 patients in the TRD sample (38%) had no negative events and 767 (20%) had negative events (Table 6). In the managed depression sample, in the same 2-year post-identification period, 2979 of 6005 patients (50%) had no negative events, and only 219 patients (4%) experienced negative events. The proportion of both samples with no negative events increases over time. The VNS patient population cannot be compared based on duration of depression to the TRD and managed depression samples directly.

Table 6.  Comparison of prevalence of co-occurring psychiatric conditions in VNS patients by outcome category.

Patients with one or more caims with the specified psychiatric conditions^a	No negative events	Marginally negative events	Negative events
Dementia	*	*	*
Alcohol-related conditions	*	*	*
Drug abuse-related conditions	0–19%	0–19%	20–39%
Schizophrenia	*	*	*
Schizoaffective disorder	*	*	*
Bipolar disorder	0–19%	20–39%	60–79%
Delusions and other psychoses	*	*	*
Anxiety, obsessive compulsive disorder, phobias	20–39%	40–59%	60–79%
Personality disorder	0–19%	0–19%	40–59%
Post-traumatic stress disorder	0–19%	0–19%	0–19%
Sleep disorder	0–19%	20–39%	20–39%

Note: Due to a data use agreement with CMS, positive cell counts less than 11 cannot be presented in the table and are represented by an *. Information that can be used to deduce cell counts less than 11 cannot be provided, resulting in reporting ranges.

^aConditions during the index quarter or any of the subsequent eight quarters were identified to maximize the probability of observing a psychiatric condition. Observations based on hospital inpatient and outpatient claims—physician office claims were not available. Sixty-one patients were excluded from this analysis due to death or removal of device because they did not have 2 full years of data post-implantation.

Once VNS patients are sorted based on the categorization of negative events, patient characteristics show differences between those with and without negative events post-implantation (Table 5). Younger patients and those eligible for Medicare because of disability are more likely to have negative events than older patients. Patients with bipolar diagnoses are much more likely to experience negative events than those without those diagnoses (Table 6). Diagnoses of anxiety, OCD, and phobias are also more common for patients with negative events, compared to those with no negative events. Hypertension, disorders of lipoid metabolisms, joint/limb/back pain, and chronic pain are more common for patients with negative events compared to those with no negative events (Table 7).

Table 7.  Comparison of comorbid condition prevalence in VNS patients by outcome category.

Patients with one or more claims with the specified psychiatric conditions^a	No negative events	Marginally negative events	Negative events
Malaise and fatigue	20–29%	20–29%	20–29%
Long-term current use of other high risk medications	20–29%	30–39%	40–49%
Diabetes	20–29%	20–29%	20–29%
Hypertension	30–39%	50–59%	60–59%
Heart disease	20–29%	30–39%	30–39%
Hypothyroidism	20–29%	30–39%	30–39%
Disorders of lipoid metabolism	30–39%	40–49%	50–59%
Asthma	10–19%	20–29%	20–29%
Anemias	10–19%	20–29%	20–29%
Chronic pain	0–9%	20–29%	30–39%
Chronic obstructive pulmonary disease	0–9%	20–29%	20–29%
Joint/limb/back pain	20–29%	50–59%	50–59%

Note: Due to a data use agreement with CMS, positive cell counts less than 11 cannot be presented in the table and some conditions that are blinded have been removed from the table. Information that can be used to deduce cell counts less than 11 cannot be provided, resulting in reporting ranges.

^aConditions during the index quarter or any of the subsequent eight quarters were identified to maximize the probability of observing a psychiatric condition. Observations based on hospital inpatient and outpatient claims—physician office claims were not available. Sixty-one patients were excluded from this analysis due to death or removal of device because they did not have 2 full years of data post-implantation.

While not shown, the VNS patients with negative events, only 10% had no evidence of psychiatric conditions other than depression in the 2-year follow-up period. Of all patients with diagnoses of personality disorder or schizoaffective disorder, more than 70% of each group was in the group with the most negative events. Among the VNS patients with diagnoses of alcohol or drug abuse related conditions, more than 60% were in the group with the most negative events. Patients with no negative events post-implantation have fewer complicating psychiatric co-occurring conditions and medical co-morbid conditions.

Comparing VNS patients in the 2-year post-implantation period to the TRD and managed depressed patient samples in the post-identification 2-year periods, VNS patient mortality is much lower than in all other groups (p < 0.0001), except for patients less than 40 years of age (Table 3). Hospitalization rates are higher (p < 0.0001) than for the managed depression group, but lower (p < 0.0001) than the TRD groups in all categories.

Cost analysis

The per patient-year costs for the VNS patient population cannot be compared directly to the TRD, managed depressed population, or general Medicare population cost because physician and DME costs were not available in 100% SAF claims for VNS patients. To enable cost comparisons between the patient populations, the physician and DME costs were removed from the per patient-year costs for the TRD and managed depressed samples. Using the same negative event categories as reported for VNS patients, the mean per patient-year costs in 2010 dollars are shown in Table 3. All depression groups studied have higher annual costs of care than the general Medicare population in 2009 ($9757). VNS patients’ post implantation costs per patient-year ($8749 without physician or DME costs) are in the range of those for managed depression patients ($8960–$7663 decreasing over 6 years post-identification period), and are much less (p < 0.0001) than costs of care for TRD patients ($13,618–$10,228 decreasing over 6 years post-identification period).

Looking at the TRD and managed depressed patients for whom all but prescription drug data are available, the cost per patient-year for TRD patients with negative events ($38,807) is 4.3-times (95% CI = 3.8–4.9) that of patients without negative events ($9116) in the 2-year post-identification period. For managed depressed patients, the cost per year is 3.5-times (95% CI = 3.1–4.0) higher for those with negative events ($28,948) compared with those without negative events ($8185). For both groups, the costs per year for those with marginal/uncertain events ($18,917 for TRD sample and $16,698 for managed depressed sample) are about twice (95% CI: TRD 1.9–2.3, MDep 1.9–2.2) that of those with no negative events. Table 8 shows the composition of these costs. While not shown, costs decline over time for patients with no negative events in both groups. For TRD patients with negative events, costs decline but only slightly over time, and for managed depressed patients with negative events, costs vary over time. While these comparisons cannot be made for VNS patients, it is reasonable to expect similar differences in the costs of care between those with and without negative events.

Table 8.  Cost distribution of patients in TRD sample and managed depression sample categorized by negative events.

	Years 1 and 2 post-identification
	No negative events		Marginal/uncertain		Negative events
TRD sample
Patients categorized by negative events [CI]	1429 (38%)	[36–39%]	1601 (42%)	[41–44%]	767 (20%)	[19–22%]
Total costs in 2-year period:	$9116	[$8375–9875]	$18,917	[$17,608–20,316]	$38,807	[$35,456–42,957]
Psychiatric inpatient hospitalizations	$0	[$0–0]	$826	[$716–940]	$14,274	[$13,281–15,316]
Medical inpatient hospitalizations	$2747	[$2327–3186]	$5959	[$5331–6611]	$7575	[$6281–9321]
Outpatient emergency room visits	$0	[$0–0]	$270	[$250–294]	$585	[$509–669]
All other outpatient visits	$1196	[$1047–1363]	$3097	[$2599–3803]	$4557	[$3989–5207]
Physician services, psychiatrists	$558	[$515–604]	$618	[$576–663]	$1745	[$1644–1850]
Physician services, primary care MD	$526	[$478–579]	$967	[$793–1203]	$2692	[$1077–5605]
Physician services, neurologists	$71	[$59–83]	$120	[$103–140]	$168	[$121–231]
Physician services, all other specialties	$2315	[$2138–2,509]	$3715	[$3462–3997]	$4040	[$3686–4451]
All other sites of service	$1703	[$1426–2,002]	$3344	[$2996–3738]	$3170	[$2674–3713]
MDep sample
Patients categorized by negative events [CI]	2979 (50%)	[48–51%]	2,807 (47%)	[45–48%]	219 (4%)	[3–4%]
Total costs in 2-year period:	$8185	[$7643–8805]	$16,698	[$15,861–17,591]	$28,948	[$25,749–32,336]
Psychiatric inpatient hospitalizations	$0	[$0–0]	$318	[$268–372]	$9723	[$8566–11,010]
Medical inpatient hospitalizations	$2664	[$2382–2965]	$6313	[$5803–6850]	$6511	[$4994–8190]
Outpatient emergency room visits	$0	[$0–0]	$252	[$237–268]	$484	[$385–597]
All other outpatient visits	$1123	[$1025–1229]	$2511	[$2334–2695]	$3257	[$2593–4008]
Physician services, psychiatrists	$184	[$175–193]	$216	[$207–226]	$1015	[$907–1131]
Physician services, primary care MD	$638	[$471–933]	$757	[$723–793]	$938	[$812–1070]
Physician services, neurologists	$65	[$56–75]	$109	[$98–121]	$162	[$110–229]
Physician services, all other specialties	$2002	[$1887–2127]	$3272	[$3125–3427]	$3366	[$2928–3827]
All other sites of service	$1509	[$1317–1720]	$2948	[$2685–3225]	$3492	[$2507–4612]

MDep, managed depressed; TRD, treatment-resistant depression.

Note: The VNS patient population cannot be directly compared to the TRD and MDep samples based on duration of depression due to the lack of detailed data for VNS patients prior to implantation.

Discussion

This is the first study to examine outcomes in Medicare patients receiving VNS therapy for TRD. These patients were largely eligible for Medicare due to disability and were considerably younger on average than the general Medicare population. Because they enter the Medicare system at a younger age, they will potentially receive healthcare services for a longer period of time. For patients receiving VNS therapy, Medicare expenditures following the VNS therapy system implant—exclusive of physician and DME costs—were much less than for a sample of Medicare beneficiaries identified as receiving care for TRD without VNS therapy.

In this study, the expenditures for the Medicare beneficiaries with adjunctive VNS therapy were actually similar to those for a sample of beneficiaries receiving treatment for depression without the characteristics of treatment resistance. When healthcare utilization is examined to identify volume of negative events associated with depression, 63% of the patients receiving VNS therapy had no negative events or only events of marginal or uncertain character. Not surprisingly, the Medicare patients receiving VNS therapy who did have volume of negative events during follow-up were more likely to have co-occurring psychiatric diagnoses than were the patients who experienced no negative events. This study also provided support for the findings from previous studies of the costs of care for TRD patients without VNS therapy5–7 that suggested average costs of care are generally several thousands of dollars higher than those calculated for the general 2009 Medicare population.

The analysis of Medicare TRD and managed depressed patients in this study confirm these reports and demonstrate more directly the difference in costs of care over a significant period of time. Costs for all groups of depression patients are somewhat higher than those for the general 2009 Medicare population. VNS patient costs ($8749) are comparable to those for managed depressed patients ($8960; 95% CI = $8555–$9381), but are significantly less than those for TRD patients ($13,618; 95% CI = $12,937–$14,342) with the severity for which VNS therapy is indicated as reflected in the TRD literature. These findings suggest that, for patients receiving VNS therapy, even with ambiguous response, costs of care are likely to be lower post-implantation than without such intervention.

The present study adds to the findings of studies referenced earlier in this paper, by providing additional outcome data that focus on healthcare utilization and healthcare costs, and by looking at the actual population treated rather than a clinically defined and controlled sample.

Study findings suggest that, among the Medicare population that received the VNS implant during the period of time during which the therapy was covered by Medicare, a substantial number of patients appeared to benefit from the therapy, in that they experienced no negative events or experienced healthcare utilization that is not clearly reflective of severe consequences of major depression during the first 2 years following implantation. Therefore, appropriately selected Medicare beneficiaries who are being treated for TRD may benefit from VNS therapy. Even if their response to VNS therapy is only partial, such patients may experience decreases in healthcare utilization after a VNS therapy system implant (compared to patients with TRD being managed without VNS therapy). This in turn can lead to lower healthcare expenditures for the Medicare system. In addition, those patients receiving VNS with evidence of other co-occurring psychiatric diagnoses, such as bipolar disorder, personality disorder, schizoaffective disorder, and post-traumatic stress disorder, had high rates of negative events, suggesting that patients with more complex diagnostic profiles will continue to have problems associated with depression and comorbid conditions.

A number of findings in this study cannot be explained without further research. The lower mortality for VNS patients compared to TRD, Mdep, and 2001–2009 Medicare FFS samples is striking, but cannot be explained directly by the data, as cause of death data are not available and other conditions associated with mortality were not examined in the study. Also the VNS population cannot be compared to the TRD and Mdep populations based on duration of illness due to the lack of detailed data for VNS patients prior to implantation. In-depth analysis of co-occurring conditions associated with negative events was not conducted to the extent done for the VNS population in the TRD and Mdep populations.

Limitations of the study

Retrospective claims analysis of the TRD population is performed here without direct evidence of failures of adequate pharmaceutical treatment, the primary clinical criteria for diagnosing TRD. However, the proxy criteria developed here characterize a population very similar to that described in smaller clinical studies in which these clinical criteria could be documented. The analyses in the population receiving VNS therapy are partially constrained by a lack of physician services and DME data and 100% SAF files for a pre-implantation period: 5% sample data were available for the pre- and post-implantation period but the sample was too small to provide a robust analysis. Medicare SAFs with 100% of claims for physician services and DME are not released for analysis. The limitations in the data for the population receiving VNS therapy also limit the ability to make full comparisons with the other populations described in this paper. The lack of complete physician service claims also limits the completeness of the examination of co-occurring and comorbid condition diagnoses for the population receiving VNS therapy.

Conclusion

This study explores the potential of using selected healthcare utilization (e.g., negative events) as health outcomes in population-based studies using administrative claims data. This approach is particularly interesting when healthcare utilization events are expected to be associated with treatment failure and/or inadequate control of depressive illness. It supports study in populations of longer-term patient experience than can be carried out through most clinical studies, and offers proxy data for the testing and clinical evaluations used in clinical studies. Given the short coverage period for the VNS therapy for the approved depression indication, this analytic approach provides a reasonable way to compare the experience of the population covered by Medicare with nationally representative samples of other non-implanted chronically-depressed Medicare beneficiaries. Interesting study results that cannot be explained within the scope of analysis provided in this study can become the basis for additional research in the future. The association of younger age with more frequent negative events could be further investigated to look at the patterns of care for younger patients that may be associated with higher rates of negative events. Having identified a TRD and comparison Mdep population within Medicare administrative claims data, further clinical description (e.g., co-occurring and comorbid conditions and more detail on health care utilization) could provide further insight into the factors that drive the high costs of care associated with these chronic conditions. This study is illustrative of the multi-layered analyses that can be conducted in administrative claims data to explore questions about conditions and services that drive high cost care in patient populations that suffer from severe and persistent mental illness.

Transparency

Declaration of funding

Cyberonics, Inc. sponsored the study through a contract with The Moran Company. Independence in the methods, analyses, and interpretation of results was a condition of the contract with The Moran Company. The authors are solely responsible for data interpretation and conclusions.

Declaration of financial relationships

RF (partner), JM, and DS (employees of the Moran Company) had no financial relationships with Cyberonics or any other entity related to this study except as the result of the Cyberonics’ contract with The Moran Company to perform the study and prepare the manuscript for submission. RF, JM, and DS have no additional information to declare. In the past 2 years, DD has received grant support from Cyberonics and Neuronetics, consulting fees from Eli Lilly, Wyeth, Pfizer, and Jazz Pharmaceuticals, and speaking fees from Pfizer, Wyeth, Neuronetics, AstraZeneca, and Bristol-Myers Squibb.

Acknowledgments

Editorial assistance was provided by Karishma Manzur, PhD and was compensated by Cyberonics, Inc.