Healthcare Utilization and Costs for Patients With Parkinson's Disease After Deep Brain Stimulation

ABSTRACT

Objective

To compare the complications, healthcare utilization and costs following DBS or medical management for patients with Parkinson's disease (PD).

Methods

We examined healthcare utilization and costs for up to 5 years between veterans with DBS and those with medical management for PD. Veterans who received DBS between 2007 and 2013 were matched with veterans who received medical management using propensity score approaches. Healthcare utilization and costs were obtained from national VA and Medicare data sources and compared using procedures to adjust for potential differences in length of follow‐up.

Results

We identified 611 veterans who had received DBS and a matched group of 611 veterans who did not undergo DBS. Among DBS patients, 59% had the electrodes and generator implanted during separate admissions. After 5 years of follow‐up, average total healthcare costs, including DBS procedures and complications, were $77,131 (95% confidence interval: $66,095–$88,168; P < 0.001) higher per person for patients who received DBS ($162,489) than patients who received medical management ($85,358). In contrast, excluding the costs of the DBS procedures and complications, average total costs were not significantly different between patients who received DBS and patients who received medical management after 5 years of follow‐up.

Conclusions

Healthcare costs over 5 years were higher for veterans who received DBS. These higher healthcare costs may reflect the costs of DBS procedures and any follow‐up required plus greater surveillance by healthcare professionals following DBS as well as unobserved differences in the patients who received medical management or DBS.

Parkinson's disease (PD) is a common progressive neurodegenerative disease.1, 2 In the United States, prevalence of PD is estimated to be 0.3% with similar prevalence rates in Europe.3, 4 Annual healthcare cost averaged $10,000 per patient.3 For PD patients with long‐term complications from dopaminergic drugs, including motor fluctuations, severe dyskinesias, medication intolerance, or refractory motor symptoms, DBS is an accepted surgical treatment for symptom management.5, 6 DBS of the STN or internal globus pallidum (GPi) have been found to be efficacious in relieving motor fluctuations and dyskinesias associated with PD, with sustained benefit.7, 8, 9, 10

Results from a previous cost‐effectiveness analysis comparing DBS with medical management found that healthcare costs are substantially higher 1 year following DBS.11 Projections beyond the trial estimated that cost differences between DBS and medical management would narrow 1 to 5 years after the DBS procedure, but widen after 5 years because of battery replacements; however, projections were sensitive to assumptions about future costs.11 Cost‐effectiveness analyses comparing DBS and medical management over a 5‐ to 15‐year time using Markov models have generally found DBS a cost‐effective alternative to medical management.12, 13 However, these long‐term cost estimates were based on extrapolations from shorter‐term follow‐up periods with assumptions about longer‐term costs of complications and follow‐up procedures (e.g., battery replacement or removal of DBS system because of infections). Data with longer‐term follow‐up have been lacking.

We previously reported that veterans with PD receiving DBS had longer average survival than a matched group of medically managed veterans.14 However, healthcare costs were not reported. In this study, we examined actual healthcare utilization and costs up to 5 years following DBS for those Veterans.

Materials and Methods

Study Design and Sample

To examine costs of DBS versus medical management, we utilized a retrospective, longitudinal, observational design where surgery (DBS) cases were matched with medically managed patients using a propensity score model. We assessed healthcare utilization and costs beginning with each patient's “index date” (defined below). We identified any patient with an International Classification of Diseases, Ninth Revision (ICD‐9) code for PD (332 or 332.0) between the federal government's fiscal year (FY) 2007 (i.e., October 1, 2006 through September 30, 2007) and FY 2013 using national VA and the Centers for Medicare & Medicaid Services (CMS) databases. To reduce the likelihood that persons did not have PD, we required at least two PD diagnoses after October 1, 2006. Patients were followed a maximum of 5 years or until October 1, 2014 (Fig. 1). We excluded patients with any DBS procedures before October 1, 2006 using ICD‐9 procedure codes and Current Procedures Terminology (CPT) codes for DBS. To exclude patients with early‐onset PD, we removed those with PD diagnoses before age 41 (n = 238). We limited the sample to patients with PD for at least 5 years. Most patients with PD receiving levodopa begin experiencing a waning of the medication effect after 5 years,15, 16 at which point DBS might be considered. To assess whether veterans had PD for at least 5 years, we determined whether they had PD medications and/or had healthcare utilization (e.g., a clinic visit, hospital stay) with a diagnosis of PD in at least 4 of the 5 previous years in VA and Medicare databases.

Figure 1.

Cohort identification. *List of PD medications used available in Supporting Information Appendix S1.

For the DBS cohort, the first DBS procedure date was the study index date. This procedure was identified with ICD‐9 or CPT codes for DBS between October 1, 2006 and September 30, 2013. We examined DBS procedure codes to assess whether patients received both the electrode implants and the pulse generator. We excluded patients without evidence for both DBS procedure components because the absence of both may indicate a subsequent repair or replacement procedure rather than the initial DBS procedure.

For the medically managed cohort, we identified patients without procedure codes for DBS. We identified the date of these patients’ first healthcare encounter (e.g., clinic visit, hospital stay) between October 1, 2006 and September 30, 2013 with an ICD‐9 diagnosis code for PD. We assessed which patients had PD medications and/or healthcare utilization for at least 4 of 5 years before that date. If the patient did not have at least 4 to 5 years of PD‐related utilization or medication use, we identified the first PD encounter in the following year and looked back 5 years from that date. We continued this process by year through September 30, 2013 to find all patients with PD meeting study criteria.

For both groups, we excluded patients in the DBS or comparison group with diagnoses of dementia, stroke, or related cerebral vascular disease, a cardiac pacemaker, cardiac defibrillator, or cochlear implant because these are contraindications for DBS.14 This study was approved by the Edward Hines Jr. VA Hospital Institutional Review Board.

Propensity Score Estimation and Matching

To address potential differences between groups, we matched DBS cases to medical management cases using propensity scores with nearest neighbor matching.17 To ensure that we had approximately equal periods of previous PD utilization for both groups at the time they were matched, we identified all medically managed cases with a PD‐related visit within 2 weeks (±) of a DBS patient's first DBS procedure. Potential matches were those medically managed cases whose duration of previous PD was the same as for corresponding DBS patient. Propensity scores were created using logistic regression to model the odds of receiving DBS based on patient age, race, Charlson comorbidity index score, diagnosis of depression in the previous year, distance to the nearest VA, distance to nearest VA Parkinson's Disease Research Education Clinical Care center (PADRECC), distance to the nearest Medicare inpatient facility, and census region.18 PADRECCs are designated VA centers for comprehensive movement disorders care, including DBS surgery. After developing propensity scores, within each strata (described above), we matched the medically managed case with the closest propensity score match to the DBS case resulting in a 1:1 match without replacement. Covariate balance was checked before and after matching using standardized differences (<0.10 considered a good match).19 The index date of each DBS patient was assigned to each matched medically managed patient.

Data Sources

Data sources included VA national Medical SAS files for inpatient and outpatient utilization, VA Managerial Cost Accounting (MCA) National Data Extracts in the VA Corporate Data Warehouse (CDW) data for VA prescription pharmacy data, and VA healthcare costs. We utilized Fee Basis (contract care) files for non‐VA care paid by the VA. Other non‐VA healthcare utilization and costs were identified using Medicare databases. Dates of death were obtained from the VA Vital Status File (VSF), which has a sensitivity of 98.3% compared with the National Death Index.20

Assessment of Utilization and Costs

All healthcare costs were included and adjusted to 2016 U.S. dollars using the Consumer Price Index. We estimated direct healthcare costs from the healthcare payers’ perspective based on VA and Medicare expenditures (described below).21 Although costs estimates were derived from different procedures for the VA and Medicare systems, they both reflect healthcare expenditures resulting from these two federal programs.

If the electrodes and generator were implanted during two hospital admissions more than 1 day apart, we classified the initial procedure as occurring in separate admissions (i.e., staged); otherwise, we included it as occurring during the same admission. Follow‐up procedures included additional surgeries to remove hardware (e.g., because of infection, battery depletion) and/or replace hardware (e.g., wire, electrode, battery). We classified any healthcare events other than the initial and follow‐up DBS procedures and complications related to those procedures as PD related if there was a PD diagnosis code or medication for PD (see Supporting Information Appendix S1 for codes and list of medications). We classified outpatient care as primary care, specialty care, mental health care, and other outpatient using provider specialty and CPT procedure codes available in both VA and Medicare systems.22

VA Expenditures

Total VA expenditures consisted of costs of VA inpatient and outpatient care and care at non‐VA facilities paid by VA. VA costs were obtained from the VA MCA National Data Extracts.23 MCA combines information from the VA's accounting and payroll system with workload information to estimate costs. PD‐related medications costs from VA pharmacies were based on the VA's acquisition and dispensing costs.24 VA payments for non‐VA care were from VA Fee Basis Files.25

Medicare Expenditures

For Medicare‐enrolled patients, cost estimates were based on Medicare expenditures. We summed the payments made to providers by beneficiaries, the Medicare program, and the primary payer (if other than Medicare).21

Statistical Analysis

Healthcare utilization and costs were compared between DBS and medically managed groups up to 5 years after the index date. We compared the numbers of outpatient visits, inpatient days, and 30‐day supplies of medications (e.g., one 90‐day supply was converted to three 30‐day supplies) between these two groups using negative binomial count models, which allow for overdispersion in the data where the variance may be greater than the mean. Because patients were entering our study between October 1, 2006 and September 30, 2013 and were followed through September 30, 2014, they may have different follow‐up lengths. To account for the differing follow‐up lengths, we included offsets in the models with the length of follow‐up time.

We estimated the difference in total cumulative costs over 5 years between groups. As with the utilization, some patients died within 5 years and some had less than 5 years of follow‐up as of October 1, 2014 (i.e., they were right censored). To accommodate death and censoring in the healthcare costs estimates, we used an inverse probability weighting technique where costs were weighted by the inverse of the Kaplan–Meier estimator of survival, such that the costs for patients with shorter follow‐up time were weighted more heavily to account for their incomplete follow‐up time.26, 27

Results

After applying exclusions, there were 620 veterans with DBS and 38,644 veterans with PD but without DBS (Fig. 1). After propensity matching, 611 veterans each were in the DBS and medically managed groups. DBS patients average 46.8 (1.0 standard deviation [SD]) months and medically managed patients average 43.8 (1.0 SD) months of follow‐up (P = 0.03). Moreover, 31.8% of DBS patients and 28.2% of medially managed patients had 5 full years of follow‐up (P = 0.17). Table 1 shows that the two groups were similar in characteristics used for propensity matching. Ages at the time of DBS and matched comparison group index date were 69.2 (7.6 SD) and 69.2 (8.7 SD) years, respectively. Approximately 71% had their DBS at a Medicare‐paid non‐VA facility.

Table 1.

Patient characteristics after propensity score matching

	DBS N = 611	Medical Management N = 611	d (%)	P Value
Age, mean (SD)	65.57 (7.64)	65.54 (8.90)	0.001	0.95
Race, N (%)
White	589 (96.40)	586 (95.91)	0.026	0.41
Black	11 (1.80)	8 (1.31)	0.040
Other	11 (1.80)	17 (2.78)	0.066
Charlson, Mean (SD)	0.46 (0.97)	0.51 (0.94)	0.019	0.34
Distance to nearest PADRECC, mean (SD)	416.57 (292.11)	451.96 (343.10)	0.039	0.05
Distance to nearest VA, mean (SD)	27.44 (30.38)	31.32 (64.43)	0.027	0.18
Distance to nearest Medicare facility, mean (SD)	9.62 (7.04)	9.62 (7.70)	0.000	0.99
Division, N (%)
1	18 (2.95)	14 (2.29)	0.041	0.71
2	60 (9.82)	60 (9.82)	0.000
3	91 (14.89)	86 (14.08)	0.023
4	65 (10.64)	74 (12.11)	0.046
5	143 (23.40)	131 (21.44)	0.047
6	27 (4.42)	21 (3.44)	0.051
7	60 (9.82)	59 (9.66)	0.006
8	63 (10.31)	61 (9.98)	0.011
9	82 (13.42)	100 (16.37)	0.083
10	2 (0.33)	5 (0.82)	0.065
Depression, N (%)	189 (30.93)	183 (29.95)	0.021	0.78

d = standardized differences, and if d < 0.10, then it is considered a good match.

Healthcare Utilization

The electrode and generator were implanted during the same admission for 41% of DBS patients (Table 2). Over 52% of DBS patients had follow‐up DBS procedures, and over 45% had complications after follow‐up procedures. Table 3 shows the complications (see Supporting Information Appendix S2 for definitions).

Table 2.

Healthcare utilization

	DBS N = 611 Mean (SE)	Medical Management N = 611 Mean (SE)	Difference Mean (SE)	95% CI	P Value
Initial procedure
Same admission, N (%)	251 (41.1)	NA	NA	NA	NA
Separate admissions (staged), N (%)	360 (58.9)	NA	NA	NA	NA
Complications during/following initial procedure
Same admission, N (%)	144 (23.6)	NA	NA	NA	NA
Separate admissions, N (%)	250 (40.9)	NA	NA	NA	NA
Follow‐up procedures, N (%)	320 (52.4)	NA	NA	NA	NA
Complications during/following follow‐up procedures, N (%)
During	49 (8.0)	NA	NA	NA	NA
Following	241 (39.4)	NA	NA	NA	NA
Other PD‐related utilization
VA outpatient, visits
Primary care	4.36 (0.14)	4.20 (0.15)	0.16 (0.20)	–0.24 to 0.56	0.43
Mental health	3.20 (0.14)	3.60 (0.17)	–0.39 (0.21)	–0.80 to 0.02	0.06
Other specialty	4.02 (0.18)	2.47 (0.10)	1.56 (0.19)	1.18 to 1.93	<0.001
Diagnostic care	1.19 (0.03)	1.11 (0.02)	0.09 (0.04)	0.02 to 0.16	0.02
Medicare outpatient, visits
Primary care	3.02 (0.20)	1.47 (0.14)	1.55 (0.24)	1.08 to 2.02	<0.001
Mental health	1.16 (0.09)	1.01 (0.01)	0.15 (0.09)	–0.02 to 0.31	0.08
Other specialty	18.81 (0.65)	6.30 (0.33)	12.51 (0.69)	11.17 to 13.85	<0.001
Diagnostic care	3.58 (0.13)	2.11 (0.08)	1.47 (0.15)	1.19 to 1.76	<0.001
Fee outpatient, visits	17.76 (3.01)	15.08 (3.07)	2.68 (3.90)	–4.96 to 10.32	0.49
Emergency department, visits
VA	1.19 (0.03)	1.15 (0.02)	0.04 (0.04)	–0.03 to 0.11	0.30
Medicare	2.29 (0.09)	1.64 (0.05)	0.65 (0.10)	0.45 to 0.85	<0.001
Fee	1.01 (0.01)	1.02 (0.01)	–0.01 (0.01)	–0.02 to 0.01	0.62
VA inpatient, days
Acute care	2.54 (0.32)	2.92 (0.35)	–0.38 (0.46)	–1.29 to 0.53	0.41
Nonacute carea	2.24 (0.39)	2.60 (0.77)	–0.36 (0.86)	–2.05 to 1.33	0.68
Medicare inpatient, days
Acute careb	7.05 (0.49)	5.08 (0.41)	1.97 (0.62)	0.77 to 3.18	0.001
Nonacute carec	11.57 (1.37)	6.70 (0.87)	4.87 (1.60)	1.73 to 8.01	0.002
Fee inpatient, days	1.16 (0.05)	1.12 (0.04)	0.04 (0.06)	–0.08 to 0.16	0.52
Pharmacy, 30‐day suppliesd
VA	52.48 (2.23)	53.91 (2.32)	–1.43 (2.99)	–7.29 to 4.43	0.63
Medicare Part D	9.30 (91.01)	8.98 (1.09)	0.32 (1.38)	–2.38 to 3.01	0.82
Other utilization (not PD related)
Outpatient, visits
VA	26.88 (1.38)	36.60 (2.01)	–9.72 (2.26)	–14.15 to –5.29	<0.001
Medicare	52.50 (2.10)	36.20 (2.10)	16.29 (2.77)	10.87 to 21.72	<0.001
Fee	11.47 (1.85)	17.36 (3.83)	–5.89 (3.92)	–13.57 to 1.79	0.13
VA inpatient, days
Acute care	1.61 (0.32)	3.84 (0.85)	–2.24 (0.88)	–3.95 to –0.52	0.01
Nonacute carea	1.33 (0.28)	1.98 (0.47)	–0.65 (0.52)	–1.67 to 0.37	0.21
Medicare inpatient, days	5.57 (0.77)	9.15 (1.10)	–3.59 (1.35)	–6.23 to –0.94	0.01
Fee inpatient, days	1.09 (0.03)	1.36 (0.09)	–0.28 (0.11)	–0.45 to –0.10	0.002
Pharmacy, 30‐day supplies
VA	127.03 (6.57)	174.18 (8.56)	–47.15 (10.27)	–67.28 to –27.03	<0.001
Medicare Part D	28.56 (2.89)	33.56 (3.91)	–4.74 (4.71)	–13.98 to 4.49	0.31

^a Days of care in VA extended care facilities (i.e., VA community living centers and domiciliaries, as well as care in community nursing homes with which the VA contract).

^b Days of care in Medicare‐reimbursed short stay hospitals.

^c Days of care in Medicare‐reimbursed long stay hospitals and skilled nursing facilities.

^d Any medication dispensed with a 30‐day or less supply was counted a one 30‐day supply; 60‐day and 90‐day supplies were counted as two or three 30‐day supplies, respectively.

Table 3.

Definition of complications

	After Initial Procedure N = 611 N (%)	After Follow‐up procedures N = 611 N (%)
DBS device complications
Wound infection at site of device	23 (3.76)	28 (4.58)
Wound dehiscence	14 (2.29)	24 (3.93)
Device failure or mechanical problems—fractured/broken lead exposed wire; wire tethering/bowstringing; lead migration	56 (9.17)	157 (25.70)
Lead removal	25 (4.09)	28 (4.58)
Lead revision	18 (2.95)	22 (3.60)
Generator removal	19 (3.11)	32 (5.24)
Generator revision	12 (1.96)	23 (3.76)
Nervous system complications from surgically implanted device	3 (0.49)	4 (0.65)
Debridement of wound (at site of implant—leads, wires, or generator—can occur with or without removing device)	36 (5.89)	39 (6.38)
Other complications of DBS surgery
Procedure related
Mental status changes	28 (4.58)	19 (3.11)
Intracranial hemorrhage	0 (0)	0 (0)
Intraventricular hemorrhage (intracerebral)	14 (2.29)	2 (0.33)
Subdural hematoma	9 (1.47)	3 (0.49)
Fall leading to injury or fracture	22 (3.60)	27 (4.42)
Cerebrospinal rhinorrhea/pneumocephalus	4 (0.65)	1 (0.16)
Bacterial meningitis	1 (0.16)	0 (0)
Encephalitis/myelitis	1 (0.16)	1 (0.16)
Meningococcal encephalopy	0 (0)	0 (0)
(any) Surgery related
Pulmonary embolism	0 (0)	0 (0)
Deep vein thrombosis	11 (1.80)	3 (0.49)
Cerebrovascular accident	7 (1.15)	0 (0)
Acute myocardial infarction	0 (0)	0 (0)
Septicemia/sepsis	4 (0.65)	2 (0.33)
Shock attributed to anesthesia	0 (0)	0 (0)
Seizures	13 (2.13)	10 (1.64)
Pneumonia	10 (1.64)	4 (0.65)
Urinary tract infection	46 (7.53)	19 (3.11)
PD‐related complications that may have been triggered by DBS/surgery
Visual disturbances	8 (1.31)	8 (1.31)
Dysphagia	39 (6.38)	27 (4.42)
Dysphasia, dysarthria	25 (4.09)	22 (3.60)
Abnormal movements	160 (26.19)	79 (12.93)
Skin sensation disturbance	6 (0.98)	3 (0.49)
Symbolic dysfunction NEC (e.g., perseveration)	5 (0.82)	4 (0.65)
Sleep disturbances	0 (0)	0 (0)
Other (mental health)
New onset of anxiety	0 (0)	0 (0)
New onset of depression	142 (23.24)	99 (16.20)
Hallucinations	11 (1.80)	9 (1.47)
Suicide (attempt/complete)	0 (0)	0 (0)
Delirium	37 (6.06)	25 (4.09)
Psychosis	26 (4.26)	14 (2.29)
OCD	1 (0.16)	0 (0)
Impulse control disorder	1 (0.16)	1 (0.16)
Apathy	0 (0)	0 (0)
Cognitive impairment/dementia	24 (3.93)	31 (5.07)
Neuromalignant (90 days)	0 (0)	0 (0)

Abbreviations: NEC, not elsewhere classifiable; OCD, obsessive‐compulsive disorder.

Compared with medically managed Veterans, DBS patients had more PD‐related outpatient visits for specialty care or diagnostic care at VA or Medicare‐covered facilities. DBS patients had more PD‐related primary care and emergency department visits to Medicare‐covered facilities. Whereas the number of PD‐related inpatient days in Medicare‐covered facilities was higher for DBS patients, the number of inpatient days was similar between the two groups at VA facilities. The number of 30‐day supplies of PD‐related medications from VA‐ or Medicare Part D–covered pharmacies was similar for both groups.

Non‐PD‐related outpatient visits from VA facilities, acute VA inpatient days, Medicare inpatient days, fee inpatient days, and VA pharmacy use were higher for medically managed patients than for DBS patients. However, DBS patients had more Medicare‐covered outpatient visits.

Healthcare Costs

The average cost of the initial DBS procedure was $40,063 per patient with complications following initial procedures costing $4,665. Over 5 years of follow‐up, subsequent procedures cost $22,591 per patient with an additional $3,764 per patient for complications (Table 4). PD‐related outpatient care costs for specialty, diagnostic, and other services from VA facilities were higher for DBS patients than for medically managed patients. Moreover, outpatient costs (except for mental health) from Medicare‐covered facilities were higher for DBS patients. PD‐related inpatient costs were higher for DBS patients receiving inpatient care from Medicare‐covered facilities.

Table 4.

Healthcare costs

	DBS N = 611 $, Mean (SE)	Medical Therapy N = 611 $, Mean (SE)	Difference $	95% CI $	P Value
Initial procedure	40,063 (786)	N/A	N/A	N/A	N/A
Complications following initial procedure	4,665 (661)	N/A	N/A	N/A	N/A
Follow‐up procedures	22,591 (1,132)	N/A	N/A	N/A	N/A
Complications following follow‐up procedures	3,764 (478)	N/A	N/A	N/A	N/A
Other PD‐related utilization
VA outpatient
Primary care	1,424 (74)	1,382 (93)	42 (119)	–191 to 274	0.72
Mental health	1,427 (135)	1,516 (136)	–89 (198)	–478 to 299	0.65
Specialty care	2,772 (239)	945 (101)	1,827 (259)	1,319 to 2,335	<0.001
Diagnostic care	36 (8)	16 (6)	20 (10)	1 to 39	0.04
Other care	2,885 (350)	1,709 (250)	1,176 (430)	333 to 2,019	0.006
Medicare outpatient
Primary care	281 (20)	134 (17)	147 (26)	95 to 199	<0.001
Mental health	9 (5)	1 (0.3)	7 (5)	–2 to 17	<0.14
Specialty care	4,581 (234)	2,073 (240)	2,508 (335)	1,851 to 3,165	<0.001
Diagnostic care	502 (32)	242 (22)	260 (39)	183 to 336	<0.001
Other care	1,158 (195)	493 (76)	665 (209)	255 to 1,076	0.001
Fee outpatient	1,937 (298)	2,036 (595)	–100 (665)	–1,403 to 1,204	0.88
Emergency department
VA	151 (34)	150 (28)	1 (38)	–72 to 75	0.97
Medicare	386 (28)	195 (18)	191 (33)	126 to 257	<0.001
Fee	14 (3)	50 (27)	–35 (28)	–89 to 19	0.20
VA inpatient
Acute carea	5,902 (1,051)	8,156 (1,607)	–2,254 (1,920)	–6,018 to 1,510	0.24
Nonacute carea	1,429 (310)	2,317 (921)	–888 (972)	–2,793 to 1,018	0.36
Medicare inpatient
Acute careb	15,333 (1,119)	9,474 (797)	5,858 (1,374)	3,166 to 8,551	<0.001
Nonacute carec	6,175 (687)	3,815 (544)	2,359 (876)	642 to 4,076	0.007
Fee inpatient	367 (104)	302 (79)	65 (131)	–192 to 322	0.49
Pharmacy
VA	4,146 (269)	4,156 (296)	–10 (399)	–793 to 773	0.98
Medicare Part D	2,144 (317)	1,288 (202)	866 (375)	131 to 1,603	0.02
Other utilization (not PD related)
Outpatient
VA	10,763 (614)	14,718 (1,003)	–3,956 (1,176)	–6,261 to –1,651	0.001
Medicare	12,026 (659)	8,724 (693)	3,302 (956)	1,427 to 5,176	0.001
Fee	2,420 (334)	2,896 (492)	–476 (595)	–1,643 to 690	0.42
Inpatient
VA	1,837 (861)	1,120 (460)	717 (976)	–1,195 to 2,629	0.46
Medicare	5,165 (690)	9,740 (1,164)	–4,575 (1,353)	–7,227 to –1,924	0.001
Fee	320 (107)	425 (111)	–104 (154)	–406 to 198	0.50
Pharmacy
VA	3,683 (291)	4,692 (347)	–1,009 (452)	–1,895 to –122	0.03
Medicare Part D	2,120 (240)	2,569 (402)	–449 (468)	–1,365 to 468	0.34
Fee Basis	0 (0)	21 (25)	–21 (25)	–70 to 28	0.40
Total costs without DBS procedures and complications	91,406 (3,300)	85,358 (4,086)	6,048 (5,253)	–4,247 to 16,343	0.25
Total costs with DBS procedures and complications	162,489 (3,874)	85,358 (4,086)	77,131 (5,631)	66,095 to 88,168	<0.001

^a Days of care in VA extended care facilities (i.e., VA community living centers and domiciliaries, as well as care in community nursing homes with which the VA contract).

^b Days of care in Medicare‐reimbursed short‐stay hospitals.

^c Days of care in Medicare‐reimbursed long‐stay hospitals and skilled nursing facilities.

Abbreviation: N/A, not applicable.

For non‐PD‐related services, costs of outpatient care from VA‐ or Medicare‐covered facilities were higher for medically managed patients than for DBS patients. Costs of inpatient care from Medicare‐covered facilities were higher for medically managed patients as were costs of medications from VA pharmacies.

After 1 year of follow‐up, average total costs, including DBS procedures and complications, were $59,129 (95% confidence interval [CI]: $54,973–$63,286; P < 0.001) higher per person for DBS patients ($74,136) than for medically managed patients ($15,007). After 5 years of follow‐up, average total costs, including DBS procedures and complications, were $77,131 (95% CI: $66,095–$88,168; P < 0.001) higher per person for DBS patients ($162,489) than for medically managed patients ($85,358; Fig. 2). Excluding costs of DBS procedures and complications, cumulative average total costs were not significantly different between DBS patients ($91,406) and medically managed patients ($85,358) at 5 years (DBS patients’ costs $6,048 higher; 95% CI: DBS patients’ cost $4,247 lower to $16,343 higher).

Figure 2.

Cumulative mean costs over 5 years of follow‐up. (A) Costs Including DBS procedures and complications. (B) Costs not including DBS procedures or complications.

Discussion

Approximately 630,000 people in the United States had diagnosed PD in 2010 with direct healthcare costs of over $14 billion ($22,000 per patient).3 Between 1.6% and 4.5% of PD patients are eligible for DBS.28 We found that healthcare costs were approximately $162,000 per patient over 5 years including and following DBS procedures. In contrast, 5‐year costs for a propensity‐score–matched cohort of medically managed patients were over $85,300. Total costs were over $59,100 higher for DBS patients 1 year following DBS than for the matched cohort of medically managed patients, and this difference persisted over 5 years of follow‐up. However, excluding costs of the DBS procedures and complications, healthcare costs were not significantly different between the groups at 5 years.

A survival advantage for DBS patients compared to medically managed patients was previously reported.14 That study found that veterans with PD who received DBS had longer survival than a matched group of veterans without DBS (mean, 6.3 vs. 5.7 years; P = 0.006).14 Recent articles examining cost‐effectiveness of DBS to best medical therapy have concluded that DBS is cost‐effective.12, 13, 29, 30, 31, 32 These articles reported incremental cost‐effectiveness ratios ranging from 6,70032 to 34,389 euros per quality‐adjusted life‐year.30 The Dams et al., Eggington, and Pietzsch et al. studies were based on actual patient‐level data that were used for Markov modeling projections to estimate cost‐effectiveness.12, 13, 32 The current study used actual patient‐level cost data over 5 years, which has not been previously reported. We found significant differences both overall and among specific components of healthcare costs between patients who received DBS and medical management over 5 years. Higher costs could reflect unmeasured differences between these groups. Cost differences could also reflect greater surveillance by healthcare providers of PD patients following DBS. With greater surveillance, there may be greater detection and treatment for both PD and other conditions than for non‐DBS patients. Another difference to previous analyses is that many studies have shown a reduction in the use of PD medication,9, 10, 11 translating to a cost benefit for DBS12, 13; however, the current study did not show a cost benefit from reductions in PD medication. We did not have information on the DBS target site. DBS of the STN has been associated with reductions in l‐dopa equivalents, whereas most studies have shown little change in PD‐related medication use follow GPi DBS.33, 34

Study strengths included using a large cohort of persons with PD across VA and non‐VA facilities. Both VA and Medicare data were used to identify the study population. However, there are limitations. First, it is based on a retrospective analysis of administrative and claims data. Although we used propensity matching to make our groups as similar as possible, we could not address unobserved differences that may influence costs and healthcare utilization (e.g., severity or duration of PD, the types and doses of PD medications that patients were on, or motor function before surgery). Second, we did not have information on quality of life; however, we previously reported that patients in our study cohort who received DBS had a modest survival benefit.14 The higher costs for the DBS group may reflect more‐intensive follow‐up and medical care. Third, if a PD diagnosis was not used, not all PD‐related visits may have been captured for the medically managed group. However, these visits would have been included with the other utilization, and costs for these visits would have been included in the total costs. Some of the PD‐related complications that may have been triggered by DBS could also occur in medically managed patients. However, these healthcare events would have been captured in the PD‐related or other utilization and costs and included in the total costs. Fourth, VA and Medicare databases did not have measures of disease severity, such as the H & Y scale or UPDRS. Fifth, our cohort was older than the median ages reported in clinical trials of DBS,33, 34 perhaps reflecting the population of persons with PD, who are typically older.1 Sixth, our cost estimates were based on expenditures from the VA and Medicare systems, which have differing payment systems. However, previous studies of the cost of care provided by VA hospitals have not shown evidence that VA hospitals were significantly more or less efficient than nonfederal hospitals, and that it was not more costly to provide care in VA facilities than in non‐VA facilities.35, 36, 37 Seventh, because the propensity score matching does not include PD disease characteristics (e.g., severity, motor response fluctuations, and cognitive impairment), the control group could be different regarding disease severity and necessary treatments, which is an important factor for disease‐related costs. Consequently, the control group may be different regarding their disease severity because they did not need/want DBS during the 5 years of follow‐up. Finally, our sample was comprised almost entirely of males, which has also been found to be related to mortality following DBS,38 and was limited to veterans with PD, which may not generalize to nonveteran populations.

Healthcare costs over 5 years were higher for veterans receiving DBS. These higher healthcare costs may reflect the costs of DBS in the United States (majority with two‐stage procedure) and any follow‐up required plus greater surveillance by healthcare professionals following DBS as well as unobserved differences in the patients who received medical management or DBS. To fully assess the economic value of DBS, both its relative benefits and costs should be taken into consideration, including the recent survival benefits reported with these data.14

Author Roles

1. Research project: A. Conception, B. Organization, C. Execution; 2. Statistical Analysis: A. Design, B. Execution, C. Review and Critique; 3. Manuscript Preparation: A. Writing of the first draft, B. Review and Critique.

K.T.S: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B

B.S.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B

F.M.W.: 1A, 1B, 1C, 2A, 2C, 3A, 3B

B.G.: 1C, 2A, 2B, 2C, 3B

Z.H.: 1C, 2B, 2C, 3B

L.C.: 1C, 2B, 3C, 3B

D.I.: 1B, 1C, 2C, 3B

K.A.F.: 1C, 2C, 3B

Disclosures

Ethical Compliance Statement: This study was approved by the Edward Hines Jr. Institutional Review Board. A waiver of informed consent was obtained for this study. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this work is consistent with those guidelines.

Funding Sources and Conflicts of Interest: The funding source for study was Medtronic. Stroupe, Smith, Weaver, Gonzalez, Huo, Cao, and Ippolito report Medtronic Neurologic grant funding.

Financial Disclosures for previous 12 months: Stroupe: Employment (Department of Veterans Affairs; Loyola University); Grants (Medtronic). Smith: Employment (Department of Veterans Affairs; Northwestern University); Grants (Medtronic). Weaver: Employment (Department of Veterans Affairs; Loyola University); Grants (Medtronic). Gonzalez: Employment (Department of Veterans Affairs); Grants (Medtronic). Huo: Employment (Department of Veterans Affairs); Grants (Medtronic). Cao: Employment (Department of Veterans Affairs); Grants (Medtronic). Ippolito: Employment (Department of Veterans Affairs); Grants (Medtronic). Follett: Employment (Department of Veterans Affairs; University of Nebraska).

Supporting information

Appendix S1. Parkinson's Disease Diagnosis Codes, DBS Codes, and Parkinson's Disease Medications. Adopted from the Parkinson's Disease Foundation website www.pdf.org “Treating Parkinson's Understanding Medications” (accessed September 15, 2017).

Appendix S2. Definition of Complications.