Temporal trends and risk factors of hospital readmission following deep brain stimulation for Parkinson disease

Abstract

Background:

Risk factors and long-term data on readmissions following deep brain stimulation (DBS) for Parkinson Disease (PD) remain unclear.

Objective:

To evaluate the temporal trend and risk factors associated with hospital readmission following DBS for PD.

Methods:

We analyzed DBS procedures for PD in the National Readmissions Database (2016–2022). The association between demographic, clinical, and hospital factors with 90-day readmission following DBS was evaluated using a multivariable logistic and Cox regression models.

Results:

Among 7535 DBS surgeries, 8.1 % individuals were readmitted within 90 days. Higher Elixhauser-Comorbidity Index and complicated diabetes mellitus were associated with a higher risk of 90-day read-mission (adjusted odds ratio 1.60, 95 % CI 1.06–2.41). Readmissions did not statistically significantly change between 2016 and 2022.

Conclusion:

The rate of non-elective readmissions following DBS for PD has not changed over time. A higher comorbidity index and complicated diabetes mellitus were identified as key risk factors.

Deep brain stimulation (DBS) is an effective therapy for Parkinson Disease (PD), resulting in meaningful improvements in quality of life [1]. Over the past few years, advances in device technology and surgical practice have contributed to the overall increase in the number of DBS surgeries for PD [2]. Although this elective surgery is generally considered safe, hospital readmissions following DBS may occur and contribute to patient morbidity and increased healthcare costs, and utilization. Despite being a common surgical procedure, national trends and contributing factors to DBS readmission rates remain unclear, limiting the possibility of interventions to improve outcomes.

The aim of this study is to identify clinical and demographic factors associated with hospital readmission risk and evaluate temporal trends in hospital readmissions after DBS for PD in the United States (US).

1. Methods

1.1. Population

We used data from the Healthcare Cost and Utilization Project (HCUP) Nationwide Readmissions Database (NRD) from 2016 to 2022 to identify adult patients with PD who underwent DBS (See Supplementary Table 1 for definitions) [3]. The NRD is a database compiling all-payer readmission data in the United States within a calendar year. The first DBS-coded admission for an individual patient during a calendar year was considered the index hospitalization. Exclusion criteria included death during index hospitalization and prolonged length of stay (>30 days), to focus on a homogeneous population with a typical post-operative course. Elective hospital readmission, such as battery placement or staged DBS procedure, was not counted as an outcome event.

1.2. Outcomes and study variables

The primary outcome of the study was hospital readmission within 90 days following discharge for the index hospitalization of DBS surgery. Patients with no readmission were censored at 90 days.

The following baseline patient characteristics were evaluated: age, sex, urban vs. rural residence, and quartile of ZIP code-level median household income. Clinical covariates included comorbidities derived from the Elixhauser-comorbidity Index, defined as continuous scores and categorical groups [4–6]. According to prior literature, complicated DM relates to end-organ damage, including neuropathy, retinopathy, or nephropathy [7]. Insurance types were reported as: private, Medicare, Medicaid, and others (unknown and uninsured) [8,9]. We also evaluated hospital-level covariates, including number of beds, teaching status, discharge disposition, and primary payer (Supplementary Material 1). Race and ethnicity are not described in the NRD database. For temporal analyses of readmission reasons, diagnoses were collapsed into three groups due to small cell counts of specific ICD-10 codes (procedure, neurologic, and medical-related).

1.3. Statistical analyses

We evaluated the trend of 90-day non-elective readmissions over time, with the hypothesis that this number would decrease over time. To produce nationally representative estimates, we applied survey weights, strata, and clusters provided by HCUP. Multivariable logistic regression was used to estimate odds ratios (ORs) and 95 % confidence intervals (CIs) for predictors of 90-day readmission following DBS. Candidate variables for adjustment were selected a priori based on biological plausibility and previous literature and included patient demographics, comorbidities, and hospital-level characteristics (Supplementary Table 2). Normality was assessed with the Hosmer–Lemeshow goodness-of-fit tests. Multicollinearity was assessed via the variance inflation factor (VIF). Time-to-event analysis was performed using Cox proportional hazard models, with the use of Schoenfeld residuals for proportional hazards assumptions testing. A two-sided p < 0.05 was considered statistically significant. As suggested by prior recommendations, multiple comparison adjustment was not applied as our analysis was hypothesis-driven and implemented regression techniques and not multiple independent test [10–12]. Moreover, our model included covariates from previously specified points, including clinical judgement and prior studies. Stata (18.0, StataCorp, College Station, TX) was used for all statistical analyses.

2. Results

A total of 7535 DBS surgeries for PD were recorded in the NRD database. Of those, 613 patients (8.1 %) were readmitted within 90 days of discharge.

Baseline characteristics of the study population are reported in Table 1. The readmitted group was older (44.5 % were age 70 or older) compared to the non-readmitted group (34 % were age 70 or older), but did not differ in sex or income distribution. The readmitted group had a higher Elixhauser-Comorbidity Index than the non-readmitted group (p < 0.001). Specific comorbidities were higher in the readmitted vs. non-readmitted group (Supplementary Table 2), including: congestive heart failure (3.8 % vs. 1.8 %, p < 0.001), pulmonary disease (12.6 % vs. 8.5 %, p < 0.001), diabetes mellitus type II (8.3 % vs. 3.6 %, p < 0.001), renal failure (6.4 % vs. 3.6 %, p > 0.001), and hypertension (7.7 % vs. 3.8 %, p < 0.001). When evaluating the readmission trend over time, the overall weighted 90-day readmission rate was 7.9 %. This rate was relatively similar between 2016 and 2022 (range 6.8 %–8.6 %), with no significant temporal trend (design-based F = 0.61, p = 0.71) (Supplementary Fig. 3). The Kaplan-Meier curve of the time-to-event model is seen in Supplementary Fig. 4. The rate of readmission in that model was also 8.1 % (95 % CI 7.5–8.8).

Table 1.

Baseline characteristics of study participants.

Characteristics	Total number of patients
	No readmission	Readmission	P-value
	6922 (91.9 %)	613 (8.1 %)
Age Groups
18–49 (ref)	372 (5.4 %)	33 (5.4 %)	<0.001
50–59	1276 (18.4 %)	84 (13.7 %)
60–69	2923 (42.2 %)	223 (36.4 %)
70+	2351 (34.0 %)	273 (44.5 %)
Female, N (%)	2145 (31.0 %)	187 (30.5 %)	0.804
Urban Residence, N (%)
Urban	6057 (87.5 %)	538 (87.8 %)	0.851
Median Income, N (%)
76th-100th percentile	2268 (32.8 %)	203 (33.1 %)	0.574
51st- 75th percentile	2024 (29.2 %)	179 (29.2 %)
26th-50th percentile	1615 (23.3 %)	131 (21.4 %)
0th-25th percentile	1015 (14.7 %)	100 (16.3 %)
Elixhauser-Comorbidity Index
Mild	2150 (31.1 %)	152 (24.8 %)	<0.001
Moderate	2188 (31.6 %)	164 (26.8 %)
High	1394 (20.1 %)	142 (23.2 %)
Severe	1190 (17.2 %)	155 (25.3 %)
Discharge Status, N (%)
Home	5953 (86.0 %)	483 (78.8 %)	<0.001
Home Health Care	757 (10.9 %)	93 (15.2 %)
Other Healthcare Facility	212 (3.1 %)	37 (6.0 %)
Primary Payer, N (%)
Medicare (ref)	4523 (65.3 %)	439 (71.6 %)	0.002
Other	2399 (34.7 %)	174 (28.4 %)
Length of stay, days
0–1	4930 (71.2 %)	394 (64.3 %)	<0.001
2–3	1718 (24.8 %)	166 (27.1 %)
4–30	274 (4.0 %)	53 (8.6 %)
Teaching Status, N (%)
Non-teaching	234 (3.4 %)	21 (3.4 %)	0.953
Teaching	6688 (96.6 %)	592 (96.6 %)
Hospital Ownership, N (%)
Government (ref)	1182 (17.1 %)	87 (14.2 %)	0.101
Private, nonprofit	5434 (78.5 %)	492 (80.3 %)
Private, investor-owned	306 (4.4 %)	34 (5.5 %)

Univariate comparison of demographic/hospital-associated characteristics and Hospital Readmission 90 days after deep brain stimulation.

The results of the multivariable logistic regression, modeling the impact of all a priori comorbidities on readmission at 90 days, are displayed in Supplementary Fig. 1. A higher Elixhauser-Comorbidity index (high or severe) was associated with significantly increased odds of readmission when compared to lower scores (mild and moderate). When evaluating each specific comorbidity, complicated diabetes mellitus (DM) remained the only significant comorbidity over time (aOR 1.60, 95 % CI 1.06–2.41, p = 0.025) (Supplementary Table 3).

The most common reason for readmission was the ICD-10 code “injury and poisoning” (33.3 %), which primarily includes postoperative complications (e.g. infection of the implant, bleed), fall-related injuries, and hardware-related issues. This was followed by primary “neurological issues” (26.8 %), followed by “cardiovascular disease “ (9.8 %) (Supplementary Table 3). In the survey-weighted Cox model, the hazard ratio of readmission for patients with moderate severity by Elixhauser was 1.06 (95 % CI 0.84–1.33, p = 0.63), for high severity was 1.51 (95 % CI 1.18–1.94, p = 0.001), and for severe was 1.83 (95 % CI 1.41–2.38, p < 0.001), compared to the mild reference group (Fig. 1). The distribution of readmission diagnoses remained stable over time with no statistical difference across the studied period (F = 1.02, p = 0.43, and survey-weighted multinomial regression, p = 0.55) (Supplementary Table 4).

Fig. 1.

90-day Readmission following deep brain stimulation surgery stratified by Elixhauser-Comorbidity index.

Weighted percentage of patients readmitted by days stratified by Elixhauser-comorbidity index.

3. Discussion

In this nationwide analysis, we found a 7.9 % of readmissions following DBS, which has remained relatively unchanged from 2016 to 2022. We also highlighted specific risk factors associated with an increased risk of hospital readmission, including a higher Elixhauser-Comorbidity Index and the presence of complicated DM.

The overall rate of readmission following DBS for PD within 90 days was slightly higher in our study (7.9 %) than in others. Rumalla et al. reported a 90-day readmission rate of 4.3 % when evaluating 3392 patients who underwent DBS for several conditions, including PD (around 70 % of all patients), essential tremor (ET), or dystonia, for a span of 9 months between January and September 2013 [13]. Readmission risk factors were age (patients older than 75 years), more than 3 comorbidities, and individuals with obesity, prior history of stroke, or perioperative hematoma. Given that our data covered 7 years, instead of 9 months, and was restricted to PD cases, rather than including all DBS surgeries, our findings are more specifically useful to the PD DBS population. The higher readmission rate in our cohort might reflect a longer observation window and the exclusion of other conditions, such as ET, often presenting at a younger age and associated with few comorbidities.

In a subsequent study, the same NRD database was evaluated between 2013 and 2014 [14]. Among the 6085 DBS procedures for PD, non-elective 30-day readmissions were relatively less common, estimated at around 4.9 % of cases. Associated risk factors included lower socioeconomic status, higher comorbidity burden, and treatment at teaching hospitals. In our study, we only replicated the finding of a higher comorbidity index as a risk factor, likely due to our extended 90-day follow-up, larger number of patients, and longer analysis window of 7 years. DeLong et al. reported a 90-day complication rate of 7.5 % among the 1757 patients who underwent DBS for PD between 2000 and 2009. The most common complications were wound infections (3.6 %), pneumonia (2.3 %), intracranial hemorrhage or hematoma (1.4 %), and pulmonary embolism (0.6 %). The readmission rate was not described in this study [15].

DM has been previously identified as a risk factor for hospital read-mission across multiple cohorts [16–18]. In patients with PD, the presence of DM has been associated with a higher risk of readmission following percutaneous endoscopic gastrostomy (PEG) placement [19]. Although the mechanism for such association is not completely clear, a possible hypothesis is that poor glycemic control might increase the risk of infections, delayed wound healing, or cause multiple metabolic derangements. Moreover, patients with DM might experience additional medical complexity through additional comorbidities and poly-pharmacy [17,18].

Notably, the stable readmission rate over time is surprising considering changes in technique and guidelines over time. For instance, this study period included the use of standardized asleep surgery workflows, surgical robots, frameless devices, and new directional and sensing leads [20]. In our study, there were no changes related to readmission diagnosis across the studied years. Additionally, while the overall number of post-operative complications and readmissions increased during the COVID-19 pandemic, this does not appear to have been the case in DBS for PD [21]. From a clinical standpoint, our findings highlight the importance of a comprehensive pre-operative assessment, with particular attention to incorporating data related to comorbidities and complicated diabetes as part of surgical planning, as well as pre-operative counseling.

There are several limitations to our study. First, any health service database includes the possibility of inaccurate coding, which may not directly reflect a patient's diagnosis. However, we believe this number is small, as patients who went for DBS likely had PD. The ICD code for PD in the NIS and NRD dataset has also been previously validated, and its sensitivity and specificity were considered high [22]. Additionally, we were unable to stratify PD cases by severity or clinical stage, as the NIS/NRD databases do not offer detailed information regarding PD diagnosis, including duration of disease or specific clinical scales scores. Although this dataset is more likely to capture readmission in different hospitals as it obtains data from insurance claims, it does not take into account readmission across state lines. Finally, we do not have detailed information on the specific DBS volumes by each center, DBS targets, and specific readmission reasons within each ICD-10 code, as this is not described in the NRD database.

4. Conclusion

In this nationwide study, unplanned readmissions following DBS surgery for PD were uncommon, although the number is higher than suggested by prior studies. Patient-related factors such as high comorbidity burden and complicated DM emerged as key predictors. Despite minor fluctuations over time and the COVID pandemic, there was no reduction in readmission rates from 2016 to 2022.

Additional research is needed to identify treatable risk factors and surgical techniques that might allow us to improve outcomes and minimize readmission risk. Further research focusing on preventing complications is necessary to avoid negative clinical outcomes, healthcare-associated costs, and utilization.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.parkreldis.2026.108196.