Risk of Schizophrenia Spectrum Disorders Following Parkinson’s Disease: A Nationwide Population-Based Cohort Study

Data sources

This study aimed to evaluate the risk of developing SSD among patients with PD and to determine whether the emergence of psychotic symptoms is independently attributable to PD itself. The analysis utilized health insurance claims and health screening data from 2012 to 2023 provided by the NHIS, which constitute a population-based dataset derived from a single-payer public healthcare system covering the entire Korean population and include a sample cohort of approximately two million individuals. The NHIS database comprises large-scale administrative data encompassing healthcare utilization records, International Classification of Diseases, 10th Revision (ICD-10) based diagnostic codes, procedures and prescription information, sociodemographic variables, and results from national health examinations, thereby providing comprehensive and high-quality health information representative of the general population [8]. In addition, the NHIS sample cohort includes detailed demographic information such as sex, age, and income level, and is constructed using stratified random sampling across approximately 1,476 strata based on these demographic variables, ensuring national representativeness at the population level [8].

This study was approved by the Institutional Review Board of Sungshin Women’s University (SSWUIRB-2025-106). As the data provided by the NHIS were fully anonymized and contained no personally identifiable information, the requirement for informed consent was waived.

Case and control group selection

To reduce reverse causation, this study implemented a washout period from 2012 to 2014, excluding individuals with prior diagnoses of PD or SSD to ensure that outcomes during follow-up represented incident cases. This approach minimizes the influence of pre-existing disease or latent symptom progression and has been emphasized as methodologically important in prior research [9]. A 3-year washout period was applied in this study based on the rationale. The observation period was then defined as 2015–2023 to accurately estimate the actual risk of developing SSD following a new diagnosis of PD.

In this study, the exposure disease, PD, was defined as having at least two diagnoses within a one-month period based on the ICD-10 codes presented in (Supplementary Table 1. According to previous research, PD has been operationally defined as a primary diagnosis of ICD-10 code G20 in conjunction with registration under the rare and intractable disease code (V124) [10]. Previous studies have also identified PD without using a V code, based on ICD-10-coded records alone, including definitions using repeated G20-coded records across years or at least one G20-coded record in administrative health data [11,12]. On this basis, a repeated-diagnosis criterion (≥2 diagnoses) was applied to minimize potential misclassification associated with single-visit diagnoses. The index date for the PD case group was defined as the date of the first recorded diagnosis.

The outcome of interest, SSD, was defined as having at least two diagnoses corresponding to one of the ICD-10 codes listed in Supplementary Table 1. Previous studies have defined schizophrenia-related outcomes either using the broader ICD-10 category of F20–F29 [13], or using narrower operational definitions such as schizophrenia defined by ICD-10 code F20 only or SSD defined while excluding F23 and F24 [14,15]. In the present study, F24 was excluded to focus on diagnostically core psychotic disorders of clinical relevance. In addition, individuals with exclusionary conditions identified during the washout period or before the index date were excluded, with detailed definitions and corresponding ICD-10 codes provided in Supplementary Table 2.

Individuals were excluded from the case group if they met any of the following criteria: 1) a diagnosis of PD during the washout period (2012–2014); 2) absence of health screening records within three years prior to the index date; 3) age younger than 20 years or older than 80 years at the index date; or 4) a prior diagnosis of SSD (ICD-10 codes F20–F29) within three years preceding the index date.

The control group was selected from individuals who participated in the National Health Screening Program between 2012 and 2023 and had no history of PD diagnosis. Propensity score matching was then performed at a ratio of 1:5 (case: control) based on matching variables including sex, age, and year of health screening, and the index date for each control was assigned to be identical to that of the matched case. Propensity scores were estimated using covariates obtainable from health screening data, including smoking status, alcohol consumption, body mass index (BMI), total cholesterol, triglycerides, high-density lipoprotein cholesterol and low-density lipoprotein cholesterol, systolic and diastolic blood pressure, fasting glucose, and income level. This approach was adopted to minimize confounding in accordance with prior methodological literature emphasizing the central role of propensity scores in observational studies [16]. In addition, missing values for variables used in the propensity score estimation were imputed using single imputation with predictive mean matching.

Variables

The analysis included key health screening variables, including sex, age, smoking status, alcohol consumption, height, weight, BMI, total cholesterol, systolic and diastolic blood pressure, fasting glucose, and income level. Each variable was extracted from the health screening record closest to the index date within the preceding 3 years, using the matched case’s index date for controls. Height and weight were treated as continuous variables, whereas the remaining indicators were categorized according to clinical relevance and incorporated into the analysis as categorical variables.

Statistical analysis

Baseline characteristics of the case and control groups were compared using appropriate statistical methods according to variable type. Continuous variables were assessed using Student’s t-test, while categorical variables were compared using Fisher’s exact test or Pearson’s χ² test, as appropriate. In addition, standardized differences were calculated to assess covariate balance between groups, with values <0.1 considered indicative of negligible imbalance. To compare the risk of developing SSD after PD diagnosis, crude incidence rates (IRs) per 1,000 person-years and incidence rate ratios (IRRs) were calculated. Cumulative incidence over time was estimated using Kaplan-Meier curves, and differences between the PD and control groups were evaluated with the log-rank test. Follow-up was censored at the earliest occurrence of (1) first diagnosis of SSD, (2) death, (3) the end of the study period (December 31, 2023) (Supplementary Figure 1). The association between PD and incident SSD was assessed using multivariable Cox proportional hazards regression models based on propensity score–matched data. The models included covariates obtainable from health screening data, including sex, age, smoking status, alcohol consumption, BMI, blood pressure, fasting glucose, blood lipid levels, and income level, and additional subgroup analyses were performed. All statistical analyses were conducted using R software (version 4.4.1; R Foundation for Statistical Computing), and a two-sided p-value<0.05 was considered statistically significant.

Overall summary of findings

Using a large population-based cohort, this study demonstrated a significantly increased risk of SSD following a new diagnosis of PD, which remained robust after adjustment for major confounders. The risk was particularly pronounced among individuals younger than 60 years and remained consistent across subgroup analyses according to age and sex. These findings suggest that psychotic symptoms following PD may reflect an increased risk of persistent psychotic disorders rather than transient medication-related effects.

Previous study

Previous studies have identified various risk factors associated with the occurrence of psychotic symptoms in patients with PD. However, important differences exist between these studies and the present study in terms of study design and outcome definition. Notably, a study using the Parkinson’s Progression Markers Initiative (PPMI) cohort [17] followed 423 newly diagnosed, drug-naïve patients with PD for a mean duration of 3–4 years and examined psychotic symptoms emerging during the disease course as the primary outcome. However, that study did not define outcomes as independent psychotic disorders meeting Diagnostic and Statistical Manual of Mental Disorders or ICD-10 diagnostic criteria, such as schizophrenia or schizoaffective disorder, and thus was limited in its ability to distinguish transient or mild symptoms from chronic psychotic conditions requiring formal clinical diagnosis. In contrast, the present study defined the outcome as SSD and evaluated the risk of incident, clinically diagnosed psychotic disorders using ICD-10 diagnostic codes.

A prospective study using the PPMI cohort followed 338 newly diagnosed patients with PD and examined the development of psychotic symptoms as the primary outcome [18]. However, that study defined outcomes as the emergence of symptoms rather than clinically diagnosed psychotic disorders, and its follow-up period was limited to approximately 5 years, restricting the evaluation of long-term risk for psychotic disorders after PD. Moreover, the absence of a washout period to exclude pre-existing or prodromal psychotic disorders limits the ability to rule out reverse causation. In contrast, this study applied a 3-year washout period and followed participants for up to 9 years to assess incident SSD, thereby providing a more precise estimation of the long-term risk of psychotic disorders following PD.

Pathophysiological mechanisms

PD is increasingly understood not as a disorder confined to degeneration of the nigrostriatal dopaminergic pathway, but as a condition involving widespread dysfunction across large-scale neural networks, including the cortico-striatal-limbic circuits. Recent reviews indicate that PD psychosis is not merely the result of dopaminergic excess, but rather reflects a complex interplay of dysregulation across multiple neurotransmitter systems—including dopamine, serotonin, and acetylcholine—along with network-level functional impairments.

Notably, the fronto-striatal-limbic circuitry plays a critical role in reality monitoring and cognitive control, and functional abnormalities within this network substantially overlap with the pathophysiology of SSD. A recent systematic review comparing psychotic symptoms in patients with PD and schizophrenia reported shared alterations in functional connectivity involving the frontal and temporal cortices as well as the striatum in both conditions [19].

Neuroimaging studies further support this hypothesis. In patients with PD accompanied by hallucinations, aberrant functional connectivity within the cortico-striatal-thalamocortical circuits has been observed, and dysfunction of the cholinergic system has been identified as a key contributor to impaired frontal cognitive control and the emergence of psychotic symptoms [19,20].

Taken together, these converging lines of evidence suggest that the increased risk of SSD observed following PD in the present study may be attributable to progressive neurodegeneration intrinsic to PD and associated network-level dysfunction. This interpretation is consistent with findings from large population-based cohort studies reporting an elevated risk of SSD after PD diagnosis [7].

Age and sex

In this study, the risk of SSD following PD appeared to be higher among younger individuals, and this pattern may suggest earlier manifestation of underlying psychosis vulnerability [21]. This interpretation is consistent with previous reports indicating that early-onset or atypical PD is often characterized by more prominent non-motor symptoms, which may confer a higher risk of psychotic manifestations [4,7]. However, because some age-stratified subgroup estimates were based on a small number of events, these findings should be interpreted cautiously and regarded as exploratory.

In sex-stratified analyses, SSD risk appeared to be higher in males, consistent with epidemiological evidence of earlier onset and poorer outcomes of schizophrenia in males [22]. The persistence of this sex-related vulnerability pattern among patients with PD may suggest that PD could amplify pre-existing psychosis risk structures rather than fundamentally altering them [23]. Although sex differences in PD likely reflect a multifactorial interplay of biological and clinical factors, given the limited number of events in some subgroups and the resulting limited statistical power, these findings should not be interpreted as definitive evidence of sex-specific differences in post-PD psychosis risk.

In joint age and sex stratified analyses, specific subgroups exhibited disproportionately higher risks, indicating that post-PD psychotic risk cannot be explained by age or sex alone. In particular, the elevated risk among younger females may suggest attenuation of the typically observed female protective effect in the context of PD, potentially reflecting altered sex-specific risk patterns and interactions with non-motor symptoms [24,25]. However, because the number of events in some age–sex intersectional subgroups may be limited, these observations should be interpreted cautiously from a hypothesis-generating perspective. Because the number of events was very small in subgroups, these estimates may be unstable, and further studies with larger subgroup sample sizes are needed to clarify whether these patterns represent true subgroup differences.

Smoking status

In analyses stratified by smoking status, relatively higher risks of psychotic disorders were observed among never smokers and former smokers. However, these findings should be interpreted with caution because some smoking-stratified subgroups had very few events, limiting statistical power. Potential selection bias and reverse causality inherent to observational studies should also be considered [26]. Moreover, given that prior literature has consistently reported smoking as a risk factor for psychotic disorders [27], the present findings warrant careful interpretation and should not be construed as evidence of a protective causal effect of smoking.

Alcohol

The higher risk observed among moderate drinkers should be interpreted with caution, as some alcohol-stratified subgroups included very small numbers of events, which may have resulted in limited statistical power and unstable estimates. In addition, observational findings on alcohol-related outcomes may be susceptible to bias and should not be overinterpreted as evidence of a causal relationship [28,29].

BMI

In BMI-stratified analyses, underweight patients with PD showed a relatively higher risk of psychotic disorders, a pattern that may be consistent with prior studies reporting associations between weight loss, low BMI, non-motor symptoms, and disease progression in PD [30]. However, because some BMI-stratified subgroups had very few events, these findings should be interpreted with caution. In PD, low BMI may be better understood as a marker of underlying disease burden and non-motor symptom severity rather than as a direct causal determinant of SSD.

Significance of the study and clinical implications

The present findings demonstrating an increased risk of SSD following PD suggest that psychotic manifestations in PD extend beyond transient non-motor symptoms and may translate into a substantive risk of psychiatric morbidity [31]. Psychotic symptoms in patients with PD have been associated with cognitive decline, increased caregiver burden, and elevated risks of hospitalization and long-term institutionalization, collectively contributing to a heightened overall clinical burden. In this context, the observed increase in the risk of SSD after PD supports the need for careful clinical monitoring and management strategies, although this interpretation should be made cautiously [32].

A key clinical implication of this study is the identification of heterogeneity in psychosis risk among patients with PD. Higher risks observed in younger patients, specific age–sex subgroups, and underweight individuals may indicate that the risk of developing SSD varies according to age, sex, and overall health status. Because several subgroup estimates were based on sparse events, they should be interpreted cautiously. Nevertheless, the overall pattern of the findings suggests that psychosis risk after PD may not be uniform across all patients. Age, sex, and BMI may therefore be considered as potentially relevant factors when assessing psychiatric risk in patients with PD. Overall, this study contributes to a broader understanding of SSD risk in PD and may help inform future research on subgroup-based risk stratification and psychiatric monitoring strategies.

Limitation and future work

Although this study has notable strengths, including the use of a large, population-based cohort to evaluate the long-term risk of SSD following PD, several limitations related to the study design and data characteristics should be considered.

First, as this study is based on an observational design, the possibility of residual confounding due to unmeasured factors cannot be fully excluded. While propensity score matching and multivariable adjustment were employed to control for major confounders, variables such as genetic factors, family history, educational attainment, and social support could not be directly assessed using the NHIS data and may have influenced the observed associations.

Second, PD and SSD were defined using ICD-10 diagnostic codes, which may introduce diagnostic misclassification. Although a repeated-diagnosis criterion was applied, heterogeneity in diagnostic practices and the lack of detailed clinical information on SSD such as symptom subtype, severity, and disease duration could not be fully addressed.

Third, this study could not directly account for key diseasespecific characteristics of PD, including clinical severity, detailed non-motor symptom profiles, and cognitive function. Although these factors are likely to be closely associated with the risk of psychotic disorders, they are difficult to quantify using health insurance claims data.

Fourth, detailed analyses of medication-related factors were limited. Although prescription information was available, time-varying effects such as antiparkinsonian drug dosage, cumulative exposure, and treatment adherence could not be fully incorporated. Consequently, the independent contribution of pharmacological factors to the risk of psychotic disorders could not be precisely disentangled.

Fifth, because health screening variables were obtained from the record closest to the index date within the preceding three years, the time interval between the health screening date and the index date may have varied across individuals. Accordingly, some baseline covariates may not have fully reflected participants’ status at the index date, and residual bias cannot be excluded.

Future studies integrating clinical, pharmacological, neuroimaging, and biological data are needed to clarify the mechanistic heterogeneity and clinically meaningful subtypes of psychosis following PD, informed by evidence of shared and distinct neural substrates between PD-related psychosis and schizophrenia [19]. In addition, further studies are needed to replicate and extend the present findings using prospective cohort designs or clinical data-based research.

Conclusion

This nationwide population-based cohort study demonstrated an increased risk of SSD following a new diagnosis of PD, which remained robust after adjustment for major confounders. The risk was particularly pronounced among younger patients, specific age and sex subgroups, and underweight individuals, and appeared to increase cumulatively over time. These findings highlight the need for long-term psychiatric risk assessment and monitoring in patients with PD that incorporates age, sex, and overall health status.