Breakthrough Study Develops Personalised Tools to Forecast Nursing Home Needs in Parkinson's Disease
A new open-access study published in Parkinsonism & Related Disorders presents the first prognostic models designed to predict the risk of institutionalisation for people with Parkinson's disease using data from multiple European incidence cohorts. The research, led by Yan Li alongside David J. McLernon, Rachael A. Lawson, Alison J. Yarnall, David Bäckström, Lars Forsgren, Marta Camacho, Caroline H. Williams-Gray, Jodi Maple-Grødem, Guido Alves, Ole-Bjørn Tysnes, Carl E. Counsell, and Angus D. Macleod on behalf of the Parkinson's Incidence Cohorts Collaboration, draws on individual participant data from 1,046 patients across six cohorts.
The full publication is available at https://www.sciencedirect.com/science/article/pii/S1353802026002178. It appeared online on 23 June 2026 and forms part of the journal's August 2026 issue.
Understanding Institutionalisation Challenges in Parkinson's Disease
Parkinson's disease is a progressive neurodegenerative condition that affects movement, cognition, and daily functioning. As the disease advances, many individuals experience increasing dependency. When home-based support proves insufficient, some transition to institutional care such as nursing homes. This shift carries significant emotional, social, and economic consequences for patients, families, and healthcare systems.
Global estimates indicate that Parkinson's affects millions, with incidence rising sharply after age 60. In advanced stages, motor symptoms like rigidity and bradykinesia combine with non-motor issues including cognitive decline to heighten care needs. Institutionalisation rates vary by region due to differences in social support structures, family caregiving traditions, and access to community services. In Europe, where the study cohorts originated, healthcare planning increasingly emphasises early identification of high-risk individuals to optimise resource allocation.
The Parkinson's Incidence Cohorts Collaboration and Data Sources
The research harnesses the Parkinson's Incidence Cohorts Collaboration, which pools prospectively collected data from newly diagnosed patients in six European studies. These include cohorts from the United Kingdom, Sweden, Norway, and other sites, ensuring representation across varied healthcare settings. The combined dataset of 1,046 participants provides robust statistical power for identifying reliable prognostic signals.
Researchers tracked participants from diagnosis onward, recording the timing of any move to institutional care. This incidence-based approach, rather than prevalence sampling, minimises bias and captures the natural history from early disease stages. The collaboration's design supports both identification of prognostic factors and development of predictive models applicable across similar populations.
Methods: Two-Stage IPD Meta-Analysis and Flexible Survival Modelling
To identify prognostic factors, the team employed a two-stage individual participant data meta-analysis. In the first stage, multivariable Cox regression models were fitted within each cohort separately. The second stage then pooled the estimates using random-effects meta-analysis to account for between-cohort heterogeneity.
For model development, investigators applied the Royston-Parmar flexible parametric survival model. This approach models the baseline hazard function smoothly while allowing time-varying effects, offering advantages over standard Cox models for longer-term predictions. Separate models were built for institutionalisation risk within seven years and within ten years of diagnosis.
Performance evaluation relied on internal-external cross-validation. This technique holds out one cohort at a time for testing while training on the remainder, providing realistic estimates of how models might perform in new but similar settings. Discrimination was measured via C-statistics, which quantify the model's ability to rank individuals correctly by risk. Calibration assessed agreement between predicted and observed risks, with recalibration applied where needed to improve fit.
Key Results: Incidence Rates and Independent Prognostic Factors
Over the follow-up period, the cumulative incidence of institutionalisation reached 37.2 percent by ten years after diagnosis. Incidence rates across cohorts ranged from 1.7 to 6.2 events per 100 person-years, highlighting variability linked to population characteristics and local care pathways.
Three baseline factors emerged as independent predictors after multivariable adjustment: older age at diagnosis, higher scores on the Movement Disorder Society-Unified Parkinson's Disease Rating Scale part 3 (the motor examination subscale), and lower scores on the Mini-Mental State Examination, a widely used cognitive screening tool. These variables retained significance across the meta-analysis, underscoring their consistent association with future care needs.
Other candidate factors, including sex, disease duration at baseline, and certain non-motor symptoms, did not demonstrate independent effects once the core trio was accounted for. The findings emphasise that both motor severity and cognitive status at diagnosis carry important long-term implications.
Photo by National Cancer Institute on Unsplash
Prognostic Model Performance and Validation Outcomes
The seven-year and ten-year models demonstrated good discrimination, with C-statistics ranging from 0.71 to 0.84 and 0.73 to 0.81 respectively across validation folds. These values indicate strong ability to distinguish higher-risk from lower-risk individuals within the studied populations.
Calibration proved more variable. Some cohorts showed under-prediction while others exhibited over-prediction of institutionalisation events. Recalibration, which involved updating the model intercept and slope coefficients using cohort-specific data, substantially improved agreement between predictions and observed outcomes.
The authors stress that these models require additional external validation before widespread clinical adoption. Differences in healthcare systems, cultural attitudes toward institutional care, and availability of home support services may affect transportability to non-European settings.
Clinical and Healthcare Planning Implications
Personalised risk estimates could enhance conversations between clinicians and patients about future care options. Individuals identified as higher risk might benefit from earlier discussions regarding advance care planning, home modifications, or community support services. Healthcare administrators could use aggregated predictions to forecast demand for long-term care beds and allocate resources more efficiently.
The study also highlights opportunities for targeted interventions. While age itself is non-modifiable, motor symptoms and cognition may respond to optimised dopaminergic therapy, physical rehabilitation, and cognitive training programmes. Future research could explore whether addressing these factors delays institutionalisation.
From a policy perspective, the 37 percent ten-year figure underscores the substantial long-term care burden associated with Parkinson's. Integrated care models that bridge neurology, geriatrics, and social services may help reduce unnecessary transitions to institutional settings.
Broader Context Within Parkinson's Research Landscape
This work builds on earlier analyses of institutionalisation incidence and risk factors conducted by the same lead author using overlapping cohorts. It represents a methodological advance by moving from descriptive epidemiology to individualised prediction.
Related efforts in the field have focused on prognostic models for mortality, dependency, and falls. The current models complement these by addressing a distinct but interconnected outcome. Collaborative networks like the one underpinning this study facilitate the large-scale data sharing essential for reliable prognostic research in relatively rare conditions.
Academic institutions with strong neurology and medical statistics departments are well positioned to contribute to and benefit from such multi-centre initiatives. Training programmes that combine clinical neurology with advanced statistical methods prepare researchers for this type of work.
Future Directions and Recommendations for Researchers
The authors recommend further validation studies in diverse populations, including non-European cohorts and community-based samples. Incorporation of additional biomarkers, genetic data, or longitudinal trajectories of motor and cognitive scores could enhance model accuracy.
Digital health tools offer one avenue for implementation. Mobile applications or electronic health record integrations could present personalised risk estimates to clinicians at the point of care. User-friendly visualisations of survival curves might aid shared decision-making.
Funding bodies and research councils have supported the underlying cohorts through grants from organisations such as Parkinson's UK and national health research agencies. Continued investment in longitudinal data infrastructure will be vital for refining these and future prognostic tools.
Opportunities for Academic and Clinical Career Development
Research of this nature creates pathways for early-career investigators in epidemiology, biostatistics, and movement disorders. Postdoctoral positions and research fellowships focused on prognostic modelling or collaborative data analysis are increasingly available at universities with established Parkinson's programmes.
Clinician-scientists who master both patient care and advanced analytics can lead translational projects that directly influence practice guidelines. Professional societies in neurology and geriatrics regularly highlight the need for expertise in predictive modelling as healthcare systems shift toward value-based and personalised approaches.
Institutions seeking to strengthen their research portfolios in neurodegenerative diseases may prioritise recruitment of faculty with experience in individual participant data meta-analysis and flexible survival modelling techniques.
Conclusion: Toward More Informed and Personalised Parkinson's Care
The development and validation of these prognostic models mark an important step toward anticipating and mitigating the challenges of advanced Parkinson's disease. By quantifying individual risk based on readily available clinical measures, the work supports more proactive, patient-centred planning. As further validation and refinement occur, these tools have the potential to improve quality of life for patients while aiding efficient healthcare delivery. Researchers, clinicians, and policymakers alike will find value in engaging with the detailed findings now available in the open-access publication.
