STRENGTHS AND LIMITATIONS OF THIS STUDY
This scoping review will identify, summarise and appraise existing multivariable prediction models for Parkinson’s disease (PD) prognostication.
Relevant research articles will be identified using a systematic search strategy that adheres to Preferred Reporting Items for Systematic Review and Meta-Analysis extension for Scoping Reviews (PRISMA-ScR) guidelines and evaluated using Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis + Artificial Intelligence (TRIPOD+AI) and Prediction model Risk Of Bias ASsessment Tool (PROBAST) reporting guidelines.
This review will include multivariable models that predict disease progression in individuals with PD using traditional statistical methods or machine learning; however, not all have been validated for use in clinical settings.
A limitation is that the literature search will be restricted to articles published in English.
Introduction
Background
Parkinson’s disease (PD) is the second most common neurodegenerative disorder and affects over 10 million persons worldwide. 1 2 PD is a progressive movement disorder that causes a combination of motor symptoms (eg, bradykinesia, rigidity, tremor, gait instability) and non-motor symptoms (eg, hyposmia, sleep and mood disturbances, orthostatic hypotension, bowel and urinary dysfunction, memory loss). 3 The exact cause of PD remains unknown but may involve a combination of genetic and environmental risk factors. 4 Unfortunately, no disease-modifying therapies are currently available to slow PD progression and there is no known cure. As there is no way to predict an individual’s disease trajectory following PD diagnosis, clinical prediction models are needed to aid in prognostication.
The heterogeneity of PD symptoms poses prognostic challenges. While neurologists use a variety of validated assessments to diagnose and quantify disease burden in both clinical and research settings, existing assessments do not predict other clinical outcomes, such as PD progression or the risk of PD-associated mortality. In these instances, prognostic models, a type of clinical prediction model, can be used to help estimate the probability that a specific event will occur in the future. 5 By providing estimates of outcome probabilities among persons with PD, prognostic models can be used to complement clinical judgement and practice guidelines, leading to better medical decision-making, individualised treatment plans and improved resource allocation.
Study rationale
Few studies have developed prognostic models for persons with PD. In addition, many published models include only one predictor variable, which are limited in their ability to measure complex relationships between variables and tend to exaggerate statistical significance. 6 Because physicians naturally integrate patient characteristics, symptoms and test results to make a clinical prediction, prognostication is inherently multivariable. 5 Thus, a summary of existing multivariable prognostic models in PD is needed. Although prior reviews have been published on the topic, there are notable methodological limitations. For example, a recent comprehensive review article published in the IEEE Journal of Biomedical and Health Informatics summarised multimodality prognostic models in PD. 7 Results were limited to studies published between 2016 and 2021 and were organised by predictor variable rather than outcome variable. 7 In addition, this review was intended for a research audience and did not emphasise the clinical relevance of individual variables or multivariable models. To expand upon this work, we propose identifying and summarising all existing prognostic models in PD and will organise our results by outcome variable, which we believe is a more clinically meaningful way to categorise results. Outcome variables will be defined as markers of disease progression that can be measured in clinical settings, including, but not limited to, general disease progression as measured by the Hoehn and Yahr scale, 8 motor symptom severity as measured by the Unified Parkinson’s Disease Rating Scale (UPDRS) Part III, 9 non-motor symptom burden as measured by the UPDRS Part I and Non-Motor Symptom Questionnaire 10 and other discrete clinical outcomes (eg, freezing of gait, dementia). We will also discuss the clinical relevance of all included variables and models.
In addition to identifying and summarising existing prognostic models, we will appraise model development (eg, validation) and model performance (eg, calibration, discrimination), allowing us to make recommendations for their use in clinical practice. A prior review of PD progression models examined quantifying changes in total UPDRS scores against time without assessing model performance. 11 A more recent systematic review of prognostic models in PD explored model performance but did not discuss the clinical relevance of qualifying models in detail. 12 Thus, we plan to address these shortcomings by providing both a comprehensive summary and assessment of current PD prediction models.
As noted in the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, the quality of reporting of multivariable prognostic models is poor, 5 and the majority of existing models have not been validated on external datasets, further limiting their clinical utility. To determine the appropriateness of these models for use in clinical practice, we will use the updated TRIPOD+Artificial Intelligence (AI) checklist to ensure that studies using traditional regression or machine learning methods are reported completely, accurately and transparently. 13 We will also use the Prediction model Risk Of Bias ASsessment Tool (PROBAST) guidelines to evaluate the quality, risk of bias and applicability of models in clinical practice. 14
We developed this scoping review protocol to outline our approach, which includes summarising existing multivariable prognostic models in PD, evaluating their methodology and identifying gaps in the literature. From there, we intend to analyse how existing models can be applied to PD clinical care and inform future research directions.
Study objectives
Using guidance from Damen et al , 15 we developed the following specific objectives for this scoping review protocol:
To identify all existing multivariable prognostic models (developed or validated) that predict disease progression in persons with PD.
To summarise results by clinical outcome variable.
To evaluate model development, internal/external validation and risk of bias.
To identify knowledge gaps and propose how existing models can be applied to PD clinical care.
Methods and analysis
We chose to conduct a scoping review of existing literature on multivariable prognostic models in PD because, to the best of our knowledge, no studies have evaluated all existing models that predict disease progression in persons with PD or summarised results based on clinical outcome and methodology. Unlike systematic reviews that generally focus on a narrow clinical question, scoping reviews employ the same systematic search methods without testing a specific clinical hypothesis. 16 Scoping reviews also provide an overview of available evidence and examine how certain research was conducted, 17 providing background information for future studies. 18 Our protocol was developed in accordance with the Preferred Reporting Items for Scoping Reviews (PRISMA-ScR) guidelines ( online supplemental file 1 ) 19 and summarised below. We began developing this protocol in May 2025 and expect to complete the project by May 2026.
Development of search strategy
The search strategy was developed using an iterative approach, with two rounds of preliminary searches and refinement based on initial search results in PubMed. The initial search strategy was guided by synonyms for ‘prediction model’. The second search strategy incorporated the term ‘progression’, which removed diagnostic models from consideration. Only titles and abstracts were reviewed, and the final terms were chosen based on their ability to maximise relevant search results. Additional details were added to the inclusion and exclusion criteria ( table 1 ).
Table 1
Review eligibility criteria based on study population, concept, context and types of evidence
GBA, glucosidase, beta, acid; LRRK2, leucine-rich repeat kinase 2.
| Inclusion | Exclusion | |
| Population | Human participants diagnosed with Parkinson’s disease, including those with idiopathic and genetic forms (eg, LRRK2 or GBA mutations) | Animal models Any human participants without a formal diagnosis of Parkinson’s disease, including those in prodromal or pre-symptomatic stages Any human participants with a primary medical diagnosis other than Parkinson’s disease (eg, atypical parkinsonism, Huntington’s disease, Alzheimer’s disease, or other forms of dementia, etc.) |
| Concept | Any study that develops and/or validates a multivariable prognostic model to predict disease progression in persons with Parkinson’s disease Any prediction model that uses traditional statistical methods or machine learning and mentions any of the following: overall model performance measures (eg, R 2 , Brier score); discrimination measures, such as C‐index, area under the receiver operating characteristic curve; classification measures (eg, sensitivity, accuracy); or the terms calibration or validation in the context of prognosis | Studies that use only one predictor variable in their prognostic model Studies that develop a diagnostic model for Parkinson’s disease rather than a prognostic model Studies that do not predict a clinical health outcome in Parkinson’s disease (eg, laboratory or imaging changes that are unrelated to symptoms or prognosis) Prediction models that only include effect estimates, such as an OR or HR |
| Context | Any geographical location or setting, including online studies | None |
| Types of evidence | Full-text research articles Full-text review articles Articles published in English | Editorial articles (eg, perspective pieces or position statements) Dissertations Abstracts or posters Articles for which we cannot obtain the full text Articles not published in English |
Eligibility criteria
Eligibility criteria were established using insights from preliminary searches. Eligible studies will include full-text research articles published in peer-reviewed journals from inception through May 2025, include all multivariable models that predict disease progression in persons with diagnosed PD using traditional statistical methods or machine learning, involve human subjects and be written in English ( table 1 ). Similar to a prior review on prognostic models for predicting clinical disease progression in multiple sclerosis, we will exclude models that only report performance measures for a single prognostic factor (ie, univariable models) or only provide effect estimates (eg, OR, HR). 20 Effect estimates quantify the causal impact of specific variables on a clinical outcome, while predictive estimates forecast future clinical outcomes based on observed data.
Information sources
Based on guidance from prior literature 21 and consultation with an expert librarian, we will query four electronic databases: PubMed, EMBASE, Web of Science and Scopus. PubMed and EMBASE are major medical databases in North America and Europe and are commonly included in reviews that inform healthcare practice. 21 In contrast, Web of Science and Scopus contain articles published in international journals and are often included in reviews to increase breadth and generalisability. 22 23
Search terms
The following search terms will be used:
PubMed: (“parkinson disease”) AND ((“prognostic”) OR (“prognosis”) OR (“prediction”) OR (“predictive”) AND (“model”)) AND (“Humans”[Mesh]).
EMBASE: ‘parkinson disease’ AND (‘prognostic’ OR ‘prognosis’ OR ‘prediction’ OR ‘predictive’ AND ‘model’) AND ‘human’.
Web of Science: ‘parkinson disease’ AND (‘prognostic’ OR ‘prognosis’ OR ‘prediction’ OR ‘predictive’ AND ‘model’) AND ‘human’.
Scopus: TITLE-ABS-KEY (‘parkinson AND disease’ AND (‘prognostic OR prognosis OR prediction OR predictive AND model’) AND ‘human’) AND (LIMIT-TO (DOCTYPE, “ar”)).
Data management and selection process
This review will be conducted in Covidence, a web-based systematic review platform. Data selection will involve three stages of screening. In stage 1, all articles obtained from database searches will be imported into Covidence and duplicates will be removed. In stage 2, article title and abstract will be independently screened by two authors for relevance. Any discrepancies will be resolved by a third author and discussed with the group to refine the inclusion and exclusion criteria as needed. Any articles that do not clearly meet exclusion criteria will be automatically moved to full-text review. In stage 3, full-text articles of all records deemed relevant or potentially relevant will be independently screened by the same two authors to determine eligibility. Any discrepancies will also be resolved by a third author and discussed with the group to refine the inclusion and exclusion criteria as needed. References of review articles will be hand-searched for additional papers. The results of the search process will be reported using a PRISMA-ScR flow diagram created in Covidence.
Data charting
Authors will develop an Excel spreadsheet for data extraction to ensure consistency in data collection. We will extract the following items from eligible studies: author name, study title, year of publication, country or countries of author affiliation, study objective, design and setting, study population, predictor variables, outcome variable, results and conclusion. This extraction tool will be modified and refined as needed throughout the extraction process, with any modifications documented in the final scoping review. This Excel spreadsheet will also be submitted with the final manuscript for publication.
Data synthesis
Results from multivariable prediction models meeting inclusion criteria will be summarised using narrative synthesis and organised by clinical outcome. Outcome variables will be grouped into three overarching categories, including motor symptoms, non-motor symptoms and functional decline, and further stratified within each category by methodology (eg, traditional methods, machine learning). Models will also be evaluated using TRIPOD+AI and PROBAST guidelines to identify strengths, weaknesses and sources of bias. More specifically, we will consider mentioning the overall model performance measures (eg, R 2 , Brier score); calibration measures; discrimination measures, such as C‐index and area under the receiver operating characteristic curve; classification measures (eg, sensitivity, accuracy); or validation techniques in the context of prognosis. 5 19 Pertinent results will be presented in tabular and graphical forms. Finally, we will identify overall themes that contribute to our goal of summarising results from existing models, identifying knowledge gaps and proposing areas of future research.
Patient and public involvement
Patients and members of the public were not directly involved in the design of this scoping review protocol. However, the authors’ experience in working with persons with PD highlighted the need to systematically compile, summarise and evaluate a list of existing multivariable prognostic models in PD that can be used to inform clinical care and prognostication.
Ethics and dissemination
Because our scoping review methodology involves the secondary analysis of data from publicly available literature, it does not require ethical approval. Review findings will be published in a peer-reviewed journal and presented at relevant conferences. By systematically identifying and summarising data from existing multivariable prognostic models in PD, organising results based on clinical outcome and appraising model development and validation, this review aims to improve clinical prognostication in PD and identify areas for future research.
Supplementary material
Footnotes
Funding: This work was supported by the National Institutes of Health (1K23AG086669-01).
Prepublication history for this paper is available online. To view these files, please visit the journal online ( https://doi.org/10.1136/bmjopen-2025-107156 ).
Patient consent for publication: Not applicable.
Provenance and peer review: Not commissioned; externally peer reviewed.
Patient and public involvement: Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
References
1. Luo Y, Qiao L, Li M, et al. Global, regional, national epidemiology and trends of Parkinson’s disease from 1990 to 2021: findings from the Global Burden of Disease Study 2021. Front Aging Neurosci. 2024;16:1498756. doi: 10.3389/fnagi.2024.1498756.
2. Dorsey ER, Bloem BR. The Parkinson Pandemic-A Call to Action. JAMA Neurol. 2018;75:9–10. doi: 10.1001/jamaneurol.2017.3299.
3. Tolosa E, Garrido A, Scholz SW, et al. Challenges in the diagnosis of Parkinson’s disease. Lancet Neurol. 2021;20:385–97. doi: 10.1016/S1474-4422(21)00030-2.
4. Ben-Shlomo Y, Darweesh S, Llibre-Guerra J, et al. The epidemiology of Parkinson’s disease. The Lancet. 2024;403:283–92. doi: 10.1016/S0140-6736(23)01419-8.
5. Collins GS, Reitsma JB, Altman DG, et al. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015;350:g7594. doi: 10.1136/bmj.g7594.
6. Talbot D, Massamba VK. A descriptive review of variable selection methods in four epidemiologic journals: there is still room for improvement. Eur J Epidemiol. 2019;34:725–30. doi: 10.1007/s10654-019-00529-y.
7. Jiang Y, Yang HJ, Kim J, et al. Power of Multi-Modality Variables in Predicting Parkinson’s Disease Progression. IEEE J Biomed Health Inform. 2025;29:1343–56. doi: 10.1109/JBHI.2024.3482180.
8. Hoehn MM, Yahr MD. Parkinsonism: onset, progression and mortality. Neurology (ECronicon) 1967;17:427–42. doi: 10.1212/wnl.17.5.427.
9. Goetz CG, Tilley BC, Shaftman SR, et al. Movement Disorder Society‐sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS‐UPDRS): Scale presentation and clinimetric testing results. Mov Disord. 2008;23:2129–70. doi: 10.1002/mds.22340.
10. Chaudhuri KR, Martinez-Martin P, Schapira AHV, et al. International multicenter pilot study of the first comprehensive self-completed nonmotor symptoms questionnaire for Parkinson’s disease: the NMSQuest study. Mov Disord. 2006;21:916–23. doi: 10.1002/mds.20844.
11. Venuto CS, Potter NB, Dorsey ER, et al. A review of disease progression models of Parkinson’s disease and applications in clinical trials. Mov Disord. 2016;31:947–56. doi: 10.1002/mds.26644.
12. Li Y, McDonald-Webb M, McLernon DJ, et al. Systematic review of prognostic models in Parkinson’s disease. NPJ Parkinsons Dis. 2025;11:266. doi: 10.1038/s41531-025-01112-x.
13. Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. doi: 10.1136/bmj-2023-078378.
14. Moons KGM, Wolff RF, Riley RD, et al. PROBAST: A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies: Explanation and Elaboration. Ann Intern Med. 2019;170:W1–33. doi: 10.7326/M18-1377.
15. Damen JAA, Moons KGM, van Smeden M, et al. How to conduct a systematic review and meta-analysis of prognostic model studies. Clin Microbiol Infect. 2023;29:434–40. doi: 10.1016/j.cmi.2022.07.019.
16. Smith SA, Duncan AA. Systematic and scoping reviews: A comparison and overview. Semin Vasc Surg. 2022;35:464–9. doi: 10.1053/j.semvascsurg.2022.09.001.
17. Munn Z, Peters MDJ, Stern C, et al. Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. BMC Med Res Methodol. 2018;18:143. doi: 10.1186/s12874-018-0611-x.
18. Grant MJ, Booth A. A typology of reviews: an analysis of 14 review types and associated methodologies. Health Info Libraries J . 2009;26:91–108. doi: 10.1111/j.1471-1842.2009.00848.x.
19. Tricco AC, Lillie E, Zarin W, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med. 2018;169:467–73. doi: 10.7326/M18-0850.
20. Reeve K, On BI, Havla J, et al. Prognostic models for predicting clinical disease progression, worsening and activity in people with multiple sclerosis. Cochrane Database Syst Rev. 2023;9:CD013606. doi: 10.1002/14651858.CD013606.pub2.
21. Aromataris E, Riitano D. Constructing a Search Strategy and Searching for Evidence. Am J Nurs. 2014;114:49–56. doi: 10.1097/01.NAJ.0000446779.99522.f6.
22. Li K, Rollins J, Yan E. Web of Science use in published research and review papers 1997-2017: a selective, dynamic, cross-domain, content-based analysis. Scientometrics. 2018;115:1–20. doi: 10.1007/s11192-017-2622-5.
23. Burnham JF. Scopus database: a review. Biomed Digit Libr. 2006;3:1. doi: 10.1186/1742-5581-3-1.