Description of the Reasons for Hospitalization of Patients With Parkinson's Disease Following Admission Through the Adult Emergency Department

The investigators performed a population-based retrospective study using the data from electronic medical records available through the data warehouse of the Greater Paris public hospitals (Entrepôt de Données de Santé (EDS) de l'AP-HP). Our study was reported according to the STROBE statement.

This study was approved and registered by the Ethics committee of the data warehouse (AP-HP CDW Scientific and Ethics Committee) under n°CSE-22-27_Emerge-Park IRB00011591. In accordance with French regulations (loi n° 2018-493 du 20 juin 2018), patients were informed that their medical records may be used for research purposes through an informational notice, and all personal identifiers were pseudonymised during the data extraction process to maintain confidentiality. No written consent was then required for this retrospective study.

The EDS currently collects data on patient's follow-up in the 39 hospitals within the AP-HP across the Île-de-France region. This warehouse contains medico-administrative data from the Medical Information System Program (PMSI), the French national hospital database. The PMSI gathers data from standardized discharge reports that document hospital visits, including diagnoses and the procedures performed in all medical units involved in the patient's care during each hospitalisation. Primary and associated diagnoses are coded using the International Classification of Diseases, 10th Revision (ICD-10). Additionally, the EDS incorporates supplementary data from multiple electronic health record systems, including drug prescriptions - stored within the ORBIS medication database and coded according to the international Anatomical Therapeutic Chemical (ATC) classification system - and medical text reports during hospital visits.

The investigators included patients aged 18 years or older, with a history of neurodegenerative parkinsonian disorders, identified through ICD-10 inpatient coding (codes 'G20', 'G21', 'G22', 'G231', 'G232', 'G233', and 'F023'), who were hospitalised in one of the 39 AP-HP hospitals via an emergency department for a minimum duration of 24 hours, between January 1, 2022, and December 31, 2023 (Fig 1). Patients who presented to the emergency department without being subsequently hospitalised were not included in the study.

To identify non-parkinsonian group patients, the investigators employed the same methodology over the same period, but without identification of an ICD-10 code relative to a parkinsonian disorder.

In a post hoc analysis given the age distribution of parkinsonian patients - with no identified patient under 50 years old and to ensure a more homogeneous population while minimizing selection bias, the investigators chose to include only individuals over 50 years old in the non-parkinsonian cohort.

The primary outcome was the principal diagnosis recorded for each hospitalisation of parkinsonian and non-parkinsonian patients. Each emergency visits that results in hospitalisation was included and considered a unit of analysis for the principal diagnosis for admission. For a single visit, a patient could be hospitalised in multiple units, with each principal diagnosis from each unit being included in the analysis.

For each patient, demographic data were collected at visit time including age, gender, length of hospital stay, number of visits per patient, and inhospital mortality. Patients were re-identified as many times as they had visits, allowing for the assessment of the frequency of emergency visits per patient as a measure of the burden of the disease.

For the analysis, all primary causes of hospitalisations were categorized into specific groups according to the motor and non-motor complications encountered in parkinsonian syndromes on existing literature and ICD-10 classifications, as follows : Neurological disorders, Falls and fractures, Infectious pneumonia, Neuropsychiatric disorders, Urogenital disorders, Constipation-related disorders, General symptoms and malnutrition, Social disability, Other infections, General comorbidities, Other causes of admission, and Sudden death.

The data collected were described using the number and percentage for categorical variables, while continuous variables were expressed as median and interquartile range (IQR).

A differential co-occurrence matrix was constructed to examine the frequency of co-occurring diagnoses between parkinsonian patients and non-parkinsonian subjects. To mitigate class imbalance, the co-occurrence matrix was transformed into percentage form, providing a pseudo-normalized representation of the associations between diagnoses. This transformation ensures that differences in diagnostic co-occurrences are not driven by sample size disparities between groups. The matrix allows for a detailed comparison of the co-occurrence patterns in parkinsonian patients themselves and also compared to non-parkinsonian patients, offering insight into the unique patterns of comorbidity in parkinsonian patients.

A radar plot was created to visualize significant features, providing an illustration that compares the extent of heterogeneity in the parkinsonian and non-parkinsonian groups.