Internal Medicine Patient Records with Somatic and Mental Comorbidities

Loaded with 28,536 internal medicine inpatient cases from a German university hospital between 2017 and 2019. It was used to build department-specific machine learning models for predicting prolonged length-of-stay, incorporating features derived from demographics and mental and somatic comorbidities.

Use Cases

• Train department-specific classification models (e.g., Random Forest) to predict prolonged length-of-stay using features like the number of somatic comorbidities and age at diagnosis.
• Analyze the influence of specific mental comorbidity subgroups, such as stress-related adjustment disorder, on hospital length-of-stay predictions.
• Compare predictive performance (AUROC) of models across six internal medicine departments using department-wise computed prolonged stay thresholds.
• Identify the most influential somatic and mental comorbidity subgroups for length-of-stay prediction through feature importance analysis.

Strengths

• Contains 28,536 patient cases, providing a substantial sample for analysis.
• Data spans three years (2017-2019) from a maximum care university hospital.
• Explicitly incorporates both mental and somatic comorbidities, a feature noted as rare in existing studies.
• Study provides concrete, department-specific thresholds for defining prolonged length-of-stay (e.g., 9 days for Cardiology, 26 for Hematology).

Limitations

• Data is from a single hospital in Germany, limiting geographic generalizability.
• The dataset is static, covering a period ending in 2019, which may not reflect current hospital practices or disease patterns.
• Specific column definitions and raw data structure are not provided in the input.

Provenance

Source

German University Hospital in Heidelberg

Collection Method

Comprised inpatient cases treated in internal medicine departments.

Time Range

2017 to 2019

Freshness

Last updated metadata indicates 2025-10-14, but underlying data covers 2017-2019.

Geography

Heidelberg, Germany