Dark Mode
US Diabetic Patient Hospitalisation Data
Patient Health Records & Digital Health
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This dataset captures a decade (1999-2008) of clinical care information from 130 hospitals and integrated delivery networks across the US. It focuses on inpatient encounters where diabetes was a primary diagnosis. Each entry details a hospital admission of a diabetic patient, with a length of stay between 1 and 14 days, and includes records where laboratory tests were conducted and medications administered. The dataset features over 50 distinct attributes related to patient demographics, hospitalisation details, medical procedures, laboratory results, and medication profiles, offering valuable insights into diabetes clinical management and patient outcomes over time.
Columns
- id: A unique patient identifier.
- encounter_id: A unique identifier for each hospital encounter.
- patient_nbr: A numerical identifier for the patient.
- race: The patient's racial background, including Caucasian, African American, and other specified categories.
- gender: The patient's sex, categorised as Female, Male, or Other.
- age: The patient's age group, presented in 10-year intervals (e.g., [70-80)).
- weight: The patient's weight, with a significant portion of missing values.
- admission_type_id: An identifier for the type of admission (e.g., emergency, urgent, elective).
- discharge_disposition_id: An identifier for the patient's disposition upon discharge.
- admission_source_id: An identifier for the source of admission.
- time_in_hospital: The duration of the patient's stay in the hospital, measured in days.
- payer_code: The code representing the patient's payer or insurance provider.
- medical_specialty: The medical specialty of the admitting physician.
- num_lab_procedures: The total number of laboratory procedures performed during the encounter.
- num_procedures: The number of non-laboratory procedures performed.
- num_medications: The total number of distinct medications administered.
- number_outpatient: The count of outpatient visits in the year preceding the hospitalisation.
- number_emergency: The count of emergency visits in the year preceding the hospitalisation.
- number_inpatient: The count of inpatient visits in the year preceding the hospitalisation.
- diag_1: The primary diagnosis recorded for the encounter.
- diag_2: The secondary diagnosis recorded.
- diag_3: The additional diagnosis recorded.
- number_diagnoses: The total number of diagnoses entered for the encounter.
- max_glu_serum: The result of the maximum glucose serum test, if performed (e.g., 'Norm', '>200', '>300', 'None').
- A1Cresult: The result of the HbA1c test, if performed (e.g., 'Norm', '>7', '>8', 'None').
- metformin, repaglinide, nateglinide, chlorpropamide, glimepiride, acetohexamide, glipizide, glyburide, tolbutamide, pioglitazone, rosiglitazone, acarbose, miglitol, troglitazone, tolazamide, examide, citoglipton, insulin, glyburide.metformin, glipizide.metformin, glimepiride.pioglitazone, metformin.rosiglitazone, metformin.pioglitazone: These 24 features indicate the status of specific diabetic medications during the encounter, with values such as 'up' (dosage increased), 'down' (dosage decreased), 'steady' (dosage unchanged), or 'no' (drug not prescribed).
- change: Indicates whether there was any change in the patient's diabetic medications during the encounter ('Ch' for changed, 'No' for no change).
- diabetesMed: A boolean indicator specifying if any diabetic medication was prescribed ('true' or 'false').
- readmitted: The outcome regarding readmission, categorised as 'NO' (not readmitted), '>30' (readmitted after 30 days), or '<30' (readmitted within 30 days).
Distribution
This dataset is provided in a CSV format, specifically
diabetes.csv, and has a file size of 19.66 MB. It contains 51 columns and approximately 102,000 records, representing individual patient encounters.Usage
This dataset is ideal for a variety of applications, including:
- Healthcare Research: Analysing trends in diabetes clinical care, patient management strategies, and outcomes.
- Predictive Modelling: Developing models to predict patient readmission rates, medication efficacy, or the impact of different treatment protocols.
- Public Health Analysis: Investigating demographic and clinical factors associated with diabetes hospitalisations in the US.
- Clinical Decision Support: Informing clinical guidelines and best practices for diabetes care based on past patient data.
Coverage
The dataset spans 10 years, from 1999 to 2008, and covers clinical care data from 130 hospitals and integrated delivery networks located across the United States. Demographically, it includes patient race (predominantly Caucasian at 75% and African American at 19%), gender (54% Female, 46% Male), and age groups, with the [70-80) age bracket being the most frequent at 26%. It is worth noting that a significant portion of 'weight', 'payer_code', and 'medical_specialty' information is not available, and 'max_glu_serum' and 'A1Cresult' frequently report 'None'.
License
Attribution 4.0 International (CC BY 4.0)
Who Can Use It
This dataset is suitable for:
- Medical Researchers: For epidemiological studies, treatment effectiveness evaluation, and understanding disease progression.
- Data Scientists & Analysts: To build machine learning models for patient risk stratification and outcome prediction.
- Public Health Professionals: To inform health policies, resource allocation, and targeted intervention programmes for diabetes.
- Healthcare Administrators: To assess hospital performance, identify areas for improvement in clinical pathways, and optimise patient flow.
Dataset Name Suggestions
- US Diabetic Patient Hospitalisation Data
- Ten-Year Diabetes Clinical Care Dataset
- Diabetes Patient Outcomes (US Hospitals, 1999-2008)
- Inpatient Diabetes Mellitus Encounters
- US Hospital Diabetes Patient Records
Attributes
Original Data Source: US Diabetic Patient Hospitalisation Data
QUALITY