Scientific Problem Statement
Unplanned Readmissions After Discharge to the Community
: A Burden on Individuals and the Nation
Unplanned readmissions of patients discharged to the community after medical or surgical treatment not only affect the quality of life for patients and their families but also impose significant financial burdens on the healthcare system.
According to a report by the Health Insurance Review and Assessment Service (HIRA) in Korea, the unplanned readmission rate within 30 days post-discharge in 2017 was 6.1%. This analysis excluded patients with cancer, mental illness, rehabilitation needs, obstetric cases, transfers, and deaths, suggesting that the actual rate among more severe cases could be higher.
In a study analyzing 165 patients discharged after stroke treatment from three university hospitals in Korea, 20.1% experienced unplanned readmissions within three months.
High Readmission Rates Among Home Health Care Patients Necessitate Multidimensional Analysis of Preventable Factors
Implemented in Korea in 1990, the home health care system aims to reduce hospital stays and manage patients’ health post-discharge through in-home nursing services, thereby preventing hospitalizations and readmissions. However, a study analyzing 1,790 patients who utilized home health care services at tertiary hospitals over three years found that 65.3% visited emergency rooms a total of 3,908 times, with 57.1% resulting in hospital admissions.
Reducing unplanned readmissions among home health care patients is crucial for enhancing patient and caregiver satisfaction, improving physical and mental health outcomes, and reducing medical expenses. Previous studies have attempted to identify factors influencing unplanned readmissions using structured data sources, but they often overlook unstructured data that may contain critical symptom information.
Utilizing Natural Language Processing to Analyze Home Health Care Notes for Identifying Readmission-Related Symptom Factors
To predict and prevent unplanned readmissions effectively, it is essential to analyze both structured and unstructured data to gain comprehensive insights into contributing factors. Nursing notes, a primary source of unstructured data, often contain symptom information not captured in structured electronic health records (EHRs). Studies have reported that approximately 50% of nursing information, including patient symptoms, is found exclusively in nursing notes.
Home health care notes, recorded by nurses during in-home visits, provide extensive information on patients’ clinical symptoms not included in structured EHRs. Analyzing these notes using natural language processing techniques is necessary to extract valuable insights from this unstructured data.
Need for Integrated Analysis of Physical, Socio-Demographic, and Nursing Service-Related Factors
Factors influencing readmissions include not only patients’ symptoms but also socio-economic status, social support, caregiver involvement, and education provided by nurses at discharge, such as medication counseling.
In Korea, existing research on home health care primarily focuses on service utilization statistics, with limited studies examining factors contributing to readmissions. Home health care services in Korea are typically prescribed by physicians upon hospital discharge or after outpatient visits in tertiary medical institutions. These services are tailored based on patients’ conditions and severity, determining the frequency and duration of visits. However, there is a lack of research, both domestically and internationally, on how nursing services impact patient outcomes, including readmissions.
Therefore, it is necessary to analyze physical and socio-demographic factors using structured data tailored to Korea’s context, identify clinical symptoms through home health care notes, and further investigate nursing-related factors that have been under-researched globally. By integrating these elements, we can comprehensively identify factors influencing patient readmissions and develop more accurate predictive models.