Exploring family caregiving experiences in heart failure: Utilization of machine-assisted text analysis
Open Access
- Author:
- Choi, Soyoung Young
- Graduate Program:
- Nursing
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- September 30, 2020
- Committee Members:
- Lisa Ann Kitko, Dissertation Advisor/Co-Advisor
Lisa Ann Kitko, Committee Chair/Co-Chair
Judith E Hupcey, Committee Member
Barbara Ann Birriel, Committee Member
Suhang Wang, Outside Member
Lisa Ann Kitko, Program Head/Chair - Keywords:
- heart failure
caregivers
family
qualitative research
machine learning
topic modeling
sentiment analysis - Abstract:
- BACKGROUND: Heart failure affects six million adults in the United States and 26 million people globally. The prevalence of heart failure is increasing as populations age and lifestyle determinants of poor cardiovascular health rise. Although the innovation of medical treatments and devices enhanced the quality of life of persons with heart failure, the complex medical regimens bring additional challenges for patients and their family caregivers. Considering the impacts of family caregivers on heart failure progress and its potential costs for healthcare in the future, researchers, healthcare providers, and policymakers need to pay attention to the challenges and needs faced by family caregivers of persons with heart failure. PURPOSE: This study was conducted to explore the caregiving experiences of heart failure family caregivers by analyzing the serial interview documents. To improve qualitative data interpretations, manual content analysis and machine-assisted text analysis were employed. Lastly, to evaluate the utilization of machine-assisted text analysis in qualitative research, this study compared the results from the machine-assisted text analysis with the results from the manual content analysis. METHODS: The secondary data used in this study was originally collected for the grounded theory research aimed to explore the end-of-life trajectories of persons with heart failure (n = 100) and their family caregivers (n = 104), and to identify their palliative care needs and the optimal time for consulting palliative care services. Predicting data saturation in qualitative interviews, the purposively sampled interview documents were analyzed by using manual content analysis (i.e., Krippendorff’s content analysis); in addition, whole interview documents were used for machine-assisted text analysis (i.e., structural topic modeling and sentiment analysis) to secure the high performance of machine learning algorithms. For the machine-assisted text analysis of this study, the R statistical software and packages were used. RESULTS: The nine main themes from the manual content analysis were: (1) accumulating knowledge and skills for caregiving; (2) losing a sense of control; (3) balancing an unstable life; (4) constructing illness memory; (5) centering the patient in daily life; (6) accepting the loss of a family member; (7) coping with grief by drawing on social support; (8) facing financial responsibility; and (9) rethinking hospice care. The structural topic modeling estimated 20 latent topics in all of the interview documents. Most topics from the structural model were relevant to the caregiving roles (e.g., monitoring and managing blood pressure; providing comfort care; observing the symptoms of heart failure; managing diabetic pain; assisting rehabilitation activities, preparing meals). In the sentiment analysis, conflicting estimations were inferred; the lexicon-based sentiment analysis counted more negative words (n = 11,748) than positive words (n = 1,116), however, the sentence-level sentiment analysis calculated higher densities in positive sentiment scores. CONCLUSION: During the illness trajectory of heart failure, the family caregivers of this study experienced repeated episodes of family members’ acute exacerbations that require emergency medical treatment or rehospitalization. The family caregivers felt unprepared, embarrassed, and stressed when they were in unpredictable situations. Methodologically, the use of machine-assisted text analysis assisted the researcher to find the possible meanings of the interview documents by maintaining a balance in qualitative interpretation, and provided insights on distinctive linguistic features of the participants’ narratives. A hybrid approach combining manual content analysis and machine-assisted text analysis can contribute to find fruitful meanings of the given texts.