Methods
Theoretical framework
The Riegel model of HF self-care,and the naturalistic decision-making theoretical underpinnings thereof, served as the guiding framework for this research on HF self-care management behaviors. Specifically, our aims were guided by evidence that decision-making in response to HF symptoms is due to interactions between the patient, the problem at hand and his/her environment and experience, and that patients vary considerably in their self-care management behaviors.Given the natural variability of HF self-care management behaviors we sought to identify distinct patterns of change in these behaviors over time.
Design
This paper addresses a primary aim of a prospective cohort study on self-care management behaviors among adults with HF; details on the study design are published.In brief, community-dwelling adults with symptomatic HF (i.e. New York Heart Association (NYHA) class II–IV) under the care of a cardiologist at an outpatient clinic that specializes in evaluating patients for advanced HF therapies were recruited to participate in a longitudinal study of HF self-care management behaviors. Written informed consent was obtained from all interested participants by study staff not directly involved in patient care. The study was approved by the appropriate institutional review board; this study conforms to the principles of the Declaration of Helsinki. At enrollment, study participants completed a questionnaire comprising socio-demographic questions, physical and psychological symptoms measures, HF self-care behaviors and HRQOL. Additional assessments of symptoms, HF self-care and HRQOL occurred at three and six months following enrollment. Questionnaire completion took place during clinic visits, by phone, or by mail according to participants’ preferences. Enrollment began in July of 2011 and follow-up was completed in June of 2013. The refusal rate was 3% and our six-month attrition rate was 9.3%.
Measurement
A socio-demographic questionnaire was used to collect data on gender, ethnicity/race, marital/partnership status, education, living with another person and employment. Clinical HF characteristics, including last known left ventricular ejection fraction (LVEF) and left ventricular internal diastolic diameter (LVIDd) and prescribed therapies, were collected during an in-depth review of participants’ electronic medical records. Comorbidities were assessed during the electronic medical record review using the Charlson Comorbidity Index.NYHA functional classification was assessed by attending HF cardiologists during clinic visits on the same day as enrollment. In order to quantify changes over time in all participants, functional classification was also measured at all three time points using the Central Assessment of NYHA functional class.
Physical symptoms were measured with the 18-item Heart Failure Somatic Perception Scale (HFSPS).Six response options range from 0 (I did not have this symptom) to 5 (extremely bothersome). Scores are calculated by summing responses; higher values on the HFSPS indicate worse physical symptoms. Depression was measured with the nine-Item Patient Health Questionnaire (PHQ9).Four response options range from 0 (not at all) to 3 (nearly every day). Scores are calculated by summing responses; higher values indicate worse depression. Anxiety was measured using the Brief Symptom Inventory (BSI).Five response options range from 0 (not at all) to 4 (extremely). The anxiety score (range = 0 to 4) is calculated by adding the ratings and dividing the total by the number of items (six); higher scores indicate higher anxiety.
Heart failure-specific HRQOL was measured with the Kansas City Cardiomyopathy Questionnaire (KCCQ), a 23-item Likert scale that quantifies physical function, symptoms, social function, self-efficacy, and quality of life. Scores range from 0 to 100 with higher scores reflecting better function.Changes in the KCCQ are responsive to changes in HF over time.The KCCQ quality of life score was used as the metric of HF-specific HRQOL. A minimally important difference has not yet been established for the KCCQ quality of life scores; however, a change of five points on the KCCQ overall summary score is predictive of mortality and other clinical events in HF.We also calculated the proportion of patients who had a five-point change in the KCCQ quality of life scores to approximate a clinically meaningful change in HRQOL.
Self-care management was defined as the participant’s ability to recognize HF symptoms when they occur, implement a treatment strategy, and evaluate the effectiveness of self-initiated remedies.The Self-care of Heart Failure Index (SCHFI) was used to measure self-care management.The SCHFI includes six self-care management items. Participants are asked How quickly did you recognize it as a symptom of heart failure? with response options ranging from 0 (I did not recognize it) to 4 (very quickly), and How likely are you to try one of these remedies? with response options ranging from 1 (not likely) to 4 (very likely). Scores on the SCHFI are standardized to range from 0 to 100 with higher levels indicating better self-care management. The SCHFI also was used to measure self-care maintenance behaviors. Ten items were provided with four response options (1 = never or rarely to 4 = always or daily); responses are standardized to range from 0 to 100 with higher values indicating better self-care maintenance.
Statistical analysis
Proportions, means and standard deviations were used to describe the sample at large. Paired t-tests were used to describe sample-level changes over time. Growth mixture modeling (GMM) was used to identify naturally-occurring trajectories of change in HF self-care management over the six months of the study. GMM is a type of longitudinal clustering that handles change as random effects. Specifically, GMM was used in this analysis to identify subgroups of patients whose self-care management differed at enrollment (i.e. different intercepts) and/or had unique patterns of change over time (i.e. different slopes). In GMM, cases are assigned to the “most likely trajectory” based on conditional probabilities. Our approach to model specification was based on common procedures;the significance of the Lo–Mendell–Rubin adjusted likelihood ratio test (LMRT),entropy (close to 100%), the proportion of sample in each trajectory (≥5%), and posterior probabilities (close to 1) were used to compare alternative models (e.g. model with two trajectories versus considering the sample as a whole).GMM was completed using MPlus V7.2 (Los Angeles, CA). No formal approach has been taken for sample size considerations in GMM; but, our n-to-k ratio exceeded sample size recommendations for closely-related factor analysis.
Unadjusted differences between observed trajectories of self-care management were quantified using t-test, Fisher’s exact test or χ2 test. Growth modeling was used to compare the intercepts (i) (i.e. enrollment values) and slopes (s) (i.e. direction and strength of change over time) in HRQOL and physical and psychological symptoms between the observed trajectories of HF self-care management. Results from growth modeling are reported in estimates, standard errors and p-values for intercepts and slopes.
Socio-demographic and clinical factors that influenced the likelihood of fitting the favorable trajectory of self-care management behavior were identified using backward stepwise (p