Study design
This was a two-group, repeated measures, randomized controlled trial. The study is registered on the Clinical Trial website (NCT01964053). The study protocol was approved by the University of Nebraska Medical Center Institutional Review Board and rural hospital ethics committee. All participants gave written informed consent. Detailed information has appeared in a previously published study protocol [16].
Study setting
The study was conducted between September 2013 and October 2015 at a rural critical access hospital (CAH). To reduce the financial vulnerability of rural hospitals and improve rural residents’ healthcare access, Centers for Medicare and Medicaid Services (CMS) created a “Critical Access Hospitals (CAHs)” designation based on the 1997 Balanced Budget Act. A certified rural CAH must have less than 25 acute care inpatient beds and be located more than 35 miles from another hospital [17].
Patient inclusion and exclusion
The principal investigator and research assistants who have ethical access at the study site were responsible for identifying potential subjects, screening for eligibility and recruitment. Eligible subjects: 1) were 21 or older; 2) had HF as one of their discharge diagnoses; 3) had New York Heart Association (NYHA) class II to IV symptoms or NYHA class I symptoms and at least one other HF-related hospitalization or ED visit in the previous year; 4) were discharged to home; 5) passed a mini-cog screen;[18] 6) understood English; and 7) had access to a phone. Patients were not eligible if they had: 1) scheduled procedures/surgeries during hospitalization; 2) depressive symptoms indicated by a score ≥ 3 on the Patient Health Questionnaire-2 (PHQ-2);[19] 3) documented diagnostic evidence of liver cirrhosis; or 4) renal failure (serum creatinine greater than 2.0 mg/dl).
Intervention and usual care
Subjects randomized to the Control Group received only usual care, the standard discharge teaching for HF that includes written and verbal information about HF self-care and scheduled follow-up doctor appointments. Subjects randomized to the Intervention Group received both usual care and the 12-week PATCH intervention. The intervention was comprised of two phases: a one-on-one in-hospital SM training session and post-discharge reinforcement sessions (twice a week for the first 2 weeks, once a week for weeks 3–6, and every other week for weeks 7–12) delivered by telephone. Intervention content was presented in a variety of formats (e.g., verbal, written, visual) with interactive ability. Besides SM workbooks, each subject was provided an SM toolkit, including a calendar for weight and salt daily logging, a step-on weight scale with large and bright readings, and an electronic pill organizer reminder alarm. Each intervention session lasted about 45–50 min. Booster sessions were administered to subjects struggling with SM at home. Subjects received the tailored intervention sessions based on activation level, pre-set goals, and specific SM needs. Intervention details were reported in another publication [16].
Outcome measures
Baseline data collection occurred prior to hospital discharge and at 3 and 6 months after discharge. The primary outcomes measured at all three times were: SM adherence (self-reported frequencies of daily weighing, following a low-sodium diet, taking prescribed medications, exercising, and attending follow-up appointments), clinical biomarkers (B-type natriuretic peptide [BNP] and urine sodium/creatinine ratio [Na/Cr]), and all-cause readmissions and ED visits measured at 30, 90 and 180 days. To assess baseline SM adherence, we asked the participants to recall specific SM behaviors in the past 12 months. The healthcare utilization data were collected from both self-report and primary care provider records. In addition to self-report measures, objective measures of physical activity were obtained using an accelerometer that subjects were asked to wear for 7 consecutive days at each assessment period.
The secondary outcomes were measured via questionnaire at baseline and 3 months to test the intervention mechanisms, including SM knowledge, self-efficacy for SM, patient activation, and SM strategies. Details about measures, instruments and their psychometric properties were reported in another publications [16, 20, 21].
Randomization, blinding and allocation concealment
Given the nature of the treatment, blinding of either subject or interventionist was impossible, but the data collector was blinded to treatment assignment. The project statistician used an on-line pseudo-random number generator to create an allocation schedule; random ordering of block sizes four and six was used to maintain even accrual through the study. Group assignments were placed in sealed envelopes and opened sequentially as patients were enrolled.
Sample size and statistical analysis
The required sample size was estimated using two-sided tests and α = .10. A liberal alpha was chosen to minimize the likelihood of overlooking promising effects in this preliminary study. For a moderate effect size (Cohen’s f = .25), 41 per group provided power of .80 for the test of the mean group difference over time. It provided similar power for a z-test of the difference in group proportions of at least .25. A target sample size of 100 patients allowed for up to 20 % attrition. Further details may be found in Young et al. [16].
For the continuous outcomes in Aim 1 (e.g., physical activity outcomes and level of BNP and urine Na/Cr), linear mixed model methods were used to compare the groups across the 6-month period, adjusting for baseline levels on the respective outcome, with tests of the difference in estimated marginal means (Group effect) and whether change from 3 to 6 months differed in the groups (Group X Time effect). These methods allow for inclusion of partial cases (missing either month 3 or month 6 follow-up data) and for flexible specification of the covariance structure of repeated measurements. However, cases missing on covariates or having only covariate (baseline) measures cannot be included.
Distributions of adherence outcomes measured as number of days per week were clearly non-normal, so responses to those questions also were categorized as non-adherent (0 days), partially adherent (1–6 days), or adherent (7 days), and groups compared using χ2. For outcomes having adherence guidelines (e.g., weighing), patients also were classified as being adherent or not and group differences in the proportion of adherent patients at the end of the intervention (3 months) and at 6 months were tested using χ2.
To evaluate immediate and extended effects of the intervention on rehospitalization and ED visits (Aim 2), a χ2 test was used to compare group proportions separately at 30 days, 3, and 6 months.
To evaluate the mechanism of the patient activation intervention (Aim 3), an independent t-test was used to compare the groups on average change in intervention components from baseline to 3 months after hospital discharge. These tests were one-tailed to correspond to the hypothesis that the PATCH intervention would increase SM knowledge, self-efficacy, activation levels, and use of SM strategies.
Effect sizes were also estimated for the estimated marginal mean difference between groups. There is no established method of estimating effect sizes in linear mixed models, so standardization (d = |M1 – M2| / SD) was carried out using the baseline standard deviation of the outcome, pooled across groups. For variables having no baseline measurement, the standard deviation from the control group at 3 months was used. For tests of intervention components, the group difference in mean change was standardized using a pooled standard deviation of the change scores.