Integrated Disease Management of Heart Failure in Primary Care; A Cluster Randomized Trial
Integrated Disease Management of Heart Failure in Primary Care; A Cluster Randomized Trial
The study population includes primary care physicians and heart failure (HF) patients attending one of over 100 family physicians in seven family health teams in Southwestern Ontario. Study purpose is to measure the effect of an integrated disease management (IDM) program for people diagnosed with HF and receiving treatment at a primary care facility. Components of IDM include HF specific patient education and self care management skills training by a heart failure educator. Study outcomes include health service use, HF symptoms, quality of life, and HF knowledge assessment compared to the usual care group.
The primary objective of this study is composite and will measure the effect of integrated disease management (IDM) on all cause hospitalizations, ED visits and mortality events. Secondary outcomes will include HF related hospitalizations, HF related ED visits, quality of life, mortality, other health service utilization, acute HF episodes, NYHA class. We hypothesize HF specific IDM implemented in primary care will be superior to usual physician-based care measured by a combined reduction in the total number of all cause hospitalizations and ED visit events.
The study population will be identified through patients attending one of 100 family physicians from 10 different family health teams (FHTs) or family health organizations (FHOs) in the Southwestern Ontario.
Study Design: A parallel cluster randomized trial design has been chosen comparing the intervention arm (patients entered on the IDM) to the control arm (patients receiving usual care). A multi-level study design is proposed, level 3 the FHT/FHO, level 2 the physician and level 1 the HF patient. We have chosen to randomize at level 2, the physician and implement the intervention at both level 2 and 1, the physician and the individual. Outcomes will be measured at individual level. Stratified randomization of physicians will be performed by FHT/FHO, giving greater balance between arms and increased power and precision by reduction of between cluster variability.
Recruitment: Physicians from the FHTs/FHOs will be invited to participate and informed consent will be obtained. The physician will be randomized to either the control or intervention group and randomization will be computer generated by FHT/FHO strata. Allocation for overall study will be 1:1 as will allocation by FHT/FHO. Each participating primary care site will identify all individuals with a HF diagnosis in their care suitable for the trial and a simple random sample will be taken from this group to obtain the desired cluster size. An initial telephone call will be arranged with the patient to discuss study details, obtain informed consent, further determine eligibility, and complete questionnaires.
Data management: As a part of the objectives of this study a POSS electronic tool has been developed, all data collected about the participants will be entered by heart failure educators and stored in a central server. Access is restricted to authorized personnel only. The POSS has been designed not only as a secure storage depot but also as a tool to standardize the data collected minimizing information bias. There is extensive data checking at the time of data entry. Data definitions are incorporated to support quality data inputs.
Sample Size: With a minimum recruitment of 50 physicians recruited and 4 participants per physician, this study would be powered to detect a minimum 36% reduction in the rate of number of hospitalizations or ED visits per person year with an attrition rate of 20%. This calculation is based on 80% power to 5% significance with an ICC of 0.05.
If 100 physicians are recruited with 2 to 3 participants per physician (and a total sample size of 280) the study will have 90% power to detect a 35% reduction in the primary outcome.
Statistical Analysis: Analysis will be on an intention to treat basis. Baseline data will be used to characterize the study population, to identify any imbalances between arms. Continuous data will be displayed as mean ±standard deviation and count (percent) for categorical variables (variables to be presented will be predetermined in an analysis plan). Due to over dispersion that occurs in this type of count data the primary outcome (and health service utilization secondary outcomes) will be analyzed using a negative binomial distribution with random effects to account for clustering and for individuals experiencing multiple events. The results will be presented as rate ratios. The secondary outcomes (change in KCCQ at 6 months and other knowledge and QoL metrics) will be analyzed at individual level using logistic regression, results will be presented as odds ratios. Reliability will be assessed by using a quadrature check and in the event of failure a generalizing estimating equation (GEE) model will be fitted.
Inclusion Criteria:
Exclusion Criteria: