Randomized controlled trial of Anticipatory and Preventive multidisciplinary Team Care

Design

This is a randomized controlled trial of Anticipatory and Preventive Team Care (APTCare) with new additional allied health providers compared with usual care (1:1). The random treatment allocation list was generated electronically by TrialStat Corporation and was concealed from all study personnel. Patients were randomized during a home visit by a research associate through an automated central telephone system. All care providers and patients were blind to the primary outcome measure of the study. Where more than 1 individual in a household was enrolled, all were randomized together to the same arm.

Setting

The study was conducted in a family health network with 8 family physicians, 5 nurses, and 11 administrative personnel serving 10 000 patients in a rural area near Ottawa, Ont. Physicians in the practice were remunerated by the publicly funded Medicare system through a blended payment formula of capitation (principally), fee-for-service, and incentives.

Selection of participants

Details of the study methodology are provided in a companion article.9 Briefly, patients were eligible if they were 50 years of age or older, rostered in the practice, and considered by their family physicians to be good candidates to benefit from additional medical resources and at risk of functional decline, physical deterioration, or experiencing an event requiring emergency services. There were no restrictions on diagnoses. Exclusion criteria included substantial cognitive impairment, language or cultural barriers, life expectancy less than 6 months, and plans to move or to be away for more than 6 weeks during the study period.

Sample size calculation

The trial was originally designed to detect a difference in emergency department visits between the APTCare and the usual care groups. Based on our original postulation, we required 120 patients in each arm. However, 5 months after the trial was initiated and before any outcome data were available, the primary outcome was deemed inappropriate owing to the low observed baseline rate of emergency department visits. Instead the objective was altered to examine differences in the quality of care (QOC) for chronic disease management (CDM) in 4 conditions (ie, diabetes, coronary artery disease [CAD], congestive heart failure [CHF], and chronic obstructive pulmonary disease [COPD]). To estimate the power of the study given a sample of 240 patients, several Monte Carlo simulations were repeated, each with 10 000 iterations of 3400 observations, yielding similar results. For these simulations, at random, 10% of the patients were assumed to be eligible for each of the maneuvers based on probabilities estimated by clinical experts. A Kruskal-Wallis test (a nonparametric test for estimating equality of medians)10 was then performed to test the difference between the 2 arms. It was determined that a sample size of 200 (allowing for dropouts) would provide 99% power to detect an absolute average difference of 25% between the 2 arms.

Intervention

The intervention consisted of care provided by a multi-disciplinary team. One pharmacist and 3 NPs were added to the family practice. The pharmacist worked full-time for the first 9 months and 1 NP worked full-time for 5.4 months, after which all worked half-time. Patients allocated to the intervention group were assigned to 1 of 3 NPs. The NPs provided coverage for one anothers’ patients when an NP was absent. The pharmacist and NPs delivered their care almost exclusively in the patients’ homes or by telephone. Both performed comprehensive chart reviews and home visits for each patient at the start of the study. The pharmacist then conducted a medication management review, identifying potential drug-related problems and actions required to address such issues. She then worked directly with the patients and in collaboration with the NPs and family physicians to address these and new drug-related problems as they arose. Each patient’s NP developed an individualized care plan in collaboration with the patient and in consultation with the pharmacist and the patient’s family physician. The care plan identified the patient’s active health issues and outlined the management goals that the patient and the team of providers would work toward over the course of the intervention.

Recruitment of patients took place between October 2004 and March 2005 and all patients were followed until March 31, 2006, for total intervention durations of 12 to 18 months (mean of 14.9 months in each arm). The study received approval from the Ottawa Hospital Research Ethics Board and is registered with CONSORT as NCT00238836.

Data collection

Chart audits were performed to review the quality of CDM, using 12 indicator chronic disease maneuvers and 6 indicator prevention maneuvers, and to evaluate the use of emergency services and hospitalization. Chart reviews for baseline and end-of-study QOC measures were performed by a foreign-trained physician. Instances in both arms in which a QOC maneuver was coded as not having been performed were reviewed by a nurse, and consensus was obtained between the 2 coders.

Outcome questionnaires were administered by a research associate in each patient’s home at baseline before randomization and again at the end of the study before withdrawing the services. Questionnaires captured quality of life (Short-Form 36 [SF-36] and health-related quality-of-life scales, both of which indicate better status with higher scores), instrumental activities of daily living evaluations (with higher scores showing more limitations in activity), and self-reported emergency department visits and hospitalization. The latter 2 items were also assessed at the midpoint to reduce recall bias, and a research associate, blinded to the treatment arm, integrated the information with that derived from the chart to obtain a final measure of service utilization. A caregiver burden questionnaire was left with each patient’s identified informal caregiver at the home visit following randomization and was mailed to the same individuals at the end of the study (higher scores indicated greater burden). Data entry of questionnaire results was performed by a research associate. Baseline data entry was validated using double data entry methods (error rate < 1%).

Analysis

Intention-to-treat analysis was performed. Patients who withdrew during the study were not contacted for surveys but were included in the QOC measures.

Primary outcome measure

Secondary outcome measures

Multivariate analyses

Multivariate linear regression models were performed for each continuous outcome measure, adjusted for age; sex; baseline measures of self-assessed health, estimated risk level, polypharmacy, frequent visits to the office in the previous year, multiple health conditions, emergency department visits in the previous year, education level, living accommodations, household income, main activities, having someone to confide in, and registration as a client of home care services; the presence of diabetes, CAD, COPD, or CHD; and baseline QOC scores.