Impact Evaluation of a System-Wide Chronic Disease Management Program on Health Service Utilisation: A Propensity-Matched Cohort Study

BACKGROUND The New South Wales Health (NSW Health) Chronic Disease Management Program (CDMP) delivers interventions to adults at risk of hospitalisation for five target chronic conditions that respond well to ambulatory care: diabetes, hypertension, chronic obstructive pulmonary disease, congestive heart failure, and coronary artery disease. The intervention consists of two main components: (1) care coordination across sectors (acute, ambulatory, and community care from both public and private sectors) and clinical specialties, facilitated by program care coordinators, and (2) health coaching including management of lifestyle risk factors and medications and self-management. These components were broadly prescribed by the head office of NSW Health, which funded the program, and were implemented by regional health services (local health districts) in ways that best suited their own history, environment, workforce, and patient need. We used a propensity-matched cohort study to evaluate health service utilisation after enrolment in the CDMP. METHODS AND FINDINGS The evaluation cohort included 41,303 CDMP participants enrolled between 1 January 2011 and 31 December 2013 who experienced at least one hospital admission or emergency department (ED) presentation for a target condition in the 12 mo preceding enrolment. Potential controls were selected from patients not enrolled in the CDMP but experiencing at least one hospital admission or ED presentation over the same period. Each CDMP patient in the evaluation cohort was matched to one control using 1:1 propensity score matching. The primary outcome was avoidable hospitalisations. Secondary outcomes included avoidable readmissions, avoidable bed days, unplanned hospitalisations, unplanned readmissions, unplanned bed days, ED presentations, and all-cause death. The primary analysis consisted of 30,057 CDMP participants and 30,057 matched controls with a median follow-up of 15 mo. Of those, 25,638 (85.3%) and 25,597 (85.2%) were alive by the end of follow-up in the CDMP and control groups, respectively. Baseline characteristics (including history of health service utilisation) were well balanced between the matched groups. In both groups, utilisation peaked just before the time of enrolment/matching, declined sharply immediately following enrolment, and then continued to decrease more gradually; however, after enrolment, avoidable and unplanned health service utilisation remained higher for CDMP participants compared to controls. The adjusted yearly rate of avoidable hospital admissions was 0.57 (95% CI 0.52 to 0.62) in the CDMP group versus 0.33 (95% CI 0.31 to 0.37) in the control group (adjusted rate ratio 1.70, 95% CI 1.62 to 1.79, p < 0.001). Significant increases in service utilisation were also observed for unplanned hospitalisations (1.42, 95% CI 1.37 to 1.47, p < 0.001) and ED presentations (1.37, 95% CI 1.32 to 1.42, p < 0.001) as well as avoidable (2.00, 95% CI 1.80 to 2.22, p < 0.001) and unplanned (1.51, 95% CI 1.40 to 1.62, p < 0.001) readmissions and avoidable (1.70, 95% CI 1.59 to 1.82, p < 0.001) and unplanned (1.43, 95% CI 1.36 to 1.49, p < 0.001) bed days. No evidence of a difference was seen for all-cause death (adjusted risk ratio 0.96, 95% CI 0.96 to 1.01, p = 0.10) or non-avoidable hospitalisations (all hospitalisations minus avoidable hospitalisations; adjusted rate ratio 1.03, 95% CI 0.97 to 1.10, p = 0.26). Despite the robustness of these results to sensitivity analyses, in the absence of a randomised control group, one cannot exclude the possibility of residual or unmeasured confounding that was not controlled for by the matching process and multivariable analyses. CONCLUSIONS Participation in the CDMP was associated with an increase in avoidable hospital admissions compared to matched controls but no difference in the rate of other types of hospitalisation or death. A possible explanation is that the program identified conditions that required participants to be hospitalised. Service utilisation decreased sharply following its peak for both groups. This finding reflects the natural tendency for high-risk patients to show reductions in use following intense phases of service utilisation and highlights that, despite the additional complexity, a carefully selected control group is essential when assessing the effectiveness of interventions on hospital use.