Effects of an interprofessional care concept in nursing homes evaluated in the SaarPHIR project: A cluster-randomized controlled trial

Abstract

Introduction: Deficits in interprofessional collaboration can lead to insufficient medical care for nursing home residents, particularly inappropriate hospitalizations. Transfers are stressful for residents, and hospital stays can lead to infections and functional decline. Increasing the role of general practitioners and improving collaboration between professionals may reduce hospitalizations. In an effort to reduce hospitalizations and improve quality of care for nursing home residents, the SaarPHIR project implemented and evaluated a complex intervention which aimed at improving cooperation between general practitioners and nurses. This paper evaluates the effectiveness of an interprofessional care concept in nursing homes.

Methods: A prospective, cluster-randomized controlled trial was conducted in Saarland, Germany, from May 2019 until July 2020 with a 15-months of follow-up, with two parallel groups and a 1:1 randomization at district level to evaluate the effectiveness of the intervention. The six administrative districts of the German federal state of Saarland were selected as randomization clusters to avoid spillover effects. The primary outcome, hospitalization, was assessed using claims data from six health insurers. Analyses were performed using generalized linear mixed models assuming both a Poisson and, for sensitivity analyses, a negative binomial distribution allowing for clustering at the nursing home level. Considering the randomized cluster level in the primary analysis would be the proper approach. However, after careful consideration, an unconventional approach was adopted to ensure the evaluation of the intervention within the complex healthcare system with a pragmatic design. The randomized cluster level was considered in sensitivity analyses. Secondary outcomes included ambulatory care-sensitive and nursing home care-sensitive admissions, mortality and hospital days. Furthermore, health economic aspects were explored by comparing costs between groups descriptively and exploratively using a generalized linear mixed model with a log-link and a gamma distribution.

Results: Twenty-eight nursing homes received the intervention (1,053 residents), and 16 nursing homes (680 residents) were assigned to usual care. Hospitalization rates did not differ significantly between groups (incidence rate ratio [IRR] = 0.94; 95% CI: 0.78-1.14). Nursing home care-sensitive admissions could be reduced in residents treated with the interprofessional care concept (IRR: 0.73, 95% CI: 0.59-0.96). No differences in mortality, number of days spent in hospital and healthcare costs were found between groups. Mean drug costs (€82.53; 95% CI: 11.79-165.06) were higher and costs for ambulatory hospital stays lower (-€40.80; 95% CI: -76.50-0.00) in the intervention group.

Conclusion: All-cause hospitalization was not significantly affected in the relatively short duration of the intervention. Nevertheless, secondary outcomes suggest some positive effects for the intervention group. However, participation in the intervention group was lower than expected at both the nursing home and resident levels, limiting the validity of the results.

Fig 1. Flowchart of the claims data provided and in-/exclusion of insured residents in the closed analyses populations.

Abbreviations: cRCT = Cluster-randomized controlled trial, LOI = Letter of interest, NH-PP-CC = Nursing home per protocol closed cohort (subgroup of NH that were actively participating in the intervention group according to the study protocol, while the control group participants remained the same), R-PP-CC = Residents per protocol closed cohort (subgroup of NH-PP-CC including intervention group residents only if they received project-specific services).

Fig 2. Boxplot of observed (crude) hospitalizations per person-year in the cRCT phase per study group.

Abbreviations: CG = Control group, IG = Intervention group.

Fig 3. Estimated base hospitalization rates per person-year (random intercept) including 95% confidence interval per nursing home and the estimated mean base hospitalization rate per person-year (red dashed line, fixed intercept) based on the primary analysis model.