Lessons Learned from a Prospective Meta‐analysis

Abstract

OBJECTIVE: To describe the unique aspects of and the lessons learned in planning and conducting a pooled analysis of multiple trials evaluating interventions to reduce functional decline in hospitalized older persons. Specific examples from the Hospital Outcomes Project for the Elderly (HOPE) meta-analysis are discussed. DESIGN: A prospective meta-analysis (PMA) that compiled and pooled data from concurrently conducted clinical trials testing related but distinct interventions. SETTING: The Data Coordinating Center for the prospective meta-analysis coordinated the collection and analysis of common outcome data from five university-affiliated hospitals and one community hospital conducting the clinical trials. PARTICIPANTS: Acutely ill hospitalized elderly participants at least 65 to 75 years old. INTERVENTIONS: Treatments being evaluated included exercise, physical therapy, a multidisciplinary geriatric care unit, a multidisciplinary in-hospital intervention with post-discharge care, a nursing-based geriatric care program, and a program to improve detection and evaluation of delirious patients. CONCLUSION: The prospective meta-analysis provides selected advantages over independently conducted clinical trials and retrospective meta-analyses. It does, however, pose special design and operational challenges that must be addressed well before initiation of the individual trials. Specific issues of concern include: maintaining scientific integrity of both the individual trials and the PMA; reaching consensus on PMA goals, what data to collect, how and when to collect them and how to maintain uniformly high quality data across all sites; defining the role of the Data Coordinating Center in a multicenter project that utilizes different trials and protocols; and establishing policies concerning analyses of the pooled data, publication of pooled analyses, and ownership of the pooled database.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 19 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.