Skip to content

Latest commit

 

History

History
70 lines (53 loc) · 5.31 KB

general-instructions.md

File metadata and controls

70 lines (53 loc) · 5.31 KB

Instructions supplied to all teams:

Procedures

to be followed by each harmonization working group:

Below is a list of the main steps to be applied while conducting the harmonization exercise:

  • Step 1 - Understand and select studies:
    1. Receive 5 datasets of 5 different studies.
    1. Review study description and dataset characteristic of each.
    1. Review metadata and related codebooks and questionnaires.
    1. Decide which studies to be included and explain inclusion and exclusion criteria.
    1. Decide if sub-selection of certain datasets is needed and why.
  • Step 2 - Define common variables and asses harmonization potential
    1. Decide on the definition and format of the common harmonized variables to be created.
    1. Assess the harmonization potential of each dataset to generate the common harmonized variable.
  • Step 3 - Process data into common format and ensure its quality
    1. When possible, transform data from each dataset into the common harmonized variables format and document processing methods/algorithms applied.
    1. Ensure the quality of the new harmonized variables created by reviewing processing methods and running descriptive statistics comparing raw and harmonized data.
  • Step 4 - Integrate harmonized variables and data into one common dataset.
  • Step 5 - Conduct bi-variant and multi-variant analysis to assess the main factors associated with current smoking status among older adults.
  • Step 6 - Report on all the steps listed above and present the results of the analysis. What to report?

Rationale and objectives

Integrative analysis of individual participant data across multiple studies is increasingly used to serve the needs of contemporary research. Adoption of such integrative approach is driven by different scientific rationales, including the usual requirements to obtain sufficient statistical power to investigate relatively rare events or complex interactions and permit rigorous comparison or replication across datasets. Furthermore, integrative agendas help maximizing the use of available data resources, increase cost-efficiency of certain research programs, and are sound catalysts for collaboration.

Investigators integrating and co-analyzing individual participant data across existing studies necessarily need to address questions related to data harmonization. Harmonization involves achieving or improving comparability of similar measures collected by separate studies or database. But, retrospective harmonization (harmonization of individual participant data across existing studies) raises specific scientific issues and can be very challenging especially since variability across study designs, methods and procedures used to collect data at different points in time is generally unavoidable.

Retrospective data harmonization, if not applied properly, might lead to substantial data measurement and assessment error. Obviously, the validity of the harmonized datasets depend on the use of a rigorous and transparent methodology through the initiation of the project and all the way until the dissemination of the harmonized output.

In order to assess the impact of different harmonization procedures on the analysis results, different harmonization working groups will harmonize, integrate, and analyse data from several real datasets. In the workshop, each group will present the methodology used and results from data analysis.

What to report

Each harmonization work group should provide a short document reporting on the questions below.
The document should be sent before the workshop to msaliba@maelstrom-research.org

  • I - Harmonization process
    • 1 - Which studies did you include/exclude and why?
    • 2 - Did you include all the participants of all datasets or no? Why?
    • 3 - What are the common harmonized variables generated? Their definitions? And formats?
    • 4 - Have you been able to generate the common harmonized variables from all the studies? If not, why?
  • II - Statistical analysis
    • 1- What are the frequencies and distribution of common harmonized variables for the pooled dataset and by each study-specific dataset?
    • 2 - What are the factors associated with current smoking? Bivariant analysis. Multi-variant analysis.
    • 3 - Results of the statistical analysis.

Deliverables

Each harmonization working group is required to provide:

A. Presentation (15 min):

    1. The harmonization approach used and reason.
    1. The steps applied to harmonize the data according to the
    1. The frequencies and distribution of the common harmonized variables created.
    1. The results of the analysis performed to answer the research question.

B. Documentation

    1. A written description of the harmonization approach, to be included in the paper for a scientific journal.

Studies

The preparatory stages of the workshop will rely on the data files from the following studies:

    1. Australian Longitudinal Study of Ageing (ALSA) - Wave 1
    1. Long Beach Longitudinal Study (LBLS)- Panel 2 - Wave 1
    1. Swedish Adoption/Twin Study of Aging (SATSA) - Q3
    1. The Irish Longitudinal Study on Ageing (TILDA) - Wave 1
    1. The Survey of Health, Ageing and Retirement in Europe (SHARE) - Israel - Wave 1

For more details on the data files please visit ./data/contents.md