Aiming for a representative sample: Simulating random versus purposive strategies for hospital selection

Area of Science:

• Healthcare research methodology
• Health services research
• Biostatistics

Background:

• Selecting a representative sample is crucial in research, but random sampling is not always feasible or optimal.
• Estimating blood use across Dutch hospitals required selecting 12 out of 89 institutions.
• This study evaluates random and purposive sampling for estimating blood use in Dutch hospitals.

Purpose of the Study:

• To compare the effectiveness of different sampling strategies for estimating blood use in Dutch hospitals.
• To determine the optimal sampling method for resource allocation and planning in healthcare.
• To provide evidence for making informed decisions on sampling strategies in similar research.

Main Methods:

• Simulated five sampling strategies: largest hospitals, maximum variation, random, regional variation, and two-region sampling.
• Used population-wide data on hospital blood use and beds for simulations.
• Compared prediction errors of each strategy against actual population values to assess quality.

Main Results:

• Maximum variation sampling yielded the lowest prediction error (15%), outperforming random sampling (19%).
• Sampling only the largest hospitals resulted in the poorest performance.
• Increasing sample size or adjusting it did not significantly alter the ranking of strategies.

Conclusions:

• Maximum variation sampling is the optimal strategy for estimating blood use in this context.
• Simulations allow for the evaluation of sampling strategies prior to study commencement when proxy data exist.
• Findings guide researchers in selecting appropriate sampling methods for improved study outcomes.