Rory Miller (RRECN, New Zealand) and colleagues have published a pair of studies that highlight how rural realities can be misrepresented in hospital data — and why researchers must be cautious when interpreting administrative datasets.
In the first study, they examined over 31,000 interhospital transfers and found that in 64% of cases, the diagnosis recorded at the rural hospital differed from the one documented at the urban hospital after transfer. Rural hospitals often used less specific codes, such as “other,” partly due to limited access to diagnostics and a lack of trained clinical coders in small facilities.
Their second study looked at how hospitalisations are counted. When related hospital events (such as transfers between departments in a hospital) are treated as separate admissions, time in ED can appear artificially short, while interhospital transfers are undercounted (e.g., 5.7% vs. 12.3% in rural R2–3 areas). Distortions like these have serious implications for how rural care is measured, funded, and compared to urban systems.
These findings echo the core message of John Ioannidis, a world-renowned medical statistician, who fired a “warning shot” in a JAMA editorial at urban researchers who draw performance conclusions from rural administrative data without understanding rural context. His message is clear: data without context can mislead, particularly when urban benchmarks are applied to rural settings.
Together, these three articles serve as a timely reminder: We must be cautious when using administrative datasets in rural health research. Both data collection and interpretation are often shaped by urban systems. This can (often subconsciously) downplay the complexity and volume of rural work.
For example:
Rural hospitals may use different coding practices, especially without trained coders.
Limited access to investigations can delay or prevent diagnostic certainty.
ED Datasets are designed to monitor urban issues (such as ambulance ramping) rather than rural issues (such as time between retrieval agreement and patient departure).
Busy clinicians, not admin staff, often enter data, increasing the risk of gaps or errors.
All of this means that what gets coded might not reflect what actually happened, and rural care can be easily misrepresented in research, funding decisions, and policy.
This matters deeply for RRECN. As a rural research network, we have a role to play in building expertise around rural data in NZ and Australia and supporting colleagues to interpret it in context. It’s part of telling a more accurate story of rural health.
Read the full article on diagnosis coding
Read the article on double-counting in transfers
Read the Ioannidis editorial
First published in the RRECN Newsletter (July 2025).
Working in rural emergency care? Join the RRECN network