Exploring discrepancies in clinical coding between rural and urban hospitals in Aotearoa New Zealand in patients who underwent interhospital transfer

Early Abstract:

Introduction: The agreement of clinical coding within the same health encounter between NZ rural and urban hospitals is unknown, and data from comparable international health systems is scarce, dated, or inconclusive. There is a reliance upon administrative datasets that store clinically coded information to complete numerous rural-urban health analyses, which inform health policy and resource allocation decisions. Anecdotally, clinical coding is often performed by clinicians or reception staff without formal coding training in many NZ rural hospitals, which would usually be completed by formally trained clinical coders in urban NZ hospitals. This study aimed to determine whether discrepancies existed between the primary diagnosis codes (PDx) assigned in the National Minimum Dataset of hospital discharges (NMDS) by NZ’s publicly funded hospitals, for patients who underwent an inter-hospital transfer from a rural to an urban hospital.
Methods: A retrospective observational study using the NMDS.NZ’s publicly funded hospitals were classified into three categories: ‘Rural hospitals’, ‘Hospitals in small urban centres’, and ‘Hospitals in large urban centres.’ Inter-hospital transfers were identified by bundling events in the NMDS into healthcare encounters. The PDx assigned at discharge from the rural hospital were compared against the codes assigned at discharge from the urban hospital, and corresponding diagnosis groups based on the World Health Organisation (WHO) chapter definitions were assigned to each code. The number and percentage, with 95% confidence intervals, of encounters where the PDx from the rural and urban hospitals were discordant were calculated.
Results: 31,691patients, from 54 publicly funded hospitals,  were included who underwent an inter-hospital transfer from an NZ rural to an urban hospital between 1st January 2015 and 31st December 2019. There were discrepancies in 64.1% (95% CI 63.5% to 64.6%) of the PDx assigned between the rural and urban hospitals, and in 32.1% (95% CI 31.6% to 32.6%) of broader diagnosis groups. In both cases, higher discrepancies existed for transfers to hospitals in small urban centres compared to hospitals in large urban centres. The most frequently assigned diagnosis group at discharge from rural hospitals fell under the non-specific group ‘Other’, constituting 24.4% of all diagnosis groups assigned by a rural hospital. For 4.8% of all healthcare encounters, a specific diagnosis group assigned on discharge from the rural hospital was subsequently changed to ‘Other’ at the urban transfer hospital. This reassignment to ‘Other’ following inter-hospital transfer occurred within every diagnosis group assigned at a rural hospital.
Conclusion: Two-thirds of primary diagnosis codes and one-third of diagnosis groups were discordant after transfer from rural to urban hospitals in NZ.  Further investigation is needed into why these discrepancies are occurring.
Keywords: Aotearoa New Zealand, clinical coding, ICD-10-AM, inter-hospital transfers, National Minimum Dataset, rural healthcare, rural hospitals.