The accuracy of administrative coding: a population-based validation study of aneurysmal subarachnoid haemorrhages, lessons for neuroscience nurses
Abstract
Background
Large-scale administrative data collection is more efficient than single centre, prospective designs for rare events such as aneurysmal subarachnoid haemorrhage (aSAH). However, there are limited studies reporting data validation of International Classification of Diseases (ICD) search algorithms for aSAH. Equally there are limited examples of utilising coding for neuroscience nursing research. Despite this neuroscience nurses play a critical and functional role in contributing to ICD coding. Nurses are the primary source of clinical data entry, they are at the point of care and complete critical pieces of documentation including admission, transfer and discharge forms as well as daily progress notes and forms. ICD coding is critical in the provision of a standardised and uniform language for classifying diseases, injuries and symptoms. Coding enables accurate health data tracking, comparative research between hospitals, institutions and regions and surveillance of diseases and mortality.
Methods
A population-based search of non-traumatic subarachnoid haemorrhages cases was undertaken from 2010 to 2014. The research was undertaken to fulfil the data collection for a PhD thesis. Using sequential combinations of the subarachnoid haemorrhage ICD-10 codes I60.0–I60.9, sensitivity, specificity and positive predictive values (PPV) were calculated on a three-stage basis. Individual access to discharge summary documentation, digital medical records (Gold standard) and data linkage with death registry records and ambulance records was used to validate aSAH cases using the three ICD-10 code algorithms.
Results
A total of 414 events were identified, including 282 non-traumatic subarachnoid haemorrhage admissions of which 172 (60.99%) were subsequently confirmed as aSAH. Medical record review of cases within ICD codes I60.0–160.9 resulted in sensitivity and specificity (95% CI) of 0.90 (0.86–0.94) and 0.23 (0.16–0.31) and a PPV of 65.1%. When analysed on a population basis the sensitivity and specificity were 0.74 (0.68–0.79) and 0.25 (0.18–0.33) and a PPV of 67.4% when using ‘160.0–160.9’. When analysing specific aneurysm morphology, the sensitivity and specificity was 0.52 (0.45–0.58) and 0.91 (0.85–0.95) with a PPV of 90.5%.
Conclusion
Applying broad administrative database search algorithms to uncommon diseases such as aSAH results in a number of inaccuracies, particularly an overestimation of results when considering the results as representative of a population or when attempting to describe specific aneurysm morphology. For nurses this highlights the importance of ensuring accuracy and completeness when documenting patient files. Whilst nurses do not generally assign ICD-10 codes, they play a vital role in ensuring that there is the necessary clinical information including (but not limited to) flowsheets, transcription of diagnostic details, narratives and care plans. This process of documentation is not only a legal requirement and part of nursing daily practice, but it plays an important role in billing, insurance claims, health service research, resource allocation, policy making and quality development as well as coding.
© 2026 Linda Nichols, published by Australasian Neuroscience Nurses Association
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.