Methodological processes in validating and analysing the quality of population-based data: a case study using the...

Area of Science:

• Health Services Research
• Perinatal Health Data Quality
• Epidemiology

Background:

• Routinely collected population-based data sets are crucial for health research.
• Assessing the accuracy of these data sets is essential for reliable findings.
• The Victorian Perinatal Data Collection (VPDC) is a key source for maternity and neonatal health information.

Purpose of the Study:

• To evaluate the accuracy of data collected in the 2003 Victorian Perinatal Data Collection (VPDC).
• To determine the reliability of 111 items related to maternity and neonatal care within the VPDC.
• To identify potential limitations in the VPDC for research purposes.

Main Methods:

• A random sample of births from the 2003 VPDC was selected.
• Data from the VPDC was compared against information in original medical records.
• Accuracy metrics, including sensitivity, specificity, and predictive values, were calculated for 111 items.

Main Results:

• Seventy-nine out of 111 items were accurate in at least 97% of cases.
• Forty-five items demonstrated accuracy in 99% or more of cases.
• Five items had accuracy below 90%, with lower sensitivities indicating potential underreporting.

Conclusions:

• The VPDC demonstrates high accuracy, particularly high specificity, making it suitable for conservative analytical approaches.
• Lower sensitivities suggest that some events may be omitted from the perinatal form, potentially masking true relationships.
• Understanding discrepancies is key to optimizing the use of population-based perinatal data.