G131 How accurate is child death overview panel data? Comparison with a cohort study dataset and cause of death classification

Abstract

Aims Child death overview panels (CDOPs) perform statutory review of all child deaths. Multi-professionals report, discuss and analyse information. Cause of death (COD) is classified by the panel using a standard hierarchical categorisation system. We sought to assess whether CDOP infant data was comparable to that collected for research purposes, by a large, local birth cohort study. We also compared CDOP COD categorisation with another classification system. Methods Cohort study infant deaths were identified and matched to pseudo-anonymised CDOP identification via NHS number. Cohort study interviewer-administered questionnaire data was accessed, and kappa values calculated, to assess agreement of variables from the two data sources. Two investigators (blinded to CDOP categorisation) processed registered COD information using CODAC (Cause of Death and Associated Conditions) Version 2. Discrepancies were discussed and resolved. Proportions of COD classification categories were compared. Results 56 cohort study infants had died and been reviewed by CDOP. Kappa values were ‘almost perfect’ (0.81–1.00) for gender, gestation, birth weight, consanguinity, deprivation and smoking in pregnancy, but ‘moderate’ (0.58) for ethnicity. For the 56 deaths, CDOP categorised COD as: 48.2% ‘Chromosomal, Genetic and Congenital Anomalies,’ 30.4% ‘Perinatal/Neonatal events,’ 10.7% ‘Infection,’ 7.1% ‘Sudden, Unexpected, Unexplained Deaths’ and 3.6% ‘Trauma and Other External Factors.’ CODAC classified COD as: 48.2% ‘Congenital Anomaly,’ 17.9% ‘Neonatal,’ 17.9% ‘Infection,’ 7.1% ‘Unknown,’3.6% ‘Intrapartum,’ 3.6% ‘Maternal’ and 1.8% ‘Placental.’ Conclusions High levels of agreement suggest CDOP information is valid, despite not being collected for research purposes. Lower levels of agreement in ethnicity are likely to reflect comparison of maternal ethnicity (cohort study gold standard) with infant’s ethnicity. Gestation at recruitment (26–28 weeks) and potential for cohort studies to under-recruit hard to reach populations may mean infants most at risk of death were excluded from the cohort. COD classification was identical for congenital anomaly (known to be high in this district) but other categories were discrepant, or different. CODAC-V2 was chosen for comparison due to its use by MBRRACE (Mothers and Babies: Reducing Risk through audits and confidential enquiries). Designed for perinatal deaths, there is more detail in classification of such, largely accounting for the described COD differences. We suggest CDOP data appears an accurate and valuable source for further examining this district’s infant mortality.