You are here » CORE INFO » Methodology » Neurological Injuries » Statistics

Statistics

Used to answer our question: 1. What neuroradiological investigations are indicated to identify abusive central neurological system injury in children?

 

The aim of the study was to analyse the relative value of an MRI examination additional to the initial CT examination, and to estimate the proportion of cases in which an additional MRI would provide supplementary information to an initial CT examination.

 

Since not all children underwent both examinations, there remains the possibility that, had MRI examinations been performed (when they were not), additional information would have been provided. In these cases, a conservative assumption was used. It was assumed that such MRI examinations would have revealed the same information as the CT investigation. The proportion of cases in which an MRI examination provided additional information was computed.


 

Used to answer our question: 2. What are the distinguishing clinical features of abusive intracranial injury in children?

 

We analysed the following clinical features:

  • Apnoea
  • Retinal haemorrhages
  • Rib fractures
  • Long bone fractures
  • Bruising to the head and/or neck
  • Seizures
  • Skull fractures

The analysis was limited by the items that authors chose to report.  Further, even when a feature was commented upon by an author, not all children in each group were examined for the feature in question.  We have used an extremely conservative imputation strategy to allow for this 1; we chose to assume that any missing data (e.g. fundoscopy) in the abusive head trauma (AHT) group would have been negative, had the child been examined.  If a data item was missing from the non-abusive head trauma (nAHT) group we assumed it would have been positive, had the child been examined (e.g. rib fractures).  Only in the case of skull fractures, a feature whose presence we suspect to be associated with nAHT, was the opposite imputation performed.  We believe that this strategy may underestimate the discriminating power of an individual feature, but that this counteracts any circularity in the data collection.

These imputations being given, we then conducted a multilevel logistic regression analysis 2, allowing not only the prevalence of abuse to vary between studies, but also the odds ratios (OR) for the features in question.  By allowing the OR to differ between studies, we again aim to minimise the risk of circularity, where an individual study may have overly relied upon a particular feature in order to arrive at the diagnosis of abuse.  For each feature, we report the estimated OR for the feature in discriminating between AHT and nAHT (with a 95% confidence interval (CI)), and a positive predictive value ((PPV) the estimated probability of abuse given the presence of this feature in a child with brain injury (with a 95% CI)).

 

Click here to open

References

  1. Van Belle G. Statistical rules of thumb. New York: Wiley-Interscience; 2002 [Wiley - details of second edition - 2008]
  2. Matthews DE, Farewell VT. Using and understanding medical statistics. Basel: Karger; 2007 [Google Books - details of fourth edition]

 

^ back to top