Prediction of Outcome from Adult Bacterial Meningitis in a High HIV Seroprevelence, Resource-Poor Setting Using a New Severity Scoring Tool, the Malawi Adult Meningitis Score (MAMS)
Bacterial meningitis in sub-Saharan African adults is associated with mortality rates in excess of 50%, is predominately caused by S. pneumoniae, is strongly HIV-associated and occurs primarily in young adults. To improve this poor outcome, a severity prediction tool is required to inform both interventional trials and clinical management, aiming to optimise inter-hospital referrals, duration of antibiotics and hospital stay. However, a previously derived European severity score for bacterial meningitis was poorly predictive when applied to Malawian data.
We utilised individual predictors of death derived from a Malawian clinical trial dataset (n 400) to develop a severity prediction tool that can be applied in a high HIV seroprevelence, resource-poor setting (Malawi Adult Meningitis Score, MAMS). Five of fifteen variables tested (Cerebrospinal Fluid (CSF) culture, CSF white cell count, Haemoglobin, Glasgow Coma Score (GCS) and pulse rate) were shown to be strongly associated with poor outcome on multivariate analysis, and were converted into a predictive tool using a nomogram. The nomogram was tested against a separate clinical trial data-set for validation (n 193).
Concordance (c statistic) of the nomogram in the validation dataset between predicted and actual outcome was 0.74 (95% CI 0.65 : 0.82), agreement 62.5%, Kappa 0.6, with an estimated sensitivity of 75% and specificity of 55%.
MAMS has equivalent power to predict outcome when applied in Malawi as the European meningitis score (EMS) when applied in Europe (EMS c statistic 0.81 (95% CI 0.74 - 0.87). Comparison of the two scores may help us understand both the pathophysiological differences and the difference in efficacy of adjunctive therapies such as corticosteroids between these two populations. MAMS has the potential to be a useful clinical and research tool.
M. Scarborough, None
K. M. B. Ajdukiewicz, None
K. Cartwright, None
B. Faragher, None
D. Lalloo, None
R. Heyderman, None
See more of: Poster Abstract Session