Derivation and validation of a novel risk assessment tool to identify children aged 2-59 months at risk of hospitalised pneumonia-related mortality in 20 countries.

INTRODUCTION: Existing risk assessment tools to identify children at risk of hospitalised pneumonia-related mortality have shown suboptimal discriminatory value during external validation. Our objective was to derive and validate a novel risk assessment tool to identify children aged 2-59 months at risk of hospitalised pneumonia-related mortality across various settings. METHODS: We used primary, baseline, patient-level data from 11 studies, including children evaluated for pneumonia in 20 low-income and middle-income countries. Patients with complete data were included in a logistic regression model to assess the association of candidate variables with the outcome hospitalised pneumonia-related mortality. Adjusted log coefficients were calculated for each candidate variable and assigned weighted points to derive the Pneumonia Research Partnership to Assess WHO Recommendations (PREPARE) risk assessment tool. We used bootstrapped selection with 200 repetitions to internally validate the PREPARE risk assessment tool. RESULTS: A total of 27 388 children were included in the analysis (mean age 14.0 months, pneumonia-related case fatality ratio 3.1%). The PREPARE risk assessment tool included patient age, sex, weight-for-age z-score, body temperature, respiratory rate, unconsciousness or decreased level of consciousness, convulsions, cyanosis and hypoxaemia at baseline. The PREPARE risk assessment tool had good discriminatory value when internally validated (area under the curve 0.83, 95% CI 0.81 to 0.84). CONCLUSIONS: The PREPARE risk assessment tool had good discriminatory ability for identifying children at risk of hospitalised pneumonia-related mortality in a large, geographically diverse dataset. After external validation, this tool may be implemented in various settings to identify children at risk of hospitalised pneumonia-related mortality.

Indoor air quality of low and middle income urban households in Durban, South Africa.

INTRODUCTION: Elevated levels of indoor air pollutants may cause cardiopulmonary disease such as lower respiratory infection, chronic obstructive lung disease and lung cancer, but the association with tuberculosis (TB) is unclear. So far the risk estimates of TB infection or/and disease due to indoor air pollution (IAP) exposure are based on self-reported exposures rather than direct measurements of IAP, and these exposures have not been validated. OBJECTIVE: The aim of this paper was to characterize and develop predictive models for concentrations of three air pollutants (PM10, NO2 and SO2) in homes of children participating in a childhood TB study. METHODS: Children younger than 15 years living within the eThekwini Municipality in South Africa were recruited for a childhood TB case control study. The homes of these children (n=246) were assessed using a walkthrough checklist, and in 114 of them monitoring of three indoor pollutants was also performed (sampling period: 24h for PM10, and 2-3 weeks for NO2 and SO2). Linear regression models were used to predict PM10 and NO2 concentrations from household characteristics, and these models were validated using leave out one cross validation (LOOCV). SO2 concentrations were not modeled as concentrations were very low. RESULTS: Mean indoor concentrations of PM10 (n=105), NO2 (n=82) and SO2 (n=82) were 64µg/m3 (range 6.6-241); 19µg/m3 (range 4.5-55) and 0.6µg/m3 (range 0.005-3.4) respectively with the distributions for all three pollutants being skewed to the right. Spearman correlations showed weak positive correlations between the three pollutants. The largest contributors to the PM10 predictive model were type of housing structure (formal or informal), number of smokers in the household, and type of primary fuel used in the household. The NO2 predictive model was influenced mostly by the primary fuel type and by distance from the major roadway. The coefficients of determination (R2) for the models were 0.41 for PM10 and 0.31 for NO2. Spearman correlations were significant between measured vs. predicted PM10 and NO2 with coefficients of 0.66 and 0.55 respectively. CONCLUSION: Indoor PM10 levels were relatively high in these households. Both PM10 and NO2 can be modeled with a reasonable validity and these predictive models can decrease the necessary number of direct measurements that are expensive and time consuming.

Provision of critical care services to HIV-infected children in an era of advanced intensive care and availability of combined antiretroviral therapy.

Intensive care facilities are always in demand in the public sector and there is constant competition for beds. Appropriate allocation of children to these resources is based on the ethical principles of distributive justice and beneficence that is determined on the presumed short-term outcome of the acute illness, long-term outcome of the underlying chronic disease and the overall demand for these facilities. At the onset of the HIV epidemic in South Africa, HIV-infected children were refused admission to the paediatric intensive care unit (PICU) on the basis of poor ICU outcomes and the lack of provision of combined antiretroviral therapy (cART) for survivors. The recent significant improvement in outcome in these patients through early recognition and treatment of HIV-related opportunistic infections, the provision of advanced organ support and the routine availability of cART suggests that the previous policy requires review. Ethical principles, the Paediatric Index of Mortality Score for each request, the quality and disability-adjusted life years and cost-effectiveness of care are all important considerations in deciding which patients should be allowed access to these limited and expensive resources. With the improved long-term outcome in HIV-infected children on cART, admission of these cases to a PICU should now be based on the prognosis of the acute illness, as with any other chronic disease such as asthma or diabetes. Withholding and withdrawing advanced life support should accord with standard protocols applied to any condition for which a child is admitted to the PICU.

In silico children and the glass mouse model: clinical trial simulations to identify and individualize optimal isoniazid doses in children with tuberculosis.

Children with tuberculosis present with high rates of disseminated disease and tuberculous (TB) meningitis due to poor cell-mediated immunity. Recommended isoniazid doses vary from 5 mg/kg/day to 15 mg/kg/day. Antimicrobial pharmacokinetic/pharmacodynamic studies have demonstrated that the ratio of the 0- to 24-h area under the concentration-time curve (AUC(0-24)) to the MIC best explains isoniazid microbial kill. The AUC(0-24)/MIC ratio associated with 80% of maximal kill (80% effective concentration [EC(80)]), considered the optimal effect, is 287.2. Given the pharmacokinetics of isoniazid encountered in children 10 years old or younger, with infants as a special group, and given the differences in penetration of isoniazid into phagocytic cells, epithelial lining fluid, and subarachnoid space during TB meningitis, we performed 10,000 patient Monte Carlo simulations to determine how well isoniazid doses of between 2.5 and 40 mg/kg/day would achieve or exceed the EC(80). None of the doses examined achieved the EC(80) in ≥90% of children. Doses of 5 mg/kg were universally inferior; doses of 10 to 15 mg/kg/day were adequate only under the very limited circumstances of children who were slow acetylators and had disease limited to pneumonia. Each of the three disease syndromes, acetylation phenotype, and being an infant required different doses to achieve adequate AUC(0-24)/MIC exposures in an acceptable proportion of children. We conclude that current recommended doses for children are likely suboptimal and that isoniazid doses in children are best individualized based on disease process, age, and acetylation status.