Innovative Analysis

poverty score

I can assist in getting most out of the data by using  innovative analytical techniques.   This is critical to fully exploit the collected information to provide policy directions. This includes the estimation of threshold values for epidemics, the transformation of households' variables into socioeconomic indices to assess health inequalities, and the identification of effectiveness factors from service data to model cost-effective management strategies.

Examples of innovative analyses includes the following:

(a) Testing the 30 cluster sampling schemes. This cluster sampling method is frequently used in household surveys to estimate coverage for specific indicators such as child immunization rates.  A computer simulation was carried out on the data that was collected on all the households of 30 villages of the Ugandan district of Mbarara.  The simulation tested the hypothesis that the 30 cluster sampling produced unbiased estimates compared with systematic random sampling (SRS). The simulation identified the indicators for which the 30 cluster methodology produced as good as estimates as the SRS.

(b) Reducing redundant information. Most household surveys collect tens of variables that are not always efficiently used and this analysis showed how to transform categorical variables into a few dimensions with their own metric. In this household survey, Multiple Correspondence Analysis was used to transform tens of variables into two factors representing 80% of the total variability. These two dimensions had a metric that could be used to identify the poorest households to assess health inequalities.

(c) Building proxy means test to identify poor households in South Africa and Ethiopia is critical to target health inequalities. These two analysis used two Living Standards and Development Surveys of the World Bank to produce a screening tool to identify poor households.    Several analytical methods were used to assign weighted score to household variables so that the total score could provide a measure of household deprivation.  The scoring system was validated against percentiles of expenditures and household malnutrition.    The results of the South African analysis were used as an example of best practices for this type of analysis by "Living Standard Analytics: Development Through the Lens of Household Survey Data".    As poverty is a major cause of health inequalities, a screening tool to identify poor households is useful to target health inqualities.

(e) Measuring Social Capital (SC) in Kampala, Uganda.   SC captures several aspects related to social interaction and social cohesion, and thus it is likely to influence health status.   The SC of individuals has been described through political activism, membership of social networks, intensity of social contacts and other categorical variables.   I applied Non Linear Principal Component and Cluster Analysis to the data from a household survey that was carried out in Kampala.  The result was to transformed the tens of categorical variables into a few indices of SC with their own metric.    As in the case of poverty, also SC might be one cause for health inequalities.  Producing better measurements to identify households with low levels of SC is likely to be useful to target health inequalities.

(f) Published data was analysed to measure the effect of active and passive iodine prophylaxis in Italy. Goitre has been declining in Italy since the 1970s and because active prophylaxis (AP) has been very limited.  It has been suggested that in Italy the decline in goitre prevalence was influenced by the silent prophylaxis (SP). SP is related to the natural increase in iodine intake because of higher consumption of iodine-rich products associated with socioeconomic development. The hypothesis tested in the present study is that SP has increased iodine intake in Italy with subsequent reduction of goitre and that such changes can be quantified. The analysis was based on surveys carried out between the 1970s and the 1990s where goitre and urinary iodine, a proxy of consumption, were measured in schoolchildren. The contribution of the SP was quantified in an annual increase in urinary iodine excretion between 2.1 and 4 mg/l, and an annual decline in goitre prevalence between 2.1 and 3.6%. In the few areas with AP, there was an annual increase in urinary iodine between 6.5 and 13.1 mg/l, while the average annual decline of goitre was between 4.4 and 10 %.   The results are important for the following reasons: i) notwithstanding the natural decline of goitre, AP is still required because it is much faster than SP to reduce iodine deficiency; and ii) any cost-effectiveness analysis of iodine prophylaxis in developed countries should be estimated at the net of the contribution already provided by the SP.