• No se han encontrado resultados

INVENTARIO DE LAS RUTAS DEL TRANSPORTE .1 Estado de la red vial del transporte

In document INFORME DE DIAGNÓSTICO (página 73-78)

This section analyses in detail the rationale for the aggregation of individual socioeconomic characteristics into a single variable and its construction. This new variable, a socioeconomic deprivation index, was decided to be the single adulthood socioeconomic indicator to be added in multivariable statistical modelling as a confounder factor. Such procedure was not adopted for parental education, a proxy for childhood socioeconomic characteristics, because its influence in health outcomes occurs at a different point of the life-course. Therefore, parental education was also decided to be kept as a separate variable to control for in further analyses.

5.4.2.1. Rationale

It is well recognised that different socioeconomic indicators —income, wealth, educational attainment and occupational group— are all related to and help explain people’s health status and that social circumstances across the life-course influence people’s health and well-being [149]. In addition, educational attainment, as an indicator or socioeconomic position, is primarily related to health through the advantages it gives people in their later socioeconomic trajectories, not simply because education encourages healthy behaviours [149].

Following on the evidence from literature, it is evident that the information collected related to socioeconomic characteristics is important to be taken into account in the evaluation of any association of interest. In fact, such information was collected, as decided a priori and confirmed in the section, due to the confounding roles it may exert on the associations of interest of this study.

The pattern of the various adulthood socioeconomic-related variables in this study is clear —the rural group tend to have a consistent pattern of lower socioeconomic position, either by showing lower rates of education attained, lower income, low assets index or higher rates of overcrowding. The opposite is observed for the urban group. Such disparity places considerable challenges on the uses of data for multivariable modelling in this study.

-160-

Ideally, as recommended by the literature [139-141, 149], all indicators should be considered separately in the statistical multivariable modelling because of their independent contribution to health outcomes. Such possibility was explored by ruling-out colinearity between different individual socioeconomic variables and results suggest that all of them could be used in statistical models to be built (data not shown).

The main challenge, however, was that, as shown already in Table V-13 and Table V-14, various indicators had very low numbers of cases in the extreme cells, e.g.

only one urban individual earns less than USD $50 dollars and only one rural individual reported to earn more than USD $450 dollars. In the same vein, 97% of rural individuals fall in the lowest tertile for possessions weighted asset index. These observations limit the spread of sufficient number of individuals in each cell or category for each variable, an important requirement to be able to run multivariable statistical models. Such concentration of characteristics into certain cells would have resulted in statistical models yielding wide confidence intervals. Even after shortening the aggregation of variables into smaller number of categories, still wide confidence intervals were observed.

The study had rich information on various proxies for socioeconomic position, but because of the patterns of the data, such richness could not be exploited to a maximum as separate variables in the statistical modelling. Thus, in terms of how to manage socioeconomic variables at the statistical modelling stage of the study, two scenarios were considered: either to choose a single indicator or to find out a reasonable way to aggregate all the available information.

It was considered an advantage of the present study to have such richness of data related to socioeconomic position. The option of selecting only one of the socioeconomic proxies would sacrifice most of the data gathered. Such alternative was discarded because, in addition to the “waste” of data collected due to non-usage, none of the indicators has been ascribed as the “best” or “gold standard” for measuring socioeconomic status.

It was thus decided to explore the maximisation of measured variables through the creation of a single proxy for socioeconomic status that could sustain multivariable modelling. There are no clear guidelines for aggregating indicators of socioeconomic position [142, 226]. However, the social sciences have a demonstrated track record of operationalising indicators, particularly for the measurement of poverty through deprivation indexes [143-153] and such have been adopted and recommended by international organisations including UNICEF [227] and the UN sponsored Expert Group on Poverty Statistics [143].

5.4.2.2. Construction of the socioeconomic deprivation index As shown in the construction of deprivation indexes elsewhere [144, 145, 148, 150], all four individual proxies for socioeconomic status —education, income, assets and overcrowding— were grouped into deprivation categories following the operational definitions set out in Table V-15. These new variables were evaluated through inter-item correlations and Cronbach's alpha, where values of 0.7 are considered appropriate for research purposes [145, 228]. The four deprivation variables showed a Cronbach’s alpha of 0.5668, and excluding overcrowding it increased to 0.6040.

The later value was considered a reasonable trade-off that would enable the use of three socioeconomic indicators into a single aggregated variable.

Following this assessment, the equally weighted deprivation scores (0, 1) of education, income and assets were summed, with a maximum score of three and a minimum of zero. Higher scores reflect the experience of a larger number of deprivations simultaneously. A cut-off of two or more deprivations was considered the threshold to define socioeconomic deprivation: nearly 90% of the rural participants were socioeconomically deprived compared to 18% and 7% in migrants and urban people, respectively. If overcrowding were to be included in the calculation of the socioeconomic deprivation index, the prevalences of experiencing simultaneously two or more deprivations would remain the same largely because of the threshold level of two or more deprivations (data not shown).

-162-

Table V-15. Operational definitions of socioeconomic deprivation

Deprivation Yes No

Education None or incomplete

primary education

Primary complete or more

Income Household income less

than USD $150 dollars per month

Household income more than USD $150 dollars per month

Overcrowding Three or more people per room

Less than three people per room

Assets Lowest tertile of

possessions weighted asset index

Middle and highest tertiles of possessions weighted asset index

Note: The categories presented in Table V-13 were used to create specific deprivation variables for each variable presented in this table

Table V-16. Distribution of specific deprivations and deprivation index by study groups

Rural Migrant Urban Missing

data

n = 201 n = 589 n = 199 n/989

Deprived by specific category, n(%)*

Education 132 (65.7%) 183 (31.1%) 13 (6.57%) 0

Income 141 (89.2%) 151 (27.2%) 38 (19.7%)

Overcrowding 94 (47%) 128 (21.9%) 55 (27.8%)

Assets 196 (97.5%) 119 (20.2%) 32 (16.1%)

Number of deprivations per individual, n(%)** 0

None 3 (1.5%) 265 (45%) 131 (65.8%)

* Categories are based on the operational definitions presented in Table V-15.

** Number of deprivations is based on the sum of individual deprivations (education, income and assets) in the same individual.

*** Socioeconomic deprivation index, as explained in the text, was calculated based on individual deprivations except overcrowding. An individual was considered as socioeconomically deprived if had two or more deprivations.

-164-

5.4.3. Remarks in relation to socioeconomic

In document INFORME DE DIAGNÓSTICO (página 73-78)