Introduction

To examine the relationship of a population's socioeconomic characteristics to its health status and use of health care services, a composite Socio-Economic Risk Index (SERI) was developed for the Population Health Information System.

From a set of socioeconomic indicators derived from public use census data, a summary index was formed from 6 indicators to generate profiles for regions of the province. The Socio-Economic Risk Index appears to be a powerful tool in clarifying what benefits in improved health status might accrue from changing the underlying inequities in amenable socioeconomic risk factors, rather than simply increasing services to regions of low health status. For more background information, please read the document titled: SERI: The Original Index and the MCHP deliverable Socio-Economic Characteristics (Population Health Information System 1991/92 - 1986 Census Version) by Frohlich et al. (1994).

SERI also served as the basis for the construction of the Socioeconomic Factor Index (SEFI). Details regarding this index can be found in the Socioeconomic Factor Index (SEFI) - Based on the 1986, 1991, and 1996 Census Data concept.

Data

Note: The MCHP deliverable, "Issues in Developing Indicators for Needs-Based Funding" (1997) details the updated SERI using a blended index from the 1986/1991 census data.

The Socio-Economic Risk Index (SERI) has been calculated for the 1986 and 1991 census data on the 60 physician service areas. This provides a more stable estimate of risk than using either year individually for calculating the index. This data is stored on the MCHP system in SAS readable format.

A SAS macro (_seri) is available on our system that will regroup this data into the Manitoba Regional Health Authorities (RHA) or Winnipeg sub-regions.

Methods

The following information is a discussion of the methods used to calculate SERI using 1986 census data. A similar method was used to re-calculate the index from the 1991 census data. The final index is based on a combination of both the 1986 and 1991 indicators.

Measures of Area Socioeconomic Characteristics

Data for measures of socioeconomic characteristics were obtained from public use files released by Statistics Canada. This library reports comprehensive data for each of the enumeration areas in Manitoba. The indicators have been grouped into six domains: dwelling characteristics, education, employment, income, mobility and social characteristics.


• Dwelling characteristics: A single measure of housing, the average market value of owner-occupied single detached dwellings, was selected for this analysis. Dwellings located on farms and on First Nations reserves are excluded from census market value estimates. In these analyses, dwelling values are interpreted as one indicator of a mixture of differences in the cost of living across regions rather than as an indicator of housing quality.
  
  
• Educational attainment: The proportion of residents in each enumeration area who attained a minimum of a high school diploma are described for three consecutive age cohorts: 25-34, 35-44 and 45-54 years of age. Younger cohorts have higher high school completion rates. At the same time, when represented at the regional level, there is little evidence that differences in high school completion rates have narrowed over time, suggesting significant persistence of regional social and economic characteristics.
  
  
• Employment: Measures from three domains of labor force activity were obtained: the percent of the labor force in each region engaged in three occupational groupings, female labor force participation and the regional unemployment rate for four age cohorts. The three occupational groupings were 1) farming, 2) manufacturing, construction and transportation and 3) managerial, administrative and scientific occupations. Unemployment rates display some of the largest regional differences among all the indicators reviewed for this report. As was also seen across cohorts in the measures of attained education, the regional unemployment pattern is consistent across age cohorts, with the most substantial regional differences occurring in the youngest age groups.
  
  
• Income: A single indicator of household economic resources, total household income from all sources, was obtained for these analyses. In addition, three measures of household housing status were implemented: the percent of all households in owner-occupied dwellings, the percent of households in owner-occupied dwellings which spent 30% or more of household income on housing, and the percent of households in tenant-occupied dwellings which spent 30% or more of household income on housing costs.
  
  
• Mobility: A single measure of population mobility was included in the set of indicators: the proportion of an area's population aged 5 years or older which moved into the area from other locations within Canada in the previous 5 years. Note that this indicator does not measure net migration, defined as in-migration minus out-migration.
  
  
• Social characteristics: Three dimensions of social characteristics were included in the selected set of indicators: an age dependency ratio, the rate of single parenthood among families with young children, and the regional distribution of French and Aboriginal language speakers. The age dependency ratio was defined as the ratio of the number of people in a region 65 years of age and older to the number of people aged 15-64. Three measures of single parenthood were selected: the percent of single-parent households among households with children aged 0-14, the percent of single female parent households among households with children aged 0-14, and the percent of single female parent households among all households with parents aged 15-24 and children aged 0-14.

Measures of Health Status

A set of four morbidity indicators, which were known to be sensitive to socioeconomic status, were chosen as a base against which to measure the explanatory power of candidate socioeconomic indicators (Dougherty et al., 1990; MacLaughlin et al., 1989; Wissow et al., 1988). In addition, we included a measure of fertility which is inversely related to socioeconomic status among younger women and which has important implications for the physical and developmental health of children.

Morbidity indicators consisted of 1) hospital admissions for injury to females, 2) hospital admissions for injury to males, 3) hospital admissions for children aged 0-4 years for respiratory infection, 4) hospital admissions for persons aged more than 65 years for respiratory infection. These measures were obtained by reviewing all hospital discharge abstracts. Only hospital separations with a primary diagnosis were selected. All admissions occurring outside the resident's area of residence were assigned back to the region of residence, and status First Nations were assigned residence on the basis of postal code.

Index Construction Methodology

As a first step in the analyses, a single index measure of the five health status indicators was constructed. Data was aggregated to the level of the municipality and each indicator was normalized by subtracting the provincial average from the observed score for each municipality and dividing the result by the variable's standard deviation (Carstairs & Morris, 1989; 1991). The index was then formed from the sum of the five normalized indicators. This sum was divided by the square root of 5, producing a scale which reads in standard deviations from the provincial mean. High positive scores on this index represent geographic areas with poorer health.

The variables derived from census data were also aggregated to the municipal level and individually normalized. The normalized variables were then correlated individually with the prototype health status index, and those measures correlating at the 0.10 level or greater were retained and entered in a stepwise multiple linear regression.

In the regression, six measures were found to explain the maximum amount of variance in the index measure of municipal health status: 1) the percentage of the labor force unemployed between the ages of 15 and 24, 2) the percentage unemployed between 45 and 54, 3) the percentage of single-parent female households, 4) the percentage of the population aged 25-34 having graduated high school, 5) the percentage of female labor force participation, and 6) the average value of owner-occupied dwellings. Other candidate measures among the socioeconomic variables were not found to add significantly to the explanatory power of these six variables.

The socioeconomic index was formed from the weighted sum of standardized forms of the six selected measures, with the regression coefficients used as weights. The final equation was:

Original equation based on 1986 data.

0.347*(percent of labor force unemployed aged 15-24)

+ 0.90*(percent unemployed aged 45-54)

+ 0.181*(percent single-parent female households)

- 0.212*(percent aged 25-34 having graduated high school)

- 0.271*(percent female labor force participation)

- 0.128*(value of owner-occupied dwellings)

Current equation based on 1986/1991 data.

0.1025*(percent of labor force unemployed aged 15-24 in 1986)

+ 0.3382*( percent of labor force unemployed aged 15-24 in 1991)

+ 0.1972*(percent unemployed aged 45-54 in 1986)

- 0.1362*(percent unemployed aged 45-54 in 1991)

+ 0.0602*(percent single-parent female households in 1986)

- 0.0019*(percent single-parent female households in 1991)

- 0.1850*(percent aged 25-34 having graduated high school 1986)

- 0.0616*(percent aged 25-34 having graduated high school 1991)

- 0.0902*(percent female labor force participation in 1986)

- 0.2822*(percent female labor force participation in 1991)

- 0.2703*(value of owner-occupied dwellings in 1986)

+ 0.1272*(value of owner-occupied dwellings in 1991)

In the final form of the index, the sum was divided by the square root of the sum of squares of the regression coefficients, to reproduce the scale in standard deviations. The coefficients obtained from the municipal level regressions were also used to construct the index at the level of the provincial health regions.

A number of sensitivity analyses were performed to establish the robustness of both indices. We restricted a correlation analysis to rural municipalities only and also tested the correlation of the two indices at the municipal level within each of eight regional aggregations. In addition, the health status index was constructed without the inclusion of the fertility measure. Finally, correlations were calculated directly between the individual health status indicators and the socioeconomic index. None of the findings from these additional analyses was inconsistent with the initial methodology.