Estimating need for physician visits with the Socio-Economic Risk Index
The Socio-Economic Risk Index (SERI)9 is a composite index of 6 measures of socioeconomic status that mark environmental, household and individual conditions placing residents of a particular area at risk for poor health and hence that are associated with a greater need for health care. The 6 variables were chosen from a pool of 23 for their strong relation to health status and use of health care resources: the proportion of people aged 15 to 24 years who are unemployed, the proportion of people aged 45 to 54 years who are unemployed, the proportion of single-parent households headed by a woman, the proportion of high school graduates aged 25 to 34 years, the proportion of female residents participating in the labour force, and the average value of a dwelling in the area. The first 3 variables are negatively related to health status (high values being associated with poor health), and the last 3 are positively related to health status. For this study, SERI values were calculated for each physician service area on a standardized scale, the provincial average corresponding to a SERI value of 0. Negative values represent areas at lower risk and positive values represent areas at higher risk. For a thorough explanation and discussion of the SERI, see Mustard and Frohlich.9
Estimating the need for physician visits
Our analysis of residents' need for physician visits was based on several assumptions: first, that in 199394, Manitobans, on average, made visits in a way that reflected the needs of their age, sex and socioeconomic characteristics; second, that people with similar characteristics require a similar number of physician contacts regardless of where they live; and third, that areas with residents in poorer health should have access to more care than areas with healthier residents, all other things being equal.
We used a 2-step process to generate estimates of the need for physician visits in each area, expressed as the average number of visits needed per resident per year (a numeric example is provided in Table A1). Step 1 accounted for factors that have been shown to influence a population's need for physician visits: differences in the age and sex make-up of the population and the socioeconomic characteristics of the area and its residents. The predicted values in step 1 were produced by a multiple linear regression analysis of actual 199394 visit rates on the age and sex distribution and the SERI values for each region (using 21 age groups and including all interaction terms). Essentially, this analysis results in a description of physician visits for the Manitoba population in 199394, according to demographic and socioeconomic factors (the model results are available on request from the corresponding author).
The second step modified the predictions from step 1 to account for differences in the health status of areas residents, as measured by premature death rates. This extra step was carried out in recognition of the desirability of providing more visits in areas whose residents were in poorer health, all other things being equal. We modelled the relation between each area's premature death rate and our step 1 estimate of need using simple linear regression, which produced a slope of 0.56 (p = 0.001). For areas in which the premature death rate was higher or lower than the provincial average (3.58 per 1000 residents aged 74 years or less) we multiplied the difference between the premature death rate for the area and that for the province by 0.56. The value obtained was then added to (if the death rate was higher than the provincial average) or subtracted from (if the death rate was lower than the provincial average) the need estimate from step 1.
Example of calculation of need values
Residents of Springfield and Grahamdale had similar actual visit rates in 199394 (see Table A1), but Springfield's age and sex characteristics and much lower SERI value resulted in a step 1 need estimate that was lower than the SERI value for Grahamdale. In step 2 the good health status of Springfield residents, as indicated by the lower than average premature death rate, resulted in a reduction of the step 1 need estimate to the final need estimate. Grahamdale's above-average premature death rate resulted in an increase from the step 1 need estimate to the final need estimate.
In a few rural areas, physicians provide a relatively small proportion of the primary care contacts (from 17% to 69%). But because these areas have some of the least healthy, highest-need populations in the province, we did not want to exclude them from the analysis altogether. To help counteract the lack of physician visits, nursing station contacts were included in the regressions used to estimate need.
[Return to text]
| CMAJ November 1, 1997 (vol 157, no 9)
/ JAMC le 1er novembre 1997 (vol 157, no 9) |
|
