Health and Income Equity

For the reader who is not familiar with public health and epidemiology terminology, brief explanations are given here. A basic textbook on epidemiology will help.
  • adjustment
  • adrenocortical axis
  • age adjustment
  • association
  • bias
  • bivariate associations
  • cohort
  • confounding, confounders
  • controlling for
  • correlation coefficient
  • cortisol
  • covariate
  • cross sectional design
  • ecological study
  • endocrine
  • epinephrine
  • Gini coefficient
  • glucocorticoids
  • health inequalities (or inequalities in health)
  • incidence
  • income inequality
  • infant mortality rate (IMR)
  • life expectancy (or expectation of life)
  • logistic regression
  • median income
  • multiple logistic regression
  • multivariate analysis
  • p value
  • prevalence
  • regression
  • social capital or social cohesion
  • socioeconomic status
  • standardized mortality rate (SMR)
  • statistical significance
  • Adjustment (see controlling for)

    Adrenocortical axis
    This is the feedback loop in primates and other mammals, in which the adrenal (next to the kidney) gland secretes a hormone ( cortisol) from its outer shell (cortex), which initiates a feedback loop in the hypothalamus (part of the brain), to suppress further secretion of that hormone. The axis refers to this loop.

    Age adjustment
    This is a summarizing procedure to make comparisons of populations with different age distributions. Age adjusted mortality rates allow comparison of death rates between populations in which one such as Florida in the US which has a large percentage of older people, with another such as Illinois, where the proportion of older people is less. Age adjusted or age standardized mortality rates are summary measures that assume the population has a standard age distribution.

    In epidemiology, this refers to a statistical relationship between one or two events or variables. In our material, this usually refers to a relationship between a measure such as life expectancy, and another, usually income distribution. The relationship in this case is negative, that is higher life expectancies are found in populations with less inequality of income distribution. No causality is implied in using the term association. The term relationship is sometimes used instead of association.

    This refers to deviation of results or inferences from the truth, or to the processes leading to such deviation. In other words, because of a certain bias in the study, the findings did not come out correctly, and an erroneous conclusion was drawn. A simplistic example would be if you compared the death rates in two populations, one young, and the other old, in a rich country, and found the older one had more deaths, then concluded that it was less healthy than the younger one. This would be a poor analysis, unless you recognized the bias that the different age structure produced in coming to the conclusion. In a poorer country, just the opposite might be true, death rates for a population of people in their 20s may be lower than for a population of children up to age 10. Good investigators attempt to describe possible bias in their studies, and correct for these whenever possible.

    Bivariate associations
    Associations between two variables.

    This is a population born during a particular historical period, that is usually followed over time. It can also mean any population, not necessarily born during the same time period, that is followed over time.

    Confounding, confounders
    This refers to a situation where two separate processes are going on that must be looked at individually. In our example for bias, age would be a confounder, that is, the investigators should have looked at the ages of the populations and separated them before drawing conclusions. An example could be the finding that people drinking coffee are more likely to have heart attacks. But people who drink coffee are more likely to smoke, so it may be that this association is just that between smoking and heart attacks. Smoking would be said to confound the relationship between coffee drinking and heart attacks.

    Controlling for
    Adjusting for is another term with the same meaning, namely that in the statistical analysis, account is taken of a factor. That is, differences in the factor are corrected for in the analysis, so that they do not account for the findings. If you correct or control or adjust for socioeconomic status in a study of lung cancer in a population, then the effect cannot be due to the poor having more lung cancer. In the studies showing that income distribution was associated with mortality by states, the relationship held true after controlling for different poverty levels, or absolute incomes, or smoking levels, in the different states. Hence these factors were not responsible for the association.

    Correlation coefficient
    This number, between plus and minus one (-1.0 to 1.0)measures the amount of agreement between two variables, meaning how close the graph drawn linking to two is to a straight line. IT is also called the Pearson Correlation Coefficient.

    A hormone secreted by the adrenal gland in the body that has numerous functions, and is elevated in stress states.

    A variable that might be possibly predictive of the outcome under study. If life expectancy is the outcome, income distribution can be considered a covariate.

    Cross sectional design
    A study looking at a defined population at one point in time. For example among the fifth US states, looking at their mortality rates in 1990 to see what relationship there is with a measure of income distribution by state represents a cross sectional design.

    Ecological study
    An investigation in which populations or groups of people, rather than individuals are looked at. Most of the studies referred to in this web page are such. An individual does not have a life expectancy, nor an income distribution, but a population, a city, state or country, does.  Ecologic studies do not allow statements to be made about individuals, just about the population.  See ecologic fallacy in the Overview and Making Causal Inferences

    A term referring to a gland that secretes its hormone directly into the bloodstream that flows through it, rather than through an opening into the intestine, as say the gallbladder, or pancreas does with digestive enzymes.

    An (endocrine) hormone secreted by the adrenal gland in its interior (medulla), that activates the acute stress, flight or fright reaction. Adrenaline is another term for it.

    Gini coefficient
    A measure of inequality, usually applied to income. It is derived from a Lorentz Curve which plots the cumulative percent of income against the cumulative percent of income recipients. It is twice the area of the curve between what would be perfect equality and the existing distribution is the Gini coefficient. A coefficient of 0 means perfect equality, while that of 1 means one unit of the population has everything and there is none for the rest. It is more sensitive to differences in the middle of the distribution, than to the ends.

    The cover term for the hormones made in the adrenal cortex, an example of which is cortisol.

    Health inequalities (or inequalities in health)
    This is the term commonly used in Europe to indicate the virtually universal phenomenon of variation of health by socioeconomic status, that is poorer people have poorer health. In the US, there is no single such term, and instead it is referred to as the socioeconomic status and health relationship.

    Incidence refers to new events, within a specific time period. An incidence rate is the number of occurrences of something, such as an illness, per unit of time (usually a year), per person, or per 1000 or per 100,000 people in that population.

    Income inequality
    A general term referring to the variation of income in a population, that is some people have more than others. The Gini coefficient is one measure of this.

    Infant mortality rate (IMR)
    The IMR is the number of deaths occurring in a population per year among infants in their first year of life.

    Life expectancy (also termed the expectation of life)
    This number, for a population , is the average number of years an individual, born today, would be expected to life if current mortality rates continued to apply. To calculate it, you need to know the mortality pattern of the population, that is the death rates in different age intervals.

    Logistic regression
    This refers to using a statistical model for a process that contains an exponential factor, and seeing what is the best fit of the data.

    Median income
    The median of a measure in a population is the number which divides the population into two equal groups, those above, and those below. Median income would then be the income value that separates the population as above

    Multiple logistic regression
    This is a kind of logistic regression in which there are many variables, including several exponential factors.

    Multivariate analysis
    This refers to a study in which there are several variables being considered simultaneously.

    P value
    Refers to the probability that the result obtained could have happened by chance. Usually refers to a number derived from a calculation in the study and is displayed as p_ 0.05 or p_ 0.01 or such. This means the likelihood of such a result by chance is less than one in twenty or one in a hundred. The custom is to consider p values of 0.05 or less to signify a significant result, one highly unlikely to happen by chance. A p value .1 or higher is more likely to be a chance event, and is accorded less significance.

    The prevalence counts the number of events in a specific population, not just new occurrences, as in incidence. The incidence of malnutrition would be all the new cases that occurred that year, while the prevalence would count the total number of people with malnutrition in that population.

    Originally used to find the best straight line that fits the data representing two variables under question. Regression analysis can also be used to indicate the process of trying to fit a mathematical relationship to the data

    Social capital or social cohesion
    Terms, that relate to the features of social organization and community life, such as civic participation, norms of reciprocity and trust in others that facilitate cooperation for mutual benefit.

    Socioeconomic status
    A descriptive term for a person's position in society, usually expressed in terms of income, education, occupation, but it could also be represented by net worth, ownership of assets such as a home, automobile, yacht, etc.

    Standardized mortality rate (SMR)
    The ratio (times 100) of the number of deaths observed in a population to the number that would be expected if the study population had the same specific rates as the standard population. The standard population is specified. This is different than the age adjusted mortality rate, in that former is a ratio (comparison of two rates, one to a standard population), while the latter is a rate (number per unit time), assuming the population had a specific age distribution.

    Statistical significance
    Some estimate in a study is said to be statistically significant if it is unlikely to happen by chance. Usually it is described as a number, or a curve fit, with a p value that is sufficiently low. Usually p=0.05 or less.

    Home Overview and making causal inferences Glossary Papers/Readings
    ©2003 Population Health Forum | Contact Us | University of Washington | School of Public Health