Now, we will build our skills for using the direct age adjustment methodology with an example. During this session, I will ask you to pause and do some work in the accompanying Microsoft Excel file. It may take you a bit of time to work through the example, but by doing so, you will build your skill set for direct age adjustment methods. The idea behind direct age adjustment is straight forward. You estimate the mortality rates in each age group of your population of interest and then apply those rates to a standard population to calculate the expected number of deaths in the standard population. Then you would total the expected number of deaths to get the overall age-adjusted mortality rate. The effect of the age distribution in your population of interest is removed in these age-adjusted mortality rates, so the rates can be compared to other populations with different age distributions. Selecting the appropriate standard population is important. For comparisons in mortality rates estimated within the US population, the year 2000 standard population is often used. This standard population is reflective of the age distribution in the US in 2000. This standard replaced the 1940, 1970, and 1980 U.S. standard populations. In just a few more years, there's likely to be another standard that will become available that's reflective of the US population. For international comparisons, the WHO standard population is based on the world average population in the calendar years of 2000 to 2025. With increased life expectancy, this standard population has the oldest age group that is greater than 100 years old. There are two other frequently used standard populations, the Segi world standard and the Scandinavian or European standard. The WHO's standard population is older than the Segi world standard, but younger than the Scandinavian standard. The two most frequently used standards are the 2000 US standard and the WHO standard population. Finally, you do not have to use any of the suggested standard populations. If you wish, you could create a standard population if you have two groups of interests by simply combining the two populations of interests. We'll talk more about this strategy a little bit later. The example that we will work through is an investigation into disparities in mortality by race and ethnicity in the state of Maryland, the state in which the Johns Hopkins University is located. We know that data from vital registration systems produce the best data available for vital statistics including mortality rates. We are able to find the Maryland Vital Statistics Annual Report online. Within this report, age, sex, and race/ethnicity stratified counts of the numbers of deaths are available, which will be used for the mortality rate numerator. We will also be able to find population counts of the people at risk for death, or more commonly put, the people who are alive in the same population from which the death came in the same time period. In the accompanying Excel Spreadsheet, you will see a tab labeled Maryland Deaths. In this tab, you will find the number of deaths among Marylanders for each age category, stratified by race, ethnicity, and sex. These data were available in the Maryland Vital Statistics Annual Report and inserted into this Excel file for you. These data will become the numerator in our mortality estimates. The data shown in this slide are not stratified by sex, rather they're the total populations of white, black, and hispanic in each of the age groups. Also in the accompanying Excel spreadsheet, you will see a tab labeled Maryland population. In this tab, you will find counts of the Maryland population for each age category, stratified by race, ethnicity, and sex. These data were again available in the Maryland Vital Statistics Annual Report and inserted into this Excel file. In the table presented here, you see the age distributions only stratified by race/ethnicity. Notice that there are differences in the age distribution by race/ethnicity. The proportion of the population at younger ages is greatest in those who are of Hispanic origin, with blacks as a close second. Conversely, the proportion of the population at older ages is greater in whites as compared with blacks and Hispanics. This can be summarized by simply stating that the white age distribution is a bit older than the black which is a bit older than the Hispanic age distribution. Oftentimes it's easier to see differences in a figure as opposed to a table with lots of numbers. In this figure, you can see the gray bars representing the age distribution of the Hispanic population, they are longest at the younger ages. The white bars representing the age distribution of the white population are longest at the older ages. Now please go to the tab labeled direct MRs in the Excel spreadsheet. For this exercise, you will insert the necessary formulas into the spreadsheet to calculate the age adjusted mortality rate overall and stratified by both race, ethnicity and sex. The gray boxes are boxes in which no number will be inserted. Notice that the number of deaths and the population estimates have been inserted for you. They are actually linked to the Maryland deaths and Maryland populations tabs. So when you want to update your calculations with updated data, you only have to input death and population estimates in one place in your spreadsheet. This technique not only saves time, but it reduces human error because you only input the estimates once into the spreadsheet file. The boxes highlighted in the Crude MR per 1,000, column need to be completed. What formula will you use? Take a minute to program the formula to calculate the crude morality rate per 1,000 in all of the columns labeled Crude MR per 1000 in the spreadsheet. Click Continue when you are ready. To calculate the Crude Mortality Rate per 1,000 population, you can program the cells in the spreadsheet to divide the number in the Deaths column by the number in the Population column. And then multiply by 1,000 to get the mortality rate per 1,000 population. Next we will estimate the expected number of deaths by multiplying our Crude MR per 1,000 by the US 2000 standard population. This estimate is also called the Cross Product in the NAPHSIS direct age adjustment tutorial that I recommend you review as another example of direct age standardization. To estimate the cross product, you multiply the Crude Mortality Rate by the US 2000 standard population. The estimated rates for the US 2000 standard population have already been inserted into the spreadsheet for you. As a side note, if we had chosen to use the approach of creating a standard population by combining two populations, we would add the cells in the population column over the different race/ethnicity groups, and then estimate the proportion of the total population in each age group. The proportion in each age group out of the total population becomes the weight, just as you see the weights from the 2000 US standard population. Take some time to insert the appropriate formula into the highlighted cells in the Expected numbers of deaths cross product cells. Click continue when you are ready. The cross product is estimated by multiplying the Crude Mortality Rate in each age group by the US 2000 standard population. The final step in estimating the age-adjusted mortality rate is to sum or add all of the expected number of deaths or cross products. I don't have enough space to show you the whole table on this one slide, so I used three periods to signal the top half of the table as missing. Be sure to include all of the cross products from the table when you sum up all the cross products and enter the sum into the highlighted cell. Take a minute to calculate the age-adjusted mortality rates for all the strata in the direct MRs tab of the spreadsheet. Click Continue when you are ready. The calculation for the age-adjusted mortality rate is adding all of the cross products in the race, ethnicity, and sex strata. Did you estimate an age-adjusted mortality rate that matches the one seen on the slide, for white Marylanders, male and female sex combined? I hope so. One last calculation before we wrap up this exercise. Notice that we also need to calculate the Crude Mortality Rate overall. This is different from the Crude Mortality Rate you calculated for each age group, which we call the age specific mortality rates. The overall Crude Mortality Rate is estimated as the total number of deaths divided by the total population for each sex and race ethnicity strata. Please estimate the Crude Mortality Rates for each of the race ethnicity, and sex strata in the direct MRs tab of the spreadsheet. Click Continue when you are ready. Did you get the same estimate of the Crude Mortality Rate per 1,000 population as shown on the slide? Again, this is for white Marylanders of both sexes. I hope so, good work. Now, you maybe wondering why we have estimated the Crude Mortality Rate when we went to all these trouble to estimate the age-adjusted mortality rate. Well, the convention is to report both the Crude and the age-adjusted mortality rates. The reason for this convention is important. The Crude Mortality Rate is the actual mortality rate being experienced by the population. It is real. This is important for public health planning. The age-adjusted mortality rate is not real. It's an estimate created after removing the influence of the age distribution on the mortality rate. The age-adjusted mortality rates are meaningful when they're compared across subgroups defined by person, place, or time. But it is not the mortality rate actually experienced by the population of interest. The figures show the Crude and age-adjusted mortality rates that we just calculated for male and female Marylanders. Notice that the Crude Mortality Rates in males show a higher mortality rate for whites compared to blacks. In the age-adjusted mortality rates, this is reversed. We see that blacks have a higher age-adjusted mortality rate than whites. Hispanics have the lowest Crude and age-adjusted mortality rates. Although our analysis does not investigate why Hispanics have lower Crude and age-adjusted mortality rates. Often the healthy migrant effect is discussed when analyses show low Crude or age-adjusted mortality rates among Hispanics. This effect is most simply explained because immigration is often a taxing experience. Those who are not as robust, health-wise, often do not immigrate. Although, this does not necessarily apply to Hispanics that are not immigrants. This healthy migrant effect does not necessarily apply to the Hispanics that are not immigrants within the population. A few quick technical notes about direct age adjustment. First, always be sure to report the standard population used. Second, remember that the age-adjusted rates are not the observed rates but are hypothetical rates. The value of these age-adjusted rates is to allow for comparison of rates in sub groups defined by person, place, or time. The third note is a word of caution. If there are less than 25 total deaths within the age group of your population of interest, indirect, not direct, age adjustment should be used. Notice in our example that there were less than 25 deaths in many of our age groups among Hispanic males and Hispanic females. We will discuss how indirect age adjustment can be used in this situation. Forth, it is not meaningful to age adjust data for small ranges of age groups. For example, if you were only interested in young adults age 18 to 24, there's no need to further break down those ages into smaller age groups and attempt to adjust for the impact of age on this very limited age range. Finally, you should not age adjust if age specific death rates do not have a consistent relationship over the time period of your study. For example, if the mortality rate among younger persons increases over time, but the rate among older persons decreases over time, adjusting for age could mask these changes in the mortality rate within each of these age groups. In this scenario, you would want to show those changes in the mortality rate over time in the different age groups, not just adjust that away, because those changes could be very meaningful. Here's a quick summary of the important calculations needed to use the direct age adjustment methodology for mortality rates. There's another version of this spreadsheet that contains all the correct formulas, so please check your work against the posted completed spreadsheet. Let's take a quick break before moving onto indirect age adjustment.