Adjusted Rates Direct Standardization Crude Rates Overall rates, e.g., obtained by dividing total cancer deaths by total population. Population A Population B 50,000 people 52,000 people 4,000 cancer deaths in a year 5,080 cancer deaths in a year crude rate = 800 per 10,000 crude rate = 977 per 10,000 Population B has a higher crude rate. Does this mean that the risk of cancer death is greater in “B”? Are there greater environmental risks in “B”? Age-Specific Rates Population “A” Age Deaths Pop. Population “B” Rate /10,000 Deaths Pop. Rate /10,000 30-39 400 10,000 400 80 2,000 400 40-49 600 10,000 600 300 5,000 600 50-59 800 10,000 800 800 10,000 800 60-69 1,000 10,000 1,000 1,500 15,000 1,000 70-79 1,200 10,000 1,200 2,400 20,000 1,200 Totals 4,000 50,000 800 (crude) 5,080 52,000 977 (crude) / / • Is it riskier to live in population “B”? • Why are the overall (crude) death rates different? Age Pop. A Age Pop. B Age is an additional factor that is affecting the comparison. (confounding) A Crude Rate Is a Weighted Average of Age-Specific Rates (Wgt.) Rate (Wgt.) Rate Age Pop. % Deaths /10k Pop. % Deaths /10k 30-39 10,000 20% 400 400 2,000 3.85% 80 400 40-49 10,000 20% 600 600 5,000 9.62% 300 600 50-59 10,000 20% 800 800 10,000 19.23% 800 800 60-69 10,000 20% 1,000 1,000 15,000 28.85% 1,500 1,000 70-79 10,000 20% 1,200 1,200 20,000 38.46% 2,400 1,200 50,000 4,000 52,000 5,080 Crude rate - 4,000/50,000 = 800/10,000 Crude rate = 5,080/52,000 = 977/100,000 Young Old Young Old A Crude Rate Is a Weighted Average of Age-Specific Rates Age 30-39 40-49 50-59 60-69 70-79 (Wgt.) Pop. % 10,000 20% 10,000 20% 10,000 20% 10,000 20% 10,000 20% 50,000 Rate Deaths /10k 400 400 600 600 800 800 1,000 1,000 1,200 1,200 4,000 Crude rate - 4,000/50,000 = 800/10,000 (Wgt.) Pop. % Deaths 2,000 3.85% 80 5,000 9.62% 300 10,000 19.23% 800 15,000 28.85% 1,500 20,000 38.46% 2,400 52,000 5,080 Rate /10k 400 600 800 1,000 1,200 Crude rate = 5,080/52,000 = 977/100,000 .20 x 400 = 80 .0385 x 400 = 15.40 .20 x 600 = 120 .0962 x 600 = 57.72 .20 x 800 = 160 .1923 x 800 = 153.84 .20 x 1,000 = 200 .2885 x 1,000 = 288.50 .20 x 1,200 = 240 .3846 x 1,200 = 461.52 SUM 800 SUM 977 The crude rate is weighted by the age distribution. What if two populations have different age distributions and age affects the cancer rate …? The Real Question How would the overall cancer mortality rates compare if the age distributions were the same? Adjustment (Standardization) If populations being compared have different distributions with respect to age, or other factors, …one can calculate adjusted rates that take into account differences in the structure of the populations being compared. The adjusted rates artificially make the two populations have identical distributions of the confounder (age, race, gender, etc.). Basically, we ask the question, “What if the population distributions were (weighted) the same with respect to the confounder? Then, how would the rates compare? Comedian Robert Klein: “I don’t know about this Florida thing. All I know is that I had two perfectly healthy 65 year old parents. They move down to Florida and then, bang, thirty years later they’re dead. I don’t know … do you think it’s something in the air or the water down there?” Death Rates In Florida & Alaska Number of deaths Total population Crude mortality rate \(per 100,000) Florida 131,902 12,340,000 Alaska 2,116 530,000 1,069 399 The crude rates are clearly different. Does this mean that it is riskier to live in Florida? If you are about to retire, would it be better to move to Alaska? Note: The Age-Specific Rates are Similar Florida % of total Age Pop. (Weight) <5 850,000 7% 5-19 2,280,000 18% 20-44 4,410,000 36% 45-64 2,600,000 21% >65 2,200,000 18% Totals 12,340,000 100% This contributes 18% to the overall rate. Crude mortality rates (per 100,000) Rate per 100,000 284 57 198 815 4,425 Alaska % of total Pop. (Weight) 60,000 11% 130,000 25% 240,000 45% 80,000 15% 20,000 4% 530,000 100% Florida 1,069 Rate per 100,000 274 65 188 629 4,350 Alaska 399 The crude rates are very different, but crude rates are weighted averages of the age-specific rates, and Florida’s population is weighted more heavily with older people. The comparison is confounded by age differences. Direct Standardization The Solution: use each population’s actual age-specific rates, but calculate a summary rate using a single (standard) age distribution (i.e. artificially weight them the same with respect to age distribution.) This is adjustment by Direct standardization. The “adjusted” rates are artificial, but they provide summary rates that can be compared without confounding by age differences. Florida As The Standard: Florida % of total Age Pop. (Weight) <5 850,000 7% 5-19 2,280,000 18% 20-44 4,410,000 36% 45-64 2,600,000 21% >65 2,200,000 18% Totals 12,340,000 100% Adjusted Mortality Rates (#1) .07 x 284 = 19.88 .18 x 57 = 10.26 .36 x 198 = 71.28 .21 x 815 = 171.15 .18 x 4,425 = 796.50 SUM Rate per 100,000 284 57 198 815 4,425 Alaska % of total Pop. (Weight) 60,000 11% 130,000 25% 240,000 45% 80,000 15% 20,000 4% 530,000 100% Rate per 100,000 274 65 188 629 4,350 .07 x 274 = 19.18 .18 x 65 = 11.70 .36 x 188 = 67.68 .21 x 629 = 132.09 .18 x 4,350 = 783.00 1,069/ 100,000 pop. SUM (Age-adjusted) 1,014/ 100,000 pop. Adjusted Mortality Rates (#2) Florida Age <5 5-19 20-44 45-64 >65 % Pop. 7% 18% 36% 21% 18% 100% Average of Florida & Alaska Distributions as the Standard: Alaska % Pop. 11% 25% 45% 15% 4% 100% Weight Rate .090 x 284 = 25.56 .215 x 57 = 12.26 .405 x 198 = 80.19 .180 x 815 = 146.70 .110 x 4,425 = 486.75 SUM 751/ 100,000 pop. Average (9.0%) (21.5%) (40.5%) (18.0%) (11.0%) 100% Weight Rate .090 x 274 = 24.66 .215 x 65 = 13.98 .405 x 188 = 76.14 .180 x 629 = 113.22 .110 x 4,350 = 478.50 SUM 707/ 100,000 pop. Age-adjusted Adjusted Mortality Rates (#3) Florida Age <5 5-19 20-44 45-64 >65 % Pop. 7% 18% 36% 21% 18% 100% Alaska % Pop. 11% 25% 45% 15% 4% 100% Weight Rate .07 x 284 = 19.88 .22 x 57 = 12.54 .40 x 198 = 79.20 .19 x 815 = 154.85 .12 x 4,425 = 531.00 SUM 797/ 100,000 pop. 1988 U.S. Population as the Standard: 1988 U.S. (7%) (22%) (40%) (19%) (12%) 100% Weight Rate .07 x 274 = 19.18 .22 x 65 = 14.30 .40 x 188 = 75.20 .19 x 629 = 119.51 .12 x 4,350 = 522.00 SUM 750/ 100,000 pop. Adjustment By Direct Standardization • Provides summary rates (all ages) that remove the unwanted effects of differences in the distributions of confounders in the populations. However, the adjusted rates are not real. (Only good for comparison.) • Standardization doesn’t always make the two rates more similar (can be more different or no difference). • It just allows a fairer comparison after ironing out some of the “other” differences that might be exaggerating or masking differences between the populations. • Direct standardization may involve more than 2 groups. Death Rates in Weymouth Versus Woburn Was there confounding by age? • Look at the crude rates. • Look at the adjusted rates. • How is the comparison affected by adjusting for a factor, such as age? • Are the apparent differences greater or smaller? – Did age differences exaggerate differences between the two groups? – Did age differences mask differences between the groups? #1 Compare The Crude & Adjusted Rates What is your interpretation? Did age differences have a confounding effect? Were the populations different after adjusting for age? 800 600 Crude: Weymouth Woburn 250/10,000 vs. 750/10,000 400 200 0 Crude 800 600 Age Adjusted: 376/10,000 vs. 383/10,000 400 200 0 Adjusted #2 Compare The Crude & Adjusted Rates What is your interpretation? Did age differences have a confounding effect? Were the populations different after adjusting for age? 800 600 Weymouth Crude: Woburn 250/10,000 vs. 750/10,000 400 200 0 Crude 800 600 Adjusted: 376/10,000 vs. 512/10,000 400 200 0 Adjusted #3 Compare The Crude & Adjusted Rates What is your interpretation? Did age differences have a confounding effect? Were the populations different after adjusting for age? 800 600 Crude: Weymouth Woburn 250/10,000 vs. 750/10,000 400 200 0 Crude 800 600 Adjusted: 306/10,000 vs. 813/10,000 400 200 0 Adjusted #4 Compare The Crude & Adjusted Rates What is your interpretation? Did age differences have a confounding effect? Were the populations different after adjusting for age? 800 600 Weymouth Crude: Woburn 250/10,000 vs. 266/10,000 400 200 0 Crude 800 600 Adjusted: 276/10,000 vs. 450/10,000 400 200 0 Adjusted Age-adjusted mortality rates, by race & ethnicity, MA 2001 Age-adjusted rate/100,000 1,200 1,050 1,000 813 800 617 600 452 400 200 0 Black non- White nonHispanic Hispanic Hispanic Asian/PI Age-adjusted to the 2000 U.S. standard population It’s The Same Population, But At Multiple Times (It’s Like Comparing Multiple Populations) 1990 1985 1980 1992