COMPARING MORTALITY BETWEEN DIFFERENT POPULATIONS Direct and indirect

COMPARING MORTALITY BETWEEN DIFFERENT POPULATIONS Direct and indirect age-adjustment Akbar soltani Assistant Professor of Medicine and Endocrinology Evidence-Based Medicine Working Team (EBMWT) Tehran University of Medical Sciences (TUMS) Shariati Hospital

Question about crude rates? n Crude mortality rate in Florida in 1981: n n 10. 9 per 1, 000 Crude mortality rate in Alaska in 1981: n 4. 4/1, 000 Is it 2. 48 times more dangerous to live in Florida? Mortality rate for osteoporosis in USA Vs IRAN?

How do we compare rates across populations? Crude rates are not helpful because … Populations differ in their age distributions Populations differ in their racial distributions Populations differ in their SES distributions

Objectives Introduction n Measures of mortality n Examples n

EXAMPLE: WHAT IS WRONG?

Risk of Fracture • In 50 years old : • Women 40% • Men 13%

Risk of Fracture in Iran (In more than 50 years old population) • Iran population in 1381: • 50 – 95 and more = 11. 5% = 65, 540, 224 7, 497, 801 • 49% = 3, 673, 922 • Risk of Fracture=40 /100 = 1, 469, 568 • All > 50 (at risk) = 1, 469, 568 + 573, 581 = 2, 043, 149

Three common forms of rates • Crude or unadjusted rates: based on cases occurring in the total observed population e. g. Crude birth rate, crude death rate • Specific or stratified rates: rates that are specific for age and sex are most likely true difference in disease risk between population e. g. sex-specific, age-specific, race-specific • Adjusted or standardized rates: if the frequency of the disease varies by both age and sex , there are 20 or more rates for comparison e. g. age-adjusted

Crude (or unadjusted) rates n n n Estimate the actual disease frequency for a population Can be used to provide data for allocation of health resources and public health planning Can be misleading if compared over time or across populations

Crude (or unadjusted) rates n For example, the number of persons nationally with diagnosed diabetes increased from 1. 6 million in 1958 to 8 million in 1995 a fivefold increase. Currently, in Illinois, approximately 500, 000 persons 18 years of age and older have diagnosed diabetes.

Category-specific (or stratified) rates n n n Provide more detailed information than crude rates about patterns of disease frequency in different populations (change in mortality with age? ) Can be used for valid comparison of populations Can be cumbersome if there is a large number of categories to compare Category-specific rates can provide general characteristics of the frequency of disease in a population. Population may not be large enough or disease common enough

How do we compare rates across populations? We compare rates across populations by putting them on an even playing field that is, we either standardize one population on another or we use an outside standard and adjust our populations to that standard.

Age-Adjustment Two types of Age-adjustment 1. Direct Method 2. Indirect Method (SMR=standard mortality ratio)

EXAMPLE-2

For example, direct age adjustment Population A Risk Population B AGE N Cases <20 21 -50 >50 100. 1 10 200. 2 40 500. 4 200 __________ 800 250 CRUDE RISK = 250/800 = 31% N Risk Cases 500. 1 50 200. 2 40 100. 4 40 __________ 800 130/800 = 16% • Crude risk indicates different risks of disease between populations. • But Age-specific rates indicate similar risks.

Direct Method : Apply risks in population B to population A. Population A Risk Population B AGE N <20 21 -50 >50 100. 1 10 200. 2 40 500. 4 200 __________ 800 250 AGE ADJUSTED RISK = 31% Cases N Risk Cases . 1 10. 2 40. 4 200 _________ 250 = 250/800 = 31%

n Example: : indirect adjustment n n n DWI’s among students attending the University of Margaritaville student population 1000 Observed DWI’s = 200 DWI’s among college students nation-wide reported to be 18% Expected = 18% * 1000 = 180 n n SMR = 200 / 180 = 1. 11 Conclude that Margaritaville students have an 11% higher rate of DWI’s compared to US

Standardized Mortality Ratio n What information is needed to calculate Standardized Mortality Ratio? Age Category College Student n Study Population 1000 Age-Specific Mortality 18/100 Expected Deaths 180 SMR = Sum of observed deaths, 200 sum of expected deaths, 180

Comparing Standardized Mortality Rates Direct standardization yields an expected rate (or standardized rate) which can then be compared to the crude rate, or to any other similarly standardized rate. Indirect standardization yields an expected number of deaths, which can then be compared to the number of actual deaths, as in the SMR, or to the expected number of deaths in another population.

MNEMONIC DEVICE n n When you use the MORTALITY RATES of the POPULATION OF INTEREST, you are DIRECTLY standardizing. When you use the MORTALITY RATES of the STANDARD POPULATION, you are INDIRECTLY standardizing.

Direct and indirect age-adjustment of death rates? • These techniques can also be used to adjust for other variables - for example, sex and ethnicity. • These techniques can also be used to adjust incidence rates and prevalence rates. • Two age-adjusted rates can be used to create a standardized rate ratio comparing two populations – such a rate ratio will be free from confounding from age.

Choice of a standard population 1. An artificial population 2. One of the study groups * 3. The sum of the study groups 4. An external population such as the population of a state or a country

EXAMPLE-3

Comparing Mortality in Different Populations Crude Mortality Rates by Race in Baltimore, 1965 Race Mortality per 1, 000 Population White 14. 3 Black 10. 2

Comparing Mortality in Different Populations Death Rates by Age per 1, 000 Population, 1965 Race Crude <1 1 -4 5 - 17 18 -44 45 -64 > 65 White 14. 3 23. 9 0. 7 0. 4 2. 5 15. 2 69. 3 Black 10. 2 31. 3 1. 6 0. 6 4. 8 22. 6 75. 9 Why is the crude death rate higher in Whites?

Comparing Mortality in Different Populations Black White Age (yr) Pop. Deaths Death rate per 100, 000 All 900, 000 862 96 900, 000 1, 130 126 30 -49 500, 000 60 12 300, 000 30 10 50 -69 300, 000 396 132 400, 000 400 100 70+ 100, 000 406 200, 000 700 350

Age Group Standard Population Black Rate Expected No. of White deaths using “Black Rate rate” 30 -49 500, 000 + 300, 000 = 800, 000 12 (12/100, 000) x 800, 000 = 96 10 (10/100, 000) x 800, 000 = 80 50 -69 300, 000 + 400, 000 = 700, 000 132 (132/100, 000) x 700, 000 = 924 100 (100/100, 000) x 700, 000 = 700 70+ 100, 000 + 200, 000 = 300, 000 406 (406/100, 000) x 300, 000 = 1, 128 350 (350/100, 000) x 300, 000 = 1, 050 2, 238 1, 830 2, 238/1, 800, 000= 0. 001243* 100, 000 = 124. 3 1, 830/1, 800, 000= 0. 010167 * 100, 000 = 101. 7 Total number of deaths expected In the standard population age-adjusted rates: Crude rates: Adjusted rates: Expected No. of deaths using “White rate” 96/100, 000 Black 124. 3/100, 000 Black 126/100, 000 White 101. 7/100, 000 White

Direct Age Standardization Black Rate /100, 000 Expected White Rate deaths using black /100, 000 rate Expected deaths using white rate Age (yr) Standard Population All 1, 800, 000 30 -49 800, 000 12 96 10 80 50 -69 700, 000 132 924 100 70+ 300, 000 406 1, 218 350 1, 050 Total expected deaths 2, 238 Age-adjusted rates: • “black” = 2, 238/1, 800, 000 = 124. 3 • “white” = 1, 830/1, 800, 000 = 101. 7 1, 830

Direct / Indirect n n Both methods of standardization are nothing more than obtaining a weighted average of category-specific rates The weighting is derived from a standard population The difference lies in the source of the weights and rates If age adjustment does not explain the differences there is an unexpected difference

EXAMPLE-4

Suppose we wish to compare the mortality rates of Iran and Sweden for 1962 -hypothetical example: No. of Crude rate Population deathsper 1, 000 Sweden 7, 496, 000 73, 555 Iran 1, 075, 000 7, 871 9. 8 7. 3

Age specific rates Age Sweden 0 -29 30 -59 60+ No. % 3, 145, 000 3, 057, 000 1, 294, 000 Rate*Crude rate 9. 8/1, 000 42% 41% 17% Iran 0 -29 30 -59 60+ 1. 1 3. 6 45. 7 7. 3/1, 000 741, 000 275, 00026% 59, 000 5% 69% 5. 3 5. 2 41. 6 *per 1, 000

Age Probability Expected of death population (Iran) (Sweden) Population 0 -20 0. 0053 30 -500. 0052 60+ 0. 0416 TOTAL 3, 145, 000 16, 668. 5 3, 057, 000 15, 896. 4 1, 294, 000 53, 830. 4 7, 496, 000 86, 395. 3 78555! What is the age-adjusted mortality rate for Iran, using Sweden as the standard population?

Compare Crude - Adjusted Crude Mortality Rates: Sweden 9. 81/1000/year Iran 7. 32/1000/year Age-adjusted rate for Iran, using Sweden's age distribution as the standard: 11. 52 / 1 000 / year What does this mean? The age-adjusted death rate for Iran is the death rate Iran would have experience in 1962 if it had the population age distribution of Sweden. Direct Standardized Rate Ratio: 11. 5/ 9. 8 = 1. 2

EXAMPLE-5

City Alpha Beta a) b) c) Age Mid-point (years) population, 1995 Number of deaths in 1995 < 15 5000 6 15 -49 > 50 10000 20000 83 486 < 15 15 -49 > 50 8000 15000 13 120 25 Calculate the crude death rate for each city. Calculate the appropriate measure(s) to compare mortality in the two cities. Is there a substantial difference in mortality between the two cities?

THANK YOU