HSS4303B – Intro to Epidemiology Jan 21, 2010 – Standardization

1. HSS4303B – Intro to Epidemiology Jan 21, 2010 – Standardization

2. Survival

3. Survival rates • _____________ is the probability of remaining alive for a specific length of time • 1 year and 5 year survival are often used as indicators of the severity of disease and the prognosis • 5 year survival rates for myelocytic leukemia is about 0.14, indicating that about 14% of the patients with acute myelocytic leukemia survive for at least 5 years after diagnosis. • Survival (S) = (A – D) / A where • A is the number of newly diagnosed patients under observation and D is the number of deaths observed in a specified period of time

5. Observation of each patient begins at diagnosis (time = 0), and continues until one of the following outcomes occurs: death, survival for 5 years, or follow-up ceases (the subject is "censored"). A patient is censored when follow-up ends prior to death or completion of a full period of observation. Follow-up could end for one of several reasons: (1) the patient decides to discontinue participation, (2) the patient is "lost" to follow-up, or (3) the study ends. Five of the six people under observation (N = 6) survive at least 2 years. Thus, the 2-year survival is 5/6= 0.83= 83%

6. Specifying Length of Survival • 1, 2, 5 years are standard, but it can be anything • For example, prostate cancer has a much higher one year overall survival rate than pancreatic cancer, and thus has a better prognosis.

7. Relative Survival Rate • RSR • Ratio of survival rate of disease in question, divided by the survival rate of the general population • Eg RSR of cancer at 2 years = (% of cancer patients who are alive @ 2 years) / (% of general population who are alive @ 2 yrs) • Why is this important? Eg, The overall 5-year relative survival rate for 1999-2005 from 17 SEER geographic areas was 89.1%. Five-year relative survival rates by race were: 90.3% for white women; 77.9% for black women.

8. Cause-Specific Survival Rate • Cause-specific survival (CSS) is a term that denotes the chances of death due to a particular condition (or cause) at a particular point of time. It takes care to exclude death due to unrelated causes in patients suffering from the cancer in question. This means that 15% of these patients are estimated to die directly due to the Hodgkin disease by 5 years. The remaining 85% are either alive or have died due to other unrelated causes. The 5-year cause-specific survival for stage IIA Hodgkin lymphoma is 85% when treated with ABVD followed by involved field radiation.

9. Problems with mortality data • Information on mortality is obtained from death certificates • Deaths are coded according to the underlying cause • Which is defined as the disease or injury which initiated the train of morbid events leading directly or directly to death or circumstances of accident or violence which produced the fatal injury • The underlying cause therefore excludes information pertaining to the immediate cause of death, contributory causes and those that intervene between the underlying and immediate causes of death • Some causes of death have better validity than others

11. Cause of death from a death certificate

12. ICD classification on death certificates • Deaths are coded by ICD classification • Drop in diabetes related deaths in 1949 were caused by changes in classification codes • Prior to 1949 the policy was to include diabetes as cause of death anywhere on the certificate lead to diabetes being mentioned on the death certificate • After 1949 only death certificates on which the underlying cause of death was listed as diabetes were coded as a death from diabetes

13. Changes in death rates from diabetes caused by changes in classification

14. Changes in the definition of disease • In 1993, a new definition of AIDS was introduced • These changes resulted in a rapid rise in the number of reported cases

15. AIDS cases in the US from 1984-2000

16. Causes of death in the early 20th century

17. Standardization

18. Consider the following… • There is concern that the nuclear power plant in Raywatville (which is a beachfront community in Florida) is causing cancer. • Death rates due to cancer are computed and compared to death rates due to cancer in a similarly-sized control community –Gomesland– in Alaska. • The rates in Raywatville are indeed higher than the rates in Gomesland • What can you conclude?

19. What is this an example of? Con…..

20. So What Can We Do To Eliminate the Uncertainty?

21. Standardization • A way to make two populations more comparable by adjusting one or both of them to conform to an external standard • Age • Sex • etc

22. Age-Adjusted Rates • if you want a measurement of mortality that can be used either to compare different populations (states, counties, cities, etc.) or to compare the mortality experience over time for one area with a changing population, it is advisable to adjust or standardize the effects of such factors as age and/or sex in these groups. • Age is the most commonly used adjustment factor • Thus age-adjusted rates are the most common form of standardization

23. Standardization • First step is to define a reference population, which serves as the “standard” against which the test populations will be set • Often select the national census • Sometimes select one of the two test populations to be the standard

24. Two Kinds of Age Standardization • Direct • Indirect • estimates the rate that would have been observed if the study population had had the same age structure as the reference group • computes the number of cases of disease that would have been expected if the disease rates from the reference population had applied in the study population. http://www.paho.org/English/SHA/be_v23n3-standardization.htm

25. Two Kinds of Age Standardization • Direct • Indirect • commonly used in reports of vital statistics (e.g., mortality) or disease incidence trends (e.g., cancer incidence). Invented in 1899. • Commonly used in studies of occupational disease or studies of place and time-limited environmental catastrophes. Can be computed from SMR (standardized mortality ratio). Invented in 1844. http://www.paho.org/English/SHA/be_v23n3-standardization.htm

26. There are three major components that are needed to perform adjusted mortality rate calculations: the number of deaths the population a "standard" population

27. Direct Standardization

28. Raywatville Gomesland Reference Pop.

29. The steps • We are going to standardize each population (Raywatville and Gomesland) individually against the reference (standard) population • Multiply the age-specific rate for the test population against the size of standard population in each age stratum • Add ‘em all up and divide by the total number of people in the standard population

30. Directly Standardized Rate for Raywatville (Miami) is…

31. Directly Standardized Rate for Gomesland (Alaska) is…

32. Compare Age-Standardized Rates • Raywatville = 6.92 deaths/ 1000 people • Gomesland = 6.71 deaths/ 1000 people Before age-adjustment: • Raywatville = 8.92 deaths/ 1000 people • Gomesland = 2.67 deaths/ 1000 people ? After age-adjustment:

33. Some Thoughts on Direct Standardization • Because we are multiplying age-specific rates by the age-specific populations in the standard population, the final age-adjusted rate is a weighted average, with the weights being the proportion of people in each age stratum of the standard population • We have a term for each such product: expected • For direct standardization, the expected number of deaths is what we get when we multiply the known rate by the standard population • The observed number of deaths is the actual number of people who died

34. Choice of Standard Pop • We chose the US national survey as an objective standard population • What if we had simply chosen to adjust the Gomesland rates by applying the Raywatville population as a standard? • Shall we try?

35. Gomesland Raywatville

37. Directly standardized rate = 4858262.2 / 562887 = 8.63 deaths / 1000 pop

38. Compare Age-Standardized Rates • Raywatville = 8.92 deaths/ 1000 people • Gomesland = 8.63 deaths/ 1000 people Before age-adjustment: • Raywatville = 8.92 deaths/ 1000 people • Gomesland = 2.67 deaths/ 1000 people ? After age-adjustment:

39. Your Homework Assignment #1 • Compute Age-Adjusted Rates in the previous example, using Gomesland as the standard population, rather than Raywatville • Then compute Age-Adjusted Rates using the hybrid population of “Gomesland + Raywatville” as the standard population • How do your conclusions differ?

40. Homework Assignment #2 Compute age-adjusted mortality rates for the early and later stages of this hypothetical disease, using a combined population (early + later) as the standard population.

41. Homework Assignment #3 Fill in the table and compue adge-adjusted rates for both Mexico and the USA, using the given standard population.

42. Indirect Age Standardization • Involves the computation of a Standardized Mortality Ratios (SMR) • SMR= observed number of deaths per year expected number of deaths per year • For indirect standardization, the expected number of deaths is what we get when we multiply the standard rate by the sample population • The observed number of deaths is the actual number of people who died

43. SMR • SMR is a ratio, therefore is given simply as a number or as a percentage, but has no units • The ratio of observed to expected deaths

44. SMR for tuberculosis

45. SMR for tuberculosis

46. Compute SMR for Occupation A:

47. Compute SMR for Occupation A: SMR = (observed deaths) / (expected deaths) = 22 / 16 = 1.38

48. Compute SMR for Occupation B:

49. Compute SMR for Occupation B: SMR = (observed deaths) / (expected deaths) = 14 / 8 = 1.75

50. Using SMR to compute age-adjusted rate Indirectly standardized rate = SMR x [crude death rate in the standard pop]