PrevMed Class of Dr. Chan

1. APPLIED EPIDEMIOLOGY Prepared by Antonio E. Chan, M.D.

13. Clinical Question Issue Question Abnormality Is the patient sick or well ? Diagnosis How accurate are tests used to diagnose disease ? Frequency How often does a disease occur ? Risk What factors are associated with an increased risk of disease ? Prognosis What are the consequences of having a disease ? Treatment How does treatment change the course of disease ? Prevention Does an intervention on well people keep disease from arising ? Does early detection and treatment improve the course of disease ? Cause What conditions lead to disease ? What are the pathogenetic mechanisms of disease Cost How much will care for an illness cost ?

19. MAJOR SIGNS MINOR SIGNS Weight loss > 10% Persistent cough > 1 month Fever > 1 month General pruritic dermatitis Chronic diarrhea > 1 month Recurrent herpes zoster General lymphadenopathy Chronic herpes simplex Oral candidiasis WHO CASE-DEFINITION FOR AIDS The presence of disseminated Kaposis sarcoma or cryptococcal meningitis or Two major signs in association with at least one minor sign

23. Level at which treatment does more good than harm - Cost In specific age groups for men and women at which treatment makes economic as well as medical sense Criteria change from time to time

27. Level at which treatment does more good than harm - Cost In specific age groups for men and women at which treatment makes economic as well as medical sense Criteria change from time to time

36. Factors influencing observed prevalence rate Increased by: Decreased by: Longer duration of the disease Shorter duration of disease Prolongation of life of patient High case-fatality rate from disease without cure Increase in new case Decrease in new cases (increase in incidence) (decrease in incidence) In-migration of cases In-migration of healthy people Out-migration of healthy people Out-migration of cases In-migration of susceptible people Improved cure rate of cases Improved diagnostic facilities (better reporting)

39. 274 CI = ------------ x 1000 = 2.3 per 1000 118,539 Example Relationship between cigarette smoking and incidence rate Stroke in a cohort of 118,539 women Never smoked 70 395,594 17.7 Ex-smoker 65 232,712 27.9 Smoker 139 280,141 49.6 Total 274 908,447 30.2 Person-years Stroke incidence rate Smoking No. of cases of observation (per 100,000 Category of stroke (over 8 years) person-years)

42. Standardization of rates (Adjustment of rates) 1. Direct adjustment of rates This requires the selection of some population, called a standard population , to which the age-specific rates for each population can be applied. 2. Indirect adjustment of rates Standardization is based on age-specific rates rather than age composition The population whose rates form the basis for comparison is referred to as the “standard population” The larger of the two populations is usually chosen as standard because its rates tend to be more stable

45. Example: Direct method Comparison of death rates in two populations by age Annual Annual Age-specific Number Crude Age Population Death rate of Death rate (years) Number Proportion (per 1000) Deaths (per 1000) (1) (2) (3) (4) (5) (6) Population A < 15 1,500 0.30 2 3 15 – 44 2,000 0.40 6 12 ≥ 45 1,500 0.30 20 30 45 All ages 5,000 1.00 45 --------- = 9.0 5,000 Population B < 15 2,000 0.40 2 4 15 – 44 2,500 0.50 6 15 ≥ 45 500 0.10 20 10 29 All ages 5,000 1.00 29 -------- = 5.8 5,000

46. Computation of Expected Number of Deaths by Direct Method Example 1 : Identical Age-specific Rates Population A Population B Age-specific Age-specific Age Standard Population Death Rate Expected Death Rate Expected (years) (A and B Combined) per 1000 Deaths per 1000 Deaths (1) (2) (3)=(2)x(1) (4) (5)=(4)x(1) < 15 3,500 2 7 2 7 15 – 44 4,500 6 27 6 27 ≥ 45 2,000 20 4 0 20 40 All ages 10,000 74 74 Conclusion : There is truly no difference between A and B in risk of death

47. Computation of Expected Number of Deaths by Direct Method Example 2 : Different Age-specific Rates Population A Population B Age-specific Age-specific Age Standard Population Death Rate Expected Death Rate Expected (years) (A and B Combined) per 1000 Deaths per 1000 Deaths < 15 3,500 2 7 2 7 15 – 44 4,500 6 27 10 45 ≥ 45 2,000 20 40 20 40 All ages 10,000 74 92 74 92 ---------- = 7.4 ---------- = 9.2 10,000 10,000 Conclusion : There is difference between A and B in risk of death

48. Example of Indirect Method Deaths by Age and Photofluorogram Reading (Whites) for Three-and-a-Half Year Observation Period, Muscogee County, Georgia, 1946 Negative for Cardiovascular Disease Suspect for Cardiovascular Age-specific Disease Age in 1946 Number of death rates Number of (years) Population Deaths per 100 Population Deaths 15 – 34 13,681 35 0.25 23 1 35 – 54 8,838 102 1.15 24 5 55 and over 2,253 149 6.61 65 14 ---------- ------- ------- ----- All ages 24,772 286 112 20 Crude death rate per 100 1.15 17.9

49. Percentage Distribution by Age of Negatives and Suspects, Muscogee County, Georgia 15 – 34 13,681 55.2 23 20.5 35 – 54 8,838 35.7 24 21.4 55 and over 2,253 9.1 65 58.0 All ages 24,772 100.0 112 99.9 Negative for Suspect for Cardiovascular Disease Cardiovascular Disease Age Percentage Percentage (years) Number of Population Number of Population

50. Calculation of Standardized Mortality Ratio for Suspects Compared with Negatives, Muscogee County, Georgia (1) (2) (3) = (1) x (2) (4) 15 – 34 23 0.25 .1 1 35 – 54 24 1.15 .3 5 55 and over 65 6.61 4.3 14 All ages 4.7 20 Death Rates per 100 Expected Deaths Observed for Persons Negative among “Suspects” Deaths Age Number of for Cardiovascular According to Rates among (years) “Suspects” Disease for Negatives “Suspects” Observed deaths 20 SMR = -------------------------- = --------- = 4.25 Expected deaths 4.7

51. No. of deaths in a year of children less than 1 year of age Infant mortality rate = ------------------------------------------------------ X F No. of live births in the same year A measure of overall health status for a given population It is based on the assumption that it is particularly sensitive to socio-economic changes and to health care intervention Other measures of mortality in early childhood are : 1. Fetal death rate 2. Stillbirth or late fetal death rate 3. Perinatal mortality rate 4. Neonatal mortality rate 5. Postneonatal mortality rate Mortality

52. Child mortality rate is based on deaths of children aged 1 – 4 years and is important because accidental injuries, malnutrition and infectious diseases are common in this age group Maternal pregnancy-related deaths in a year Maternal mortality rate = ------------------------------------- Total births in the same year Life expectancy is the average number of years an individual of a given age is expected to live if current mortality rates continue Mortality

54. DIAGNOSIS Clinical question: How accurate are tests used to diagnose disease ? Diagnostic test – the objective is to diagnose any treatable disease present Characteristics of a diagnostic test Reliable – gives the same measurement when repeated more than once Valid - measures what it intends to measure Accurate – correctly determines those with disease and those without Easy to use – can be performed by other people without difficulty Not expensive – affordable Safe and acceptable

56. Cut-off points 80 90 100 110 120 130 140 150 160 170 Normal Group Abnormal Group B l o o d L e v e l ( mg / 100 ml )

57. DIAGNOSIS Clinical question: How accurate are tests used to diagnose disease ? a + c b + d a + b + c + d a + b c + d DISEASE Present Absent TEST Positive a b Negative c d

58. Validity of a diagnostic test a = no. of true positives, b = no. of false positives c = no. of false negatives, d = no. of true negatives Sensitivity = probability of a positive test in people with the disease = a/(a + c) Specificity = probability of a negative test in people without the disease Positive predictive value = probability of the person having the disease when the test is positive = a /(a + b) Negative predictive value = probability of the person not having the disease when the test is negative = d / (c + d)

59. 62 87 37 112 149 Group A  -Hemolytic Streptococcus on Throat Culture Present Absent Clinical Diagnosis of Strep Pharyngitis Yes 27 35 No 10 77

63. Effect of Sequence is Serial Testing: A Then B versus B Then A Prevalence of Disease Number of patients tested 1000 Number of patients with disease 200 (20% prevalence) Sensitivity and Specificity of the Tests Test Sensitivity Specificity A 80 90 B 90 80 Sequence of Testing Begin with Test A Begin with Test B Disease Disease + - + - A + 160 80 240 B + 180 160 340 - 40 720 760 - 20 640 660 200 800 1000 200 800 1000 240 Patients Retested with B 340 Patients Retested with A Disease Disease + - + - B + 144 16 160 A + 144 16 160 - 16 64 80 - 46 144 180 160 80 240 180 160 340

65. Trade-Off between Sensitivity and Specificity when Diagnosing Diabetes Blood Sugar Level 2 hr after Eating Sensitivity Specificity (mg/100 mL) (%) (%) 70 98.6 8.8 80 97.1 25.5 90 94.3 47.6 100 88.6 69.8 110 85.7 84.1 120 71.4 92.5 130 64.3 96.9 140 57.1 99.4 150 50.0 99.6 160 47.1 99.8 170 42.9 100.0 180 38.6 100.0 190 34.3 100.0 200 27.1 100.0

67. Positive Test 0 20 40 60 80 100 100 80 60 40 20 0 Prevalence of Disease (Percentage) Predictive value (Percentage) Negative Test

86. LIKELIHOOD RATIO NOMOGRAM

96. DISEASE Clinical question: How accurate are tests to diagnose disease?

100. EARLY DIAGNOSIS NATURAL HISTORY OF DISEASE (FOUR STAGES) T I M E EARLY USUAL BIOLOGIC DIAGNOSIS CLINICAL ONSET POSSIBLE DIAGNOSIS OUTCOME Recovery Disability Death D X

101. EARLY DIAGNOSIS CRITICAL POINTS IN THE NATURAL HISTORY OF DISEASE 1 2 3 CP CP CP EARLY USUAL BIOLOGIC DIAGNOSIS CLINICAL ONSET POSSIBLE DIAGNOSIS OUTCOME Recovery Disability Death D X

104. EARLY DIAGNOSIS CRITICAL POINTS IN THE NATURAL HISTORY OF DISEASE Data modified from S. Shapiro. Evidence of screening for breast cancer from a randomized trial (Suppl.) 39:2772, 1977 Breast cancers diagnosed early in the Health Insurance Plan Study Age at diagnosis Percentage with positive axillary nodes 40-49 50-59 60+ Total Mode of early diagnosis Only by mammography 6 (19%) 27 (42%) 11 (31%) 44 (33%) 16% Only by clinical exam 19 (62%) 26 (40%) 14 (38%) 59 (45%) 19% Detected by both modes 6 (19%) 12 (18%) 11 (31%) 29 (22%) 41% 31 (100%) 65 (100%) 36 (100%) 132 (100%)

105. EARLY DIAGNOSIS CRITICAL POINTS IN THE NATURAL HISTORY OF DISEASE Some results of the H.I.P. randomized trial of early diagnosis in breast cancer Data modified from S. Shapiro. Evidence of screening for breast cancer from a randomized trial, Cancer(Suppl.) 39:2772, 1977 Deaths per 10,000 women per year From breast cancer From all other causes From cardiovascular disease 40-49 50-59 60-69 Control women 2.4 5.0 5.0 54 25 Experimental women 2.5 2.3 3.4 54 24

117. Types of epidemiological study (Observational studies) Study subjects With outcome Without outcome Population at risk Defined population Onset of study TIME No direction of inquiry Cross-sectional study (Prevalence study)

121. Types of epidemiological study (Observational studies) CASES (people with disease) CONTROLS (people without disease) Exposed Not exposed Exposed Not exposed Population direction of inquiry T I M E Design of a case-control study

128. Types of epidemiological study (Observational studies) Population People without the disease Exposed Not exposed disease no disease disease no disease direction of inquiry T I M E Design of a cohort study

129. Past Present Future Cohort Follow-up assembled Historical cohort Cohort Follow-up assembled Concurrent cohort

132. Table arrangement and formula for relative risk (RR) Disease No Disease Total Risk factor present A B A + B Risk factor absent C D C + D Total A + C B + D A / (A + B) RR = ----------------- C / (C + D)

137. Advantages and disadvantages of different observational study designs Probability of: selection bias NA medium high low recall bias NA high high low loss to follow-up NA NA low high confounding high medium medium low Time required low medium medium high Cost low medium medium high Ecological Cross- Case- Cohort sectional control

138. Applications of different observational study designs Investigation of rare disease ++++ - +++++ - Investigation of rare cause ++ - - +++++ Testing multiple effect of + ++ - +++++ cause Study of multiple exposure ++ ++ ++++ +++ and determinants Measurements of time ++ - + +++++ relationship Direct measurement of - - + +++++ incidence Investigation of long - - +++ - latent periods Ecological Cross- Case- Cohort sectional control

139. Types of epidemiological study (Experimental studies) Non-participants Do not meet Selection criteria Potential participants Participants Non-participants Control Treatment Study population Randomization Invitation to participate Selection by defined criteria Design of a randomized clinical trial

142. Relative ability of different types of study to “prove” causation Type of study Ability to “prove” causation Randomized controlled trials Strong Cohort studies Moderate Case-control studies Moderate Cross-sectional studies Weak Ecological studies Weak

150. Cause Clinical question: What conditions lead to disease ? What are the pathogenetic mechanisms of disease ? INCREASED SUSCEPTIBILITY INGESTION OF CHOLERA VIBRIO CHOLERA Causes of cholera Exposure to contaminated water Effect of cholera toxins on bowel wall cells Genetic factors Malnutrition Crowded housing Poverty Risk factors for cholera Mechanisms for cholera

152. SUFFICIENT CAUSES U A B U A E U B E I II III

161. SUSCEPTIBLE HOST INFECTION TUBERCULOSIS Exposure to Mycobacterium Tissue Invasion and Reaction Crowding Malnutrition Vaccination Genetic Risk Factors for Mechanisms of Tuberculosis Pathogenesis Tuberculosis Distant from Outcome Proximal to Outcome Causes of tuberculosis

167. Relationship between cigarette smoking and incidence rate of stroke in a cohort of 118,539 women Never smoked 70 395,594 17.7 Ex-smoker 65 232,712 27.9 Smoker 139 280,141 49.6 Total 274 908,447 30.2 Smoking Person-y ears Stroke incidence rate category No. of cases of observation (per 100,000 of stroke (over 8 years) person-years)

176. Could it be due to selection or measurement bias Could it be due to confounding? Could it be a result of chance? Could it be causal? Apply guidelines and make judgment ASSESSING THE RELATIONSHIP BETWEEN A POSSIBLE CAUSE AND OUTCOME No No Probably not

180. Treatment Clinical question: How does treatment change the course of disease? DECIDING ON THE BEST THERAPY

199. Clinically relevant outcomes in a randomized trial of clofibrate in the prevention of coronary heart disease PLACEBO CLOFIBRATE Average change in serum cholesterol (%) +1 - 9 Non-fatal myocardial infarctions per 1000 subjects 7.2 5.8 Fatal and nonfatal myocardial infarctions per 1000 subjects 8.9 7.4 Total deaths per 1000 subjects 5.2 6.2

204. Comparing the conclusions drawn from a clinical trial with the true state of affairs w x y z TP=true positive; FP= false positive; FN= false negative; TN= true negative THE CLINICAL TRIAL IS THE DIAGNOSTIC TEST The true state of affairs Drug A is better than drug B Drug A is no better than drug B Conclusion drawn from a clinical trial Drug A is better than drug B TP Correct FP Error Drug A is no better than drug B Error FN Correct TN

205. Naming the erroneous conclusions from a clinical trial w x y z The true state of affairs Drug A is better than drug B Drug A is no better than drug B Conclusion drawn from a clinical trial Drug A is better than drug B TP Correct (1-  = power) FP Type I error (risk of making this error=  =P value) Drug A is no better than drug B Type II error (risk of making this error=  ) FN Correct TN

209. Occurrence of death, stroke, or other major complications How might these benefits be expressed in terms of clinical significance ? Patient status at entry Adverse event rates Placebo P Active RX A Prior target organ damage .22 .08 No prior organ damage .10 .04

210. Occurrence of death, stroke, or other major complications These relative risk reductions mean that the risk of death, stroke, or other complications of hypertension was reduced by almost two-third through active treatment Patient status at entry Adverse event rates Relative Risk Reduction RRR Placebo P Active A (P – A) -----------= RRR P Prior target organ damage .22 .08 .22 - .08 ----------- = 64% .22 No prior organ damage .10 .04 .10 - .04 ----------- = 60% .01

212. Occurrence of death, stroke, or other major complications The decimal form of absolute risk reduction is foreign to most clinicians Patient status at entry Adverse events Absolute risk reduction ARR Placebo P Active A RRR P – A = ARR Prior target organ damage .22 .08 64% .22-.08=.14 No prior organ damage .10 .04 60% .10-.04=.06

214. Occurrence of death, stroke, or other major complications Patient status at entry Adverse events Number Needed to Treat (NNT) Placebo P Active A RRR ARR 1 ----- = NNT ARR Prior target organ damage .22 .08 64% .14 1 ---- = 7 .14 No prior organ damage .10 .04 60% .06 1 ---- = 17 .06

215. The effect of different baseline risks and relative risk reductions on the number needed to treat Baseline risk (with no treatment) Relative risk reduction on treatment 50% 40% 30% 25% 20% 15% 10% .9 .6 .3 2 3 7 3 4 8 4 6 11 4 7 13 6 8 17 7 11 22 11 17 33 .2 .1 .05 10 20 40 13 25 50 17 33 67 20 40 80 25 50 100 33 67 133 50 100 200 .01 .005 .001 200 400 2000 250 500 2500 333 667 3333 400 800 4000 500 1000 5000 667 1333 6667 1000 2000 10000

219. Was the trial big enough to show a relative risk reduction of  25% if it had occurred ? Observe rate of events in the experimental group .95 .90 .85 .80 .75 .70 .65 .60 .55 .50 .45 .40 .35 .30 .25 .20 .15 .10 .05 Observed rate of events in the control group .95 .90 .85 .80 .75 .70 .65 .60 .55 .50 .45 .40 .35 .30 .25 .20 .15 .10 .05 14 27 68 391 11 18 38 110 1057 14 25 54 185 4889 11 18 33 78 326 13 22 44 112 635 11 16 28 57 165 1524 13 20 35 75 250 6349 10 15 24 43 99 402 12 17 28 53 132 722 13 20 33 65 180 1607 10 15 22 38 79 254 11 16 25 44 98 381 12 18 28 50 121 634 13 19 30 57 1296 10 13 20 33 64 196 4537 10 14 20 34 71 261 10 14 20 35 78 371 10 13 20 34 80 589 12 17 30 74 1245

220. Was the trial big enough to show a relative reduction of  50% if it had occurred ? Observed rate of events in the experimental group .70 .65 .60 .55 .50 .45 .40 .35 .30 .25 .20 .15 .10 .08 .06 .04 .02 Observed rate of events in the control group .98 .95 .90 .85 .80 .75 .70 .65 .60 .55 .50 .45 .40 .35 .30 .25 .20 .15 .10 .08 .06 .04 .02 14 24 50 165 5803 12 19 37 102 921 14 26 58 236 12 19 38 108 995 10 15 27 63 256 12 21 41 116 1059 16 29 66 268 13 22 43 120 1082 11 17 30 68 270 13 22 42 119 1059 11 17 30 66 260 13 22 42 113 987 11 16 28 62 239 13 20 38 102 867 10 15 26 55 205 12 18 33 86 699 13 22 45 160 10 15 26 64 482 11 17 32 102 254 2017 14 25 66 131 453 12 20 44 76 179 1313 10 16 31 47 87 274 12 22 30 47 97 561

224. CONFIDENCE INTERVAL IN A “NEGATIVE” RANDOMIZED TRIAL Control (Placebo) Experimental (aspirin) Patients with recurrent strokes n c = 252 p c = .07 n E = 253 p E = .09 Absolute risk reduction = p c – p E = .07 - .09 = -.02 Relative risk reduction = (p c – p E ) / p c = .02 / .07 = -29%

225. CONFIDENCE INTERVAL IN A “NEGATIVE” RANDOMIZED TRIAL = = From to

244. Outcomes of Disease (the Five Ds) Death A bad outcome if untimely Disease A set of symptoms, physical signs, and laboratory abnormalities Discomfort Symptoms such as pain, nausea, dyspnea, itching, and tinnitis Disability Impaired ability to go about usual activities at hoe, work, or recreation Dissatisfaction Emotional reaction to disease and its care, such as sadness or anger