epidemiology 4: Measures of Mortality Direct & Indirect Age Adjustment Validity & Reliability Prognosis

Percent Agreement Observed

(a + d) / (a+b+c+d) * 100 N = (a+b+c+d) [(a) / (a+b+c)] * 100 Sometimes another approach for calculating percent agreement is to exclude cell 'd', which reflects the complete agreement for negative findings on the part of both observers Calculated to determine if the high number of concordant negative findings masks significant disagreement between observers in the identification of patients positive for a disease

questions assked about SMR

. So we are able to answer our 1st question, "How many deaths would we expect in each age-group and overall IF the individuals in Population A and Population B had the same mortality experience (i.e. experienced the same mortality rate) as the individuals in the same age-group of the STANDARD POPULATION?" So we are able to answer our 2nd question, "Is the overall mortality experience of Population A and Population B higher than that for the individuals in the STANDARD POPULATION?"

Guidelines for Interpreting Kappa

>0.75 Excellent Agreement 0.40-0.75 Intermediate/Good Agreement <0.40 Poor Agreement

Adjusted Rates

A rate whereby some statistical adjustment has been calculated to produce an adjusted or standardized rate when it is important to 'remove the effect' of some characteristic Most frequently applied for age differences, but can be applied for any characteristic (gender, race/ethnicity, socioeconomic status

Relevance of False Positive/False

Burden of false positive results Return for more sophisticated or invasive tests Health care system - more expensive Patient - greater anxiety & concern Burden of false negative results Dependent on nature & severity of disease Intervention interval may be missed completely or delivered too late

indirect age adjustment

Calculate the Standardized Mortality Ratio (SMR) based on: [Observed Number of Deaths per year / Expected Number of Deaths per year]

direct age adjustment (more in depth)

Calculate the age-adjusted mortality rate based on: [Total Expected Number of Deaths for each population / Total Standard Population] Age-adjusted rate Mexico 639/100,000 or 6.39/1,000 United States 574/100,000 or 5.74/1,000

Approaches for Describing Prognosis

Case-fatality [Number of persons dying during a specified time period after a specific disease onset or diagnosis / Number of persons with the disease] X 100 Best used to evaluate short-term acute conditions as measurement becomes difficult the longer a disease progresses or the further out from diagnosis 5-year survival Observed survival Life table approach - based on pre-determined intervals Kaplan Meier - based on exact points of withdrawal for death, etc. Median survival time Preferred over mean survival because it is: Less affected by extremes Able to use fewer deaths than data based on mean survival (i.e. calculation can be made on the deaths of half the group Relative survival Compares survival of a group to what is expected if they did NOT have the disease

cohort and cohort effect

Cohort A group of individuals followed or traced over a period of time Cohort Effect relates back to a cohort Most commonly refers to a 'cohort' of individuals identified at a specific period of time & followed over part or all of their life span Sharing of a common temporal experience, such as birth year Sharing may also refer to common societal experiences or exposures Reflects how major social, cultural, or political influences & events influences a cohort's perspective May be associated with disease/illness patterns Is a change in mortality due to a 'cohort effect'? Important to decisions that impact health policy

crude mortality rate (more in depth)

Considerations when comparing crude mortality rates Not appropriate for comparison of different populations because age significantly impacts mortality data & is dependent on the population If the numerator uses the sum of the number of deaths across multiple years, the denominator should use the sum of the population over the same time-period If the average annual number of deaths is reported in the numerator, either the average annual population or the population in a single year in the middle of the time-period should be used in the denominator

What are the Two Approaches for Age Adjustment: Direct Method

Direct Method A standard population is used to eliminate the effects of age between populations 'holds constant' age composition of 2+ populations so a comparison of rates can be made in which the possible effect of age is eliminated Direct adjusted rates do not reflect true mortality risk because the: age-specific rates are actual rates in the population, but standard population is hypothetical Most frequent method for comparing rates across different populations

Data Required for Direct & Indirect Approach for Age Adjustment: direct method

Direct Method (Standardized Rate) Age-specific rates in the 'index' population under study Size of 'standard' population in each age group Total number of deaths (could be cases of disease) in the 'standard' population

crude mortality rate (more in depth)

General indicator of health status of a population Reflects differences In the 'force of mortality' or 'hazard rates' or the 'risk of death Age composition of population

Cross Tabulation of a Clinical Test with Disease Status

In "clinical" epidemiology, investigators have developed methods of determining the performance of various criteria for detecting disease. The process of detection is often referred to as "screening". If one had a "perfect" measure of the true presence of a disease, one could measure the performance of a particular diagnostic test or set of criteria by comparing its ability to discriminate people with disease (true positives) from those without disease (true negatives), against the error rates (false positives and false negatives). This slide shows how the clinical epidemiologist sets up a four-fold table to make this comparison. The performance of the test is measured by two different statistics derived from the table: Sensitivity (ability of the test to detect "true positives"); Specificity (ability of the test to detect "true negatives"). Additional statistics are the "positive predictive value" and "negative predictive value". true positive disease and test positive (a) ---------------------- false negative disease but test negative (c) ---------------------false positive no disease but test positive (b) ------------------- true negative no disease and test negative (d) = sensitivity a/(a+c) = specificity = d/(b+d)

What are the Two Approaches for Age Adjustment: Indirect Program

Indirect Method A standard population is used to eliminate the effects of age between populations Age-specific rates for known standard population are applied to each age-group in the index population Most frequently used when numbers of deaths for each age-specific stratum are unknown Most frequent method to evaluate mortality for an occupationally exposed population compared to the general population

Data Required for Direct & Indirect Approach for Age Adjustment: indirect method

Indirect Method (Indirect Standardized Mortality ratio (SMR)) Age-specific rates in a 'standard' population Size of the 'index' population in each age group Total number of deaths (could be cases of disease) in the 'index' population

What Contributes to Reliability?

Intersubjective variation Time of day, season of year Conditions or environment Intraobserver variation Time of day Subjective level of coding Interobserver variation Lack of a clear operational definition, which makes it difficult for an objective set of rules to be followed allowing for increased subjectivity Level of experience Conditions or environment may influence intrasubject variation related for example to a blood-draw which may vary by time of day.

kappa statistic

Kappa = (PAO - PAE) / (1 - PAE) PAO = [(a + d) / (a+b+c+d)] * 100 PAE = ((a + b)*(a+c) /N + ((b+d)*(c+d)) /N) /N or PAO, Percent Agreement for Observed PAE, Percent Agreement for Expected N = (a+b+c+d)

A Reliability Study

Objective To assess intra- and interobserver variability in the measurement of aortic and common iliac artery diameter as measured by computed tomography (CT) in subjects with normal & abnormal aortas Design Reproducibility study to assess results measured by 3 radiologists with varying experience Methods Measurements taken of aortic diameter at multiple areas using CT Results Intra-observer variability varied between radiologists but in general was 94% Inter-observer variability was 82% Both intra-observer and inter-observer variability increased with increasing vessel diameter & were largest in patients with abdominal aortic aneurysms Conclusions Interobserver variability of CT measurements of aortic & common iliac artery diameter is not negligible & should be taken into account when making clinical decisions When assessing change in aortic diameter, previous CT-scans should be reviewed simultaneously as a routine to exclude interobserver variability.

Percent Agreement (Concordance Between Two Categorical Variables)

Percent agreement provides a measure of interobserver variation Objective Maximize A & D Minimize c & b

Predictive Values of a Test & Measure of Screening Program Feasibility

Positive predictive value Given that screening is positive, what is the probability that a patient has the disease? (Number of people who test positive & have disease / Number of people who test positive for disease) * 100 Negative predictive value Given that screening is negative, what is the probability that a patient does NOT have the disease? (Number of people who test negative & do NOT have disease / Number of people who test negative for disease) * 100

Factors Associated with Positive Predictive Value

Prevalence of disease in tested population Higher or increased prevalence associated with gain in predictive value Therefore, screening programs are most productive & efficient among high-risk target populations because disease prevalence will be higher Specificity of the test when the disease is infrequent Increase in specificity associated with gain in predictive value This is true because the majority of the population does not have disease or as Gordis notes, they fall to the right of the vertical line in the 2 x 2 table, therefore when prevalence is low even a small increase in the specificity of a test (correctly identifying those without disease) will yield a greater gain in predictive value than an increase in the sensitivity because there are so many more people withou the disease than those with the disease

What is Prognosis?

Prognosis refers to the course of a disease or condition & is a description or prediction for the likely course of the condition Timepoint Sometimes hazy because actual date of diagnosis may not be clear Endpoints Death Diagnosis to recurrence Time to functional impairment Time to change in quality of life

Why Are Quality of Life Measures (QOL) Important?

QOL attempts to measure the 'total' impact of a disease on a patient Used to determine if specific treatments enhance or detract from a patient's QOL Used to establish priorities for scarce health resources

direct age adjustment (more in depth) (rate ratio)

Rate ratio of rate for Mexico to rate for U.S. 6.39/5.74=1.11 Mexico's death rate is 11% higher than

What Is Burden of Disease?

Refers to the mortality, morbidity, & resource cost of care & treatment Main interest is how the burden of different diseases compare Is heart disease a bigger killer than cancer? Which diseases afflict women more often than men? Which diseases impact the cost of health care

examples of SMR / what does SMR mean

SMR for Population A Observed=120 Expected=195.5 SMR= 120/195.5 or 0.61 * 100 = 61 SMR for Population B Observed=30 Expected=69.5 SMR= 30/69.5 or 0.43 * 100 = 43 Notice that the Standard Population Age-specific Mortality Rate/1,000 is used for both Population A and Population B when multiplying by the stratum-specific population estimate (this is the actual size of the population within each age group) of each of the 2 index populations (A and B). The calculated result = the 'Expected Deaths in the Index Population' for each age-group

SMR (standardized mortality ratio)

SMR for Population A Observed=120 Expected=195.5 SMR= 120/195.5 or 0.61 * 100 = 61 SMR for Population B Observed=30 Expected=69.5 SMR=30/69.5 or 0.43 * 100 = 43 Notice that the Standard Population Age-specific Mortality Rate/1,000 is used for both Population A and Population B when multiplying by the stratum-specific population estimate (this is the actual size of the population within each age group) of each of the 2 index populations (A and B). The calculated result = the 'Expected Deaths in the Index Population' for each age-group Remember that the point of the standardization (regardless of whether direct or indirect) is to eliminate the effects of age between populations. In this case we use the indirect because we have: We do not have the age-specific stratum deaths So, we apply the age-specific rates for known population for each age-group in the Index Population

Standardization of Rates

Standardization methods adjust for the effects of age (or other factors) in the comparison of rates between populations Standardized rates can be used to compare the 'burden of disease' across populations Ratio of 1 indicates that populations being evaluated are similar in terms of the disease under study Standard population varies depending on the populations to be compared National population for evaluation of geographic regions World population, Continent population, Country population Sum of populations under examination Choice of standard population is important because different results may occur

Direct Age Adjustment

Steps for direct standardization Evaluate the age-specific rates to ensure there are no major differences across age strata Select a 'standard' population for adjustment For mortality statistics, the year 2000 U.S. population is commonly used for U.S. data Apply the specific rate from each age-specific stratum to the stratum-specific standard population to calculate the expected number of deaths

Indirect Age Adjustment

Steps for indirect standardization Total the number of observed deaths in each age group of the index population Select a 'standard' population to use age-specific rates for adjustment Apply the age-specific rate from the standard to the stratum-specific population estimates of the index population of interest to calculate the expected number of deaths Calculate the SMR based on the ratio of observed number of deaths per year to the expected number of deaths per year

reliability

The degree of stability exhibited when a measurement is repeated under identical conditions Also referred to as reproducibility or repeatability

validity

The degree to which a method measures what it purports to measure For screening tests it is the ability of a test to distinguish between those who have a condition and those who do not Sensitivity is the ability of a test to identify correctly those who have a disease or condition (Number of people who test positive / Number of people with disease) * 100 Specificity is the ability of a test to identify correctly those who NOT have a disease or condition (Number of people who test negative / Number of people disease-free) * 100

Balance Between False Negative & False Positive

The false negative value is kept low, especially for a disease that has major consequences if the appropriate intervention time period is missed. The false negative value is kept low by increasing the specificity.

Adjusted Mortality Rate

•The adjusted rate is calculated by using a summarizing procedure to reduce the effect of age or some other factor due to differences in rates across

Crude Mortality Rate

•[Number of TOTAL deaths from all causes for specified time period / Total population for same geographic area for specified time period] X 100,000 •Denominator represents person-years at risk