COPHP: Quantitative Methods - Week 1

COPHP: QUANTITATIVE METHODS Winter Quarter Week 1 Topics What is epidemiology? Measures of occurrence Absolute measures of excess risk Outbreak Investigation 101 What is Epidemiology? “study” “upon” epidemiology Who? When? Where? Why? “people” How? Language matters Tips for learning epi Ask question...

COPHP: QUANTITATIVE METHODS Winter Quarter Week 1 Topics What is epidemiology? Measures of occurrence Absolute measures of excess risk Outbreak Investigation 101 What is Epidemiology? “study” “upon” epidemiology Who? When? Where? Why? “people” How? Language matters Tips for learning epi Ask questions Practice Topics What is epidemiology? Measures of occurrence Absolute measures of excess risk Outbreak Investigation 101 Measures of Occurrence Measure Description Possible Value Example Count Ratio Number Division of two unrelated numbers Divisions of two related numbers; numerator is a subset of the denominator Division of two numbers; time is always in the denominator 0–infinity 0–infinity Bicycle accidents since 2010 Bicycle accidents per bicycles sold Bicycle accidents per bicyclist Proportion Rate 0–1 or 0%–100% 0–infinity per unit of time Bicycle accidents per year Count Number of people with event (e.g., disease, injury, etc.) Difficult to compare populations of unequal size or over different intervals Useful for rare events Count COVID cases in the U.S. Motor vehicle collisions in Washington State Lung cancer cases in an HMO Ratio Numerator & denominator represent different populations Does not account for population size Difficult to compare populations over different intervals Ratio Men to women Bicycles to people Swimming pools to students Proportion Numerator is part of denominator # cases per # persons in population (e.g., %, per 1000 persons, etc.) Accounts for size of at-risk population over specific time period Can be compared across different population & time periods Useful when onset difficult to measure Proportion 35% of Washington households own a firearm 73% of King Country adults have a college degree 4 out of 5 dentists recommend… Rate Numerator & denominator have different units (denominator always includes time) # cases per population time at-risk Expressed as # cases per person-time (e.g., per 100 persons-years) Accounts for size of at-risk population over time Can be compared across different populations & time periods Indicates risk of becoming affected over time, so useful to estimate future burden of disease Rate Miles per hour Heart attacks per month Drownings per year Special Cases Epidemiology Jargon Measures of Disease Frequency Prevalence Incidence Proportion Proportion within a given time period or rate Covers a snapshot in time More commonly used in chronic disease outcomes Flexible in duration/time period covered Better for injury event outcomes Prevalence Prevalence = # 𝑜𝑓 𝑝𝑒𝑜𝑝𝑙𝑒 𝑎𝑓𝑓𝑒𝑐𝑡𝑒𝑑 𝑖𝑛 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 # 𝑜𝑓 𝑝𝑒𝑜𝑝𝑙𝑒 𝑖𝑛 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 x 10n (i.e., per 10n people) Proportion Numerator is part of the denominator Expressed as number per persons (e.g., per 1000 people or %) The probability of a case being detected during a prevalence survey depends on the duration of disease Prevalence 60% of King County residents are vaccinated for COVID 15% of pregnancies have a miscarriage 3 per 10,000 people experience surgical complications Cumulative Incidence Estimates the risk of an event # 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠 𝑖𝑛 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 CI = # 𝑎𝑡−𝑟𝑖𝑠𝑘 𝑝𝑒𝑜𝑝𝑙𝑒 𝑖𝑛 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 over a specified time period Proportion without dimensions Numerator is part of the denominator Expressed as number per persons (e.g., per 1000 people or %) Time is not part of the denominator, it just describes the period of observation Cumulative Incidence There were 74 heart attacks per 10,000 adults in the U.S. between 1990–1999 10 people out of 1000 on a week-long cruise ship got norovirus Of the ~70,000 residents remaining in New Orleans for the month following Hurricane Katrina, approximately 640 experienced a joint sprain or fracture Incidence Rate Estimates the risk of an event # 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠 𝑖𝑛 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 IR = # 𝑎𝑡−𝑟𝑖𝑠𝑘 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑥 𝑡𝑖𝑚𝑒 Expressed as a Rate Numerator is number of observed events Denominator is number of people observed * duration of observation Very useful for comparing groups that are observed separately Incidence Rate 17 suicides per 100,000 person-years 8 collisions per 1,000 driver-miles 1.2 injuries per 10,000 worker-hours How Numbers Relate Putting it All Together Incidence and Prevalence Incidence—New Cases Prevalence—Current Cases Former Cases—Death or Recovery Incidence and Prevalence Prevalence = Incidence x Duration Infectious disease with short duration High incidence, steady prevalence Chronic disease with long duration Lower incidence, high prevalence Topics What is epidemiology? Measures of occurrence Absolute measures of excess risk Outbreak Investigation 101 Excess Risk Differences in risk tell us what could potentially be changed to improve health outcomes Want to quantify the magnitude of the difference Helps us prioritize potential interventions Measures incorporate statistical and clinical significance Can be measured on absolute or relative scales Type of Measure ABSOLUTE What is the difference in disease risk in the exposed group compared to unexposed groups? Example: Rate of pedestrian injury is 7 per 10,000 higher for crossings happening without a crosswalk. RELATIVE How much/how many times higher is the risk for disease in the exposed group compared to unexposed groups? Example: Rate of pedestrian injury is 4x as common in unmarked crossings compared to crosswalks Absolute Measures Measured as a difference in rates (risk difference) or prevalence (prevalence difference) Helpful when looking at common outcomes or describing to policy makers Calculated: Directly in cross-sectional, cohort, or RCTs Inferred from case-control studies RD = Ie - Iu Relative Measures of Excess Risk We may want to better understand the risk of disease in one group relative to another group Relative Risk/Risk Ratio/Prevalence Ratio RR/PR = 𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑜𝑟 𝑃𝑟𝑒𝑣𝑎𝑙𝑒𝑛𝑐𝑒 𝑖𝑛 𝐺𝑟𝑜𝑢𝑝 𝐴 𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑜𝑟 𝑃𝑟𝑒𝑣𝑎𝑙𝑒𝑛𝑐𝑒 𝑖𝑛 𝐺𝑟𝑜𝑢𝑝 𝐵 OR/Odds Ratios are a similar, but slightly different relative measure More details next week… Topics What is epidemiology? Measures of occurrence Absolute measures of excess risk Outbreak Investigation 101 Purpose Limit scope and severity of immediate threat to public health Prevent future outbreaks Identify new vehicles of infection Identify new pathogens Outbreak Investigation 1. 2. 3. 4. 5. 6. 7. Detect Find Generate Test Solve Control Decide Definitions in Infectious Disease Attack Rate: the proportion of at-risk people who get sick, cumulative incidence Case Fatality Ratio: proportion of cases who die from their illness Incubation Period: time between infection and symptom onset Definitions in Infectious Disease Suspected Case: typical clinical symptoms of infectious agent; no epidemiologic or laboratory information Probable Case: typical clinical symptoms AND epidemiologic link; no laboratory confirmation Confirmed Case: laboratory confirmation of suspected infectious agent https://outbreaktools.ca/background/case-definitions/ Epidemic Curves Graphical representation of disease frequency over time # of cases on Y-axis Time on X-axis Distribution of cases can help inform cause and source of an outbreak Once a pathogen is identified, the known incubation period can lead us to find the source Epidemic Curves: Propagated Outbreak Measles Cases by Date of Onset in Aberdeen, South Dakota October 15, 1970 – January 16, 1971 Indicates person-to-person transmission Each peak is one incubation period apart https://www.cdc.gov/training/quicklearns/epimode/index.html Epidemic Curves: Point Source Cryptospordiosis Cases Associated with a Child Care Center by Date of Onset in Port Yourtown, Washington | June 1998 Single, one-time point of exposure Most cases occur within one incubation period of infection event https://www.cdc.gov/training/quicklearns/epimode/index.html Epidemic Curves: Continuous Source Salmonellosis Cases Exposed to Contaminated Salami by Date of Onset, United States December 2009 – January 2010 Ongoing exposure to the same source over a longer period Doesn’t always have defined peaks https://www.cdc.gov/training/quicklearns/epimode/index.html Summary Key Points Terminology is important The “best” measure is context specific Rates include time Multiple ways to compare the frequency of disease in different groups Epidemic curves can tell us a lot about the cause of an outbreak