Epidemiology Introduction PDF

1. Introduction to epidemiology, history of epidemiology, measures of disease occurrence Theodore Lytras Assistant Professor of Public Health Let us ask ourselves: How do we know smoking / air pollution / lack of excercise is associated with disease? Let us ask ourselves: H...

1. Introduction to epidemiology, history of epidemiology, measures of disease occurrence Theodore Lytras Assistant Professor of Public Health Let us ask ourselves: How do we know smoking / air pollution / lack of excercise is associated with disease? Let us ask ourselves: How do we know smoking / air pollution / lack of excercise is associated with disease? How do we know that a new lipid-lowering medication is effective in preventing cardiovascular disease? Let us ask ourselves: How do we know smoking / air pollution / lack of excercise is associated with disease? How do we know that a new lipid-lowering medication is effective in preventing cardiovascular disease? How do doctors know that COVID-19 vaccination is effective in preventing COVID-19 disease and death? Let us ask ourselves: How do we know smoking / air pollution / lack of excercise is associated with disease? How do we know that a new lipid-lowering medication is effective in preventing cardiovascular disease? How do doctors know that COVID-19 vaccination is effective in preventing COVID-19 disease and death? We learn that: by comparing disease rates between groups of people more generally, by studying the distribution of deaths, diseases and risk factors, and their associations What is epidemiology? Definition “the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems” What is epidemiology? Definition “the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems” Epidemiology is a Greek word: Epi meaning “upon” Demos meaning “people” Logos meaning “study of” Meaning: the study of what befalls a population What is epidemiology? Definition “the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems” Frequency, both absolute (counts) and relative to the size of the population (proporion) Patterns, with respect to time, place and person (person = personal characteristics, e.g. age, sex, etc) What is epidemiology? Definition “the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems” “Health-related states”: not just diseases, for example: injuries, causes of death, birth defects, behaviours (e.g. tobacco use), biological markers of disease, positive health states, etc To encompass everything, we use the term outcome What is epidemiology? Definition “the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems” “Determinants”: causes, risk factors or preventive factors that influence the ocurrence of disease or other outcomes biological, chemical, physical, social, cultural, economic, genetic and behavioural In epidemiology we call all these factors exposures What is epidemiology? Definition “the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems” All epidemiological studies refer and to a specific study population Defined by identifiable characteristics of time, place and person, e.g. age, occupation, etc Questions we may use epidemiology for: How often does a disease occur? What are the consequences of having a disease? What is the life expectancy of people with disease? What factors are associated with an increased risk of disease? How accurate are tests to diagnose a disease? How does treatment change the course of disease? Does early treatment improve the course of disease? Questions we may use epidemiology for: How often does a disease occur? What are the consequences of having a disease? What is the life expectancy of people with disease? What factors are associated with an increased risk of disease? How accurate are tests to diagnose a disease? How does treatment change the course of disease? Does early treatment improve the course of disease? Every doctor must have a good understanding of epidemiology It is the “basic science” of both clinical medicine and public health It is our main tool for generating actionable clinical knowledge – no matter your specialty – in the context of Evidence-Based Medicine Uses of epidemiology Initially applied to study and control of communicable diseases (late 19th – early 20th century) Subsequently applied to link environmental exposures with disease Later applied to chronic non-communicable diseases (e.g. cancer, heart disease) and also injuries, especially in middle- and high-income countries Even more recently: molecular and genetic epidemiology, pharmacoepidemiology, etc Exposures = biochemical markers, gene variants (allelles), drugs Sources of epidemiological data for study Population census Vital statistics Births / deaths – as reported in population registry offices Hospital records + primary care records Public health surveillance Mostly (though not exclusively) about infectious diseases – mandatory notification and other surveillance systems Chronic disease registries e.g. cancer registries, occupational disease registries Epidemiologic research ad-hoc data collection, e.g. in longitudinal studies “Big data” Electronic Health Records (EHR), e-prescription databases, etc Objectives of epidemiology 1. Determine mortality and disease burden, its patterns and evolution over time 2. Identify the aetiology (causes) of disease and its associated risk factors 3. Study the natural history and prognosis of disease 4. Evaluate preventive and therapeutic interventions (and diagnostic strategies as well) 5. Develop public health policy Objectives of epidemiology 1. Determine mortality and disease burden, its patterns and evolution over time 2. Identify the aetiology (causes) of disease and its associated risk factors 3. Study the natural history and prognosis of disease 4. Evaluate preventive and therapeutic interventions (and diagnostic strategies as well) 5. Develop public health policy Three general categories of epidemiological studies Descriptive epidemiology Describe patterns of disease Analytical epidemiology Examine associations between diseases and risk factors Experimental epidemiology Examine the effectiveness of experimental interventions  Descriptive epidemiology When and where cases of disease appear? Any common group characteristics? (age, sex, etc) Reveals patterns that may otherwise not have observed =⇒ hypothesis generation Analytical epidemiology Testing those hypotheses; comparing groups of persons to find associations between exposures and outcomes e.g. comparing rates of asthma attacks between persons exposed to air pollution and those who aren’t or, compare consumption of a suspect food item between cases of gastroenteritis and healthy people exposure ←→ outcome (works both ways!) Experimental epidemiology Randomise a group of people, so that some receive an intervention, and other do not; compare them with respect to an outcome History of epidemiology History of Epidemiology Hippocrates (4̃00 BC): First to attempt to explain disease from a natural rather than a supernatural viewpoint “On airs, waters and places” – first to suggest environmental and host factors (behaviours) might influence the occurrence of disease, and that altering those might change its course History of Epidemiology Hippocrates (4̃00 BC): First to attempt to explain disease from a natural rather than a supernatural viewpoint “On airs, waters and places” – first to suggest environmental and host factors (behaviours) might influence the occurrence of disease, and that altering those might change its course John Graunt (1662): (a haberdasher!) Published a landmark analysis of mortality data and its patterns First to correctly observe (among other things): the high mortality of infants and children at the time (1/3 died by the age of 5) the higher male-to-female ratio in births (14 to 13) and deaths that the plague claimed many more deaths than attributed to it Elected to the Royal Society – but eventually died in poverty & obscurity! History of Epidemiology William Farr (1838 – 1883): “Compiler of Abstracts” to the General Register Office in England Systematized the collection, study and reporting of vital statistics data, including routinely collecting the cause of death: allowing for the first time to study mortality rates of different occupations Described the “bell-shaped curve pattern” of deaths during an epidemic (the “epidemic wave” of today) History of Epidemiology William Farr (1838 – 1883): “Compiler of Abstracts” to the General Register Office in England Systematized the collection, study and reporting of vital statistics data, including routinely collecting the cause of death: allowing for the first time to study mortality rates of different occupations Described the “bell-shaped curve pattern” of deaths during an epidemic (the “epidemic wave” of today) John Snow (1854): (not Jon...) The father of field epidemiology An English anesthesiologist who administered chlorophorm to Queen Victoria (!) during childbirth, and successfully investigated the 1854 cholera outbreak in Soho, London, without knowledge of the causative agent of cholera (or even germ theory!) Cholera in the 19th century The prevailing theory for cholera at the time was “miasma theory”: a cloud (“miasm”) of foul air that clung low on the surface of the earth William Farr subscribed to this theory, and even collected data to support it −→ Cholera in the 19th century The prevailing theory for cholera at the time was “miasma theory”: a cloud (“miasm”) of foul air that clung low on the surface of the earth William Farr subscribed to this theory, and even collected data to support it −→ John Snow disaggreed, and thought that the cholera “poison” was transmitted through water. By talking to people on the field, he observed that most cases during the 1854 outbreak clustered around the water pump at Broad street John Snow’s spot map of the 1854 London cholera outbreak John Snow’s spot map of the 1854 London cholera outbreak John Snow’s first investigation By talking to the authorities about his observations, he convinced them to disable the Broad street pump by removing its handle This is credited for ending the outbreak – although it was already in decline at the time (and Snow himself recognized it) Spot maps as an epidemiological tool are enduring to this day... John Snow’s second investigation Snow then observed that death rates from cholera were highest in two London districts served by two water companies (Southwark & Vauxhall, and Lambeth) Both received water from the river Thames, but the Lambeth company had moved its intake point upstream, therefore avoiding sewage (that were discharged directly into the river...) John Snow’s second investigation John Snow’s second investigation Snow compared mortality rates between districts supplied by either company, or both: He then investigated further the area that was supplied by both, and identified the water supplier for each house John Snow’s second investigation After the outbreak people were still reluctant to accept Snow’s theory (the fecal-oral route was unpalatable) and still believed miasma However Snow’s observations were so convincing that William Farr, as Registrar General, required his district registrars to record which water company supplied each house when a cholera death occurred Later (in 1866) when investigating another outbreak Farr issued orders that water should be boiled before drunk John Snow’s second investigation After the outbreak people were still reluctant to accept Snow’s theory (the fecal-oral route was unpalatable) and still believed miasma However Snow’s observations were so convincing that William Farr, as Registrar General, required his district registrars to record which water company supplied each house when a cholera death occurred Later (in 1866) when investigating another outbreak Farr issued orders that water should be boiled before drunk Epidemiology is the way to convince people... Good (*) epidemiology can be very impactful to promote changes in practice of both clinical medicine and public health * by the standards of the time Ignaz Semmelweis and handwashing Semmelweis was a mid-18th-century obstetrician who became interested in “childbed fever” (puerperal fever), which was a major cause of death of women during childbirth Germ theory was not invented yet – theories abounded! Semmelweis was placed in charge of the first (of two) obstetric clinics in the General Hospital in Vienna (1846). He observed the 1st clinic had twice the mortality of the 2nd clinic (16% vs 7%) The 2nd clinic was staffed by midwives and the 1st clinic by physicians and medical students... Ignaz Semmelweis and handwashing Physicians and students began their days performing autopsies in women who had died of childbed fever, and then would proceed to perform deliveries of babies (without any handwashing in-between...) Ignaz Semmelweis and handwashing Physicians and students began their days performing autopsies in women who had died of childbed fever, and then would proceed to perform deliveries of babies (without any handwashing in-between...) Semmelweis surmised the hands of physicians and students were transmitting “disease-causing particles”, and implemented a hand hygiene intervention (washing and scrubbing with lime juice) This rapidly lowered mortality from 12.2% to 2.4%, comparable to the 2nd clinic! Ignaz Semmelweis and handwashing What Semmelweis did is called today a “controlled before-and-after” (or “interrupted time series”) study by modern epidemiologists The “Doctors study” of smoking by Doll and Hill Started in 1951, including registered British doctors, with follow-ups in 1957, 1966, 1971, 1978, 1991 and 2001 Showed that lung cancer, coronary heart disease and overall mortality was much more common among smokers The “Doctors study” of smoking by Doll and Hill The “Doctors study” of smoking by Doll and Hill Measures of disease occurrence Summary of terms used Know your terminology well! Count Number of individuals meeting a criterion (case definition) Proportion A/(A + B) Number of individuals meeting a criterion / total number of people in the population or group Ratio X/Y comparing two unrelated but similar quantities (same units) Rate A/t or A/((A + B) ∗ t) Count or proportion of new cases per unit time. Often used with person-time as denominator. Risk P Probability of an event occurring (essentially a proportion: cases / people at risk) Odds P/(1 − P) An alternative expression of probability. Probability of occurrence / probability of non-occurrence Hazard The instantaneous probability of an event happening at time t, given that it has not already happened. Used in survival analysis Some examples 100 students per instructor 20% of the population are smokers 5 tuberculosis cases per 100,000 population during 2018 5 TB cases per 100,000 population per year, between 2010–2020 4 days of rest per day of work Some examples Ratio 100 students per instructor Proportion 20% of the population are smokers 5 tuberculosis cases per 100,000 population Proportion during 2018 5 TB cases per 100,000 population per year, Rate between 2010–2020 Ratio 4 days of rest per day of work Odds P probability of occurrence Odds = = 1−P probability of non-occurrence What is the probability of throwing a 6? What are the odds of throwing a 6 ? What are the odds of throwing anything but a 6 ? Conceptually, odds = probability (expressed in a different form) Used for “technical” reasons (case-control studies, logistic regression) Probability vs odds P Odds = 1−P A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 What was the exposure and what was the outcome in this study? A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 What was the risk of testing positive for influenza among unvaccinated patients? A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 What was the risk of testing positive for influenza among vaccinated patients? A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 What were the odds of being vaccinated, among patients testing positive for influenza? A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 What were the odds of being vaccinated, among patients testing negative for influenza? A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 Divide these odds between them. This is an Odds Ratio A further example Seasonal influenza vaccination and hospitalization due to influenza Population: 7724 persons that were hospitalized with respiratory symptoms over 6 winter seasons, and were laboratory tested with PCR for influenza Some were vaccinated, and others were not Influenza(–) Influenza(+) Total Unvaccinated 3499 2752 6251 Vaccinated 1057 416 1473 Total 4556 3168 7724 What were the odds of testing positive, among vaccinated and unvaccinated patients? Divide the two odds between them. What do you notice? Recap Risk is a proportion and a probability Proportions (or percentages) include everyone as the denominator Ratios compare two similar in nature but distinct quantities Rates are counts or proportions per unit time Recap Risk is a proportion and a probability Proportions (or percentages) include everyone as the denominator Ratios compare two similar in nature but distinct quantities Rates are counts or proportions per unit time However, there’s a problem with rates: What do you do when people have not been followed up for the same time? e.g. what is the death rate among people vaccinated against COVID-19 ?? Rate & person-time (Person-years, or person-months, etc) Rate & person-time (Person-years, or person-months, etc) 3 cases / 87 person-months = 3.45 per 100 person-months = 0.4 per person-year Hazard: an instantaneous measure of risk Hazard is the instantaneous risk of something happening to you at time t + 1, provided that you’ve survived until time t e.g. we have 20 COVID-19 patients who are admitted in an ICU (t = 0). 10 have survived until day 5, 9 have survived at day 6, and 8 have survived at day 7. What is the hazard of death at day 5 and at day 6? Measures of disease occurrence Fundamentally there’s just two of them, each with two “flavours” Incidence Counts new cases of the disease (or outcome) Cumulative incidence Incidence rate Essential to distinguish between these two! Prevalence Counts existing cases of the disease (or outcome) Point prevalence Period prevalence Although distinction is not often made between the latter two... Cumulative incidence New cases over a period of time Initial population at risk It’s a proportion Goes from 0 to 1 You must define the time frame! Synonyms: incidence proportion, risk (!), incidence (avoid) Cumulative incidence New cases over a period of time Initial population at risk It’s a proportion Goes from 0 to 1 You must define the time frame! Synonyms: incidence proportion, risk (!), incidence (avoid) Population must be free of the outcome (i.e. at risk) in the beginning! Incidence rate New cases over a period of time Person-time at risk It’s a rate Goes from 0 to +∞ You must report the person-time units Synonyms: incidence density, incidence (usually, but avoid) Prevalence Counts existing cases at a specified time (or period) It is a proportion Prevalence = Incidence rate × duration of disease Prevalence = 1 / 5 cases Prevalence = 2 / 5 cases Prevalence Counts existing cases at a specified time (or period) It is a proportion Prevalence = Incidence rate × duration of disease Prevalence = 1 / 5 cases Prevalence = 2 / 5 cases Prevalence Counts existing cases at a specified time (or period) It is a proportion Prevalence = Incidence rate × duration of disease Prevalence = 1 / 5 cases Prevalence = 2 / 5 cases Doubling the duration of disease doubles the prevalence An example: prevalence & incidence Can you explain this chart? Prevalence = Incidence rate × duration of disease This is important to remember! As doctors, we try to have our patients live longer =⇒ prevalence of chronic disease increases Important to lower incidence = prevent new cases of chronic disease however possible The importance of specifying the population Both for incidence and prevalence Our measures might be referring to the general population, or to subsets thereof (e.g. age groups, occupational groups, or based on any other characteristic) The importance of specifying the population Both for incidence and prevalence Our measures might be referring to the general population, or to subsets thereof (e.g. age groups, occupational groups, or based on any other characteristic) Example A country has 5,800,000 population of whom 200,000 are MSM (men who have sex with men). There are 6,600 persons currently living with HIV in that country, of whom 3,600 are MSM. What is the prevalence of HIV in the general population, and among MSM in the country? The importance of specifying the population Both for incidence and prevalence Our measures might be referring to the general population, or to subsets thereof (e.g. age groups, occupational groups, or based on any other characteristic) Example A country has 5,800,000 population of whom 200,000 are MSM (men who have sex with men). There are 6,600 persons currently living with HIV in that country, of whom 3,600 are MSM. What is the prevalence of HIV in the general population, and among MSM in the country? Population: 6,600 / 5,800,000 = 1.14 per 1000 persons Among MSM: 3,600 / 200,000 = 18 per 1000 persons A special case: Attack Rate (AR, %) Used in the context of a disease outbreak with closed   populations: Number of cases during outbreak Population at risk   Despite the name, it is not a rate!! Rather a... A special case: Attack Rate (AR, %) Used in the context of a disease outbreak with closed   populations: Number of cases during outbreak Population at risk   Despite the name, it is not a rate!! Rather a proportion (cumulative incidence). A special case: Attack Rate (AR, %) Used in the context of a disease outbreak with closed   populations: Number of cases during outbreak Population at risk   Despite the name, it is not a rate!! Rather a proportion (cumulative incidence). A classic example (the outbreak in Oswego county, 1940) Among 75 persons attending a church supper, 46 fell ill with diarrhea. (Attack Rate = 46/75 = 61%) 54 people ate vanilla ice cream, among whom 43 fell ill. Among 21 people who did NOT eat ice cream, only 3 fell ill. ARice cream = 43/54 = 80% ARno ice cream = 3/21 = 14% A special case: Attack Rate (AR, %) Used in the context of a disease outbreak with closed   populations: Number of cases during outbreak Population at risk   Despite the name, it is not a rate!! Rather a proportion (cumulative incidence). A classic example (the outbreak in Oswego county, 1940) Among 75 persons attending a church supper, 46 fell ill with diarrhea. (Attack Rate = 46/75 = 61%) 54 people ate vanilla ice cream, among whom 43 fell ill. Among 21 people who did NOT eat ice cream, only 3 fell ill. ARice cream = 43/54 = 80% What if you compare the ARno ice cream = 3/21 = 14% two using a ratio? Measures of the occurrence of death Mortality rate Number of deaths per population per unit time Crude (all-cause) vs cause-specific mortality Usually annual, per 100,000 or 1,000 population (at mid-year) Synonym: mortality Case Fatality Rate Proportion of deaths among people with a certain disease (and over a period of time) NOT a rate! Denominator = ill persons Measures of the occurrence of death Mortality rate Number of deaths per population per unit time Crude (all-cause) vs cause-specific mortality Usually annual, per 100,000 or 1,000 population (at mid-year) Synonym: mortality Case Fatality Rate Proportion of deaths among people with a certain disease (and over a period of time) NOT a rate! Denominator = ill persons Example During the year 2020, a country of 10,700,000 people recorded 140,000 COVID-19 cases and 5,650 COVID-19 deaths. What is the mortality rate and the CFR? Mortality rate = 5,650 / 10,700,000 = 52.8 deaths per 100,000 population CFR = 5,650 / 140,000 = 4% Measures of the occurrence of death Mortality rate Number of deaths per population per unit time Crude (all-cause) vs cause-specific mortality Usually annual, per 100,000 or 1,000 population (at mid-year) Synonym: mortality Case Fatality Rate Proportion of deaths among people with a certain disease (and over a period of time) NOT a rate! Denominator = ill persons Can you have one high and the other low? Sure you can! Malaria, Influenza ↑ incidence and mortality rates, despite ↓ CFR (≈0.5% malaria, influenza usually lower) Ebola Thankfully very rare and thus ↓ mortality rate, despite ↑↑ CFR (>80%) Measures of the occurrence of death Mortality rate Number of deaths per population per unit time Crude (all-cause) vs cause-specific mortality Usually annual, per 100,000 or 1,000 population (at mid-year) Synonym: mortality Case Fatality Rate Proportion of deaths among people with a certain disease (and over a period of time) NOT a rate! Denominator = ill persons Can you have one high and the other low? Sure you can! Malaria, Influenza ↑ incidence and mortality rates, despite ↓ CFR (≈0.5% malaria, influenza usually lower) Ebola Thankfully very rare and thus ↓ mortality rate, despite ↑↑ CFR (>80%) −→ In general: Mortality rate = Incidence rate × CFR Direct age standardization (or: Direct age adjustment) Mortality is affected by the age structure of the population Standardization allows us to appropriately compare mortality between different populations It involves calculating age-specific rates and applying them to age strata of a common population. An example: Direct age standardization (or: Direct age adjustment) Mortality is affected by the age structure of the population Standardization allows us to appropriately compare mortality between different populations It involves calculating age-specific rates and applying them to age strata of a common population. An example: Later period has a higher death rate Direct age standardization (or: Direct age adjustment) Mortality is affected by the age structure of the population Standardization allows us to appropriately compare mortality between different populations It involves calculating age-specific rates and applying them to age strata of a common population. An example: But note how there are more older people in the later period... Direct age standardization (or: Direct age adjustment) Mortality is affected by the age structure of the population Standardization allows us to appropriately compare mortality between different populations It involves calculating age-specific rates and applying them to age strata of a common population. An example: Early period has a higher age-adjusted death rate Indirect age standardization This is used in computing Standardized Mortality Ratios (SMRs) Observed number of deaths in a group, divided by expected number of deaths if they had the age-specific mortality rates of the general population Frequently used to compare the mortality of occupational groups to that of the general population. An example: Indirect age standardization This is used in computing Standardized Mortality Ratios (SMRs) Observed number of deaths in a group, divided by expected number of deaths if they had the age-specific mortality rates of the general population Frequently used to compare the mortality of occupational groups to that of the general population. An example: Thank you!