PUBH 613
Years of Life Lost (YLL)
(The number of deaths) X (The standard life expectancy at the age at which death occurs) It is often used to give an idea of the relative contribution of a condition to health. The more years of life lost attributed to a particular condition, the more important it is to population health.
Use of surveillance data
- Estimate magnitude of problem - Determine geographic distribution - Detect epidemics/ define a problem - Evaluate control measures - Monitor changes in infectious agents, and other exposures - Detect changes in health practices - Facilitate planning
National Surveys to assess the health status of the population
- Publicly available - Globally the DHS is carried out in 90 developing countries to monitor health: Locally the US has a number of survey: 1. Behavioral Risk Factor Surveillance System (BRFSS): is the US nation's premier system of health-related telephone surveys that collects state data about U.S. residents regarding their health-related risk behaviors, chronic health conditions, and use of preventive services. 2. National Health and Nutrition Examination Survey(NHANES): is a program of studies designed to assess the health and nutritional status of adults and children in the United States. The survey is unique in that it combines interviews and physical examinations.
Notifiable Disease Surveillance
- Reporting mandated by law/regulation ex: mumps/measles - Passive surveillance ex: up to the provider or lab to complete the report - Active surveillance The health department actively seeks data from provider/laboratory. - MMWR (US) - weekly report of notifiable diseases (since 1961)
Cross-section study
-Compare exposures between people WITH and WITHOUT disease -Select from a defined population, often population representative Obtain information about a group of people in ONE point in time. For example their habits and their disease status. Typically they assess incidence and prevalence. Selected from a defined population using a sampling frame and being population representative. Pay attention to pop. representative pop. based and community based. Only pop representative MEANS pop representative. Ex: BMI with diabetes - Easily available (no waiting for disease) a study being conducted for routine surveillance so we don't need to worry about waiting for disease incidence we just use prevalence. Can evaluate associate with many exposure - Hard to distingusih case from effect no temporality. _Susceptible to major biases
Strengths and Weaknesses DAGS
-Make assumptions EXPLICIT and unambiguous - draw and review the DAG -Drawing the assumptions makes them clearer than explaining them in words -Show we only have to worry about two major sources of biases (confounding and selection bias) when thinking about studies of cause and effect. -Make it more obvious we have no models of population health to help dreaw the DAGs
RCT
-Randomize exposure -No selection bias -No confounding bias - No information bias Possibility of loss to follow up Not feasible for all exposures non modifiable Harmful but can always reduce rahte than increas exposure Noncompliance lifestyle long follow up May not generalize if the intervention operates by a mechanism that is only relevant in some sub-groups
Problem of confounding
1. Impossible to measure comprehensively all known potential confounders - Cannot ask people comprehensively about all confounders: may not answer honestly/accurately 100s of question 2. Impossible to ask questions on UNKNOWN and common causes of exposure even though they may exist. - Beta Carotene -HRT -Vitamin Supplement (non have been substantiative) can afford them and more health conscious
Directed Acyclic Graph DAG
A causal diagram displaying causal relation between variables. Directed: Each direct causal effect of one variable on another is represented by an arrow is represented by an arrow with its TAIL at the CAUSE and its HEAD at the EFFECT Acyclic: No cycle as no variable can affect itself Explicitly depicts underlying assumptions. Gives us a template we can use when we encounter new situations. Essentially, we need to account for things that cause gray hair and death if we want to get the true causal effect of gray hair ON death.
Prevalence
A measurement of all individuals affected by the diease at a patricular time.
Case reports and Cases Series
Always pay attention bc they can be the start of something BIG. Importance: - Early detection of unusual occurrences or presentations - Preliminary source of information for potential exposures Limitations -Do no have comparison group -Usually very small sample size -Often do not systematically collect data (rely on medical charts)
Avoid COnfounding
COnduct a study where confounders play less of a role, but the study is still open to unknown confounding
Confounder
Common cause of exposure and outcome 1. SEP (socio-economic position) - many exposures like diet, life style, environmental exposures 2. Health status - Many exposures Diet life style -Many outcomes -- most disease 3. Health consciousness -Many exposures -Diet, life style, environmental exposures -Many outcomes -Most diseases Ex: not seeing effects of qual's eggs but a person ABLE to eat qual's eggs We get a misleading association
Observational - Ecological Study
Correlate population level attributes: - Chocolate consumption & Nobel laureates Issues: - May not apply at the individual - Many such correlations exist May be more relevant to disproving associations
Detect changes in health pracitces
Could be something like changes in vaccination rates, the use of opiods
Selection bias
Cross sectional study Ensure no selection on exposure and outcome Case-control study do not select controls on exposure Cohort study Ensure no selection of the cohort on exposure and outomce avoid loss to follow up CAUSAL inference from observation studies requires strong assumptions
Disability-Adjusted Life Year (DALY)
DALY = YLL (The number of deaths) X (The standard life expectancy at the age at which death occurs) + YLD [The number of incident cases in that period is multiplied by the average duration of the disease and a weight factor that reflects the severity of the disease on a scale from 0 (perfect health) to 1 (dead) ]
Food-born botulism reported cases by year, US 1982-2002
Each time the numbers go up, there is an outbreak investigation to find out what caused the problem and then remove it from the food chain. So here the botulism was caused by seafood, baked potatoes and chili-sauce. Once they are removed the problem goes away.
Observational Studies
Ecological Case Reports Cross sectional Case control Cohort - Exposure not under control of research Used when assignment of exposures is NOT ethical Too expensive Otherwise difficult Open to biases such as confounding and selection
Selection Bias
Estimate EXCLUDES some of the effect due to selection or other factors Solution: can be complex, and may involve selecting a new sample. May need to start again w new sample Ex: Measured exposure and/or disease (information bias) healthy worker (healthy worker bias) Hospital Patient (Berkson's bias 1946)
Selection Bias Ex 4
Ex: Study of survivors could be harmful (older people who were treated for higher blood pressure lived longer)
Internal Validity
For a study of cause and effect, criteria for internal validity are: -No confounding -No selection biase Are these the same criteria as for internal validity of a risk prediction model - No preducting risk and assessing cause and effect are completely different questions so they have different criteria for internal validty
Health Metrics
Health metrics and evaluation helps identify the best strategies to build a healthier world. By measuring health, tracking program performance, and finding ways to maximize health system impact. It provides a foundation for informed decision-making that ultimately will lead to better health for people worldwide.
Estimate magnitude of problem
How many are affected
Experimental
In vitro In vivo Randomized controlled trail RCT Research assigns exposure or treatment to a group of study Assignment usually randomized Maximizes control over influence of other factos Improved measurement of exposure Expensive can be unethical
Breast cancer screening in the US 1989-1998
Mammogram uptake over time in three states. Rates are going up, Kentucky is always lagging New York and Washington, so it would be a good idea to focus more effort on Kentucky. However, in the meatime researchers have begun to question the value of breast cancer screening for all, as for example (Pashayan N, Morris S, Gilbert FJ, Pharoah PDP. Cost-effectiveness and Benefit-to-Harm Ratio of Risk-Stratified Screening for Breast Cancer: A Life-Table Model.
Differential Misclassification
Manifestation of selection Bias - Error related to both exposure and outcome -Overweight people tend to under report consumption of unhealthy foods 0Association of depression with self rated physical health -People who are depressed may rate their physical health poorly
Quasi Experimental
Natural experiments (instrumental variable analysis) **Hard to Find
Internal validity requires
No confounding and no selection biase, but confounding and selection biase are the major sources of biar in observational studes - May generate spurious associations - May conceal real causes -Easily occur in observational studies -May be easier to conceptualize using DAGs
Blood Pressure Non Differential Misclassification
Non differential: Stress + Blood Pressure - Blood Pressure Station - Fail to allow the five minutes rest bc inpatient -Blood pressure measured incorrectly (THIS IS INDEPENDENT OF THE STRESS EXPOSURE) the way it is measured -Adding random noise - Biases toward to NULL
Facilitate planning
Once a problem has been identified, then try to solve it, for example it could be the measles vaccination rate in New York
Evaluate control measures
One of many methods, have to check if it is a valid method
Information Bias
Poor measurement Classified as differential and non-differental biasses -non differential if it is unrelated to the occurrence or presence of disease - if the misclassification of exposure is different for those with and without disease it is differential
Case series
Reporting describing characteristics of more than one patient with a given disease - 5 young homosexual men w Pneumocystics at 3 LA hospita Exposure: Sexual behavior Outcome: PCP (Pneumocystis carinii pneumonia) With a case series we have no comparison group so it is hard to put this information in perspective and make sense of it
Case control Study
START with the DISEASE. THEN they COMPARE EXPOSURES between people WITH AND WITHOUT THE DISEASE. Ex: To compare exposures to people with and without lung cancer. With exposure had a ten fold difference? SMOKING Cannot usually obtain incidence bc we started with the cases and may NOT know how they compare to the underlying population. We may compare ODDS of EXPOSURE in the cases to ODDS of EXPOSURE in the controls which gives an odds ratio. -Easy for clinicians to use - Ideally best to use INCIDENT cases (i.e. NEWLY DIAGONSED) Prevalent --> Living with disease who might have different determinants. The reasons why someone survives might be different from why they contracted the disease. -Determine PAST EXPOSURE (questionnaire, medical records, hospital databases, disease registries, death certificates) -Compare exposure in cases and controls using odds ratio - may have controls per case typically 1:1 to 4:1 to increase power Strengths: -Fast and relatively cheap (no waiting for disease) - Efficient - requires fewer study participants -Efficient for rare diseases, or diseases with long latency period -Can evaluate associate with many exposures - when disease is rare, Odds ratio can be interpreted/approximates the risk ratio and can be interpreted as such. Weakness: -NOT efficient if exposures are rare -Susceptible to makor biases -Temporal order exposure/disease not alays clear to establish -Can be operationally difficult to revruit control participants Uses: -Good for outbreak investigation -Often used for genome wide association studies
Population Representative
Sampling a representative sample of the population.
Community Based
Sampling from the community. Typically population based and community based samples are not representative of the underlying population.
Population Based
Sampling from the population. Typically population based and community based samples are not representative of the underlying population.
Selection Bias: Defintion
Selecting on a common effect of exposure and outcome also called "collider" biase -Common and influential sources - selecting out people who have already died from the exposure - odler people and sick poeple -Does population represteative study include poeple who have already died? -Attribute of the associate being estimated -Arises from creating a spurious link between exposure and outcome often through a "common effect" -If I select people based on two characteristics I may create a spurious relation between those two characteristics even if non exists. Ex: Coffee and Stomach Pancreatic cancer: Selecting controls with stoach disorders means the controls are largely non-coffee drinkers, so it looks as if coffee drinking causes pancreatic cancer. Ex: The full harms of the genetic variant cannot be observed particularly for studies recruited at older ages. Dead and not available for selection
Selection Bias ex
Selection bias ex: Everyone who died from hemorraghic stroke due to their alcohol use are ALREADY DEAD and missing from the sample We need an arrow from alcohol to stroke (purple = causal relationship being CONSIDERED) Heavy alcohol use might prevent selection into the study. Selection into the study (two arrows meet here) create an extraneous link between heavy alcohol and stroke. SO we cannot obtain the TRUE effect of alcohol on STROKE (on older people). The box indicates the selection is an extraneous link.
Emic-Etic/Universal-Culture-Specific
Specifically, 'etic' refers to research that studies cross-cultural differences, whereas 'emic' refers to research that fully studies one culture with no (or only a secondary) cross-cultural focus.
Cohort Study
Start with a group of people COHORT. And follows up with the group of people in the cohort. Compare exposures between people who do and do not develop disease Select a sample usually without disease, possibly on exposure, and identify subsequent disease. We can compare INCIDENCE of the exposed with INCIDENCE of the unexposed which gives an RISK RATIO. - Population based - Occupation based - Selected based on exposure -Geographic - Demographic or other common characteristic -Births at a particular time -Exposure-based (whether or not they had the exposure of interest) - Start with a group of people usually WITHOUT disease Weaknesses: -Expensive & length (especially prospective cohorts) -Inefficient for rare disease (need to wait a long time for enough cases to accrue) -Loss to follow up decrease the sample size -Susceptible to major baises Strengths: -Clear temporal order of exposure with disease -Can directly measure incidence - Good for RARE exposures (can oversample exposed grup -Good exposure assessment -Possible to look at multiple outcomes (compare risk of many outcomes for any given exposure)
Detect epidemics/ Define a problem
Tend to have an epidemic of influenza every winter, also have had epidemics of measles in New York recently
Years Lost due to Disability (YLD)
The number of incident cases in that period is multiplied by the average duration of the disease and a weight factor that reflects the severity of the disease on a scale from 0 (perfect health) to 1 (dead).
Total Life Expextancy
The number of years a baby born today would be expected to live. The longer the better.
Quality Adjusted Life Year
The quality-adjusted life year or quality-adjusted life-year QALY is a generic measure of disease burden, including both the quality and the quantity of life lived. One QALY = One year in perfect health
Monitor changes in infectious agents, and other expisures
Track exposures, people have measured "bugs" on the subway, antimicrobial resistant (AMR) bacteria in the food chain, particularly in meat, or pesticides or pollutants)
Generate hypotheses, stimulate research
Trends can always spark an idea, or more often identify problem disease (issue with cardiovascular disease) or practices (opioid epidemic)
Confounding
Type of bias: Exposure and disease may be correlated due to other attributed rather than the exposure causing disease -Mistaking symptoms for causes Ex: Grey hair correlates with death because aging causes grey hair and death Selection bias: Exposure and disease may be correlated due to the sample selection rather than the exposure causing disease Selection bias: High alcohol use correlates with lower risk of hemorrhagic stroke because people who have already died of hemorrhagic stroke due to high alcohol use are already dead and missing from the sample.
Confounding bias
When estimates include effects of same extraneous factors that cause exposure and outcome (i.e. common exposures of exposure AND outcome) Ex: common cause of LOW VIT D and DEATH is ILL-HEALTH Solution - account for those factors. MUST TAKE ILL HEALTH INTO ACCOUNT IN ORDER TO NOT GET A MISLEADING ESTIMATE
Determine geographic distribution
Where is it
Observational - Case report
ex: 40 year old premenopausal women w/ pulmonary embolism Exposure: Oral contraceptive use Outcome: Pulmonary embolism Can you say much from this case report about whether or not OC use causes Pulmonary Embolism? No.
