Social Epidemiology

Potential Outcomes

100

Why do Kaufman and Cooper argue that we can't we apply the counterfactual definition of a causal effect to social factors?

Due to the nature of social factors which are either attributes of individuals or components of social relations, it is not plausible to randomize even hypothetically, which makes epidemiologic contrasts inappropriate.

200

What is the difference between homogeneity, consistency and exchangeability and how does it relate to neighborhood effects in social epidemiology?

Homogeneity (Holland); exchangeability (Greenland & Robins); Consistency (Rubin)- exchangeability means no confounding;the comparison groups are comparable and the actual exposure does not predict the counterfactual outcome.

300

Which assumption does imperfect compliance affect and which type of analytical approach can be implemented to resolve this?

What is identifiability and Instrumental Variables

400

Why would it be important to have a well-defined intervention?

The PO under a different intervention could take a different value (SUTVA violation)

500

It is possible to define potential outcomes for aggregate parameters. A policy of using anti-coagulation therapy rt-PA in hospital emergency rooms was implemented on ischemic stroke and hemorrhagic stroke patients. How can a true summary causal effect that is null arise?

The individual-level behavior that gave rise to the aggregate measure would be masked. (Kaufman, 2007, 361)

Life Course Epi

100

Which conceptual model does this DAG portray?

Accumulation of risk

200

What is the difference between lifecourse and social epidemiology?

Life course approach to chronic disease epidemiology is the study of long-term effects on chronic disease risk of physical and social exposures during gestation, childhood, adolescence, young adulthood and later adult life (Ben-Shlomo and Kuh 2002, p.285). Social epidemiology is the branch of epidemiology that studies the social distribution and social determinants of states of health (Berkman and Kawachi 2000, p.6)

300

Name 3 limitations of conducting life-course epidemiology studies.

Loss-to-follow-up, Measurement issues with correlated data, Retrospective reports, Time-varying exposure, Challenge to identify direct effect of childhood conditions not mediated by SEP: separate direct and indirect effects

400

What are the benefits to approaching chronic disease from a life-course epidemiology perspective?

It presents a new conceptualization of health and disease that allows us to examine multiple, possibly intersecting pathways between life in utero and adulthood. One of life course epidemiology’s greatest conceptual contributions to the field is its focus on the timing of exposure. For example, the critical period model suggests that there are periods of life when an individual is particularly vulnerable to certain exposures

500

What is the difference between a critical period and a sensitive period?

Critical window: a period of time occurring early in life that can have life-long and irreversible consequences (p.7); a limited time window in which an exposure can have adverse or protective effects on development and subsequent disease outcome (and only during this time window) Limb development in relation to maternal thalidomide use Perturbation at a critical period of development causes permanent alterations with lifelong, irreversible consequences Lack of folic acid during pregnancy can cause neural tube defects Drinking alcohol during pregnancy can lead to fetal alcohol syndrome Number of sweat glands that are set to become active is determined by an adaptation to the local environment at birth and has long-long consequences Sensitive period: a time period when an exposure has a stronger effect on development and hence disease risk than it would at other times If you learn a language before age 12, accent will be better and easier to learn Exposure to domestic violence increases chance of becoming perpetrator Poor nutrition when young makes child more susceptible to obesity

Thy Neighborhoods

100

What is the difference between endogenous factors and a contextual factors?

Endogenous factors are neighborhood characteristics that are determined by the individual characteristics of the residents. Contextual effects are exogenous neighborhood attributes that are not defined by the individual units. Endogenous effects may not be identifiable in observational studies (Manksi 1995) and therefore the effects may not be distinguished because once we condition on confounders we estimate conditionally independent effect. Ex. neighborhood food supply Diez Roux (2004)

200

What coefficient in multilevel models map the relationship between emergent context effects and the compositions of neighborhoods?

What is cross level coefficient (Oakes,2004, 1936)

300

What is the difference between contextual effects and integral effects?

Contextual effects emerge from with-in neighborhood interactions and integral effects emerge from structural things like toxic waste, parks, trash cans, etc.

400

What is the relationship between propensity score matching and estimating neighborhood effects?

The micro-model adjustment for neighborhood selection bias of the multi-level analysis approach is equivalent to propensity score matching of individuals on neighborhood exposure.

500

If assessing the causal relationship between proximity to idling bus stops and asthma, should low individual SES be treated as a confounder or mediator or both?

SES can be both a confounder and mediator of the relationship.

Methodological Problems

100

When focusing on studies that seek to identify effects of social variables, many authors argue for (Complete this motto): "no causation without ________."

Manipulation.

200

What methodological problem is depicted in this figure?

What is off support inference

300

In comparing two contrasting neighborhoods, investigators find that cases of lyme disease have been rising in an affluent neighborhood (with many avid golfers) while remaining negligible in an adjacent poor neighborhood. What is a methodological problem that arises?

What is structural zero - an entry in a contingency table that is certain to be zero, whatever the sample size, because it corresponds to an impossible outcome

400

What is the paradox of using the PO approach when adjusting for covariates in social epidemiology?

Kaufman says " It is common practice in social epidemiology to adjust for multiple social factors when attempting to isolate the "independent effect, but In this real world, where real people live, these other factors are not held constant...there can be no meaningful interpretation of an 'independent' effect, since in does not correspond to any consequence of exposure that can be isolated from the mechanism of treatment assignment"

500

A crucial problem in research on neighborhood characteristics and health is that neighborhood effects are by definition endogenous to the compositional characteristics of neighborhoods. Why is social stratification not a good idea?

Any attempt to separate out the "independent" effects of neighborhood characteristics and individual-level socio-economic indicators is futile- social stratification confounds comparisons (Diex-Roux, Oakes)

Methodological Solutions

100

What types of social causes are best treated as intermediate outcomes? (Provide example)

Social factors whose etiology can be traced back to events with intervention potentials. (Greenland 2005; Kaufman 1999)

200

Which proposed alternate strategy for progress in Social Epidemiology methods posits a solution to the violation of SUTVA?

System Based Modeling resolves this impasses of stability of treatment and inter-unit dependency by modeling inter-individual transmission of social process using observed data and feed-back loops. Kaufman (1999)

300

List 3 reasons why multivariate summary measures are considered superior to single unit summaries?

1) Allows for comparison of explicit "valuations" for different outcomes; 2) Allows for concurrent evaluation of economic cost, absolute and proportional effects; 3) Assess benefit of pure outcome removal approaches

400

Why is the local average treatment effect of a randomized trial in social epidemiology still estimatable in the presence confounders of the neighborhood exposure and disease relationship?

Using the randomized assignment mechanism as an IV tool, under a set of strong assumption, we can estimate the average effect exposure on disease in the "complier" group. The treatment effect on the treated can be estimated, or at least the effect can be bounded.

500

Can randomized community trials allow us to estimate contextual effects of neighborhoods?

Randomized trials allow for an exogenous intervention to modify compositional aspects of the neighborhood independent of endogenous factors.