cross-sectional studies and pmr studiescross-sectional incidence study in the base population 3...

24
Cross-Sectional Studies and PMR Studies 200A November 5, 2009

Upload: others

Post on 24-Jun-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Cross-Sectional Studies and PMR Studies

200ANovember 5, 2009

Cross-sectional study – a survey

An observational study in which all variables are measured at a single point in time

Cross-sectional studies are used to estimate prevalence of diseases and frequencies of exposures.

Diseases of short duration will not be well presented since prevalence is a function of incidence and duration

All in the population at that time

A simple random sample. Give all a number 1-1000

=763764765=

Select 100 first random numbers to the study

Often much more complicated with random sampling from different segments of the population by sex, ethnic groups, socio- economic studies, etc.

Often using differing sampling fractions.

A study of peripheral vascular disease (PVD) in Scotland and smoking

Measures of association?Interpretation?

SmokingEver Never All

PVDNo PVD

23 81704 1291

312995

All 1727 1299 3026

Because exposure and disease are assessed at the same time, cross-sectional studies may not be able to establish that exposure preceded onset of the disease process.

Cross-sectional studies can examine multiple exposures and multiple diseases but they often have limited scientific value. Most important problems:

E N MID

S

+ 1000 20 80

+ - 1000 30 70

E D

E Duration

E Cases Cohorts

+ 80 75- 70 75

150 150OR= 80/70

75/75 =1.14

An investigator would like to see whether smoking is associated with hearing loss. She distributes a questionnaire to a random sample of adults living in Los Angeles asking about smoking history and prevalent hearing loss. Results demonstrate that 3 times as many smokers than non-smokers suffer from hearing loss. What type of study is this?

Time

Observe at one moment in time, but

may ask about period

of time

Cross-sectional study

WeaknessesTemporal ambiguityExposure status vulnerable to measurement errorPossible selection biasDifficult to distinguish prevalent cases from incident casesInefficient for studying rare diseases or diseases that are highly fatal

Cross-sectional Study

StrengthsRelatively easy, time-efficient, and cost-efficient Useful for screening new hypothesesUseful for health planning

Cross-sectional Incidence StudyCross-sectional Incidence Study• Did you experience the onset of depression during

the past month?• Did you come down with a cold during the past two

weeks?• Did a diarrhea episode (i.e., three loose stools in a 24

hour period) start during the past week?

Ask

Original cohort (i.e., source population)

loss loss loss loss

Asking question in the study population only of the survivors

Time period of reflection

Cross-Sectional Incidence Study

Subjects are sampled from a population and asked to recollect incident disease over a specified time period.Does not strictly result in an incidence rate, but may approximate the rate in the (unobserved) source population.Members of the source population who develop the outcome may be less likely to be observed in the study.In other words, disease status may affect the probability of being selected (sampled) in the study.

Cross-Sectional Incidence Study

Suppose there is a population of 6 people in 1995 that we want to follow for 5 years to measure incident disease.How would our estimate of incidence change if we try to survey this group in 2000 instead of following them prospectively?For example, “Were you diagnosed with diabetes in the last 5 years? If so, when?”

1995 1996 1997 1998 1999 2000

Time

X-Sectional Study

Subject 1

Subject 2

Subject 3

Subject 4

Subject 5

Subject 6

Source Population

Dead or lost to follow-up

Diagnosed with diabetes

[------------------------------------------------------------------- 5 Years --------------------------------------------------------------------]

Cross-Sectional Incidence Study

In the base population 3 incident cases occurred in 21 person-years at-risk for 6 subjects (0.14 cases/PY)In the cross-sectional study 1 incident case was reported in 14 reported person-years at-risk for 3 observed subjects (0.07 cases/PY)People developing diabetes were more likely to die or be lost tofollow-up than those not developing diabetes.This resulted in an underestimate of the incidence of diabetes.This type of problem can occur when:

the risk period is longlosses to follow-up or deaths are frequentlosses/deaths are associated with developing the outcome of interest

Proportional Mortality Study