Descriptive study

Hierarchy of epidemiological study design

Epidemiology is the study of the distribution and determinants
of disease frequency in human population.
Epidemiology: Principles and Methods

by Brian MacMahon

1996

10
The distribution of
Disease considers

Who is getting disease?

Where is disease occurring?
When is disease occurring?
11
Descriptive Epidemiology
Cases

Person Time
Place

Who? Where? When?

12
Descriptive study is to subset collected data or some survey data, included the
data from experiment, to groups by different places, different time or different
population character and descriptive the frequency of disease and health. Helps to
build the relationship between risk factors and disease or health, offer etiological
clues.

13
Descriptive epidemiology

 Descriptive epidemiology is concerned with the variations of morbidity

and mortality in a community.
 The information is not only useful in health care planning, but also can
provide clues to the potential causes of disease.
Highlights

 Introduction
 Steps in designing and conducting a survey
 Sampling
 Data gathering approaches
 Biases in health surveys
 Advantages and pitfalls associated with cross-sectional studies
Introduction

 A cross-sectional study (CSS), also called a survey or prevalence stu

dy, is an investigation that systematically collects information and do
es not use intervention.
 Notice: In epidemiological research information collected through su
rveys can be very broad and is not limited to diseases. This informati
on can also consist of attributes that affect disease risk, knowledge or
attitudes that affect health behaviors and the use of health services.
Introduction

 Depending on the research questions to be answered and statistical te

chniques being used, a cross-sectional study can be descriptive, anal
ytical or both.
 The term "survey" is used for the remaining of this chapter.
Mechanism Purposes and applications

Describes the distributing of disease or health events

within populations
To provide etiological clues
To identify high risk groups in the population
To evaluate the effect of surveillance ， vaccination
and disease prevention and control

27
Introduction

 Application of health surveys: important sources of health informati

on for evidenced based health care planning and provision, preventat
ive and clinical medicine.
 Examples (P49)
Introduction

Characteristics and categories

Characteristics
•No control group
•Particular point of time
•Limit to ensure the causal relationship

Categories
•Category
•Census
•Sampling survey

29
Steps in designing and conducting a survey

 A good plan is a prerequisite for a good survey study.

Defining survey objectives Defining survey variables Defining study population

Collecting data Deciding data collection questionnaire Deciding data collection methods

Analyzing and interpreting data Disseminating research results

Steps in designing and conducting a survey

 Defining survey objectives : before a study is even planned, the inv

estigator must clarify the purpose of the study and why it is importan
t. (Is the purpose of the study to estimate the prevalence of certain di
sease or risk factors? Will the study results help disease prevention o
r health care prevision?) After clarifying the initial thoughts of the st
udy, it's important to conduct a literature review and make sure you a
re not duplicating other's work.
Steps in designing and conducting a survey

 Defining survey objectives : It is likely that you may expect your su

rvey to answer a number of research question. In this case, it would
be helpful to list these questions in order of priority. Research questi
ons must be specific and able to be answered through the proposed st
udy.
 Example: P50
Steps in designing and conducting a survey

 Defining survey variables: there are two categories of information t

hat can be collected in a health survey: disease occurrence or health s
tatus variables, and exposure variables.
 Whether disease occurrence or health status can be adequately meas
ured in a health survey directly affects the overall validity of the stud
y.
Steps in designing and conducting a survey

 Defining survey variables: As most health surveys are based on ans

wers to questionnaire questions, measuring disease occurrence or he
alth status in a large population in a short time is more challenging in
comparison to case-control or cohort studies. Certian self-reported h
ealth conditions are more reliable than others.
 Examples: P51
Steps in designing and conducting a survey

 Defining survey variables: There are a number of procedures that c

an be used to improve or estimate the accuracy of self-reported disea
se status in health surveys. The least expensive way is to improve the
clarity of probing survey.
 Examples: P51
Steps in designing and conducting a survey

 Defining survey variables: The issues related to measuring exposur

es in health surveys are similar to those discribed in case-control and
cohort study design chapters. They can come from questionnaires, re
cords, laboratory tests, physical examinations and special procedures.
Steps in designing and conducting a survey

 Defining study population: a study population is the group of peop

le that is surveyed. Depending on the research questions to be answe
red, the study can be all members of certain group or a sample of the
population.
 Example: P51
 In literature, the study population from which the sample is chosen is
also referred to target population or parent population.
Steps in designing and conducting a survey

 Deciding data collection method: a survey may be conducted in va

rious manners, such as face-to-face interviews, self-completed questi
onnaires in a designated venue (schools, hospitals and factories), tele
phone interviews, postal service (mail surveys), website-based and o
thers.
Steps in designing and conducting a survey

 Deciding data collection method: the method of data collection is o

ften linked with the study objectives, research questions, characterist
ics of the target population, and research budget. Besides, the decisio
n in determining the mode of data collection often requires balancing
the pros and cons associated with each method.
Steps in designing and conducting a survey

 Deciding data collection method

Interview
Collecting data through person-to-person interviews dates back to the
mid-19th century. In 1842, John Snow traced the cholera outbreak in
London to a contaminated water well on the Broad Street through int
erviewing local residents.
Steps in designing and conducting a survey

 Deciding data collection method

Interview
There are two major types of interviews used to collect information: fa
ce-to-face and telephone interviews. Notwithstanding internet based
or YouTube based interview is likely to appear, it has not mainstream
ed yet and would not be considered as separate type in this chapter. T
hus, only the two classical data collections are being discussed.
Steps in designing and conducting a survey

 Deciding data collection method

Interview
Advantages: A face-to-face interview enables personalization of the st
udy to the participant and is often associated with a higher response r
ates. Interviewers can offer clarifications and assistance for those wh
o have difficulties in comprehending the survey question and finishi
ng the survey independently. Additionally, interviewers are able to re
cord the time and quality of the interview, which may be useful for d
ata analyses.
Steps in designing and conducting a survey

 Deciding data collection method

Interview
Disadvantage: Interview surveys engender extra costs form hiring and
training interviewers. For a large population based survey, the costs c
an be substantial. Another pitfall associated with the interview surve
y is that it is open to possible manipulation or interviewer bias. Inter
viewers’ knowledge about the study may consciously and sub-consci
ously affect the way they approach and ask the study participants qu
estions. Finally, personal interviews become difficult and inefficient
when it comes to accessing respondents in certain geographical areas.
Steps in designing and conducting a survey

 Deciding data collection method

Mail survey
Advantage: As a mail survey (postal survey) does not need interviewer
s and uses identical questionnaires, it is free from interviewer biases.
It is often cheaper compared to the interview-administrated survey a
nd offers a great accessibility to wide geographic areas. It provides s
ome convenience of time to the participants; thus it is more suitable f
or long surveys.
Steps in designing and conducting a survey

 Deciding data collection method

Mail survey
Disadvantages: Mail surveys are notoriously associated with low response rates. F
or example, a number of physician surveys conducted in the United States had r
esponse rates around 30% to 40%. It happens from time to time that participants
may interpret the same questions differently and mail surveys do not work well i
n low socioeconomic populations. Finally, the likelihood of unanswered questio
ns is high as some participants only answer those they think are important to the
m. In comparison with face-to-face interview surveys, postal surveys are more li
kely to be affected ny the general literacy level of the study population.
Steps in designing and conducting a survey

 Deciding data collection method

Wed-Based survey and electronic mail
There has been a tremendous increase in internet use in the past decad
e and consequently the way we communicate with one another has c
hanged dramatically. E-mails, websites, list serves, chat rooms, and n
ewsgroups allow people almost instant communication. Inevitably,
e-mail and website based health surveys have emerged in main strea
m health research. Data collection through the internet has its unique
advantages over other survey modalities.
Steps in designing and conducting a survey

 Deciding data collection method

Wed-Based survey and electronic mail
Advantages: The most salient advantage of the online survey is its low
costs. As the cost of computer hardware and software continues to de
crease, the costs for conducting health surveys is getting lower. The
data are automatically stored thus, it is time saving.
Steps in designing and conducting a survey

 Deciding data collection method

Wed-Based survey and electronic mail
Advantages: A second advantage is that Internet-based survey research
er to reach thousands of people with common characteristics in a sho
rt amount of time, despite possibly being separated by great geograp
hic distances.
Steps in designing and conducting a survey

 Deciding data collection method

Wed-Based survey and electronic mail
Advantages: It works particularly well for often related to social stigm
a and it takes advantage of the ability of the Internet to provide acces
s to groups and individuals who would be otherwise difficult, if not i
mpossible, to reach. For example, information regarding individuals’
sexual behaviors or sexually transmitted diseases(such as syphilis an
d HIV) is often difficult to obtain or biased through telephone and pe
rson-to-person interview because they are stigmatized offline.
Steps in designing and conducting a survey

 Deciding data collection method

Wed-Based survey and electronic mail
Disadvantage: One limitation of the web-survey is that it very unlikely,
if not impossible, to obtain a representative sample of the general po
pulation. Obviously, people without access to internet cannot possibl
y participate a web-based survey. Additionally, multiple answers fro
m the same individual can occur. As well, it is very difficulty to verif
y certain information provided through internet.
Steps in designing and conducting a survey

 Deciding data collection method

Mixed survey approach
Two or more approaches are used in one study.
.
Steps in designing and conducting a survey

Sampling
 sampling refers to the process and method of selecting study particip
ants from all potential participants in a particular group or populatio
n.
 The sample size is the number of subjects required to address the res
earch question of interest.
Steps in designing and conducting a survey

Sampling
 The study population refers to the collective of individuals in the stu
dy of interest. It can be composed of medical doctors, university stud
ents or the entire population in a given geographic area or administra
tive body.
 If a study’s sampling is flawed, the results will not be useful and
even be misleading.
Steps in designing and conducting a survey

54
Steps in designing and conducting a survey
Sampling
Study sample
Classification of Sampling

Probability Non-probability
sampling sampling

Simple
Systematic Stratigied Cluster Convenience Quota Purposive Snowball
random
Steps in designing and conducting a survey

56
Steps in designing and conducting a survey
Random Sampling -Chance sampling

57
Steps in designing and conducting a survey

Random Sampling -simple random sampling

 the basic building block and the most intuitive one among all sampli
ng techniques.
 the sample is a form of probability, in which each individual in the p
opulation has an equal chance for being included in the sample. Theo
retically, this technique is free from bias.
Steps in designing and conducting a survey

Random Sampling-simple random sampling

 In modern time, the usual approach is the use of computer.
 Example: P54 Table 4.1.
Steps in designing and conducting a survey

Random Sampling -simple random sampling

60
Steps in designing and conducting a survey

Random Sampling-simple random sampling

 Despite being often used as an example, the flip of a coin may not co
nstitute a random selection as it is hard to imagine that any coins wo
uld have a perfect weight distribution on both sides.
Steps in designing and conducting a survey

Random Sampling -simple random sampling

62
Steps in designing and conducting a survey

Random Sampling-systematic sampling

 In this sampling method, study subjects are selected from the populat
ion at a regular interval, that is every k th (e.g. third, fifth, tenth) subje
ct is selected.
 Example: P54 Table 4.1.
Steps in designing and conducting a survey

64
Steps in designing and conducting a survey

Random Sampling-stratified sampling

 In this sampling method, researchers first divide the study populatio
n into strata, or groups based on certain characteristics (e.g. age grou
p, gender and province) and random sample is drawn from each strat
um.
Steps in designing and conducting a survey

Random Sampling-stratified sampling

Advantages compared to simple random sampling
1. It can ensure a better representation for each subgroup.
2. When the population is divided into sub-groups, more precise estim
ates of the population parameters of the population parameters can b
e obtained.
Steps in designing and conducting a survey
Example: stratified sampling

67
Steps in designing and conducting a survey

Random Sampling-cluster sampling

 Cluster sampling is a sampling technique where the entire population
is divided into groups, or clusters and a random sample of these clust
ers are selected. All observations in the selected clusters are included
in the sample.
 It can be understood as simple random sampling as described except
that the basic sampling unit is a cluster or group of people instead of
individual.
Steps in designing and conducting a survey

Random Sampling-cluster sampling

 Cluster sampling is typically used when the researcher can not get a
complete list of the members of a population they wish to study but c
an get a complete list of groups or clusters of the population.
 It is also used when a random sample would produce a list of subject
s so widely scattered that surveying them would prove to be far too e
xpensive.
 Example: P55
Steps in designing and conducting a survey

Random Sampling-cluster sampling

 Difference between cluster sampling and stratified sampling: the for
mer takes a random sample of clusters (often small groups) and inclu
des every subjects in the selected clusters; the latter divide the popul
ation into large strata (often large group of people) and takes random
a subset of individuals from each stratum.
Steps in designing and conducting a survey

Random Sampling-mixed sampling

 Utilization in combination of two or more sampling methods
 Multistage sampling is a typical form of mixed sampling, in which la
rger sampling units are first selected from a population, and then sma
ller units are sampled from within each larger sampling unit.
 Multistage sampling is commonly used in larger national surveys.
 Example: P55
Steps in designing and conducting a survey

Nonprobability sampling
•In probability sampling, as describe previously of a subject
being selected can be known and the representativeness of sample
for the population can be estimated.
•However, in certain circumstances, probability sampling may
not be feasible and the researchers may decide to employ non-
probability sampling.
Steps in designing and conducting a survey

Nonprobability sampling
•In contrast to probability sampling, the probability of a person b
eing selected is unknown and, consequently, the claim for the repres
entativeness of the population cannot be made, and varies.
Steps in designing and conducting a survey

Nonprobability sampling
•Not-withstanding its limitations in presenting the population,
nonprobability sampling is also an important tool and sometimes can be
uesd to glean important insights.
It is often used in pilot or feasibility studies before large scale population
surveys are undertaken.

74
Steps in designing and conducting a survey

75
Sample size determination

How big should a sample be?

76
Sample size determination
Determine the sample size
Formulas for quantitative data( mean)

Z  2 2
n 2
d
Formulas for qualitative data （ proportion ）

2
Z PQ
n
d2
77
Sample size determination

78
Sample size determination

Factors influencing sample size

Type I error: The probability of incorrectly rejecting the null hypothesis and declar
ing that a difference exists when it does not. In medical research, it is a conventi
on to set this error at 5%(P=0.05), it is called statistically significant. The corres
ponding two tailed Zα value under the standard normal distribution curve is 1.96.
Sample size determination

Factors influencing sample size

Type II error: The probability of incorrectly accepting the null hypothesis and decl
aring that a difference does not exists when it does. The corresponding one taile
d Zα value under the standard normal distribution curve is 0.84.
Sample size determination

Factors influencing sample size

Statistical power: "1-β" is the probability of detecting a true difference and is calle
d statitical power or simple power. The power of a study is influenced by a num
ber of factors including the frequency of the conditions under study, the magnitu
de of the expected difference, and sample size.
Sample size determination

Factors influencing sample size

Degree of variability: refers to the average difference among subjects with respect
to the attributes being measured in the population. The more heterogeneous a po
pulation, the larger the sample size required to obtain a given level of precision.
The less variable a population, the smaller the sample size.
Sample size determination

Factors influencing sample size

Level of precision: It is sometimes called sampling error, is the range in which the
true value of the population is estimated to be.
Data gathering approaches

 A questionnaire is an essential tool for epidemiological

findings collecting information in survey studies.
 Epidemiological findings are often based on responses to
questions in the questionnaire.

84
Questionnaire

 A well-designed questionnaire meets the research objectives.

 A good questionnaire is able to obtain the most complete and
accurate information possible.
 It would keep the interview brief and to the point and be so
arranged that respondent(s) remain interested through the
interview.

85
Questionnaire

 Survey question can be either closed-ended or open-ended.

 In closed-ended question, possible answers are provided for
participants to chose. The usual types for closed-end question
include dichotomous, multiple choice, rating, and ranking.

86
Questionnaire

 Open-ended question generally consist of a blank space where

the study participants can describe their answer.
 Example: P59

87
Data sort and analysis

Check the integrity and accuracy of data

Ø Classify and verify data by standard

88
Target Population or Sampling Frame
Census or Sampling Survey

Sample

Disease Disease Non-disease Non-disease

And and and and
Exposure Non-exposure Exposure Non-exposure

89
90
Example of Cross-Sectional Study
Hypothesis:

Obesity is a risk factor for knee osteoarthritis

Sample:

100 retirees living at “University Village”

91
Medical exam + X-rays to diagnose osteoarthritis of the knee

Osteoarthritis
+ -
+ 40 10 50
Obesity

- 20 30 50

92
Prevalence of osteoarthritis among obese subjects:
40/50 = 0.8
Prevalence of osteoarthritis among non-obese subjects:
20/50 = 0.4
Prevalence ratio = 0.8/0.4 = 2.0

93
Obese subjects are two times more likely to
have osteoarthritis of the knee than non-obese
subjects.

94
Bias in health surveys

Sample biases: are systematic arrors between the study sample and its target popul
ation.
Bias according to the Last dictionary

 Deviation of results or inference from the “truth”, or processes

leading to such deviations.

 Any trend in the collection, analysis, interpretation, publication,

or review of data that can lead to conclusions that are
systematically different from the “truth”.

96
Biases and control

Common type of bias

•Selection bias
•Non-response bias
•Survivor bias
•Report bias
•Measurement bias
Control of bias
Bias in health surveys

Selection bias: selection bias occurs when individuals do not have an equal opport
unity of being chosen. It normally is not a concern in probability sampling, but s
ometime can be problematic.

Example: P60
Bias in health surveys

Non-responce bias:
 some eligible individuals may choose not participate in the study for various rea
sons.
 Non-response may pose a threat to the generalization of study results.
Bias in health surveys

Non-responce bias:
 A high non-response proportion means a reduced sample size and precision.
 When non-respondents are significantly different from participants with respect
to diseases or risk factors of interest, the results can be biased.
 People with lower social economic status are less likely to take part in health sur
vey.
Bias in health surveys

Interviewer bias
 In an interviewer-administered survey, an interviewer's opinions and prejudice
may either consciously or subconsciously affect the way how she/he conducts th
e interview.
 This is more likely to happen with open-ended questions.
Bias in health surveys

Interviewer bias
 Therefore, it is important for researchers to make sure that the interviewers cond
uct the interview with total objectivity and respondents are not influenced by an
y outside source in their responses.
 For this reason, interviews should be conducted by well trained and qualified int
erviewers.
Advantages and pitfalls

Advantages
 Less expensive
 Useful for planning services
 Shows relative distribution of conditions
 Shows interrelatedness of attributes and conditions
 Does not rely on individuals who present for medical treatment
Advantages and pitfalls

Disadvantages
Do not establish the true temporal sequence of events.
Does not yield incidence or true relative risk.
They are not feasible for the cases of death, recovery and short
course.
Not effective if the disease is rare.