Question

What do you think are ethical issues relevant to statistical data collection and reporting?

Answer 1

There are several ethical issues which must always be considered when planning any type of data collection. Data collection always costs someone something. It may cost health workers' time and energy to complete surveillance forms. It certainly costs the health coordinating organization money and time to collect, analyze, interpret, and disseminate surveillance data and results. Surveys are even more resource intensive. Data collection also costs the people in the population from which the data are collected a certain amount of time, discomfort, and potential harm.

In addition, implementing or revising programmes in response to the conclusions drawn from data collected always cost manpower, time, money, and other resources. And if the conclusions are wrong because the data were poorly collected, these resources, which could have been used otherwise, may be wasted or inefficiently employed.

Therefore, before beginning the planning process, be sure that the results of the data collection will:

• Truly be needed,
• Be disseminated widely,
• Be used to implement or revise a program, and
• Use the least intrusive and costly data collection method possible

Nonetheless, keep in mind that data collection in emergency situations is necessary to guide program decisions. Collection of data necessary for this purpose should not be delayed if the data collection poses only minimal risk to individuals or groups

Regardless of the type of data collection, it is absolutely necessary to gain the approval of the community from which the data will collected. This is nice to state in theory, and everyone always agrees with it, but it is also true on a very practical level

Although many consider the reporting of research
data to be a purely objective task, the availability
of numerous statistical analysis techniques and re-
porting techniques presents numerous options,
some of which are more scientifically valid and eth-
ical than others.
External forces may negatively impact the hon-
est and accurate reporting of research data. Re-
searchers may wish to present their work in the
most favorable light to improve chances of accep-
tance for publication or presentation. Although all
of the following examples may not necessarily be
immoral or unethical, a few potential examples of
inaccurately reporting data may include:
1. Failing to include number of eligible partici-
pants. For example, if a large number of eligible
participants refused consent, or were not included
for other reasons, failure to report this may falsely
mislead the reader to believe that study partici-
pants are representative of the entire body of eli-
gible participants.
2. Inaccurate reporting of missing data points. For
example, in a study with a large number of partic-
ipants, but high fractions with missing data points,
the reporting of data could be manipulated to cam-
ouflage missing data points.
3. Failing to report all pertinent data. Some re-
searchers purposely neglect to report sections of
data that are inconsistent, or unexplainable, with
other results. This unscientific selection of data
points is misleading.
4. Failing to report negative results. Although pos-
itive results appear more interesting, and may
carry increased probability of acceptance for pub-
lication, negative results are scientifically as im-
portant to report. A failure to demonstrate a sta-
tistically significant difference between groups is
not a failure to demonstrate meaningful results.
Demonstration of negative results may be valuable
to the medical community, and may obviate the
need for additional financial or time investments
from other researchers. Particularly in this era of
cost containment, the demonstration of a lack of
benefit of a new therapy may be valuable.
5. Allowing research sponsors to influence report-
ing of results. For example, pharmaceutical spon-
sors may pressure researchers to report only re-
sults favorable to their product or may prohibit
presentation or publication of the results alto-
gether.
6. Inappropriate graph labels. For example, show-
ing only a fraction of the y-axis scale, or unclear
axis labeling, may magnify small differences be-
tween data points.
7. Reporting percentages rather than actual num-
bers. Although this may be appropriate in some
cases, if the intent is to deceive, it represents in-
accurate and perhaps unethical reporting.
8. Reporting results of inappropriately applied sta-
tistical tests. For example, several statistical tests
may be applied to the same data set, and the re-
searcher may elect to report only the test that
yielded the most favorable results, rather than de-
termining a priori the most appropriate statistical
tests to use.
9. Reporting differences, when statistical signifi-
cance is not reached. For example, although the
p-value is not significant, some researchers report
a ‘‘trend’’ toward significance. This is inaccurate,
as, per convention, unless otherwise stated, a p-
value $ 0.05 is accepted as not significant; that is,
a difference was not found. Some authors suggest
that reporting confidence intervals is a superior
method of reporting, because of improved infor-
mational content, in a more explicit and precise
format.1,2
10. Reporting no difference, when power is inade-
quate. If the study sample size is too small, a type
II error may be committed. Although power of 0.8
is acceptable by convention, this allows a 20%
chance of incorrectly accepting the null hypothesis.
In some cases, researchers should consider being
more fastidious, and strive for a power of 0.9 or
higher. In any case, alpha and beta should be con-
sidered prior to beginning the study .