Secondary Data Analysis in Health Care Research
Secondary Data Analysis,Economical Category,Analysis of Secondary Data,Question of Using Clinical Nursing,Sample Biases of Clinical Database,Caveats Needed for Data Analysis.
Secondary Data Analysis
Secondary data analysis uses the analysis of data that the analyst
was not responsible for collecting or data that was collected for a different
problem from the one currently under analysis.
The data that are already
collected and archived in some fashions are referred to as secondary
information (Stewart, DW, &Kamins , 1993). Statistical meta analysis might
be considered a special case of secondary analysis (see Meta-analysis).
Economical Category
Secondary information is an inexpensive data source that
facilitates the research process in several ways. It is also useful for
generating hypotheses for further research. It is useful in comparing findings
from different studies and examining trends.
Steward and Kamins (1993) point
out that population data sets, such as Bureau of the Census data, may be used
to compare sample to population characteristics in order to examine the
representativeness of the study sample.
Analysis of Secondary Data
The analysis of secondary information is a useful strategy for
learning the research process. The secondary data sets that have used optimum
sampling techniques provide an optimum resource for students by virtue of the
quality of sampling and the time and expense involved in data collection.
Given
that students are expected to understand, explain, and defend the data set in
terms of purpose, sample selection, methods, and instruments, only the
real-life collection and recording of data remain unexperienced by the student.
A further virtue of using the analysis from secondary information while
learning to do research is that it protects the pool of potential research
participants and agencies for participation in studies conducted by qualified
researchers.
Every research study is conducted with a specific purpose in mind.
Delimitations are specific to the original study and introduce specific types
of sampling and other bias into the original study.
Operational definitions may
not be replicable in a second study. For learning purposes, differences in the
original study and data set can be handled through careful critique processes
by students. However, the biases and differences that exist may be too extreme
to permit a valid secondary analysis outside the practice situation.
Archived data sets are rarely held in the form of raw data because
the data is usually summarized. The summarization may or may not be appropriate
for the research question under consideration for secondary analysis. To
analyze such data further confounds results beyond acceptable limits.
Question of Using Clinical Nursing
The question of using clinical nursing data sets for secondary analysis
comes with the advent of clinical nursing information systems. The use of
clinical databases as research data sets must be carefully examined. One
difficulty is that restricted data resources force clinicians to carefully
choose which data to collect. These data are usually not identical with what
the researcher needs.
Sample Biases of Clinical Database
Beyond data restrictions another major difficulty is that the
sample biases of clinical databases and research data sets for randomized
control studies are different. This difference in bias of the data from
clinical databases and randomized controlled trial research data sets can be
exploited as a strategy for doing cross-design synthesis.
However, this special
case aside, the issue is that of sample representativeness. The research sample
is selected for a specific reason, with specific delimitations in mind, to be
representative of the general population.
In contrast, the clinical population
from which the clinical data set is drawn is representative only of that type
of patient or client on whom data is being collected in that location and
rarely, if ever, typical of the general population or even all persons with
that clinical problem.
For example, patients with congestive heart failure in
Alabama are not necessarily representative of patients with congestive heart
failure in New England or California. The same is true of patients with
congestive heart failure in a community hospital versus those in a teaching
hospital in the same county.
Caveats Needed for Data Analysis
These caveats need close evaluation of data sets to be used for
secondary analysis. The information needed for such evaluation must be archived
along with the data set. Such information includes study purpose; data
collection details, such as who collected the data, when, and where; sampling
criteria and delimitations; known biases; operational definitions; and methods
of data collection.
Traditionally, nursing has not archived research data sets of its
own for use in teaching or secondary analysis. Nursing students and nurse
researchers do use large government databases, but none are collected
specifically by nurse researchers to answer nursing research questions.
This is
a problem to the extent that learning takes place best when examples and
experiences relate closely to daily (nursing) experience. Certainly, problems
peculiar to but not exclusive to nursing research are more easily taught with
examples from real life.
This is a problem also to the extent that nursing
research data sets can, in fact, generate new knowledge, whether by reanalysis
or by stimulation of further investigation and hypothesis generation.
Sigma Theta Tau International has begun a program to archive
selected research data sets of nurse researchers. The project is still in its
infancy, with acquisition and dissemination policy still under study (see Data
Stewardship). Descriptions of the research study will be required to fulfill
criteria for data set evaluation mentioned above.