At ProQOL.org we believe the continuing development of our field is in continuing our research.
The ProQOL is the world's most commonly used measure of the positive and negative aspects of helping others who have experienced great sorrow or traumatic stress. There are hundreds of published documents about research and use with the ProQOL. You may wish to consult PubMed (free access), the U.S. National Library of Medicine of the Institutes of Health, the PILOTS database (free access) of the VA National Center for PTSD, Google Scholar (free access) and PsychInfo (access fee may apply) of the American Psychological Association. A bibliography of over 1,000 papers through 2010 can be found on this site.
Can I use the ProQOL?
If you want to use the
ProQOL for research you are free to do so under the terms and conditions
below. If you need a specific written permissions can be accessed at the Request Use Permission page.
The ProQOL measure may be freely copied and used as long as (a) author
is credited, (b) no changes are made other than those authorized below,
and (c) it is not sold. You may substitute the appropriate target group
for / [helper] / if that is not the best term. For example, if you are
working with teachers, replace / [helper] /with teacher. Word changes
may be made to any word in italicized square brackets to make the
measure read more smoothly for a particular target group.Additionally
you are granted permission to convert the ProQOL into other formats such
as a computerized or taped version for the visually impaired.
Conducting Statistical Analyses with the ProQOL
Do the norms published in the manual apply to my group? The norms in the manual are the best available at this time. We
do not have norms for specific groups. In general we believe that the
main group norms can apply to any helping profession. We cannot
guarantee this is true but it is the best available information we have
at this time.
Should I use the raw scores or grouping scores (high, normal or low risk)? If you are doing research with the ProQOL it is probably best
to use the raw scores rather than the categorical ones (e.g. high, normal and low risks). We always recommend that researchers use more specific, continuous data when it is possible. Because the categorical scores are so non-specific, error can be introduced by using them.
It may be best to use the categorical scores if
you are interested in how people group across the scores, for example,
in ANOVA where you have categorical comparisons. You may choose to create your own groups based on the raw scores. For example, you may wish to create groups based on 10 percentile ranges rather than on the three used for the self-score method.
The most important thing to do is to understand your question of
interest. This is probably not your hypothesis. The Question of Interest
is what you want to know and the hypothesis is a way of setting
up your analysis to answer your question. For example, you may be
concerned about the neonatal nurses in your hospital. You are guessing
the job is more stressful on those nurses. You may be wondering if a
special training course would be useful. Your hypothesis would be
Neonatal Nurses experience greater stress or Neontaal nurses do not
experience more stress than general practice nurses. If your hypothesis
was significant, you probably would chose to implement a program.
How do I compare my sample's scores to the scores published in the manual? There are several ways of computing scores to compare to the manual.
Below are two of the most common ways. If this does not make sense to
you please find someone with statistical experience who can assist you.
You can compare your group to itself. If you go this route you use
the means and standard deviations of your own group. You may wish to
compare the scores of your group to the means and standard deviations
reported in the manual. If you want to do this you use your raw scores
and compare them to the raw scores in the manual. This is often done
using a one sample t-test comparing the means and standard deviations of
your group to the norms reported in the manual. The center of the sample data is represented by the mean (X bar) and the norms are represented by mu. The variance of the sample data are represented by the standard deviation (SD) and the population variance (sigma squared). You may need the Standard Error
of the Mean (SEM), which is also reported in the manual.
You can also do an ANOVA or a variant of that by making one group the
raw scores from the manual and the second group your scores. You can
decide how you wish to compare your scores. In general people use the
t-test noted above. Another alternative, although less statistically
satisfying and one that introduces error into the analysis, is to simply assign the mean and standard deviation of the
manual score as a hypothetical group. Because there will be no variance
in this "group" you make from the manual scores, you will have to
introduce some variance into your hypothetical sample. This can be done
in multiple ways, including setting the means and standard deviations by
the SEM or using the variance. You will have to vary the scores but
your final mean, standard deviation and SEM should average to the manual
numbers. There are some computer programs that will do this type of
calculation for you or you can do it using a program like Excel or by
What does it mean if my results are different from the published norms?
If you are comparing your sample's raw scores to the published raw
scores and there are differences, you probably have a group that is
different from the larger sample. This can be good or baffling. For
example, if you have a sample of nurses and their scores differ from the
general samples reported in the manual, you probably have found a true
Sometimes there are differences but they do not make theoretical sense.
In those cases it is useful to check your analyses types or the computer
code. It is also important to make sure that each of your data points
are aligned with the right subject and so forth. It is very easy to get
data misaligned. You certainly will not be the first researcher that
this has happened to! In an early data collection on the ProQOL, the average number of children for the subjects was 42.3 children. A quick look at the data showed that the column for number of children was mixed up age of participant.
My alpha reliabilities are different than what is published in the manual, what does that mean?
If your alpha reliability is different than the ones reported in the
manual, the ProQOL works differently in your specific sample than the
aggregate sample from the manual. If your alpha reliabilities are substantially lower than what is
reported in the manual the measure may be a poor one for your sample and care should be used in interpreting your results. If
the alpha reliability is higher, the ProQOL is a particularly useful and reliable measure
for your sample. You may find that some scales are better for your sample
than others. You must decide based on your reliability scores, if you
are comfortable with using a particular scale with your sample.
My factor structures are different than what is published in the manual, what does that mean?
If your factor analytic structure is different from the manual, consider
three things. The first is the simplest. You should check to see that you used the same type of factor analysis. For example, principal componants analysis will yield different structures than a common factor analysis. You can read more about this below in the factor analysis section.
Second, you may be observing the recognized overlap between the Depression
and STS scales that can lead to a two factor solution or a three factor solution with fewer items. If you observe that you can measure these scales with
fewer items, you are correct. The items that could be deleted are included for user friendliness and the items are
statistically neutral, that is, they do not change the statistical
characteristics of the scale. More information about this issue is in the next section which discusses the correlation between the Burnout and STS scales.
The third, and probably the most important thing is to examine the similarity of your sample with the data used to create the ProQOL. Are the two comparable? It may be that your factor structure is based on your sample and may or may not apply to the population used for the ProQOL development. For example, if your sample is of teachers in high-risk scools the factor structure describes your sample of teachers in high-risk schools, not the universe of people who could take the ProQOL. The data bank includes people from many fields, males and females and multiple countries.
There are high correlation between the Burnout and STS Scales, what does that mean and why should I use both scales?
Clinical treatment and previous research has identified a
strong relationship between depression and Traumatic Stress; depression
and trauma are frequent co-travelers. The correlation between the two
scales acknowledges this overlap assigning as best a possible the unique
variance of each. However, since theoretically items about depression
and trauma can belong to both scales the two scales share variance.
Items that load highly on both scales are also retained for practical
reasons. Based on the information we received from focus groups, it is
easier for people to understand three subscales with 10 items each than
one subscale with 10 items, one with 5 and one with 7. Another
practical reason is writing computer code for statistical analysis. When
you have the same number of items on each subscale you do not have to
adjust the denominator across comparative analyses.
The items that are included to balance each scale to 10 items are statistically neutral to the item to scale and factor structures. That is, they make no statistical difference if they are or are not included. They are in effect place holder items.
I want to do factor analysis on the ProQOL, how to I do that?
Below is a quick summary of Factor analyses and Principal Componants
Analysis. Readers may be interested in the following two papers: Principal Componants Analysis and Factor Analysis. There is no relationship between the authors of these papers and the ProQOL.
Depending on your question of interest, It may be
useful to conduct a factor analysis. The data reported in the manual are
analyzed using Common factor analysis (CFA) and multigroup factorial
analysis (MFA), not a Principal Components Analysis (PCA). All of these analyses can be
useful but they have different goals and mathematical procedures so it
is important to match your analysis to your question of interest. Each are discussed below.
Principal Components Analysis (PCA)
In PCA, the goal is to extract the variance accounted for
among the variables measured iteratively. PCA creates its synthetic variables (e.g. the factors or componants) iteratively. The procedure first identifies the factor that accounts for
the most variance. For the second step it subtracts the variance accounted for by
the first component and then finds the componant that accounts for the
most variance in the remaining variance. The creation of componants (factors) continues until all of the variance is accounted for. In this way it is like a step-wise
regression. Thus the first componant (factor) is drawn from all data. The second is
drawn from all of the data minus the variance used in the first componant (factor) and so
A PCA may produce 5, 10, 15 or more componants. At what
point is it useful to stop selecting components? How many does it take
to truly represent the data collected? How many componants (factors) account for the variance? For this test an egin value analysis or scree plot may be useful in determining when it is appropriate to stop including componants. For example, if the first componant (factor) accounts for 65% of
the variance and the second 23% followed by a few componants that together
account for the remaining 22%, is it better to use the first two or should you use more componants? The
first two componants (factors) tell you what happens with 88% of the variance and adding
other factors may not be useful and might even introduce error variance
into your equations.
Because the goal of the test is to extract the
first componant (factor) based on its accounting for the most variance and the
remaining variance the principal component with the most variance
accounted for, rotating the components (factors) is not only unhelpful
it literally alters the goal of PCA.
Common Factor Analysis
In common factor analysis (CFA) the synthetic variables (factors) are
extracted simultaneously, as opposed to PCA where they are extracted iteratively based on the size of the variables' contributions. In CFA the
factors are computed based on similarity of variance within any one
variable. That is, a single variable is parsed out to the different
factors based on how it contributes to the factors. For example, the
variance for gender may be split into three factors in varying
proportions. The correlation of gender to Factor 1 could be .36; to
Factor 2, it could be .19; and to Factor 3 it could be .07. The decision
how many factors to use for the variance in depends on your theoretical
expectation. In general, most people would use the .36 but not the .07 as
it is a very small, probably inconsequential contribution. You can
examine this further by leaving that variable out and seeing if the
structure of the factors are similar or different. In theory because the variable loaded highly on Factor 1
it should change. Factor 2 might change and Factor 3 probably would not since there was such a tiny contribution made by the variable (.07).
extracts the factors simultaneously, there are an infinitely likely
number of potentially equally likely solutions. Rotation is useful here
in that it presents you with an outcome that is most likely the most
commonly occurring. A scree plot may be very useful in identifying factor structures.
In this analysis I used SPSS with GLS and ML
extractions, varimax rotation with three models considered with set two,
three and four factor solutions. A three-factor model was selected
because it fit with the theory and because there was unique contribution
for all three factors. Burnout and STS might be collapsed given that
they are both statistically and theoretically correlated based on
depression-type symptoms, but they retain unique values that separate
the two, primarily the element of fear associated with STS.
Multigroup Factorial Analysis
The data are also analyzed using a
multigroup factorial analysis. In MFA data are compared based on their
CFA structure across groups. For example, Factors 1 and 2 may be
analyzed across gender and profession. If the factors are different for
gender and not different for professions, it is important to treat
gender as one grouping variable with two subgroups, males and females. Since there are no
differences across professions, these can be conglomerate into a single
An example of how to use this in an analysis can be seen in a multigroup analysis of variance or a nested ANOVA. Scores on the ProQOL are the dependent variables (scale scores) and your Independent (grouping)
variables would be gender and professions. You would need to have
subgroups (male and female) for gender but not for professions. For example, if your
professions group included nurses and teachers you might find that male
helpers might be different than female helpers. In this analysis data
were analyzed across groups expected to be different; e.g., gender and
professions as well as year of data collection and across randomly drawn samples from the larger
You can participate in the continued development of the ProQOL We invite
you to consider donating a copy of your raw data to the ProQOL
databank. We depend on these donations to build the psychometric
properties and norm the ProQOL. Please know that your donated data will
never be published so that it can be identified with your research. For
example, we have collected data from multiple studies of nurses. The
data are merged as nursing data, not specific project data.To find out
more about contributing to the ProQOL effort, go to the Donate Data page.
The ProQol is a volunteer effort. As our time allows, we consult
with student and professional colleagues in regard to their research on
professional quality of life, compassion satisfaction, burnout and
compassion fatigue. Many of our consultations are ones where
cross-cultural issues and issues of indigenous people are paramount.
While we cannot always donate our time, as time and resources permit, we
do work pro bono.
At times, students ask us to serve on their theses and dissertation
committees or to act as content experts for their advanced degrees. We
are happy to discuss these options with you and participate as time and