Retrospective Versus Prospective Cohort Study Designs - Lippincott ...

WJ3504_391-394.qxp 6/23/08 10:13 PM Page 391
J Wound Ostomy Continence Nurs. 2008;35(4):391-394.
Published by Lippincott Williams & Wilkins
CE
WOUND CARE
Retrospective Versus Prospective
Cohort Study Designs for Evaluating
Treatment of Pressure Ulcers
A Comparison of 2 Studies
Michael Clark
The effect of interventions designed to help prevent or treat
pressure ulcers can be assessed through a number of study
designs including retrospective and prospective cohort studies.
This article highlights the strengths and weaknesses of these
2 approaches to data collection and analysis. Retrospective
studies provide for analysis of large amounts of data with less
investment, while prospective cohorts may capture clinically
relevant variables missing from retrospective data sets.
Prospective studies may also gather data in a more consistent
and accurate manner. However, ensuring comparability
between the various study groups (patient groups managed
with different products or interventions) remains a challenge
for both prospective and retrospective cohort studies. In retrospective
cohort studies, allocation may be based on arbitrary
reimbursement decisions, while prospective cohort designs may
mask unequal distribution of key risk factors.
■ Introduction
A growing body of evidence exists that focuses on the treatment
of pressure ulcers. Considering randomized controlled
trials (RCTs) alone, there are now at least 95 such studies reported
worldwide. 1 Fifteen of these studies compare different
pressure-redistributing support surfaces. They vary widely in
methodologic quality, and a clinical guideline panel commissioned
by the UK National Institute for Health and Clinical
Excellence conclude that most are weak. 1 One of the surprising
problems affecting these studies is the failure to consider
the size of the pressure ulcer at the time of study recruitment.
For example, 6 of 15 RCTs comparing various support surfaces
failed to report the initial size of the pressure ulcer. Among the
7 studies in which initial ulcer size was reported, there was no
apparent comparability in the size of the wounds allocated to
the different interventions. Failure to control for or report pressure
ulcer size when selecting subjects is a significant weakness
in methodology, since it is well known that large pressure ulcers
demonstrate a more dramatic reduction in surface area as
compared to smaller pressure ulcers. 2 Such observations
prompted a recommendation in the early 1990s from the US
National Pressure Ulcer Advisory Panel 3 for investigators to
stratify subjects based upon the initial size of their pressure
ulcer. More than a decade later it appears that this recommendation
has been largely overlooked to the detriment of
our ability to interpret study findings.
While RCTs are considered the “gold standard” in comparative
studies, prospective 4,5 or retrospective 6 cohort
studies also provide valuable data regarding the effects of
various interventions on pressure ulcer healing. Each of
these designs has both strengths and weaknesses that must
be considered when interpreting study results. One
advantage of prospective designs is the ability to control
factors such as wound measurement techniques. This
inability to control specific factors is a limitation of retrospective
analyses of administrative data sets. Retrospective
analyses may suggest the effect of specific interventions but
the accuracy of the analysis is limited by the methods used
to calculate wound size. For example, simple measures of
wound length multiplied by width (the most commonly
available indication of wound size in clinical records) 7 may
overrepresent the surface area of pressure ulcers by up to
30%. 8 This article addresses the strengths and weaknesses
of retrospective and prospective cohort studies, with implications
for interpretation of study results.
■ Retrospective Cohort Studies
One advantage of retrospective studies is the ability to
analyze data obtained from large numbers of subjects with
a comparatively small investment of time and modest
cost. The advantages and disadvantages of this approach
Michael Clark, PhD, Senior Research Fellow, Wound Healing
Research Unit, School of Medicine, Cardiff University, Cardiff,
United Kingdom.
Corresponding author: Michael Clark, PhD, Wound Healing
Research Unit, School of Medicine, Cardiff University, Cardiff CF14
4XN, United Kingdom (wsrmc@cf.ac.uk).
Copyright © 2008 by the Wound, Ostomy and Continence Nurses Society J WOCN ■ July/August 2008 391

WJ3504_391-394.qxp 6/23/08 10:13 PM Page 392
392 Clark J WOCN ■ July/August 2008
can be illustrated by review of a large-scale retrospective
cohort study reported by Ochs and colleagues 6 in 2005.
Ochs and colleagues 6 compared pressure ulcer healing
rates among nursing home patients managed with group 1,
group 2, and group 3 support surfaces. The 3 groups of support
surfaces are based on US Centers for Medicare &
Medicaid Services guidelines for coverage of support surfaces
in the home healthcare setting. They can generally be classified
as nonpowered support surfaces (group 1), powered
support surfaces (group 2), and air-fluidized therapy
(group 3). Data were extracted through a retrospective analysis
of clinical notes, with formal reliability checks between
the trained data extractors; this attention to interrater reliability
represents a strength of the study. However, the key
outcome measure, wound size, was calculated from reported
maximum wound length multiplied by width. The inability
to ensure the accuracy of these recorded measurements is a
limitation of this methodologic design.
Data were obtained from administrative data sets. The
following criteria were used for inclusion in the study:
(1) the resident had to be present in the facility for at least
14 days; (2) have 1 or more pressure ulcers; and (3) placed on
a group 1, 2, or 3 support surface. This yielded a total sample
of 664 subjects drawn from a potential population of
2,486 adult nursing home residents. However, it was not
clear how many long-term care facilities contributed this
total of 664 residents, and no comments are provided describing
any potential differences in participating facilities,
such as nutritional support and topical wound treatments,
that may have influenced study results.
Data were then analyzed in 2 different ways: (1) analysis
by person, defined as the rate of healing for the largest pressure
ulcer per resident over a minimum of 5 days and (2)
episode analysis, defined as the rate of healing for all pressure
ulcers per resident tracked over 7- to 10-day episodes. This
latter approach, which broke each resident’s stay into discrete
7- to 10-day blocks (or episodes), required a sophisticated
statistical analysis that attempted to account for
potential bias inherent where multiple episodes from multiple
ulcers are combined. Despite the mathematical sophistication
of this episode analysis, the independence of each
episode from other episodes gathered from the same nursing
home resident is doubtful. If a pressure ulcer is healing
in one episode, it may be expected that similar progress
would be seen in the next and subsequent episodes. In this
summary of the study by Ochs and colleagues, 6 the validity
of the episode analysis is weakened by 2 factors: (1) the potential
for lack of independence between episodes and (2)
the potential for error introduced when measuring relatively
small changes in wound size between 2 relatively close
points in time (7–10 days apart), especially given the inexact
measure of wound surface area.
In analyzing the data generated in their study, the
researchers acknowledged that the 3 study groups lacked
baseline comparability. 6 For example, residents allocated to
group 3 surfaces were younger (mean age 67.9 years) than
residents assigned to group 1 and group 2 surfaces (group 1,
mean age 79.3 years; group 2, mean age 77.4 years). In
addition, residents who had been placed on group 3 support
surfaces typically had more pressure ulcers, and their
ulcers were of larger size and greater severity as compared to
residents placed on group 1 and group 2 support surfaces.
This variability in baseline group characteristics is understandable,
since group 3 surfaces (air-fluidized beds) are typically
reserved for patients with multiple and/or severe
pressure ulcers. However, the lack of comparability among
the study groups reflects a frequently encountered challenge
of using retrospective data when allocation to treatment
groups is based on prevailing treatment and
reimbursement patterns rather than random assignment.
For example, given the more severe nature of the pressure
ulcers among residents managed with group 3 support
surfaces, it was not surprising that their pressure ulcers appeared
to heal faster. Twelve years ago the US National
Pressure Ulcer Advisory Panel 3 noted that initial pressure
ulcer size is a predictor of healing rate, with larger ulcers
demonstrating a more dramatic reduction in dimensions.
The regression analyses conducted by Ochs and colleagues 6
reaffirmed this observation; they found that initial pressure
ulcer size was the key predictor of healing rate, and commented
that the “marked effect of initial pressure ulcer size
masked other factors with a potential impact on the healing
rate.” While retrospective cohort studies have the potential
to identify important trends in pressure ulcer management,
the inability to form comparable groups at baseline weakens
their ability to compare the efficacy of specific interventions,
such as the ability of a specific pressure redistribution surface
to promote pressure ulcer healing. This was the case of Ochs
and colleagues, 6 where the initial allocation of support surfaces
based on clinical considerations alone resulted in the
largest pressure ulcers being allocated to one treatment arm
(group 3). This inevitably led to the observation that their
wounds healed faster based on ulcer size rather than differences
in the efficacy of the pressure redistribution surface.
Ochs and colleagues 6 correctly concluded that this challenge
is ideally resolved by conducting a prospective study that
matches wound sizes among groups at the onset of the
study. So where healing rates are to be calculated and compared,
retrospective studies may fail to identify clinically relevant
differences between interventions.
Another limitation of retrospective cohort studies is
their inability to assign each treatment option frequently
enough to ensure the ability to perform a meaningful
comparison to other treatments. Instead, specific interventions
may occur only rarely in available data sets. For
example, while multiple subjects in the study reported by
Ochs and colleagues 6 were managed with group 1 and
group 2 support surfaces, very few were managed with alternating
pressure overlay support surfaces (Table 1).
Sparse employment of a key intervention, such as an
alternating pressure overlay support surface, precludes any
meaningful analysis of its efficacy.

WJ3504_391-394.qxp 6/23/08 10:13 PM Page 393
J WOCN ■ Volume 35/Number 4 Clark 393
TABLE 1.
Number of Nursing Home Residents Allocated to
Different Group 1 and 2 Support Surfaces a
Support Surface Group 1 Group 2
Foam 350 0
Water/gel 83 0
Alternating pressure overlays 16 0
Low air loss 0 62
Powered pressure reducing 0 35
Powered air overlay 0 12
Nonpowered advanced 0 16
pressure surfaces
Unreported 14 0
Total number 463 125
a
The total number of residents allocated to group 2 surfaces (n 125)
exceeds the total number of residents noted by Ochs and colleagues 6 to
have received group 2 surfaces (n 119).
■ Prospective Cohort Studies
In contrast to retrospective cohort studies, prospective
cohort studies frequently require a considerable commitment
of time and a robust budget. One example of a
prospective cohort study that illustrates the challenges
and benefits of collecting outcome data prospectively was
reported by Clark and associates. 4,9 Their study initially
involved 4 UK hospitals and 1 US hospital. It required 10
full-time data collectors (in the United Kingdom) to gather
data from a cohort of adult patients (16 years old) who
stayed in hospital for more than 2 days, were able to provide
consent (or assent was available from relatives), and were
not cared for on psychiatry, ophthalmology, gynecology,
pediatrics, obstetrics, or psychiatric wards. Data were gathered
over a 2-year period from 2,507 UK hospital patients,
representing 29,611 total patient-days. A further 1,202
subjects were recruited at the US site; but differences between
the countries in the availability of some data precluded
combination of the 2 data sets. Specifically, data at
the US hospital were gleaned from prospective review of
medical and nursing records rather than direct patient
observation. The primary focus of this prospective study
was the effectiveness of support surfaces on pressure ulcer
prevention, although data were also collected upon
subjects who entered the cohort with existing pressure
damage. A wide range of support surfaces was used in the
study including low-air-loss, air-fluidized, foam, gel, and
alternating pressure-redistributing mattresses and overlays.
The study cohort involved 218 subjects with pressure
ulcers; 100 were present on admission and the remaining
118 developed in the hospital, representing an incidence
rate of 42.6 people developing pressure ulcers per 10,000
patient-days.
Prospective data collection allowed for the inclusion of
variables relevant to pressure ulcer development that may
not be captured in a retrospective review, which is limited
by the accuracy and comprehensiveness of previously
recorded data. For example, Clark and associates 4,9 assessed
the condition of the bed mattress and found that 15%
(n 382) of the patients on foam mattresses received inadequate
support (bottomed out) and 15 of the alternating
pressure mattresses and overlays exhibited alarm signals.
This level of detail is unlikely to be captured from retrospective
data sets. Rather, it is only likely to be measured when
full-time data collectors are available at each participating
center. Interestingly, the condition of the mattress was not
associated with a higher incidence of pressure ulcers.
While prospective cohort studies offer advantages over
retrospective studies, one weakness remains: the inability
to form treatment groups that are truly comparable at
baseline. If the study groups are not comparable, the data
gathered may fail to reflect key factors impacting study
outcomes, including the treatment options under evaluation.
This remains a principal weakness of both retrospective
and prospective cohort studies, for without true
randomization at baseline it remains unlikely that
differences in the comparison groups can be effectively
controlled for, allowing selective evaluation of the treatment
options under evaluation.
TABLE 2.
Comparison of the Benefits and Disadvantages of Prospective and Retrospective Cohort Studies Compared With
Randomized Controlled Trials
Randomized Controlled Trial Retrospective Cohort Study Prospective Cohort Study
Baseline comparability Lack of baseline comparability Lack of baseline comparability
Narrow, defined population Wide population, potentially better Wide population, potentially better reflecting
reflecting “real-world” care
“real-world” care
High internal validity Perhaps, low internal validity Perhaps, low internal validity
Expensive Less costly Expensive
Ability to determine data items to be Lack of ability to capture key variables Ability to determine data items to be captured
captured
missing from the data set
Effect can be attributed to intervention May not be able to attribute results May not be able to attribute results to the
to the intervention
intervention

WJ3504_391-394.qxp 6/23/08 10:13 PM Page 394
394 Clark J WOCN ■ July/August 2008
■ Conclusion
Prospective and retrospective cohort studies appear to offer
valid alternatives to the RCT when exploring the effect of
interventions used in pressure ulcer prevention or treatment.
However, the relative strengths and weaknesses of the
2 designs must be borne in mind; retrospective cohorts may
be generated at relatively low cost but may miss variables of
interest and fail to adequately reflect specific interventions
within the cohort. In contrast, prospective cohorts are able
to include specific variables and data can be collected with
greater confidence in its reliability and accuracy, but require
considerable investment of time and money and are therefore
more difficult to perform (Table 2). In addition, both
designs are limited by the difficulty of establishing truly
comparable baseline groups. This challenge of baseline comparability
suggests that there remains a strong place for wellconducted
RCTs when exploring the impact of new
interventions for pressure ulcer prevention of healing.
■ ACKNOWLEDGMENT
The author has no significant ties, financial or otherwise,
to any company that might have an interest in the publication
of this educational activity.
KEY POINTS
✔ Retrospective studies are relatively low cost but frequently
fail to report on specific interventions.
✔ Prospective designed studies require significant financial
input but collect more valid and reliable data.
✔ Initial wound size is a strong predictor of healing.
■ References
1. UK National Institute for Health and Clinical Excellence.
Pressure ulcers: the management of pressure ulcers in primary
and secondary care. Clinical guideline CG029. http://www.
nice.org.uk/page.aspxo/CG029&c/skin. Published 2005.
Accessed November 1, 2006.
2. Brown GS. Reporting outcomes for stage IV pressure ulcer healing:
a proposal. Adv Skin Wound Care. 2000;13(6):277–283.
3. Xakellis GC, Maklebust JA. Template for pressure ulcer research.
Adv Wound Care. 1995;8(1):46–48.
4. Clark M, Benbow M, Butcher M, et al. Collecting pressure ulcer
prevention and management outcomes; part 1. Br J Nurs.
2002;11(4):230–238.
5. Clark M, Benbow M, Butcher M, et al. Collecting pressure ulcer
prevention and management outcomes: 2. Br J Nurs. 2002;
11(5):310–314.
6. Ochs RF, Horn SD, van Rijswijk L, Pietsch C, Smout RJ.
Comparison of air-fluidized therapy with other support surfaces
used to treat pressure ulcers in nursing home residents.
Ostomy Wound Manage. 2005;51(2):38–68.
7. Eager CA. Monitoring wound healing in the home health
arena. Adv Wound Care. 1997;10(5):54–57.
8. Schubert V. Measuring the area of chronic ulcers for consistent
documentation in clinical practice. Wounds. 1997;9(5):
153–159.
9. Clark M. Models of pressure ulcer care: costs and outcomes.
Br J Healthc Manage. 2001;7(10):412–416.
Call for Authors: Wound Care
• Review articles, case studies, case series, and original research reports focusing on the potential role of unprocessed
honey in wound healing
• Review articles or original research reports focusing on the antibacterial properties of silver
• Continuous Quality Improvement projects, research reports, or institutional case studies focusing on innovative
approaches to reduce facility-acquired pressure ulcers
• Case studies, case series, review articles, and original research reports focusing on topical therapies for pressure
ulcers, vascular ulcers, or neuropathic (diabetic foot) ulcers
• Original research reports focusing on the histologic and clinical effects of negative pressure wound therapy