Lauren Silver HPM 582 Brief Writing Assignment #2; 2-23-2009

Proposed Research Design. Ideally, a randomized prettest-posttest control group design would produce
the most valid results, however, ethical and budgetary issues will preclude using this design (i.e.,
Stoppadrop should not be withheld from patients who otherwise could receive it and such a design would
be cost-prohibitive). Conseuently, since !CC" prefers a prospective study, a nonrandomized prettest-
posttest noneuivalent comparison group design is proposed and li#ely will yield more valid results than
alternative uasi-e$perimental designs (%igure &).
Site selection. 'he comparison and intervention groups should include both rural and urban patients
to control for patient differences attributable to location and related demographic factors. 'hus, outcomes
among the rural intervention and comparison groups will be compared and li#ewise for the urban groups.
(iven the recent arrival of Stoppadrop onto the mar#et, sites also will be chosen based on whether
Stoppadrop is offered to breast cancer patients on ad)uvant chemo following surgery. *dditionally,
prettest and posttest data collection should ta#e place at the same rural and urban sites in order to control
for+to the e$tent possible+physician-related factors.
,atient recruitment. Comparison group patients will be recruited from among those who are eligible
for Stoppadrop at the sites not offering Stoppadrop, and intervention group patients will be selected from
among those who are eligible at the sites offering Stoppadrop. 'his recruitment method will help ensure
that ethical guidelines are met since Stoppadrop will not be withheld from patients intentionally as part of
the study design. -nly patients whose insurance covers Stoppadrop will be selected. 'hose who are
uninsured but are willing to pay the high cost for the drug may introduce selection bias with respect to
their motivation level to complete chemo.
-utcome measures. %or the comparison groups, the proposed outcome measure is the .total / of
patients on ad)uvant chemo only who drop out prior to completing 0 cycles, but could be clinically
recommended for Stoppadrop1total / of patients who begin ad)uvant chemo only but could be clinically
recommended for Stoppadrop2. %or the intervention groups, the proposed outcome measure is the .total /
of patients on ad)uvant chemo with Stoppadrop who drop out prior to completing 0 cycles1total / patients
who begin ad)uvant chemo with Stoppadrop2. 'he analysis would compare the prettest-posttest changes in
drop-out rates for the intervention and comparison groups. If Stoppadrop reduces the probability that
patients will discontinue chemo, we would e$pect to see a larger, statistically significant reduction in the
Lauren Silver HPM 582 Brief Writing Assignment #2; 2-23-2009
drop-out rate for the intervention groups and no change (or very little that is not statistically significant) in
the drop-out rate for the comparison groups.
Threats to Internal Validity. If properly implemented, this particular design is effective at controlling for
a number of threats to internal validity. %irst, in terms of history threats, it is unli#ely that patients from
the same location would e$perience different e$ternal events that also could e$plain variation in
outcomes. Second, if the time interval between the start and end of the study is #ept to a minimum,
maturation threats should not pose a problem. 'hird, the more similar the intervention and comparison
groups are, the less li#ely outcome differences will be due to regression rather than to the effect of
Stoppadrop. %inally, to the e$tent that attrition occurs, statistical analyses can verify whether and how
those who drop out of the study differ from those who do not. Selection bias poses the ma)or challenge to
internal validity.
Selection bias. 3ealth system factors may be controlled to the e$tent that all patients receive care at
the same type of system, such as all fee-for-service practices or all managed care organizations. 3owever,
there are potential sources of patient- and physician-related selection bias that should be addressed. Is
there a difference between physicians and1or sites that offer and do not offer Stoppadrop (e.g., physicians
offering Stoppadrop may be more li#ely to coach and encourage their patients to complete the entire si$-
wee# cycle of chemo)4 *re patients receiving care at the sites offering Stoppadrop different from patients
see#ing care at non-Stoppadrop sites (e.g., do some patients choose to receive care at sites offering
Stoppadrop with the intention to increase their li#elihood of completing chemo4)4 If so, then these
patients will differ from patients in the comparison group based on their motivation level to complete
chemo. 'he above e$amples illustrate how selection bias may ma#e it difficult to attribute reduced drop-
out rates to Stoppadrop as opposed to underlying differences in the intervention and comparison groups.
Complications, Questions, Uncertainties. *lthough Stoppadrop has been on the mar#et for only si$
months, it is possible that only a small number of sites are not yet offering the new drug, which would
pose problems obtaining a sufficient sample size for comparison groups. It may not be possible to collect
real-time baseline data on patients at sites where Stoppadrop is offered, so what alternative sources of
data on drop-out rates might be used4 'he very #nowledge of being followed while on chemo as part of
this study may influence patients5 motivation to complete their chemo and, thus, potentially bias results.
Lauren Silver HPM 582 Brief Writing Assignment #2; 2-23-2009
Interventio
n Group
Compariso
n Group
Urban
Sites
Urban
Sites
Rural Sites
Rural Sites
Prettest Prettest
Stoppadrop
Baseline drop-out rate
Posttest Posttest Posttest drop-out rate
Compare prettest-posttest change in drop-out rates
Figure 1.
(rural-to-rural and urban-to-urban comparisons)