In response to our article there have been a number of valid
questions raised:
(1) Joel Hay questions whether excluding patients with cancer or a
neurological diagnosis would change the results. We have re-analysed our
primary outcome either removing these patients or adding in a co-variate
adjusting for the presence or absence of these diagnoses. Neither
analysis altered the rate ratio found.
(2) Joel Hay also questions whether it was appropriate to recruit
from hospital. This source was chosen as it identified a group who on
discharge, are likely to have drug changes. A number of UK studies have
highlighted this finding. We therefore considered this group to be
vulnerable to medication errors (duplication, unclear changes to
medication etc.) and therefore potential beneficiaries of our
intervention.
(3) We agree with David Leopold that a reduction of mortality rate by
one quarter over six months would have been a spectacular result. Our
study was not designed to demonstrate such a change. We specifically
powered our study on the basis of hospital admissions data, not mortality.
A further study would need to recruit over 2,540 patients in order to have
80% power to investigate a potential 25% reduction in mortality.
(4) Michael Scott suggests that we too readily dismiss the
possibility that our intervention improved adherence. This was not our
intention. We have no evidence as to whether we improved adherence, and
we simply commented that whilst some investigators have demonstrated
improved adherence through medication review, others have not. We agree
that one of our aims was to improve adherence.
(5) Michael Scott also questions whether a meaningful medication
review can be conducted without access to a medical record. We agree that
access to such a record is ideal. Nonetheless, gaining access to such
records outside hospital, or general practice is difficult. We set out to
ask a pragmatic health service question based around a service that had
already been implemented in pilot form by Norfolk Social Services. In
hindsight, it may have been enhanced by access to clinical records though
this would have been difficult to negotiate. Importantly, within the new
pharmacy contract, whilst it will pay community pharmacists to undertake
medication reviews, it is unlikely that there will be access to clinical
notes until IT systems are considerably more advanced than is currently
the case. Such pharmacy-based reviews will have access to far less
information than our HOMER pharmacists as they will not be able to witness
a patient’s social circumstances, or check on drug storage/hoarding.
(6) Richard Davies may be justified connecting the increased hospital
admissions with the increase in GP home visiting. Further analysis has
demonstrated that amongst those admitted to hospital the rate of GP home
visits was almost double that of patients not admitted (rate ratio 1.98,
95% C.I. 1.56 to 2.50, p<0.001), although the direction of this
association is uncertain.
(7) Abdullah Mohammed suggests that our study design should have gone
beyond answering a single question. Again space limited what we could
present. We have investigated cause of admission in both groups. Only
two differences were found: there were eight readmissions (3.4%) amongst
intervention patients with endocrine abnormalities (these involved fluid
and electrolyte abnormalities, or hypoglycaemia). This compared to only
one equivalent control admission (0.6%), [Fisher’s exact test = 0.05]. In
contrast, there was an excess of surgical complications in the control
group (3.9% vs. 0.9%, p=0.02). However, it should be stressed that
testing for these differences involved multiple statistical testing (21
different causes were investigated). Since neither of these differences
were predicted, or pre-specified in advance these results should be
treated with caution. Overall, data on cause of admission do not shed
useful light on our main finding.
(8) David Green, Michael Scott and Duncan Petty would have liked
further information about our intervention. We regret that in the space
available we could only cursorily describe this. We plan to publish more
descriptive detail in due course.
(9) In response to David Green, we feel that any medication review
programme, including that within the new pharmacy contract, should either
be based on existing evidence, or should be the subject of further
evaluation. Our study we believe, serves as a useful reminder that
plausible interventions do not always have intended effects.
Competing interests:
None declared
Competing interests:
No competing interests
07 March 2005
Richard Holland
Senior Lecturer in Public Health Medicine
On behalf ot the HOMER trial investigators
School of Medicine, Health Policy & Practice, University of East Anglia, Norwich, NR4 7TJ
Rapid Response:
Authors' response to comments on the HOMER trial
To the Editor
In response to our article there have been a number of valid
questions raised:
(1) Joel Hay questions whether excluding patients with cancer or a
neurological diagnosis would change the results. We have re-analysed our
primary outcome either removing these patients or adding in a co-variate
adjusting for the presence or absence of these diagnoses. Neither
analysis altered the rate ratio found.
(2) Joel Hay also questions whether it was appropriate to recruit
from hospital. This source was chosen as it identified a group who on
discharge, are likely to have drug changes. A number of UK studies have
highlighted this finding. We therefore considered this group to be
vulnerable to medication errors (duplication, unclear changes to
medication etc.) and therefore potential beneficiaries of our
intervention.
(3) We agree with David Leopold that a reduction of mortality rate by
one quarter over six months would have been a spectacular result. Our
study was not designed to demonstrate such a change. We specifically
powered our study on the basis of hospital admissions data, not mortality.
A further study would need to recruit over 2,540 patients in order to have
80% power to investigate a potential 25% reduction in mortality.
(4) Michael Scott suggests that we too readily dismiss the
possibility that our intervention improved adherence. This was not our
intention. We have no evidence as to whether we improved adherence, and
we simply commented that whilst some investigators have demonstrated
improved adherence through medication review, others have not. We agree
that one of our aims was to improve adherence.
(5) Michael Scott also questions whether a meaningful medication
review can be conducted without access to a medical record. We agree that
access to such a record is ideal. Nonetheless, gaining access to such
records outside hospital, or general practice is difficult. We set out to
ask a pragmatic health service question based around a service that had
already been implemented in pilot form by Norfolk Social Services. In
hindsight, it may have been enhanced by access to clinical records though
this would have been difficult to negotiate. Importantly, within the new
pharmacy contract, whilst it will pay community pharmacists to undertake
medication reviews, it is unlikely that there will be access to clinical
notes until IT systems are considerably more advanced than is currently
the case. Such pharmacy-based reviews will have access to far less
information than our HOMER pharmacists as they will not be able to witness
a patient’s social circumstances, or check on drug storage/hoarding.
(6) Richard Davies may be justified connecting the increased hospital
admissions with the increase in GP home visiting. Further analysis has
demonstrated that amongst those admitted to hospital the rate of GP home
visits was almost double that of patients not admitted (rate ratio 1.98,
95% C.I. 1.56 to 2.50, p<0.001), although the direction of this
association is uncertain.
(7) Abdullah Mohammed suggests that our study design should have gone
beyond answering a single question. Again space limited what we could
present. We have investigated cause of admission in both groups. Only
two differences were found: there were eight readmissions (3.4%) amongst
intervention patients with endocrine abnormalities (these involved fluid
and electrolyte abnormalities, or hypoglycaemia). This compared to only
one equivalent control admission (0.6%), [Fisher’s exact test = 0.05]. In
contrast, there was an excess of surgical complications in the control
group (3.9% vs. 0.9%, p=0.02). However, it should be stressed that
testing for these differences involved multiple statistical testing (21
different causes were investigated). Since neither of these differences
were predicted, or pre-specified in advance these results should be
treated with caution. Overall, data on cause of admission do not shed
useful light on our main finding.
(8) David Green, Michael Scott and Duncan Petty would have liked
further information about our intervention. We regret that in the space
available we could only cursorily describe this. We plan to publish more
descriptive detail in due course.
(9) In response to David Green, we feel that any medication review
programme, including that within the new pharmacy contract, should either
be based on existing evidence, or should be the subject of further
evaluation. Our study we believe, serves as a useful reminder that
plausible interventions do not always have intended effects.
Competing interests:
None declared
Competing interests: No competing interests