Udvalget for Landdistrikter og Øer 2019-20
ULØ Alm.del
Offentligt
2125261_0001.png
RESEARCH/Original Article
Journal of Telemedicine and Telecare
2015, Vol. 21(7) 377–384
!
The Author(s) 2015
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1357633X15572202
jtt.sagepub.com
A cohort study following up on
a randomised controlled trial of a
telemedicine application in COPD patients
Anne-Kirstine Dyrvig
1
, Oke Gerke
2
,
Kristian Kidholm
1,3
and Hindrik Vondeling
2,4,5
Abstract
Introduction:
The studies that constitute the knowledge base of evidence based medicine represent only 5%–50% of patients
seen in routine clinical practice. Therefore, whether the available evidence applies to the implementation of a particular service
often remains unclear. Chronic obstructive pulmonary disease (COPD) is no exception.
Methods:
In this article, the effects of implementing a telemedicine intervention for COPD patients were analysed using data
collected before, during, and after a randomised controlled trial (RCT).
More specifically, regression techniques using robust variance estimators were used to analyse whether the use of telemedicine,
patient age, and gender could explain the risk of readmission, length of hospital admission, and death during a five-year
observation period.
Results:
Increased risk of readmission was significantly related to both use of telemedicine and increased age in three sub-
periods of the study, whereas women showed a more pronounced risk of readmission than men only during and after the RCT
period. The number of days admitted to hospital was higher for patients using telemedicine and being of older age. Risk of death
during the observation period was decreased for patients using telemedicine and for female patients and increased for elderly
patients. No interaction between intervention and time period was observed.
Statistically significant relationships were identified between use of telemedicine and risk of readmission, days admitted to
hospital, and death.
Discussion:
Research on effect modification in telemedicine is essential in designing future implementation of interventions as
it cannot be taken for granted that effectiveness follows from efficacy.
Keywords
External validity, telemedicine, evidence based practice, efficacy, effectiveness
Date received: 1 June 2015; accepted: 30 June 2015
Background
In principle, evidence based medicine should be provided
to all patients. However, most guidelines on translating
evidence into practice are largely based on randomised
controlled trials (RCTs) that may include only between
5 and 50% of patients seen in routine practice.
1–3
In other words, due to strict inclusion criteria, the
external validity of many trials may be low. Formally,
external validity has been defined as ‘whether the results
[of randomised controlled trials] can be reasonably
applied to a definable group of patients in a particular
clinical setting in routine practice’.
4
This definition high-
lights the potential efficacy of an intervention (therapeutic
benefit under ideal circumstances) versus effectiveness
(therapeutic benefit under everyday life circumstances
5
).
Few studies have been carried out to investigate the
differences between outcomes in patients in trials com-
pared to real-life, non-enrolled patients.
2,6–10
However, it
is acknowledged that efficacy studies are more likely to
obtain favourable results than effectiveness studies, and
that the difference can be attributed to contextual
1
Centre for Innovative Medical Technologies, Odense University Hospital
and University of Southern Denmark
2
Centre of Health Economics Research (COHERE), Department of Nuclear
Medicine, Odense University Hospital, University of Southern Denmark
3
Department for Quality, Research and HTA, Odense University Hospital
4
Center for applied services research and technology assessment, University
of Southern Denmark
5
Department of Health, Technology and Services Research (HTSR),
University of Twente, Enschede, the Netherlands
Corresponding author:
Anne-Kirstine Dyrvig, MScPH, Centre for Innovative Medical Technologies,
Odense University Hospital and University of Southern Denmark, Sdr.
Boulevard 29, entrance 216 st. th., DK-5000 Odense, Denmark.
Email: [email protected]
1
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0002.png
378
circumstances,
5
for example socio-demographic status and
gender.
A systematic literature review on external validity, which
was conducted for this article, resulted in 12 articles (details
on the search and selection strategy are provided in the
supplementary material). These articles presented empirical
data on external validity without restrictions on any par-
ticular clinical field. Eight of the articles presented data on
the proportion of patients seen in routine practice who were
considered eligible for randomised trials. In the field of
asthma and
chronic obstructive pulmonary disease
(COPD), 4%–6% of patients were included;
1,11
of patients
with myocardial infarction, 50% were included;
12
and with
stroke/ischaemic attack, 25%–67% were included.
7
Among
patients reported as eligible for medication for lowering
blood pressure, 60% were actually included,
13
50% were
included in a nicotine dependency trial,
3
and less than 20%
of enrolees were part of a trial on anxiety.
14
Finally, in the
only article describing the external validity of a telemedi-
cine trial, Riper et al.
9
found that their intervention on
alcohol dependency was externally valid, with no difference
in the results of their trial compared to the real-life
effectiveness.
With these differences in test results versus implemen-
tation in practice, it is clear that further understanding of
the modifications of effects that cause implementation
response to differ from that of RCTs will improve real-
life interventions.
15
In Denmark, hospitals are obliged to develop internal
but publicly available guidelines for treatment.
16
From
local guidelines it appears that a telemedicine intervention
is recommended for patients admitted with COPD.
17
(The
telemedicine intervention is described in detail in
Rasmussen et al.
17
and Sorknaes et al.
18
It consisted of
real-time video-consultations with hospital nurses for
approximately 30 mins each day during the first week
after discharge. The consultation included measurements
of blood oxygen level and airflow.) In the guideline, it is
mentioned that an RCT has been conducted, that patients
are positive towards the technology, and that no statistic-
ally significant differences were identified for the risk of
readmission or risk of death.
17
The guideline does not
recommend restrictions on the use of telemedicine based
on the RCT criteria for inclusion.
18
Thus, in the periods
before and after the RCT there were no descriptions of the
characteristics of patients to consider for the telemedicine
intervention, and the decision was then left to clinical
judgement. The impact of this potential difference in selec-
tion criteria on the overall effectiveness of the intervention
constituted the main focus of this cohort study.
In the present study, the effectiveness of the telemedi-
cine intervention provided to patients admitted to hospital
due to exacerbation of their COPD was measured during
three different periods of time. The effectiveness was mea-
sured in terms of the risk of readmission, number of days
spent in hospital, and the risk of dying from any cause.
Risk of readmission and risk of death are commonly used
as outcome measures throughout the literature on
Journal of Telemedicine and Telecare 21(7)
COPD.
19
The number of days spent in hospital was used
as it was hypothesised that the intervention could lead to
earlier discharge. This hypothesis could not be tested in
the RCT because of ethical considerations,
18
so it was
reported within this study instead. The aim of the study
was to investigate the effectiveness of a telemedicine inter-
vention in COPD patients during, before, and after an
RCT while adjusting for age and gender.
Methods
The study was a cohort study covering the time span from
1 January 2009 to 31 December 2013, comprising three
periods:
.
Pre-RCT: observations registered before introduction
of the RCT (1 January 2009–31 April 2010).
.
During RCT: observations registered during the period
of the RCT (1 May 2010–31 October 2011).
.
Post-RCT: the group of observations registered after
the RCT ended (1 November 2011–31 December 2013).
Included patients were admitted to hospital because of
an acute exacerbation of their COPD. COPD was diag-
nosed according to the GOLD guidelines (global initiative
for chronic obstructive lung disease).
20
In the guideline,
COPD is defined as:
Chronic Obstructive Pulmonary Disease (COPD), a
common preventable and treatable disease, is character-
ized by persistent airflow limitation that is usually progres-
sive and associated with an enhanced chronic
inflammatory response in the airways and the lung to nox-
ious particles or gases. Exacerbations and comorbidities
contribute to the overall severity in individual patients.
Thus, the condition is complex and multifactorial, and
the diagnostic guidelines are revised on a five-year basis.
21
While airflow as measured by spirometry is a prerequisite
for the diagnostic procedure, it is further qualified by the
assessed risk and number of symptoms.
20,21
Population and assignment to intervention
The population studied was composed of all patients
admitted to Odense University Hospital (OUH),
Denmark, during the years 2009 to 2013 with acute
exacerbation of COPD registered as primary diagnosis
on the basis of ICD-10-codes (international classification
of diseases).
22,23
Patients were classified within each period
into those who received telemedicine intervention subse-
quent to their admission to hospital (cases) and those who
did not (controls), where the telemedicine intervention was
registered in the hospital administrative system. During
the RCT phase, a group of patients selected on the basis
of criteria for inclusion was asked to participate in the
study. Half of the patients included were randomly
assigned to the intervention versus control group.
2
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0003.png
Dyrvig et al.
In comparison, in the periods before and after the RCT,
the patients were selected for intervention on the basis of
clinical judgement (incorporating the knowledge gener-
ated from the RCT regarding the latter). So there were,
theoretically, different ways of selecting patients for tele-
medicine intervention in the different periods of time. The
control group included all patients not receiving the inter-
vention throughout the study period and was, therefore,
not restricted to selected patients.
379
no) was analysed by means of logistic regression and days
of admittance to hospital by negative binomial regression
(both using robust variance estimators when appropriate).
In case of interaction between telemedicine intervention
and time period, stratified analyses by time periods were
done. Survival analyses were performed using Cox pro-
portional hazards regression, which took into account
the difference in time (measured in days) that each indi-
vidual contributed to the analyses. The time measurement
was started when a person entered the cohort by admit-
tance to hospital and ended at discharge or when the
person died (all-cause mortality). In a sensitivity analysis,
missing inclusion dates, due to outpatient visits, were
imputed by taking the start date of the respective period
as replacement. This was done in 1438, 1218, and 1293
cases in the respective three periods (pre-RCT, during
RCT and post-RCT).
Explanatory variables in all regression models were
telemedicine intervention (yes/no), age, gender, and
period (pre-RCT, during RCT and post-RCT).
Significance level was 5% (two-sided testing). All ana-
lyses were carried out using STATA/IC 13 (StataCorp Lp,
College Station, Texas 77845 USA).
Outcomes
Outcomes were:
.
Readmissions (defined as admission to hospital subse-
quent to a previous admission within a period of 42
days);
.
Number of days being admitted to hospital;
.
Death during period of observation.
Data were extracted from a hospital administrative
system, so that selection of variables was based on (a)
availability of data, (b) data previously used in telemedi-
cine studies on COPD,
19
and (c) outcomes of the previous
efficacy study (RCT).
18
These also comprised the demo-
graphic covariates age and gender.
Ethics
This project was approved by the Danish Data Protection
Agency and needed no approval from the ethics
committee.
Data management
The dataset was composed of observations representing
either an admission to hospital or an ambulatory visit.
Ineligible observations were excluded prior to the statis-
tical analysis as follows. Observations in patients under
the age of 30 years were excluded from the dataset
(N
¼
51) because COPD usually develops over a period of
20 years, thus indicating an age older than 30 is necessary
for patients to suffer from COPD. In addition, a number of
observations had registrations of admission and discharge
from hospital, but lacked registrations such as age or treat-
ment codes (N
¼
26) and, therefore, had to be excluded.
One observation was deleted due to an inconsistency
(the patient had been registered as admitted to hospital
later than her date of death).
Finally, 17 observations were excluded from the data
set due to abnormally long hospital admissions (duration
over 90 days) because, according to the professor of the
department, admissions longer than this were not caused
by the COPD diagnosis.
24
Results
The patient administrative system provided data on
11,303 patients admitted to hospital during the five-year
period covering 2009 to 2013. While one observation rep-
resented either an admission to hospital or an outpatient
visit, each person could account for several observations.
While 8257 (73.05%) patients contributed only to one
period, the remaining patients contributed to two periods
(N
¼
2130 (18.85%)) or to all three periods (N
¼
916
(8.10%)), see Table 1. In total, the patients contributed
Table 1.
Distribution of how 11,303 patients contributed to either
single time periods alone or to multiple time periods (N¼15,265
occurrences).
During
Pre-RCT RCT
Post-RCT Frequency Percentage Cumulative
X
X
X
X
X
X
X
X
X
X
X
X
3403
1628
3226
839
533
758
916
30.11
14.40
28.54
7.42
4.72
6.71
8.10
30.11
44.51
73.05
80.47
85.19
91.90
100.00
Statistical analyses
Demographics were described for all groups by means and
standard deviations for continuous variables (or medians
and ranges in case of skewed data) and proportions with
respective percentages for categorical variables.
Differences on demographic aspects between intervention
and control groups were assessed by Student’s
t-test
and
by the

2
test where appropriate. Risk of readmission (yes/
RCT: randomised controlled trial.
3
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0004.png
380
Journal of Telemedicine and Telecare 21(7)
Table 2.
Demographic and clinical characteristics of patients.
Telemedicine
mean (SD)
72.85
71.99
71.80
N
113
127
394
N
41
35
153
N
102
103
406
Mean
1.51
1.54
1.79
Mean
16.67
11.87
12.97
Mean
4.43
3.42
3.11
(9.31)
(9.60)
(9.83)
(%)
(54.85)
(60.48)
(57.18)
(%)
(19.90)
(16.67)
(22.21)
(%)
(49.51)
(49.05)
(58.93)
(SD)
(2.58)
(2.69)
(3.09)
(SD)
(17.75)
(13.69)
(16.01)
(SD)
(4.23)
(2.75)
(2.89)
Control
mean (SD)
71.68
71.97
72.87
N
2804
2027
2425
N
1463
823
1278
N
1281
868
943
Mean
0.51
0.45
0.34
Mean
8.01
6.45
6.82
Mean
4.14
3.58
4.02
(11.72)
(11.75)
(12.16)
(%)
(51.12)
(51.56)
(51.12)
(%)
(26.67)
(20.94)
(26.94)
(%)
(23.35)
(22.08)
(19.88)
(SD)
(1.75)
(1.83)
(0.96)
(SD)
(13.38)
(11.71)
(11.80)
(SD)
(6.72)
(6.40)
(6.59)
Number of observations
p-value
0.16
0.98
0.03
0.29
0.01
0.003
0.03
0.14
0.008
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.53
0.72
0.0004
Telemedicine
206
210
689
206
210
689
206
210
689
206
210
689
206
210
689
206
210
689
206
210
689
Control
5485
3931
4744
5485
3931
4744
5485
3931
4744
5485
3931
4744
5485
3931
4744
5485
3931
4744
5485
3931
4744
Age
Pre-RCT
During RCT
Post-RCT
Gender (f)
Pre-RCT
During RCT
Post-RCT
Fatalities
Pre-RCT
During RCT
Post-RCT
Readmission
Pre-RCT
During RCT
Post-RCT
Number of readmissions
Pre-RCT
During RCT
Post-RCT
Days admitted
Pre-RCT
During RCT
Post-RCT
Average days admitted
Pre-RCT
During RCT
Post-RCT
Note: Individual patients could be represented only once within each period, but were allowed to contribute to different periods over time.
RCT: randomised controlled trial
with 89,050 observations in terms of hospital admissions
and outpatient visits.
Table 2 summarises the distribution of demographic
variables (age and gender) and outcome variables (fatal-
ities, readmission, number of readmissions, days admitted,
and average number of days admitted) separately for
patients receiving the telemedicine intervention and for
those who did not. Data are presented for each of the
three periods of time (pre-RCT, during RCT and post-
RCT).
Age differed between the intervention and the control
group only in the last period of measurement (post-RCT,
71.80 years vs. 72.87 years,
p
¼
0.03). There were signifi-
cantly more females in the intervention group in the
during RCT period (60.48% vs. 51.56%,
p
¼
0.01) and
in the post-RCT period (57.18% vs. 51.12%,
p
¼
0.003).
Smaller proportions of fatalities were observed in the
telemedicine group throughout all periods of time, but
to a statistically significant degree only in the pre-RCT
phase (19.90% vs. 26.67%,
p
¼
0.03) and the post-RCT
phase (22.21% vs. 26.94%,
p
¼
0.008). Every second
patient in the intervention group was readmitted to hos-
pital, whereas this was the case for only every fourth to
fifth patient of the control group (p
<
0.0001 in all three
periods). Patients of the intervention group were, on
average, 1–1.45 times more often readmitted to hospital
than patients of the control group (p
<
0.0001 in all three
periods). Moreover, patients receiving telemedicine were
admitted to hospital longer than those who did not
receive telemedicine (between 5.42 and 8.66 days,
p
<
0.0001 in all three periods), but the average stay at
hospital was significantly shorter in patients using tele-
medicine in the post-RCT period (3.11 vs. 4.02 days,
p
¼
0.0004).
Readmission
In the pre-RCT group, the odds ratio (OR) for a readmis-
sion in patients with telemedicine intervention was 3.18
(95% CI 2.40–4.22,
p
<
0.0001; see Table 3). During the
4
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0005.png
Dyrvig et al.
381
Table 3.
Odds ratio of readmission for patients.
Telemedicine
Period
Pre-RCT
During RCT
Post-RCT
OR
3.18
3.44
6.04
95% CI
2.40–4.22
2.59–4.57
5.09–7.16
p-value
<0.0001
<0.0001
<0.0001
Age
OR
1.023
1.03
1.022
95% CI
1.018–1.029
1.02–1.04
1.017–1.028
p-value
<0.0001
<0.0001
<0.0001
Gender (female)
OR
1.11
1.25
1.15
95% CI
0.99–1.26
1.08–1.45
1.01–1.31
p-value
0.09
0.003
0.04
Number of
observations
5691
4141
5433
Each patient could contribute only once per period, but could contribute in different periods of time. Due to stratification by period, no adjustment for the
correlation structure was done.
OR: odds ratio; RCT: randomised controlled trial.
Table 4.
Negative binomial regression on days admitted to
hospital.
IRR
Telemedicine
Age
Gender (female)
Pre-RCT
During RCT
Post-RCT
1.97
1.017
1.01
–
0.80
0.84
95% CI
1.82–2.15
1.014–1.021
0.95–1.08
–
0.74–0.85
0.79–0.90
p-value
<0.0001
<0.0001
0.70
–
<0.0001
<0.0001
Table 5.
Cox proportional hazard regression on death during the
study period.
HR
Telemedicine
Age
Gender (female)
Pre-RCT
During RCT
Post-RCT
Interaction
Telemedicine
Â
RCT
Telemedicine
Â
post-RCT
0.50
1.035
0.84
–
0.95
1.0
–
1.05
1.14
95% CI
0.37–0.68
1.032–1.039
0.78–0.89
–
0.87–1.04
0.93–1.08
–
0.66–1.66
0.80–1.63
p-value
<0.0001
<0.0001
<0.0001
–
0.25
0.98
–
0.83
0.46
Observations: 15,265, adjusted for 11,303 clusters.
IRR: incidence rate ratio; RCT: randomised controlled trial.
RCT, patients receiving telemedicine were more likely to
experience a readmission compared to patients not receiv-
ing telemedicine (OR 3.44, 95% CI 2.59–4.57,
p
<
0.0001).
In the post-RCT group, patients receiving telemedicine
were more likely to be readmitted to hospital (OR 6.04,
95% CI 5.09–7.16,
p
<
0.0001).
In all three periods, higher age was significantly asso-
ciated with the risk of experiencing a readmission (OR
between 1.022 and 1.03,
p
<
0.0001). Females were more
likely to be readmitted to hospital than males during all
three periods of time. In the pre-RCT period, the differ-
ence was not statistically significant (OR 1.11, 95% CI
0.99–1.26,
p
¼
0.09), during the RCT period (OR 1.25,
95% CI 1.08–1.45,
p
¼
0.003) and in the post-RCT
period (OR 1.15, 95% CI 1.01–1.31,
p
¼
0.04) it was
statistically significant.
Observations: 11,316 in 8040 patients; deaths: 3729; time at risk: 9561.29
person-days.
HR: hazard ratio; RCT: randomised controlled trial.
0.95–1.08,
p
¼
0.70). The group of patients admitted
during the RCT experienced a decreased rate for days
admitted to hospital in comparison to the pre-RCT
group (IRR 0.80, 95% CI 0.74–0.85,
p
<
0.0001).
The same applied for the post-RCT group when com-
pared to the pre-RCT group (IRR 0.84, 95% CI 0.79–
0.90,
p
<
0.0001).
Death
Patients receiving telemedicine had half the risk of dying
during the five-year observation period of those who did
not (hazard ratio (HR) 0.50, 95% CI 0.37–0.68,
p
<
0.0001; see Table 5). Increased age increased the risk
of death by approximately 3.5% per year (HR 1.035, 95%
CI 1.032–1.039,
p
<
0.0001), and females were less likely to
die during the study period than males (HR 0.84, 95% CI
0.78–0.89,
p
<
0.0001). No statistically significant differ-
ences were observed between the pre-RCT and the
during RCT and post-RCT periods. No significant inter-
action between intervention and period of time was
observed.
When imputing missing start dates for outpatient visits
with the purpose of conducting a sensitivity analysis,
patients receiving telemedicine still had a lower risk of
Days spent admitted to hospital
Patients receiving the telemedicine intervention, compared
to patients who did not receive the telemedicine interven-
tion, had a rate 1.97 times greater for days of admission
(incidence rate ratio (IRR) 1.97, 95% CI 1.82–2.15,
p
<
0.0001; see Table 4). Age was positively related to
days spent in hospital, indicating that an increase in age
by one year was associated with an increase of 1.7% in the
rate of days spent at hospital (IRR 1.017, 95% CI 1.014–
1.021,
p
<
0.0001). Female gender was not significantly
related to days spent in hospital (IRR 1.01, 95% CI
5
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0006.png
382
Journal of Telemedicine and Telecare 21(7)
history of disease or socio-economic status. Further ana-
lyses including these variables would be highly relevant
contributions to the knowledge of external validity of tele-
medicine for COPD patients.
Another important limitation in the data extraction pro-
cedure could be the request for data. In the case of this article,
the criterion for entry into the cohort was COPD as primary
diagnosis. Different registration procedures imply that not
all usage of the telemedicine equipment was included in the
used data set. This is quite likely in that only 1105 observa-
tions of usage were registered over the period. Whether the
possibly missed registrations would affect the outcomes posi-
tively or negatively remains unknown.
Previously, only one published article has reported
differences between efficacy and effectiveness of a tele-
medicine intervention.
9
A web-based self-help interven-
tion was tested in an RCT, and subsequently the
intervention was made available to the general public.
Demographic differences between the RCT population
and the implementation population were identified.
Nevertheless, outcomes within the two populations were
similar, and the authors concluded that the external val-
idity of the RCT was high. However, this article did not
have sufficient data from which to draw conclusions on
the effectiveness in routine practice, and the results were
thus not comparable to those presented by Riper et al.
9
Methods of excluding observations within our cohort
are debatable. All excluded observations were, however,
thoroughly considered before the decision to exclude them
was made. Fifty-one observations with an age below
30 years were excluded on the basis of their age. The 26
observations that did not provide data supplementary to
the days of admission and discharge were disregarded,
since the lack of additional information would have
excluded them from the analyses in any case. The 17
admissions with duration of more than 90 days were
regarded as typing or registration errors. In the depart-
ment treating COPD patients at OUH, admissions of this
length do not occur.
24
Finally, one observation was
removed due to a register error since the date of admission
was registered as occurring after the date of death.
All-cause mortality was used in the survival analyses.
In the data, only the date of death was provided, not the
cause. There were two reasons for including mortality
from any cause in the analysis. First, a number of
causes of death can be difficult to distinguish from the
COPD diagnosis, e.g. breathlessness, which is caused by
the COPD, may imply concomitant heart failure or ven-
tricular arrhythmia causing sudden cardiac death. Second,
there was no reason to believe that deaths from causes
unrelated to COPD would be distributed differently
between the intervention group and the control group.
Although it might have further enlightened the findings
with more information on death causality, we do not
expect that the distribution of cause of death differed
between the groups in a way that would affect the results.
The decision to assign a patient to the intervention
group whenever he/she had received telemedicine at least
Table 6.
Sensitivity analysis of Cox proportional hazard regression
on death during the study period, imputing missing start dates for
outpatient visits.
HR
Telemedicine
Age
Gender (female)
Pre-RCT
During RCT
Post-RCT
Interaction
Telemedicine
Â
RCT
Telemedicine
Â
post-RCT
0.65
1.051
0.90
–
0.82
0.91
–
1.22
1.27
95% CI
0.47–0.88
1.047–1.054
0.84–0.96
–
0.76–0.90
0.84–0.98
–
0.77–1.93
0.89–1.81
p-value
0.006
<0.0001
0.001
–
<0.0001
0.01
–
0.40
0.18
Observations: 15,265 in 11,303 patients; deaths: 3793; time at risk: 16,033.35
person-days.
HR: hazard ratio; RCT: randomised controlled trial.
dying during the five-year observation period than those
who did not (HR 0.65, 95% CI 0.47–0.88,
p
¼
0.006), see
Table 6. Increased age increased the risk of death by
approximately 5% per year (HR 1.051, 95% CI 1.047–
1.054,
p
<
0.0001), and females were less likely to die
during the study period than males (HR 0.90, 95% CI
0.84–0.96,
p
¼
0.001). During the RCT and in the post-
RCT period, lower risk of death was observed than in
the pre-RCT period (HR 0.82, 95% CI 0.76–0.90,
p
<
0.0001 and HR 0.91, 95% CI 0.84–0.98,
p
¼
0.01,
respectively), but no significant interaction between inter-
vention and period of time was observed.
Discussion
To our knowledge, this is the first study on the routine
practice effectiveness of telemedicine used in COPD
patients. It was based on a large dataset including hospital
admissions over a five-year period, corresponding to
11,303 patients contributing with a total of 15,265 occur-
rences during the three time periods.
Our results showed that patients who received
telemedicine were (a) more likely to be readmitted to hos-
pital during the study period, (b) more likely to be
admitted for a longer time, and (c) less likely to die
during the study period than patients not receiving tele-
medicine. All analyses were adjusted for patients’ age and
gender. In our sensitivity analysis, the risk of death from
any cause was lower for people admitted during the two
latter periods of the study, i.e. during RCT and post-RCT,
as compared to those admitted prior to the RCT. This
may, however, simply be due to the reduced follow-up
time in the latter phases of the study as compared to the
pre-RCT phase.
Due to limitations in the availability of data, the selec-
tion criteria for the telemedicine intervention in the during
RCT period compared to periods of non-restricted imple-
mentation was determined for age and gender, but not
other relevant covariates such as severity of COPD,
6
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0007.png
Dyrvig et al.
once at any time point during the respective time periods
is likely to have had an impact on analyses, since we there-
fore analysed the data according to the intention-to-treat
principle. This is a very conservative approach in this set-
ting, and readmissions as well as days admitted to hospital
are, therefore, likely to be overestimated in the interven-
tion group. However, a favourable effect on survival in
patients having received telemedicine at some point
could still be observed. When patients were – during a
specific time period – sometimes admitted while receiving
telemedicine and sometimes not, differentiation of effects
was challenging. Therefore, we did not see an alternative
to our conservative approach of applying the intention-to-
treat principle, which in turn is likely to be the method of
choice in similar, other investigations.
This paper constitutes, to the best of our knowledge,
the third quantitative, scientific contribution on one inter-
vention for COPD patients. The first publication was a
cohort study of 100 consecutively selected patients that
were assigned to intervention or control in a 1:1 ratio on
the basis of geographic location of their residence.
25
The
first study identified a statistically significant protective
effect of the telemedicine intervention (HR: 0.25, CI:
0.09–0.69). The results were, however, discussed to be sub-
ject to selection bias. The second publication was an RCT
of 266 patients, which identified no statistically significant
differences in risk of readmission between intervention
and control groups.
18
Finally, this third publication – a
cohort study of the effectiveness – found a statistically
significant increased risk of readmission among the recipi-
ents of telemedicine compared to non-recipients.
This discrepancy among findings is highly interesting
and highlights both the divergence among results of dif-
ferent methodological approaches and the need for
follow-up data collections on interventions implemented
on the basis of scientific studies.
A number of further analyses are necessary for any
final conclusions on the effectiveness of telemedicine for
COPD patients. The most important data that would fur-
ther clarify the issue include severity or history of the
COPD, comorbidities, experience with the intervention,
number of days until first readmission, residential area,
and socio-demographic status. It was not possible to
include these variables in the current research project,
but further knowledge about these indicators could help
clarify the relationship between the intervention and
patient outcomes.
383
intervention had a higher risk of readmission and simul-
taneously a lower risk of death when the intervention was
applied in routine practice. This is in contrast to findings
in other fields in which a clearer pattern emerges: everyday
effectiveness has been shown to be less beneficial than sug-
gested by efficacy studies. At this point, it is not clear why
this discrepancy was observed. It may be due to chance or
caused by unknown differences between telemedicine and
other clinical fields. To elucidate this possibility requires
further research. Relevant analyses should include add-
itional explanatory variables and/or the matching of
patients on the basis of criteria for inclusion. Follow-up
of cohorts is a relatively low-cost procedure of potential
advantage to many patients falling outside the eligibility
criteria for RCTs and of major importance for the inter-
pretation of studies like the present one. For this reason,
relevant follow-up should be included in the design phase
of future RCTs.
Acknowledgements
We would like to thank Professor Jørgen Vestbo for his kind
help in explaining procedures within the Department of
Respiratory Medicine at Odense University Hospital.
Authors’ contributions
AKD suggested the research strategy, collected data, carried out
all analyses and drafted the text. OG supported strategy for and
conduct of analyses and reviewed the text. KK reviewed the
strategy for analyses and the text. HV was the main supervisor
of all aspects of preparing and writing the manuscript.
All authors have read and approved the final version of
this manuscript.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The authors received no financial support for the research,
authorship, and/or publication of this article.
References
1. Travers J, Marsh S, Williams M, et al. External validity of
randomised controlled trials in asthma: to whom do the
results of the trials apply?
Thorax
2007; 62: 219–223.
2. Koeth O, Zahn R, Gitt AK, et al. Clinical benefit of early
reperfusion therapy in patients with ST-elevation myocardial
infarction usually excluded from randomized clinical trials
(results from the maximal individual therapy in acute myo-
cardial infarction plus [MITRA Plus] registry).
Am J Cardiol
2009; 104: 1074–1077.
3. Le Strat Y, Rehm J and Le Foll B. How generalisable to
community samples are clinical trial results for treatment of
nicotine dependence: a comparison of common eligibility cri-
teria with respondents of a large representative general popu-
lation survey.
Tob Control
2011; 20: 338–343.
Conclusion/perspectives
Knowledge of routine practice effectiveness is relevant for
many current and future patients treated in healthcare
systems throughout the world. Along with Riper et al.,
9
this article suggests that the external validity of telemedi-
cine applications for COPD patients may differ from tele-
medicine applications in other clinical fields. Riper et al.
9
found external validity to be high. The current article
shows diverging results in that patients receiving the
7
 ULØ, Alm.del - 2019-20 - Endeligt svar på spørgsmål 61: MFU spm. om de nyeste undersøgelser af effekterne af KOL-kufferten, til sundheds- og ældreministeren
2125261_0008.png
384
4. Rothwell PM. External validity of randomised controlled
trials: ‘‘To whom do the results of this trial apply?’’
Lancet
2005; 365: 82–93.
5. Singal AG, Higgins PD and Waljee AK. A primer on effect-
iveness and efficacy trials.
Clin Transl Gastroenterol
2014; 5:
e45.
6. Hutchinson-Jaffe AB, Goodman SG, Yan RT, et al.
Comparison of baseline characteristics, management and
outcome of patients with non-ST-segment elevation acute
coronary syndrome in versus not in clinical trials.
Am
J Cardiol
2010; 106: 1389–1396.
7. Maasland L, van Oostenbrugge RJ, Franke CF, et al.
Patients enrolled in large randomized clinical trials of anti-
platelet treatment for prevention after transient ischemic
attack or ischemic stroke are not representative of patients
in clinical practice: the Netherlands stroke survey.
Stroke
2009; 40: 2662–2668.
8. Rasoul S, Ottervanger JP, Dambrink JH, et al. External
validity of ST elevation myocardial infarction trials: the
Zwolle studies.
Cathet Cardiovasc Interv
2007; 69: 632–636.
9. Riper H, Kramer J, Conijn B, et al. Translating effective
web-based self-help for problem drinking into the real
world.
Alcohol Clin Exp Res
2009; 33: 1401–1408.
10. van der Lem R, van der Wee NJ, van Veen T, et al. Efficacy
versus effectiveness: a direct comparison of the outcome of
treatment for mild to moderate depression in randomized
controlled trials and daily practice.
Psychother Psychosom
2012; 81: 226–234.
11. Travers J, Marsh S, Caldwell B, et al. External validity of
randomized controlled trials in COPD.
Respir Med
2007;
101: 1313–1320.
12. Koeth O, Zahn R, Gitt AK, et al. Clinical benefit of early
reperfusion therapy in patients with ST-elevation myocar-
dial infarction usually excluded from randomized clinical
trials (results from the maximal individual therapy in acute
myocardial infarction plus [MITRA Plus] registry).
Am J
Cardiol
2009; 104: 1074–1077.
13. Pedone C and Lapane KL. Generalizability of guidelines
and physicians’ adherence. Case study on the Sixth Joint
Journal of Telemedicine and Telecare 21(7)
National Commitee’s guidelines on hypertension.
BMC
Publ Health
2003; 3: 24.
Hoertel N, Le Strat Y, De Maricourt P, et al. Are subjects in
treatment trials of panic disorder representative of patients
in routine clinical practice? Results from a national sample.
J Affect Disord
2013; 146: 383–389.
Wyatt JC and Sullivan F. eHealth and the future: promise or
peril?
Br Med J
2005; 331: 1391–1393.
Sundhedsvæsenet I-IfKoAi. Akkrediteringsstandarder for
sygehuse. IKAS, 2013.
Rasmussen O, Sorknaes AD, Svenningsen H, et al.
TeleKOL
kuffert – visitation og anvendelse,
2nd ed. OUH – Svendborg
Sygehus: Odense University Hospital, 2014.
Sorknaes AD, Bech M, Madsen H, et al. The effect of real-
time teleconsultations between hospital-based nurses and
patients with severe COPD discharged after an exacerbation.
J Telemed Telecare
2013; 19: 466–474.
McLean S, Nurmatov U, Liu JL, et al. Telehealthcare for
chronic obstructive pulmonary disease: Cochrane review
and meta-analysis.
Br J Gen Pract
2012; 62: e739–e749.
Vestbo J, Hurd SS and Rodriguez-Roisin R. The 2011 revi-
sion of the global strategy for the diagnosis, management
and prevention of COPD (GOLD) – why and what?
Clin
Respir J
2012; 6: 208–214.
Vestbo J, Hurd SS, Agustı´ AG, et al. Global strategy for the
diagnosis, management, and prevention of chronic obstruct-
ive pulmonary disease.
Am J Respir Crit Care Med
2013;
187: 347–365.
World Health Organization.
International classifications of
diseases (ICD).
Geneva: World Health Organization, 2015.
Vestbo J, Titlestad I, Dommerby V, et al.
OUH,
Patientforløbsbeskrivelse
for
Kronisk
Obstruktiv
Lungelidelse. 1.6 ed.
Denmark: Odense University
Hospital, 2014.
Vestbo J. Personal communication.
Sorknaes AD, Madsen H, Hallas J, et al. Nurse tele-consul-
tations with discharged COPD patients reduce early
readmissions – an interventional study.
Clin Respir J
2011;
5: 26–34.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
8