Remote symptom monitoring with patient-reported outcome measures in outpatients with chronic kidney disease (PROKID): a multicentre randomised controlled non-inferiority study

Abstract

Background
The rising incidence of chronic kidney disease (CKD) is straining the capacity of the outpatient clinics. Remote healthcare delivery might improve CKD follow-up compared to conventional face-to-face follow-up. Patient-reported outcomes (PROs) are used to empower remote follow-up and patient engagement. The consequences of shifting from face-to-face follow-up to remote outpatient follow-up on kidney function, health resource utilisation and quality of life remain unknown.

Methods
We conducted a multicentre pragmatic non-inferiority trial at three outpatient clinics in the Central Denmark Region. A total of 152 incident outpatients with CKD were randomised (1:1:1) to either PRO-based-, PRO-telephone follow-up or Standard of Care (SoC). The primary outcome was the annual change in kidney function measured by the slope of the estimated glomerular filtration rate (eGFR). The non-inferiority margin was eGFR 2.85mL/min/1.73m2/year. Mean differences were estimated using intention-to-treat, per protocol, and random coefficient models.

Results
Mean eGFR slope differences between PRO-based and SoC were -0.97 mL/min/1.73m2/year (95% CI -3.00;1.07) and -1.06 mL/min/1.73m2/year (95% CI -3.02;0.89) between PRO-telephone and SoC. Non-inferiority was only established in the per-protocol analysis due to confidence intervals exceeding the margin in the intention-to-treat group. Both intervention groups had fewer outpatient visits: -4.95 (95% CI -5.82; -4.08) for the PRO-based group and -5.21 (95% CI -5.95; -4.46) for the PRO-telephone group. We found no significant differences in quality of life, illness perception or satisfaction.

Conclusion
Differences in the eGFR slope between groups were non-significant, and results on non-inferiority were inconclusive. Thus, transitioning to remote PRO-based follow-up requires close monitoring of kidney function. Reducing patients' attendance in the outpatient clinic was possible without decreasing either quality of life or illness perception.
ClinicalTrials.gov Identifier: NCT03847766

Full text

Clinical Kidney Journal , 2024, vol. 17, no. 7, sfae176
https:/doi.org/10.1093/ckj/sfae176
Advance Access Publication Date: 14 June 2024
Original Article
ORIGINAL ARTICLE
Remote symptom monitoring with patient-reported
outcome measures in outpatients with chronic kidney
disease ( PROKID) : a multicentre randomised
controlled non-inferiority study
Birg ith Eng elst Grove
1 ,2
, Liv Marit Valen Schougaard1
, Frank Mose
3
,
Else Randers4
, Niels Henrik Hjollund1 ,2 ,5
, Per Ivarsen2 ,6 ,∗
and Annette De Thurah
2 ,7 ,∗
1
AmbuFlex – Centre for Patient-reported Outcomes, Gødstrup Hospital, Herning, Denmark,
2
Department of
Clinical Medicine, Aarhus University, Aarhus, Denmark,
3
Department of Renal Medicine, Gødstrup Hospital,
Herning, Denmark,
4
Department of Internal Medicine, Viborg Regional Hospital, Viborg, Denmark,
5
Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark,
6
Department of Renal
Medicine, Aarhus University Hospital, Aarhus, Denmark and
7
Department of Rheumatology, Aarhus
University Hospital, Aarhus, Denmark
∗These authors share last authorship.
Correspondence to: Birgith Engelst Grove; E-mail: bigcri@rm.dk
ABSTRACT
Background. The increasing incidence of chronic kidney disease ( CKD) is straining the capacity of outpatient clinics.
Remote healthcare delivery might improve CKD follow-up compared with conventional face-to-face follow-up.
Patient-reported outcomes ( PROs) are used to empower remote follow-up and patient engagement. The consequences of
shifting from face-to-face follow-up to remote outpatient follow-up on kidney function, health resource utilisation and
quality of life remain unknown.
Methods. We conducted a multicentre pragmatic non-inferiority trial at three outpatient clinics in the Central Denmark
Region. A total of 152 incident outpatients with CKD were randomised ( 1:1:1) to either PRO-based, PRO-telephone
follow-up or standard of care ( SoC) . The primary outcome was the annual change in kidney function measured by the
slope of the estimated glomerular ltration rate ( eGFR) . The non-inferiority margin was an eGFR of
2.85 ml/min/1.73 m2
/year. Mean differences were estimated using intention-to-treat ( ITT) , per protocol and random
coefcient models.
Results. Mean eGFR slope differences between PRO-based and SoC were −0.97 ml/min/1.73 m2
/year [95% condence
interval ( CI) −3.00–1.07] and −1.06 ml/min/1.73 m2
/year ( 95% CI −3.02–0.89) between PRO-telephone and SoC.
Non-inferiority was only established in the per-protocol analysis due to CIs exceeding the margin in the ITT group. Both
intervention groups had fewer outpatient visits: −4.95 ( 95% CI −5.82 to −4.08) for the PRO-based group and −5.21 ( 95% CI
−5.95 to −4.46) for the PRO-telephone group. We found no signicant differences in quality of life, illness perception or
satisfaction.
Received: 27.9.2023; Editorial decision: 7.5.2024
© The Author(s) 2024. Published by Oxford University Press on behalf of the ERA. This is an Open Access article distributed under the terms of the Creative
Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/) , which permits non-commercial re-use, distribution,
and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com
1

2B.E. Grove et al.
Conclusion. Differences in the eGFR slope between groups were non-signicant and results on non-inferiority were
inconclusive. Thus, transitioning to remote PRO-based follow-up requires close monitoring of kidney function. Reducing
patients’ attendance in the outpatient clinic was possible without decreasing either quality of life or illness perception.
ClinicalTrials.gov identier: NCT03847766
GRAPHICAL ABSTRACT
Keywords: chronic kidney disease, outpatient care, patient-reported outcome measures, randomized controlled trial,
remote symptom monitoring
KEY LEARNING POINTS
What was known:
• During the last decade, there has been increasing outpatient activities and the use of telephone consultations.
•Remote symptom monitoring using patient-reported outcomes ( PROs) has been investigated in other populations of chronic
and malignant diseases, but its potential in patients with chronic kidney disease ( CKD) remains unknown.
This study adds:
•No differences in the patients’ kidney function across the intervention groups were found and patients’ quality of life,
satisfaction and illness perception were not affected.
•Implementing PRO-based remote follow-up reduced face-to-face consultations but increased the number of telephone con-
sultations.
•Remote PRO-based follow-up had no negative effect on other CKD markers.
Potential impact:
•Close monitoring and an awareness of preserving kidney function is crucial in transitioning from the standard of care to
remote follow-up.
•Remote PRO-based follow-up decreases the need for face-to-face outpatient visits.

Remote symptom monitoring in outpatients with CKD 3
Incident patients to the
renal outpatient clinics
Assessed for eligibility (n=1260)
Excluded (n=1108):
• Not meeting inclusion criteria (n=940)
• Declined to participate (n=122)
• Ended follow-up (n=46)
Analyzed
• Intention to treat (n=51)
• Per-protocol (n=37)
Discontinued intervention (n=14)
• Deceased (n=5)
• Initiated dialyses (n=2)
• Withdrawn by clinician (n=3)
• Comorbidity, patient wish (n=1)
• Long-term hospitalisation (n=2)
• Ended follow-up (n=1)
Discontinued intervention (n=14)
• Deceased (n=4)
• Patient wish for phone consult (n=2)
• Withdrawn by clinician (n=1)
• Cancer diagnosing (n=1)
• Psychiatric diagnosis (n=1)
• Ended follow-up (n=5)
Allocated to face-to-face consultation
(standard of care) (n=47)
Analyzed
• Intention to treat (n=47)
• Per-protocol (n=33)
Allocation
Analyses
Follow-up
Randomized (n=152)
Enrollment
Allocated to PRO-telephone follow-up
(n=54)
Allocated to PRO-based follow-up
(n=51)
Discontinued intervention (n=8)
• Deceased (n=2)
• Comorbidity, patient wish (n=1)
• Withdrawn by clinician (n=2)
• Other attendance follow-up (n=1)
• Ended follow-up (n=2)
Analyzed
• Intention to treat (n=54)
• Per-protocol (n=46)
Figure 1: CONSORT ow diagram.
INTRODUCTION
Globally, the incidence of people with chronic kidney disease
( CKD) and the demand to deliver more healthcare is increas-
ing [1 ,2 ]. In Denmark, patients with CKD are referred to spe-
cialist care for diagnosis and treatment in hospitals when
the estimated glomerular ltration rate ( eGFR) reaches 30–
40 ml/min/1.73 m2
[3 ]. Providing care for people with CKD stages
3b–5 requires monitoring kidney function, symptom burden and
overall well-being [4 ,5 ]. Patients’ health conditions may be cap-
tured using patient-reported outcomes ( PROs) collected through
disease-specic questionnaires [6 ,7 ]. Using PROs in outpatient
care may provide additional information about patient percep-
tions of their health [8 ,9 ]. When regular outpatient visits are re-
placed with disease-specic questionnaires, it is termed ‘PRO-
based remote follow-up’ [10 ]. The effects of using PRO in remote
care have been investigated in other populations and have been
shown to improve symptom control [11 ,12 ], patient–clinician
communication [13 ], satisfaction and supportive care [14 ,15 ]
and decreased outpatient appointments [16 ,17 ]. Evidence for
the effects of remote care for patients with CKD is not yet es-
tablished, even though studies have shown that, e.g., telephone
consultations have been widely adopted as a safe option for
receiving care during the coronavirus disease 2019 ( COVID-19)
pandemic [18 ].
To investigate the efcacy and safety of remote symptom
monitoring in patients with CKD, we conducted a multicen-
tre randomised non-inferiority study ( PROKID) with two differ-
ent intervention groups compared with a standard-of-care ( SoC)
group. The intervention groups were either entirely managed by
PROs or PROs supported by telephone consultations. The pri-
mary endpoint was non-inferiority in the difference of the eGFR
slope. Secondary endpoints were the difference in quality of life
( QOL) , illness perception and the use of healthcare resources.
We hypothesised that the eGFR slope change was non-
inferior between patients in the remote PRO-based follow-up
groups compared with patients receiving the SoC.
MATERIALS AND METHODS
Study design and participants
The PROKID study is a multicentre non-inferiority randomised
controlled study carried out at Aarhus University Hospital,
Gødstrup Hospital and Region Hospital Central, Viborg, Den-
mark. Patients were included from January 2019 until August
2021 with 18 months of follow-up. Patients were eligible if they
were newly referred to the renal outpatient clinic, ≥18 years
of age, with an eGFR of 10–40 ml/min/1.73 m2
and did not ex-
hibit cognitive dysfunction. The main exclusion criteria were
projected risk of progression to end-stage kidney disease ( ESKD)
within 12 months, inability to answer a questionnaire or suf-
fering from terminal illness ( Fig. 1 ) . The design and procedures
of the study have been published previously [19 ]. The Con-
solidated Standard of Reporting Tria l ( CONSORT) Extension for
Non-inferiority [20 ] and CONSORT PRO [21 ] extension checklists
were followed ( Supplementary Table S1) . The study was con-
ducted following the Helsinki Declaration, and the Danish Data

4B.E. Grove et al.
Protection Agency granted permission to store and use conden-
tial data ( no. 1-16-02-873-17) . Ver bal and signed written consent
was obtained from each patient before enrolment.
Randomisation
Participants were randomly allocated ( 1:1:1) to either PRO-based,
PRO-telephone or SoC. Computer-generated randomisation was
used and project nurses carried out randomisation after en-
rolment and the patients completed a baseline questionnaire.
Blinding was not possible.
Interventions
All patients in the study were followed for 18 months with six
planned contacts and had blood samples taken at a local clinic or
hospital before contact. Irrespective of group allocation, patients
were allowed to initiate contact between visits.
Patients randomised to the PRO intervention responded to a
disease-specic questionnaire either on paper or electronically
through a generic web PRO system before each consultation
[6 ,22 ] that included self-reported weight and blood pressure ( BP) .
These PRO data were available to physicians in the electronic
health record [6 ].
PRO-based follow-up
The questionnaire was used as a decision aid together with
other clinical data for triaging patient care and contact. A
physician assessed the questionnaires, provided feedback on
responses and blood tests by secure email and called the patient
or scheduled a face-to-face visit if necessary or if it was desired
by the patient.
PRO-telephone follow-up
The questionnaire was used as communication support and
a symptom monitoring aid during the telephone consultation.
The patients’ responses, results of blood tests and BP and weight
were discussed.
SoC ( control group)
Patients receiving the SoC were seen by the physician and had
BP and weight measured by the nurse in the outpatient clinic.
Primary outcome
The primary outcome was the difference in the slope of eGFR
per year between the intervention groups and the SoC group.
Secondary outcomes
The secondary outcomes were the difference in other CKD
markers [urine albumin:creatinine ratio ( UACR ) , plasma potas-
sium, plasma phosphate and BP], ESKD, hospitalisations and
PROs, including health-related quality of life ( HRQOL) , measured
by the EuroQoL ve dimension ( EQ-5D) index and EQ-5D visual
analogue scale ( VAS) [23 ], and illness perception ( IP) measured
by the Brief Illness Perception Questionnaire ( BIPQ) [24 ]. Health-
care evaluation comprises condence, satisfaction, involvement
and safety measured by single items from the Danish Cancer So-
ciety’s Patient-Reported Experience Measure questionnaire [25 ].
Registered contacts included all outpatient telephone consulta-
tions and face-to-face consultations. Outpatient visits included
all face-to-face consultations. Additional information on the
nature of resource utilisation was obtained from the medical
records and captured in a database ( REDCap) [26 ]. An overview
of outcomes has been published [19 ].
Sample size
Based on a literature review, we assumed that the expected
loss in eGFR would be ≈5 ml/min/1.73 m2
/year in all groups
[27 ]. The sample size was calculated using a non-inferiority
margin of 2.85 ml/min/1.73 m2
/year between the groups [19 ,
27 ]. This estimate was based on existing literature [28 ], clinical
judgements [20 ] and the assumption that the study participants
would present a consistent eGFR during the follow-up period
[27 ]. Given 90% statistical power and a P -value of 0.05, we
needed 34 patients in each group to detect non-inferiority group
differences. To examine secondary outcomes and account for
attrition, a total of 152 patients were enrolled.
Statistical methods
All randomised participants were included in the intention-to-
treat ( ITT) analyses. Per-protocol analysis included patients who
completed all six contacts in their allocated group. Each inter-
vention group was compared with the SoC group. Normally dis-
tributed baseline data were presented with means and standard
deviations ( SDs) ; otherwise, medians and interquartile ranges
( IQRs) were reported. Sum scores followed guidelines for han-
dling missing items for each score.
Kidney function
The primary outcome was the mean difference ( MD) and a two-
sided 95% condence interval ( CI) in a change in kidney function
measured as the slope of eGFR within and between the groups. A
random coefcient mixed model was used [29 ]. Before the anal-
yses, outliers were identied and longitudinal plots of the data
over time were constructed for visual presentation. Model as-
sumptions were checked by comparing observed and expected
within-subject SDs and correlations and by inspecting plots of
residuals versus tted values and QQ plots. The model included
xed and random ( time) effects for the intercept and coefcient.
We performed a sensitivity analysis adjusting for sex, age and
comorbidity.
Secondary outcomes
CKD markers were analysed by calculating the MD from baseline
to the end of follow-up and compared between groups. When-
ever possible, an ITT approach was used. PRO data were analysed
in the per-protocol population. Between- and within-group dif-
ferences were calculated using a linear mixed regression model.
Longitudinal plots of the PRO data over time were constructed
to visualise the presentation of data. Between-group differences
in the categorical variables, such as the utilization of health-
care, were summarised and reported as numbers and percent-
ages. The MD between groups was assessed by linear regres-
sion. Due to an expected skewed distribution, 95% CIs will be
found using the bootstrap method with 1000 replications [30 ].
Two-sided P -values < .05 were considered to indicate statistical
signicance.

Remote symptom monitoring in outpatients with CKD 5
Tab l e 1: Baseline characteristics.
Var i abl es To t a l ( N = 152) SoC ( n = 47)
PRO-telephone
follow-up ( n = 54)
PRO-based follow-up
( n = 51)
Age ( years) , median ( IQR) 74 ( 68–79) 74 ( 64–79) 74 ( 68–78) 75 ( 68–80)
Male, n( %) 98 ( 64) 32 ( 68) 31 ( 57) 35 ( 68)
eGFR ( ml/min/1.73 m2
) 28.67 ( 6.22) 28.29 ( 6.60) 28.98 ( 6.49) 28.72 ( 6.43)
Systolic BP ( mmHg) 133 ( 21) 133 ( 21) 132 ( 20) 136 ( 21)
Diastolic BP ( mmHg) 75 ( 12) 77 ( 12) 75 ( 13) 75 ( 12)
Pulse ( bpm) 72 ( 13) 74 ( 15) 71 ( 12) 71 ( 13)
Charlson Comorbidity Index 1.96 ( 1.33) 2.02 ( 1.24) 1.91 ( 1.28) 1.96 ( 1.48)
Plasma phosphate ( mmol/l) 1.16 ( 0.25) 1.21 ( 0.30) 1.15 ( 0.22) 1.15 ( 0.24)
Plasma albumin ( g/l) 37.70 ( 3.01) 38.09 ( 3.16) 37.65 ( 3.28) 37.38 ( 2.56)
UACR ( mg/ml) , median ( IQR) 75 ( 476) 106 ( 558) 62 ( 574) 56 ( 230)
Plasma potassium ( mmol/l) 4.36 ( 0.60) 4.28 ( 0.43) 4.45 ( 0.71) 4.32 ( 0.60)
Haemoglobin ( mmol/l) 7.83 ( 1.10) 7.90 ( 1.32) 7.81 ( 0.98) 7.8 ( 1.02)
Education, n( %)
Low ( < 10 years) 37 ( 24) 10 ( 21) 14 ( 26) 13 ( 25)
Medium/high ( ≥10 years) 100 ( 66) 32 ( 68) 35 ( 65) 33 ( 65)
Missing 15 ( 10) 5( 11) 5( 9) 5( 10)
Labour market afliation, n( %)
Employed 20 ( 13) 9( 19) 7( 13) 4( 8)
Retired 119 ( 78) 33 ( 70) 43 ( 80) 43 ( 84)
Missing 13 ( 9) 5( 11) 4( 7) 4( 8)
QOL ( EQ-5D index)
Median ( IQR) 0.878
( 0.755–0.959)
0.815
( 0.660–0.939)
0.912
( 0.819–1)
0.858
( 0.758–0.952)
Mean ( SD) 0.82 ( 0.21) 0.78 ( 0.19) 0.86 ( 0.21) 0.81 ( 0.20)
Missing, n( %) 5( 3) 3( 6) 1( 2) 1( 2)
QOL ( EQ-5D-VAS)
Median ( IQR) 70 ( 50–80) 62.5 ( 50–80) 80 ( 60–85) 70 ( 50–80)
Mean ( SD) 67.36 ( 21.09) 63.98 ( 20.23) 70.84 ( 22.76) 66.64 ( 19.72)
Missing, n( %) 5( 3) 3( 6) 1( 2) 1( 2)
Illness perception score 37.73 ( 11.95) 40.59 ( 10.71) 34.42 ( 12.03) 38.56 ( 12.35)
Missing, n( %) 14 ( 9) 5( 10) 6( 11) 3( 6)
Val u es are presented as mean ( SD) unless stated otherwise.
RESULTS
From January 2019 to August 2021, 320 newly referred patients
with CKD were found eligible for possible inclusion in the
PROKID trial and a total of 152 ( 48%) accepted participation. A
total of 25 ( 16%) patients left the study or died, 11 ( 7%) after ran-
domisation, leaving 116 ( 76%) patients for the per-protocol anal-
yses ( Fig 1 ) . Non-completer analyses showed that dropouts were
older ( P = .04) and had a higher level of comorbidity ( P = .01) and
a lower level of concentration ( P = .02) ( Supplementary Table S2) .
Demographic and clinical characteristics at baseline were bal-
anced ( P < .05) between the groups ( Tabl e 1 ) .
Kidney function
We found no statistical differences in the eGFR slope across
the intervention groups. Nevertheless, as the lower limit of
the CIs extended the non-inferiority threshold in both PRO-
based intervention groups, non-inferiority was not established
for the primary outcome in the ITT analysis ( Fig. 2 ) . A differ-
ence in eGFR slope when comparing each of the PRO interven-
tion groups with the SoC was −0.97 ml/min/1.73 m2
/year ( 95%
CI −3.00–1.07) and −1.06 ml/min/1.73 m2
/year ( 95% CI −3.02–
0.89) , respectively ( Tabl e 2 ) . As the CI includes zero and the
non-inferiority margin, the results in the ITT population are
inconclusive. In the per-protocol analysis, non-inferiority was
established. The adjusted analyses reached the same results
Favours
standard of care
Favours PRO-based
remote follow-up
PRO-tele vs. SoC
PRO-based vs. SoC
PRO-tele vs. SoC
PRO-based vs. SoC
ITT Analysis
PP Analysis
–2.85
eGFR (mL/min/1.73 m2/year)
–2 –1 0 1 2
Figure 2: Forrest plot displaying the difference in the slope of eGFR between the
PRO intervention groups and the SoC. ITT and per-protocol analyses.
( Supplementary Table S3) . While a decrease in the eGFR slope
was noted within all groups, only in the PRO-telephone group
was a statistically signicant decrease observed, with an eGFR
slope of −1.45 ml/min/1.73 m2
/year ( 95% CI −2.77 to −0.14)
( Tabl e 3 and Fig. 3 ) .

6B.E. Grove et al.
Tab l e 2: Differences in the change in the eGFR slope between the PRO-based intervention groups and the SoC 18 months after randomisation.
PRO-based versus SoC PRO-telephone versus SoC
Intervention n
Change in eGFR
a
( ml/min/1.73 m2
/year)
( 95% CI) P -value
Change in eGFR
b
( ml/min/1.73 m2
/year)
( 95% CI) P -value
ITT 152 −0.97
( −3.00–1.07)
.35 −1.06
( −3.02–0.89)
.28
PP 116 −0.62
( −2.58–1.35)
.54 −0.86
( −2.74–1.02)
.36
PP: per-protocol.
a
The estimated mean difference in eGFR slopes between the PRO-based remote follow-up and SoC group.
b
The estimated mean difference in eGFR slopes between the PRO-telephone and SoC group.
Random coefcient mixed models were used.
Tab l e 3: Decline in eGFR within each of the intervention groups 18 months after randomisation.
SoC PRO-telephone PRO-based
Intervention
eGFR decline
( ml/min/1.73 m2
/year)
( 95% CI) ( n = 47/33)
eGFR decline
( ml/min/1.73 m2
/year)
( 95% CI) ( n = 54/46)
eGFR decline
( ml/min/1.73 m2
/year)
( 95% CI) ( n = 51/37)
ITT −0.39
( −1.85–1.07)
−1.45*
( −2.77 to −0.14)
−1.36
( −2.78–0.06)
PP −0.20
( −1.63–1.23)
−1.06
( −2.28–0.15)
−0.82
( −2.17–0.54)
PP: per-protocol.
Linear mixed model regression was used. A negative value means a decrease in eGFR.
*Statistically signicant ( P < .05) .
Usual follow-up
(control group)
PRO-based
PRO telephone
93Baseline 6 12 15 18
Time since randomisation (months)
eGFR
32
30
28
26
24
Change from baseline score
Figure 3:
Change in eGFR within and between the intervention groups.
Healthcare utilisation
Patients who received the PRO-based intervention had signi-
cantly fewer contacts with the outpatient clinic than those who
received the SoC [MD −3.04 ( 95% CI −4.43 to −1.64) ]( Ta ble 4 ) .
The total number of contacts to the outpatient clinic was high-
est in the PRO-telephone group, with a median of 7( IQR 2) con-
tacts, followed by a median of 6( IQR 3) contacts in the SoC group
and a median of 3( IQR 4) registered contacts for the patients
in the PRO-based group ( Ta ble 4 ) . Both intervention groups had
fewer outpatient visits compared with the SoC, with an MD of
−4.95 ( 95% CI −5.82 to −4.08) for the PRO-based intervention
and −5.21 ( 95% CI −5.95 to −4.46) for the PRO-telephone group.
Accordingly, patients in the intervention groups had more fre-
quent telephone consultations compared with the SoC, with
an MD of 1.99 ( 95% CI 1.19–2.80) for the PRO-based interven-
tion and 5.80 ( 95% CI 5.02–6.58) for the PRO-telephone group.
Supplementary Table S4 outlines an overview of the type of
contact.
HRQOL
A higher not statistically signicant difference in self-reported
outcomes ( EQ-VAS) was seen in the PRO-based follow-up group

Remote symptom monitoring in outpatients with CKD 7
Tab l e 4: Healthcare utilisation during the 18-month follow-up period among outpatients with CKD ( ITT population = 152) .
Var i abl es SoC ( n = 47)
PRO-telephone
( n = 54)
PRO-based follow-up
( n = 51)
Mean difference
a
PRO-telephone versus
SoC ( 95% CI)
Mean difference
a
PRO-based versus
SoC ( 95% CI)
Follow-up time ( months)
Tota l 708 888 762
Median ( IQR) 18 ( 6) 18 ( 0) 18 ( 6) 1.38 ( −0.50–3.26) −0.12 ( −2.22–1.97)
All registered contacts
Tota l 315 394 191
Median ( IQR) ( range) 6( 3) ( 0–19) 7( 2) ( 0–17) 3( 4) ( 0–20) 0.59 ( −0.66–1.84) −3.04 ( −4.43 to −1.64)
Contacts per month, mean ( SD) 0.45 ( 0.16) 0.45 ( 0.15) 0.24 ( 0.20) −0.01 ( −0.07–0.06) −0.22 ( −0.29 to −0.14)
Outpatient visits
b
, median
( IQR) ( range)
6( 2) ( 5–13) 0( 1) ( 0–5) 0( 1) ( 0–12) −5.21 ( −5.95 to −4.46) −4.95 ( −5.82 to −4.08)
Outpatient visits per month,
mean ( SD)
0.39 ( 0.10) 0.04 ( 0.08) 0.06 ( 0.11) −0.35 ( −0.38 to −0.31) −0.33 ( −0.38 to −0.29)
Telephone consultations
c
,
median ( IQR) ( range)
0( 1) ( 0–8) 6( 2) ( 6–14) 3( 3) ( 0–10) 5.80 ( 5.02–6.58) 1.99 ( 1.19–2.80)
Telephone consultation per
month, mean ( SD)
0.07 ( 0.13) 0.40 ( 0.10) 0.19 ( 0.17) 0.33 ( 0.38–0.38) 0.13 ( 0.07–0.19)
Additional contacts
d
, median
( IQR) ( range)
1( 2) ( 0–11) 1( 2) ( 0–11) 1( 3) ( 0–14) 0.22 ( −0.72–1.16) 0.37 ( −0.67–1.40)
a
The ITT MDs and 95% CIs were obtained after linear regression by using the bootstrap method with 1000 replications [30 ].
b
Including the scheduled visits in the usual follow-up group ( control group) .
c
Including the scheduled telephone consultations in the PRO-telephone group.
d
Number of additional contacts between scheduled contacts.
as compared with the SoC [MD 4.56 points ( 95% CI −3.55–12.67) ]
( Supplementary Table S5) . Patients in the SoC gr oup r e ported
the lowest QOL from the onset of the study and did not change
in level during follow-up ( Fig. 4 ) . We found no between-group
differences in the EQ-5D index score.
Illness perception
Patients in the PRO-based intervention groups reported a more
threatening view of their illness during follow-up compared
with the SoC group, although none of the difference was
statistically different ( Fig. 4 and Supplementary Table S5) .
Healthcare evaluation
No statistically signicant differences were found between
the SoC and intervention groups regarding evaluation of
healthcare ( satisfaction, involvement, safety and condence) .
Fig. 4 outlines the development over time ( numbers shown in
Supplementary Table S5) . More than 95% of the patients in all
three groups answered that they were ‘very satised/satised’
with their mode of control.
No differences in the CKD markers such as BP and UACR were
found ( Supplementary Table S6) .
DISCUSSION
We hypothesised that a change in kidney function, measured by
the slope of eGFR, was non-inferior between patients in the two
PRO-based intervention groups compared with patients receiv-
ing the SoC. We found no signicant differences in the slope of
eGFR, and results regarding non-inferiority were inconclusive.
Patients in the PRO-based intervention groups had an overall re-
duction of face-to-face consultations and more telephone con-
sultations. No signicant differences between the groups were
found in the clinical data, patients’ QOL, illness perception and
health service evaluation.
In the ITT analyses, we did not reach non-inferiority in the
slope of eGFR, as the lower bound of the CIs reached the non-
inferior limit. In the per-protocol analyses, the slope of the eGFR
decline between the SoC and intervention groups was not sig-
nicantly different and all the estimates were within the non-
inferiority margin. Thus the per-protocol analyses demonstrated
non-inferiority and are considered equally important [31 ]. The
overall results regarding non-inferiority were inconclusive [20 ].
A commonly used minimum clinically important differ-
ence in eGFR has been reported in various studies ranging
from 1 to 5 ml/min/1.73 m2
/year [28 ,32 ]. Thus the differences
across the intervention groups ranging from eGFR −0.97 to
−1.06 ml/min/1.73 m2
/year with CIs reaching the non-inferiority
limit, represent a clinically meaningful difference. However,
the non-inferiority limit must be determined carefully and
will always depend on the specic context and population
[33 ]. During the follow-up period, the decrease in eGFR within
the intervention groups was modest, ranging from −0.39 to
−1.45 ml/min/1.73 m2
. These ndings were lower than those
from a recent meta-analysis, including CKD 3–5 cohorts [34 ]. We
recruited patients with an eGFR of 10–40 ml/min/1.73 m2
, how-
ever, the contrast may be caused by the low-risk CKD population
in our study.
Patients in the PRO-based intervention groups had signi-
cantly fewer contacts and fewer outpatient visits than the SoC
group. Effectiveness in terms of utilization of healthcare in a
remote care intervention has been investigated in other stud-
ies, reporting a lower number of outpatient visits in the inter-
vention groups among patients with rheumatoid arthritis [17 ,
35 ], inammatory bowel disease [36 ,37 ], type 1 diabetes [38 ]
and cancer [39 ]. Our study stands out from the other research
endeavours, as the PRO questionnaire assessed patients’ need
for contact. Combining the PRO questionnaires and blood sam-
ples formed the basis for triaging patient care and the type of

8B.E. Grove et al.
Illness perception
Higher score reflects a more
threatening view of illness
Quality of life
Line going up means
increased quality of life
Involvement
Line going up means
increased feeling of involvement
Satisfaction
Line going up means
increased satisfaction
Confidence
Line going up means
increased confidence
Safety
Line going up means
increased feeling of safety
Usual follow-up
(control group)
PRO-basedPRO telephone
Very muchSomehow
HighSomewhat
Extremely goodGood
Safety
Very muchSomehow
Involvement
EQ-5D index score
61218
Time since randomisation (months)
61218
Time since randomisation (months)
61218
Time since randomisation (months)
6Baseline 12 18
Time (months)
6Baseline 12 18
Time (months)
61218
Time since randomisation (months)
0.85 42
40
38
36
34
0.80
0.75
Figure 4: Change in PROs during the 18 months of follow-up. Per-protocol population = 116.
contact. In evaluating the resource utilisation of this interven-
tion, it is essential to note that patients in the SoC group were
consulting both a physician and a nurse, whereas patients in the
intervention groups primarily interacted with a physician, and
only saw a nurse if a nursing task was deemed relevant by the
physician. A process evaluation following the PROKID trial out-
lines the distribution and nature of visits [40 ].
Self-rated QOL at baseline was high, but lower than that of
the general Danish population at 0.90 ( SD 0.16) [41 ]. No signi-
cant differences were observed in the change in the QOL within
or between the groups, and only minor changes were detected.
Our ndings showed a non-signicant increase in IP com-
pared with the SoC. Nonetheless, we argue that knowledge of the
individual’s illness representation can be valuable in healthcare
settings. It may help healthcare providers to tailor their commu-
nication and treatment plans accordingly [42 ].
Qualitative studies have found that both patients and physi-
cians were overall condent in using a remote approach [40 ,
43 ]. This nding was supported in our study, as > 95% of
the patients reported condence and satisfaction with this
mode of follow-up. No statistically signicant differences in
patient satisfaction and health service evaluation across the
intervention groups were found. If patients felt uncondent
or unsatised with the follow-up mode, we would presum-
ably have seen a higher non-adherence rate. Non-adherence
to health technology interventions is a well-known problem
[44 ]. We had a reasonably high completion rate among the
participants, as 116/152 ( 76%) completed the study according
to the protocol. The implementation strategy, which in-
volved the active participation of clinicians and patients dur-
ing the development phase, may have contributed to this
outcome [19 ,22 ].
Strengths and limitations
One strength of this study lies in its pragmatic randomised
design, which enabled the production of feasible results by
aligning the research with real-life conditions in routine clinical
care.
Among the 320 patients eligible for the study, 46 ended
follow-up after the rst consultation, leaving 152 ( 48%) agreeing
to participate. We recruited newly referred patients from three
different outpatient clinics in three hospitals and the study pop-
ulation had similar characteristics to the source population [40 ],
which enhanced the external validity. However, we excluded
patients who were expected to have a rapid decline in kid-
ney function. This decision was made to encompass patients
with relatively stable kidney function and to minimise attrition
induced by the risk of reaching ESKD. This could potentially
impact the external validity and lead to a risk of being under-
powered. A potential risk of selection bias might occur, as only
patients who were capable of completing a questionnaire, con-
sidered to have a stable disease pathway and possessed cog-
nitive abilities were included. The study should be approached
with caution when attempting to apply its ndings to an older
population with greater comorbidity, as the individuals who dis-
continued participation were older and had more underlying
health conditions.
A major limitation arose from the inconclusive results re-
garding non-inferiority, hindering the ability to denitively con-
clude the non-inferiority of PRO-based interventions over the
SoC. A total of 36 patients left the study. Non-adherence to
the intervention may have inuenced the results, potentially
leading to an underestimation of the intervention’s effect com-
pared with what would have been observed if all patients had

Remote symptom monitoring in outpatients with CKD 9
followed the intervention [45 ]. The underlying reasons for the
loss of follow-up were similar between the groups and unrelated
to the intervention. Data for the primary outcome were complete
except for the deceased ( n = 11) . Thus the risk of bias due to
missing data was low. Moreover, the attrition analysis revealed
no difference regarding the primary outcome. However, several
patients ended follow-up in the outpatient clinic, especially in
the SoC group, often due to their preference for remote care con-
sultations [40 ] and probably as a consequence of the COVID-19
pandemic. Conversely, COVID-19 heightened the motivation and
engagement among physicians, as they benetted from the re-
mote monitoring of patients in the trial. Even post-pandemic,
remote monitoring will remain crucial for providing and triag-
ing healthcare services. Thus PRO-based interventions may help
older adults by continuously monitoring their health status and
reducing the burden on outpatient clinics.
CONCLUSION
In conclusion, differences in the slope of eGFR across the
intervention groups were non-signicant and our results regard-
ing non-inferiority were inconclusive. Implementing PROs in re-
mote care for patients with CKD may substitute for or replace
some of the traditional outpatient visits without compromis-
ing patients’ QOL, satisfaction or IP. In changing healthcare de-
livery modes, where remote patient management is increasing,
remote PROs may help improve care. Even though the results
were ambiguous regarding the non-inferiority of eGFR, we rec-
ommend close monitoring and tightened focus on maintain-
ing kidney function during the shift from traditional outpatient
follow-up to remote monitoring.
SUPPLEMENTARY DATA
Supplementary data are available at Clinical Kidney Journal online.
ACKNOWLEDGEMENTS
The patients and clinical staff who participated in the PROKID
study from Aarhus University Hospital, Gødstrup Hospital and
Viborg Hospital are gratefully acknowledged by the authors for
their valuable contributions. The project was approved by the
Danish Data Protection Agency ( ref 1-16-02-873-17) .
FUNDING
Karen Elise Jensen Foundation.
AUTHORS’ CONTRIBUTIONS
All authors were involved in drafting or reviewing the article,
with careful attention to the actual intellectual content. All au-
thors approved the nal version to be submitted for publication.
The rst author had complete access to all study data and is
accountable for maintaining the integrity and precision of the
data analysis.
DATA AVAILABILITY STATEMENT
The data underlying this article will be shared upon request to
the corresponding author.
CONFLICT OF INTEREST STATEMENT
The authors declare no conicts of interest.
REFERENCES
1. Levey AS, Schoolwerth AC, Burrows NR et al. Comprehensive
public health strategies for preventing the development,
progression, and complications of CKD: report of an expert
panel convened by the Centers for Disease Control and Pre-
vention. Am J Kidney Dis 2009; 53 :522–35. https://doi.org/10.
1053/j.ajkd.2008.11.019
2. GBD Chronic Kidney Disease Collaboration. Global, regional,
and national burden of chronic kidney disease, 1990–2017:
a systematic analysis for the Global Burden of Disease
Study 2017. Lancet 2020; 395 :709–33. https://doi.org/10.1016/
S0140-6736(20)30045-3
3. Dansk Nefrologisk Selskab, Dansk Paediatrisk Selskab,
Dansk Selskab for Klinisk Biokemi. Kronisk nyresyg-
dom: Analysemetoder og klinisk evaluering. Rekomman-
dationer for vurdering af glomerulær ltrationsrate og al-
buminuri. 2015. http://www.dskb.dk/Clubs/CommonDrive/
Components/GetWWWFile.aspx?leID=50003
4. Levey AS, Coresh J. Chronic kidney disease. Lancet
2012; 379 :165–80. https://doi.org/10.1016/S0140-6736(11)
60178-5
5. Almutary H, Bonner A, Douglas C. Symptom burden in
chronic kidney disease: a review of recent literature.
J Renal Care 2013; 39 :140–50. https://doi.org/10.1111/j.1755-
6686.2013.12022.x
6. Hjollund NHI, Larsen LP, Biering K et al. Use of patient-
reported outcome ( PRO) measures at group and patient lev-
els: experiences from the generic integrated PRO system,
WestChronic. Interact J Med Res 2014; 3 :e5.
7. Hjollund NHI, Larsen LP, de Thurah AL et al. Patient-reported
outcome ( PRO) measurements in chronic and malignant
diseases: ten years’ experience with PRO-algorithm-
based patient-clinician interaction ( telePRO) in AmbuFlex.
Qual Life Res 2023; 32 :1053–67. https://doi.org/10.1007/
s11136-022- 03322-9
8. van der Veer SN, Aresi G, Gair R. Incorporating patient-
reported symptom assessments into routine care for peo-
ple with chronic kidney disease. Clin Kidney J 2017; 10 :783–7.
https://doi.org/10.1093/ckj/sfx106
9. Greenhalgh J, Gooding K, Gibbons E et al. How do patient
reported outcome measures ( PROMs) support clinician-
patient communication and patient care? A realist synthe-
sis. J Patient Rep Outcomes 2018; 2 :42. https://doi.org/10.1186/
s41687-018- 0061-6
10. Schougaard LM, Larsen LP, Jessen A et al. AmbuFlex:
tele-patient-reported outcomes ( telePRO) as the basis for
follow-up in chronic and malignant diseases. Qual Life Res
2016; 25 :525–34. https://doi.org/10.1007/s11136-015- 1207-0
11. Basch E, Deal AM, Kris MG et al. Symptom monitoring with
patient-reported outcomes during routine cancer treat-
ment: a randomized controlled trial. J Clin Oncol 2016; 34 :557–
65. https://doi.org/10.1200/JCO.2015.63.0830
12. Denis F, Basch E, Septans A-L et al. Two-year survival
comparing web-based symptom monitoring vs routine
surveillance following treatment for lung cancer. JAMA
2019; 321 :306–7. https://doi.org/10.1001/jama.2018.18085
13. Carfora L, Foley CM, Hagi-Diakou P et al. Patients’ experi-
ences and perspectives of patient-reported outcome mea-
sures in clinical care: a systematic review and qualitative

10 B.E. Grove et al.
meta-synthesis. PLoS One 2022; 17 :e0267030. https://doi.org/
10.1371/journal.pone.0267030
14. Chen J, Ou L, Hollis SJ. A systematic review of the impact of
routine collection of patient reported outcome measures on
patients, providers and health organisations in an oncologic
setting. BMC Health Serv Res 2013; 13 :211. https://doi.org/10.
1186/1472-6963- 13-211
15. Kotronoulas G, Kearney N, Maguire R. What is the value of
the routine use of patient-reported outcome measures to-
ward improvement of patient outcomes, processes of care,
and health service outcomes in cancer care? J Clin Oncol
2014; 32 :1480–501.
16. Aiyegbusi OL, Hughes SE, Peipert JD et al. Reducing the
pressures of outpatient care: the potential role of patient-
reported outcomes. J R Soc Med 2023; 116 :44–64. https://doi.
org/10.1177/01410768231152222
17. de Thurah A, Stengaard-Pedersen K, Axelsen M et al. Te le-
health followup strategy for tight control of disease activ-
ity in rheumatoid arthritis: results of a randomized con-
trolled trial. Arthritis Care Res 2018; 70 :353–60. https://doi.org/
10.1002/acr.23280
18. Bokolo AJ. Application of telemedicine and eHealth tech-
nology for clinical services in response to COVID19 pan-
demic. Health Technol 2021; 11 :359–66. https://doi.org/10.
1007/s12553-020- 00516-4
19. Grove BE, Ivarsen P, de Thurah A et al. Remote follow-up
using patient-reported outcome measures in patients with
chronic kidney disease: the PROKID study—study proto-
col for a non-inferiority pragmatic randomised controlled
trial. BMC Health Serv Res 2019; 19 :631. https://doi.org/10.
1186/s12913-019- 4461-y
20. Piaggio G, Elbourne DR, Pocock SJ et al. Reporting of non-
inferiority and equivalence randomized trials: extension
of the CONSORT 2010 statement. JAMA 2012; 308 :2594–604.
https://doi.org/10.1001/jama.2012.87802
21. Calvert M, Blazeby J, Altman DG et al. Reporting of patient-
reported outcomes in randomized trials: the CONSORT PRO
extension. JAMA 2013; 309 :814–22. https://doi.org/10.1001/
jama.2013.879
22. Grove BE, Schougaard LMV, Ivarsen PR et al. Patient-reported
outcome measures for clinical decision-making in outpa-
tient follow-up: validity and reliability of a renal disease
questionnaire. J Patient Rep Outcomes 2021; 5 :107. https://doi.
org/10.1186/s41687-021-00384-0
23. Rabin R, De Charro F. EQ-5D: a measure of health status from
the EuroQol Group. Ann Med 2001; 33 :337–43. https://doi.org/
10.3109/07853890109002087
24. Broadbent E, Petrie KJ, Main J et al. The Brief Illness Percep-
tion Questionnaire ( BIPQ) . J Psychosom Res 2006; 60 :631–7.
25. Danish Cancer Society. Kræftpatienters behov og oplevelser
med sundhedsvæsenet i opfølgnings- og efterforløbet,
Kræftens Bekæmpelses Barometerundersøgelse, danish:
Kræftens bekæmpelse, 2019.
26. Harris PA, Tay l o r R, Thielke R et al. Research electronic
data capture ( REDCap) —a metadata-driven methodology
and workow process for providing translational research
informatics support. J Biomed Inform 2009; 42 :377–81. https://
doi.org/10.1016/j.jbi.2008.08.010
27. Meijer E, Drenth JP, d’Agnolo H et al. Rationale and de-
sign of the DIPAK 1 study: a randomized controlled clini-
cal trial assessing the efcacy of lanreotide to halt disease
progression in autosomal dominant polycystic kidney dis-
ease. Am J Kidney Dis 2014; 63 :446–55. https://doi.org/10.1053/
j.ajkd.2013.10.011
28. Heerspink HJL, Jongs N, Chertow GM et al. Effect of da-
pagliozin on the rate of decline in kidney function in pa-
tients with chronic kidney disease with and without type 2
diabetes: a prespecied analysis from the DAPA-CKD trial.
Lancet Diabetes Endocrinol 2021; 9 :743–54. https://doi.org/10.
1016/S2213-8587(21)00242-4
29. Janmaat CJ, van Diepen M, Tsonaka R et al. Pitfalls of linear
regression for estimating slopes over time and how to avoid
them by using linear mixed-effects models. Nephrol Dial
Transplant 2019; 34 :561–6. https://doi.org/10.1093/ndt/gfy128
30. Hesterberg TC. What teachers should know about the
bootstrap: resampling in the undergraduate statistics cur-
riculum. Am Stat 2015; 69 :371–86. https://doi.org/10.1080/
00031305.2015.1089789
31. Le Henanff A, Giraudeau B, Baron G et al. Quality of reporting
of noninferiority and equivalence randomized trials. JAMA
2006; 295 :1147–51. https://doi.org/10.1001/jama.295.10.1147
32. Mayne TJ, Nordyke RJ, Schold JD et al. Dening a min-
imal clinically meaningful difference in 12-month es-
timated glomerular ltration rate for clinical trials in
deceased donor kidney transplantation. Clin Transplant
2021; 35 :e14326. https://doi.org/10.1111/ctr.14326
33. Cuzick J, Sasieni P. Interpreting the results of noninferiority
trials—a review. Br J Cancer 2022; 127 :1755–9. https://doi.org/
10.1038/s41416-022-01937-w.
34. Janmaat CJ, van Diepen M, van Hagen CC et al. Decline of
kidney function during the pre-dialysis period in chronic
kidney disease patients: a systematic review and meta-
analysis. Clin Epidemiol 2018; 10 :613–22. https://doi.org/10.
2147/CLEP.S153367
35. Seppen B, Wiegel J, Te r Wee MM et al. Smartphone-assisted
patient-initiated care versus usual care in patients with
rheumatoid arthritis and low disease activity: a randomized
controlled trial. Arthritis Rheumatol 2022; 74 :1737–45. https:
//doi.org/10.1002/art.42292
36. Appel CW, Pedersen SC, Nielsen AS et al. Telemedicine based
on patient-reported outcomes in management of patients
with inammatory bowel disease in a real-life setting—a be-
fore and after cohort study. Scand J Gastroenterol 2022; 57 :825–
31. https://doi.org/10.1080/00365521.2022.2041083
37. de Jong MJ, van der Meulen-de Jong AE, Romberg-Camps
MJ et al. Telemedicine for management of inammatory
bowel disease ( myIBDcoach) : a pragmatic, multicentre, ran-
domised controlled trial. Lancet 2017; 390 :959–68. https://doi.
org/10.1016/S0140-6736(17)31327-2
38. Laurberg T, Schougaard LMV, Hjollund NHI et al. Random-
ized controlled study to evaluate the impact of exible
patient-controlled visits in people with type 1 diabetes: the
DiabetesFlex Trial. Diabet Med 2022; 39 :e14791. https://doi.
org/10.1111/dme.14791
39. Riis CL, Jensen PT, Bechmann T et al. Satisfaction with care
and adherence to treatment when using patient reported
outcomes to individualize follow-up care for women with
early breast cancer –a pilot randomized controlled trial.
Acta Oncol 2020; 59 :444–52. https://doi.org/10.1080/0284186X.
2020.1717604
40. Grove BE, de Thurah A, Ivarsen P et al. Remote symp-
tom monitoring using patient-reported outcomes in pa-
tients with chronic kidney disease: process evaluation of
a randomized controlled trial. JMIR Form Res 2024; 8 :e48173.
https://doi.org/10.2196/48173
41. Jensen MB, Jensen CE, Gudex C et al. Danish population
health measured by the EQ-5D-5L. Scand J Public Health
2023; 51 :241–9. https://doi.org/10.1177/14034948211058060

Remote symptom monitoring in outpatients with CKD 11
42. Schwartz MN, Rimland CA, Quinn KA et al. Utility of the brief
illness perception questionnaire to monitor patient beliefs
in systemic vasculitis. J Rheumatol 2020; 47 :1785–92. https://
doi.org/10.3899/jrheum.190828
43. Grove BE, Val en Schougaard LM, Ivarsen P et al. Remote
follow-up based on patient-reported outcomes in patients
with chronic kidney disease: a qualitative study of patient
perspectives. PLoS One 2023; 18 :e0281393. https://doi.org/10.
1371/journal.pone.0281393
44. Wiegel J, Seppen B, van der Leeden M et al. Adherence to tele-
monitoring by electronic patient-reported outcome mea-
sures in patients with chronic diseases: a systematic review.
Int J Environ Res Public Health 2021; 18 :10161. https://doi.org/
10.3390/ijerph181910161
45. Sterne JAC, Savovic J, Page MJ et al. RoB 2: a revised tool for as-
sessing risk of bias in randomised trials. BMJ 2019; 366 :l4898.
https://doi.org/10.1136/bmj.l4898
Received: 27.9.2023; Editorial decision: 7.5.2024
©The Author( s) 2024. Published by Oxford University Press on behalf of the ERA. This is an Open Access article distributed under the terms of the Creative
Commons Attribution-NonCommercial License ( https://creativecommons.org/licenses/by-nc/4.0/) , which permits non-commercial re-use, distribution,
and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com