The impact of self-monitoring in chronic illness on healthcare utilisation: a systematic review of reviews

A total of 2114 references were retrieved. After exclusions based on title alone 320
full articles were retrieved and after screening 17 articles reporting 16 different
systematic reviews or meta-analyses were selected for possible inclusion (Fig. 1). A list of excluded reviews can be found in Additional file 3. One review had been published twice, as a Cochrane review and again as peer reviewed
journal article 28], 29]. Both containing the same data, only the journal article was included 28]. One additional review 30] was identified as a result of reference list searches, resulting in a total of 17
reviews in this overview. Of the 320 full text articles reviewed by two authors, there
was disagreement on 17 (5.31 %) of these, discussion between the two reviewers resolved
14 of these and 3 were taken to the third reviewer for discussion.

Fig. 1. PRISMA Flowchart

Review characteristics

The characteristics of the included reviews can be found in a table in Additional
file 4. Table 1 provides a summary of this data, with the most frequently evaluated interventions,
measures of healthcare utilization and monitored data. The reviews synthesised intervention
effectiveness across three chronic conditions; hypertension (n?=?2), chronic obstructive pulmonary disease (COPD) (n?=?2) and heart failure (n?=?13). The overall CCA across all five long-term conditions was 4.10 %, which represented
slight overlap 23]. There was however, significant variation between long-term conditions (Table 1). Eight reviews included only randomised controlled trials (RCT), the remaining reviews
included a combination of study designs.

Table 1. Summary of included systematic reviews and meta-analyses

Intervention characteristics

Electronic Additional file 5 is a table with detailed information about the characteristics of the interventions.
In COPD the interventions were action planning and telehealthcare and in both hypertension
and heart failure telemonitoring. Self-monitoring was facilitated through the use
of technology in 15 reviews, in which patients took measurements and then transmitted
data to a healthcare professional for interpretation and adjustment. Assessing the
interventions according to the three components of self-monitoring: awareness, interpretation
and response 6], 7]. In only two reviews, one of self-management in heart failure 31] and the other action planning in COPD 32], did interventions consist of all three components. In all other reviews only awareness
was achieved.

Control conditions

On the whole the articles that included studies with a control group generally provided
a poor description of the content. Two articles failed to detail the content of any
control groups 33], 34]. For a majority there was no consistency in what the intervention group was compared
to and for others they were compared to a mixture of usual care and/or an active control
group. In most cases however, the definition of usual or standard care was not described
or where it was it was not consistent across primary research studies.

Assessment of review quality

The methodological quality of the 17 reviews varied (Fig. 2), but was generally good (median score?=?5). The most common methodological problems
were failure to report conflicts of interest, lack of integration of study quality
into the conclusions of the review, exclusion of studies based on their publication
status, not providing a list of included or excluded studies and not assessing the
likelihood of publication bias.

Fig. 2. Distribution plot of the quality of review articles

Intervention effectiveness

Whilst the reviews in heart failure and COPD synthesised the evidence in relation
to a range of healthcare utilisation outcomes, the two reviews in hypertension reported
the effects in relation to GP attendance only.

Hospitalisation

Thirteen of the 17 reviews reported the effects of either telehealth (n?=?12) or action planning (n?=?1) on disease-specific and/or all-cause hospitalisation, 11 in heart failure and
2 in COPD.

The one article on action planning was a low quality meta-analysis, in which patients
with COPD were actively involved in adjusting their treatment or seeking medical advice
in response to their symptoms. This meta-analysis failed to find any significant effect
on rates of hospitalisation (n?=?2, weighted mean difference (WMD)?=?0.16, 95 % Confidence Interval (CI) -0.09–0.42)
32] and all of the primary research studies in this review had methodological limitations,
increasing the risk of bias.

The reviews of telehealth were more positive. Seven meta-analyses evaluated the evidence
with regards to telemonitoring and/or structured telephone support in either COPD
or heart failure. The use of technology to support self-monitoring was associated
with up to 27 % reduction in total all-cause or disease-specific hospitalisations
compared to control conditions (see Table 2 for specific results) 28], 35]–40]. All these reviews were rated of moderate or high quality. The quality of the primary
research studies within them however, ranged from introducing a low risk of bias 35], 36] to introducing a significant amount of methodological bias 29], 38], 40]. Whilst human-to-human structured telephone support led to a significant reduction
in heart failure-related hospitalisations, the one review on human-to-machine structured
telephone support failed to have any effect 35]. In contrast to what might be expected, subgroup analyses within this moderate quality
review indicated that telemonitoring with medical support available only during office
hours was associated with a greater reduction in hospitalisations than when medical
support was available 24/7 35].

Table 2. Results of the meta-analyses in relation to hospitalisation for technology enabled
self-monitoring

Five further systematic reviews, of either low or moderate quality, concluded that
there was a positive trend towards a reduction in all-cause and disease-specific hospitalisation
in favour of telemonitoring for patients with heart failure 30], 33], 41]–43]. Only two of these systematic reviews rated the quality of the primary research studies,
both suggested that the studies were of good quality 30], 33].

Readmissions

Rates of readmission in heart failure were reported in one meta-analysis and three
systematic reviews. The meta-analysis, of moderate quality, found that self-management
in which patients were taught to seek medical assistance in response to their symptoms,
reduced the odds of all-cause and disease-specific readmission by up to 54 % (n?=?5, Odds Ratio (OR)?=?0.59, 95 % CI 0.44–0.80; n?=?3, OR?=?0.44, 95 % CI 0.27–0.71; respectively) 31]. The quality of the primary research studies included in this review however, varied
significantly. Conclusions of the three systematic reviews, which were either of low
or moderate quality, indicated an association between telemonitoring and fewer readmissions
to hospital 42]–44]. As result of the quality of the primary research studies in these systematic reviews
the authors concluded that further, more methodological robust trials were needed
before widespread adoption of telemonitoring should take place.

Length of hospital stay

The conclusions drawn by the authors of eight systematic reviews in heart failure,
in relation to the number of days patients spent in hospital, were mixed. Telemonitoring
was associated with a reduction in the length of hospital stay in three low to moderate
quality systematic reviews; both within the intervention group over time and also
when compared to a control group 42]–44]. Only one review 44] rated the strength of the evidence, which was considered to be very heterogeneous.
The five remaining moderate to high quality systematic reviews failed to find an association
between telemonitoring and time spent in hospital 30], 35]–37], 40]. All reviews judged the primary research studies to be of least fair quality with
a low risk of bias. Except one, which failed to rate quality of the primary research
studies 37].

Accident and emergency (A E) attendance

Seven reviews; three meta-analyses and four systematic reviews synthesised the evidence
in relation to A E attendance. Action planning in COPD, which involved patients
measuring, interpreting and responding to their data, was not found to have any significant
effect on visits to A E in one meta-analysis (n?=?2, WMD?=??0.01, 95 % CI ?0.12–0.10, p?=?0.85) 32]. However, the primary research studies in this review did include some risk of bias
and the review itself was of low quality.

There was a mixed picture in regards to telehealth in heart failure, further muddied
by the poor quality of these reviews. Whilst a meta-analysis of telemonitoring in
heart failure failed to find any effects on A E attendance (n?=?4, Risk Ratio (RR)?=?1.04, 95 % CI 0.86–1.26, p?=?0.67) 37], the systematic reviews in heart failure 30], 40], 42], 43] concluded that telehealth was associated with fewer all-cause and disease-specific
A E visits. Within these reviews the strength of the evidence was either rated as
fair 30], 40] or was not rated at all 37], 42], 43]. More promisingly, a moderate quality meta-analysis of telehealth in COPD 28] found that odds of attending the A E department were significantly reduced in the
telehealthcare compared to control group (n?=?3, OR?=?0.27, 95 % CI 0.11–0.66, p?=?0.005), however, a majority of the primary research studies in the review included
significant risk of bias.

Outpatient visits

One systematic review, of moderate quality, assessed the impact of telemonitoring
in heart failure 40] and concluded that home telemonitoring was associated with increased visits to specialist
outpatient services.

GP visits

The impact of self-monitoring on the frequency of GP visits was reported in one meta-analysis
and two systematic reviews, all rated low quality. The meta-analysis in COPD found
no significant difference between action planning and usual care in scheduled (n?=?1, mean difference (MD)?=??0.50, 95 % CI ?4.06–3.06, p?=?0.78) or unscheduled GP visits (n?=?1, MD?=??0.20, 95 % CI ?1.55–1.15, p?=?0.77) 32]. The primary research studies in this review also had a number of methodological
limitations. The two systematic reviews in hypertension, one of home-based blood pressure
monitoring 34] and the other of telemonitoring 45], both found no impact on GP visits. Neither of these systematic reviews assessed
the quality of the primary research studies.

Home visits

Two systematic reviews in heart failure reported weak and inconsistent effects for
telemonitoring on the frequency of home visits 40], 43]. Although one of these reviews suggested a reduction in home visits, there was no
quality assessment of the primary research studies and the review itself was of poor
quality 43]. The other systematic review, of moderate quality, concluded that home telemonitoring
was associated with a greater number of home care visits 40].