Adherence to therapeutic splint wear in adults with acute upper limb injuries: a systematic review
Non-adherence with therapeutic splinting in acute hand injury can reduce treatment beneﬁts, increase risk of disability and bias assessment of treatment efﬁcacy. This systematic review aims to critically analyze the literature on splinting of acute upper limb injuries to identify key factors that could inﬂuence patient adherence with splint wear.
Trials were identiﬁed from searches of EMBASE, MEDLINE, CINAHL (to June 2009) and reference lists of articles and relevant reviews. Search terms used were patient compliance/adherence behavior, splint/s, orthosis/es and brace. Where possible, randomized controlled trials or prospective cohort studies were sought, and then cross-sectional and retrospective studies if the former were not available. Studies speciﬁcally addressing chronic conditions were excluded. All relevant trials were assessed for methodological quality by the author using explicit criteria. Data were extracted using a standardized form designed by the author.
Six studies (one randomized controlled trial, two cross-sectional analytic surveys and three retrospective ﬁle reviews) involving 490 people were included. Owing to the heterogeneity of studies synthesis is narrative rather than quantitative. There was no consistent correlation between adherence and age or gender. One study found a correlation with patient perception of positive effect, and one found negative correlations with agitation and brain injury severity.
Studies found were generally of varied quality and may be susceptible to bias. This is a ﬁeld with little published scientiﬁc evidence, and future research should measure adherence relationships with socioeconomic, health-care system, therapy- and patient-related characteristics.
The terms ‘adherence’ and ‘compliance’ are often used interchangeably in medical and therapeutic literature, but have different connotations and inferences. A detailed description of the differences is discussed by Meichenbaum and Turk1 and can be summarized as follows:
Compliance is the extent to which patients obey and follow instructions, prescriptions and proscriptions outlined by their treating health practitioner;
Adherence implies an ‘active, voluntary and collaborative involvement by the patient in a mutually acceptable course of behavior to produce a preventative or therapeutic result’.(p. 20)1
The term ‘adherence’ is intended to be non-judgmental, and does not imply blame on the part of the patient, health-care practitioner or treatment.2
Hand Therapy Vol. 15 No. 1 March 2010
Adherence is the ‘most unpredictable, least controllable variable in medical intervention [and] can strongly sway the outcome of any treatment’3 (p. 31).
It is well accepted that consumer non-adherence with medical or therapeutic treatment can reduce treatment beneﬁts, affect recovery, increase the risk of disability and bias assessment of treatment efﬁcacy.2,4
The literature on the determinants of adherence with health-care interventions is extensive. The multi-dimensional adherence model (MAM) published by the World Health Organization in 2003 following a major critical review of the evidence identiﬁes ﬁve dimensions that inﬂuence adherence.5 The ﬁve dimensions of the MAM are (1) socioeconomic, (2) health-care system-related, (3) condition-related, (4) treatment-related and (5) patient-related factors (Figure 1). It differs from many preceding models by removing the focus on patient-related factors, stating that it is a ‘misconception that adherence is a patient-driven problem’ (p. 26). This model contends that ability to follow treatment is impacted by more than one factor, and interventions to improve adherence need to address all relevant factors. Although this model was designed for long-term therapies, its key concepts apply equally to acute conditions and it provides clinically useful guidelines (referred to as ‘lessons learned’) that are readily applied to the ﬁeld of hand therapy.
Practitioners (for example, occupational therapists, physiotherapists and orthotists) who prescribe and ﬁt splints, braces or orthotics are particularly reliant on patient adherence in achieving desired outcomes after acute musculoskeletal injuries. In their case, adherence covers a variety of behaviors including:
The ﬁve dimensions of adherence. Source: World Health Organization (WHO), 2003 (reproduced with permission from WHO, Geneva)
(1) entering and continuing a therapy programmes
(2) attending assessment and follow-up appointments
(3) correct wear of prescribed and ﬁtted splints, braces or orthoses
(4) correct performance of home-based therapy programmers (which may include exercise, rest, edema management strategies)
(5) avoidance of risk behaviors (e.g. overuse of the injured limb during recovery stages).
This review is focused on point 3 above, as there is evidence to support that adherence to prescribed splint wear in acute conditions results in superior recovery and prevention of deformities in tendon,6–10 nerve11–13 and bony injury,14–17 and prevention of contractures post burns.18,19
Non-adherence with splinting in acute injury can result in increased health system costs by increasing the need for difﬁcult secondary surgical procedures,14,20 medical, nursing and allied health support.1,21,22 Although there are several systematic reviews of com-pliancy/adherence studies in chronic conditions, including examining splint wear in patients with rheumatoid arthritis,23,24 there are no systematic reviews addressing splint adherence in acute injuries.
The objective of this review is to identify key factors that could inﬂuence patient adherence with splint wear in acute upper limb injury. Factors will be grouped and dis-cussed using the MAM. Outcomes other than splint adherence were not examined as part of this review.
Relevant articles were identiﬁed from a search of Ovid MEDLINE (1970 to June 2009), Ovid CINAHL (1970 to June 2009) and EMBASE (1970 to June 2009) – encom-passing Cochrane database of systematic reviews and conference papers.
The following search strategy was used by the author to search CINAHL and was modiﬁed as necessary for MEDLINE and EMBASE:
(1) compliance.mp. (mp ¼ title, subject heading word, abstract, instrumentation)
(2) patient compliance/or patient adherence or
‘ADHERENCE BEHAVIOR (IOWA NOC)’/
(3) splint*.mp. or exp Splints/
(4) orthosis.mp. or exp Orthoses/
(6) hand*mp or wrist* or ﬁnger* or thumb* or elbow*mp
(7) 1 or 2
(8) 3 or 4 or 5
(9) 6 and 7 and 8.
The search was restricted by age (adults: aged 18 years plus) and date (1970 to June 2009), and publication in English. The reference lists of relevant review articles and all included studies were examined to identify further studies.
Types of studies
Where possible, randomized, quasi-randomized or clini-cally controlled trials or prospective cohort studies were sought, and then cross-sectional and retrospective studies if the former were not available.
Types of participants
People aged 18 years and over with acute (i.e. ,3 months post injury) bone, tendon or nerve injury, including symptoms of newly diagnosed nerve compression injury, of the hand, wrist or forearm.
Types of intervention
All studies that measured adherence with splint or orthosis wear were included. Co-interventions (such as exercise programes) were allowed.
Studies speciﬁcally addressing chronic conditions (such as rheumatoid or osteoarthritis, osteoporosis or chronic tendonitis) were excluded. Case studies were also excluded, as were studies that did not examine reasons for non-adherence. A list of excluded studies is available on request.
Methods of review
After excluding all articles not relating to upper limb injuries in adults, 38 articles were identiﬁed by the search strategy, and the title and abstracts of potentially relevant papers were reviewed by the author.
Full text of articles that speciﬁcally addressed factors associated with non-adherence with splinting or bracing for acute injuries (i.e. bone, tendon or nerve injury of ,3 months since onset) were obtained by the author, and were then evaluated for quality using explicit criteria.
For all included studies, relevant data were extracted by the author into standardized forms to capture the fol-lowing information: (1) study design, (2) number of sub-jects and their associated injury/condition, (3) adherence measure used, (4) presence of explicit deﬁnitions for adherence, (5) whether a statistical comparison was done between adherent and non-adherent groups and (6) study results.
Owing to the heterogeneity of studies (differing adherence measures and variables used in analysis), synthesis is narrative rather than quantitative. Statistics are therefore reported as published in the original studies. When reported, the 95% CI is used; if it was not available, a
P value is reported.
Description of included studies
Six studies involving 490 people and published over a 19-year period (1987–2008) were deemed suitable for inclusion in this review (see Figure 2 for ﬂow diagram of inclusion/exclusion of studies). All papers addressed upper limb splinting. One was a quasi-randomized con-trolled trial (with participants allocated to groups depending on the last digit of their Social Security number), two were cross-sectional analytic surveys and three were retrospective ﬁle reviews. Settings included the USA,7,14,25 Europe,13 UK26 and Australia.27
Most studies involved people with acute hand injuries or nerve compressions who were living at home and
Hand Therapy Vol. 15 No. 1 March 2010
returning to normal daily living and work; however, data for one study27 was drawn from inpatients with coexisting acute brain injury. Duration of observation varied from 15.4 days (average length of hospital stay in O’Brien and Bailey27) to 12 months13 with most being between four and nine weeks. One study did not indicate the duration of observation.14
The mean age for most participant groups was in the ﬁfth decade, with one study conducted in a veteran’s center having an older group (mean age 60 years),25 another with a mean age of 30 years26 and one failing to describe the total group adequately.14 Interestingly, most participants were men (234 in total compared with 56 women, or 80% of sample). While this generally reﬂects the gender split in acute trauma-related hand injuries
(68% men in reference28), the study examining adherence in people with carpal tunnel syndrome had a men:women ratio of 16:1, which conﬂicts with the usual gender pattern of 1:329 suggesting an unrepresentative sample.
All had reasonably clear criteria for determining adherence with splinting, with half electing a dichotomous (i.e. yes/no) measure and half including grades of adherence (complete, partial, non-adherence).
While most studies conducted a comparative analysis of factors associated with adherent and non-adherent groups, two14,25 did not, and one made only passing reference to factors found to be statistically unrelated.7 Only two studies used multivariate analysis to examine relationships between adherence and other variables.13,27
Table 1 summarizes the included studies as well as their ﬁndings.
Studies were assessed for quality using the Critical Review Form for Quantitative studies published by McMaster University, Canada.30 This tool was used as it allowed for the differing methodological designs of studies included in this review and it has been used in several allied health evidence reviews.31–33 Studies were evaluated using the form, which incorporated the headings shown in Table 2.
Each included study was assessed on whether it met the requirements listed under each criterion. Those requirements that were met were rated as ‘yes’ and awarded one point. If the requirement was inadequately addressed or completely overlooked, it was given a rating of ‘no’ and received no points. For headings that were inappropriate for a particular research design and did not reﬂect the quality of the article, the ‘not applicable (n/a)’ option was checked.
The total number of points received by a study (out of a maximum of 15) was then calculated and is represented in Figure 3.
Most studies found high adherence rates (i.e. 75% or more; mean ¼ 85.17%) with splint wear instructions, with the obvious exception of the study focusing on those with concurrent acute brain injury,27 which found an adherence rate of 60.5%. Sandford et al.26 initially reported a very low adherence rate (33%) but this reﬂected a very stringent deﬁnition (any instance of splint removal over the ﬁrst 4 weeks counted as non-adherence, whether it be for hygiene, to get dressed or because of discomfort). This was acknowledged in their paper, and using Groth’s7 deﬁnition of ‘secondary compliance’, they adjusted the ﬁgure to 83%.
Factors associated with adherence
Results are presented using the ﬁve dimensions of the MAM and are summarized in Table 3.
Social and economic
While Hall14 concluded that age ,27 years, male gender, unemployment, injury sustained in a ﬁght and alcohol consumption at time of injury were signiﬁcantly associ-ated with non-adherence with splinting in acute hand fractures, no statistics were presented, and there appears to be no between-group comparison for adherent versus non-adherent groups. Numbers were also low in the non-adherent group (only 12 out of 200 participants) so results should be interpreted with caution. Sandford
et al.26 found a signiﬁcant correlation for male gender, but results are possibly explained by low numbers of women enrolled in the study (11% of sample).
O’Brien and Bailey27 found no signiﬁcant correlation for age (OR 1.0 [0.98, 1.03], P ¼ 0.95), gender (OR 0.62 [0.14, 2.62], P ¼ 0.51), occupation type (OR 0.87 [0.63, 1.21], P ¼ 0.41) or race/cultural factors (OR 0.99 [0.31, 3.14], P ¼ 0.99). Groth et al.7 and Sandford et al.26 also found no correlation for age, but neither published their statistical analyses.
Health care and system
This factor was not examined by any of the studies included in this review.
Sandford et al.26 found no correlation with adherence and injury type (ﬂexor versus extensor tendon) or dominant versus non-dominant-hand injuries. Paternonstro-Sluga et al.13 completed a logistic regression and also found no signiﬁcant relationship between adherence and diagnosis (type of peripheral nerve injury) or hand dominance. Neither study published speciﬁc statistics for non-signiﬁcant results. No other studies explored the relationship between injury type and severity with adherence.
One study found evidence of a strong association between patient perception of positive effect and adherence with a day-time functional splint25 (OR 54.1 [2.106, 1]), but this factor was not measured in any other included studies. Sandford et al.26 found splint discomfort to be the fourth most common reason for splint removal, and Walker25 noted that one participant did not wear their splint due to interference with work, but neither paper reported speciﬁc statistics for these ﬁndings.
This dimension includes such factors as physical sensory or cognitive impairment, psychological factors, such as low motivation, lack of understanding of the condition (and/or the need for treatment), beliefs about side-effects, stress and negative views about medicine. Only one study27 collected data on patient factors, ﬁnding that duration of post-traumatic amnesia (an index of brain injury severity) (OR 0.94 [0.89, 0.99], P ¼ 0.04) and the presence (OR 0.15 [0.05, 0.44], P ¼ 0.001) and duration of agitation (OR 0.73 [0.60, 0.90], P ¼ 0.003) were signiﬁ-cantly associated with non-adherence.
Although the literature shows that social and economic factors in general are not consistently associated with adherence rates, some such as race (and cultural beliefs), poverty, illiteracy, unemployment, lack of social supports, distance from treatment outlet, family dysfunction, cost of travel/treatment and age have shown relationship with adherence to long-term therapies.5 While age was found to be a signiﬁcant factor in one study in this review,14 it was considered to be of poor quality and highly susceptible to bias. Three other studies found no relationship with age7,26,27 suggesting that this is unlikely to be a predictive factor.
In this review, only one study27 included education and occupation type in their analysis, but found no signiﬁcant correlation. The literature on the impact of the person’s level of education on therapy adherence in general is equivocal; Groth et al.34 stated that more literate people were more likely to understand their condition and hence comply with treatment,34 whereas Sluijs et al.35 found highly educated patients to be more likely to be non-compliant with home exercise programmes, but did not speculate on why this was so.
One study in this review14 found a link between unemployment and non-adherence but, as stated pre-viously, concerns about quality make it a poor source of evidence.
No studies in this review examined factors related to the health-care team and system. This may be an important gap, as there is some evidence that a good patient–provi-der relationship can improve splint adherence in chronic conditions,24,36 and positive feedback from the therapist can improve exercise adherence in acute conditions.35 There is relatively little research into factors that can have a negative effect, such as poorly developed services, overworked and poorly trained health-care providers and lack of continuity of care.5
Condition-related factors include severity of symptoms, level of disability, prognosis, rate of progression and the availability of effective treatment. This review found similar adherence rates in patients with acute tendon and bony injury. This is not surprising given that these con-ditions will result in similar levels of short-term disability and overall would have the expectation of a good return to function. Interestingly, the study that focused on nerve repairs (which have a longer recovery time and a lower likelihood of a full recovery) also had a similar adherence rate.
The treatment-related dimension includes complexity, duration, immediacy of beneﬁt, interference with life-style, side-effects and frequent changes to treatment pathways. It also encompasses the availability of support to deal with the above factors.
One study13 found that immediate beneﬁt from wearing the splint was the only factor signiﬁcantly associated with splint adherence, concluding that this highlights the need for good patient education: ‘the better an individual is informed of the potential positive effect, the better it will be realized’ (p. 93).
Ensuring splints are comfortable and aesthetically acceptable to the patient is also a key issue, but was examined by only one of the included studies26 who found ‘discomfort’ was one of the four most common reasons for splint removal. Previous research has found that splint comfort37,38 and the visual appearance (and visibility to others) of the splint is important to the wearer and can inﬂuence adherence.19,39,40 For example, a modiﬁed splint for axilla burns in Indian population claimed that it had greater patient acceptance due to
‘aesthetic appeal over the currently available aeroplane splints, as this could be worn comfortably within one’s garment’19 (p. 502).
The patient-related dimension includes physical, sensory and psychological aspects of the patient. In this review, one study of people postbrain injury27 showed lower rates of adherence with splinting than all other studies (60.5%compared with overall mean of 85.17%). Even in the absence of an identiﬁed cognitive impairment, however, it is important to recognize the patient’s ability to under-stand and remember hand therapy instructions.41–43 One study (not included in this review) followed 28 unimpaired patients postﬂexor–tendon repair42 and found that only 42.5% recalled instructions (including ‘do not remove your splint’) without the need for a cue. Another author (an experienced hand therapy practitioner) recommends the use of the mini-mental status examination with elderly clients to determine if memory problems exist so that the therapeutic approach can be amended as necessary.41
Patient beliefs and attitudes about their condition
(particularly their own power to inﬂuence the outcome) have been found to have an effect on adherence in chronic conditions,35,44 but were not examined by any studies in this review.
Prevalence of non-adherence
Most estimates of non-adherence with medical or thera-peutic treatment range from 30% to 60%45 with a high degree of variability depending on the type of treatment. Non-adherence is reportedly rare in treatments for acute-onset conditions requiring direct medication, high supervision and monitoring (e.g. chemotherapy for cancer)1 and higher in chronic disorders where there is little discomfort or perceived risk from the disorder, and where lifestyle changes are required.
Acute versus chronic condition
This review found higher overall rates of splint adherence in acute injuries (75%) than the comparable literature for chronic conditions such as rheumatoid arthritis (rates of 25–65%),23,24 which parallels the evidence on adher-ence with prescription medicine in acute versus chronic illness.1,5 The difference may be partly explained by the
fact that splinting for chronic conditions is mainly for palliative purposes, so there are no immediate perceived dangers associated with non-adherence as there are in acute injuries.
This is also consistent with ﬁndings in adherence studies examining other aspects of therapy. For example, a study examining a multidisciplinary pain management programme for people with chronic pain found that only 12% adhered to the full programme.46 Chronic con-ditions may also be often associated with co-morbidities such as depression and substance abuse, and these play an important role in modifying the individual’s ability to adhere to treatment.
Limitations of this review
This review was completed by a sole author, so there is the possibility of bias in the inclusion and exclusion of studies, and in the quality ratings given to studies. In addition, grey literature sources such as registers of clinical trials and dissertations were not searched. Finally, the included studies did not use comparable adherence measures or variables, limiting the possibilities for pooling data.
Methodological barriers to the study of adherence
Published prevalence ﬁgures must be interpreted with caution as adherence is a construct that is difﬁcult to measure for several reasons. Firstly, many studies of adherence are reliant on the patient’s self-report and patients may be unwilling to admit non-adherence.1 There may be selection bias in many studies of adherence, as respondents are by deﬁnition compliant with requests for information and may be unrepresentative of the typical patient population.
There is also the potential for performance bias; patients’ behaviour may change if they know their adherence is being monitored. One way researchers are attempting to limit the reliance on patient report is to embed sensors in the splint or brace that can accurately calculate hours of wear and/or exercises per-formed while wearing the splint6,47 although this method is still rarely used in hand splinting.
Secondly, adherence is conceptualized, deﬁned and measured differently by researchers. Adherence tends to be treated as a dichotomous variable when there are, in effect, varying levels of non-adherence which may span from (a) never-adhered to any aspect of treatment; (b) adhered to some but not to other aspects; to (c) initially adhered but relapsed over time. One deﬁnition of non-adherence that may be clinically useful is ‘The point at which the desired preventive or desired therapeutic result is unlikely to be achieved’ (Gordis, 1976 in reference1,p.31).
Finally, adherence can be context-dependent. A person may manage well when surrounded by cues and remin-ders (for example, during their inpatient hospital stay) but may lose motivation or forget to adhere to their therapy programme when they return home.
It is established that poor adherence to splinting leads to worse outcomes for the patient and increasing costs to the
Hand Therapy Vol. 15 No. 1 March 2010
health-care system.2,4 We also know that adherence is an important modiﬁer of treatment effectiveness.5
Implications for research
Many studies of splinting adherence in people with mus-culoskeletal injuries have a number of limitations that reduce the usefulness of their ﬁndings. These limitations include the failure to use multivariate analytic methods to study factors associated with adherence and the failure to use a theoretical model to select the variables measured. Future research should be designed according to an estab-lished and validated adherence model and analysed using multivariate analysis.
To be of best value to practitioners, it is recommended that speciﬁc data be collected on socioeconomic variables (see Table 3), distance from treatment centre, length of follow-up, continuity of care (e.g. did the same therapist provide treatment or were there multiple therapists involved?) and patient ratings of complexity of treatment regimen, patient–therapist relationship and interference with lifestyle/activities of daily living/work.
Measures of adherence should include length of time the splint was worn (as a percentage of recommended wear time) as well as number of therapy sessions attended (as percentage of number scheduled). Options for recording splint wear time objectively include embedded sensors. Where this is impractical, separate splint wear diaries completed by the patient and their partner/carer may yield a more accurate measure of splint adherence.
Implications for practice
This review found no consistent relationship between splint adherence and socioeconomic and condition-related factors, suggesting that there is little to be gained from adapting treatment based on these variables in isolation. It is vital that patients are supported throughout their therapy, and not blamed for ﬂuctuating or poor adherence.5
There was some evidence that treatment
therapy-related factors such as immediacy of beneﬁt, splint comfort, and minimizing interference with lifestyle and daily living activities can improve splint adherence.
Acknowledgements: Thanks to Ted Brown and Louise Farnworth, Monash University who are supervising my PhD.
Competing interests: None declared.
Accepted: 7 July 2009