Direct discharge of patients with simple stable musculoskeletal injuries as an alternative to routine follow-up: a systematic review of the current literature

Traditionally, all patients with musculoskeletal injuries are referred to a fracture clinic for further review and treatment, after initial assessment in an Emergency Department (ED). Consequently, fracture clinics are often characterized by the referral of large numbers of unselected patients, many of whom have minor injuries that do not require intervention. This leads to long waiting times, recurrent unnecessary reviews, and a high workload that inevitably has consequences for patient experience, staff morale, training, and quality of care [1].

A Virtual Fracture Clinic (VFC) model has been introduced in several hospitals worldwide as an alternative model of fracture care to regulate access to fracture clinics [2, 3]. This model is increasingly used in the United Kingdom (UK), but also in the Netherlands, Norway, Australia and New Zealand [3]. The VFC model comprises two main components [2].

First, direct discharge (DD) of patients with relatively simple stable musculoskeletal injuries. This means patients are discharged without subsequent review or repeated imaging, supported by self-removable orthoses, discharge leaflets, and a telephone helpline. Second, the establishment of an individualised management plan for all other patients during a daily consultant-led VFC review. This process should further streamline outpatient care and ensure that each patient is seen at the right time by the most appropriate person [2].

The DD protocols were developed based on studies showing that for several minor self-limiting injuries, casting offers no benefit over functional treatment [2], and on the assumption that patients with these injuries require reassurance and information, but do not need to attend a fracture clinic routinely. However, for DD to be a useful and acceptable alternative to routine follow-up, patient outcome, patient experience and complication rates should at least remain comparable, while healthcare utilizations and consequently costs are ideally reduced. Despite several independent studies that were conducted since DD was first established in 2011 [2], an overview of all current evidence regarding DD is not available.

The objective of this study was to systematically review the literature on the logistic and financial benefits of DD of patients with simple stable musculoskeletal injuries, as well as their patient-reported outcome, experience, and adverse outcomes. To illustrate the possible treatment of a patient with a simple stable musculoskeletal injury, both before and after DD, two case samples are provided below.

This systematic review was planned, conducted, and reported using PRISMA guidelines [4]. A study protocol was registered with the PROSPERO register prior its commencement (registration number: 120779) [5].

Both comparative (i.e., routine care before DD protocols were adopted compared to DD) and non-comparative studies (i.e., a DD cohort only) were considered if they featured DD of one or more musculoskeletal injuries in an adult, paediatric, or mixed population. Case reports and abstracts were excluded. There were no restrictions regarding the timing of the study, nor the duration of follow-up. Only articles reported in English were included.

Direct discharge was defined as scheduling no routine follow-up appointment after the ED visit. This could either take place directly after the ED visit (ED DD) or after a daily ‘virtual’ review (VFC DD, Fig. 1). A single fracture clinic visit, shortly after attending the ED, was interpreted as ED DD if the sole purpose of this visit was application of a removable splint/orthosis. Studies were excluded if any further information or assessments were part of this visit. Studies were also excluded if reporting the potential effects of DD, without actually discharging patients directly. Both prospective and retrospective studies were included if reporting one of the following outcomes:

Logistic outcomes included: (1) proportion of patients discharged directly, (2) number of follow-up appointments, and (3) number of repeat radiographs during follow-up. If the number of follow-up appointments was not reported, a study was also included if it reported a return rate instead (i.e., the proportion of patients that returned to the hospital despite being discharged directly).

Financial outcomes included any report on costs, for instance healthcare costs, societal costs etc.

Patient outcomes included any patient-reported experience/outcome measures (PREMs/PROMs) and adverse outcomes.

Medline, Embase, CINAHL, Cochrane Library and Web of Science databases were searched from inception to 15 January 2020. A limited update literature search was performed on 6 August 2020. Reference lists of included studies were scanned to ensure literature saturation.

The electronic search strategy was developed by a health librarian and peer-reviewed by another librarian. Medical subject headings (MeSH) were used in Medline and complemented by text words related to DD and (virtual) fracture clinic redesign. This search strategy was then translated for the other databases. The PICO strategy and the full electronic search of Medline are reported in Tables S1 and Table S2, respectively.

Two review authors (THG and JV) independently screened the titles and abstracts yielded by the search against the inclusion criteria. Full texts were obtained if studies appeared to meet the inclusion criteria, or in case of uncertainty. All reasons for exclusion were recorded. Reviewing authors were not blinded to the journal titles, study authors and institutions.

A data extraction sheet was developed and pilot-tested on five randomly selected included studies, and refined accordingly. Data from the included studies were extracted by one author (THG) and checked by another (JV).

The following data items were collected for all included studies: country; year; design; sample size; injury/injuries studied; eligibility criteria for DD; population (adult/paediatric); study period. If reported, the distribution of patients across the various parts of the VFC model (i.e., ED DD; VFC DD or VFC follow-up, Fig. 1) was extracted and summarized in Table S3. Information on the type of immobilization used before and after DD protocols were implemented was extracted and summarized in Table S4.

The following data items were extracted if available: the proportion of patients discharged directly; number of follow-up appointments; return rate; number of repeat radiographs; costs; functional outcome score; any other PROMs or PREMs measured; any adverse outcomes reported. Method of assessment, timing and response rate (if applicable) were collected for all outcomes of interest. Missing information was scored as ‘not reported’. Authors were contacted if further information or confirmation of data was required.

In some studies, only part of the intervention cohort consisted of patients who were discharged directly (e.g., if the effects of the whole VFC model as a whole was assessed, rather than the effect of DD in particular; Fig. 1). For those studies, if possible, we extracted logistic outcome and adverse outcome data only for the patients who were discharged directly (DD cohort). If this was not possible, data were only extracted if > 75% of patients within the intervention cohort were discharged directly. Costs and PREMs/PROMs data were extracted and only included in the main analysis if reported specifically for patients discharged directly (DD cohort).

Two reviewers (THG and JV) independently assessed risk of bias using the Cochrane risk-of-bias tool for randomized controlled trials [6], and the Newcastle–Ottawa Quality Assessment Scale (NOS) for non-randomized studies [7]. A modified NOS was used to assess non-comparative studies. Following the manuals of the tools, studies were scored as either having a “low”, “medium”, “high” or “unclear” risk of bias.

Disagreements regarding study selection, data collection or risk-of-bias assessment were resolved through discussion or by consulting a third author (JCG).

To summarize the results, various outcome-specific summary measures were estimated based on the extracted data, including: (1) logistic outcomes: DD rate (proportion of patients discharged directly); mean number of follow-up appointments, and in case of a comparative study the mean reduction; mean repeat radiographs, (2) financial outcomes: costs as reported, and in case of a comparative study, the absolute and proportion difference in euro (€) and %, respectively, (3) patient outcomes: functional outcome score using a validated multi-item questionnaire; satisfaction with treatment; number and rate of adverse outcomes divided in non-union, secondary surgery, and other.

We chose not to pool data via meta-analysis due to high levels of clinical and methodological heterogeneity. For all outcomes, findings were, therefore, presented narratively and in summary tables.

The search identified 5872 unique records, 5668 of which were excluded after screening title and abstract. A further 184 studies were excluded after reading full texts. Six additional studies were included after scanning reference/citation lists and an updated search. Figure 2 shows the selection process and an overview of the reasons for exclusion.

Two studies were based on the same cohort [2, 8]. In both studies, the only outcome relevant for this systematic review was the DD rate. Therefore, we did not exclude one as duplicate, but merged the results of these studies in the corresponding table. Another study described results for three types of injuries separately [9]. Therefore, the results of this study were summarized per injury separately in all tables.

Ten studies (38%) compared DD to routine care [8‐17], whereas 16 (62%) studies did not have a control cohort [1, 2, 18‐31]. Twenty-four studies (92%) were conducted in the UK and two in New Zealand (8%) [25, 26]. The patient population was described in 22 studies (85%), whereas 4 (15%) did not describe the study population. One study included an adult population, seven a paediatric population, and 14 a mixed population. Twelve studies included a single injury (46%) and 14 (54%) multiple injuries. The 26 studies included a total of 38,506 patients, 3832 of which were assessed before implementing any changes (control cohort) and 34,674 patients thereafter (intervention cohort). Of these 34,675 patients, 11,133 were discharged directly (DD cohort).

A detailed description of the eligibility criteria for DD was reported in ten studies (Table 1) [9‐11, 13, 16, 17, 25, 26, 28, 31].

Risk of bias was “low” in fifteen studies, “high” in four studies and “unclear” in the remaining seven studies (Table S5).

In 10 studies, the intervention cohort consisted exclusively of patients discharged directly. Hence, a DD rate could not be estimated. In the remaining 16 studies, the DD rate ranged from 18.2 to 97.9% (Table 2). Of these 16 studies, three reported a DD rate as a proportion of all patients that attended the ED with a musculoskeletal injury. The DD rate in these studies was 26.7%, 59.5% and 33.3%, respectively [1, 8, 30].

The mean number of follow-up appointments was reported in eleven studies, and ranged from 1.00 to 2.08 in the control cohort vs. 0.0–0.33 after implementing DD (Table 3). In the comparative studies, the mean reduction of follow-up appointments ranged from 1.00 to 1.78 appointments per patient.

Twelve studies determined a return rate of patients after DD. Brooksbank et al. found 19.7% of patients that sustained a mallet finger injury to return after DD [21]. All other studies reported a return rate of less than 10%.

The mean number of repeat radiographs was reported in five studies [9, 10, 17, 22, 24, 25]. Of the two comparative studies, Mackenzie et al. reported a mean reduction of 0.34, 0.17 and 0.79 radiographs per patient, in patients with fifth metacarpal neck, fifth metatarsal and radial head fractures, respectively [9]. Seewoonarain et al. found a mean reduction of 1.00 radiograph per patient with a torus wrist fracture [17]. In three non-comparative studies, the mean number of repeat radiographs in the DD cohort was 0.0 based on a return rate of 0% [22, 24, 25].

Six studies estimated costs with reductions ranging from €69 to €210 per patient after implementing DD (Table 4) [9, 13, 17, 22, 25, 26].

Seven studies assessed functional outcome using a validated questionnaire, including four non-comparative and three comparative studies (Table 5). The three comparative studies found equal functional outcome scores before and after implementing DD [9, 11, 13]. Of them, Mackenzie et al. reported significantly better QuickDASH scores at 6 months within the DD subgroup of patients with fifth metacarpal fractures [9]. All four non-comparative studies reported good recovery in terms of functional outcome based on QuickDASH scores within the DD cohort.

Satisfaction with treatment was reported by nine studies, five of which were comparative and four were non-comparative (Table 6). Of the comparative studies, Bansal et al. found that patients with fifth metacarpal neck fractures were more satisfied after DD compared to patients that were followed-up [5.1 vs. 7.0 on a 1 (very dissatisfied) to 10 (very satisfied) rating scale] [11]. Three studies reported no difference in satisfaction before and after implementing DD, without providing any rates [13, 15, 16]. Mackenzie et al. found high satisfaction rates before and after implementing DD (95% vs. 98% using a yes/no “are you satisfied with treatment” question) [9]. The four non-comparative studies all assessed satisfaction rates based on a Likert satisfaction scale, and satisfaction ranged from 84.9 to 100% [21, 24, 25, 29].

All other PROMs and PREMs that were reported in the individual studies are summarized in Table S6. These included, amongst others, satisfaction with recovery and whether patients had visited other clinicians such as their general practitioner for the treatment of their injury.

Non-union-rate was reported by three studies (Table 7), ranging from 0.0 to 0.9% in the control cohort vs. 0.0–2.3% in the DD cohort [9, 12, 20]. Secondary surgery rates were reported in two comparative studies and three non-comparative studies, ranging from 0.0 to 1.1% in the control cohort [9, 12], [9, 12, 18, 20, 28] vs. 0.0–2.3% in the DD cohort [9, 12, 18, 20, 28]. Four non-comparative studies reported that ‘no adverse outcomes’ occurred [22, 23, 27, 31].

This systematic review supports the idea that patients with certain simple and stable injuries can be discharged directly from the ED without compromising patient outcome. This suggests that DD offers an opportunity to alleviate fracture clinic workload by reducing unnecessary appointments and consequently healthcare costs. This will allow physicians to spend more time on patients with more complex injuries, teaching, training, or improving standards of care.

Frequently, studies, including several systematic reviews performed recently [32‐34], report on the effects of the VFC model as a whole. This model includes both DD of simple stable musculoskeletal injuries, as well as a daily VFC review, consequently including more complex injuries that require follow-up (Fig. 1) [1, 2, 8, 33]. This limits the ability to independently assess the feasibility, efficacy and safety of DD of simple stable musculoskeletal injuries as a solitary concept. To our knowledge, this is the first systematic review to focus exclusively on DD of simple stable musculoskeletal injuries. Herewith, it provides an extensive and critical evaluation of all evidence currently available.

The DD rate as a proportion of all patients in the ED with a musculoskeletal injury ranged from 18 to 59.5% [1, 2, 8, 14, 30]. This is a remarkably large variation, and despite the studies’ lack of detail on their DD eligibility criteria, this variation is most likely caused by a combination of: (1) differences in the definition of ‘all musculoskeletal injuries’ (i.e., including contusions, wounds, soft-tissue injuries, or not), (2) differences in the kinds of injuries discharged directly, and (3) differences in the period since the DD protocols were first implemented [31]. Regardless of this variation, when implemented, DD will concern a large number of patients with musculoskeletal injuries that are seen frequently in an ED. This is an important factor to determine the logistic level of impact on a fracture clinic. Among the included comparative studies in this systematic review, the mean reduction in the number of follow-up appointments after DD ranged from 1.00 to 1.78 after DD.

All studies that estimated financial effects found DD to reduce healthcare costs. However, cost-analyses were limited to relatively simple estimations of fracture clinic costs, such as material costs, radiology costs and staffing costs. Hence, other important cost categories, such as other healthcare costs, (unpaid) productivity costs, and possibly informal care costs were not included. While it seems evident that healthcare costs reduce when healthcare utilization reduces, these results should be interpreted in a national context, as different healthcare payment systems are in place in each country. Furthermore, full economic evaluations, preferably from a broader healthcare perspective or a societal perspective, are needed to estimate the cost-effectiveness of DD compared with usual practice.

All comparative studies reported ‘as good’ or ‘better’ functional outcome and satisfaction with treatment in the DD cohort compared to patients treated before DD protocols were implemented. The non-comparative studies also reported high levels of satisfaction and satisfactory functional outcome. Of the included studies, Bhattacharyya et al. reported a relatively high mean QuickDASH after DD of patients with a clavicle fracture [18], but this is within the range of the normative values of this questionnaire [35].

Non-union and secondary surgery rate were reported in fifth metatarsal fractures [9, 12, 20], fifth metacarpal fractures [9], radial head fractures [9, 28], and clavicle fractures [18]. These rates were low and comparable in all cohorts. The DD model is established around the idea that a large proportion of patients are well able to manage their recovery independently, if adequately instructed. This model also appreciates that some patients will have concerns or persisting pain, and an even smaller number might develop complications like non-union. However, these problems would have probably also occurred despite routine follow-up, and the majority of patients recover without any issues. Follow-up should, therefore, not solely serve as a safety net to identify those patients with concerns, or complications that might occur in 1–2%. Rather, our results emphasize the importance of instructing patients when to contact the hospital, and of providing an open access helpline in case of any concerns. This helpline should always be part of the DD model, with subsequent face-to-face review in a fracture clinic if necessary.

This review should be regarded in light of the following limitations. First and foremost, there was high clinical and methodological heterogeneity amongst the included studies. As a consequence, we were not able to perform a meta-analysis.

Second, most studies were non-randomized and are therefore prone to selection bias, especially in retrospective cohort studies. Randomization at an individual patient level might not always be feasible for treatment redesigns like DD. However, other methods to reduce confounding effects of systematic differences in baseline characteristics were not used, including institutional cluster-randomization, or advanced statistical techniques such as a propensity score matching or weighting.

Third, only seven studies (27%) assessed functional outcome within the DD cohort using a validated questionnaire, only nine studies (35%) assessed patient satisfaction with treatment and only ten studies (38%) assessed adverse outcomes. Moreover, there was a large variety of other patient-reported outcomes/experiences measured, with methodology of assessment ranging from use of Likert scales, simple yes/no questions, and 1–10 rating scales. Furthermore, with regard to logistic outcomes, the extent to which the included studies assessed whether patients visited their GP or another hospital/clinician for further treatment was limited (Appendix Table S6). However, only non-comparative studies assessed this, while comparative studies would be needed to indicate whether DD increases GP visits or visits to another hospital/clinician.

Fourth, several studies have reported high numbers of patients discharged directly, but this far exceeds the number of patients in which logistic, financial and patient outcomes have actually been evaluated. To illustrate, Glasgow Royal Infirmary have discharged 3802 patients in the first year alone [2], while a later study reported that 30,000 patients were treated successfully since the implementation of their VFC pathway, 65% of which were discharged directly [36]. White et al. studied a cohort of 12,069 patients, 3222 of which were discharged directly [1]. Despite these figures, this systematic review included ‘only’ 2137 patients in the DD cohort to assess logistic outcomes, with even smaller sample sizes for patient outcomes. In other words, there appears to be a gap between clinical practice and evidence base. Additionally, in 62% of the included studies, a clear description of the eligibility criteria for DD was not included and often limited to ‘simple’, ‘minor’ or ‘stable’ injuries. Altogether, this complicates implementation and external validation of DD in other hospitals.

Fifth, most studies were conducted in the UK; hence, the generalizability to other countries might be limited depending on the similarity of their healthcare system with that of the UK, e.g., whether extensive low-threshold public healthcare is available. Different baseline levels of effectiveness of care prior to DD, based on local protocols, might cause logistic outcomes to be different, and patient’ mindset might result in different patient experiences or outcome, as well as patient’s acceptance of DD without further care.

Last, most studies did not report a priori sample size calculations based on the minimal clinical important difference of a predefined primary outcome, such as satisfaction or function. Despite the significant reductions in appointments, sample sizes were often relatively small and, therefore, lacked statistical power to determine a change in patient experience, outcome or complications.

Future studies on DD should be prospective, comparative and include subgroup analysis of each injury eligible for DD. We propose the minimum set of outcome variables of such studies to include: mean number of follow-up appointments, whether patients visited their GP or other hospital, functional outcome using a validated questionnaire, satisfaction using visual analogue scales as well as Likert point scales, and non-union/secondary surgery rates after at least 1 year.

Furthermore, future studies should also focus on fine-tuning the DD treatment protocols by assessment of outcomes within specific patient sub-groups (i.e., based on age, comorbidity, injury subgroup, etc.). If such analyses indicate that specific patient characteristics are predictive of, for example, high levels of return for follow-up, dissatisfaction or low functional outcome, treatment protocols should be adjusted accordingly. Preferably, a multicentre database is be established to this end, as it is likely necessary to have a relatively large sample size in order to have sufficient power to conduct such subgroup analyses.

Based on the high remodelling capacity and low rate of non-union in children [37], it is highly likely that DD is also a safe alternative for several stable paediatric injuries. However, Robinson et al. were the only authors to report on DD of paediatric injuries, other than paediatric clavicle and torus wrist fracture, exclusively within a paediatric cohort [31]. Future studies could focus on the identification of additional minor and stable injuries that can be discharged directly, both in the adult and paediatric population.