A model of access combining triage with initial management reduced waiting time for community outpatient services: a stepped wedge cluster randomised controlled trial

Excessive waiting times for care can be a problem for both patients and health services [1]. Access issues are often associated with emergency departments [2, 3] and surgical procedures [4, 5]. However, sub-acute and community-based services also suffer from the constant pressure of lengthy waiting lists [6–9]. Delays in access to care for these services have been associated with poorer patient outcomes [10, 11], anxiety [12], and reduced engagement with services [13, 14].

Delays in care can result in waiting lists or queues. Delays, and hence waiting lists, are the result of a disparity between demand for a service and the capacity available to meet this demand [15]. Queueing theory is the equation that defines the relationship between demand, capacity and wait time [16]. Queues or waiting lists are formed when demand is higher than capacity [16]. Shortages in markets are often corrected through a “price signal”, but this mechanism is ineffective in public health services where consumers face subsidised prices and costs to consumers do not change in response to rise in demand. On the supply side, individual providers may not be responsive to price signals due to wages paid as salary or other government regulation. Hence, alternative strategies are needed to mitigate the adverse effects of long waiting times.

One strategy is to use short-term strategies such as immediate and temporary increases in supply to “clear” a waiting list; however, this does not resolve the underlying problem and waiting lists are likely to simply recur [17]. Another common strategy is to focus attention on managing the waiting list, for example, by dedicating staff to monitoring patients on the list, setting up data systems to record waiting list data or creating complex sorting or triage systems to prioritise patients according to need. These systems can help the most urgent patients to access timely care, but often do not assist in reducing overall waiting time; conversely, they may contribute to the problem by diverting resources from clinical care to administrative processes associated with managing the waiting list [17, 18].

In contrast, promising results have been reported from strategies that address patient flow by reducing complexity in booking systems, combining triage with initial management and/or actively managing the relationship between supply and demand [18, 19]. Trials in this area have focussed primarily on emergency departments (for example by placing a senior physician at triage to commence treatment and quickly manage simple cases) [20] and primary care settings (the Advanced Access approach, for example, reduces time for pre-booked appointments, opening sufficient space for patients needing same-day or next-day consultations with their local doctor) [21]. Preliminary evidence is also available from isolated studies in single services to suggest that patient flow interventions that utilise one or more of these elements may be effective in other types of community outpatient services [22–25].

Long waits for services are a problem for community outpatient services with negative consequences for patients and service providers. Of the various strategies that have been tried to reduce waiting time, there are several elements that show promise of effectiveness but evidence is limited to small, single-site studies or from extrapolation of evidence from emergency departments and primary care settings. It is not known whether a model of care that brings these key elements together can be successful in reducing waiting time across a variety of community-based outpatient health services. This trial aimed to determine whether a model of access that combines triage with initial management and allows supply to be responsive to demand fluctuations can be used to reduce waiting time across multiple community outpatient services.

The STAT model (Specific Timely Appointments for Triage) is designed to reduce waiting times for community outpatient services by booking patients directly into protected assessment appointments and combining triage with initial management as an alternative to a waiting list and triage system. A constant rate of patient flow is maintained and calculated to match the rate of referral, and service providers are encouraged to make priority decisions about ongoing treatment in the context of demand. This is the first time that this model has been trialled on a broad scale with multiple services. Findings suggest that the STAT model accounted for a 34% reduction in waiting time after controlling for clustering by site, similar to results of pilot trials conducted in community rehabilitation (42% reduction in waiting time) [22] and physiotherapy outpatients (22% reduction in waiting time) [25]. Thus, this is a feasible way to reduce waiting time across a broad range of community outpatient services, resulting in improved access to care and increased patient flow.

Reductions in waiting time achieved with STAT also appear to be comparable with other patient flow initiatives reported in community outpatient settings, although direct comparison is difficult due to heterogeneity in the ways in which wait times are measured. For example, there was a 23% reduction in median waiting time for a prosthetics clinic through changing scheduling to a modified walk-in system, rather than scheduled appointments [24]; Lynch et al. achieved a 70% reduction in the number of people on a waiting list for mental health services with an intervention that addressed the residual waitlist in combination with new approaches to treatment and triage [23]; and Maddison et al. described a reduction in waiting time for musculoskeletal services despite an increase in referrals through creation of a back pain pathway [34]. In contrast to these studies that all described interventions developed specifically for the services in which they were conducted, the current trial provides evidence of a structured approach that can be used to reduce waiting time across a broad range of settings.

Stability of demand is an important element in the effectiveness and sustainability of the STAT model. Similar to the Advanced Access developed for use in primary care [21], STAT is based partly on the observation that many services have a relatively stable demand, indicated by a waitlist that shows little variation in length over time. Patients are therefore entering the service at a similar rate, but always several weeks or months behind. If the backlog can be cleared and patients brought into a service at a rate that is consistent with the rate of demand, it follows that the service should be able to maintain patient flow without a waiting list developing. However, a sustained increase in demand is one notable risk to the model’s success. This was observed in the current trial as, despite all sites recruited to the trial having reported reasonably stable waiting lists over the previous 2 years, data from two of the sites showed an increase in referrals of more than 50% over the equivalent period in the first compared with the second year of the project, and another site experienced an increase of 25–50%. Despite this, substantial reductions in waiting times were still observed across the eight sites in the trial. This implies that efficiencies driven by STAT may be able to compensate for some increase in demand, but it is likely that a point is reached where additional strategies (such as tightening eligibility criteria or increasing supply) are needed to address the imbalance between supply and demand to achieve ongoing reductions in waiting time. It is also possible that in some services reduced waiting time may stimulate demand, leading to an increase in people seeking the service [1].

In addition to reductions in waiting time, another important outcome was the observed reduction in variability of wait time (Fig. 2). Consistent with a previous pilot study evaluating the STAT model in physiotherapy outpatients [25], the greatest benefits of this intervention appear to have come to those who were previously waiting the longest. This reduction in variability may provide an explanation for how STAT was effective in reducing waiting time overall. This finding is important, as one of the criticisms of traditional waiting lists and triage systems is the risk that low priority patients are continually pushed down the list by those with higher priority ratings, sometimes to the point where they never get seen [8].

The intra-cluster correlation of 0.05 observed for the primary outcome of waiting time was substantially higher than the estimate of 0.01 that was used to determine the sample size in the protocol [27]. This suggests that there was a higher degree of variability between the clusters than originally anticipated and that the impact of the intervention varied to some degree across sites. Given the diversity of sites included in the trial, it is not surprising that there may be site-specific factors that influence the success of the model. The STAT model requires a significant shift in the way that clinicians prioritise their workload, and response to change is likely to have differed to some degree across sites. It is possible also that the STAT model may be more applicable to some settings than others. The planned exploration of the perceptions of key stakeholders at sites where STAT was implemented using qualitative methods will provide insights on the human and service factors that influenced success.

One component of the STAT model is a one-off strategy to reduce the existing backlog prior to implementing the model. A small investment of resources was allocated to each site to facilitate this, and it could be argued that the observed reductions in waiting time were simply a reflection of short-term changes directly related to those additional resources. However, previous literature has shown that single injections of resources, without changes to service delivery, are unlikely to make a sustainable difference to waiting times [35, 36]. STAT is also consistent with other waiting time initiatives that have advocated approaches combining one-off backlog reduction strategies followed by the implementation of patient flow interventions [21, 24]. In the current trial, a comparison of mean waiting times for a small sample of consecutive patients entering the service immediately before and after the waitlist reduction strategies provides some indication of their impact. The 33% reduction in waiting time observed at the conclusion of implementation of targeted waitlist reduction strategies was consistent with the 34% reduction measured across the entire trial. This would suggest that the initial gains made during the backlog reduction strategies were maintained by the STAT model, regardless of the timing of the intervention and relative length of the follow-up period within the stepped wedge design, which continued for up to 10 months.

One perceived risk of an intervention that allows patient flow into a service at a steady rate is the possibility that a “hidden” waitlist is created, where patients receive a first appointment promptly but then wait for a second appointment. The current trial showed no difference in the time from first to second appointment when considering the data across all sites. This finding suggests that concerns about secondary delays were unfounded and that clinicians were prioritising second appointments equally.

There was an increase in the proportion of patients who failed to attend at least one appointment in the intervention period, which was surprising given that failure to attend rates have previously been negatively associated with waiting time [13]. A possible explanation is that patients in the intervention group were given information about their appointment time soon after referral rather than being placed on a waiting list for an interim period. Although overall waiting time reduced during the intervention period, the time between being given an appointment and the appointment itself increased. For example, where previously a patient might wait 6 weeks to receive notification of an appointment 1 week later, with STAT this same patient receives notification after 1 week for an appointment 3 weeks later. Forgotten appointments become more likely and could be mitigated by strategies such as SMS reminders [37].

This trial was conducted in eight community outpatient sites that differed from each other in a number of ways. They provided a range of services to patients ranging from infants to the frail elderly, some treated chronic conditions and others provided short-term follow-up to acute injuries. All sites, however, shared the common features of providing non-emergency services to patients over a series of outpatient appointments. These observations suggest that the STAT model is likely to be generalisable to a wide range of outpatient services provided that they have these features. STAT encourages clinicians to change the focus of decisions about patient priority; rather than triage decisions influencing access to the service, prioritisation is instead directed at the rationing of resources for ongoing treatment. In order for this to work, there needs to be some flexibility in the way that services are delivered. For example, clinicians working in these types of services can choose to see patients less frequently, for shorter appointments, or move patients from individual to group sessions during times of high demand. Results of this trial suggest that STAT is likely to be applicable to any non-emergency outpatient service with stable demand and flexibility in service delivery decisions, regardless of the type of service provided.

A major strength of this trial is the use of a stepped wedge cluster randomised controlled trial design, an emerging trial design in health service delivery evaluations that offers several benefits over other parallel cluster designs [26, 38]. The rigour of this method and the involvement of multiple sites offers clear advantages over other commonly used methods for evaluation of patient flow or waiting list interventions, such as single-site studies [23, 39], quality improvement methods [40, 41] or retrospective analyses of health service datasets [42, 43]. The ICC for the primary outcome of waiting time was larger than hypothesised in our sample size estimation (ρ = 0.058 versus ρ = 0.01) [27], possibly due to greater variability than expected between sites. There were many possible sources of variation, including differences between clinicians, management of each of the sites, socioeconomic characteristics of the patient population and complexity of patient needs that were not measured in the trial. Despite this, all sites provided services for patients living in the community and the analysis took account of clustering. Further adequately powered studies could investigate the effect of variation, in particular the effect of STAT on subgroups of patients, such as those from lower socioeconomic backgrounds and those with more complex health needs [44, 45].

The stepped wedge design of this trial meant that the lengths of control and intervention data collection varied between sites. As a result, there were some differences in the characteristics of patients in the pre and post intervention data driven by differences between the services, but this was accounted for by clustering in the analysis. This aspect of the design also meant that the last site to receive the intervention had a follow-up period of only 3 months. It is possible that this was not long enough to measure the true effect of the intervention. Conversely, over a longer follow-up period, sustainability of the intervention may come into question, as the effect of the initial injection of resources wears off and support from the research team is withdrawn. The relatively short follow-up time (particularly for the last site to receive the intervention) is a limitation of the current trial, and further research is required to look at longer-term outcomes. A further challenge of the trial design was that it provided for little flexibility in the timing of implementation of the intervention, reducing backlogs and embedding new processes into practice. It is therefore likely that greater benefits may be achieved when implementing STAT without these limitations.

There were some minor deviations from the protocol due to the characteristics of the services selected for inclusion and availability of the required data. We intended to collect data on the number of patients on the waiting list at key time points for each service to assess the fidelity of the implementation strategies to reduce backlog at each site. It was not possible to collect these data across all sites due to differences in the way that waiting list data were recorded. Instead, we analysed waiting time for a sample of 20 consecutive patients at each time point rather than counting the number of patients on the waiting list. We also intended to analyse the number of group and individual appointments across sites and time periods to see whether the new model of care led to increased use of group appointments, as observed in a previous trial [22]; however, this was not necessary as the majority of the sites selected for inclusion in the trial did not offer group appointments as a treatment option.

A model of access and triage based on evidence-based strategies known to improve patient flow was successfully implemented on a broad scale, involving eight community outpatient services. Waiting time was reduced by 34%, and waiting time variability also decreased substantially, suggesting those people previously waiting the longest were likely to benefit most. This trial also demonstrated that evaluation of patient flow initiatives previously limited to single-site studies, quality improvement projects or retrospective analysis of health service data can be conducted using rigorous research methods to produce high-quality evidence for health care service providers.