We have reported throughput data on almost 8,000 patients referred to four NHS sites operating stepped care services for common mental health problems, mapping patients' entry and exit from the systems as well as recording the types of therapeutic inputs received. Our data illustrate the considerable variation in the design and implementation of care systems in response to the recommendation for stepped care in the NICE guidelines , which lacked explicit detail concerning the optimal model for delivery.
The variation in models was significant. Although it is helpful to place them on an operational continuum from stepped to stratified, this categorisation does not do sufficient justice to their complexity and diversity. While the design of the systems as 'stepped' or 'stratified' is a key dimension that dramatically influences the performance of stepped care systems, staff availability and professional referral behaviour can subvert initial plans in important ways. For example, lack of experienced workers meant that site C became essentially a low-intensity-only service, whereas site D saw a 'stepped' model bypassed by a significant number of patients, although this might also be influenced by patient population characteristics, such as the proportion of patients diagnosed with post traumatic stress disorder where low-intensity treatments are not known to be effective.
In contrast, however, there were some interesting similarities in patient flow between sites. The levels of attrition between referral and assessment have been observed previously in stepped care services [14–16] and are a well-known phenomena in psychological therapies services [17–19]. In the two sites where there was closer balance between low- and high-intensity provision (sites B and D), we observed a consistent 'stepping up' rate of less than 10%, similar to that reported in one of the first Improving Access to Psychological Therapies (IAPT) demonstration sites over three years .
Scheduled completion rates for all treatments were variable between sites and steps (Figures 2 and 3). Site D achieved less treatment completion at high-intensity, where only one-third of cases were recorded as completing their high-intensity treatment, although our data are limited by project end dates as cited earlier. Nonetheless, low-intensity treatment is intrinsically shorter than high-intensity treatment, and it is conceivable that patients will be more likely complete a shorter course of low-intensity treatment than the longer high-intensity option. This observation requires further investigation since the concept that treatments should be 'least restrictive' for the patient [5, 9, 20] is at the heart of the original concept of stepped care.
Our study is of four 'early implementer' sites attempting to reconfigure their services to more closely reflect NICE guidelines . As a consequence, the data may have limited generalisability. Missing demographic data (a product of poor local information systems and haphazard clinical data entry highly prevalent in UK mental health services) limits knowledge of the characteristics of the populations served by these sites.
A more significant limitation is our inability to utilise clinical outcome data to ascertain the outcomes for patients, rather than their flow through the systems. Clinicians and information systems were unable to provide us with this data. Recent evaluations of Australian 'Better Access' mental health systems have similarly suffered from a lack of outcome data, basing published evaluations on a mere 15% of all patients treated . However, it should be noted that the bulk of treatments delivered were evidence-based, as summarised in the NICE guidelines , and our analysis assumes that outcomes would be broadly in line with those reported in the guidelines. However, this remains to be confirmed, especially in the context of a stepped care system where patients may receive a number of treatments in a serial fashion.
That clinical services do not use routine data is of greater concern, given the supposed centrality of outcome monitoring in stepped care to support clinical decision-making and ensure that services are responsive when patients do not benefit from initial 'steps.' The assessment of treatment effect to aid clinical decision-making appears deficient in these services. However, the use of formal psychometric measures is not the only way to assess treatment progress. Clinical judgement may be applied by mental health workers using clinical interview assessments. Clearly, for research purposes this is less easy to quantify without routine outcome measures, but their absence does not necessarily invalidate the stepped care process given our patient pathway data. Whether the use of such measures within a formal clinical decision-making algorithm would have reduced the variation in patient pathways could be a potential subject for further research.
We were also ignorant as to any additional service options available to patients and referrers outside the services we studied. Other resources available may alter who is referred to these services and we cannot assume that the proportions of patients 'needing' step two or step three interventions would be the same in all sites.
The study highlights variability in the implementation of stepped care for depression. This is to be expected, to the degree that the NICE guidelines were not explicit about a number of issues, and provided no formal 'blueprint' for the organisation and delivery of services. Our observations of variability are confirmed by the report on the first-year IAPT service . Of course, some local variation is desirable, but the very different service delivery models may not be desirable in the long term. It is important that evidence from implementation studies such as that reported here impact on later iterations of the depression guidelines, to maximise standardisation where appropriate and ensure that models of service delivery have a solid basis in evidence, complementing the evidence base relating to the health technologies which are delivered within these services.
Our data suggest that the principal driver of patient flow through stepped care systems is the allocation to initial treatments. The rate of stepping up was low, no matter how the patients were assessed or how many were allocated directly to high-intensity treatment. Not dissimilar proportions of patients were stepped up in systems which allocated large numbers directly to high-intensity treatments, allowed referrers to make direct referral to high-intensity treatment or direct most patients to low-intensity treatment. The two services which stepped fewer patients from low- to high-intensity treatment included one where lack of resources led to very little high-intensity treatment provision, and another where initial allocations to high-intensity treatment was almost 50% of referrals assessed.
Although service planners may seek to design services that reflect their desired balance between stepping and stratification, they must be aware that patient flow is highly sensitive to other factors, including the background of the workers at each step of the service. Triage or assessment by a professionally qualified workforce may lead to more people receiving high-intensity treatment, providing that option is available. Service managers may need to plan on the basis that whatever the initial allocation rates of patients to low- or high-intensity treatments, providing sufficient high-intensity resource is available, less than 10% of patients may be stepped up from low- to high-intensity treatment. It is, therefore, important to resource all available steps sufficiently to allow patients to be stepped up from low- to high-intensity treatment and to prevent situations arising where patients might be inappropriately 'held' at a low-intensity step.
Finally, stepped care systems do not seem to differ from the often observed attrition rates to psychological therapies at all stages in the patient pathway. Access to care has not traditionally received the same research focus as issues of treatment effectiveness. That is now beginning to change [7, 23, 24] but there is an urgent need to understand the reasons for these levels of attrition and ensure the findings are used to inform the design of stepped care systems in the future.
Although implementation science is relatively new to the evidence-based movement, its use in ensuring that effective treatments and organisational models are put in place consistently is now recommended by research funders . Stepped care would seem to be a prime example of a recommended organisational system idea being interpreted and applied in very different ways. The application of core principles of implementation science (such as targeted planning, implementation strategies and clear activity specification) around service organisation is urgently required.