Tailoring interventions to implement recommendations for the treatment of elderly patients with depression: a qualitative study

Only 50 % of patients with depression receive care in accordance with guidelines [1, 2]. Many factors may impede or facilitate adherence and determine whether a patient receives appropriate care. These factors are referred to as determinants of practice [3]. Knowledge about determinants of practice can guide efforts to develop and choose interventions that are tailored to address those determinants and more effectively implement guidelines. Applying and increasing knowledge about effective strategies for implementing guidelines can potentially reduce the gap between scientific evidence and clinical practice.

The aim of the Tailored Implementation for Chronic Diseases (TICD) project was to directly compare alternative approaches in the tailoring process and subsequently assess the effectiveness of tailored implementation interventions [3]. The Norwegian part of TICD addressed elderly patients with depression [4]. Elderly patients with depression have an increased risk of a chronic course, and the prognosis is worse as compared with younger adults [5, 6]. Evidence indicates that healthcare professionals use longer time to diagnose depression and initiate adequate treatment in elderly patients [7].

Adherence to guidelines for depression improves patient outcomes [8, 9]. A logical step to improve adherence to guidelines is to identify significant determinants of practice and tailor implementation interventions to address these factors. Tailored interventions are more likely to improve professional practice than no intervention or dissemination of guidelines alone [10]. However, it is uncertain how best to tailor interventions. Thus, there is a need to compare different ways of identifying determinants and developing implementation strategies to address those determinants.

We conducted a systematic review of 13 clinical practice guidelines for the management of depression [11]. With the help of a reference group (see “Acknowledgments”), we prioritised six recommendations that we wanted to implement (Table 1). Depression in the elderly is frequent, affecting 10–16 % of people over 65 years, and complex, triggered by social, psychological, and biological factors [12, 13]. Acknowledging this complexity, the recommendations addressed the need for a coordinated combination of interventions, including pharmacotherapy, psychotherapy, self-help strategies, social strategies, and coordination of care. As a result of the prioritising process in a previous part of this project only treatment issues, and not diagnostic, were chosen [14]. Although the evidence for the effectiveness of antidepressants in depression has been questioned [15], there are systematic reviews that indicate that antidepressants are beneficial in combination with psychotherapy in elderly patients with severe depression [16, 17].

We identified 352 determinants of practice for the six recommendations using a multi-methods approach, and prioritised 99 determinants that we wanted to address by tailoring interventions to facilitate adherence to the recommendations [14]. The aim of this article is to describe how we developed implementation interventions based on these determinants of practice. We will evaluate the effectiveness of these interventions in a randomised trial [4], and conduct a process evaluation to examine the validity of the tailoring methods that were used [18]. Partners in the TICD project are conducting parallel studies addressing the implementation of guidelines for different chronic conditions in four other European countries. The sequential steps in the TICD project, the Norwegian publications focused on improving treatment of elderly patients with depression, and the cross-country TICD publications are summarised in Table 2.

The TICD group developed a common protocol for the study. Each of the five countries in the TICD project selected between eight and 30 determinants for discussion in focus groups. The methods, setting, study sample and group interviews are presented in detail elsewhere [19, 20]. Here we briefly describe how we conducted our study in Norway.

We (EA, SF and AO) selected 22 of the 99 prioritised determinants of that we had identified in a previous study [14]. We used the tools provided by Flottorp and colleagues (TICD Worksheet 3: development of an implementation strategy) [21]. We independently assessed each of the 99 determinants by evaluating its likely impact (3 = major impact, 2 = moderate impact, 1 = minor impact) and the effect of the likely impact on adherence using a 7 point scale (−3 = major reduction in adherence, 0 = no effect in adherence and +3 = major increase in adherence) [21, 22]. This process yielded a product of the likely impact of the determinant and the likely effect on adherence (range −9 to +9). We then discussed potential implementation strategies, the likely impact of the implementation strategy, the feasibility of the implementation strategy and whether the strategy should be targeted (only implemented for selected GPs, practices or communities where the determinant could be identified) or adjusted. In addition, we assessed each determinant in light of what we could accomplish within the resources and the timeframe of the project. We also assessed what might be realistic based on our knowledge of the Norwegian primary healthcare system. Thus, the prioritised determinants were not a result of the scoring process alone. We resolved disagreement by discussion.

We developed a draft of a plan with 55 interventions that addressed all the 22 prioritised determinants and the six recommendations. We grouped the 55 interventions in the plan in six different categories (Box 1).

We grouped all interventions that the participants suggested into the six categories in the drafted plan. As described above, the intervention plan by the research team included 55 interventions that addressed all the prioritised determinants and could be fit within the six intervention categories [14]. This is presented in the logic model (Fig. 1; “Appendix”).

The logic model is a construct that connects the planned interventions to the determinants and the assumed effects of the interventions. One or more interventions may address one or more determinants.

We presented the selected 22 determinants to the groups (Table 3). We did not present our draft plan to the participants, to avoid influencing their thinking.

Thirty-one people consented to participate. Two did not show up (one sick, the other gave no reason). Thus, 29 people participated in the various group sessions (five GPs, four implementation researchers, six primary care nurses, six representatives from professional and voluntary organisations, five quality improvement officers and three relatives of elderly patients). Making personal calls was a practical and effective strategy to recruit participants. Three people living in other parts of Norway participated in the group interviews using Skype. The group sessions took place from October to December 2013. The groups were enthusiastic, relaxed and creative. Written feedback from 23 (80 %) of the participants indicated that they experienced the group sessions as meaningful. The limited time was a challenge and several claimed that they should have prepared better prior to the meeting. One Skype participant found it difficult to contribute to the structured part of the session for technical reasons. A majority (70 %) said that their participation was useful and 68 % said it was satisfactory. None indicated that their participation was unsatisfactory or not useful.

The six groups yielded approximately 450 suggested interventions, of which many were related to each other and to suggested interventions in the drafted plan. We found that approximately 70 suggestions contained statements or attitudes rather than interventions (such as “Lack of available services is more important than GPs’ time constraints”.). This left 379 suggestions of interventions for further analysis. In the first four group interviews, we presented the recommendations in the following order: social contact, collaborative care plan, depression case manager, counselling, mild depression and severe depression. There were 127 suggestions for interventions to improve adherence to the recommendation on social contact, 68 for collaborative care plan, 54 for depression case manager, 40 for counselling, 47 for mild depression, and 43 for severe depression. The groups with representatives of professional and voluntary organisations and quality improvement officers generated the most suggestions, (106 and 96 respectively). The group with nurses generated 67 suggestions and the implementation researchers 36. The groups with GPs and relatives focused mainly on three selected recommendations. They generated 41 and 33 suggestions respectively.

We reduced the number of interventions from 379 to 65 based on our assessments of their likely effectiveness and feasibility. Of these, 28 were added or modified after the group sessions (18 modifications and 10 new interventions). We determined that 18 of the interventions should be adapted to municipalities or practices (Table 4). In Table 4, we present the following for each recommendation: the research group’s suggestions for interventions prior to the group interviews (first column), modifications of these suggestions and new suggestions from the group interviews (second column), whether the intervention should be adapted to each municipality or practice and, if so, how (third column), and the determinant at which each intervention was targeted (last column).

We removed suggestions that could not be tested in the planned randomised controlled design [4], such as “dissemination in the Norwegian Electronic Medical Handbook” and “using newspapers to inform patients”. Some suggestions were related, although addressing different recommendations and determinants. We were able to group several interventions together. We reduced the number of interventions presented in the logic model from 65 (the sum of interventions suggested by the research team and the groups) to 52 by combining all similar or related interventions that addressed different determinants or recommendations as presented in the logic model (“Appendix”). Each intervention is numbered in the order it appears the first time. When a closely related intervention is noted later in relation to a different recommendation or determinant, it is given the same number and a sequential lowercase letter. Thus intervention 1a is closely related to 1b, 1c and 1d, although they may not appear in numerical order. An intervention could address a single determinant, such as this suggestion: “Discuss GPs’ urge to do something and the view that prescribing antidepressants is a simple action in outreach visits”, which addressed this determinant: “GPs want to do something, and prescribing drugs is easy”. An intervention could also address several determinants, such as: “Provide structured referral forms for psychotherapy”, which addressed these two determinants: “GPs do not have the expertise to provide psychotherapy” and “There is a lack of health professionals who can provide this type of service”.

Several interventions could also address the same determinant. For example, this determinant: “There is a lack of other types of services that makes it difficult not to prescribe antidepressants in mild depression” was addressed by these three interventions: “Provide training in counselling to health professionals”, “Provide CME approved courses in counselling to GPs”, and “Provide courses as e-learning courses”.

We have conducted group interviews with several stakeholder groups to inform our decisions about how to tailor implementation interventions to improve adherence to clinical practice guidelines for elderly patients with depression. We developed a draft plan consisting of 55 interventions that addressed determinants of practice for the six recommendations, organised in six domains: resources for the development of a collaborative care plan, resources for GPs and other healthcare professionals, resources for patients and their relatives, outreach visits to GPs, educational resources for GPs and web-based services. The plan covered many of the interventions that the groups suggested. However, the groups added many new ideas, and they modified approximately half of the interventions suggested in the draft plan.

We included several stakeholder groups, to achieve a purposeful sample of healthcare professionals, relatives of elderly patients with depression, implementation researchers and others that might be able to suggest effective interventions to address the identified determinants of practice. This approach to tailoring an intervention to prioritised determinants was standardised across five countries and disease groups in the TICD project [4, 25‐28]. We are not aware of any other project that has addressed tailoring of implementation interventions in this comprehensive manner, using a check-list systematically to identify and prioritise determinants of practice and to identify interventions that could address them [21]. However, due to the complexity of our recommendations, the large number of prioritised determinants and the limited time available for the interviews, it was not feasible for us to address each of the suggested interventions in the systematic way stated in the common TICD protocol.

The recommendations that we prioritised addressed several levels of the healthcare system, from the patients and their relatives to the healthcare professionals and the healthcare administration in the municipalities. The use of the TICD checklist to prioritise determinants and interventions made it possible to analyse the results in a systematic way [21]. Nonetheless, the results from this part of the analysis were assessments based on our considerations and judgments. An alternative strategy would be for representatives from the stakeholder groups to do this assessment. The wide range and the large number of interventions that the groups discussed within a limited time may have compromised more detailed and structured discussions, and may have resulted in superficial assessments for some determinants or interventions.

The number of suggested interventions for each determinant and recommendation varied. Recommendations presented early in the session appeared to yield the most suggestions. These recommendations addressed mainly the community and municipalities, while the last recommendations addressed clinicians. It is possible that the nature of the first guidelines generated more suggestions. It is also possible that there were fewer suggestions for the recommendations presented later in the interview because of exhaustion in the groups. One solution to this could have been to present the recommendations in a different order for each of the groups.

The large number of suggested interventions addressed only six recommendations, whereas clinical practice guidelines frequently contain many more recommendations. There is a risk that guideline developers will experience information overload, if they try to use this approach for a full guideline.

We excluded suggested interventions that could not be evaluated in our planned cluster randomised controlled trial. Thus, we omitted potentially useful dissemination channels suggested by the groups, such as media and e-resources that are popular among healthcare professionals (The Norwegian electronic health library, the Norwegian Directorate of Health’s web site and the Norwegian Electronic Medical Handbook).

Determinants of practice related to depression guidelines are numerous and apply to all levels of the healthcare system [14, 29, 30]. Relatively few studies on improving the care of patients with depression have described the development of a systematically planned intervention tailored to address identified determinants. Shirazi and colleagues [31] demonstrated that tailoring an educational intervention, based on GPs’ readiness-to-change (high-low), improved GPs’ performances in hypothetical (role-playing) consultations as compared with controls. Verhaak and colleagues [32] found that disability (particularly disability that affects participation, self-care and social activities) had a major impact on depression in the elderly. One might argue that the interventions that we planned considered this aspect to a limited degree only. Nevertheless, we addressed social withdrawal and frailty in our planned interventions. Furthermore, their findings indicated that the effect of disability on depression was largest among the younger elderly (those between 60 and 70 years). We included patients 65 years or older in our study. In a randomised controlled trial based on a psychological theoretical framework, Baker and colleagues [33], identified obstacles to adherence among 34 GPs, and tailored their intervention to each practitioner. They found that this strategy improved assessment of suicide risk and depression, assessed with Beck’s Depression Inventory. They found no difference for anti-depressant therapy or utilisation of psychotherapeutic services. Addressing clinicians individually to identify determinants of practice is an attractive approach, but rarely realistic in large-scale efforts to implement clinical practice guidelines. We deemed this approach unfeasible.

In a joint analysis of the studies to tailor interventions in the TICD project, Huntink and colleagues [20] found no relationship between the total number of suggested interventions and the number of unique suggestions (interventions only suggested by one group).

The extent to which there are similar determinants of practice in other settings and the extent to which similar interventions would be appropriate in other settings is uncertain. However, many of the same determinants are likely to be similar in other settings. The approach that we used to develop a package of tailored implementation interventions was both feasible and efficient. Those interested in tailoring interventions to implement guidelines can use the TICD checklist [21] and the interview methods that we used in this study.

We are evaluating the effectiveness of the tailored interventions that we have developed in a randomised trial [4]. We will assess whether we identified the most important determinants and selected appropriate interventions to address those in a process evaluation [18].

There is a paucity of research comparing different methods of identifying determinants of practice and tailoring implementation interventions to address those. To our knowledge, the TICD project is one of the first projects that have done this [19, 20, 34]. We need more research to improve the methods used to tailor interventions and to understand how best to do this. This includes evaluation of strategies to prioritise suggested interventions, given the abundance of suggestions that is possible, as illustrated by this study.