De-implementation of low value castration for men with prostate cancer: protocol for a theory-based, mixed methods approach to minimizing low value androgen deprivation therapy (DeADT)

It is useful to consider a conceptual model for theory-based qualitative barrier assessment (Aim 1), quantitative prioritization (Aim 2), and piloting of de-implementation strategies tailored to provider behavior change techniques (Aim 3). As illustrated in Fig. 1, we highlight several TDF domains and constructs in our conceptual model that may contribute to organizational, provider, and patient behavior in the setting of ADT for localized prostate cancer. In addition, our qualitative approach allows for flexibility as we conceptualize the main issues when it comes to chemical castration. Last, the quantitative discrete choice methods (Aim 2) create significant opportunities to examine interactions among domains and constructs allowing us to select, tailor, and pilot the most informed organizational and individual level de-implementation interventions during Aim 3.

This study is being conducted across several sites in the Veterans Health Administration with the majority of study activities at two sites: the VA Health Services Research & Development Center for Clinical Management Research at the VA Ann Arbor Healthcare System and the Rogel Cancer Center at the University of Michigan.

We will identify potential study participants using methods successfully used in prior studies of VA providers and prostate cancer survivors [67, 68]. Specifically for the Aim 1 semi-structured provider interviews, we will purposefully sample 10–12 urologists from our integrated delivery system facilities with the highest use of castration as primary treatment to achieve thematic saturation. Likewise, we will also sample 2–4 urologists from facilities with the lowest use of castration. Any urologist who has experience caring for prostate cancer patients on ADT and expresses interest in prostate cancer care will be eligible to participate in semi-structured interviews. We will consider interviewing non-physician providers (e.g., nurse practitioners) if they prescribe a significant amount of ADT. To better understand patient perspectives on not initiating or stopping castration with ADT, we also plan to conduct a limited number of patient interviews (~ 6) from high outlier sites.

For the Aim 2 DCE, we will distribute invitations to the national VA Urology listserv which includes approximately 250 VA urologists. We anticipate this survey will be available over the intranet, as an online survey, as well as available in paper and pen form if preferred. We will use a modified Dillman technique to enhance response rates similar to prior successful survey research [69]. Our engagement with the national VA Urology listserv and the American Urological Association Annual Meeting, alongside operational support from the VA Urology Surgery Advisory Board, should ensure that at least the minimum number of providers needed participate. If needed, we can pursue other organizations (e.g., American Urological Association, Society of Urologic Oncology) given our prominent urologic oncology team.

For Aim 3, we will conduct pilot testing at two sites during years 3 and 4 of this proposal. This will take place at the VA Ann Arbor Healthcare System and University of Michigan’s Rogel Cancer Center Urologic Oncology clinics. The number of patients initiating primary ADT at each site should be adequate for piloting based on preliminary data. However, the primary purpose of the pilots is to assess for issues described in more detail below [70], not effectiveness, so the number of patients on ADT is not critical to success for pilot outcomes.

This proposal received Institutional Review Board approval from the VA Ann Arbor Healthcare System and the University of Michigan. Aim 1 initially involves analyses of retrospective administrative data to identify facilities, their characteristics, and ADT utilization rates. Study data will come from the VA Corporate Data Warehouse and its Oncology database. This is a centrally maintained national electronic data repository consisting of cancer registry, laboratory, pharmacy, utilization, and administrative data for all patients receiving care in the delivery system. Reported data will be aggregated to the facility level which will not permit identification of individual patients or providers. Next, we will conduct semi-structured interviews with providers and a limited number of patients from facilities with the highest ADT treatment rates for localized prostate cancer. We will audiotape and analyze this data in Aim 1. The risks of the proposed semi-structured interviews are limited, but do include psychological stress related to discussions of appropriateness of prostate cancer treatment with ADT and possible tensions given the low levels of evidence supporting ADT use in localized disease. Another risk is the potential breach of confidentiality of this interview data. Precautions will be taken to protect against any such breach, making the risk low, following standard operating procedures at our Center. Potential human subjects’ risks for the national survey in Aim 2 are low including an email and minimal disruption to provider work flow. With the exception of some demographic information (e.g., age in 5-year increments), the participant information and responses will be kept anonymous. Aim 3 risks for pilot site providers include disruption of clinical workflow and a low likelihood of psychological distress. These risks will be site- and intervention-dependent with a research assistant on-site to assist with recruitment, acceptability of the intervention, feasibility in clinical practice, and data collection and outcome assessment issues. The right of subjects to discontinue their involvement in the research at any time will also be fully disclosed during the informed consent process.

For the qualitative portions of Aim 1, in person and telephone interviews will be piloted with providers at the VA Ann Arbor Healthcare System followed by 12–16 urologists from several other VA Medical Centers as described. Medical Center Directors and Chief Urologists will be contacted via email and brief telephone calls initially and will also be provided information on the nature and purpose of the study, as well as the types of questions their providers will be asked. With approval, urologists who express an interest in participation will be consented by a trained research assistant, who will provide them with information on the nature and purpose of the study, as well as the types of questions that they will be asked (focused on ADT treatment for prostate cancer). Written or recorded phone consent will be obtained. Those who are eligible will then be called at a later date for a formal interview lasting approximately 45 min. Interviews will be conducted by the study team and will be audiotaped. We will use similar recruitment methods for our limited number of patient interviews.

For the survey portion of Aim 2, providers will be identified through a national urologist email listserv and contacted/recruited via email to participate in a DCE survey. An informed consent form will be included, highlighting the nature and purpose of the proposed research as well as sample questions. Participants will be asked to electronically sign the consent form, and will be clearly offered the option of not participating.

Urologists at the pilot sites in Aim 3 will be given an overview of the site-specific intervention, highlighting the nature and purpose of the proposed research, as well as what they should expect to encounter during routine clinical care should they agree to proceed with the study. Pilot site urologists will be asked to sign the consent form by the site research assistant, if they agree to participate, and will be clearly offered the option of not participating.

The goal of Aim 1 is to clarify barriers and facilitators to stopping castration with ADT as primary prostate cancer treatment using an individual behavior change framework, the theoretical domains framework (TDF) [71]. This approach will identify key barriers to de-implementation of low value ADT-based castration. We will conduct semi-structured interviews with urologists to clarify preferences for (i.e., facilitators) and barriers to stopping ADT use. This will prepare us for development of a theory-based DCE among a national sample of urologists in Aim 2 to quantify the relative importance of barriers, and to direct intervention strategy tailoring to increase acceptability and effectiveness. Embedding use of the TDF within a DCE is particularly innovative.

The goal of Aim 2 is to then prioritize barriers and facilitators to de-implementation of chemical castration with ADT discovered in Aim 1. The highest priority barriers will need to be addressed during strategy development and tailoring for our pilot interventions in Aim 3 to support acceptability and feasibility in practice. We will accomplish this using a DCE. This stated preference method drives marketing strategy development based on stakeholder preferences for a given practice and is a promising applied approach for health care optimization [59‐62, 76, 77]. Real-world DCE examples include prioritizing provider preferences related to hospital consultant work and electronic medical record use [78, 79].

In our DCE, the barriers and themes with the highest frequency and most conflicting beliefs across respondents identified in Aim 1 will be refined and presented as the “attributes” and associated levels will be developed. In short, we will use data obtained from Aim 1 to develop TDF-based choice sets for inclusion in a national urologist DCE. Once we have the most important, not just most common or conflicting, themes and barriers based on a national urologist DCE, we can select the most effective evidence-based behavior change techniques to direct de-implementation tailoring efforts in Aim 3.

In a DCE, stakeholders, in this case urologists prescribing ADT, choose between hypothetical alternatives described by themes and barrier characteristics identified from Aim 1 (i.e., attributes and levels). The five or six highest frequency and most conflicting themes based on Aim 1 findings will be varied across hypothetical examples. Through systematically varying a set of levels in a series of choice sets where providers are asked which option they most prefer, we gather critical data on castration preferences and tailoring for our pilot interventions. This results in a preference structure where certain attributes (barriers) are most important across the respondent sample. The data obtained on the levels of attributes shows which direction the attribute is most favored. In an example of whether high or low physician autonomy or level of evidence is driving ADT treatment decisions, for each choice set given, we will ask respondents to choose the situation that they would most prefer when prescribing ADT for their prostate cancer patients (choice A, choice B, or neither). We may ask participants: “Please select the situation that you most prefer when prescribing ADT as primary treatment.” Candidate themes, attributes, and levels include, for example, the ability to make the final recommendation (levels: yes, no), and the amount of clinical time required (levels: 15 min, 30 min). We will use an opt-out option to ensure realistic scenarios. Data from this DCE survey will identify leading themes and attributes (barriers).

There are several potential implementation strategies to de-implement low value castration within the broad categories we identified for this proposal. We focus on formulary restriction and informed decision-making because of their difference in key attributes, including level (organizational vs. individual), likelihood of quick success vs. long-term sustainment, and effort required by clinicians. We describe some possible strategy design features briefly for each type in Table 2. While several options exist, there is no existing evidence to inform the best approach from the provider perspective. Aims 1 and 2 will inform which of these strategies is likely to be most acceptable to clinicians, and provide data needed to tailor them. For example, we do not know how a blunt formulary restriction intervention would be received by providers considering primary ADT treatment.

While formulary restriction of ADT for localized prostate cancer seems warranted, we may find that it is widely considered unacceptable to providers and patients. Nor do we know how shared or informed decision-making can be efficiently operationalized in a clinical setting for patients considering castration for localized disease. By tailoring each strategy using behavior change techniques and barrier solutions derived from Aims 1 and 2, we believe we can design implementation interventions that will be accepted by providers, but still allow us to test differences in the widely varying mechanisms of action. We will refine these approaches through robust efforts and finding from Aims 1 and 2 (Table 1) and the expertise of our trans-disciplinary investigative team.

Based on findings from Aims 1 and 2, Aim 3 pilot work plays a critical role to help us understand the acceptability, feasibility, and scalability of these complex interventions in preparation for a full-scale randomized de-implementation evaluation trial. In fact, the UK Medical Research Council guidance indicates piloting is essential to complex intervention development and testing prior to large-scale evaluation [47]. The main goal of both pilots will be to decrease castration rates for patients with localized prostate cancer, in a way that is acceptable to patients and to the clinicians who treat these patients. We are purposely choosing intervention strategies from opposite ends of the behavior change continuum because of their evidence-based potential to change provider behavior. Specifically, we are selecting one approach (formulary restriction policy) that operates at the organizational level and is widely perceived as a forcing function, giving providers little leeway to exercise judgment. The other, physician/patient shared or informed decision-making, operates at an individual and dyadic level, and is perceived as maximizing the opportunity for discussions between patients and providers. The first approach requires little to no learning on the part of providers, while the second requires considerable upfront learning (“cost” to the provider and possibly also to the patient). This approach sets up a testable hypothesis for our subsequent comparative effectiveness trial, that a blunt de-implementation policy may be effective in the short term but that it will lose its effects as providers learn work-arounds. Conversely, a shared or informed decision-making approach to de-implementation might take longer to observe measurable decreases in castration rates, but its effects will create sustainable change as providers internalize and routinize this clinical practice.

While the next steps for this work will be a cluster randomized comparative effectiveness trial, setting up this complex trial will create opportunities for dissemination. We will publish at least one manuscript per research aim in peer-reviewed journals, as well as submit at least one abstract to clinical, quality improvement, and/or implementation research meetings. We will convene a Steering Committee and update this group through quarterly phone calls as we progress through the research plan. In addition, we will present our findings to the American Society of Clinical Oncology, Society of Urologic Oncology, and the Association of VA Hematology/Oncology as an opportunity to include a de-implementation of low value cancer care theme in their annual meeting agendas. We will brief relevant operational partners annually including the VA National Program Director for Oncology and the VA National Urology Surgery Advisory Board. We will also share our findings and recruit for our cluster RCT at the Annual AUA Meeting.