The Delphi method was adopted. The study design followed the “Guidance on Conducting and Reporting Delphi Studies (CREDES)” (20). Expert consensus was achieved through repeated consultations. Each expert was invited to rate the importance of a series of pre-defined conditions independently (21). They were also encouraged to suggest any changes, such as modification, deletion, or addition in open text boxes. The aggregated anonymous rating results were then fed back to the participants in subsequent consultations. This approach allowed participants to express their opinions free from peer pressure. The consultations were repeated until a consensus was reached among the participants.

The Delphi method is effective in establishing expert consensus when the published literature is not able to provide a robust conclusion. While expert consensus is deemed as a weak type of evidence in evidence-based medicine, its validity is supported by “wisdom of crowds:” research shows that groups can make good judgements under certain conditions (22). A recent systematic review suggests that a sample size in the low to medium double-digit range is adequate for a Delphi study (23). However, Delphi studies can now be run online with a very large panel size, as shown by Japan's NISTEP Delphi survey (24). Delphi method has been used in developing measures to assess the quality of primary care services in China (25). In the current study, the Delphi method was adopted because there is a paucity of literature documenting facilitators and barriers in community detection and management of MCI.

The study protocol was approved by the Human Ethics Committee of La Trobe University (HEC20125) and the Medical Ethics Committee of Yangpu Hospital, Shanghai, China (LL-2019-SCI-004).

The Delphi consultations involved three stages: expert panel formation, development of consultation rating scales, and data collection and analysis (20).

Data were analyzed using Microsoft Excel.

Response rate [(Number of returned responses)/(Number of invitations)×100%] was calculated to reflect the enthusiasm of the eligible experts in participating in this study. Demographic characteristics of the participants were described using frequency distributions. The degree of authority of the consultation results was measured using the authority coefficient (Cr), which was determined by the reported familiarity (Cs) and the judgement foundation (Ca) of the participants:

In line with the literature (35–37), Cs scores were interpreted as: 0.9 = very familiar; 0.7 = familiar; 0.5 = somewhat familiar; 0.3 = a little familiar; 0.1 = unfamiliar. We used the scoring matrix that has been widely used in China and validated for Delphi studies on primary care services to calculate summed scores (36, 37): theoretical analysis (0.3 = a great deal; 0.2 = moderate; 0.1 = little); work experience (0.5 = a great deal; 0.4 = moderate; 0.3 = little); referring to literature (0.1 = a great deal; 0.1 = moderate; 0.1 = little); and intuition (0.1 = a great deal; 0.1 = moderate; 0.1 = little). A Cr value ≥ 0.7 was regarded as an indication of reasonable authority (36–38).

The degree of consensus of the participants was assessed using the coefficient of variation (CV) and the percentage of participant agreement with a rating of 4 (important) or 5 (essential). Mean and median scores were calculated to reflect the importance ratings. Retention of the domains, subdomains, and items required a median ≥ 4, mean ≥ 3.5, CV <0.25, and ≥ 80% agreement with a rating of 4 (important) or 5 (essential), in accordance with the literature (39–41).

Eighty percent (24/30) of the invited experts participated in the first round of consultations. Two were not available after completing the first round of consultations, resulting in a reduction of the sample size to 22 for the second round. Over half of the participants were GPs and in the age group 30 to 39 years. The vast majority were women (>75%), resided in Shanghai (>87%), and had a master's degree qualification (>72%). Over two-thirds had worked for 10 to 19 years. About two-thirds of the participants had a senior professional title (Table 1).

On average, the study participants had an authority coefficient (Cr) of 0.9 (SD = 0.1), with the smallest value exceeding the cut-off point of 0.7 (Appendix 2). Over 91% of study participants were very familiar or familiar with the study topic. About 71% made their ratings based on work experience to a great deal, compared with 42% reporting a moderate level of use of theoretical analysis. Intuition played a limited role in the expert judgement, with one-third reporting a moderate impact and two-thirds reporting little impact (Appendix 3).

The first round of consultations resulted in a removal of one sub-domain: “access to essential medicines.” Only 67% of participants rated it as important or essential, well below the cut-off point of 80%. The participants recommended the removal of this sub-domain simply because there are no effective medicines to treat MCI. The overall structure of the domains and sub-domains remained largely intact, despite the recommended re-naming of six sub-domains, removal of one sub-domain, and subdivision of two domains and two sub-domains. System support was divided into two: organizational effects and environmental support. “Healthcare setting-related factors” and “task-related factors” associated with patient engagement were moved and integrated into “organizational efforts.” More detailed descriptions regarding the sub-domains under “engaged patients” and “system support” were proposed by the experts and were adopted in the second round of consultations (Table 2).

The study participants recommended significant changes to the items in the first round of consultations, which included merging 11 items, dividing four items, removing two items, and adding 12 new items (Appendix 4). Eight items failed to reach the required 80% agreement on importance, three of which were removed as suggested by the participants. The removed items included two items measuring the removed sub-domain “access to essential medicine,” which also failed to meet the consensus criteria. Another removed item measured the “lack of control over non-pharmacological interventions by GPs” (63% agreement, 0.25 CV). Two items (“inconvenient location of health services” and “lack of testing/assessment facilities”) under the sub-domain “health care setting-related factors” were retained for the second round of consultations despite a lower than 80% agreement, since participants recommended re-categorizing and re-phrasing these items. Similarly, “peer pressure,” “perceived stigma,” and “impact on GPs from the task” were retained for the same reason (Table 3).

Twelve items were added for the second-round consultations following the recommendations of the study participants. These included “patient cooperation,” measuring opportunity for GPs; “patient follow-up,” measuring GPs' behavior; “financial concerns” and “attention to MCI demonstrated by GPs,” measuring “patient perceived barriers”; and “inclusion of MCI screening in a routine health check-up list for older adults,” measuring “health service arrangements.” In addition, four items were added to measure the added sub-domain “support of family caregivers,” and three items were added to measure the added sub-domain “patient health literacy” (Table 3).

Four items were suggested by the study participants to be sub-divided using more specific descriptions. “Confidence” of GPs in detecting and managing MCI was divided into three aspects: cognitive function assessment, training certificate in MCI detection/management, and communication with patients regarding the MCI diagnosis. “Patient awareness of cognitive disorder” was divided into two: “patient perceived susceptibility to MCI” and “patient perceived seriousness of MCI.” “Electronic screening scale” was divided into “electronic screening scale” and “electronic referral request.” “Investment in infrastructure” was divided into “establishment of a memory clinic” and “investment in intervention facilities” in addition to “investment in screening/diagnostic facilities.”

Eleven items were merged with others because of their inherent links. For example, two duplicated items measuring “caregivers' support” under “opportunity of GPs” and “engaged patients,” respectively, were merged. Belief in the value of MCI detection and the effectiveness of MCI intervention were pulled together. “Information alert system” covered both “follow-up alert” and “screening alert.” Individual responsibility and coordination were deemed equally important as an indication of effective teamwork. The three items measuring “disease-related factors” were amalgamated into the sub-domain “patient attitudes toward MCI.” All matters relating to service procedures (e.g., location, time, referral, and rewards) were amalgamated into the sub-domain “management support.”

Almost all items were rephrased using more specific descriptions after the first-round consultations. The second-round consultation rating scales contained 4 domains, 15 sub-domains, and 47 items, all of which met the consensus criteria (Table 3). No further changes were recommended by the study participants. Figure 1 summarizes the entire Delphi process.

GPs' capacity (5 items), opportunity (4 items), motivation (3 items), and behavior (5 items) were all deemed relevant or important by the expert panel of our study for MCI detection and management. Knowledge and skill acquisition are always the first step to building capacity for new medical services (42). The study participants agreed that GPs need to feel confident about cognitive function assessment and communication (disclosure) with patients. Confidence was considered to be associated with having completed a training certificate in MCI detection/management. These results are consistent with the findings of previous studies. A report from a global panel of clinicians and cognitive neuroscientists on current challenges that hamper widespread cognitive performance assessment has shown that GPs may struggle with choosing appropriate tools for assessing cognitive function (43). It is not uncommon for primary care providers to feel a lack of confidence in disclosing MCI diagnosis to patients for a variety of reasons. For example, GPs may fear that disclosing a diagnosis of cognitive impairment might damage their relationship with patients because of the potential negative impacts on the patient receiving the diagnostic information (44).

Apart from the availability of technologies and time, the expert panel suggested “patient cooperation” may provide opportunity for GPs to conduct MCI detection and management. Empirical evidence from previous studies has shown that high levels of patient cooperation are indeed associated with improved health outcomes, enhanced patient satisfaction, and better adherence to treatment (34, 45).

The expert panel agreed that belief in the value of MCI care, a sense of responsibility, and rewards such as career advancement and remuneration are motivators of GPs to conduct detection and management of MCI. These factors align well with a contemporary motivation theory. In a recent systematic review, belief in the value of work and responsibility were labeled as intrinsic motivation factors, while income remuneration and career advancement were considered extrinsic driving factors (46).

The expert panel agreed that GPs can provide a range of MCI-related services, including screening, referral, communication, intervention, and follow-up evaluation. According to a global working group comprising international experts on MCI and Alzheimer's disease, GPs should coordinate post-diagnostic interventions and track the progression of MCI to early Alzheimer's disease (43).

The expert panel agreed that patient engagement is associated with patient attitudes (2 items), perceived barriers (3 items), health literacy (3 items), doctors' encouragement (3 items), and the support of family caregivers (4 items). These results are supported by evidence derived from other studies. The HBM proposes that the health actions of patients are triggered by perceived susceptibility to the disease and the seriousness of the condition (42). Unfortunately, memory loss has commonly been considered as part of normal aging among people living with MCI (10, 47), leading to inaction.

Our study identified major patient-related barriers to MCI care, including difficulties in engaging patients in non-pharmacological interventions, social stigma, and financial concerns. Indeed, MCI interventions may not take effect until after 3 months of intervention, according to a systematic review (48). Patients may find it difficult to comply with long-term MCI interventions (49). Adding to the complexity of achieving effective patient engagement is the social stigma attached to MCI, which can lead to internalized shame or social isolation (50). In a health system yet to achieve universal health coverage, such as in China (51), it is understandable that “financial concerns” can also become a hurdle for accessing MCI-related services. Some therapies such as cognitive intervention therapies are not covered by social health insurance programs in China.

The expert panel agreed that patients may engage in MCI detection and management to various degrees, depending on their health knowledge, self-assessed health, and financial and living conditions. These results align with the definition of health literacy from a systematic review (52): “Health literacy is the ability of an individual to obtain and translate knowledge and information in order to maintain and improve health in a way that is appropriate to the individual and system contexts.”

The role doctors play in encouraging patients to engage in MCI detection and management activities is shaped by the individual commitment of the doctor, the doctor-patient relationship, and patient trust in the doctor, according to the expert consensus found in our study. Widespread distrust and poor doctor-patient relationships have recently attracted serious concern in China (18). However, according to a qualitative study in the United Kingdom, even if patients place high trust in their doctors, they may still dismiss concerns about MCI (53).

The expert panel agreed that family caregivers also play a role in patient engagement, which is associated with their knowledge about MCI, family relationships, self-perception of their role as caregivers, and the support available to them. Empirical evidence from the United States shows that a lack of knowledge and coping strategies to support people with MCI are common in family caregivers (54), although good family relationships can encourage patients to engage in daily activities that are beneficial to MCI (55). A systematic review found that family caregivers tend to be reluctant to acknowledge their responsibilities relating to MCI care due to predictable “role strains” (56). As a result, family caregivers may need some additional support (57).

The current study's expert panel categorized organizational efforts into information support (3 items), infrastructure support (3 items), management support (2 items), and teamwork (3 items). These are closely aligned with China's efforts in strengthening community management of chronic diseases, such as hypertension and diabetes (14). Indeed, information sharing is essential to ensure continuity and coordination of care, which is critical for improving patient safety and patient care outcomes, especially for those with chronic diseases (18).

The expert panel endorsed memory clinics as needed infrastructure for MCI detection and management, in addition to investment in screening instruments and intervention facilities. A study in China shows that the number of memory clinics in tertiary hospitals is considered by health professionals to be inadequate, and the situation is much worse in primary care settings (58). Empirical evidence from some high-income countries indicates that memory clinics can provide primary care providers with accessible, efficient, and cost-effective tools to handle memory problems (59).

Management support (in terms of service procedure and performance assessment) is needed to incentivise primary care workers to commit to MCI-related services, according to our expert panel. This perhaps reflects the broad management culture in China. Pay for performance based on the volume of services provided has been widely adopted as an instrument to incentivise health workers (60), despite criticism about the potential for perverse incentives (61).

GP-led teamwork has been a predominant model in community health services in China (13), which was also a necessary condition for managing MCI endorsed by our expert. The provision of “essential medicines,” however, was deemed irrelevant to MCI management by the expert panel. In addition to the lack of effective medicines to treat MCI (62), medicines listed in China's essential medicines list have been widely available and affordable from a variety of health facilities (14). This may have led the experts to believe that essential medicines are irrelevant to MCI.

The expert panel of our study was concerned about not only the health service arrangements in place for detection and management of MCI (2 items) but also the broad social environment in which health services occur (2 items). They endorsed the call to include MCI- and dementia-related services into the national package of essential public health services (63). Unfortunately, the current package promulgated in 2017 covers only hypertension, diabetes, psychosis, and tuberculosis (64). Many CHCs are providing regular health check-ups for older people free of charge. The expert panel agreed that MCI screening should be included in the check-up list. This result aligns well with the recommendations of the report from the organization Alzheimer's Disease China (ADC) (65).

Public awareness of MCI and community involvement in MCI detection and management were endorsed by the expert panel of our study for measuring the supportiveness of the social environment in detecting and managing MCI. Establishing a “friendly community” has been widely accepted internationally as a strategy to enable people with cognitive impairment to feel supported and be integrated into their local communities (66). High public awareness and social support can also be used in fighting the social stigma attached to MCI (50).