A narrative synthesis scoping review of life course domains within health service utilisation frameworks

Approach

A narrative synthesis approach was used to identify and amalgamate the relevant framework information from appropriate studies for this paper. A narrative synthesis uses a textual rather than a statistical approach for analysing results and drawing conclusions. This words-based approach is more suitable to investigating the life course domains within HSU.

A scoping review is a descriptive review, which is a broader and less defined search than a systematic review. It is suitable for literature that is large, complex or heterogeneous. A scoping review can summarise the extent, variety and characteristics of findings on a specific topic, in this case the use of life course domains within HSU, and identify gaps in knowledge (Tricco et al., 2018).

Search strategy

The search spanned a 20-year time frame (1995–2015); 1995 was selected as the starting point for the review as this was the year the major revision of the Andersen model was published in the academic literature. Email alerts were set up in the databases to ensure the review was current. A summary of the review criteria is presented in Table 1.

Databases

PubMed, Cinahl Plus, Emerald, PsycINFO, Web of Knowledge, and Scopus were used for the search. The key search terms used were health services utilization OR health services utilisation OR health services accessibility AND (framework OR model); and related MESH terms in PubMed. The searches in Scopus and Web of Knowledge were restricted to journal title and included titles relevant to health services, health policy, health management, value in health, health planning.

An example of a search string that was used in PubMed for a custom year span of 1995–2015 is:

((((health service* utilization) OR health service* utilisation) OR health service* Access*) AND Framework) OR model

Search procedure

Application of the broad search string indicated above returned 12,126 articles. See Figure 2 for the flow diagram of results.

Figure 2. Flow chart for systematic review search of models or frameworks to study health service utilisation.

At stage one and two of the screening process, many articles were excluded based on the title, abstract or full text not meeting key inclusion criteria, leaving 1698 articles for further review.

The full text of the 1,698 articles were accessed and reviewed for the framework or model used in the study. Articles were excluded at this stage (stage 2) if statistical modelling alone as opposed to a conceptual or theoretical model for understanding health services utilisation was referenced, resulting in 551 articles remaining. If a framework was named or a new framework was proposed this was also recorded. As noted earlier only one article purposively selected with the occurrence of a framework was included. A total of 164 frameworks with an associated primary article remained after this third stage of screening.

To ensure that the identified frameworks were unique and separate frameworks, and were applied in an HSU context, each full text article (n=164), whether quantitative or qualitative, was read and confirmed for inclusion by MOD (and confirmed by EB). Further duplicates or adaptations of the Andersen model were excluded, resulting in 64 frameworks remaining. Six additional frameworks were subsequently identified through email alerts, and once reviewed against the same criteria, the final number of frameworks included in the review was 70. This list of frameworks was reviewed and consensus agreed for inclusion by the entire review team.

See Table 4 for a sample of the frameworks identified. The complete list of the 70 frameworks is available on Zenodo: https://doi.org/10.5281/zenodo.2590439 (O’Donovan et al., 2019).

Frameworks

For the purposes of this review, the Behavioural model and any iteration will be referred to as the Andersen model/framework. The terms framework and model were used interchangeably and taken to be equal. Of note is that the context for use of the framework was to understand health services utilisation. Thus, the frameworks were identified as ‘used by researchers’, regardless of whether the original genesis or intent of the framework was to understand or examine this particular phenomenon.

The frequency with which the identified models were used was not recorded, as the purpose of the review was to identify the range rather than prevalence/utilization of specific frameworks. Hence one paper using the identified framework was purposively selected for further analysis. These selected articles met the criteria for the systematic review most clearly and incorporated the most complete use of the framework.

The unit of analysis was the individual’s use of services. The terms health services, utilisation and population are defined in Table 2. Inpatient, surgery and diagnostic testing were not included in this review.

Screening

There were three levels of screening, led by the first author (MOD) with consultation meetings and review at each stage involving the entire review team (the remaining authors). The review team covered expertise across the areas of HSU (PMcC), health systems and qualitative research methods (EB), health and disability statistics (HM, PMcC) and intellectual disability and health (MMcC, PMcC). Each of these experts had previous experience in conducting systematic and scoping reviews. The review team systematically discussed the merit of each framework in relation to the aim of the review as well as the context of its application to understanding HSU. This process was on-going until a consensus was reached on the final list of frameworks for inclusion.

Mapping

Once the frameworks were identified they were mapped against a matrix of HSU categories and variables. The Andersen model was the basis for ’a-priori’ categories and any variables not easily aligned with the Andersen model were categorised as ‘other specified’. The a-priori categories were the predisposing, enabling and need variables of the model. The rationale for this approach was to explore whether the Anderson model incorporated all the categories and variables of the other HSU models and if not, what these ‘missing’ variables were.

The Anderson model and the ‘other specified’ categories were then mapped to the life course domains of human agency, location, relationships and interdependencies, and temporality (Elder Jr & Rockwell (1979)). Table 3 explains each of the life course domains. This mapping exercise explored whether the frameworks incorporated a life course approach and could therefore be used in exploring HSU from a life course perspective. See Table 5 for details of the mapping.

This process of meta-synthesis was conducted by two of the authors (MOD and EB) and reviewed by the entire review team. It involved a process of sense-making and developing a cohesive structure with which to present the findings of the extraction.

Andersen and the life course

Andersen’s framework (Andersen, 1973; Andersen 1995) provides a useful basis for understanding and exploring predictors of health services utilisation. This is illustrated by the frequent application and adaptation of the framework to different health service contexts. In addition to the application of the Andersen model, which is the dominant guiding framework in the study of HSU, this review indicates that HSU in some contexts, with some populations and specific service settings is not sufficiently addressed through the Andersen model. This is evidenced by the emergence of such a range of alternative frameworks.

When assessing HSU in the context of the life course theory, Andersen is limited in its incorporation of the domains of the life course. Most notably, through the absence of issues related to human agency and temporality, limited reference to location, and some engagement with relationships and interdependencies. Though the traditional HSU model does not include core domains, recent reviews of the model indicate that the application of the model in this way has begun. For example, the domains of choice, decision-making and control are not explicitly evident in the Andersen framework, but the health locus of control concept was noted by Babitsch et al. (2012) as used in one study as a predisposing factor. This shows the potential to adapt the model to take on aspects of a life course perspective.

The particular strength of the Anderson model is very useful for considering the individual’s interaction with the health system at one point in time for one condition/scenario. Temporality as envisaged within the life course is not addressed within the model, although ‘time interval’ does feature within phase two of the model. This time interval refers to time within the health system, such as waiting time. However, the alternative frameworks propose ways to capture the temporality of health and temporality of health care need and use over the life course.

Strong across both Anderson and non-Anderson models, is the recognition of the importance of systems of support around the individual to engage with and access the health system. Relationships and interdependencies are how this is relayed in the life course approach and it is apparent in both.

Andersen identifies family and community level variables as important enabling resources to HSU with social and emotional supports recorded by Babitsch et al. (2012). Guilcher et al. (2012) recommend the inclusion of quality of relationships/emotional support with informal and formal carers to future adaptations of Andersen. Andersen introduces the concept of ‘site’ of the health service in version two and ‘external environment’ in version three of his model, again related to the health system, and are location-related variables (Andersen, 1995).

Applications of the Andersen model have used broad place variables at the level of geographic location and community characteristics such as distance travelled, urban/rural distinction and region. There is an absence of a more refined place variable such as type of living setting, condition of housing and ease of navigation between home and wider neighbourhood/community space. Place and home have been shown to be important non-determinants of health (Field & Briggs, 2001; Macintyre et al., 2003) and yet have limited expression in Andersen’s model. The model is primed to further develop these domains to reflect the life course more fully.

In the review of Andersen’s model by Guilcher et al. (2012), it is recommended that ‘home layout’ be included in future access studies for people with spinal cord injury. Applications of Andersen’s model have incorporated the idea of distance travelled to health services, but Andersen does not specify access to transport as a factor in influencing HSU.

Culture is another aspect of the life course incorporated into HSU frameworks. Babitsch et al. (2012) include culture as an additional domain when mapping previous applications of Andersen. Culture is not a specific domain in the Andersen model but can manifest under the social structure aspect of the pre-disposing factors. In addition, cultural factors have been identified as enabling in application of the Andersen model. It is important to consider that the role of other people and cultural and social norms, may in fact be prohibitive to using health services and engaging with the health system.

This scoping review has identified a number of frameworks that illustrate how elements of the life course can be incorporated into the study of health service utilisation, although no framework included all aspects. The additional frameworks identified in this review, demonstrate commonality in variables with the Andersen model. These commonalities are illustrated in Table 5 (column 2).

The unique variables identified in these frameworks have a strong propensity towards the human agency aspect of the life course and position the individual as a more active participant in their health care and life situation. There is also added focus on the timing of care and temporality overall which is not easily identified in Andersen and which the current analysis considers absent.

The life course and HSU

The WHO (2002) perspective of the individual as an active participant in their health and healthcare, is evidenced in the findings of this scoping review. The identified frameworks positioning the individual at the centre of their own lives and in particular their health experience. Choice, decision making and the individual demonstrating agency over their health experience and HSU experience is explicit.

Selecting among competing priorities is a judgement that individuals face in everyday life and in the healthcare context. Such competing decisions also exist at a health system and health policy level. This cost/benefit judgement will implicitly impact on if, when, how and the type of health services accessed by individuals and decisions to forgo required health services. Influencing factors at these decision points include the social network and connections that the individual has, as well as family, community and broader cultural and societal beliefs and values.

In particular, the role that environmental factors play in the HSU experience of minority or more vulnerable populations, such as health equity, healthcare reaching and cultural and social capital (Nauenberg et al., 2011) is highlighted, and includes issues of discrimination and stigma as well as legal status of minority groups within the community (Lee et al., 2009).

Social interconnectedness (Levy et al., 2014), psychosocial (Nauenberg et al., 2011) and affective supports (Witucki & Wallace, 1998) are also important influences on HSU. Availability of supports, in addition to the availability of services, is another strong theme within the reviewed frameworks and align with the interdependency and relationship domain of the life course. Andersen’s predisposing domain could incorporate social relations, but the context or nature of the role of social relations in HSU is unclear within the model.

The importance of the provider-patient relationship and ability to communicate and establish rapport in the HSU process was noted and highlighted, particularly for immigrants who were non-English speaking (Jo et al., 2008). Broader issues of human rights, feminism, and stigma are included within the interdependency domain as they impact on the status and roles an individual has within a given society and how others in society interact with the individual.

Cultural beliefs and folk values also present a type of relationship and interdependency that informs HSU for some people. The influence of the individual’s support networks and the inter-relationship between the individual and the broader cultural belief system within which they live or have lived has been shown to influence type of health care, such as decision to use traditional or complementary health services compared with mainstream health services. The heightened focus on cultural factors and distinguishing culture from community is an important addition.

Andersen does not address cultural factors as distinct from community, however reviews of the Andersen model have identified the application of culture in the use of the model (Babitsch et al., 2012) under the rubric of contextual factors. Similarly, social relationships are not explicitly mentioned in the Anderson model, but are implied through its application.

Adopting a life course perspective to HSU enables a disaggregation of the influence of the family unit, which changes over the trajectory of the life course. This impacts on the use of health services through availability of support from other people to access services. Others knowledge, attitudes and experience of the health system, predisposition to mainstream or alternative, complementary or culturally specific values, and beliefs of appropriate health system to engage with, can also affect HSU. The inclusion of living status by Toloo et al. (2013) in examining HSU, that is whether the person lives alone or with someone else (partner, children, family etc.), highlights the link between place and people.

Under location, the need to include a more refined and ‘local’ place variable to better understand HSU is highlighted. Though the location of the health care interaction needs consideration, so does the type of housing, structural issues, living situation, neighbourhood and community factors. These place variables, which were identified and differ to that of site of health service and distance to travel, which are included in Andersen, are shown to be useful additions. Transportation (Messer et al., 2013; Shaw et al., 2013) highlights the relevance of how an individual can navigate the space they live in to access services. Alborz et al. (2005) also indicate service setting as an important contributor to HSU, specifying the need to set characteristics of opening times and waiting times in understanding use. In addition, the home setting and social connections within the home are also relevant locational factors. Where an individual lives and who an individual lives with, as well as how geographically close to networks of support they are, acknowledges the role of the social determinants of health (Dahlgreen & Whitehead, 1991; Emerson et al., 2011).

Temporality is the fourth domain of Elder’s life course perspective but there is limited focus on time and the impact of life events over time in the frameworks. The absence of time illustrates the continued tendency to view health access at one point in time. Of the four models with some reference to temporality, only one explicitly named time. However, life changes and life events were identified in others as potential trigger points for health service use. The inclusion of a time variable acknowledges that health is impacted by early life events as is current access and use of appropriate services (Norris & Aiken, 2006). HSU is not static. It is a complex process, involving multiple service interactions at any one time. Interaction with one element of the health system may be a trigger for later interactions with the same or other aspects of the system. Hser et al. (2007) introduces the concept of duration of use in addition to frequency and highlights the importance of capturing HSU data over multiple time points.

As noted above, not all of the frameworks identified are traditional HSU models but have been applied to understand, explain or predict service utilisation in the health sector. Of note, the health belief model was widely used when the service focus was screening and health checks, with studies of emergency care and ambulance service use more likely not to use a framework. It could be implied that the move beyond HSU frameworks is an indication of the limits of current HSU frameworks, including Andersen.

The identified models of HSU add to the understanding of HSU by including the individual as an active participant in health and utilization of services. This is demonstrated through the focus on decision-making and choice elements. The individual utilization of healthcare does not exist in isolation. HSU is demonstrated as informed by the geographical, social, temporal and cultural context within the person’s life, and is more deeply explored in these models.

In addition to the differences between these models and the Andersen model, some similarities were identified, with many of the reviewed frameworks including variables that map to the traditional domains of the Andersen model; predisposing, enabling, need, health system characteristics and outcome variables. Most notable was the prevalence of predisposing and enabling variables. However nearly 40% (n=27) of the frameworks included variables that did not easily map to Andersen’s framework. See Table 5, column 3 to identify the variables that were unique to other frameworks.