Classification of theories according to level of abstraction
7.1.3 Theory Development Strategies
A researcher must use a mixed methods approach whenever they begin the research process with an inductive approach and later test the hypothetical model with quantitative methods (see Chap. 4). As seen in Fig. 7.2, it is possible that the researcher will need to return back to the inductive strategy and/or data at some point during the theory development process. A relevant example will be presented later in this chapter.
7.1.4 The Structure of Theory
Theory comprises concepts, their definitions and relational statements [2, 4–6]. Hence, a concept is the basic unit of theory. Concepts provide symbolic statements about the observed phenomenon, and are formulated in words that enable people to attach meaning to phenomena that can be directly or indirectly seen, heard, tasted, smelled or touched. In this context of theory, a concept may be a word (e.g. grief, power, pain), two words (e.g. job satisfaction, role strain), or a phrase (e.g. maternal role attachment, health-promoting behaviour). The concepts can be abstract (social support, personality) or concrete (e.g. chair, red colour) depending on the level of theory. However, if a researcher is interested in quantitatively measuring a concept, or verifying their structure, they need to operationalise the concepts so that they can be empirically tested (See Chap. 9).
Adherence is an active, intentional and responsible process in which patients work to maintain their health in collaboration with health care staff.
Collaboration means working with health care providers such as physicians and nurses.
Collaboration means that a patient regularly comes to meet physicians. If the patient cannot come, he/she will book a new appointment. The patient and physician plan the care together during these appointments.
Empirical indicators (which can, for example, be items in an instrument)
(1) I regularly visit my doctor or nurse; (2) My doctor works with me in planning a treatment that will suit my life; (3) My nurse works with me in planning a treatment that will suit my life.
Statements and hypotheses can indicate the connections between presented concepts. Statements provide suggestions of relationships on the general level, while a hypothesis is formulated to test the connections between concepts. The testing of a hypothesis can be extended to the testing of a theory (see Chap. 9). Theories are tested to study the reliability and validity of the theory, assess the usefulness of the theory, or further develop the theory [4, 6, 12].
7.2 The Creation of Concepts Through Content Analysis
As mentioned before, concepts are the basic element of theory. Inductive content analysis is a useful method for creating concepts that will be tested later or integrated into a theory. Chapter 2 described how content analysis can be applied to create concepts that will answer the research question, while this chapter will explain how these concepts can be used to create a hypothetical model.
7.2.1 The Theory Development Process: From Inductive to Deductive
This section provides a brief example of the theory development process. The first part of the section demonstrates how a hypothetical model can be created based on content analysis while the latter part of the section describes how quantitative methods can be used to create an instrument for testing the hypothetical theory, and how these results underlie the creation of theory. Theory testing is presented in more detail in Chap. 9. This example requires some understanding of inductive content analysis (see Chap. 2) as the concepts included in the presented model were identified through content analysis.
188.8.131.52 Phase 1: Inductive Development of a Hypothetical Model
184.108.40.206 Phase 2: Development of an Instrument to Test the Hypothetical Model
An instrument was then developed to test the hypothetical model. The items of the instrument were derived from the open codes (a product of content analysis) that were organised under the concepts included in the hypothetical model. The items in the questionnaire were formatted according to the original open codes. For example, one of the items was “diabetes causes me to worry about the future”. The instrument was then tested with data collected from 12- to 17-year-old adolescents with type 1 diabetes (n = 91), after which factor analysis was performed and face validity, correlation coefficients as well as Cronbach’s alpha were calculated to assess the validity and reliability of the instrument (see Chap. 8 for the instrument development process). A comparison of the adolescents’ self-evaluated adherence to health regimens and long-term blood sugar test results served as another measure of validity. The results of these assessments showed that the instrument is valid and demonstrates high reliability [7, 14].
220.127.116.11 Phase 3: Testing the Hypothetical Model
To test the model, data were collected from 346 adolescents with type 1 diabetes using the previously developed instrument. During this stage, a LISREL analysis—which is based on linear structural models—was used to create a MIMIC (multiple indicators multiple causes) model. A MIMIC model consists of two parts: a measurement model, which defines the relationships between a latent variable (here adherence) and its indicators; and a structural model, which specifies the relationships and causal effects between the latent variables and indicators. According to the MIMIC model, adherence was defined by self-care behaviour, responsibility for care, intention to perform self-care and collaboration with health care professionals. Furthermore, adherence to health regimens was shown to be strongly influenced by motivation, the results of care and having enough energy and willpower for care, while a sense of normality and the fear experienced by young diabetics exerted a weaker, yet positive, effect on adherence .
18.104.22.168 Phases 4: Expansion of the Model
The theory development process may revert from a deductive approach to an inductive approach when a researcher returns to the original data (Fig. 7.2). This happened in the presented example, as the researcher was not content with the content of the MIMIC model. Analyses that use a MIMIC model rely on the statistical testing of a hypothetical model, and in this case, the researcher felt that certain factors which were not identified in the MIMIC model, for example, support from health care providers and parents, are nevertheless important to determining adherence to health regimens. For this reason, the researcher re-analysed the qualitative data and quantified concepts for statistical analysis (see Chap. 4). These quantified concepts were then used to construct a model of good adherence and the related factors. The connections between concepts were analysed by cross-tabulation and stepwise discriminant analysis, with the results showing that support from parents, friends and health care providers do not directly explain adherence, but explain motivation and energy and willpower to take care of oneself, both of which directly explain adherence . This is why these factors do not exist in the MIMIC model, which only verifies direct connections between concepts .
22.214.171.124 Phase 5: Construction of a Theoretical Model of Adherence
In the next step of the theoretical model development process, the MIMIC model (phase three) and the model of good adherence (phase four) were combined to create a theoretical model of adherence among adolescents with type 1 diabetes. The theoretical model was then tested. However, this required certain modifications to the validated instrument. The reason for this is that the validated instrument was based on the hypothetical model, which includes more concepts than the theoretical model.
126.96.36.199 Phase 6: Development of an Instrument that Can Be Used to Test the Theoretical Model and Build Theory
The hypothetical and theoretical models included different amounts of concepts because some of the concepts in the hypothetical model were not verified through statistical analyses. The items included in the modified instrument should only measure concepts which are in the theoretical model . The data collected from adolescents with type 1 diabetes in phase three was used to test the instrument. The items included in the modified instrument were chosen based on a LISREL analysis, with the requirement that the covariance coefficient >0.40. For example, the initial instrument included five items that measure motivation, but the modified instrument would only include items that have a covariance coefficient >0.40. Following this process, the modified instrument included 32 items that measure the concepts presented in the theoretical model. An additional 12 background questions were also added. Content validity was assured by two diabetes nurses and five adolescents with type 1 diabetes. Based on this evaluation, nine more items (a total of 41 items) and one more background question (a total of 13 background questions) were added to better cover the content of the concepts presented in the theoretical model. The instrument was then tested with 13- to 17-year-old adolescents with type 1 diabetes (n = 30), with the data analysis demonstrating satisfactory correlation coefficients (>0.40). As the aim was to test the theoretical model using not only adolescents with diabetes, but also adolescents with arthritis, epilepsy and asthma, the self-care items were modified to be suitable to these other patient groups, after which content validity was tested with clinical experts as well as adolescents with asthma (n = 10), epilepsy (n = 10) and arthritis (n = 10). This resulted in an instrument that can be used to test a theoretical model of adherence among adolescents with type 1 diabetes, asthma, epilepsy and arthritis.