The TPB, along with other social cognitive models [42] and specifically in relation to health actions [43], assumes that any given behaviour is a functional outcome of the attitude – intention – behaviour sequence, which makes the model suitable for understanding and predicting volitional health-preserving or improving behaviour. The efficacy of the TPB in relation to health-related behaviour is dependent on the particular behaviour (i.e. adhering to health improvement or engaging in risky activities), age, time-frame and measurements [44]. Improvement made in order to increase the predictive power of the TPB model is abundant, including descriptive norms [45], subjective norms [46], self-identity [47], self-efficacy and locus of control [48], behavioural control [49], anticipated regret [50,51], desires and emotions [52], moral norms and anticipated affect [53], social cognition properties [54], prototypes and willingness [55], conscientiousness [56] and goals and its properties [57]. The variations of the augmented TPB models acknowledges the complexity of factors underlying the ultimate behavioural choice but nevertheless assume that the behaviour is predicted by some combination of social cognitive factors about the behaviour. Whilst each added variable incrementally increases the predictive power of the TPB model, it inadvertently introduces some unexplained variances with each predictor, thus the price to be paid for the improvement manifests in complexity. Meta-analyses have shown that beliefs and attitudes in the TPB model predict 39% of the behavioural intention and 27% of the actual behavior, with notably stronger prediction when behavior was based on self-reports [58]. In order to close the gap between behavioural intention and actual behaviour, Gollwitzer and Sheeran [59] proposed the addition of implementation intention.

Approaching behaviour from a different angle, the Control Theory [60] proposes that behavioural choices are driven by goal pursuits, where desired goals induce intentions but also serves as reference values against which the progress of achieving the goals are constantly monitored and adjusted if necessary. The allowance for incremental approaches to the desired goals connects goal theories of behaviour [57] and implementation intentions [59] via self-regulation by which people control their feelings and impulses to ensure behavioural outcomes that are perceived to be desirable. Self-regulation is a complex process that includes long-term perspectives in constant monitoring attainments [60] as well as resisting temptation with the view of pursuing long term goals [61]. The self-regulatory effect is thought to rely on a limited energy resource that can be temporarily depleted, hence exerting effects on other behavioural choices made simultaneously or shortly after [62].

Investigations of the role of self-regulation in long-term medication adherence [63] revealed that self-regulation not only accounted for the largest explained variance in the self-determination theory-based model but also was the only variable that showed significant correlation with both self-reported adherence to prescribed medication and pill-count. Self-managed care based on a self-regulation framework has been shown to be effective in health-maintenance among people with chronic health conditions including improvement in condition, along with better physical and psychosocial functioning [64]. Other examples for interventions based on self-regulation theory being successful are related to reducing risk of coronary heart disease [65], reducing depression among people suffering from rheumatoid arthritis [66] and end-stage renal disease [67].

Self-regulation is present in peoples’ daily lives because discrepancies between expectations and reality in situations people encounter are inevitable. Successful adaptive functioning is based on effective self-regulation by which people control their impulses, thoughts, feelings and adjust their behaviour to close or minimize the gap between their expectations and present experiences. The information processing models of self-regulation focuses on the process of constant comparison and feedback between the current and the desired, informing the necessary adjustment required to the behaviour [68]. However, the way people carry out this task is a reflection of their individual dispositions. General personality dimensions and specific self-referent personality traits are thought to be inseparably intertwined with self-regulation [69,70]. General personality dimensions such as conscientiousness [71], impulsiveness vs. constraint [72] and specific self-referent traits such as self-efficacy and self-esteem [73] are key components of the self-regulative process.

Every step of the self-regulative process, which can manifest in active initiation and maintenance of a goal-oriented and effortful activity or inhibition of impulsive behaviour that could jeopardize reaching the desired state, is influenced by the personality factors. Individual differences determine people’s dispositional capacity for self-regulation, as well as affecting the dynamics of the regulatory processes [69]. Stable individual differences determine the self-regulatory strategies individuals automatically employ [74]. Personality traits affect how ambitious the goals are; how discrepancies are viewed; and what self-regulatory strategies are employed. Personality also influences whether people exhibit flexibility in behaviour to achieve the desired goal or increase effort in following the behavioural path they set themselves on, initially assuming that the discrepancy between the desired and current state is to be bridged by trying harder or one is motivated to find alternative ways to reach the desired goals. Integrating trait and process variables self-regulation offers a more comprehensive account of self-regulation [75].

Demographic factors such as gender and age moderate the effect of personality factors on goal setting, self-regulation and behavioural antecedents. Although personality trait changes occur most dominantly in young adulthood (20-40years of age), they gradually change at a very slow pace over the lifespan: self-control generally increases with age [76-78], along with conscientiousness [79], self confidence and emotional stability [78]. Gender differences are small relative to individual variation within genders and broadly conform with gender stereotypes of males being more assertive than females who perceive themselves higher than males on agreeableness, warmth and neuroticism while being open to feelings, rather than ideas [80] but no difference exists in impulsiveness, locus of control or orderliness [81]. Males possess slightly higher general self-esteem compared to females [82]. On the contrary, ethnicity within the same culture does not seem to exert a significant influence on behaviour but rather culture is seen as aggregate personality profiles [83] that manifest in typical behaviour of most, but not all, people in a given country [84]. Although cultural differences in broad personality traits have been consistently shown in cross-cultural studies [85], within culture differences in personality traits by ethnicity were found to be generally negligible [86].

The integration of the relevant goal-intention-behaviour theories is depicted in Figure​Figure1.1. It shows a conceptual map that underpins the current investigation in adherence to prescribed exercise and treatment regime among HIV positive patients. Light blue arrows show the behavioural intention model based on TPB. Arrows coloured in dark blue and dark burgundy lay out the path for the goal intention model, where explicit and implicit goal intentions are either an antecedent (Goal intention → Behavioural intention → Adherence, shown in dark blue) or a moderator (Behavioural intention → Goal intention → Adherence, shown in dark burgundy) of the behavioural intention. In the latter case, we expect the strength of the path between behavioural intention and adherence being diminished or weakened by the inclusion of goal intention between behavioural intention and adherence [87].

A person’s ability to follow a long-term interest via self-regulation plays a vital role in medical treatment, particularly in conditions that require prolonged health management (e.g. chronic conditions or HIV). Although many segments of self-regulation appear to be in the unconscious and automatic domain, health improvement or maintenance requires conscious, effortful and goal-oriented self-regulatory behaviour. That is not to say that unconsciously held motivators do not play an important role in consciously self-regulated behaviour. Dual process models [88] suggest the existence of a two-tier system where the higher-order system is responsible for reflective and planned behaviour whilst the lower-order system responds to associative cues at the moment [89].

Automatic processes occur without intention or awareness whereas controlled processes are based on intentional effort to either seek information in order to choose the appropriate action or to inhibit automatic processes in order to execute the selected action. Clearly, implicit automatic and explicitly controlled processes are qualitatively different entities that independently contribute to regulating human behaviour [90,91]. Research in implicit social cognition provides ample evidence that thought processes outside the conscious domain influence behaviour and that the processes under conscious control constitute only a small fraction compared to what is present in decisions about behaviour [92]. The distinction between automatic implicit and self-controlled explicit processes has been evidenced in motivation [93], attitudes, stereotypes, self-esteem and self-concept [94,95] with considerable individual differences in awareness of and ability to control automatic processes [92,96] being evidenced. Despite this, most policy development narrow-mindedly assumes that behavioural choices are exclusively under the controlled processes when devising preventive measures and interventions to encourage the desirable behaviour [97]. In order to improve the effectiveness of such measures, social policies aiming to behavioural change should incorporate assessment of both implicit and explicit processes [97].

Therefore for the current project it is proposed that explicit assessment is complemented with implicit associations [98] where appropriate. It is known that individuals may be biased in how they see themselves and this bias is often not recognised by the individual [99]. Self-reported, explicit assessments of individual differences are particularly prone to distortion, which may or may not be deliberate. A current research stream in psychology focusing on the uncontrolled, unconscious processes [95] provides a promising avenue to compliment the explicit assessments. Based on the widespread definition of attitude as a relatively stable tendency to evaluate an attitude object with some degree of like or dislike, implicit attitudes are defined as associative processes reflecting this tendency [100]. Associative evaluations are the result of stimulus driven, uncontrolled, unintentional, goal independent or unconscious processes [101] and as such, they do not require respondents to be aware of these attitudes and may be free of self-presentation distortion in some key cognitions such as conscienciousness [102]. Recent research has shown that implicit association adds to the predictive power of explicit attitude measures [103] via one of eight possible mechanisms [104]. Based on the assumption that implicit associations either predict or moderate the relationship between antecedents and actual behaviour, Implicit Association Tests (IAT) have been applied to understand consumer behaviour [105-107], job satisfaction among nurses caring for drug addicts [108], doping behaviour [109] academic performance [110], suicide attempts [111,112], addiction [113], dietary change [114], dental hygene [115] or HIV-risk sexual behaviour [116], illustrating the widespread use of the IAT concept. In the current project, using the Brief IAT (BIAT) concept [117], implicit measures include self-efficacy, attitude and norms. Implicit tests use exactly-matched content to explicit measures ensuring that both explicit and implicit measures are equally and directly related to the concept [118,119].

Sample size calculation was based on the results from a previous study [17]. Finding statistically significant difference/relationship with medium effect size (d=0.3) at alpha error probability=0.05, 1 beta error probability=0.8, two tail analysis for the FAHI Quality of Life (QL) measure [129] between two groups (independent sample): n=44 (combined sample); FAHI QL measure change in the adherent group (repeated measure): n=10 (combined sample); and fitness index change in the adherent group (repeated measure): n=74 (combined sample). The difference in fitness baseline measure between adherent and non-adherent patients was very small.

The recruitment is staggered and continuous. The total timeframe for each patient involved in three testing sessions is 25weeks, including post testing and allowing for delays owing to personal circumstances. Each testing session will last for approximately 90 minutes with short breaks between the implicit assessment test blocks or as required during the completion of the questionnaire and physical testing. To avoid cognitive fatigue, computerised and paper psychological tests are interspersed between the elements of the physical tests.

Participants are asked to donate a hair sample at T2. The hair sample (a diameter of 3 to 4 mms, approximately 50 hairs), 3-4cm in length and preferably untreated is taken by the attending physiotherapist cutting near the skin surface using a pair of blunt edge scissors in appropriate privacy to observe local patient dignity standards. Depending on hairstyle, if possible, cuts are taken from different locations of the head to avoid bald patches. Hair samples are catalogued by the ID code and stored in a paper envelope tested for the presence of selected Combination Antiretroviral Therapy (cART) drugs and used to verify adherence to medical treatment.

Data on individual patients’ medical record, physical assessment, attendance, psychological tests and hair sampling are linked via numerical codes. Coding is undertaken by the physiotherapy researchers. Other researchers in this project only have access to the coded (hence anonymous) data set. Data collection and analysis are completely separated to negate any researcher bias. Owing to the staggered start-finish of the data collection, the full dataset is not analysed until the data collection is complete.

The project has no foreseeable risk to the participants and their identity is fully protected. The computerised and self-reported psychological tests are coded by the attending physiotherapist and although they present a cognitive challenge, they are completed in segments between the elements of the physical tests. The physical assessment and the 10-week exercise programme are part of the prescribed treatment. None of the physical tests are to exhaustion.

Involvement in an exercise programme could potentially cause temporary pain, discomfort and/or inconvenience but it is part of the standard treatment and the Trust has a generic risk assessment for gym based exercise. In this study, the risks are minimised by prescribing clinically indicated simple exercises that are suitable for each participant and that each participant is observed doing the exercises under the supervision of a state-registered physiotherapist. Participants may feel inconvenienced at attending pre- and post assessment sessions, however, their time is compensated with gift vouchers. The research design does not have any risk related to its activities.

Testing: computerised testing could have potential acute cognitive fatigue that may affect executive function in an immediate follow-up task that would also require executive control. This has no bearing on normal functioning and should recover within 10 minutes. Although the questionnaire contains some sensitive questions as part of the standard quality of life assessment (e.g. satisfaction with sexual life), they will be analysed collectively at the group level (adherent/non-adherent) and will not be discussed face to face with the participant.

The researchers have no conflict of interest. Psychological testing is added to the exercise programme provided within standard care. Researchers will have no direct benefit or detriment from the outcome of this study. Knowledge generated in this project will serve future studies into treatment adherence, which in turn may lead to increased functional capacity and physical fitness, better heath and improved quality of life. Although there is no direct benefit to participants from the research component (psychological testing), their contribution will lead to a better understanding of motivating factors and barriers to adherence and inform future intervention strategies and support. The level of service provided (standard care) is not affected whether patients take part or not in this study.