Knowledge about the adherent or non-adherent patient is predominantly sought through analysis of the many 'factors' which are potentially predictive of adherence or non-adherence. These include socio-demographic factors; attitudes and beliefs; the nature and effects of the medication; the quality of the doctor-patient relationship; and social isolation and support. Typically, significant associations with adherence/non-adherence are sought from descriptive studies.

This approach has a number of limitations. Its reduction of the social world into a set of standardised indicators disguises the complexity of forces at work. It also fails to understand the subject of the research - the treatment-taking individual - as an active, decision-making agent. The application of this approach to understanding adherence to HAART has produced a large number of papers, but their results often lack consistency and in some cases are contradictory. The following review draws attention to these problems.

An emerging approach within HIV treatments adherence research is to look at deliberate and accidental non-adherence to medication. Drawing on a patient-centred approach, researchers try to identify how patients understand adherence and their own behaviour as a precursor to a more comprehensive understanding of the causes of non-adherence. This review also presents prominent examples of this type of adherence research, which utilises self-report and patient narratives.

Sociodemographic factors

It is unusual for socio-demographic factors to be associated with adherence to treatment for any chronic illness (Eraker 1984). In the context of HIV, many studies have looked for a relationship between sociodemographic characteristics and non-adherence, but results have been inconsistent, pointing to the difficulty in predicting non-adherence from sociodemographics.

Non-adherence occurs in all groups and classes of people, and it is widespread and unpredictable. Furthermore, adherence varies between individuals who share certain sociodemographic features, identities or behaviours, as well as within the same individual over time (Carrieri 2002).

In a review of adherence to HIV clinical trials, none of the following variables were found to be consistently predictive of non-adherence: age, race/ethnic group, sex, marital status, religion, socioeconomic status, level of education, or occupation (Besch 1995).

However recent research into HIV treatment adherence outside the clinical trial setting presents a slightly more varied picture:

• Gender is consistently unrelated to adherence (Rodriguez-Rosaldo 1998; Weidle 1999; Johnstone 1998; Anderson 1999; Holzemer 1999).

• Younger age has been linked to non-adherence in two large studies (Klosinski; Anderson 1999; Howard 2002). Other studies have not shown any age-related differences (Singh 1996; Rodriguez-Rosaldo 1998; Weidle 1999; Holzemer 1999).

• Race/ethnicity has been shown to be predictive of adherence, although most of the relevant studies were undertaken in American cities where patterns of ethnicity are likely to be very different from in the UK.

Black race and/or ethnic minority has been associated with poorer adherence in several studies (Muma 1995; Ohmit 1998; Singh 1996; Laine 2000; Golin 2002; Power 2003; Johnson 2003). However, other American studies have not found this association (Johnstone 1998; Holzemer 1999). Singh and Muma speculate that their results are likely to reflect a combination of factors including differences in socio-economic status, language and health beliefs; conflicting concepts of illness; less developed networks of educational and social support structures; and the effects of the social stigma that may accompany diagnosis with HIV infection.

Language and cultural barriers within the health system may also contribute to poor adherence (Murphy 2003). A recent large UK study found lower adherence associated with Black African ethnicity, but again there was no evidence for the specific causes of this association (Anderson 1999).

A history of intravenous drug use has frequently been associated with adherence (Nakashima 1998; Ohmit 1998; Stone 1998; Moatti 2000; Cook 2001; Johnson 2003) although in a recent large descriptive study, these differences were shown to be modest (Turner 1998).

There is some evidence to suggest that current illicit drug use, rather than previous drug use, is associated with poor adherence (Pratt 2001; Golin 2002; Howard 2002; Palepu 2003; Power 2003). However, several other studies have not shown any association between adherence and drug use (Broers; Carrieri; Ferrando; Weidle; Anderson 1999; Holzemer 1999). One study of alcohol and drug abuse has found that current drug use is associated with poor adherence but when people stop using drugs they don't have ongoing adherence problems (Lucas 2002). Another study which looked at people with a history of illicit drug use, found that active drug use and fewer than two meals per day were associated with poor adherence (Gebo 2003). Not surprisingly, active drug use which undermines adherence is associated with poorer response to therapy (Palepu 2003).

Illegal or injected drugs are not the only culprits; many studies have found that high alcohol consumption is associated with poor adherence (Spire 2002; Duran 2001; Brigido 2001; Murphy 2002; Lucas 2002; Halkitis 2003; Golin 2002; Howard 2002; Power 2003; Carrieri 2003). A study of 1910 people conducted in California calculated the risk of non-adherence associated with particular drugs and patterns of drug use. While cocaine, amphetamines (e.g. speed, ice, ecstasy) or frequent heavy alcohol use more than doubled the risk of non-adherence, use of marijuana, sedatives or moderate or heavy alcohol consumption increased the risk of non-adherence by half (Tucker 2003).

An American study which followed people over three years has provided a new perspective on the possible association between recreational/illegal drug use and poor adherence. This study found that lack of stability, rather than drug-taking or the presence or absence of certain demographics per se, significantly increased the odds of non-adherence. The study involved 435 HAART-treated individuals who were registered with the New York State Treatment Adherence Demonstration Project between January 1999 and March 2002. The study looked at five factors: drug use; stress levels; housing status; belief in the efficacy of HAART, and alcohol use.

A relationship between all the five factors and non-adherence was found, but more importantly the investigators found that it was lack of stability in housing, stress and drug and alcohol usage which most significantly increased the odds of non-adherence. Patients who started to use drugs in the course of the study had an odds ratio of non-adherence of 3.18 and individuals who started to drink alcohol frequently had an odds ratio of non-adherence of 2.92. Adherence in stable drug users, who reported drug use at baseline and follow-up visits was, by contrast, not significantly different from the level recorded in non-users (Tesoriero 2003).

Several other recent studies have found that social or family problems often play a part in the failure to take medications as prescribed. A French study has reported similar findings to the New York study, with social instability reducing adherence to medication among ex-users but not among current addicts (Bouhnik 2002). Being homeless in the previous year (Johnson 2003) and serious social or family problems (Palmer 2003) have also been linked to poor adherence.

As the work on instability suggests, there are likely to be a multiplicity of underlying factors which contribute to the association between drug use and poor adherence. Qualitative studies have described several factors: ambivalence towards the treatments; destructive coping strategies (such as relapsing into using street drugs); confusion about the dosing schedule; and worries about interactions with street drugs (Demas 1995; Freeman 1996; Spire 2002). However, it is likely that intravenous drug users are principally influenced by factors similar to other populations in their ongoing adherence to therapy (Jones 1998).

Other demographic factors which have consistently not shown any relation to adherence include education level (Weidle 1999; Klosinski 1998; Singh 1996; Stone 2001; Anderson 1999); being in paid employment (Anderson 1999; Singh 1996); religious affiliation (Johnstone 1998; Singh 1996); and socio-economic status (Bangsberg 2001; Rubin 1997; Klosinski 1998). Lower income and less education have been linked to poor adherence in one study (Golin 2002).

The difficulty of making sense of specific demographic differences (such as those partially seen for race/ethnicity and intravenous drug use) demonstrates the weakness of these data in general, for such differences are likely to encompass a range of potential causes of non-adherence.

Equally, the lack of any statistical difference may disguise specific causes: although gender consistently fails to be associated with levels of adherence to HIV, qualitative work on women's experience in maintaining adherence has described gender-specific problems that women face in managing complex treatment regimens, not least the demands of parenting and child-care (Acero 1997; Murphy 2002). But such problems are invisible when the experience of all women is tested against the experience of all men.

Mental health and psychological well-being

Psychological and psychiatric problems have been commonly found to be associated with lower levels of adherence to HIV medications. However care needs to be taken to distinguish between the many different meanings and definitions employed in this literature, ranging from clinically defined psychiatric illness to aspects of quality of life and self-reports of personal feelings.

Among studies which use a clinical definition of mental illness, three studies have found differences in adherence to HIV medication between those who either did or did not require psychiatric consultation (Ferrando 1996; Broers 1994; Spire 2002). However, some individuals with serious mental illness manage excellent adherence to treatment, so factors other than psychiatric conditions may be contributing to non-adherence (Wagner 2003). There is evidence that different psychiatric conditions have different impacts on adherence: borderline personality disorder has been linked to non-adherence (Palmer 2003), as have depression (Carrieri 2003) and social phobia, but anxiety disorders may improve adherence to antiretroviral medication (Ingersoll 2004).

Psychological inventories are commonly used to assess mental health. One study found differences on three scales, measuring depression, emotional disturbance and coping (Singh 1996). However a study specifically investigating psychological well-being and HIV treatment adherence (Holzemer 1999) which used five different scales measuring depression, psychological distress, and aspects of quality of life found treatment adherence to be related only to quality-of-life measures such as 'feeling comfortable and well-cared for' and 'having a meaningful life'. Other studies have found adherence associated with 'a highly satisfactory emotional life' and 'excellent mood' (Martini 2000); and non-adherence associated with 'extreme levels of anxiety' (Klosinski 1998). 'Feeling depressed' defined by patient self-report rather than clinical diagnosis has been associated with poor adherence (Pratt 2001).

Severity of depressive symptoms was correlated with non-adherence in a study of 135 Italian HIV-positive patients, but was more strongly correlated with complaints about impairment of sexual performance (OR 1.05 per scale point versus OR 6.62) (Ammassari 2004).

A study of 456 gay and bisexual men in New York has found that men who cope with problems through avoidance, drink a lot of alcohol and have difficulty in communicating with sexual partners about HIV miss more doses of HIV medication than other HIV-positive men (Halkitis 2003). The authors suggested that the desire to escape from reality is triggering poor adherence.

Social isolation and support

There is considerable general literature which draws attention to the associations between social isolation or social support with treatment adherence (Levy). Although operating principally as an indicator of the quality of the immediate social environment, social isolation is also closely linked to the theoretical framework of psychological well-being.

One large quantitative study found that two measures of social support (SF36 and a report of perceived social support) were both associated with adherence, although the measures only explained a modest amount of the differences between adherent and non-adherent participants (Holzemer 1999; Bakken 2000). Lack of social support has also been linked to poor adherence in a cross-sectional study of 445 people in the French APROCO cohort (Spire 2002), while substantial social support has been associated with good adherence (Malcolm 2003). Living alone has been found to be associated with non-adherence (Morse 1994; Pratt 2001), as has the lack of a close confidant (Pratt 2001); whereas living with a partner or perceived satisfactory support from a partner has been associated with improved adherence (Anderson 1999; Power 2003). While this suggests intimate relationships are important in adherence, broader support from family and friends may not have a significant impact on adherence (Power 2003).

Although these results indicate a general principle about the importance of social support in maintaining adherence, they do little to explain what kind of support is appropriate. A qualitative study has found that although participants valued social support, especially in overcoming side-effects and the difficulties of taking the drugs, those who provided support both assuaged and created problems. In particular, many participants reported feeling under intense pressure from peers, family and medical providers to take therapy, resulting in a fear of failing to meet expectations and an unwillingness to be open about problems that were encountered (Kunches 1998). Also, another study has found adherence associated with excellent family life but non-adherence associated with contacts with other seropositive people (Martini 2000).

Attitudes and beliefs

Most research into attitudes and beliefs uses the 'Health Belief Model' (Rosenstock; Becker). This model posits that the likelihood that clinical advice will be followed is a function of a range of individual perceptions:

• Individual perception of susceptibility to disease.

• Individual perception of disease severity including presence or absence of symptoms.

• Individual perception of the benefits resulting from accepting and following medical advice.

• Individual perception of the barriers impeding acceptance and following of medical advice.

Most of the HIV treatment literature is directed to the last two variables: the perceived benefits of therapy and perceived barriers to taking therapy. It has been argued that the perceived severity of the disease is not likely to be predictive for severe diseases such as HIV (Rimer), and recent research in the UK indicates that HIV is still almost universally considered to be a very serious condition (Anderson 1999). This belief may inform non-adherence. Specifically, the belief that if people take antiretroviral medications now, the drugs may not work later has led some people to skip doses or take drug holidays (Gifford 2000).

Despite the evidence that HIV/AIDS is still regarded as a serious condition, some research shows that the perception that AIDS is no longer a serious disease since the development of antiretroviral medications has been associated with poor adherence (Williams 2000).

Stage of disease may be an important factor shaping people's level of adherence to therapy. A study of 72 people at various stages of HIV infection found that those with most advanced disease perceived a higher risk if they did not comply with their regimen, compared to those with less advanced disease, and they were also more adherent (Gao 2000).

An early study of adherence to AZT monotherapy, which tested all four belief factors, found only that negative perceptions of the benefits of treatment were associated with non-adherence (Muma 1995). Such attitudinal barriers to the use of AZT have been found in other studies (Siegel 1997; Smith 1996).

As treatments optimism surged in the mid-to-late 1990s positive attitudes to therapy were seen to be associated with better adherence (Martini 2000; Jones 1998; Williams 2000; Wenger 1999). However, experiences of short- and long-term toxicities continue to inform beliefs about the harmful effects of HAART and may be driving some non-adherence (Walsh 2001; Duran 2001b; Powell-Cope 2003). The ability to manage side-effects may protect against non-adherence (Johnson 2003).

To complicate matters further, a study has reported the surprising result that sceptical attitudes towards therapy were associated with better adherence (Weiss 1998). The authors speculate that this may be because those who feel more optimistic feel less personal responsibility or that those who are more sceptical identify a bigger role for themselves in getting results.

Beliefs about doctors also affect whether people access health care and their adherence to prescribed therapy. A study of the relationship between stereotypes of physicians and adherence among low income, HIV-positive people found that people who held negative stereotypes about physicians sought care less often when sick and were less likely to adhere to recommendations for treatment (Bogart 2004).

A study of the attitudes and beliefs of patients with excellent adherence to antiretroviral therapy have been compared with the attitudes of people with suboptimal adherence. Beliefs expressed more commonly among those with very good rates of adherence included the need for 90-100% adherence for medication efficacy, and the desire to stay healthy as their key motivator (Malcolm 2003).

Knowledge and understanding

Better treatments knowledge has been consistently associated with improved adherence, although the content of this knowledge varies between studies. An accurate understanding of the purpose of therapy has been strongly associated with adherence (Durante 2003; Weiss 2003).

Lack of knowledge and information about HIV medications and a lack of understanding of HIV disease have also been associated with poor adherence (Williams 2000). Knowledge of dosing appears to improve over time but poor knowledge of medications and dosing eight weeks after starting treatment is associated with poor adherence and lower levels of literacy (Miller 2003).

In a large study in the UK, adherence was associated both with knowledge of the potential for viral resistance if treatments are not taken properly, and with knowledge of the risks of cross-resistance (Anderson 1999).

These results point to the importance of understanding the rationale for adherence as well as the basic mechanics of treatment-taking. They also suggest that health professionals can use questions which focus on knowledge about HIV to assess a person's readiness to take treatments (Weiss 2003).

Medications and the experience of illness

Regimen complexity is commonly assumed to affect adherence. There are three key aspects to regimen complexity which need to be considered: absolute number of pills; number of doses, and food and drink restrictions associated with dosing.

The total pill count may have a variable influence on adherence. One study (in a prison setting) has shown an association between fewer medications and improved adherence (Mostashari 1998) but this is contradicted by studies which have shown a greater number of medications associated with higher adherence (Samet 1992; Singh 1996; Delago 2003). Other studies looking at the effects of the number of pills have found no difference in adherence (Rodriguez-Rosaldo 1998).

The crucial factor in regimen complexity may not be the number of medications but the number of doses that have to be taken every day (Singh; Porter 1969). This is borne out in HIV therapy in several studies (Eldred 1998; Paterson 2000; Golin 2002; Howard 2002), although others have found no association between adherence and either the number of doses or the number of pills (Martini 2000; Lorenzen 1998; Carrieri 2001). The use of fewer adherence aids such as pill boxes and timers is also associated with poorer adherence (Golin 2002).

A large UK study found no associations between adherence and any aspect of regimen formulation: numbers of medications, doses or pills (Anderson 1999).

However, the complexity of some anti-HIV combinations, including separate dosing times and food and drink restrictions, may impact on adherence. A cross-sectional study found that only 63% of participants understood the food restrictions and dosing frequency of their current regimen (Stone 2001). Not surprisingly, simpler regimens were understood by a greater proportion of patients. Although many people may take their required number of doses per day, fewer are in line with time, food and dosing restrictions (Paterson 2000). Social factors such as work, discrimination and whether medications fit into a daily schedule may further undermine adherence to complex dosing and eating schedules (Adam 2003; Johnson 2003).

Experience of disease and medications also impacts on adherence. Researchers have studied the following factors: how long a person has been taking therapy for; stage of HIV disease, and experience of side-effects.

The general adherence literature typically stresses the close association between duration of therapy and worsening adherence, but evidence from HIV is more equivocal. Reductions in adherence with the duration of therapy have been seen in some recent HIV studies (Rodriguez-Rosaldo 1998; Mannheimer 2002; Stone 1998; Howard 2002), but not in others (Lorenzen 1998; Tuldra 1999; Anderson 1999). Periods of instability or change increase the risk of non-adherence to HIV medication over time (Tesoriero 2003).

There is contradictory evidence regarding the impact of duration or stage of HIV disease upon adherence to therapy. Several studies have reported that stage of disease does not influence adherence (Lorenzen 1998; Pratt 2001; Tebas 1998; Anderson 1999). However one study found that people with advanced HIV were more adherent to HAART (Gao 2000), as did one early study of AZT monotherapy (Samet 1992). Other studies suggest that people who have received an AIDS diagnosis or had lower CD4 counts are more adherent than people who have less severe immune damage (Delago 2003; Howard 2002).

However, the more specific variable of the experience of side-effects has been associated with poor adherence (Durante 2003; Halkitis 2003; Tuldra 1999; Spire 2002; Duran 2001b), although not consistently: two studies have not found any association (Martini 2000; Anderson 1999). Being able to manage side-effects has been linked to improved adherence (Johnson 2003).

The doctor-patient relationship and the quality of care

Quantitative studies have reported better adherence by those with a more stable doctor-patient relationship (Patullo 1998); by those with greater perceived professional support (Bakken 2000); by those who are highly satisfied in their relationship with their physician (Martini 2000;Heckman 2004); and by pregnant women with an adequate - rather than inadequate - standard of prenatal care (Laine 2000).

Studies with stronger qualitative elements have described better adherence amongst those with greater trust and satisfaction with their doctors (Munzenberger 1997; Malcolm 2003; Murphy 2003); those that felt their doctor spent adequate time with them during their visits (Durante 2003); and those who had a good relationship with the medical team (Jones 1998). Patients who hold negative views of doctors seem to be less satisfied with the care they receive and less likely to adhere to treatment (Bogart 2004).

However, most of the literature which attends to the quality of relationships with professionals in maximising adherence either takes the principle for granted and describes the appropriate character of these relationships, or seeks to evaluate specific interventions.

Several studies have reported that the experience of the physician also impacts on adherence. The more experience the physician has of prescribing antiretroviral drugs, the more likely patients are to be highly adherence (Delago 2003). This may reflect better communication about how HIV treatments work and ways to maximise adherence.

Taking the patient's perspective

Studies which seek to identify predictive factors of adherence/non-adherence are consistent with the medical research tradition which seeks to understand specific problems by identifying general traits of groups of affected people. The alternative approach is to seek the perspective of adherent/non-adherent individuals and assess the particular personal, social and psychological circumstances of their behaviour.

This can produce surprising results. For example, in a qualitative study of people referred to a rheumatology outpatient clinic, the authors argue that adherence may simply not be an issue for patients. They suggest that taking drugs is not about obeying doctor's orders but about weighing up the costs and benefits of taking particular medications as they are perceived within the contexts and constraints of everyday life (Donovan & Blake).

The conflict between managing the therapy and managing the demands of every day life is a strong theme within qualitative patient-centered studies. A study of people with epilepsy found that individuals regulated their medication as part of their own process of gaining control over their condition and the difficulties of their everyday lives; consequently what appears as non-adherence from a medical perspective may actually be the patient's attempt to assert greater control over the disorder of his or her life (Conrad 1985).

In HIV therapy, however, individuals have less leeway to adjust their regimens according to their own needs. The seriousness of the disease and the unusually harsh implications of non-adherence demand a much larger commitment than other chronic disease treatments.

Without doubt, the most common reason that people taking HIV therapy give for non-adherence is simply forgetting, identified in numerous studies (Walsh 2001; Arabe 1998; Bachiller 1998; Carrieri 2001; Gifford 2000; Lorenzen 1998; Nakashima 1998; Ohmit 1998; Paiva 1998; Stone 2001; Weidle 1999; Anderson 1999).

Forgetfulness encompasses a range of mistakes, predominantly either forgetting a dose or forgetting to take the pills when going out. However, forgetfulness is often linked to a failure to plan for changes in routine. Qualitative studies of highly active antiretroviral therapy (HAART) have described the process of integrating the demands of the regimen into daily life, rather than adjusting the regimen to suit daily life.

Problems with integrating the regimen into the daily routine and coping with changes in routine have been reported in a variety of ways: 'can't work them into my schedule' (Nakashima 1998); away from home/change in daily routine (Gifford 2000); 'too much trouble' (Johnstone 1998); situational obstacles (Paiva 1998); being asleep at the time to take medication (Carrieri 2001; Anderson 1999; Powell-Cope 2003; Walsh 2001); difficulties taking the medications in the workplace (Burgos 1998) and when socialising (Golin 2002; Walsh 2001), and the conflicting demands of medication, work and food requirements (Adam 2003; Walsh 2001). Being 'too busy' was reported by several studies (Carrieri 2001; Gifford 2000; Stone 2001; Anderson 1999; Powell-Cope 2003).

Individuals may report several reasons for non-adherence, and the more reasons a person reports, the less adherent they are likely to be (Walsh 2001).

One study suggested that adherence was typically achieved when pill-taking became routine (Anderson 1998). Although some participants in the study perceived the extent of daily pill-taking to be daunting prior to starting therapy, once the pattern had been incorporated into a personal daily routine it was no longer perceived as such, although some expressed a sense of fatigue with the inescapable daily round of pill-taking.

Those who were unable to successfully establish the regimen as a routine had the greatest problems with adherence. These were people who typically suffered from side-effects or who had practical problems with dietary requirements such as having to constantly drink large amounts of water.

Most of the studies which contribute to our understanding of the patient's perspective of antiretroviral therapy are qualitative. They typically describe self-reported reasons for missed doses. These data suggest that people taking antiretroviral therapy are not actively choosing to vary or miss doses, but rather they are usually making mistakes contrary to their intentions. Alternatively, non-adherence may be a situational to difficult circumstances. Sometimes, issues of confidentiality, discrimination and prejudice influence people's decision not to take their medication (Powell-Cope 2003; Adam 2003).

One study has drawn particular attention to the challenges presented by dietary restrictions: lunch time doses were often hard to take, especially if a substantial meal was required; conversely those who had to maintain an empty stomach found this could make work difficult. The study suggested that full adherence to drugs with dietary restrictions was low (50%) compared to over 95% of those taking drugs with no dietary restrictions (Hedge 1998).

Drug intolerance, toxicity or side-effects were also frequently reported as reasons for non-adherence (Arabe 1998; Bachiller 1998; Burgos 1998; Gifford 2000; Johnstone 1998; Nakashima 1998; Ohmit 1998; Paiva 1998; Stone 2001; Weidle 1999; Duran 2001; Spire 2002). One study reported that nausea at the prospect of taking the drugs was more often a reason for non-adherence than sickness itself (Anderson 1999). In another study, participants reported not taking therapy in order to avoid problems with side-effects in key situations, such as when starting a relationship or when going for a job interview (Burgos 1998).

These results provide an entirely different perspective on the 'problem' of adherence from the factor analyses described above. Here non-adherence is constructed not as a phenomenon that can be predicted by assessing an extensive range of measurable variables, but as a product of everyday events, expected and unexpected, desired and undesired.

References are listed at the end of Interventions to improve adherence in Anti-HIV therapy: Adherence.