Health beliefs and knowledge of patients and doctors in clinical practice and research

Patient Education and Counseling 26 (1995) 99-106

Health beliefs and knowledge of patients and doctors in clinical practice and research

Clare Bradley

Deparrnlerlt of‘ Psyd1o~og.v. Royal Holloway, University of’ London. Egham Hill, Eglmn, Surrey TW.20 OEX, UK

Abstract

This paper calls for a renewal of interest in monitoring and improving patients’ knowledge about their condition and self-care. Such knowledge is essential for patients with chronic disorders such as diabetes. Small or non-existent relationships often observed between knowledge levels and metabolic outcomes in diabetes can be seen to result from inappropriate investigation of these relationships. Knowledge about different aspects of diabetes management needs to be related to relevant outcomes and metabolic control cannot be expected to be related to every aspect of knowledge about diabetes self-care. The importance of health beliefs and perceived control in determining treatment choice and efficacy of treatment is considered. The implications of patients’ and doctors’ varying knowledge and beliefs for their preferences for different treatment options are examined and the need to consider such preferences when designing, developing and evaluating new treatments is emphasised.

Keywords: Diabetes; Knowledge; Health beliefs; Perceived control; Clinical trials

1. Introduction

Knowledge has been defined as a ‘shared set of tested and verified beliefs’ [l]. Though this is a useful working definition, in practice the distinc- tion between knowledge and beliefs is difficult to make. In my main area of research, diabetes management, and also in other areas of medicine such as renal medicine [2], it appears that health professionals may share a set of beliefs not all of which have been tested and verified. However, they are inclined to refer to these beliefs as knowledge rather than beliefs. Furthermore, health professionals, even those working in the same team, often differ in the beliefs they hold.

Over a decade ago, when I was first developing diabetes-specific measures of health beliefs and cognitions with a view to understanding individual differences in choice and efficacy of treatment [3], there appeared to be more interest in measuring patients’ knowledge than in measuring their beliefs about diabetes. Now there seems to have been a shift of interest away from measuring patients’ knowledge to focusing on their beliefs. Although knowledge measures could usually be relied upon to show improvements in knowledge following education sessions, they were disappointing when it came to predicting consequent improvements in diabetes control.

073%3991/95/$09.50 0 1995 Elsevier Science Ireland Ltd. All rights reserved SSDI 0738-3991(95)00725-F

100 C. Bradley / Patient Educ. Cows. 26 (1995) 99-106

2. Knowledge

2.1. Knowledge measures It was originally expected, not unreasonably,

that increased knowledge of diabetes would be associated with improved diabetes control. How- ever, people with high levels of knowledge often had poor blood glucose control. This gap between knowledge and behaviour was taken by some researchers to indicate that patients’ knowledge was not as important as their beliefs or motiva- tion to apply knowledge. Others suggested that the knowledge measures were not valid. Before moving on to consider the importance of beliefs, it is worth considering the way the validity of knowledge measures has been tested.

Developers of knowledge measures have usually expected that knowledge levels would be corre- lated with measures of blood glucose such as Haemoglobin Al (HbA,) measures. However, I suggest that such expectations are not reasonable given the nature of the knowledge measures and the limitations of HbA, measures. Take for example the widely used measures of knowledge of diabetes designed and developed by Dunn’s group [4]. If we consider the individual knowledge items separately we can see that: (1) some aspects of knowledge cannot be expected to relate to blood glucose control (e.g. knowledge about foot care); (2) other aspects of knowledge might affect experience of hypoglycaemia but such episodes are usually transient and would have little if any effect on HbA, levels. Thus ignorance of the causes of hypoglycaemia would not be expected to be reflected in higher blood glucose levels. Indeed knowledge of the causes of hypoglycaemia might well lead to increased blood glucose in an attempt to avoid the increased risks of hypoglycaemia associated with tighter control; and (3) yet other aspects would relate to blood glucose control under specific circumstances and would only relate to HbA, under certain conditions, e.g. lack of knowledge of sick-day rules would not affect HbA, levels unless the patient became ill.

Looking at the latest versions of the three par- allel forms of this knowledge measure [4] and excluding items such as those described in l-3 above, I would only expect there to be significant

relationships between HbA, and eight, seven and five out of the 15 items on each of the forms A, B and C respectively. Thus attempts to relate the total of all 15 scores to HbA, are unlikely to be successful.

2.2. Patients’ knowledge Even with those items which we might expect to

relate to HbA,, the relationship is not straightfor- ward. Although we might expect improved blood glucose control to follow the acquisition of knowledge about diabetes management, it is clear that when patients recognise they have HbA, levels that are too high, or when they develop a problem such as a foot ulcer, they are more likely to learn about that specific problem and related aspects of diabetes management at that time, rather than at an earlier stage when all was well. Thus, impressive levels of knowledge can be found among people with poor blood glucose control and complications. It may, nevertheless, have been the case that poor knowledge at an earlier date was a causal factor leading to poor metabolic control and complications.

None of the above points are intended as criti- cisms of the knowledge scales themselves which contain a sensible collection of well-designed items dealing with important questions that people with diabetes need to know about. The point is that lack of knowledge about one of the items included in the questionnaire may have very different implications from lack of knowledge about one of the other items. We need to devise precise ways of validating specific knowledge items or small subsets of knowledge items. Rather than examining correlations between knowledge and outcomes assessed at the same time, we should look at those people with poor knowledge of some aspect of diabetes management and predict that those patients who are helped to improve this aspect of their knowledge will have fewer problems at a later date than those with whom there has been no intervention. It is no good looking for evidence of validity in patients who already have foot ulcers because they may have increased their knowledge since the ulcer developed. It is also important to recognise that well- informed patients may not be those with the best

C. Bradley / Patient Educ. Couns. 26 (1995) 99-106 101

diabetes control because having the knowledge may not be enough. There are many reasons why patients may be unable or unwilling to apply their knowledge about diabetes. This is where beliefs can be so crucial. But first there is the issue of doctors’ knowledge and the knowledge of other health professionals to be considered.

2.3. Doctors’ and health professionals’ knowledge When we have given lengthy knowledge ques-

tionnaires, such as the Charing Cross questionnaire, to diabetes care teams, we have found a disturbing lack of agreement about the ‘correct’ answers. A great deal of discussion may be necessary before reaching a consensus. This may be because some members of the team do not know enough about issues they regard as outside their special area of expertise. Thus, for example, some diabetologists are surprisingly ignorant about di- etary issues. Providing the dietitians agree on the correct answers and their shared views are based on valid evidence, the answers of the dietitians can be said to represent correct knowledge rather than beliefs. However, there are other instances where doctors disagree amongst themselves about the ‘correct’ answers to more traditionally ‘medical’ questions. When this happens, doctors’ knowledge may be out of date or otherwise inadequate, or the questions or response options need to be modified. New treatment techniques may mean that items are no longer relevant or appropriate. For example, the following item was included on the Diabetes Health Profile [5]. ‘When should you give insulin injections?’ The options offered included ‘5- 10 min before a meal’, ‘15-30 min before a meal’, ‘35-40 min before a meal’, ‘lo-15 min after a meal’. With the advent of very fast acting insulins, recommendations are changing and will be dependent on the type of insulin being used. A knowledge question about timing of insulin injections now needs to elicit information about the type of insulin being used and that information has to be taken into account in judg- ing the correctness of the response option chosen. Furthermore, the response options need to cover all the varieties of insulin in use among the patient population concerned. Looking back at knowledge questionnaires designed in the 1980s such as

the Charing Cross questionnaire, we can see that many items are no longer suitable. There have been changes in insulins and insulin delivery systems (giving more flexibility about meal times). There have been changes in the wide range of food products now available (salad creams are now not all high in fat) and of increasing importance are the moves towards tailoring treatment regimens to suit individual patients and their lifestyles.

Where the health professionals do not agree on the correct answers to a knowledge questionnaire, appropriate action needs to be taken to resolve the disagreement. This may be achieved by one of the following: (1) making up any knowledge deficits in some of the health care professionals by discussing discrepancies and sharing information; (2) where a correct answer cannot be verified because of lack of evidence, the item may be redefined as a belief rather than knowledge and excluded from the questionnaire; and (3) changing the question or the response options so that the members of the health care team can agree on the correct answer.

Though modification of questionnaires is not to be undertaken lightly and precautions are needed [6], sensible modifications are preferable to assess- ment of knowledge which is not being consistently or adequately taught by all members of the health care team.

In sum, knowledge is essential to diabetes self- care and lack of essential knowledge is likely to result in poor outcomes. Improvements to previ- ously inadequate aspects of knowledge should lead to improvement in the relevant outcomes. Care is needed in testing the validity of knowledge items to ensure that precise predictions appropriate to specific items or selected subsets of items are tested.

3. Beliefs

3.1. Beliefs about diabetes and its complications If people with poorly controlled diabetes be-

lieve that they will not be able to improve their control without destroying their quality of life, they are unlikely to try and do so without appropriate encouragement and support to help them

102 C. Bradley 1 Patient Educ. Couns. 26 (1995) 99-106

change their behaviour and related beliefs. Ac- quiring an understanding of belief systems and how they can work, can be helpful in identifying and overcoming destructive patterns of beliefs and facilitating and encouraging constructive beliefs. Although there may be resistance to change, beliefs are far more readily changed than personality factors. Often beliefs can be instantly changed by supplying some key piece of information. For example, a belief that the effort demanded suc- cessfully to manage insulin-dependent diabetes is not worthwhile, may be altered when information is given about the complications known to be associated with poor control and about the benefits in terms of risk reduction which the Dia- betes Control and Complications Trial (DCCT) showed can follow the introduction of intensified treatment [7]. Such beliefs may be held by health professionals as well as patients and, even in the light of the DCCT results, there will be patients and health professionals who believe that they personally would not find intensified treatment possible or sufficiently worthwhile.

In order to make sense of the bewildering range of beliefs of patients and health professionals, it is helpful to develop an understanding of the available models of beliefs and to gain some insight into how beliefs can affect behaviour. There are various models designed for differing circumstances to predict a range of behaviours. Leclercq introduces elsewhere in this volume, a model called PROACTA, which draws on concepts from several earlier models. I will mention here three models which have usefully guided my work in diabetes and which are increasingly being com- bined and used by other researchers. These are the health belief model, the social learning model and attribution theory. I will refer to the questionnaires designed to measure the concepts derived from the models and illustrate the importance of considering the interactive effects of beliefs when attempting to understand and predict behaviour.

The health belief model (HBM) guided con- struction of diabetes-specific health belief scales in work which started in 1982 [3,8]. The HBM at the time suggested that there were four major belief factors which determine whether or not an individual will follow the treatment recommended.

These were beliefs about: (1) benefits of treatment; (2) barriers to treatment; (3) severity of the disorder; and (4) vulnerability to the disorder.

It is clear that patients are more likely to follow a treatment regimen if they believe there are many important benefits of treatment and no insupera- ble barriers. In the case of diabetes where the disorder is already present, beliefs about severity and vulnerability can be measured as beliefs about the short- and longer-term complications of diabetes, including beliefs about severity of and vulnerability to hypoglycaemic episodes on the one hand and retinopathy and other long-term complications on the other.

The ways in which beliefs about severity and vulnerability may exert influence are far less clear than the ways in which beliefs about benefits and barriers can affect behaviour. As with knowledge of diabetes, so too with beliefs. Knowledge and beliefs can influence outcomes but outcomes can also influence knowledge and beliefs. Although feelings of vulnerability to a severe complication might usually be expected to result in action, the improved control which may follow may then result in reduced feelings of vulnerability. It may then appear that reduced vulnerability is associated with improved control. These dynamic inter- relationships between beliefs and outcomes have been discussed elsewhere [9,10] and indicate that the predictions of the health belief model are not static and universal, but are dynamic and dependent on particular circumstances, including the education and counseling patients receive.

When testing the predictive validity of measures of knowledge or beliefs, we need to look at changes across time following interventions, to improve knowledge or to change beliefs. Thus we would predict that patients with poor diabetes control would make efforts to improve their diabetes control if we: (1) encourage beliefs in the benefits of treatment (e.g. to avoid future complications); (2) overcome perceived barriers to treatment (e.g. by increasing regimen convenience and flexibility); (3) provide accurate and immediate feedback about poor blood glucose control and ensure that patients know about the associated risks of complications in a supportive, enabling context, where patients and health professionals

C. Bradley / Patient Educ. Courts. 26 (1995) 99- 106 103

work together to take constructive action; and (4) discourage attempts to diminish the perceived severity of complications while encouraging beliefs that risks can be reduced by even small improvements in diabetes control. The missing link here is that the patterns of beliefs to be shaped will still only lead to constructive action to improve diabetes control if patients expect their actions to be successful. Recognising this, HBM researchers, have suggested that self-efficacy beliefs need to be incorporated into the HBM [lo]. Self-efficacy beliefs are related to beliefs about perceived control, which was measured in a sepa- rate set of scales described below.

3.2. Perceived control of diabetes The Perceived Control of Diabetes scales have

proved to be excellent predictors of some important outcomes. The scales were influenced by earlier work on locus of control measurement (which had its origins in social learning theory) [12- 141 and more recent developments in measures of attributional style (which derive from attribution theory) [15]. The Perceived Control of Diabetes scales were developed because the concept was so clearly relevant to diabetes and there were reasons to believe that diabetes-specific measures would be more useful than generic measures in understanding the behaviour of people with diabetes and in predicting outcomes [16].

People with diabetes who attribute diabetes control to good or bad luck, are favouring a chance locus of control. Such people would not be expected to be well motivated to follow a de- manding treatment regimen which depends on their own actions and decisions. Those people who favour an internal locus of control and attribute diabetes control to their own efforts, would be more likely to adhere to such a regimen. The Perceived Control of Diabetes scales ask people with diabetes to consider the most likely causes of each of a series of diabetes-related outcomes, some of which are positive (e.g. regaining good diabetes control after a period of poor control) and some negative (e.g. becoming over- weight). For each diabetes-related outcome the causes offered were then rated by respondents on seven subscales which combine to provide three

composite measures labelled personal control, medical control and situational control. High scores on personal control indicate strong beliefs in personal responsibility, personal control and the foreseeability of diabetes-related outcomes. High medical control scores indicate strong beliefs in doctors’ responsibility and control over treatment outcomes. High situational control scores indicate strong beliefs in chance factors and the role of other people and circumstances in determining outcomes.

Wallston and Wallston’s speculative health locus of control typology offered a way of improving our understanding of the relationships between locus of control and health behaviour by looking at patterns of locus of control beliefs and their implications for specific situations. The typology was originally constructed for scales developed by the Wallstons: the Multidimensional Health Locus of Control (MHLC) scales [14]. The MHLC scales include three subscales: internality, powerful others and chance. These subscale scores may be regarded as equivalent to the Perceived Control of Diabetes scales’ scores for personal control, medical control and situational control. The typology was readily adapted to diabetes management [16] and the predictions made are summarised in Table 1. We applied the Wallstons’ typology to the responses of tablet-treated patients completing the Perceived Control of Dia- betes scales. The ordering of the types in relation to the outcomes of HbA, and weight, fitted predictions very well. Believers in control had the best measures of HbA, and weight, followed closely by those with ‘pure’ personal control. At the other extreme, the double externals had the worst measures of HbA, and, together with the ‘pure’ medical control externals, also fared badly in terms of excess body weight. Kruskal-Wallis tests for significance showed significant differences between the types for percentage ideal body weight (P < 0.02) and HbA, levels (P < 0.05). When subjects’ well-being scores were examined in relation to the Wallstons’ typology, no significant differences were found between the groups. However, there were significant differences in treatment satisfaction (P < 0.001) with, as expected, the most satisfied patients being the be-

104

Table 1

C. Bradley / Patient Educ. Couns. 26 (199.5) 99-106

Summary of main predictions from Wallston and Wallston’s typology and predicted consequences for diabetes management”

Type

Believers in control

‘Pure’ personal control

‘Pure’ medical control externals

‘Pure’ situational control externals Double externals

Scale scores Predicted consequences for diabetes management

Personal Medical Situational control control control

High High Low Good use of personal resources and health service resources

High Low Low Good use of personal resources but may not recognise when these are inadequate

Low High Low Poor use of personal resources and unrealistic expectations of health service resources

Low Low High Fatalistic: poor use of all resources Low High High Poor use of personal resources, unrealistic

expectations of health service resources and an element of fatalism

a Modified from Bradley (1994). In: The Handbook of Psychology and Diabetes. Reproduced (with amendments) by permission of Harwood Academic Publishers.

lievers in control and ‘pure’ personal control patients. The least satisfied patients were the double externals. These findings demonstrate well why measures of powerful others, or medical control, have often failed to be useful predictors in previ- ous research. Strong perceptions of medical control may be advantageous in combination with strong perceptions of personal control. However, attributions to medical control are not constructive in the absence of attributions to personal control where diabetes is concerned and where self-management is central to the success of treatment.

3.3. Perceived control and health beliefs as predictors of patients’ choice and eficacy of insulin treatment

The most clearcut evidence of the value of our measures in understanding outcomes was provided by the studies of insulin-treated patients where the scales were used to predict choice and efficacy of treatments [9,18]. The measures were used in a Sheffield feasibility study where patients were offered the choice of continuous subcutaneous insulin infusion (CSII) pumps, intensified conventional (injection) treatment (ICT) or conventional treatment (CT). The scales significantly predicted patients’ choice and efficacy of treat-

ment. Of particular interest was the way the perceived control scales predicted response to CSII. Rather less than a third of the patients wished to try CSII (approx 100 patients). Of these about one third were using CSII 5 years later. However, some patients were attracted to CSII with unrealistic expectations that the pump would manage their diabetes for them (as indicated by responses to the Perceived Control of Diabetes scales). After 12 months of CSII use, these patients had higher glycosylated haemoglobin levels, indicating greater impairment of diabetes control, than patients who expected to take more personal responsibility for their treatment [9]. Furthermore, responses to the scales helped to explain the reasons for an alarming increase in life-threatening diabetic ketoacidosis (DKA) among pump users. Those who experienced DKA while using CSII, were those who expected that the treatment would control their diabetes for them and had less sense of personal control over their diabetes. In reality, the effectiveness of the pump was dependent on their actions. The reasons for patients holding unrealistic expectations of CSII controlling their diabetes for them could be traced back to the information given to patients at the start of the study when they were told that ‘The evidence suggests that CSII will control your diabetes bet-

C. Bradley / Patient Educ. Cows. 26 (1995) 99-106 105

ter than injections’. Thus patients received the impression that CSII pumps would control their diabetes for them and those with stronger beliefs in medical control with less strong beliefs in personal control found this impression of CSII at- tractive. Such patients would be expected to be less likely to detect metabolic problems and to take emergency action when needed. Bradley et al. [ 181 suggested that over-optimistic expectations of CSII pumps may have encouraged patients to assume that they would have less need to take personal responsibility for their diabetes when they changed from injection treatment to CSII, though in reality they needed to take greater care.

4. Treatment preferences and clinical trial design

The Sheffield feasibility study provided a clear illustration of a way in which the knowledge and beliefs of health care professionals can influence the way patients are recruited into trials. In turn, patients’ knowledge and beliefs about the treatment options will influence their treatment choices and, in the case of randomised controlled trials (RCTs), determine whether or not they decide to participate in a trial. Yet when we look at papers which report clinical trials comparing treatments, it is the exception rather than the rule for papers to provide details about what patients were told on recruitment about the treatment under study, or to include discussion about patients’ motives for participating in the trial. However, if patients believe that a new treatment in a RCT is preferable to the control treatment which is the treatment they are currently using, and if that new treatment is only available within the trial (as was the case in the early 1980s for CSII pumps), it is likely that patients who agree to take part in the RCT will tend to be those who preferred the idea of the new treatment [19]. (Those who preferred injections could have their preference outside the trial without risk of being assigned to their non- preferred treatment). Patients’ choice will depend on what they were told about a particular trial. If, as in the Sheffield study, patients are offered the vision of a medical solution without a balance of emphasis on the need for increased personal effort, a particular subgroup of patients will be

attracted into the trial (with a predominance of pure medical control external patients and double externals). If, on the other hand, it is made clear (with the hindsight of the Sheffield findings) that CSII can be a useful tool but is one which is dependent for its effectiveness on the user, a quite different subgroup of patients is likely to be recruited (those who are believers in control and pure personal control patients). We can see that the medical outcomes of a trial are likely to be dependent on which type of patient tends to be recruited and on the methods of recruitment used. The unwary researcher, especially those who included measures of psychological processes, may all too readily generalise inappropriately from the findings obtained with one inadvertently highly selected subgroup of eligible patients, to the patient population as a whole.

It is a common experience that different clinical trials produce different results and conclusions and carry quite different implications. If we take account of patients’ beliefs and preferences when recruiting them into trials and if we provide the necessary information to allow them to make informed choices, we will increase the chances of understanding differences in trial outcomes. With better understanding of the impact of patients’ knowledge and beliefs on their behaviour and associated treatment outcomes, recruitment procedures can be designed which encourage preferences for new treatments (whether in a clinical trial or in clinical practice) among patients with appropriate preferences and beliefs and discourage those patients with inappropriate beliefs and preferences who are unable or unwilling to adapt in the light of the new information offered. Pa- tient-centred partially randomised clinical trial designs provide a methodology for clinical trial comparisons of treatments where patients have strong views and where different patients prefer different treatments [20]. Elsewhere I have encouraged creative use of other clinical trial designs and measurement of patients’ psychological outcomes and processes (including beliefs and preferences) alongside traditional medical variables [21]. It is an all-too-common belief of medical doctors that beliefs and preferences need only be considered once medical efficacy of treatment has been estab-

106 C. Bradley / Patient Educ. Couns. 26 (1995) 99-106

lished. Such beliefs need to be carefully reconsid- ered in the light of the evidence, some of which is referred to above. Medical efficacy cannot be de- termined in isolation. Beliefs and preferences need to be studied from the early stages of designing, developing and evaluating each new treatment and their potential influence recognised through- out the process of evaluating new treatments, where efficacy is likely to be dependent on patients’ preferences and behaviour.

References

[I] Marteau TM, Johnston M. Health professionals: a source of variance in health outcomes. Psycho1 Health 1990; 4: 4758.

[2] McGee HM, Bradley C, editors. Quality of life following renal failure. Chur, Switzerland: Harwood Academic, 1994.

[3] Bradley C et al. Development of scales to measure perceived control of diabetes mellitus and diabetes-related health beliefs. Diabetic Med 1984; 1: 213-218.

[4] Beeney LJ, Dunn SM, Welch G. Measurement of diabetic knowledge: the development of the DKN scales. In: Bra- dley C, editor. Handbook of psychology and diabetes. Chur. Switzerland: Harwood Academic, 1994.

[5] Meadows KA, Wise PHW. A national profile of knowledge and psychosocial status in 2239 Type 1 diabetic patients. Diabetic Med 1992; 9 Suppl 1: 48A.

[6] Bradley C. Adapting scales and procedures: the limits of reliability and validity. In: Bradley C, editor. Handbook of psychology and diabetes. Chur, Switzerland: Harwood Academic, 1994.

[7] DCCT Research Group. The effect of intensive treatment of diabetes on the development and progression of long- term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993: 329: 9777986.

[8] Lewis KS, Bradley C. Measures of diabetes-specific health

beliefs. In: Bradley C, editor. Handbook of psychology and diabetes. Chur, Switzerland: Harwood Academic, 1994.

[9] Bradley C et al. The use of diabetes-specific perceived control and health belief measures to predict treatment choice and efficacy in a feasibility study of continuous subcutaneous insulin infusion pumps. Psycho1 Health 1987; 1: 133-146.

[IO] Shillitoe RW, Christie MJ. Determinants of self-care: the health belief model. Holistic Med 1989; 4: 3-17.

[Ill Rosenstock IM et al. Social learning theory and the health belief model. Health Educ Q 1988: 15 (2): 175-183.

[I21 Rotter JB. Generalised expectancies for internal versus external control of reinforcement. Psycho1 Monographs 1966; 80: 1-28.

[I31 Wallston KA et al. Development of the multidimensional health locus of control (MHLC) scales. Health Educ Monographs 1978; 6: 160- 170.

[14] Levenson H. Distinctions within the concept of internal- external control. Proceedings of the 80th Annual Conven- tion of the APA, 1972; 7: 261-262.

[15] Peterson C et al. The attributional style questionnaire. Cog Therap Res 1982: 6: 287-300.

[16] Bradley C. Measures of perceived control of diabetes. In: Bradley C, editor. Handbook of psychology and diabetes. Chur, Switzerland: Harwood Academic, 1994.

[17] Wallston KA, Wallston BS. Who is responsible for your health? The construct of health locus of control. In: Sanders G, Suis J, editors. Social psychology of health and illness. New Jersey: Erlbaum, 1982.

[18] Bradley C et al. Predicting risk of diabetic ketoacidosis in patients using continuous subcutaneous insulin infusion. Br Med J 1986; 293: 242-243.

[I91 Bradley C. Clinical trials - time for a paradigm shift? [editorial]. Diabetic Med 1988; 5: 107-109.

[20] Brewin CR, Bradley C. Patient preferences and randomised clinical trials. Br Med J 1989; 299: 3133315.

[21] Bradley C. Designing medical and educational intervention studies: a review of some alternatives to conventional randomized controlled trials. Diabetes Care 1993: 16: 509-518.

Health beliefs and knowledge of patients and doctors in clinical practice and research

Documents

Transcript of Health beliefs and knowledge of patients and doctors in clinical practice and research