SELF-MEDICATION WITH ANTIBIOTICS: AN ASSOCIATION …

MASTERS THESIS

SELF-MEDICATION WITH ANTIBIOTICS: AN ASSOCIATION BETWEEN INFORMATION

PROVISION AND LOWER LEVELS?

ANNA-SOPHIE DE JONG

SUPERVISORS

DR. IOANA VAN DEURZEN (TILBURG UNIVERSITY)

PROF. PETER ACHTERBERG (TILBURG UNIVERSITY)

EXTENDED MASTERS SOCIOLOGY

DECEMBER 2015

TABLE OF CONTENTS Abstract…………………………………………………………………………………..……2

1.1 Introduction……………………………………………………………………………….3

1.2 Results of Eurobarometer 2013………………………………………………….…….6

2. Theory……………………………………………………………………………….….…14

3. Methods…………………………………………………………………………….…..…21

4. Data Analysis……………………………………………………………………….…….26

5. Results…………………………………………………………………………….………29

6. Discussion and conclusions…………………………………………………….…..…..38

7. Bibliography…………………………………………………………………….……..….42

8. Appendix ………………………………………………………………..…...…….……..45

9. Annex…………………………………………………….………………………………..54

2

Abstract The current thesis aims to investigate whether there is an association between information

provision of antibiotics and the behaviour of those who self-medicate with antibiotics. Self-

medication is the use of drugs without medical guidance, in response to self-diagnosed

disorders or symptoms, or the continued or sporadic use of a prescribed drug for recurrent or

chronic symptoms of disease (Kunin, 1978). Insight into the behaviour of those who self-

medicate with antibiotics is valuable as this form of antibiotic use contributes to the misuse of

antibiotics, which further contributes to the issue of antibiotic resistance. I analysed data from

the ‘Antimicrobial resistance and causes of non-prudent use of antibiotics in human

medicine’ research project, conducted by the Netherlands Institute for Health Services

Research (NIVEL). A total sample of 1,846 respondents across seven European countries

who self-medicated with antibiotics in the past 18 months were surveyed through a

computer-assisted telephone interview in 2015. Data was analyzed by use of a logistic

regression. The results show the association between information provision and self-

medication to be weak at most, however due to the selective sample the results are not

generalizable. Future research should include data on individuals who have not self-

medicated with antibiotics in order to determine real effects.

3

1.1. Introduction

In recent decades, a growing body of evidence has advanced our medical and social

knowledge of antibiotics and the ability of microorganisms to transform and develop immunity

to them. Antibiotic resistance is the resistance of a microorganism to an antibiotic drug that

was originally effective for treatment of infections caused by it (WHO, 2014). Evidence of the

consequences of antibiotic resistance and methods to delay, reduce or avoid it are found in

both medical and social research fields. On the social level, both individual and societal

implications are of potential concern. On the individual level, antibiotic resistance poses the

risk of an increase in the quantity and intensity of complicated treatments due to limited

treatment options (Levy, 2005). On the wider societal level, consequences include avoidable

healthcare costs and new infectious diseases. The European Centre for Disease Prevention

and Control (ECDC) estimates that each year in EU countries, 25,000 deaths are directly

attributable to multi-drug resistant infections and 1.5 billion lost due to extra in-hospital and

outpatient costs and productivity losses due to absence from work and patients who died

from their infection (ECDC, 2009). Combining this with an evident increase in globalization

(Denis et al., 2006), international transmission enables a worldwide threat of emerged

diseases (Grigoryan et al., 2007). While not the focus of this thesis, the severity of the

consequences of antibiotic resistance highlights the importance of identifying and combating

the factors which cause it.

Previous studies have shown that on the national level, the prevalence of antibiotic

resistance is positively correlated with prescribed outpatient drug use (non-hospitalised drug

use, i.e. drug use in primary care) (Goossens et al., 2005; Albrich el al., 2004). Cause for

concern then is the fact that outpatient use accounts for more than two-thirds of antibiotic

sales globally (Llor & Cots, 2009) and it has been found that more than 40% of prescriptions

for antibiotics are more or less inappropriate (WHO, 2014). More significantly, there is

evidence for a correlation between outpatient use and antibiotic resistance in Europe

(Adriaenssens et al., 2011; Goossens, 2005). These figures, however, report levels of

prescribed antibiotics. What is missing from empirical research is information on self-

medication with antibiotics and determinants explaining the variations in levels, while we

know that this is substantial in several countries (Lopez-Vazquez et al., 2012).

Self-medication is the use of drugs without medical guidance, in response to self-diagnosed

disorders or symptoms, or the continued or sporadic use of a prescribed drug for recurrent or

chronic symptoms of disease (Kunin, 1978). Self-medication occurs most frequently through

over-the-counter (OTC) sales, left-over supplies and drugs procured from family or friends

4

(Berzanskyte et al., 2006). Although OTC sales of antibiotics are largely illegal, this does

occur (Markovic-Pekovic, 2012). Misuse of antibiotics may increase antibiotic resistance in a

community’s commensal flora (microorganisms within a relationship whereby one benefits

from the other while not benefiting or hurting the other) by exerting a selective pressure on

the skin, gut and upper respiratory tract, favouring bacteria resistant to the antibiotics

(McNulty et al., 2007). Therefore these forms of antibiotic use constitute a primary form of

irrational use of medicine, as antibiotics represent one of the most prescribed drugs

worldwide (Donkor et al., 2012).

In 2011 the European Union put in place a communication strategy in response to the threat

of antimicrobial resistance (EU, 2011). The goals of this strategy are to prevent the spread of

microbial infections, undertake research into effective ways to combat resistance and ensure

the appropriate use of antimicrobials. Communication, education and training form the core

of the strategy. The importance of educational campaigns throughout the EU is argued in

light of widespread public misconceptions regarding the nature and appropriate use of

antibiotics. Also in 2011, The World Health Organisation attributed World Health Day to

increasing awareness of the issue of antibiotic resistance and released a 6 point policy

package addressing how to combat resistance (WHO, 2011). Preventing self-medication is

presented as a core action in tackling resistance by requiring the enforcement of

prescription-only use. Five suggestions are put forward in the area of public education

promotion of antibiotics and their use. Suggestions emphasise the role of prescribers and

dispensers in educating patients on the correct use of antibiotics and proposes the use of

targeted public health campaigns. They also suggest to introduce the correct use of

medicines in health education components of both school curricula and adult education

programs. The European Centre for Disease Prevention and Control also dedicates urgency

to the issue of antibiotic resistance and proper use through the organisation of the annual

European Antibiotic Awareness Day (EAAD) (ECDC, 2005). The EAAD aims to increase

awareness of the public threat of antibiotic resistance and the importance of the correct use

of antibiotics. As part of this strategy, the ECDC aims to monitor both levels of public use of

antibiotics and knowledge about antibiotics.

In light of the theme of education found in the initiatives and research focused on combatting

antibiotic resistance, this research will seek to discover how differences in interaction

between healthcare professionals or media outlets and patients may contribute to levels self-

medication with antibiotics. Specifically, I will investigate whether receiving information on the

use of antibiotics is associated with lower levels of self-medication. The overarching research

question of this thesis is:

5

‘How do differences in information provision by healthcare professionals or media outlets to

patients who self-medicate with antibiotics affect the level of self-medication?’.

The purpose of this research question is to determine whether there is an association

between receiving information and lower levels of self-medication. It also extends to

determine which type of information and source of information produces a stronger

relationship. The research will focus on seven European countries: Cyprus, Estonia, Greece,

Hungary, Italy, Romania and Spain. These countries were selected as they were identified by

a special Eurobarometer report on antimicrobial resistance as countries where levels of self-

medication are high and public knowledge of antibiotics is low (Eurobarometer, 2013). The

results will be relevant for the research project ‘Antimicrobial resistance and causes of non-

prudent use of antibiotics in human medicine’ (ARNA), conducted by the Netherlands

Institute for Health Services Research (NIVEL) as interventions to address the non-prudent

use of antibiotics will be developed for the seven countries.

In the next section, I will present some relevant results of this Eurobarometer survey (TNS

Opinion & Social, 2013) to establish a context for the analysis. The selected results highlight

the importance of educating the public on antibiotic use, especially in geographic areas

where levels of self-medication are high and knowledge low - as is the case for the 7

countries under study.

In Chapter 2 I will discuss the theoretical background underlying this research, and as the

arguments build, the hypotheses are presented. A number of theories are utilised to

formulate two hypotheses and one exploratory hypothesis. I will also present a conceptual

model for clarity.

In Chapter 3 I will outline the methods of the research including descriptions of the data

collection, the sample and the questionnaire used.

In Chapter 4 I will describe the statistical analysis including both the operationalization of the

variables and the regression performed.

In Chapter 5 I will present the results of the analysis. This will begin with descriptive statistics

followed by a comparison of the background characteristics of those who self-medicated with

antibiotics and those who did not, in order to determine whether those who self-medicate are

a selective group. I will also present the results of the statistical analysis.

In Chapter 6 I will reflect on the results and provide conclusions of the research and a

discussion of its limitations and implications.

6

1.2. Results of Eurobarometer 2013

As part of the ECDC strategy, the Directorate-General for Health and Consumers

commissioned an EU wide survey on antimicrobial resistance published in 2013. The

purpose of the survey is to monitor public knowledge and use of antibiotics. The report

covers the 28 EU Member States (Table 1.1) with a total of 27,680 respondents.

Respondents who represent different social and demographic background were interviewed

face-to-face in their mother tongue.

Respondents were asked to indicate true or false to the following four statements:

- Antibiotics kill viruses. (correct answer: False)

- Antibiotics are effective against colds and flu. (correct answer: False)

- Unnecessary use of antibiotics makes them become ineffective. (correct answer: True)

- Taking antibiotics often has side-effects, such as diarrhoea. (correct answer: True).

According to the findings of this survey, only just over a fifth of Europeans correctly answered

their four questions about antibiotics, while the European average of correct answers is 2.4

out of 4.

While 84 percent know that the unnecessary use of antibiotics makes them become

ineffective, 49 percent are unaware that they are ineffective against viruses (with an

additional 11 percent indicating they don’t know); and 41 percent are unaware that they are

ineffective against colds and flu. This becomes interesting in light of evidence which shows

that self-limiting, virus-caused infections are a leading reason to self-medicate with antibiotics

(Grigotyan et al., 2007; Väänänen et al., 2006). A geographical distinction is evident in the

results; all nine countries in which a majority indicated correctly that antibiotics do not kill

viruses are in northern or western Europe (See Figure 1.1). The highest proportion of correct

responses is from Sweden at 74 percent, while the rest range from 51 to 59 percent. All

countries under study here fall below the EU27 average of 40% correct responses; Hungary

(38%), Estonia (36%), Italy (33%), Spain (29%), Greece (25%), Cyprus (21%), and Romania

coming last at 15%. Education proves to be a significant factor in that on average, only 27

percent of those whose education ended before or at the age of 15 (lowest education level)

gave a correct response, in contrast to 52% of those whose education ended at or before the

age of 20 (highest education level) (Table 2.1). Obtained information and the source of this

information also have an effect. Those who received information from a healthcare

professional answered this question correctly 10 percent more than those who did not

receive information (44% and 34% respectively). In addition, media campaigns prove to be

7

particularly effective at raising awareness, as 55% of those who received information from

such a source gave a correct answer.

Table 1.1. Country abbreviations

Abbreviations Countries Abbreviations Countries Abbreviations

Countries included in study

AT Austria

LU Luxembourg CY Cyprus

BE Belgium

MT Malta

EE Estonia

CZ Czech Republic NL The Netherlands EL Greece

BG Bulgaria

PL Poland

ES Spain

DK Denmark

PT Portugal

HU Hungary

DE Germany

RO Romania

IT Italy FR France

SI Slovenia

RO Romania

HR Ireland

SK Slovakia IE Italy

FI Finland

LT Lithuania

SE Sweden LV Latvia

UK The United Kingdom

Figure 1.1. For each of the following statements please tell me whether you think it is true or false. Antibiotics kill viruses.

*Adapted from: Eurobarometer (2013: 25)

0

10

20

30

40

50

60

70

80

90

100

SE FR LU DK

NL FI UK

BE IE HR SI

EU2

7

HU DE EE CZ IT PL

SK ES AT

LV LT EL MT

BG CY

PT

RO

Percentage of correct responses to the statement; 'Antibiotics kill viruses'

KEY

European countries EU average Countries included in present analysis

8

Table 2.1. For each of the following statements please tell me whether you think it is true or false. Antibiotics kill viruses. By education, received information, and information source.

EU27 (n= 26,680)

Correct answer Incorrect answer

40% 49%

Education (End of)

15- 27% 58% 16 - 19 37% 51% 20 + 52% 41% Still studying 40% 50% Received information

Yes 52% 41% No 34% 53% Information source

Advice from a professional 44% 50% From media/campaigns 55% 37% Did not receive information 34% 53%

*Adapted from: Eurobarometer (2013: 27).

A similar pattern is observed for the results of those who are aware that antibiotics are not

effective against colds and flu, although knowledge of this is better, with Estonia and Italy

both scoring above the average at 54% and 52% correct responses respectively (Figure 2.1).

These are followed by Spain (44%), Hungary (37%), Greece (34%), Romania (33%), and

Cyprus (24%). Again, the highest scoring countries are mostly northern and western with

Sweden scoring the highest at 77 percent, followed by Denmark, Finland, the Netherlands

and the UK all scoring 70 percent or higher. A similar educational gap as described above is

observed, with 41% of those with the lowest level of education giving the correct response

compared with 63% of those with the highest level of education (Table 3.1). Again, obtained

information proves to be important. 71 percent of those who received information from a

media source and 58 percent of those who received information from a healthcare

professional answered this question correctly, versus 45 percent of those who received no

information.

9

Figure 2.1. ‘For each of the following statements, please tell me whether you think it’s true or false. ‘Antibiotics are effective against cold and flu.’’


Table 3.1. ‘For each of the following statements, please tell me whether you think it’s true or false. ‘Antibiotics are effective against cold and flu.’’ By education, received information, and information source.

EU27 (n= 26,680)


52% 41%

Education (End of)



Advice from a professional 58% 38% From media/campaigns 71% 25% Did not receive information 45% 46% *Adapted from: Eurobarometer (2013: 32)

While a vast majority of respondents are aware that antibiotics may become ineffective

following unnecessary use, obtained information shows interesting intersections (Table 4.1).

94% of those who received information from a media source and 85% of those who received

information from a healthcare professional answered this question correctly compared to

80% of those who received no information. 11% of those who did not receive information

0

10

20

30

40

50

60

70

80

90

100

SE DK FI NL

UK

BE

LU IE FR CZ SI HR EE SK IT

EU2

7 ES DE LT LV MT

HU EL PL

RO AT

BG PT

CY

Percentage of correct responses to the statement;

'Antibiotics are effective against cold and flu'

KEY


10

were unable to answer this question compared to only 4% of those who received information.

While all these results regarding information and knowledge of antibiotics do not imply

causation, they do support the principle that propagating information concerning the correct

use of antibiotics is essential in tackling widespread misconceptions regarding appropriate

use.

Figure 3.1. ‘For each of the following statements, please tell me whether you think it’s true or false. ‘Unnecessary use of antibiotics makes them become ineffective.’’


Table 4.1. ‘For each of the following statements, please tell me whether you think it’s true or false. ‘Unnecessary use of antibiotics makes them become ineffective’’. By education, received information, and information source.

EU27 (n= 26,680)


84% 8%

Education (End of)



Advice from a professional 85% 10% From media/campaigns 94% 3% Did not receive information 80% 9% *Adapted from: Eurobarometer (2013: 36).

0

10

20

30

40

50

60

70

80

90

100

SE DK SI NL EL FR CY

CZ FI

MT

UK

LU SK BE ES DE

HR PL IE LT

EU2

7

AT EE PT

LV BG

HU IT RO

Percentage of correct responses to the statement; 'Unnecessary use of antibiotics makes them become ineffective'

KEY


11

Results of the public’s knowledge of the fourth and final statement: ‘Taking antibiotics often

has side-effects, such as diarrhoea’ show disparity from the results of the previous

statements. Although two thirds (66%) gave a correct response, there is clearly more

uncertainty as 19 percent of respondents were unable to give a response. In addition the

geographic divide that is evident in the previous results (better knowledge in the Northern

and Western countries compared to the Southern and Eastern countries) is not as

distinguished in the results to this statement. Three of the five highest scoring countries are

Eastern European countries: Poland (78% correct responses), Estonia (77%) and Slovakia

(75%). Similar to the previous results, those who received information are more likely to give

a correct response than those who did not (73% correct responses versus 62% respectively).

Media campaigns proved to have less of an impact on knowledge of this statement

compared to the previous statement and compared to advice from a healthcare professional.

Results to this statement by education level are not presented in the report.

Figure 4.1. ‘For each of the following statements, please tell me whether you think it’s true or false. ‘Taking antibiotics often has side-effects, such as diarrhoea.’’


0

10

20

30

40

50

60

70

80

90

100

PL

LU EE FI SK SI EL AT

CY

DK LT BG IT DE

MT

BE

HR FR

EU2

7 LV UK CZ

PT

HU SE NL

ES IE RO

Percentage of correct responses to the statement;'Taking antibiotics often has side-effects, such as diarrhoea'

KEY


12

Table 5.1. ‘For each of the following statements, please tell me whether you think it’s true or false. ‘Taking antibiotics often has side-effects, such as diarrhoea’’. By received information, and information source.

EU27 (n= 26,680)


66% 15%

Received information


Advice from a professional 75% 13% From media/campaigns 72% 13% Did not receive information 62% 16% *Adapted from: Eurobarometer (2013: 40)

While these results prove interesting, they still beg the question why is increased awareness

of antibiotic use and misuse expected to be associated with lower levels of self-medication

with antibiotics, among those who self-medicate? Although a clear correlation is found

between receipt of information and levels of objective knowledge, this is not to imply

causation, particularly considering the positive relationship of some socio-demographic

variables –i.e. education – with both variables. It is therefore impossible to go further to argue

that information provision will be associated with patient behaviour. This expectation comes

from results which show that when information is provided by a healthcare professional,

patients are more likely to change their behaviour – than when it is provided by a media

outlet - in accordance with this information or advice. This point will be expanded upon in the

theory section.

On average, only a third of respondents remembered receiving information about not taking

antibiotics unnecessarily in the last 12 months. Of the 33 percent who received information,

almost a fifth received it from a media source as opposed to only 11 percent from a

healthcare professional. Media sources are most common in France (52%), Belgium (35%)

and Luxembourg (32%), and lowest in Romania (7%) and Hungary and Portugal (both

5%). Variation in advice from professionals ranges from roughly a fifth in Luxembourg,

Romania and Italy to 6 percent in Portugal and Spain and 5 percent in the Netherlands. The

case of Spain is particularly interesting as we can see a decrease of 13 percentage points in

advice from the media and 12 percentage points in advice from a professional since 2009.

Those residing in southern European countries and those with poorer levels of knowledge

about antibiotics are more likely to have their views changed by such information. A pattern

can be drawn from the results in that countries with a higher level of correct knowledge in the

general population also have lower levels of self-medication. In that sense, a possible

method of reducing overuse or misuse of antibiotics could be through educating those who

13

use them by informing them of appropriate use and the consequences of misuse. Education

interventions have been found to reduce absolute rates of antibiotic consumption (AHRQ,

2006).

The importance of the source of information is also highlighted in the Eurobarometer (2013)

study, as 80 percent of those who received information from a healthcare professional

reported that they will consult a doctor about the use of antibiotics in the future compared

with only 69 percent of those who received information from media campaigns. While a

striking majority (94%) stated that they would choose to see a medical professional in order

to receive reliable information about antibiotics, there was also a stronger preference to visit

a doctor (87%) than a pharmacist (49%).

These results coincide with the argument that while media campaigns are effective at

disseminating information, healthcare professionals are more effective in transforming

behaviour change and as such, the role of healthcare professionals in educating patients

about the use and misuse of antibiotics is very important if we are to change their behaviour.

This study found no socio-demographic differences on the question of where the

respondents obtained their last course of antibiotics. This is due to an overwhelming majority

on average who indicated that they obtained them with a prescription. Therefore, low

response rates for the self-medicating answers (antibiotics left over from a previous course

or obtained without a prescription) imply that differences in these categories are not

statistically significant. This is one reason to conduct an in-depth analysis of those who did

self-medicate (ARNA had samples of 400 persons per country), in order to determine

differences in this category.

14

2. Theory

The use of antibiotics by patients is firmly controlled through a variety of mechanisms

including controlled clinical trials, regulation and release of pharmaceuticals on the

commercial market, and the controlled distribution by physician prescription through licensed

pharmacies. The goal of these successive steps is to increase the likelihood that patients

receive the most appropriate and efficacious medication for specified indications. It is the

final steps of this process which are of concern for this thesis. The drug dispensing and

consumption processes are intimately involved with human factors such as prescribing

practices of the physician, expectations of the patient and their interaction in general.

Arguably these human factors have the ability to distort and negate the effectiveness of the

entire process which precedes the interaction of the patient, physician and medication (Hulka

et al., 1976). More recently, Awad and Aboud (2015) cited the complex interaction of human

factors and relationships which inevitably have an effect on consumption and use patterns,

including inadequate informing of patients by physicians, patients perceptions of the patient-

prescriber communication, patients’ knowledge, expectations and beliefs. They concluded

that these factors perpetuate the misuse of antibiotics.

Many interdependent factors have been shown to produce an increase in the non-prudent

use of antibiotics. Previous studies from across the globe have shown improper use of

antibiotics and noncompliance with the regimen to be strongly associated with public

knowledge and awareness of the subject (Pavyde et al., 2015). In order for self-medicators of

antibiotics to make an informed decision to change their behaviour, they must first be given

some scientific knowledge or incentives to correctly use antibiotics. McNulty et al. (2007)

found that awareness of public campaigns in England among both self-medicators and

prescribed users was associated with better knowledge of antibiotics. This was reflected in

the 42% versus 24% correct responses to statements, particularly the statement that

antibiotics do not cure most coughs and colds. This is highly relevant, as treating throat

symptoms or bronchitis is found to be the top medical as opposed to practical reason for self-

medicating (Grigoryan et al., 2006). McNulty et al. (2007) also found a significant association

between knowledge of antibiotics and being more likely to finish a course of antibiotics.

However, knowledge was also associated with being more likely to take antibiotics without

being advised by a healthcare professional.

There remains the question: why is it expected that an increase in knowledge would result in

actual behaviour change? This argument is derived from the Rational Choice Theory

(Hechter & Kanazawa, 1997). The main assumption of this theory is that people are

15

individuals with agency who act rationally in response to the information made available to

them. It is individuals who take action, and according to this theory it is assumed that this

action is based on rationally evaluated, self-maximizing conclusions. Actions are also based

on optimality. Individuals base their actions on both their preferences and the opportunities or

constraints present to them. In this way, individuals do the best they can given their

circumstances (Abell, 2000). However rationality is the most dominant assumption of the

theory. According to this assumption all individuals behave in the manner which is perceived

as the most self-promoting or follow actions which benefit themselves most. Applied to the

context of self-medication with antibiotics: informing the public can rectify ignorance about

antibiotic resistance and proper use. This information will arouse concern which in turn will

materialise into behaviour change which yields the most favourable outcome for the

individual. This argument is logically applied to the context of antibiotic misuse as there are

many fundamental issues about which the public is misinformed (Eurobarometer, 2013). An

explanation of these issues may lead to a change in behaviour as patients evaluate the

benefits of not self-medicating, both directly for themselves (e.g. because antibiotics used

incorrectly may result in adverse side-effects) and / or indirectly through benefiting society

(e.g. if antibiotic resistance were not an urgent issue to be dealt with, more resources would

be available to develop treatments for other illnesses or diseases).

This paradigm of rational action is best exemplified by two theories; the theory of reasoned

action (Ajzen & Fishbein, 1980) and the theory of planned behaviour (Ajzen, 1985). Both

theories explain behaviour through the attitudes towards the behaviour and the intention to

adopt it. Intention is the foremost predictor of behaviour as it is a function of the attitudes

towards the behaviour. These attitudes are determined by individual beliefs concerning the

outcome of such behaviour, including an evaluation of this outcome. This is the rational

calculation. The intention to adopt the behaviour is also determined by subjective norms.

These norms come from the individual’s beliefs of the expectations of their significant others

and the intention to meet said expectations. In addition, both the intention and the behaviour

are determined by the individual’s perceived control over the behaviour. That is, their

perceptions of their own ability to perform the behaviour. (Bartiaux, 2008) The theory of

planned behaviour adds to this with the assumption that only specific attitudes regarding the

behaviour are expected to predict it. By placing beliefs, both behavioural and normative

beliefs, at the beginning of the causal model, the fundamental role of information in

determining behaviour is highlighted. Information has the potential to modify these beliefs

and, as consequence, the attitudes to the behaviour and the intention to behave in a certain

way (Bartiaux, 2008).

16

Although not directly a source of self-medication, inappropriate prescribing by a physician

facilitates self-medication through left-overs, especially where antibiotics are dispensed in full

packages as is the case in the seven countries under study. Effective communication

between healthcare professionals and patients is fundamental to limit inappropriate

prescribing and dispensing of antibiotics. Stevenson et al. (1999) found that doctors who

provided information and education to patients were successful in reducing antibiotic

prescriptions. While some GPs may prescribe inappropriately in the belief that patient

satisfaction primarily stems from such an outcome, this has not been found to be the case.

On the contrary, patients were found to be content to leave without a prescription if an

explanation was provided (Stevenson et al. 1999). In addition, previous experience of being

prescribed an antibiotic may give the patient ample confidence to use said antibiotic without

seeking medical advice. This and evidence from Jenkins et al. (2003) that 62% of

consultations resulted in at least one kind of problem such as a patient getting less than what

they wanted, a mismatch in patient and doctor perceptions, unwanted prescriptions,

unnecessary prescriptions, inappropriate prescribing, patient non-adherence or problems

with medication, highlight the importance of effective communication in the consultation

process. If the general public does not have a clear understanding of the potential

consequences of misusing antibiotics, a new social norm for antibiotics as a last resort is

improbable (Pinder et al., 2015).

From this we can derive the first hypothesis;

H1: Of those who self-medicate, the level of self-medication will be lower among those who

have received information than among those who did not receive information.

A second point of interest is whether the type of information received by patients who self-

medicate with antibiotics is associated with their level of self-medication. Although to my

knowledge there is no previous literature detailing which type of information may motivate a

patient to not self-medicate with antibiotics, a hypothesis will be derived based on Protection

Motivation Theory (PMT) (Rogers, 1975) and the usefulness of negativity in conveying

information (Geer, 2006).

The starting assumption is that in health communication, the level of fear arousal may

increase the perceived seriousness of the issue at hand, and the perceived susceptibility to

being adversely affected by said issue. This effect is explained through Protection Motivation

Theory (PMT), based on the works of Lazarus (1966) and Leventhal (1970). PMT argues that

in response to fear-evoking information, individuals exercise either adaptive or maladaptive

17

coping responses as the result of two processes. Maladaptive coping responses are those

which pose a health risk to the individual including the absence of a behaviour (e.g. missing

a check-up) and behaviours which may result in consequences (e.g. smoking). In the context

of self-medication with antibiotics, an adaptive coping response would be to only use

antibiotics when they are prescribed while an example of a maladaptive coping response

would be to deny the issue.

The two processes which result in either of these responses are 1) health threat appraisal

and 2) coping appraisal. During the threat appraisal process, factors related to the evaluation

of the threat are assessed. In the context of health behaviour, perceived susceptibility to the

threat and the perceived severity of the threat are estimated. Both evaluations are expected

to reduce the adoption of maladaptive responses. In this sense, fear arousal is beneficial in

that it enhances the perceived severity and susceptibility and thereby increases the

protection motivation. During the coping appraisal process, those factors relevant to the

evaluation of coping responses are appraised. These are both the response efficacy - the

likelihood that carrying out the behaviour will result in a reduction of the threat - and self-

efficacy, taken from Bandura’s social learning theory (1977). Self-efficacy is one’s belief in

one's own ability to successfully adopt or complete the behaviour. According to PMT,

adaptive behaviour is most likely when both response efficacy and self-efficacy are high.

In addition to Protection Motivation Theory, the benefits of utilising negative messages to

convey information with the goal of influencing behaviour have been argued and documented

by Geer (2006). In the past, the dominant view was that positive messages were key in

changing behaviour. This came from the idea that providing alternative solutions to problems

was a constructive way to motivate. However an informed decision on any topic requires

knowledge of both good and bad aspects. Without the knowledge of the bad aspects, we put

ourselves liable to risk. In addition, Geer argues that negative information is additionally

beneficial in that it is more likely to be backed up by evidence. This is because positive

information is more readily accepted by the public whereas negative information demands

evidence to solidify its credibility. In turn, evidence based arguments are more likely to

change behaviour. Applied to the context of self-medication with antibiotics, it could be

argued that when patients who self-medicate receive information on the side-effects of

incorrect antibiotic use, they will be more motivated and therefore more associated with lower

level of self-medication. This negative information is more likely to arouse fear in the patient

resulting in the rational decision to protect themselves by not self-medicating with antibiotics.

Information on side-effects will also be backed up by evidence, certainly more so than

18

general information on antibiotics, specifically the operationalization of general information in

this thesis, as will be discussed in chapter 4.

Additionally, it is also fair to consider that since most self-medication occurs for the medical

reason ‘to treat colds or coughs’, and knowledge of the fact that antibiotics are ineffective

against these ailments is low (Eurobarometer, 2013), the effect of receiving information

regarding side-effects may be the strongest as it includes the ‘information concerning

conditions in which not to take an antibiotic’.

From this hypothesis 2 is derived:

H2: Among patients who self-medicate, the effect of receiving information on ‘side-effects’

will be most associated with lower levels of self-medication.

A third interesting aspect under investigation is whether the source of the information has an

association with lower levels of self-medication with antibiotics. Knowledge of this can

potentially influence policy. It is beneficial to know whether efforts should be spent on

governmental interventions such as campaigns or whether the focus should be on the

communication between the healthcare professional and the patient, such as part of a

healthcare-professional oriented intervention. While it has been shown that media campaigns

are more effective at disseminating information regarding antibiotics, clinicians have been

found to be more effective at changing the behaviour of their patient (Pinder et al., 2015).

Arguably what is important here is how the patients receive the information from their source

and whether they perceive their source as credible or not.

Previous research has repeatedly cited physicians as the most preferred and trusted

sources, ranked by patients (Narhi, 2007; Mayer et al., 2007; Rutten et al., 2005; Hesse et

al., 2005). However although they are preferred, they are not necessarily the most frequently

used source. Patients have also been found to utilize information leaflets and the internet

more than physicians (Narhi, 2007; Hesse et al., 2005). Hesse et al. (2005) found that 50

percent of patients cited physicians as their preferred first point of contact regarding specific

health information, however only 11 percent actually went to their physician first, compared

to 48 percent who sought information online first. This could be an indication that the degree

of distrust of the internet as a source of information is not as strong as previous surveys have

concluded. Perhaps the convenience of the internet outweighs the concerns about the quality

of the information; or perhaps people have found strategies to identify and avoid less

trustworthy information; or perhaps it is a combination of these (Eysenbach, 2008). Genni et

19

al. (2006) argue that patients use the information they find online as a discussion point with

their doctors. In this way, although patients generally find media information trustworthy, they

still seek confirmation from their doctor. Arguably then, it is this professional confirmation

which has a stronger influence over people’s attitudes and consequently their beliefs

concerning antibiotics and antibiotic resistance, than information found directly from

media/internet.

From this a third and final (exploratory) hypothesis is derived:

H3: When information is received from a healthcare professional, levels of self-medication

among those who self-medicate will be lower than when the information is obtained from the

media/internet.

20

*Note: Cells with a dashed border are not tested in the analysis.

Figure 5.1.: Conceptual Model* (Source: Adapted from Ajzen (2006))

21

3. Methods

Data will be used from the ‘Antimicrobial resistance and causes of non-prudent use of

antibiotics in human medicine’ (ARNA) research project, conducted by the Netherlands

Institute for Health Services Research (NIVEL). The project runs for 2 years and consists of a

broad analysis of the whole of the EU followed by an in-depth analysis of seven countries by

use of an additional questionnaire. These seven countries were selected as they were found

by the Eurobarometer (2013) to have high levels of self-medication with antibiotics. These

are: (percentages indicate the share of antibiotics used for self-medication of all antibiotics

collected) Romania (20%), Greece (16%), Hungary (8%), Spain (8%), Italy (5%), Cyprus

(10%), and Estonia (7%) (Eurobarometer, 2013: 12). Although the percentage of self-

medication found in Italy is equivalent to that of the EU27 average (Eurobarometer, 2013:

12), it was included in the ARNA study due to an explicit request for inclusion by the

European Commission, one of the funders of the research project. The EC requested the

inclusion of Italy in order to confirm the findings of the Eurobarometer for Italy which were

suspected to be incorrect.

Data Collection

Data collection was outsourced to the Dutch survey company TNS NIPO. This company has

a database of thousands of screened respondents. For each survey, a random selection is

made from the database. For the current study, the sample was selected using Random

Digital Dialling (RDD) in order to conduct Computer-assisted telephone interviews (CATI).

This type of surveying has been found to be preferable to self-administered surveys in terms

of both quality and quantity of responses (Bowling, 2005). In Cyprus, Estonia, Greece,

Hungary, Italy and Romania, the sample was selected randomly through telephone numbers

which were randomly generated. In Spain, the sample was selected randomly through

national telephone directories. TNS NIPO employed the triple C method to collect the data in

the seven countries. This means that data collection was coordinated from one central

location in Amsterdam in one central database, and each country was assigned its own

Triple C team who conducted the telephone interviews and sent the data back to the central

database. Each country received the same questionnaire, translated into the respective

national language.

Overall, data was collected between December 2014 and February 2015. In Cyprus and

Greece, data collection began in December and ended January, lasting roughly six weeks. In

Estonia and Romania, data collection took place between January and February, also taking

22

roughly six weeks. In the remaining countries - Hungary, Italy and Spain - data collection ran

from December to February and lasted about ten weeks.

A target of 400 self-medicating respondents from each country was set for the ARNA project.

A sample was selected in each country in order to reach this target. If the target was not met

after the whole sample was contacted, the parameters of the sample were extended.

Contacts were stopped once the target of 400 self-medicating respondents was met.

Implications of this method on the results will be discussed in Chapter 6. Difficulties in

collecting 400 self-medicating respondents in Estonia led to the conclusion that a sample 200

for this country would suffice.

Initially, 165,694 contacts were made across all seven countries (Figure 6.1). This initial

sample was based on the incident rates of self-medication found in the Eurobarometer

(2013) report. Based on these levels of self-medication, the costs for the fieldwork were

calculated, including how many contacts should be made in order to reach the target of 400

respondents per country. From the 165,694 initial contacts, 65,103 were successful. That

means that 100,591 contacts were unsuccessful, either the phone number did not work, the

call was unanswered or the respondent refused to participate. Information on the non-

respondents has not been provided by TNS NIPO to date.

The 65,103 respondents were screened by two questions. First they were asked ‘Have you

or your child(ren) taken any antibiotics orally such as tablets, powder or syrup in the last 18

months?’. Those who answered ‘no’ to this question were dismissed from the interview

process. The 29,647 respondents who answered ‘yes’ were screened further with the

question ‘How did you obtain these courses of antibiotics in the last 18 months?’. Those who

answered that they obtained them with a prescription were only asked their sex and age.

Respondents who answered that they obtained the antibiotics without a medical prescription

continued with the full interview. This resulted in a total of 2,601 full interviews consisting of

400 respondents from Cyprus, Greece, Hungary, Italy and Spain, 200 respondents from

Estonia and 401 respondents from Romania. As this sample of self-medicators also included

respondents who answered the interview on behalf of their child(ren), said respondents were

removed from the data for the analysis (Table 5.1). This means that from the 2,601 self-

medicating respondents, 755 gave either the response ‘yes my child(ren) have’ or ‘yes, I and

my children have’. These respondents were excluded from the analysis which results in a

total sample of 1,846 respondents who self-medicated with antibiotics in the past 18 months.

23

Figure 6.1. Flow diagram of data collection

165,694 respondents (Total sample)

CY 1,882

EE 6,312

EL 1,487

ES 6,303

HU 6,676

IT 4,687

RO 2,300

CY 23,410

EE 26,780

EL 28,358

ES 17,182

HU 26,688

IT 31,312

RO 11,964

65,103 respondents (Total interviews)

CY 3,423

EE 16,779

EL 3,330

ES 11,795

HU 17,205

IT 9,313

RO 3,936

29,647 respondents (who used antibiotics with or without prescription)

2,601 respondents (who used antibiotics without medical prescription for their last

course, including children)

CY 400

EE 200

EL 400

ES 400

HU 400

IT 400

RO 401

1,846 respondents (who used antibiotics without medical prescription)

CY 315

EE 164

EL 295

ES 256

HU 261

IT 269

RO 271

24

Table 5.1. Responses to the question ‘Have you or your child(ren) taken any antibiotics orally such as tablets, powder or syrup in the last 18 months?’

Total Cyprus Estonia Greece Hungary Italy Romania Spain

Yes, I have 1,831 (70%) 315 (79%) 164 (82%) 295 (74%) 261 (65%) 269 (67%) 271 (68%) 256 (64%)

Yes, my child(ren) have 313 (12%) 37 (9%) 13 (7%) 54 (14%) 58 (15%) 72 (18%) 35 (9%) 44 (11%)

Yes, my children and I have 442 (17%) 48 (12%) 23 (12%) 50 (13%) 81 (20%) 49 (12%) 91 (23%) 100 (25%)

No 15 (0.6%) 1 (0.25%) 10 (3%) 4 (1%)

Total 2,601 400 200 400 400 400 401 400

Table 6.1. Data collection in the seven European countries.

ABBR. Countries Method Sample selection Number of interviews

Fieldwork dates Population

CY Cyprus Telephone interviews-CATI Random through telephone numbers 400 03/12/2014 16/01/2015 23,410 EE Estonia Telephone interviews-CATI Random through telephone numbers 200 21/01/2015 26/02/2015 26,780 EL Greece Telephone interviews-CATI Random through telephone numbers 400 15/12/2014 20/01/2015 28,358 EE Spain Telephone interviews-CATI Random through national telephone

directories 400 09/12/2014 19/02/2015 17,182

HU Hungary Telephone interviews-CATI Random through telephone numbers 400 16/12/2014 23/02/2015 26,688 IT Italy Telephone interviews-CATI Random through telephone numbers 400 03/12/2014 09/02/2015 31,312 RO Romania Telephone interviews-CATI Random through telephone numbers 401 07/01/0215 20/02/2015 11,964 Total 2,601 165,694

25

Representativeness of the data

In Cyprus, 41% of respondents came from the Capital, Nicosia, while the remaining 59%

were divided between four other smaller cities in the south (25%), southeast (17%),

southwest (7%) and east (10%). In Greece, all 400 participants were sourced from Athens

with roughly a fifth to a quarter from each geographical district (north, south, east and centre

and west). In Hungary, respondents were sourced from seven different geographical regions

with the largest section (29%) coming from Central Hungary and the smallest (9%) coming

from Southern Transdanubia. In Italy, data collection occurred in four regions; one third of

respondents came from the south and islands, one quarter from the north-west and a fifth

from both the centre and north-east region. Data collection was mostly divided by region in

Spain. Respondents ranged from 0.6% from La Rioja to 17.3% in Andalusia and were spread

over 18 regions including the Capital and the Canary Islands. In-depth descriptions of the

location(s) of data collection in Estonia and Romania are not available to date.

Questionnaire

A structured questionnaire of 34 questions was developed in English and translated for each

country. Topics in the questionnaire included:

1. Respondents’ and their children's antibiotic use without a prescription, the source of

said antibiotics, when they were taken and the reason why they were used without a

prescription.

2. Other people’s use of antibiotics without a prescription, including friends, family,

neighbours, partner.

3. Type and dose of antibiotics used

4. Medical reasons for using the antibiotics without a prescription

5. Whether or not information was received on how to prudently use the medication and

if so what information was obtained.

6. Knowledge of antibiotics

7. Experienced positive consequences and negative side-effects.

8. Role of the general practitioners and pharmacists in over-the-counter use of

antibiotics - how they prevent and simulate this.

26

4. Data Analysis

Operationalization

The variables of interest for this study are those concerning the level of self-medication of

respondents and the information they received regarding the use and side-effects of

antibiotics. This includes both type and source of information. In addition, relevant

background variables of the respondents are controlled for. A country indicator is included in

the analysis to determine any variation between the countries.

The level of self-medication with antibiotics is defined as the number of courses the

respondent has used without a prescription in the past 18 months. In the questionnaire,

respondents could indicate ‘1 course’, ‘2 - 3 courses’ or ‘4 or more courses’. Preliminary

analysis showed weak effects when the dependent variable was coded as three categories

(Table 9.2, Appendix). This is due to the distribution of the data: the small n-values of those

indicating ‘2 - 3 courses’ or ‘4 or more courses’ compared to those indicating ‘1 course’ of

self-medicated antibiotics. Therefore, for this analysis, level of self-medication is

dichotomously coded with 0 indicating ‘1 course’ of self-medicated antibiotics in the past 18

months and 1 indicating ‘2 or more courses’.

The three key independent variables are whether the respondent received information, the

type of information received and the source of this information.

Respondents were asked whether or not they had received any information on how to use

the last course of antibiotics that they self-medicated with in the past 18 months. This is

coded dichotomously.

Respondents were asked what information they obtained about how to use the last course

of antibiotics that they self-medicated with. The types of information respondents could

indicate that they received are: i) Complete the full course, ii) Only take the antibiotics over

the prescribed period at the correct dose, iii) Do not give the antibiotics to anyone else, iv)

Return left-over antibiotics to the pharmacy, v) Advice about how to act if the condition

persists or the infection becomes worse, vi) Information about conditions under which the

antibiotics should not be taken, information concerning: vii) Possible allergies, viii) Side-

effects, and ix) Drug interaction, x) Other, and xi) Don’t know / refuse.

These response categories are used to create the three ‘types of information’ variables;

proper use, general information, and side-effects.

The variable proper use is created by grouping the response categories i) Complete the full

course and ii) Only take antibiotics over the prescribed period and at the correct dose.

27

The variable general information is created by grouping the responses iii) Do not give them

to anyone else, iv) Return left-overs to a pharmacy and v) How to act if the condition persists.

The variable side-effects is created by grouping the response categories vi) Information

about when antibiotics should not be taken, vii) Information about allergies, viii) Information

about side-effects and ix) Information about drug interaction.

The ‘Other’ and ‘Don’t know / refuse’ categories are recoded as missing.

Respondents were also asked to indicate the source of this information. Answers were

initially recorded as i) Pharmacy staff, ii) General practitioner, iii) Other healthcare

professional, iv) Internet, v) National campaign, vi) Other media, vii) Other and viii) Don’t

know / refuse.

These responses are grouped according to the two information sources of interest for this

study: Healthcare professional and the Media.

Response categories i) Pharmacy staff, ii) General practitioner and iii) Other healthcare

professional are grouped together to create the variable Healthcare professional.

All the categories iv) Internet, v) National campaign and vi) Other media are grouped

together to create the variable Media.

‘Other’ and ‘Don’t know / refuse’ responses were recoded as missing.

Background variables of interest are Education, Employment status, Health insurance, Self-

reported health, Smoking status and Presence of a longstanding illness or condition.

Education is the second background variable, operationalized as ‘Low’, ‘Medium’ or ‘High’.

Health insurance is the next control variable. Respondents could specify whether they are

covered by ‘Private insurance’, ‘Public insurance’, ‘Community / social insurance’ or ‘Other

health insurance’. These four types are grouped together so that Health insurance is

operationalized as ‘Insured’ or ‘Uninsured’.

Self-reported health was originally recorded as ‘Excellent’, ‘Good’, ‘Fair’, ‘Bad’ or ‘Very bad’.

These categories are grouped to make three; ‘(Very) good’, ‘Fair’ or ‘(Very) bad’ self-reported

health categories.

Presence of a longstanding illness or condition is the final background variable. Here the

respondents could specify whether they suffer from ‘Asthma’, ‘Chronic obstructive pulmonary

disease’, ‘Emphysema’, ‘Diabetes mellitus’, ‘Cardiovascular disease’, ‘Hypertension’ or

‘Other’. These are grouped together to indicate the presence of a chronic illness.

These background variables were selected, as preliminary analysis suggested potential

significance in the analysis while several other background variables were excluded.

28

Although education did not prove to be significant in the preliminary analysis, it was included

in the logistic regression due to the results of the Eurobarometer (2013) which suggested an

intersection of education, information and knowledge of antibiotics.

Analysis

When the dependent variable is dichotomously coded as is the case for this research, a

logistic regression rather than a multiple regression or discriminant analysis is suitable

(Hosmer & Lemeshow, 1989; SPSS, 1989). Therefore, the chosen method of analysis is a

logistic regression, followed by a Poisson regression as a robustness check. In order to

perform the Poisson regression, the dependent variable is changed from a binary variable to

a three category variable. Consequently, level of self-medication is recorded as 1 course, 2-3

courses or 4 or more courses. Although significant variation between the countries is not

expected, the final model of the logistic regression is run separately for each country to give

further explanation to the effects found in the grouped regression model.

The building of the models for the logistic regression analysis is as follows:

Model 1 consists of background variables; area of residence, education, employment status,

health insurance, smoking, perceived health, chronic illness, age and sex and a country

indicator to provide insight into potential variation between the countries. This model will give

an idea of the characteristics of the sample and an initial sense of why respondents may be

more susceptible to self-medication with antibiotics.

In Model 2 the impact of generally receiving information is added. This will test hypothesis 1

by indicating the general effect of receiving information.

In Model 3, the impact of general information is removed and is replaced by the three

different types of information. This will test hypothesis 2 by determining whether the three

types of information received are equal in terms of their effect on self-medication with

antibiotics.

In Model 4 the effect of the source of information is added. If the effects observed in model 3

change or disappear, this can be interpreted as a stronger effect of the source of information,

rather than the type of information obtained. In that case, the source of information would be

considered more important than the type of information in reducing levels of self-medication

with antibiotics. This model will also test the explanatory hypothesis, as it will tell which

source of information produces a stronger effect in reducing levels of self-medication.

All analyses are carried out by the procedures in STATA/SE 13.

29

5. Results

First I will present the results of the descriptive analyses, followed by the results of the

logistic regression. Tables of the preliminary analysis results (multilevel regression),

descriptive results and robustness check results (poisson regression) can be found in the

appendix (Tables 1.2 – 10.2).

Descriptives

From preliminary results, it was found that the variance in levels of self-medication is as

follows: of the entire sample of self-medicators (n= 1,808), 67 percent stated that they self-

medicated with one course of antibiotics in the past 18 months, while 33 percent stated 2 or

more courses. Three quarters (74%) had received advice or information on how to use the

antibiotics while one quarter had not (n= 1,833). Four fifths received advice regarding the

proper use of antibiotics (80%), two fifths had received general information (41%) and two

fifths received information on side-effects (42%) (n= 1,294). A majority (89%) received this

information from a healthcare professional while 5 percent received it from a media outlet (n=

1,351).

Figure 7.1. Respondent’s level of self-medication in past 18 months (n= 1,808).

In terms of sample characteristics, half have achieved a medium level of education (50%),

more than a third achieved a high level of education (36%), while the remaining 14 percent

are lowly educated (n= 1,831). Almost three fifths report (very) good self-perceived health

(58%), one third report fair health while the remaining nine percent report (very) bad health

0

10

20

30

40

50

60

70

80

90

100

1 course 2 + courses

Level of self-medication

30

(n= 1,832). One third of the sample have a chronic illness (35%) or condition while two thirds

do not (n= 1,829).

Almost two thirds of the studied sample are female (65%) while slightly more than two thirds

are male (35%). This is comparable to the sample of those who used antibiotics with a

prescription in the past 18 months, of whom three fifths are female (61%) and two fifths male

(Table 7.1). Of the sample of self-medicators, 30 percent are between the ages of 18 and 40,

half of the respondents are between the ages of 41 and 60 while the remaining 16 percent

are aged 65 or over. This is also comparable to the sample of respondents who did not self-

medicate with antibiotics. Of this sample, one quarter are aged between 18 and 40, half are

between the ages of 41 and 60 while the remaining quarter are 65 years old or older (Table

7.1). From this, it is determined that those who self-medicate with antibiotics are not selective

on the basis of sex or age.

Table 7.1. Comparison of sex and age between those who did not self-medicate with antibiotics and those who did self-medicate.

Self-medicators Prescribed users

N (%) SD Min. Max. N (%) SD Min. Max.

Sex 0 1 0 1

Male 666 (36%) 24,132 (39%)

Female 1,180 (64%) 38,361 (61%)

Age 16.42 15 99 16.27 2 100

0-17 7 (0.4%) 212 (0.3%)

18-40 539 (30%) 14,559 (24%)

41-64 927 (50%) 31,614 (51%)

65+ 364 (20%) 15,624 (25%)

Regression 1.1 Logistic Regression.

To test the three research hypotheses regarding the relationship between receiving

information on the use of antibiotics and respondents’ levels of self-medication, a two

predictor logistic model is fitted to the data. Results of the regression analysis are presented

in Table 8.1. As the interpretation of the logistic coefficient is not as straightforward as in the

case of interpreting a multiple linear regression coefficient, the conventional Beta coefficient

is rewritten in terms of the odds of engaging in high levels of self-medication with antibiotics

(Pyke & Sheridan, 1993). This is defined as the ratio of the probability that high levels of self-

medication will occur, to the probability that it will not. Factors with values greater than one

indicate that the odds of high levels of self-medication with antibiotics are increased, and

factors with values less than one indicate that the odds are decreased.

31

According to model 1, the odds of self-medicating more with antibiotics are positively related

to having a chronic illness (OR: 1.47, 95% CI: 1.16-1.85). In other words, respondents with a

chronic illness are more associated with higher levels of self-medication with antibiotics than

those who do not have a chronic illness (Table 8.1). All other background characteristics

show no significant relationship with levels of self-medication with antibiotics. In addition, little

country variation is found as only the Estonian dummy variable produces a significant

association with lower levels of self-medication with antibiotics (OR: 0.48, 95% CI: 0.30-

0.75).

To test hypothesis 1, the variable of generally receiving information is added to the model

(model 2). When this is done the positive association of having a chronic illness remains

stable. The effect of generally receiving information is surprisingly associated with higher

levels of self-medication with antibiotics. That is, respondents who received information on

how to use the last course of antibiotics that they self-medicated with are more likely to report

higher levels of self-medication. However this effect is only borderline significant (OR: 1.25,

95% CI: 0.99-1.58), meaning, Hypothesis 1: ‘Of those who self-medicate, the level of self-

medication will be lower among those who have received information than among those who

did not receive information.’ is rejected. In addition, country variation remains relatively stable

in model 2 compared to model 1.

In order to test hypothesis 2, the general effect of receiving information is removed from the

model and the three different types of information are added (model 3). When this is done

only one type of information bears a significant relationship to the prediction of respondents’

levels of self-medication: proper use. Respondents who received information regarding the

proper use of antibiotics are associated with lower levels of self-medication (OR: 0.74, 95%

CI: 0.56-0.98). Although not significant at the p<0.05 level - yet borderline significant - the

effect of receiving information regarding side-effects is associated with higher levels of self-

medication with antibiotics. Receiving general information is associated with lower levels of

self-medication, however it is also not significant at the p<0.05 level. This indicates that

receiving either of these types of information is not associated with lower levels of self-

medication among patients who self-medicate. Thus from these results Hypothesis 2 ‘Among

patients who self-medicate, the effect of receiving information on ‘side-effects’ will be most

associated with lower levels of self-medication.’ is rejected. Receiving the more negative

information, side-effects, did not yield a significant result on respondents' level of self-

medication. However further investigation could be warranted on the effect of receiving

information on proper use. Additionally, in model 3 the effect of having a chronic illness on

respondents’ level of self-medication remains constant. The dummy variable for Spain

32

produces a significant association with lower levels of self-medication with antibiotics (OR:

0.58, 95% CI: 0.37-0.91), while the rest of the country variables remain relatively constant.

In order to test the third and final hypothesis, the two sources of information are added to the

model (model 4). Both sources of information are not significant at the p<0.05 level. In other

words, the source of the information on how to use antibiotics does not play a significant role

in the relationship between receiving information and respondents’ level of self-medication. In

addition, both sources produce an odds ratio higher than 1. From these results, hypothesis 3

‘When information is received from a healthcare professional, levels of self-medication

among those who self-medicate will be lower than when the information is obtained from the

media/internet.’ is rejected. It can also be seen from model 4 that the effect of receiving

information regarding the proper use of antibiotics remains constant (OR: 0.74, 95% CI: 0.56-

0.98), as does the effect of having a chronic illness. In addition, the country variation remains

constant.

33

Table 8.1. Logistic regression analysis results.

Model 1 - Basic Model Model 2 - Hypothesis 1 Model 3 - Hypothesis 2 Model 4 - Hypothesis 3

OR 95% CI OR 95% CI OR 95% CI OR 95% CI

Education (ref Low)

Medium 1.1 (0.8-1.52) 1.08 (0.79-1.49) 1.2 (0.82-1.74) 1.2 (0.82-1.75) High 0.95 (0.68-1.33) 0.94 (0.67-1.32) 1.04 (0.69-1.55) 1.03 (0.69-1.55) Perceived health (ref (very) good)

Fair 1.08 (0.86-1.37) 1.09 (0.86-1.38) 1.27 (0.97-1.67) 1.27 (0.97-1.67) (Very) Bad 1.44 (0.98-2.11) 1.46 (0.99-2.14) 1.54 (0.99-2.42) 1.54 (0.99-2.42) Chronic Illness

1.47* (1.16-1.85) 1.46* (1.15-1.84) 1.36* (1.03-1.79) 1.36* (1.03-1.79)

Age (ref 0-17 yrs)

18-40 1.83 (0.35-9.66) 1.87 (0.36-9.91) 4.2 (0.48-36.89) 4.3 (0.49-37.79) 41-64 1.15 (0.22-6.02) 1.17 (0.22-6.18) 2.52 (0.29-22.05) 2.57 (0.29-22.56) 65+ 0.98 (0.18-5.20) 1.01 (0.19-5.40) 1.97 (0.22-17.5) 2.02 (0.23-18.00) Sex (ref female)

Male 1.16 (0.94-1.43) 1.16 (0.94-1.43) 1.11 (0.87-1.42) 1.11 (0.87-1.43) Received Information

1.25 (0.99-1.58)

Type of Information

Proper Use

0.74* (0.56-0.98) 0.74* (0.56-0.98)

General Information

0.87 (0.66-1.17) 0.87 (0.65-1.16)

Side-Effects

1.3 (0.98-1.70) 1.29 (0.98-1.71) Source of Information

Healthcare Professional

1.07 (0.74-1.56)

Media

1.1 (0.63-1.95) Country

(ref Cyprus) Estonia 0.48* (0.30-0.75) 0.47* (0.30-0.74) 0.45* (0.26-0.72) 0.44* (0.26-0.73) Greece 0.86 (0.61-1.22) 0.88 (0.62-1.25) 0.7 (0.46-1.08) 0.7 (0.46-1.08) Hungary 1.32 (0.93-1.87) 1.31 (0.92-1.86) 1.11 (0.74-1.67) 1.11 (0.74-1.68) Italy 0.89 (0.61-1.31) 0.89 (0.60-1.30) 0.88 (0.56-1.39) 0.88 (0.56-1.39) Romania 1.2 (0.85-1.69) 1.19 (0.84-1.68) 1.01 (0.68-1.51) 1.01 (0.68-1.51) Spain 0.8 (0.56-1.15) 0.81 (0.56-1.16) 0.58* (0.37-0.91) 0.58* (0.37-0.91)

N 1,801 1,789 1,324 1,324

* p<0.05

34

Regression 1.2 Logistic Regression by country

To investigate this picture further, models 2 and 4 were run separately for each country.

Results for this analysis are displayed in Table 9.1. The aim of this is not to investigate each

variable separately per country but to use the country analysis to explain the effects found in

the combined analysis. Hypothesis 1, ‘Of those who self-medicate, the level of self-

medication will be lower among those who have received information than among those who

did not receive information.’ was rejected on the basis that the variable indicating that

respondents received any information on the use of antibiotics did not yield a statistically

significant effect on their level of self-medication. When run separately for each country, it

can be seen that only in Cyprus does receiving any information produce a statistically

significant effect on respondents’ level of self-medication (OR: 2.26, 95% CI: 1.21-4.21). In

Cyprus, patients who received any information on how to use antibiotics are associated with

higher levels of self-medication with antibiotics. Only in Spain is receiving any information

associated with lower levels of self-medication with antibiotics, however this effect is not

significant at the p<0.05 level.

Hypothesis 2, ‘Among patients who self-medicate, the effect of receiving information on ‘side-

effects’ will be most associated with lower levels of self-medication.’ was also rejected.

Information regarding the side-effects of antibiotics was found to produce a borderline

significant result. When run by country, it is found that this is driven by a strong effect in

Cyprus and is associated with higher levels of self-medication (OR: 2.42, 95% CI: 1.13-5.20).

Only in Hungary and Spain was receiving information regarding the side-effects of antibiotics

associated with lower levels of self-medication with antibiotics, however both results are not

significant at the p<0.05 level. The effect of receiving information regarding the proper use of

antibiotics was found to be significantly associated with lower levels of self-medication with

antibiotics in the combined analysis. When run individually for each country, this effect does

not prove to be significant at the p<0.05 level in any country, however in Cyprus it is

borderline significant (OR: 0.53, 95% CI: 0.26-1.06). This shows the value of running the

pooled analysis, rather than only separate country analyses.

Hypothesis 3, ‘When information is received from a healthcare professional, levels of self-

medication among those who self-medicate will be lower than when the information is

obtained from the media/internet.’ was also rejected. Neither source yielded a significant

result in the combined analysis. When run separately for each country, a healthcare-

professional source does not yield a significant result in any country. A media source

35

produces a strongly significant result in one country only - Spain - where it is associated with

higher levels of self-medication (OR: 8.86, 95% CI: 1.93-40.58).

36

CYPRUS** ESTONIA*** GREECE**** HUNGARY***** ITALY****** ROMANIA******* SPAIN********

OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI

Education (ref Low)

Medium 3.91* (1.3-11.7) 0.62 (0.24-1.58) 0.7 (0.16-3.04) 1.37 (0.54-3.5) 1.08 (0.46-2.53) 0.92 (0.34-2.48) 0.43 (0.14-1.39) High 2.15 (0.68-6.76) 1 (omitted) 0.73 (0.17-3.18) 1.19 (0.44-3.2) 1.04 (0.38-2.86) 0.79 (0.27-2.3) 0.19* (0.05-0.77) Perceived health (ref (very) good)

Fair 1.24 (0.54-2.87) 1.58 (0.51-4.90) 1.76 (0.72-4.32) 0.94 (0.48-1.86) 2.49 (0.88-7.03) 1.03 (0.53-2.04) 0.94 (0.33-2.66)

(Very) bad 0.24 (0.05-1.09) 2.3 (0.49-10.86) 8.95* (1.91-42.01) 2.2 (0.81-5.99) 3.19 (0.71-14.23) 1.58 (0.47-5.26)

Chronic Illness

1.34 (0.67-2.69) 1.27 (0.42-3.82) 1.29 (0.58-2.84) 1.32 (0.66-2.62) 1.32 (0.57-3.08) 0.77 (0.33-1.81) 2.46* (1.02-5.98)

Age (ref 0-17 yrs)

18-40 - - - - 4.3* (1.34-13.78) 1.06 (0.4-2.82) - - 2.03 (0.68-6.05) - -

41-64 0.37* (0.18-0.73) 1.6 (0.50-5.11) 1.56 (0.54-4.53) 1.07 (0.48-2.39) 0.45 (0.16-1.26) 1.71 (0.64-4.57) 0.33* (0.14-0.78) 65+ 0.82 (0.30-2.22) 0.74 (0.14-3.83) 1 (omitted) 1 (omitted) 0.25* (0.07-0.82) 1 (omitted) 0.34 (0.10-1.12)

Sex (ref female)

Male 0.79 (0.43-1.45) 0.99 (0.37-2.63) 2.88* (1.32-6.26) 1.003 (0.53-1.89) 0.79 (0.33-1.87) 1.38 (0.76-2.51) 1.41 (0.67-3.01)

Received information 2.26* (1.21-4.21) 2.36 (0.73-7.63) 1.09 (0.63-1.90) 1.22 (0.58-2.59) 1.78 (0.98-3.27) 1.04 (0.56-1.95) 0.59 (0.31-1.12)

Type of information

Proper use 0.53 (0.26-1.06) 0.65 (0.19-2.18) 0.8 (0.36-1.78) 1.17 (0.56-2.45) 0.51 (0.20-1.26) 0.78 (0.36-1.69) 0.94 (0.31-2.88)

General information 0.41 (0.15-1.1) 0.7 (0.21-2.36) 1.48 (0.68-3.2) 1.48 (0.77-2.84) 1.6 (0.66-3.86) 0.63 (0.28-1.39) 0.48 (0.14-1.67)

Side effects 2.42* (1.13-5.20) 1.83 (0.56-5.94) 1.03 (0.48-2.19) 0.998 (0.52-1.93) 1.34 (0.6-3.02) 1.66 (0.75-3.66) 0.83 (0.26-2.68) Source of information

Healthcare professional

0.9 (0.29-2.78) 1.54 (0.49-4.82) 0.71 (0.17-3.01) 0.6 (0.29-1.24) 1.67 (0.28-9.75) 1.36 (0.44-4.18) 3.3 (0.73-14.87) Media 1 (omitted) 2.04 (0.43-9.74) 0.19 (0.02-2.21) 1.19 (0.42-3.4) 1 (omitted) 0.28 (0.05-1.49) 8.86* (1.93-40.58)

N 232 123 176 212 173 208 176

* p<0.05 ******* Age dropped (predicts success perfectly) ******** Perceived health !=0 predicts success perfectly. Perceived health dropped.

** media != 0 predicts success perfectly. Media dropped *** Education dropped (predicts success perfectly)

**** Age dropped (predicts success perfectly)

***** Age !=0 predicts failure perfectly. Age dropped

****** Media !=0 predicts failure perfectly. Media dropped

Table 9.1. Logistic regression Models 2 and 4 run by country results.

37

Regression 2. Poisson Regression

Following the logistic regression, a Poisson regression was conducted as a robustness test.

A Poisson regression is chosen due to the distribution of the data. Poisson regressions are

suitable for modelling count data. Those who indicated 1 course of antibiotics in the past 18

months are over-represented compared to those who indicated 2 - 3 courses and again

compared to those who indicated 4 or more courses.

Building the models for the Poisson regression mirrors that of the logistic regression. Results

from the Poisson regression show that the logistic regression is robust and the conclusions

remain the same (Table 10.2, Appendix).

38

6. Discussion and Conclusions

This study aimed to answer the research question: ‘How do differences in information

provision by healthcare professionals or media outlets to patients who self-medicate with

antibiotics affect this level of self-medication?’. Data was used from the ‘Antimicrobial

Resistance and Causes of Non-Prudent Use of Antibiotics in Human Medicine’ research

project conducted by NIVEL. Data was collected in 7 European countries (Cyprus, Estonia,

Greece, Hungary, Italy, Romania and Spain) which were identified by the Eurobarometer

(2013) as countries with high levels of self-medication with antibiotics. This data was deemed

appropriate for this study as the selected countries were also found by the Eurobarometer to

have low levels of knowledge regarding antibiotics. Results from the Eurobarometer (2013)

found an association between receiving information on the use of antibiotics and correct

knowledge of antibiotics. The present research aimed to answer whether receiving

information on antibiotics is also associated with the behaviour of an individual who self-

medicates with antibiotics. In addition, it aimed to investigate whether the specific type of

information and source of information produce a stronger association with this behaviour. In

order to answer the research question two hypotheses and one explanatory hypothesis were

tested. Before discussing the results, a limit of the research is discussed, to provide clarity to

the conclusions drawn.

A limitation of the present research is the nature of the dependent variable. 1 course of self-

medicated antibiotics in the past 18 months is the lowest level respondents could indicate.

Therefore there is no reference to respondents who used antibiotics but did not self-medicate

with them. This sample bias means that the results are not generalizable to whole

populations but can only be applied to a subgroup of antibiotic users, although it was found

that self-medicators are not selective on the basis of sex and age. Strengths of the study

include a large sample size, which was well distributed geographically. In addition, the

method of data collection – computer assisted telephone interviews - has been found to be

optimal in terms of quality and reliability of responses, as opposed to self-administered

surveys (Bowling, 2005).

Hypothesis 1; ‘Of those who self-medicate, the level of self-medication will be lower among

those who have received information than among those who did not receive information’ was

based on rational choice theory, exemplified through the theory of reasoned action (Ajzen &

Fishbein, 1980) and the theory of planned behaviour (Ajzen, 1985). Based on the results of

the logistic regression, this hypothesis was rejected. A possible explanation for this surprising

result is that respondents may have received information but had forgotten it. Arguably,

39

information is better remembered when it has influenced one’s behaviour. If the information

did not have an impact on the respondent’s behaviour, then it is likely that they forgot the

information or forgot receiving information in general. An additional, more concrete

explanation is a statistical one. Only data on respondents who had previously self-medicated

with antibiotics was collected. Therefore, due to the nature of the dependent variable (level of

self-medication) what is being tested in the analysis is not a causal reduction effect but rather

an association. With this in mind, one cannot entirely reject the theory behind the hypothesis

before an analysis is carried out with data on those who did not self-medicate with antibiotics.

Hypothesis 2; ‘Among patients who self-medicate, the effect of receiving information on ‘side-

effects’ will be most associated with lower levels of self-medication.’ was based on Protection

Motivation Theory (Rogers, 1975) and the usefulness of negativity in conveying information

(Geer, 2006). From this, it was expected that negative information would have the strongest

association with lower levels of self-medication. This is because when the severity of an

issue is highlighted, people behave in a rational, protective manner and because negative

information is more often validated by facts than positive information it is therefore more

motivating to change behaviour. Based on the results of the regression analysis, this

hypothesis was also rejected. This is a particularly surprising result considering the

operationalization of the variable side-effects. Information about when antibiotics should not

be taken, allergies, side-effects and drug interaction would theoretically evoke a rational,

behaviour changing thought in the patient. Again the issue of the nature of the dependent

variable should be considered. What was analysed was the respondents’ intensity of self-

medication. Perhaps the results would be different if data on those who did not self-medicate

was included. We cannot completely reject the theory behind the hypothesis because

although evidence for it was not found in this study, perhaps evidence would be found if the

sample was representative.

Although information on the side-effects of antibiotics did not bear an association with

respondents’ level of self-medication, obtaining information on the proper use of antibiotics

was found to be associated with less self-medication. A possible explanation for this can be

given following the arguments of Geer (2006). Drawing on the results of previous studies

concerning methods to inform the public on topics which the common person is not an

expert, Geer argues that simple messages are the most effective. The public are not

interested in spending time learning intricate details but are more receptive to messages

which are straightforward and uncomplicated. Considering the operationalization of the three

types of information in this research, it could be argued that information on the proper use of

antibiotics (complete the full course and only take antibiotics over the prescribed period and

40

at the correct dose) is the most explicit and easiest to understand. Therefore respondents

are more receptive to this type of information and it has a stronger association with their level

of self-medication with antibiotics. In addition, the non-existent effect of obtaining general

information could be explained through the operationalization of this variable. No knowledge

of how to act if the condition persists is unlikely to encourage a patient to self-medicate

further.

The (exploratory) hypothesis 3 ‘When information is received from a healthcare professional,

levels of self-medication among those who self-medicate will be lower than when the

information is obtained from the media/internet.’ was also rejected as neither source yielded

a significant relationship with the respondents’ level of self-medication with antibiotics.

Although a disappointing result, it could be an indication that governments need not invest

time and energy in public media campaigns, and instead, efforts should be focused on the

supply side of self-medication with antibiotics. That is, interventions should be directed

towards pharmacists who allow over-the-counter selling of antibiotics which one would

normally need a medical prescription to obtain. Or perhaps thorough discussion with patients

at pharmacies and in GP offices should be reinforced. The effect of having a chronic illness

or poorer health suggests that this method would be feasible in reducing self-medication as

these persons would have more frequent and regular contact with healthcare professionals

than people with better health.

A possible explanation for the contradictory results found in this study compared to those

found in the Eurobarometer (2013) is that this study assessed the factors that explain some

level of self-medication with antibiotics (i.e. 1 course of antibiotics is the base), while the

Eurobarometer assessed the factors that explain any level of self-medication (i.e. no self-

medication is possible). Consequently, the findings are not directly comparable.The

Eurobarometer also only included respondents who used antibiotics in the past 12 months

while this study included respondents who used antibiotics in the past 18 months. In addition,

the Eurobarometer (2013) study assessed 27 EU countries, while the present study only

focused on 7 EU countries.

Antibiotic resistance continues to be an issue of immediate concern and it is fundamental

that the public is made aware of the basic facts of antibiotics and their use. Therefore

education about antibiotics, specifically regarding the proper use of antibiotics, through

school curricula could be a possible venture for the countries in this research. However

further research is also warranted into the reasons for self-medication with antibiotics.

41

Because this thesis has shown the association between self-medication and information to

be weak at most, other explanations should be sought. Logically, there could be economic,

and cultural factors involved. For instance, if future research finds that people primarily self-

medicate with antibiotics for cost-saving reasons (i.e. to avoid paying for an appointment with

a GP) then public educational interventions would not be the most effective. In this case,

pharmacist based interventions could be more appropriate.

42

7. Bibliography Abell, Peter. (2000) ‘Sociological Theory and Rational Choice Theory,’ in Turner, B. S. The

Blackwell Companion to Social Theory, second edition, Blackwell Publishers, Malden, Massachusetts.

Adriaenssens, N., Coenen, S., Versporten, A., Muller, A., Minalu, G., Faes, C., and Goossens, H. (2011). ‘European Surveillance of Antimicrobial Consumption (ESAC): outpatient antibiotic use in Europe (1997–2009).’ Journal of Antimicrobial Chemotherapy, 66(6): 3-12.

Agency for Healthcare Research and Quality (AHRQ). (2006) ‘Closing the quality gap: a critical analysis of quality improvement strategies.’ Technical Review 9, Publication 04(06)-0051-4. Maryland. Agency for Healthcare Research and Quality.

Ajzen, I. (2006). ‘Constructing a Theory of Planned Behaviour Questionnaire.’ Retrieved October 2015 from the World Wide Web: http://www.people.umass.edu/aizen/pdf/tpb.measurement.pdf.

Ajzen, I., and Fishbein, M. (1980) Understanding attitudes and predicting social behaviour. (NJ): Englewood Cliffs: Prentice-Hall.

Ajzen, I. (1985) ‘From intentions to actions: a theory of planned behaviour’. In: Kuhl, J., and Beckman, J. (eds). Action control: from cognition to behaviour. Heidelberg: Springer. p. 11-39.

Albrich, W. C., Monnet, D. L., and Harbarth, S. (2004) ‘Antibiotic selection pressure and resistance in Streptococcus pneumoniae and Streptococcus pyogenes.’ Emerging Infectious Diseases; 10: 514−7.

Awad, A. I., and Aboud, E. A. (2015) ‘Knowledge, Attitude and Practice towards Antibiotic Use among the Public in Kuwait.’ PLoS ONE, 10(2), e0117910. Source: http://doi.org/10.1371/journal.pone.0117910. [Accessed 1 March 2015].

Bartiaux, F. (2008) ‘Does environmental information overcome practice compartmentalisation and change consumers’ behaviours?’ Journal of Cleaner Production; 16 (11): 1170–80.

Bandura, A. A Social Learning Theory. Englewood Cliffs, N.J.: Prentice-Hall. Berzanskyte, A., Valinteliene, R., Haaijer-Ruskamp, F. M., Gurevicius, R., and Grigoryan, L.

(2006) ‘Self-medication with antibiotics in Lithuania.’ International Journal of Occupational Medicine and Environmental Health; 19: 246–253.

Bowling, A. (2005) ‘Mode of questionnaire administration can have serious effects on data quality’. Journal of Public Health. 27(3): 281-291.

Denis, C., McMorrow, K. and Roger, W. (2006) Globalisation: trends, issues and macro implications for the EU. Directorate General Economic and Financial Affairs, European Commission. European Economy – Economic Papers, 254.

Donkor, E. S., Tetteh-Quarcoo, P. B., Nartey, P., and Agyeman, I. O. (2012) ‘Self-Medication Practices with Antibiotics among Tertiary Level Students in Accra, Ghana: A Cross-Sectional Study.’ International Journal of Environmental Research and Public Health, 9(10): 3519–29.

ECDC and EMA (2009) The bacterial challenge: time to react. Joint Technical Report. Stockholm.

Eurobarometer (2013) Antimicrobial Resistance. Special Eurobarometer 407. Source: http://ec.europa.eu/health/antimicrobial_resistance/docs/ebs_407_en.pdf. [Accessed: 15 January 2015].

European Centre for Disease Prevention and Control, “European Antibiotic Awareness Day”, Source: http://ecdc.europa.eu/en/eaad/Pages/Home.aspx. [Accessed: 25 November 2015].

European Commission (2011) Communication from the Commission to the European Parliament and the Council, Action plan against the rising threats from Antimicrobial Resistance, COM 748, 15 November 2011 (http://ec.europa.eu/dgs/health_consumer/docs/communication_amr_2011_748_en.pdf). [Accessed: 3 November 2015].

http://doi.org/10.1371/journal.pone.0117910

http://ec.europa.eu/health/antimicrobial_resistance/docs/ebs_407_en.pdf

http://ecdc.europa.eu/en/eaad/Pages/Home.aspx

http://ec.europa.eu/dgs/health_consumer/docs/communication_amr_2011_748_en.pdf

http://ec.europa.eu/dgs/health_consumer/docs/communication_amr_2011_748_en.pdf

43

Eysenbach, G. (2008) ‘Medicine 2.0: Social Network, Collaboration, Participation, Apomediation, and Openness’. Journal of Medical Internet Research; 10(2) e22. doi: 10.2196/jmir.1030.

Geer, J. G. (2006) In Defense of Negativity: attack ads in presidential campaigns. Chicago, University of Chicago Press.

Genni, M., Newnham, W., Burns, I., Snyder, R. D. Dowling, A. J., Ranieri, N. F., Gray, E. L., and Lachlan, S. A. (2006) ‘Information from the Internet: Attitudes of Australian Oncology Patients’. Internal Medicine Journal; 36: 718–23.

Goossens, H., Ferech, M., Vander Stichele, R. and Elseviers, M. ESAC Project Group. (2005) ‘Outpatient antibiotic use in Europe and association with resistance: a cross-national database study.’ Lancet; 365: 579−87.

Grigoryan, L., Burgerhof, J. G., Degener, J. E., Deschepper, R., Lundborg, C. S., Monnet, D. L., and Haaijer‐Ruskamp, F. M. (2007) ‘Attitudes, beliefs and knowledge concerning

antibiotic use and self‐medication: a comparative European study.’ Pharmacoepidemiology and drug safety, 16(11): 1234-43.

Grigoryan, L., Burgerhof, J. G. M., Haaijer-Ruskamp, F. M., Degener, J. E., Deschepper, R., Monnet, D. L., Di Matteo, A., Scicluna, E. E., Bara, A. C., Lundborg, C. S., and Birkin, J. (2007) ‘Is self-medication with antibiotics in Europe driven by prescribed use?’ Journal of Antimicrobial Chemotherapy; 59: 152-6.

Hechter, M. and Kanazawa, S. (1997) ‘Sociological Rational Choice Theory’. Annual Review of Sociology 23: 191-214.

Hesse, B. W., Nelson, D. E., Kreps, G. L., Croyle, R. T., Arora, N. K., Rimer, B. K., et al. (2005) ‘Trust and sources of health information: The impact of the internet and its implications for health care providers: Findings from the first Health Information National Trends Survey’. Archives of Internal Medicine; 165: 2618–24.

Hosmer, D. W , and Lemeshow, S. (1989) ‘Applied Logistic Regression’. Statistics in Medicine; 10(7): 1162-3.

Hulka, B. S., Cassel, J. C., Kupper, L. L., and Burdette, J. A. (1976). ‘Communication, compliance, and concordance between physicians and patients with prescribed medications’. American Journal of Public Health, 66(9): 847–53.

Jenkins, L., Britten, N., Stevenson, F., Barber, N., and Bradley, C. (2003) ‘Developing and using quantitative instruments for measuring doctor-patient communication about drugs. Patient Education and Counselling; 50(3): 273-8.

Kunin, C. M. (1978) ‘Problems of antibiotic usage: Definitions, causes and proposed solutions.’ Annals of Internal Medicine; 89: 802–805.

Lazarus, R. S. (1966). Psychological Stress and the Coping Process. New York: McGraw-Hill.

Leventhal, H. (1970). ‘Findings and Theory in the Study of Fear Communications’. In: L. Berkowitz, (ed). Advances in Experimental Social Psychology. 5: 111-86.

Levy, S. B. (2005). ‘Antibiotic resistance—the problem intensifies.’ Advanced Drug Delivery Reviews, 57(10), 1446-1450.

Llor, C. and Cots, J. M. (2009) ‘The sale of antibiotics without prescription in pharmacies in Catalonia, Spain’. Clinical Infectious Diseases 48(10): 1345-9.

Lopez-Vazquez, P., Vazquez-Lago, J. M. and Figueiras, A. (2012) ‘Misprescription of antibiotics in primary care: a critical systematic review of its determinants. Journal of Evaluation in Clinical Practice. 18(2): 473-84.

Marković-Peković, V. and Grubiša, N. (2012) ‘Self-medication with antibiotics in the Republic of Srpska community pharmacies: pharmacy staff behavior.’ Pharmacoepidemiology and Drug Safety; 21: 1130–33.

Mayer, D. K., Terrin, N. C., Kreps, G. L., Menon, U., McCance, K., Parsons, S. K., et al. ‘Cancer survivors information seeking behaviors: A comparison of survivors who do and do not seek information about cancer’. Patient Education and Counseling; 65: 342–50.

http://dx.doi.org/10.2196%2Fjmir.1030

44

McNulty, C. A., Boyle, P., Nichols, T., Clappison, P., and Davey, P. (2007) ‘Don’t wear me out – the adult’s knowledge of and attitudes to antibiotic use’. Journal of Antimicrobial Chemotherapy; 59: 727–38.

Narhi, U. (2007) ‘Sources of medicine information and their reliability evaluated by medicine users’. Pharmacy World and Science; 29: 688–94.

Pavyde E., Veikutis V., Maciuliene A., Maciulis V., Petrikonis K., and Stankevicius E. (2015) ‘Public Knowledge, Beliefs and Behavior on Antibiotic Use and Self-Medication in Lithuania’. International Journal of Environmental Research and Public Health; 12(6): 7002-16.

Pinder, R. J., Berry, D., Sallis, A., Chadborn T., et al. (2015) Antibiotic prescribing and behaviour change in healthcare settings: literature review and behavioural analysis. London, UK, Publisher: Department of Health & Public Health England, 2014719.

Rutten, L. J. F., Arora, N. K., Bakos, A. D., Aziz, N., and Rowland, J. (2005) ‘Information needs and sources of information among cancer patients: A systematic review of research (1980–2003)’. Patient Education and Counseling; 57: 250–61.

Rogers, R. W. (1975) ‘A protection motivation theory of fear appeals and attittude change’. Journal of Psychology; 91: 93-114.

Stevenson, F. A., Greenfield, S. M., Jones, M., Nayak, A., and Bradley, C.P.(1999) ‘GPs perceptions of patient influence on prescribing’. Family Practice; 16: 255–61.

Väänänen, M. H., Pietilä, K., and Airaksinen, M. (2006) ‘Self-medication with antibiotics–does it really happen in Europe?’ Health Policy; 77: 166-71.

World Health Organisation (2014) Antimicrobial Resistance. Fact Sheet. Source: http://www.who.int/mediacentre/factsheets/fs194/en/. [Accessed: 1 March 2015].

World Health Organisation (2011) ‘Regulate and promote rational use of medicines, including in animal husbandry, and ensure proper patient care’. World Health Day 2011: policy briefs. Source: http://www.who.int/world-health-day/2011/presskit/whd2011_fs4_animal.pdf?ua=1. [Accessed: 4 September 2015].

http://www.who.int/mediacentre/factsheets/fs194/en/

http://www.who.int/mediacentre/factsheets/fs194/en/

http://www.who.int/world-health-day/2011/presskit/whd2011_fs4_animal.pdf?ua=1

http://www.who.int/world-health-day/2011/presskit/whd2011_fs4_animal.pdf?ua=1

45

8. Appendix 1 Descriptives tables Table 1.2. Descriptives (Total sample)

Variables Min Max % Valid cases

Level of self-medication (D.V.) 0 1

1,841

1 course

65%

2 + courses

33%

Don't know / refuse

2%

Received information 0 1

1,846

Yes

73%

No

26%

Don't know / refuse

1%

Type of information

1,353

Proper use 0 1 79%

General information 0 1 38%

Side-effects 0 1 41%

Don't know / refuse 0 1 4%

Source of information

1,353

Healthcare professional 0 1 88%

Media 0 1 5%

Don't know / refuse 0 1 0.2%

Age

15 99

1,846

< 17

0.4%

18-40

29%

41-64

50%

65+

20%

Refuse

0.5%

Sex

0 1

1,846

Female

64%

Male

36%

Education 1 3

1,846

Low

14%

Medium

50%

High

36%

Don't know / refuse

0.8%

Chronic Illness 0 1

1,846

Yes

35%

No

65%

Don't know / refuse

0.9%

Perceived health 1 3

1,846

(Very) good

58%

Fair

33%

(Very) bad

9%

Don't know / refuse 1%

46

Table 2.2. Descriptives (Cyprus) Variables

Min Max % Valid cases


315

1 course

65%

2 + courses

34%

Don’t know / refuse

1%

Received information 0 1

315

Yes

75%

No

24%

Don't know / refuse

0.6%

Type of information

237

Proper use 0 1 70%



Don't know / refuse 0 1 7%

Source of information

237


Media 0 1 1%

Don't know / refuse 0 1 -

Age

18 85

315

< 17

-

18-40

33%

41-64

49%

65+

18%

Refuse

0.3%

Sex

0 1

315

Female

55%

Male

45%

Education 1 3

315

Low

14%

Medium

45%

High

41%

Don't know / refuse

-

Chronic Illness 0 1

315

Yes

37%

No

63%

Don't know / refuse

0.3%


315

(Very) good

74%

Fair

19%

(Very) bad

7%

Don't know / refuse -

47

Table 3.2. Descriptives (Estonia)



164

1 course

77%

2 + courses

21%

Don't know / refuse

2% Received information 0 1

164

Yes

80%

No

19%

Don't know / refuse

1% Type of information

131

Proper use 0 1 81%



Don't know / refuse 0 1 6% Source of information

131


Media 0 1 7%


Age

18 80

164

< 17

18-40

23%

41-64

57%

65+

20%

Sex

0 1

164

Female

60%

Male

40% Education 1 3

164

Low

5%

Medium

55%

High

40%

Don't know / refuse

- Chronic Illness 0 1

164

Yes

41%

No

59%

Don't know / refuse

-


164

(Very) good

42%

Fair

46%

(Very) bad

12%


48

Table 4.2. Descriptives (Greece)



295

1 course

71%

2 + courses

29%

Don't know / refuse

- Received information 0 1

296

Yes

61%

No

39%

Don't know / refuse

0.3% Type of information

181

Proper use 0 1 73%



Don't know / refuse

- Source of information

181


Media 0 1 4%

Don't know / refuse

-

Age

16 87

296

< 17

0.7%

18-40

25%

41-64

54%

65+

19%

Refuse

1% Sex

0 1

296

Female

71%

Male

29% Education 1 3

296

Low

10%

Medium

44%

High

46%

Don't know / refuse

- Chronic Illness 0 1

296

Yes

35%

No

65%

Don't know / refuse

0.3%


296

(Very) good

72%

Fair

24%

(Very) bad

5%

Don't know / refuse 0.7%

49

Table 5.2. Descriptives (Hungary)



261

1 course

56%

2 + courses

42%

Don't know / refuse


261

Yes

84%

No

14.5%

Don't know / refuse


219

Proper use 0 1 79%



Don't know / refuse 0 1 4% Source of information

219


Media 0 1 8%


Age

17 90

261

< 17

0.4%

18-40

29%

41-64

51%

65+

19%

Refuse

0.8% Sex

0 1

261

Female

70%

Male

30% Education 1 3

261

Low

14%

Medium

56%

High

30%

Don't know / refuse

0.4% Chronic Illness 0 1

261

Yes

48%

No

51%

Don't know / refuse

1%


261

(Very) good

54%

Fair

31%

(Very) bad

14%


50

Table 6.2. Descriptives (Italy)



279

1 course

66%

2 + courses

30%

Don't know / refuse


279

Yes

65%

No

34%

Don't know / refuse

1% Type of information

182

Proper use 0 1 82%



Don't know / refuse

3% Source of information

182


Media 0 1 0.6%

Don't know / refuse

0.6%

Age

18 99

279

< 17

18-40

14%

41-64

53%

65+

33%

Refuse

0.4% Sex

0 1

279

Female

73%

Male

27% Education 1 3

279

Low

28%

Medium

48%

High

20%

Don't know / refuse

4% Chronic Illness 0 1

279

Yes

30%

No

67%

Don't know / refuse

3%


279

(Very) good

17%

Fair

68%

(Very) bad

11%

Don't know / refuse 4%

51

Table 7.2. Descriptives (Romania)



271

1 course

58%

2 + courses

39%

Don't know / refuse


275

Yes

79%

No

20%

Don't know / refuse


218

Proper use 0 1 81%



Don't know / refuse


218


Media 0 1 6%

Don't know / refuse

0.5%

Age

15 81

275

< 17

1%

18-40

44%

41-64

39%

65+

16%

Refuse

0.7% Sex

0 1

275

Female

59%

Male

41% Education 1 3

275

Low

13%

Medium

48%

High

39%

Don't know / refuse


275

Yes

28%

No

72%

Don't know / refuse

0.7%


275

(Very) good

60%

Fair

31%

(Very) bad

9%

Don't know / refuse 0.4%

52

Table 8.2. Descriptives (Spain)



256

1 course

70%

2 + courses

29%



256

Yes

72%

No

27%



185

Proper use 0 1 87%





185


Media 0 1 6%


-

Age

16 99

256

< 17

0.4%

18-40

34%

41-64

52%

65+

13%

Refuse

0.4% Sex

0 1

256

Female

57%

Male

43% Education 1 3

256

Low

11%

Medium

55%

High

33%



256

Yes

26%

No

74%


0.8%


256

(Very) good

80%

Fair

17%

(Very) bad

3%

Don’t know / refuse 0.4%

53

Regression Tables Table 9.2. Multilevel logistic regression (Preliminary analysis)

Model 1 Model 2 Model 3 Model 4

Variables Coef. P-Value Coef. P-Value Coef. P-Value Coef. P-Value

Area -0.01 0.489 -0.01 0.609 -0.01 0.511 -0.01 0.494

Education (ref. Low) -0.03 0.153 -0.03 0.164 -0.02 0.411 -0.02 0.422

Employed -0.04 0.212 -0.04 0.182 -0.03 0.459 -0.03 0.451

Health insurance 0.02 0.741 0.01 0.759 -0.04 0.446 -0.04 0.446 Smoker 0.04 0.181 0.05 0.168 0.06 0.155 0.05 0.157

Perceived health (ref. (very) good) 0.06 0.018 0.07 0.011 0.1 0.002 0.1 0.002

Chronic Illness 0.08 0.024 0.08 0.031 0.05 0.199 0.05 0.206

Age (ref. 0-17 yrs) -0.08 0.001 -0.08 0.001 -0.09 0.002 -0.09 0.002

Sex (ref. female) 0.07 0.019 0.08 0.015 0.05 0.137 0.05 0.136

Information

0.06 0.083 Proper use

-0.12 0.004 -0.12 0.004

General information

-0.05 0.224 -0.05 0.227

Side effects

0.05 0.192 0.06 0.18 Healthcare professional

0.001 0.982

Media

-0.04 0.636

N 1,630 1,621 1,198 1,198

Table 10.2. Poisson regression (robustness test)

Model 1 Model 2 Model 3 Model 4

Variables Coef. P-Value Coef. P-Value Coef. P-Value Coef. P-Value

Education (ref. Low) -0.02 0.483 -0.02 0.494 -0.01 0.711 -0.02 0.731 Perceived health (ref. (very) good) 0.05 0.196 0.05 0.174 0.07 0.086 0.07 0.086

Chronic Illness 0.06 0.169 0.06 0.185 0.05 0.361 0.05 0.366

Age (ref. 0-17 yrs) -0.06 0.062 -0.06 0.069 -0.06 0.077 -0.06 0.078

Sex (ref. female) 0.04 0.354 0.04 0.338 0.03 0.566 0.03 0.541

Information

0.05 0.297 Proper use

-0.08 0.158 -0.08 0.147

General information

-0.03 0.639 -0.03 0.622

Side effects

0.04 0.503 0.04 0.481 Healthcare professional

0.03 0.711

Media

0.01 0.959

N 1,769 1,758 1,299 1,299

54

9. Appendix 2. Syntax *Countries: *1=Cyprus *2=Estonia *3=Greece *4=Hungary *5=Italy *6=Romania *7=Spain sort country destring, replace *DROP RESPONDENTS WHO USED AB WITH PRESCRIPTION + CHILDREN drop if v5last==. drop if v3course_2==1 | v3course_2==2 | v3course_2==3 | v3course_2==4 drop if v3course_3==1 | v3course_3==2 | v3course_3==3 | v3course_3==4 **BACKGROUND** *AREA* recode v25area (4=.a) tab v25area summarize v25area *EDUCATION* recode v26edu (4=.a) summarize v26edu tab v26edu *JOB* recode v27job (10=.a) recode v27job (9=.a) tab v27job gen employed=0 replace employed = 1 if v27job==1 | v27job==2 | v27job==3 replace employed = . if v27job==. replace employed = .a if v27job==.a tab employed label define employedlabel 0"no" 1"yes" label values employed employedlabel codebook employed *HEALTH INSURANCE* *Yes, covered by national health insurance* tab v28ins_1 *Yes, covered by private health insurance*

tab v28ins_2 *Yes, covered by community/social healthcare insurance* tab v28ins_3 *Other* tab v28ins_4 *Not insured* tab v28ins_5 *Don't know/Refuse* tab v28ins_6 gen healthinsurance=0 replace healthinsurance = 1 if v28ins_1==1 | v28ins_2==1 | v28ins_3==1 | v28ins_4==1 replace healthinsurance = . if v28ins_1==. tab healthinsurance codebook healthinsurance *SMOKE?** tab v30smoke recode v30smoke (3=.a) gen smoke=0 replace smoke = 1 if v30smoke==1 replace smoke = . if v30smoke==. replace smoke = .a if v30smoke==.a label define smokelabels 0"no" 1"yes" label values smoke smokelabel codebook smoke **PERCEIVED HEALTH?** recode v31health (6=.a) codebook v31health gen Phealth=0 replace Phealth = 1 if v31health==1 | v31health==2 replace Phealth = 2 if v31health==3 replace Phealth = 3 if v31health==4 | v31health==5 replace Phealth = . if v31health==. replace Phealth = .a if v31health==.a label define Phealthlabel 1"(very)good" 2"fair" 3"(very) bad" label values Phealth Phealthlabel codebook Phealth summarize Phealth tab Phealth **CHRONIC DISEASE?** *Asthma* tab v32chro_1

55

codebook v32chro_1 *COPD* tab v32chro_2 *Emphysema* tab v32chro_3 *Diabetes Mellitus* tab v32chro_4 *Cardiovascular disease* tab v32chro_5 *Hypertension* tab v32chro_6 *Other* tab v32chro_7 *No* tab v32chro_8 *Don't know/Refuse* tab v32chro_9 codebook v32chro_9 gen chronic=0 replace chronic = 1 if v32chro_1==1 | v32chro_2==1 | v32chro_3==1 | v32chro_4==1 | v32chro_5==1 | v32chro_6==1 | v32chro_7==1 replace chronic = . if v32chro_1==. replace chronic = .a if v32chro_9==1 tab chronic *AGE* tab v33age_1 *remove 999 years answer* replace v33age_1=. if v33age_1==999 replace v33age_1=. if v33age_1==0 codebook v33age_1 recode v33age_1 (min/17=1)(18/40=2)(41/64=3)(65/max=4) tab v33age_1 codebook v33age_1 *SEX* tab v34sexe_1 recode v34sexe_1 (3=.a) codebook v34sexe_1 gen sex=0 replace sex=0 if v34sexe_1==2 replace sex=1 if v34sexe_1==1 replace sex = .a if v34sexe_1==.a label define sexlabel 0"female" 1"male" label values sex sexlabel tab sex **HOW MANY COURSES W/OMP?** recode v3course_1 (4=.a)

tab v3course_1 by country: tab v3course_1 gen course=0 replace course=1 if v3course_1==2 | v3course_1==3 replace course=0 if v3course_1==1 replace course=. if v3course_1==. label define courselabel 1"2+courses" 0"1 course" label values course courselabel tab course summarize course **ADVICE ABOUT HOW TO USE THE ANTIBIOTIC?** recode v17advice(3=.a) tab v17advice codebook v17advice gen information=0 replace information=0 if v17advice==2 replace information=1 if v17advice==1 replace information=. if v17advice==. replace information=.a if v17advice==.a label define informationlabel 0"no" 1"yes" label values information informationlabel tab information codebook information **WHAT INFORMATION DID YOU OBTAIN?** gen properuse=0 replace properuse = 1 if v19info_1==1 | v19info_2==1 replace properuse = . if v19info_1==. | v19info_2==. tab properuse codebook properuse gen generalinfo=0 replace generalinfo = 1 if v19info_3==1 | v19info_4==1 | v19info_5==1 replace generalinfo = . if v19info_3==. | v19info_4==. | v19info_5==. tab generalinfo gen sideeffects=0 replace sideeffects = 1 if v19info_6==1 | v19info_7==1 | v19info_8==1 | v19info_9==1 replace sideeffects = . if v19info_6==. | v19info_7==. | v19info_8==. | v19info_9==. tab sideeffects

56

**FROM WHO OR WHERE?** gen HCProfesh=0 replace HCProfesh = 1 if v18where_1==1 | v18where_2==1 | v18where_3==1 replace HCProfesh = . if v18where_1==. tab HCProfesh gen media=0 replace media = 1 if v18where_4==1 | v18where_5==1 | v18where_6==1 replace media = . if v18where_4==. tab media codebook media *ANALYSIS 1ST ATTEMPT* *Multivariate logistic regression* *Model 1 - control variables manova v3course_1 = v25area v26edu employed healthinsurance smoke Phealth chronic v33age_1 sex mvreg *Model 2 - received information manova v3course_1 = v25area v26edu employed healthinsurance smoke Phealth chronic v33age_1 sex information mvreg *Model 3 - Replace received info with type of info manova v3course_1 = v25area v26edu employed healthinsurance smoke Phealth chronic v33age_1 sex properuse generalinfo sideeffects mvreg *Model 4 - Add source of info manova v3course_1 = v25area v26edu employed healthinsurance smoke Phealth chronic v33age_1 sex properuse generalinfo sideeffects HCProfesh media mvreg *ANALYSIS 2ND ATTEMPT* *Multi level mixed-effects linear regression *Model 1 mixed v3course_1 c.v25area c.v26edu c.employed c.healthinsurance c.smoke c.Phealth c.chronic c.v33age_1 c.sex || country: *Model 2 mixed v3course_1 c.v25area c.v26edu c.employed c.healthinsurance c.smoke

c.Phealth c.chronic c.v33age_1 c.sex information || country: *Model 3 mixed v3course_1 c.v25area c.v26edu c.employed c.healthinsurance c.smoke c.Phealth c.chronic c.v33age_1 c.sex properuse generalinfo sideeffects || country: *Model 4 mixed v3course_1 c.v25area c.v26edu c.employed c.healthinsurance c.smoke c.Phealth c.chronic c.v33age_1 c.sex properuse generalinfo sideeffects HCProfesh media || country: *Poisson regression *Model 1 mepoisson v3course_1 c.v26edu c.Phealth c.chronic c.v33age_1 c.sex || country: *Model 2 mepoisson v3course_1 c.v26edu c.Phealth c.chronic c.v33age_1 c.sex information || country: *Model 3 mepoisson v3course_1 c.v26edu c.Phealth c.chronic c.v33age_1 c.sex properuse generalinfo sideeffects || country: *Model 4 mepoisson v3course_1 c.v26edu c.Phealth c.chronic c.v33age_1 c.sex properuse generalinfo sideeffects HCProfesh media || country: *ANALYSIS 3RD ATTEMPT* *lOGISTIC REGRESSION WITH DICHOTOMOUS DEPENDENT VARIABLE* *Model 1 - control variables logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex country *Model 2 - received information logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex information country *Model 3 - Replace received info with type of info logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex properuse generalinfo sideeffects country *Model 4 - Add source of info

57

logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex properuse generalinfo sideeffects HCProfesh media country *ANALYSIS 4TH ATTEMPT* *BY COUNTRY: lOGISTIC REGRESSION WITH DICHOTOMOUS DEPENDENT VARIABLE* *Model 1 - control variables by country: logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex *Model 2 - received information by country: logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex information *Model 3 - Replace received info with type of info by country: logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex properuse generalinfo sideeffects *Model 4 - Add source of info by country: logit course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex properuse generalinfo sideeffects HCProfesh media *ANALYSIS 5th ATTEMPT* *lOGISTIC REGRESSION WITH DICHOTOMOUS DEPENDENT VARIABLE* *Model 1 - control variables logistic course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex i.country *Model 2 - received information logistic course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex i.information i.country *Model 3 - Replace received info with type of info logistic course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex properuse generalinfo sideeffects i.country *Model 4 - Add source of info logistic course i.v26edu i.Phealth i.chronic i.v33age_1 i.sex properuse generalinfo sideeffects HCProfesh media i.country

SELF-MEDICATION WITH ANTIBIOTICS: AN ASSOCIATION …

Documents