Article Text

Download PDFPDF
Factors that impact on emergency department patient compliance with antibiotic regimens
  1. Jiong Ho1,
  2. David McD Taylor2,
  3. Miguel S Cabalag1,
  4. Antony Ugoni1,
  5. Michael Yeoh2
  1. 1University of Melbourne, Melbourne, Victoria, Australia
  2. 2Emergency Department, Austin Hospital, Melbourne, Victoria, Australia
  1. Correspondence to A/Professor David McD Taylor, Emergency Department, Austin Health, Studley Road, Heidelberg, Victoria 3084, Australia; david.taylor{at}


Aim To investigate factors that impact upon compliance with antibiotic regimens among patients in the emergency department (ED).

Methods This was a prospective cohort study of patients prescribed antibiotics in a single ED. Patients were identified by witnessing the consultation, medical records and ‘after hours’ prescriptions. Data were collected on demographics, presenting condition, usual medications, antibiotic regimen and instructions given. At follow up 7 days later, data were collected on compliance, antibiotic cost and packaging, side effects, difficulty with the regimen and other medical advice sought. The association between compliance and predictor variables was examined using multivariate logistic regression.

Results 192 patients had complete data for analysis. Using two definitions of compliance (100% and ≥80% of prescribed doses), antibiotic compliance was 80% and 93%, respectively. Unemployment was negatively associated with 100% compliance (OR 0.24, 95% CI 0.07 to 0.78) and taking ≥2 regular medications was positively associated with 100% compliance (OR 4.2, 95% CI 1.2 to 15.5). No variable was associated with compliance at the ≥80% level. However, patients who were female, employed, born overseas, better educated, prescribed a single antibiotic or who had a longer course, a single dose per day, medication rather than a prescription and tablets rather than capsules tended to be more compliant. Forgetfulness, improvement of symptoms and side effects were the main reasons for non-compliance.

Conclusion Compliance was better than reported elsewhere. The good compliance among patients taking ≥2 regular medications may relate to their established medication routines. Scope exists for ED pharmacists to intervene with patients ‘at risk’ of poor antibiotic compliance.

  • Emergency departments
  • clinical management
  • pneumonia/infections
  • emergency care systems
  • paramedics
  • respiratory

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Close adherence to the prescribed timing and dosing of antibiotics is desirable for a number of reasons. First, treatment failure may ensue if irrational antibiotic consumption occurs—for example, erroneous timing of doses,1 premature cessation of the course2 and use of ‘left over’ antibiotics.3 4 This can lead to reduced therapeutic concentrations and antibacterial activity and treatment failure.5 Antibiotic non-compliance should be considered if a patient shows less than satisfactory improvement.6–9

Second, antibiotics in low concentrations may exert selective pressures on pathogens.10 If patients alter their dosage regimen, the serum antibiotic concentration may be suboptimal to inhibit some bacterial growth. Consequently, only the most susceptible strains are killed and resistant progenies may gradually be created. This pattern is well documented in the emergence of multidrug resistant Mycobacterium tuberculosis.11

Third, the combination of treatment failure and antibiotic resistance ultimately translates to additional healthcare costs.2 11 These include the requirement for novel and more expensive antibiotics, prolonged hospitalisation, increased consultation time and wasted resources on unused medications.12

Finally, poor antibiotic compliance has been linked with antibiotic misuse. A large UK survey observed that 16% of patients had kept an unfinished course of antibiotics from previous infections.3 An American study4 reported that 17% of emergency department (ED) patients had used antibiotics in the previous 12 months without consulting a physician, most commonly for symptoms of upper respiratory tract infection.

Antibiotic compliance in the ED and general practice (GP) settings has been reported to be as low as 30–40%.2 13 Unfortunately, differences in methodology and definitions of compliance make comparison of studies within and across different settings difficult.7 Ideally, 100% (‘absolute’) compliance, where all the prescribed antibiotic doses are taken, will be achieved.13 However, some authors report that 80% (‘clinically significant’) compliance, where ≥80% of the doses are taken, is more clinically relevant.13–15

There is substantial literature relating to various forms of ED patient compliance such as follow-up appointments,16–18 prescription filling16 19 and obtaining follow-on prescriptions.20 However, there is a relative paucity of ED studies that have specifically examined antibiotic compliance and its determinants. We aimed to determine antibiotic compliance rates and the factors that impact upon compliance among patients who are prescribed antibiotics in the ED and then discharged home. This information may inform strategies aiming to improve patient compliance rates in the ED.


This was a prospective cohort study undertaken in a tertiary referral mixed (adult/paediatric) ED with an annual census of approximately 70 000 patients. Patients were enrolled between September–November 2007 and February–April 2008. This study was authorised by the hospital's Human Research and Ethics Committee.

Patients were included if they were aged ≥18 years, were prescribed a course of oral antibiotics and were discharged home. Patients were excluded if they refused to participate, had communication difficulties (eg, language, deafness) or cognitive impairment, did not self-dispense the antibiotics (eg, nursing home residents), were lost to follow-up or if they had a treatment change before completing the antibiotic course.

The primary study end points were compliance with the antibiotic regimen at two levels: total compliance (100% of prescribed doses taken) and clinically significant compliance (≥80% of doses). The point at which the patient deviated by >20% from the prescribed doses has been shown to give a sensitivity for detecting true non-compliance of 82% and a specificity for detecting true compliance of 91%.21 Secondary end points were the nature of the non-compliance (failure to commence the course, premature termination and missing doses), the reasons for non-compliance and antibiotic side effects.

A convenience sample of patients was enrolled. The principal investigator, present in the ED mainly between 09:00 and 17:00 on weekdays, identified most patients by witnessing the ED consultation. Antibiotic prescriptions deposited in the ‘after hours’ drug cupboard, discharge diagnoses on the ED electronic database and ED pharmacists all assisted in identifying additional patients. No patient was advised of the study while they were in the ED. In order to avoid a potential Hawthorne effect (modification of behaviour resulting from observation), patients were only invited to enrol during a follow-up telephone call. This was made at least 1 day after the antibiotic course should have been complete. Patients were considered lost to follow-up if they could not be contacted within 1 month of the ED presentation. Up to eight telephone calls were made to contact patients.

A data collection document was specifically designed for the study. Data from an explicit medical record review by the principal investigator were entered into the first section. These comprised patient demographics, diagnosis, number of usual medications, antibiotic regimen (drug, frequency of doses, length of treatment), instructions given and whether the patient had a regular GP. The second section was completed during the telephone follow-up. Following an explanation of the nature and purpose of the study, patients were invited to enrol and verbal informed consent was sought. For those agreeing to participate, the principal investigator administered a structured questionnaire. Data were collected on compliance with the antibiotic regimen, cost and packaging of the antibiotic, difficulty in fitting the antibiotic into their daily schedules, side effects, perception of illness severity, whether they had sought other medical advice following discharge, level of education and use of complementary and alternative medicines.

Sample size was based on comparison of compliance rates. It was believed that an absolute difference of 15% in compliance between two patient groups of interest was clinically significant and a conservative estimate of compliance was set at 50% for one group. Accordingly, a sample size of 368 patients was sought (equal number of patients in each group, level of significance 0.05, power 0.8).

Separate data analyses were undertaken based on compliance at the 80% and 100% levels. Univariate analyses preceded logistic regression analyses. Results are presented as ORs and corresponding 95% CIs. All analyses were performed using Stata 1999 statistical software (Stata Corporation, College Station, Texas, USA).


During the data collection periods 321 patients were screened for enrolment. Of these, 192 had complete data for analysis (figure 1). Treatment courses of antibiotics were prescribed for 162 (84.4%, 95% CI 78.3 to 89.1) of these patients (urinary infection 52, cellulitis 30, chest infection 17, tonsillitis 11, abscess 11, other infection 41) and treatment dose prophylactic courses for 30 (15.6%, 95% CI 11.0 to 21.7) patients (laceration 15, fracture 5, bite 4, other 6).

Figure 1

Flow chart of patient recruitment.

Overall, compliance rates were high: 153 patients (79.7%, 95% CI 73.2% to 85.0%) reported 100% compliance and 178 patients (92.7%, 95% CI 87.8% to 95.8%) reported ≥80% compliance (table 1). Univariate analyses showed that compliance at the 100% level was significantly greater if the number of doses per day was small, if there was a follow-up appointment and if the antibiotic was formulated as a tablet. Multivariate analyses showed that 100% compliance was significantly greater if the patient was employed and if they usually took two or more medications.

Table 1

Univariate and multivariate analyses results of total and clinically significant compliance

Univariate analysis showed that no variable was associated with compliance at the ≥80% level. However, trends were apparent. Patients who were female, employed, born overseas, better educated and prescribed a single antibiotic or who had a longer course, a single dose per day, medication rather than a prescription and tablets rather than capsules tended to be more compliant. The small number of non-compliant patients at the ≥80% level (n=14) precluded multivariate analyses.

Omission of single doses was the most frequently observed pattern of non-compliance. Although a high antibiotic compliance rate was recorded, many patients took longer than the prescribed number of days to complete the course. Two patients failed to commence the course. Some patients also reported dosing irregularities where they took more than the prescribed number of tablets per day or less than the prescribed days to complete the course. Irregular interdose intervals and regular daily omission of doses were also frequently reported. Premature termination of the course was also noted. One patient restarted the antibiotics after stopping them prematurely. Some patients had more than one pattern of non-compliance—for example, patients with shift work commitments had irregular timing between doses as well as omission of doses.

The reported reasons for non-compliance included forgetfulness (27 cases), improvement of symptoms (11), side effects (9), difficult regimen (8), hectic lifestyle and work (7), concerns of drug-drug interactions (2) and not having the antibiotics with them (2). Another common reason for non-compliance was that the antibiotic was ‘thought to be redundant’. Of the 15 patients who received two antibiotics, 2 patients were non-compliant because the shorter course ran out and they forgot to finish the longer one. Patients also reported other life stressors and beliefs that medications were harmful. Similar to the patterns of non-compliance, patients often cited a combination of reasons for non-compliance. For example, patients frequently reported forgetting antibiotics simultaneously with improvement in their condition or a hectic lifestyle causing forgetfulness.


From the patients' point of view, decision-making regarding compliance is a rational and logical process where clinical content of the condition is only one among many factors associated with compliance.22 In this study the compliance rates were surprisingly high. Our 80% compliance rate (92.7%) was similar to that reported in a Sydney ED (95%).23 However, the Sydney study definition of compliance—“whether or not the patients took the antibiotics ‘as prescribed’”—was not specific and a direct comparison cannot be made. Other ED studies have reported lower compliance rates. In an American municipal ED,2 only 31% of patients completed the 10-day antibiotic course, 41% stopped the antibiotic early and 28% failed to start treatment. Unfortunately, the authors did not specifically define compliance. Lam et al15 also reported an overall compliance rate of 69%, where 78% took ≥80% of the antibiotics but 9% failed to start treatment.

No ED study has evaluated antibiotic compliance using the 100% compliance threshold. Importantly, antibiotic compliance rates decreased from 92.7% to 79.7% with this more stringent definition. This was expected as resolution of symptoms can discourage 100% compliance.2 24 It was possible that patients equated the resolution of symptoms to full recovery or bacteriological eradication and ceased the antibiotics early.

It is difficult to explain the reasons for the high antibiotic compliance rate found in this ED setting as only two variables were associated with 100% compliance on multivariate analysis. The first was employment, which is consistent with a GP study.14 Otherwise, the effect of this variable on antibiotic compliance has not been reported in the ED literature.

The second variable was taking two or more regular medications. Although speculation, these patients are accustomed to a medication routine and have a better understanding of the need for compliance. Only one study in the emergency medicine literature has identified ‘too many tablets’ as a determinant between patients with 80% compliance and those who failed to commence therapy.15 However, the definition of ‘too many tablets’ was ambiguous as no differentiation was made between ‘too many regular medications’, ‘too many antibiotics’ or ‘too frequent dosing’. It is not therefore possible to compare our findings directly.

The trends found in our study may be important. Compliance tended to be inversely related to dose frequency at both the 80% and 100% compliance levels. This has been reported across a range of different clinical settings.1 7 13 22 25 One study reported a fall in compliance from 93% to 77% when the regimen changed from once daily to thrice daily.13 Also, complex regimens of antibiotics, including longer courses, have been consistently shown to discourage compliance.1

The trend towards better compliance among those given or recommended to have a follow-up appointment may relate to a perception of increased disease severity or a possible Hawthorne effect. In one Australian GP setting,14 patients who perceived their conditions to be more severe tended to be more compliant. Since two-thirds of our patients perceived their condition to be severe and very severe, it is likely that this encouraged compliance. Notably, however, perception of disease severity was not a variable independently associated with compliance.

In the ED, antibiotics may be dispensed in various ways—prescription, full course of antibiotic tablets/capsules or as a starter pack with instructions to obtain a follow-on prescription. Other ED studies have examined compliance with prescription filling and obtaining follow-on prescriptions.16 19 20 These have generally revealed low compliance rates. In a randomised controlled trial,19 patients were either given a fully paid for prescription that could be filled by a designated pharmacy or the full antibiotic course itself. Only 74.2% of patients in the pharmacy group filled the prescription. In another study,20 patients were given a 3-day starter pack of antibiotics and were all instructed to obtain a follow-on prescription from either the hospital pharmacy or their GPs; only 48% of patients did so.

In our study, patients who received antibiotic tablets directly from the ED tended to have better compliance than those who received a prescription. It has been suggested that direct antibiotic dispensing in the ED may enhance compliance by preventing patients from not obtaining the medication.19 While economic cost may reduce the filling of a prescription,2 no ED study has suggested that it directly influences antibiotic compliance. In our study, patients who had out-of-pocket costs had similar compliance rates to those who did not. Finally, patients who are discharged during pharmacy opening hours may have better compliance because the vicinity of the hospital pharmacy reduced the inconvenience of filling the prescription.2

Tablets—especially when not in blister packaging—tended to promote compliance. The effect of formulation has not been reported in the adult ED setting. However, in children, the taste of syrup has been identified as a predictor of compliance.26 27 One researcher also suggested that large difficult-to-swallow tablets promoted general medication non-compliance and that elderly patients with poor dexterity experienced difficulty with packaging such as child-safe containers.

Patients not born in Australia tended to have better compliance; an explanation for this is not immediately apparent. This has not been reported previously in the emergency medicine literature. Language is not necessarily correlated to country of birth and has not been reported as a determinant of compliance.16 23

Education has not been reported as a determinant of general medication or antibiotic compliance in other clinical settings.1 7 However, a UK survey found that the less well educated were less knowledgeable about antibiotics.3 This could lead to a lack of awareness on the importance of completing antibiotics, resulting in non-compliance.

Interestingly, patients prescribed antibiotics for ≥6 days tended to have better compliance than those who received a 5-day course. This is contrary to other studies where shorter courses of antibiotics or other drugs were associated with better compliance.1 3 9 26 This finding is difficult to explain but might be affected by confounding variables such as perception of disease severity or the advice given.

Patients who experienced side effects from the antibiotics tended to be as compliant as those who did not. However, other studies have frequently reported that side effects discourage antibiotic compliance.1 2 9 15 28 Our finding may relate to the small numbers of patients who experienced side effects or the mild nature of the side effects.

In this study, antibiotic compliance tended to be similar between those who could and those who could not nominate a GP. This is inconsistent with an ED study by Dinh et al23 which suggested that GP nomination was linked to improved follow-up management of patients and could promote compliance.

The designation of the treating doctor was used as a surrogate for age in our study. Patients who were seen by a resident tended to be less compliant. This is inconsistent with the finding that patients seen by younger doctors are more likely to be compliant.14

This study has important limitations. Forty-nine of the 241 enrolled patients had incomplete data for analysis with missing data points distributed across all variables. While these exclusions had the potential for selection bias, the basic demographics variables (age, gender, employment) of those included and excluded in the analysis did not differ (p>0.05). Hence, this potential bias is unlikely to have affected the results.

As only a proportion of the ED consultations were witnessed, six variables were excluded from analysis and their impact is therefore unknown. Convenience sampling over two separate periods and the use of various means to identify patients may have introduced selection bias. However, there is no reason to believe that compliance (either extent or direction) was affected by the timing of the ED presentation or enrolment technique. Despite a lengthy study period, the availability of the principal investigator was the major reason why a small number of patients was screened.

As patients were initially unaware of the study, the Hawthorne effect is unlikely to have affected the results. The patients' self-report of compliance may have been affected by measurement and prevarication bias and this may have resulted in patients overestimating their compliance rate. As a non-judgemental atmosphere facilitates accurate compliance rates,6 9 an understanding was provided in the follow-up interviews that ‘errors in taking medication were expected’.6 Finally, as prevarication would have conferred little benefit to the patient, the magnitude of this bias was probably small. Recall bias was probably minimised as most patients were followed up soon after their expected course completion.

The investigators who undertook the interviews (JH and MC) were aware of the study end points. While this may have introduced interviewer bias, the questions were mainly ‘closed’ and there was limited opportunity for investigator interpretation to have affected the results.

The 80% and 100% rates are crude compliance end points and take no regard of the accuracy of interdose timing, double dosing for forgotten doses or the absolute times for course completion. These variables may be as clinically important as crude compliance rates and deserve further investigation.

The sample size calculation was based on an anecdotal expectation of a compliance rate of 50%. During data collection it was apparent that the overall compliance would be considerably greater than that on which the sample size calculation was based. The study was terminated with the expectations that, at least, trend information would be generated. Importantly, had the sample size been based on an expected compliance rate of >50% (eg, 75%) and still with a clinically significant absolute difference of 15% between the groups, a smaller sample size would have been required. While larger studies may show additional variables statistically associated with compliance, the clinical significance of these associations may be limited. As a single-site study, its external validity is questionable.


Antibiotic compliance among the ED patients examined was good. Only employment and taking two or more regular medications were significantly associated with compliance rates. These findings and other apparent trends may inform the manner in which patients are instructed in the use of these drugs. Specifically, the importance of completing antibiotics as prescribed should be emphasised to those who may be at risk of non-compliance. It is recommended that subsequent studies investigate other components of compliance, especially interdose variability.



  • Funding The study was funded by a grant from the University of Melbourne, Australia. However, the University played no part in the design or execution of the study, data analysis or manuscript preparation.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Austin Health, Victoria, Australia.

  • Provenance and peer review Not commissioned; externally peer reviewed.