Community-acquired pneumonia (CAP) is a common, potentially life-threatening disease that is associated with much morbidity, mortality and use of healthcare resources. Recognition of the consequences of CAP and unexplained variation in quality of care has resulted in the development of clinical practice guidelines in various countries.^1–4 Several papers have reported underperformance with respect to key recommendations of these guidelines and have shown that poor physician adherence may be associated with poorer patient outcome.^5–7 However, implementation of such guidelines has not consistently resulted in improved antibiotic use in CAP.^8,9,10

The limited ability of strategies to change physician prescribing behaviour may be due to a lack of understanding about specific factors impeding and facilitating optimal performance in CAP. Studies have shown that implementation strategies are more likely to be effective if they focus directly on problems in care provision and factors that influence change.¹¹ Surveys of internists’ attitudes toward clinical guidelines in general report barriers such as a lack of familiarity with or confidence in the guideline. Internists said they were worried about effects of guidelines on their clinical autonomy, on healthcare costs and on satisfaction with daily clinical practice.^12–15 For CAP guidelines, a questionnaire has clarified that physicians’ low awareness may account for poor compliance.¹⁶ In another study, professionals reported that a large variety of barriers inhibited successful implementation of a critical-care pathway for CAP.^17,18 These studies all focussed mainly on professional knowledge and attitudes.

We used in-depth interviews and small group sessions to qualitatively study the whole spectrum of patient, care-provider, system and guideline barriers that impede judicious antibiotic treatment for CAP. We discussed six key recommendations from guidelines on antibiotic treatment for CAP and used a validated framework to standardise the reporting of barriers.^19,20 This model suggests that physicians fail to adhere to guidelines in the presence of an internal barrier that has a cognitive (awareness or knowledge) or affective (attitude or motivation) component, or in the presence of an external barrier (patient, guideline and environmental factors) that restricts the professionals’ ability.

Data obtained with these qualitative techniques will help us to better understand which barriers we should overcome and will enable us to generate hypotheses for potentially effective strategies to improve physician adherence.

METHODS

Study design

We conducted semistructured interviews to understand the barriers to optimal performance with respect to six key recommendations of antibiotic treatment for CAP (table 1).

Participants

We selected care providers with all levels of experience from various professional backgrounds and hospital settings (purposive sampling²¹). To do so, we asked medical directors of three secondary care hospitals in the South East of the Netherlands (two non-university teaching and one non-teaching hospital) to provide an exhaustive list of residents and specialists in internal medicine and pulmonology, medical microbiologists and clinical pharmacists. In the three hospitals a total of 42 residents, 20 specialists, 6 microbiologists and 5 clinical pharmacists were working. We then randomly selected professionals from this list. An invitational letter was sent to 12 residents, 6 specialists, 3 microbiologists and 3 clinical pharmacists.

Procedure

One trained interviewer performed the semistructured interviews. The interviewer (JS) was a resident in internal medicine, with no relationship to the hospitals or interviewees that were selected. Clinical doctors were asked to present a clinical case: we asked them to select their most recent patient with CAP who had been admitted in the four weeks preceding the interview. If no such patient could be found, the interviewer presented a previously prepared “dummy” patient before the interview. This was also done before the interviews of all non-clinical care providers (medical microbiologists and clinical pharmacist). All sessions were audiotaped. New interviews took place until no new information was gleaned.

Questions

The questions were open-ended and linked to the clinical case history as closely as possible. They focussed on perceived barriers to appropriate use of antibiotics in CAP as described in six key recommendations, and covered both diagnostic and therapeutic aspects of the process of care in a logical order (table 1). These recommendations had been systematically selected by an expert panel from national guidelines edited by the Dutch Working Party on Antibiotic Policy (SWAB) and the National Society for Respiratory Physicians (NVALT), international guidelines from the Infectious Diseases Society of America (IDSA), American Thoracic Society (ATS), British Thoracic Society (BTS) and the European Respiratory Society (ERS) and a systematic review of the literature.²² An interview guide for each key issue was developed, and then it was adapted after two pilot interviews with senior residents. The interview guide contained questions clarifying potential barriers to optimal antibiotic use at all possible levels (patient, doctor, system, and guideline).

Analysis of barriers

All audiotaped interviews were transcribed verbatim. Two researchers (JS and MH) independently reviewed the manuscripts and marked comments about barriers to adherence. Remarks of professionals were compared and classified into categories of potential barriers to physician adherence according to a conceptual model developped by Cabana et al.^19,20 This model suggests that physicians fail to adhere to guidelines in the presence of an internal barrier that has a cognitive (awareness or knowledge) or affective (attitude or motivation) component, or in the presence of an external barrier that restricts the professionals’ ability (see fig 1). The external barrier may contain factors relevant to the patient, the guideline or the environment. For example, the remark that “doctors often do not know whether they have to wait for collection of a sputum culture before starting antibiotic therapy” implies a knowledge-based barrier (a lack of awareness or familiarity with the guideline recommendation). However, “Sometimes administration of antibiotics will have to wait until the doctor finds time to start an IV drip” would be classified as an external, organisational factor (lack of time). The two reviewers discussed all the remarks that they had individually highlighted and classified until consensus was reached. They consulted a third researcher (SN) to make a formal judgement about differences in classification. If controversy remained, the comment was considered ambiguous and was excluded. Our principal aim was to describe the whole spectrum of possible barriers rather than quantifying their relative importance, so all types of barriers that were mentioned are presented in table 2 with their most representative remark(s). However, the barriers that were mentioned most often by our interviewees are discussed more in detail in the results section.

RESULTS

Interviews took place in 2003. All invited professionals agreed to take part. Eighteen professionals (9 residents, 6 consultants, 2 microbiologists and 1 clinical pharmacist) were interviewed until no information that we hadn’t received in preceding interviews was given. In our first interview sessions with residents, we discovered that for some recommended processes of care, barriers were mainly attributable to practical, organisational difficulties needing a multidisciplinary solution from those most closely involved. We therefore decided to add two interactive small group interviews with a nurse from the emergency department, a nurse from a pulmonology ward and a resident.

Before performing our first interview with a clinical pharmacist, we already doubted whether he/she was a “care provider relevant to the process”. Due to the specific characteristics of our guideline, the pharmacist’s contribution to our analysis of barriers was essentially limited to two recommendations: adjustment of antibiotic dose to renal function and switch therapy. In addition, a clinical pharmacist was reported to be seldomly present at ward meetings of general internal medicine and pulmonology in our recent survey at Dutch secondary care hospitals and thus may have little insight as to the reasons for non-adherence to guideline recommendations.²⁴ After interviewing the first clinical pharmacist, we decided not to recruite another one as our our doubts were confirmed.

The mean age of the 24 participants was 34.5 years (range 25–56); 10 participants were women. We encountered a wide spectrum of possible barriers to optimal antibiotic treatment for CAP. Table 2 presents transcripts of comments, grouped by theoretical barriers to adherence. All types of barriers, along with the most representative remarks, are presented. Barriers that were mentioned most often (⩾7 times) by different interviewees were considered frequent. In the table, these barriers are shown in bold text. Analysis of our data revealed that each single recommendation elicited its own specific pattern of barriers. We present the most important patterns for each of the recommendations.

Barriers to prescribing empirical antibiotic therapy adherent to the guidelines

Professional barriers to adherent prescribing included a doubt in outcome expectancy and predominantly reported by residents and specialists (“You never know … penicillin has a very narrow spectrum … I would not feel at ease treating my patient with only that” [specialist pulmonology S3]) and a lack of agreement with guidelines (“In the Netherlands, we always want to start with a very narrow spectrum, preferably with penicillin, and we add erythromycin only if a patient is really deteriorating … why not turn it around? Why not start with broad-spectrum therapy and tailor it down to narrow-spectrum therapy as soon as the culture results become available?” [resident internal medicine R6]). External barriers were mentioned to a lesser extent and were mainly related to the social context in which professionals operate: “Out of courtesy to colleagues, no criticism of the chosen antibiotic regimen is made at end-of-shift meetings” [resident internal medicine R4].

Barriers to timeliness of administration of antibiotics

Most interviewees (residents, specialists, microbiologists and nurses) mentioned external barriers related to organisational factors (for example, substantial delays in delivering laboratory results to the emergency department, antibiotics not present on the ward, IV drip not started). However, barriers were also created by the physician’s lack of knowledge about the impact that timely antibiotic administration can have on patient outcome and a lack of agreement with the guideline (for example, several specialists and residents stated “this rule only applies to the patient with CAP who is severely ill”). Some remarks combine different barriers: “Ward nurses prioritize non-medical issues (such as diet and social setting) during intake, leaving prescribed medication, including IV antibiotics, to the last or postponing administration until regular medication rounds” [awareness of importance and social-organisational context, specialist internal medicine S2].

Barriers to adjusting the dosage and dosing interval to accommodate renal function

Reasons for omitting dosage adjustment to renal function were mainly attributable to lack of awareness—for example, that antibiotic dosage should be adjusted for patients with reduced renal function; that a moderately elevated serum creatinine concentration could conceal a significantly decreased renal clearance in elderly patients. Calculating the creatinine clearance with the Cockroft-Gault formula for every patient was regarded as time consuming and unpractical by residents (external organisational barrier). According to medical specialists, only half of the residents were able to calculate creatinine clearance with this formula (lack of familiarity).

Barriers to streamlining therapy

Doctors said that they felt uncertain about tailoring empirical broad-spectrum antibiotics to narrow-spectrum antibiotic therapy (once culture results were known), especially when a patient had been very ill on admission. “Never change a winning team” was quoted by several residents and specialists and observed by medical microbiologists. Organisational barriers also apply: “The results only become available 3 to 5 days after culturing, and due to weekends and poorly computerized reporting systems, the time between availability and notification of the results becomes even greater”[specialist pulmonology S6] and “When patients do well with the initial treatment, streamlining therapy is postponed until the supervisory ward round, which is generally held only once or twice a week” [resident internal medicine R4]. Inexperienced residents, but not specialists, appreciated spontaneous interference with prescribed antibiotic therapy by non-clinicians (medical microbiologists and clinical pharmacists).

Barriers to the intravenous-oral switch

Treating a patient intravenously provided some clinicians with a subjective feeling of security, especially when a patient was severely ill on admission (attitude-based barrier). Identical oral formulae are not available for some broad-spectrum antibiotics (ceftriaxon and cefotaxim). This complicates switching to an oral antibiotic when no culture results are available: “I find it difficult to select an oral alternative with the same spectrum” was stated by several residents (knowledge-based barrier). Social pressure may also postpone a timely switch: “Residents tend to wait until the supervisor’s ward round before taking decisions” [nurse ward, N2]. Finally, clinicians said that nurses played a facilitating role in the IV-to-oral switch, asking for the IV drip to be discontinued at nearly every ward round (organisational facilitator).

Barriers to blood and sputum culturing

Most residents and specialists mentioned that blood culturing was easily forgotten for elderly patients with CAP (who are often afebrile) or for severely ill, hypothermic patients: “We only request blood cultures if a patient with CAP has a fever” [lack of guideline applicability to patient, specialist pulmonology S1]. One nurse remarked: “I will give the patient a receptacle for sputum, but I will not wait until he/she produces it … it may then remain at a bedside table (or in the ward’s fridge) for days” [nurse ward N1]. This reveals a variety of barriers: lack of awareness of the importance of careful and efficient handling of sputum; lack of motivation; lack of communication between doctors, nurses, and patients; and organisational constraints due to limited opening hours of microbiology labs. Influence of non-clinical professionals (for example, medical microbiologists) regarding the availability of diagnostic tests (for example, urine antigen testing for Legionella spp) is considered undesirable by most clinical specialists (barrier of social interference).

DISCUSSION

A large variety of barriers to key recommendations on antibiotic usefor CAP was reported by our interviewees. Each recommendation elicited its own pattern of barriers that should be overcome. Non-adherence to guidelines for empirical antibiotic therapy was mainly attributable to physicians’ negative attitudes towards the guideline. For another recommendation (“timely administration of antibiotics”), logistical and organisational factors were reported to be the most important barriers. Improving performance on either recommendation would obviously require a different approach.

Previous studies have reported barriers to CAP guideline (or critical-care pathway) adherence as a whole,^16,17,25 rather than to its various recommendations or they focus on only one recommendation.²⁶ In addition, these studies do not consider the views of care providers other than clinical specialists. In our study, some very important barriers were, however, suggested by other care providers: nurses in the emergency department mentioned that some of the antibiotics recommended by the local guideline were not directly available at the emergency department (for example, due to a shortage of space to stock medication). This barrier was not perceived by physicians. In addition, our choice to interview physicians of various educational levels resulted in a variety of perceptions, which may have been overlooked when only medical specialists were interviewed. This clarified which professional levels our improvement strategy should target for some recommendations.

Since there is a general pattern in the type of barriers for each recommendation, improvement strategies should focus on this pattern. While many studies describe barriers to guideline implementation, there is only little practical advice on how they should be translated into practice changing strategies.²⁷ It seems logical to provide an educational intervention in a situation where lack of knowledge is an important barrier, but this remains far less clear for other perceived barriers: for example, how can a social structure in a hospital or a department be changed? How can attitude towards a guideline be changed? Tailoring of an intervention can be difficult because there is no one-to-one relationship between the objectives and interventions, and empirical evidence on links between specific interventions and specific objectives is limited.

In our guideline implemantation study (submitted), we tried to develop interventions in a structured manner, using an exploraratory method.^27,28 For each factor impeding adherence, the best possible intervention was chosen, using evidence of effectiveness when available, as well as experience, common sense and creativity.

In our case, improving guideline adherence for the prescription of empirical antibiotic therapy should aim at changing a physician’s attitude rather than improving knowledge about the guideline. An intervention should be directed towards both specialists and residents, because—in line with a previous study¹⁸—our results suggest that residents do not decide independently about antibiotic policies in Dutch hospitals. There is a perceived lack of evidence justifying recommendations and a lack of confidence in the guideline developers, which leads to a lack of agreement with current guideline recommendations. This, along with the reported social pressure among professionals at end-of-shift meetings might reflect the ongoing discussion about benefits and doubts concerning empirical regimens of broad-spectrum antibiotics covering both atypical and classic pathogens.^29,30 A tailored intervention to remove these barriers would seem logical: involve clinical specialists in actively developing local guidelines based on the available evidence, organise small group discussions on appropriate prescribing and produce a clearly written and unequivocal critical-care pathway.

In contrast, for other recommendations, such as “timely administration of antibiotics”, guideline adherence was predominantly impeded by external, organisational barriers. Some of these barriers may be dealt with easily: make the antibiotic available in the emergency department and integrate conflicting ward protocols. Other external barriers, such as either doctors’ or nurses’ lack of time, are often more difficult to address as they need interventions on a higher organisational or political level.

In this paper we have aimed to describe the whole range of barriers one could potentially identify for pneumonia guideline recommendations. We realise that barriers may be quite variable throughout different hospital settings. It is clear that several of the external, organisational barriers (such as the timing of a “supervisory ward round”) may only apply to a local hospital setting, but another Dutch study reported similar results and this suggests that our findings are—at least—not confined to these three hospitals but represent a Dutch hospital setting.¹⁸ This paper, however, also describes the patterns of barriers that surface when analysing the outcomes of all interviewees. One might think that these patterns identified were unique to the institutions studied. Our findings, however, correspond with findings from other qualitative research where similar antibiotic practice beliefs were found.^17,26

A qualitative approach is the best method for exploring the reasons and hidden motives for non-adherence to clinical guidelines. However, this study may have some limitations. First, there is a chance that we selected mainly cooperative care providers who underreported potential barriers, which may have limited generalisability. As our interviews were conducted in a non-confrontational setting, we believe that care providers were less likely to give “professionally acceptable” or “socially desirable” responses. We deliberately choose to let interviews be performed by a resident in internal medicine (JS). Due to the (sometimes complex) clinical case study that was discussed, someone with a medical background was considered necessary. The interviewees were aware of the fact that the interviewer was medically trained, but they did not know the interviewer in a personal or professional context. They did not know his clinical specialty, but they knew that the interviewer was a researcher. As far as we know there has been no research performed to evaluate whether interviews performed by different types of professionals lead to different results. More methodological research on this topic is required.

Secondly, we realise that barriers that are reported by care providers may be different from those observed in real practice. However, results from a study on determinants of adherence to guidelines on antibiotic use for CAP in a sample of 498 patients confirm many of the barriers presented in this paper.³¹

In summary, we find that each key recommendation for the optimal antibiotic treatment of CAP meets its own pattern of barriers that must be overcome. This finding suggests that future improvement strategies should focus on different types of interventions for different aspects of the guideline.