Background Data of the German Competence Network for Community-Acquired Pneumonia showed a gap between the recommendations of the national guideline for management of community-acquired pneumonia (CAP) and the routine care. We developed and evaluated an implementation strategy to improve the quality of care of patients with CAP.
Method A prospective, randomised, controlled trail was conducted within CAPNETZ. In four local clinical centres (LCC), the guideline was implemented by different strategies. The other four LCC served as control group. Indicators for guideline adherence comprised initial site of treatment, initial antibiotic treatment and duration of antibiotic treatment. As patient-related factors, we assessed the effect of guideline implementation on 30-day mortality and length of hospital stay.
Results Active guideline implementation yielded an increased proportion of guideline adherence to the length of antibiotic treatment in outpatients (+9.2%), the recommended antibiotic treatment (+5.6%) and duration of antibiotic treatment in inpatients (+5.0%) compared with baseline. In contrast, the proportion of patients in the control group, treated according to the guideline, decreased in the same period by 7.9%, 2.9% and 4.7%, respectively. None of these results was statistically significant. Decrease of CAP-related mortality was higher in the intervention group compared with the control group (2.9% vs 0.5%, ns).
Conclusion This study showed improvements in the process of care after implementation of a guideline for treating CAP. Further strategies, such as quality improvement cycles and medical practice audits, may enhance this effect.
- Community-acquired pneumonia
- quality of care
- randomised controlled trail
- clinical guidelines
Statistics from Altmetric.com
- Community-acquired pneumonia
- quality of care
- randomised controlled trail
- clinical guidelines
Community-acquired pneumonia (CAP) is associated with high morbidity, mortality and use of healthcare resources.1 2 In Germany, the annual incidence is estimated to be 6–8 cases/1000 inhabitants.3 There is abundant evidence of variation in the management of this condition.4–9 Several guidelines for the treatment of CAP patients have been developed. Guidelines should assist practitioners' and patients' decisions about appropriate healthcare to improve the quality of care, to reduce variability in clinical care and, finally, to reduce healthcare costs.4 10–13 However, systematic reviews document that neglecting of guidelines is common practice.12 14–16
In 2005, the German guidelines for management for adult CAP patients was published.17 Data from the German competence network for CAP (CAPNETZ), described in detail elsewhere,18 showed a gap between guideline recommendations and treatment of CAP patients in the network: low-risk patients were often (38%) hospitalised despite recommendations for outpatient treatment, and about 8% of high-risk patients were treated as outpatients. Examination of antibiotic treatment revealed both unnecessary overtreatment of outpatients and inappropriate undertreatment of inpatients. However, adherence to CAP guidelines could result in reduced need for hospitalisation,19 shorter hospital stays,7 20 shorter duration of antibiotic treatment,20 21 lower mortality20 22 23–25 and, finally, in lower costs.26 Several barriers have been discussed in explaining non-adherence with guidelines including a doubt in outcome expectancy, a lack of agreement with the guideline, a lack of physicians' knowledge, patient preferences or subjective feeling of security.6 Therefore, our study was aimed at developing an appropriate implementation strategy to disseminate the German guidelines on the management of CAP in adults in general practices and hospitals. Furthermore, we conducted a randomised controlled trial within CAPNETZ to evaluate the implementation strategy regarding the adherence to the guideline and the outcome of CAP.
Methods and materials
The competence network CAPNETZ was funded by the German Ministry of Education and Research (Bundesministerium für Bildung und Forschung (BMBF)). The long-term goal was to improve the quality of treatment for CAP patients.27 Within this network, experts and clinicians from different fields collaborated. The network consisted of nine local clinical centres (LCC). A personal tutor (PT) recruited new CAP cases on the basis of defined criteria: age ≥18 years, an infiltrate diagnosed by chest x ray and, in absence of this, clinical symptoms consisting of a body temperature ≥38.3°C (rectal) and cough, purulent sputum or positive auscultation. Criteria for exclusion were an acquired or therapeutically induced immune deficiency, florid tuberculosis or an infection of possible nosocomial acquisition. Data of the recruited cases were entered by the PT in-time, electronically using a standardised electronic report form (case report form) in a central database. Validity and consistency checks were performed by an independent party prior to data analysis. This concept allows the patient recruitment process to be followed continually, the identification of problems in-time and the quality of the project to be managed (for more information, see also http://www.capnetz.de).
Within CAPNETZ, a prospective, randomised, controlled trial was carried out. In February 2007, all LCC, except one including only hospitalised patients, were computer-based, randomly (block size four) assigned to an intervention (IG) or control group (CG). In the IG, an active implementation of the guideline was carried out. After a training period of 1 month, process of care after guideline implementation (1 April 2007 to 29 February 2008=t1) was compared with the treatment before (1 September 2006 to 28 February 2007=t0). Furthermore, after t1, participating doctors received a postal questionnaire to determine potential barriers to guideline adherence.
At the beginning of the study, we invited all active reporting general practitioners (GPs) and physicians of the participating hospitals to attend an educational lesson where the concept of the study and key elements of the CAP guideline was presented (table 1). In addition, the GPs and physicians received a poster, a short printed version and an electronic version of the guideline. GPs not attending the educational lesson were visited in their practice, informed about the study and received the implementation material. In t1, all GPs received reminders including a comparison of the treatment before and after guideline implementation every second month. In October and November 2007, a second educational seminar was offered.
In the control group, the PTs of the LCC were informed about the study, but the CAP guideline was not actively implemented.
Process-of-care and outcome measurement
Adherence to the guideline was analysed for the following variables: initial site of treatment, empiric initial antibiotic treatment and duration of antibiotic treatment. Outcome measurement included length of hospital stay, all-cause 30-day mortality and pneumonia-related 30-day mortality.
Statistical analyses were done using SPSS V.15.0 for Windows. Descriptive statistics included frequencies, percentages, means and standard deviations (SD). To compare IG and CG, χ2 for categorical variables were employed, and Student's t test and a non-parametric test (Mann–Whitney U test) for continuous variables. To compare the process-of-care and outcome parameters, we also conducted χ2 for categorical variables and non-parametric tests (Wilcoxon rank sum test) for continuous variables. To control for potential confounders, we conducted logistic regression analyses taking adherence to the guideline (no/yes) as dependent variable and age, sex, tobacco consumption and severity of CAP measured by the CRB-65, confusion, respiratory rate, blood pressure, 65 years of age and older (CRB-65) scoring system28 as independent factors. Adjusted odds ratio (OR) and their 95% CI) were calculated.
The effect of the intervention could be influenced by an interaction between time period and IG. To analyse a multidimensional contingency table, a loglinear analysis should be conducted.29 Due to the asymmetric research question, we choose the logit loglinear analysis to derive proof of the null hypothesis H0 postulating no differences between the saturated and unsaturated models with only second-order interactions.
A significance level of 5% was used throughout.
Six hospitals and 17 sentinel practices of the IG were included in the study reporting 238 CAP cases at baseline (mean age 57.8 (19.0) years, 46.2% men) and 275 CAP cases in the t1 (mean age 56.0 (18.1) years, 58.5% men). Patients in t1 were significantly older compared with patients in t0. In addition, the proportions of men, smokers and outpatients were significantly higher in t1 than in t0 (table 2).
In the CG, five hospitals and 17 sentinel practices were included reporting 302 CAP cases in t0 (mean age 61.1 (17.6) years, 58.9% men) and 348 cases in t1 (mean age 61.2 (18.2) years, 57.5% men). No statistically significant differences were observed between patients of the CG in t0 and t1.
Comparisons between IG and CG revealed significant differences concerning initial site of treatment and tobacco consumption in t0 as well as in t1. According to the higher proportion of patients in the CG treated as inpatients, significantly more patients had a higher CRB-65 score compared with IG patients (table 2).
Adherence to the pneumonia guideline
At baseline, IG and CG did not differ significantly concerning the main indicators for guideline adherence. In t1, the proportion of guideline conformity increased in the IG concerning the duration of antibiotic treatment in outpatients. The average duration was 9.6 (3.6) days, whereas patients of the CG received the antibiotic treatment for an average of 10.7 (3.6) days. Concerning the antibiotic treatment and its duration in inpatients (mean 11.2 (4.8) days in t1, CG 10.6 (4.2)), the proportion of patients treated according to the guideline increased as well (table 3). In the CG, only the proportion of outpatients treated with the recommended antibiotic treatment increased marginally in t1.
Results were heterogeneous between the several LCC in both IG and CG (data are shown in table E1).
In period t1, IG and CG differed significantly in the proportion of patients treated at the recommended site of treatment (71.7% vs 59.5%, p=0.004), the duration of antibiotic treatment in outpatients (71.1% vs 53.6%, p=0.015) and the initial antibiotic treatment in inpatients (66.2% vs 53.4%, p=0.016). The adjusted odds of IG patients being treated at the recommended site of treatment in t1 were significantly elevated, relative to the CG (OR=2.3, 95% CI 1.5 to 3.5). The adjusted odds of inpatients receiving the appropriate antibiotic treatment was also significantly elevated in the IG relative to inpatients of the CG (OR=1.8, 95% CI 1.1 to 2.8) as well as the adjusted odds for outpatients receiving antibiotic treatment for the guideline-compliant duration, relative to outpatients of the CG (OR=3.0, 95%CI 1.6 to 6.1). A significantly higher proportion of outpatients included in the IG received in t1 all of the three guideline-recommended processes of care than outpatients in the CG (28.0% vs 12.1%, p=0.001). The same could be observed in inpatients (22.7% vs 16.3%, p=0.187).
In t0, mean duration of hospital stay was 10.7 (7.6) days (IG) and 11.4 (9.5) days (CG), respectively (p=0.503). In t1, a marginal decrease could be observed in both IG and CG (IG: −0.7 days, p=0.784; CG: −0.5 days, p=0.768). The 30-day overall mortality was 5.2% in the IG and 2.9% in the CG (p=0.198) in t0. After guideline implementation, the 30-day overall mortality rate in the IG was 3.6% and in the CG 3.8%. Differences in the CAP-related mortality for IG and CG did not reach statistical significance, neither in t0 nor in t1.
Logit loglinear model
No statistical differences were found between the saturated and the unsaturated models including only the second-order interactions regarding the main indicators for guideline adherence such as initial site of treatment (p=0.325), empiric initial antibiotic treatment in outpatients (p=0.387), duration of antibiotic treatment in outpatients (p=0.109), empiric initial antibiotic treatment in inpatients (p=0.148) and duration of antibiotic treatment in hospitalised patients (p=0.352).
Barriers to guideline adherence
The questionnaire which was sent to the participating sentinel practices revealed that only three of the responders (n=17) used the interactive compact disk, seven used the poster, and 15 used the short printed version frequently for supporting treatment decisions. The main reason mentioned for non-adherence was an inadequate care at home (n=7). Further reasons were that CRB-65 score did not include underlying diseases (n=6), dissenting opinions of patients (n=4) and the difficulty to change one's behaviour (n=6). None of the responder mentioned costs as reason for non-adherence.
The results of our study suggest a higher adherence to the German guideline for the management of CAP patients after active implementation. The proportion of guideline conformity increased concerning the indicators “duration of antibiotic treatment in outpatients” (+9.2%), “antibiotic treatment in inpatients” (+5.6%) and its duration (+5.0%), whereas in the CG, a decrease could be observed in all the said indicators (−7.9%, −4.7% and −2.9%, respectively) except the “antibiotic treatment in outpatients” (+0.6%). Nevertheless, the logit loglinear analyses showed no effect of intervention or time period on the process of care. In addition, no significant effect could be observed on the 30-days overall mortality, the CAP-related mortality and the length of hospital stay.
Our study differs from previous investigations in this area in several aspects. We recruited patients not only in hospitals but also in sentinel practices, which allowed us to estimate the effect on the process of care in both. Most other studies were restricted to hospitalised patients only.20 23 25 26 Within CAPNETZ, we conducted a randomised, multicentre trial including a large population size which resulted to a higher generalisability. We used a standardised definition of CAP and standardised identical instruments when collecting data. Finally, we analysed preintervention data and compared changes in the IG with changes in the CG with “usual care”. This allowed us to draw conclusions regarding the effect of guideline dissemination.
Our study design may have limited the magnitude of the observed effects. The IG consisted of 17 sentinel practices and six hospitals. Each sentinel practice reported up to 20 patients at baseline and follow-up (median 3). Therefore, the number of CAP cases per practice was too small to obtain significant changes. It might be easier to implement a guideline in one hospital only. After guideline dissemination in one Spanish hospital, a statistically significant improvement was observed concerning the appropriate antibiotic treatment, the coverage of atypical pathogens and treatment duration.20 Since our trial was conducted within CAPNETZ, a higher awareness of the guideline in the reporting practices could be expected. This also applies to the sentinel practices of the CG and might explain the small differences. A further reason for the non-significant changes might be that the intensity of guideline implementation was too low. Only 12 practitioners visited the first educational seminar and four visited the second one. The implementation of audits or a continuous quality improvement cycle might increase the effect on physicians' behavioural change. A study determining the effect of increasing the intensity of implementing pneumonia guidelines revealed a higher proportion of low-risk patients treated as outpatients in the high-intensity group as compared with the low-intensity group.30 Results of single-site studies showed positive results of continuous quality improvement and quality management on clinical performance.16 However, an area-wide, intensive implementation strategy reaching all GPs is cost-intensive. For the realisation of our study, only limited financial resources were provided.
The limited influence on the behaviour of the practitioners may be another reason. The development of implementation strategies does not insure their use in practice. We could only offer the guideline in several ways, but finally, the practitioners decide how to treat the patient. The knowledge about potential barriers is essential in choosing the appropriate implementation strategies. Barriers to adherence can be grouped under the headings “knowledge”, “attitudes” and “behaviour”.15 For CAP guidelines, a questionnaire clarifies that physicians' lack of awareness accounts for poor compliance.31 This was not applied in our study, where we conducted an active dissemination of the guideline. Nevertheless, barriers to prescribing empirical antibiotic treatment adhering to the guideline could explained by a doubt in outcome expectancy and a lack of agreement with the guideline.6 In our study, most of the physicians rated guidelines as helpful to improve the quality of care (n=13), as good educational tools (n=10) and as a helpful guide in decision making (n=16). Only a few described guidelines as “cookbook” medicine (n=4) or as oversimplified (n=5). The most reported reason for non-adherence was related to the patient. An inadequate care at home resulted in the hospitalisation of low-risk patients.
We observed consistent improvements in guideline adherence after active implementation of the guideline compared with CG. In addition, we highlighted heterogeneity in clinical practice. Further strategies, such as medical practice audits or quality improvement cycles, may enhance the effect of guideline dissemination.
This study was supported by the German Medical Assembly [Bundesärztekammer] by grant 06-69 and the German Federal Ministry of Education and Research [Bundesministerium für Bildung und Forschung (BMBF)] by grant 01KI0103-105.
See Commentary, p 471
Thanks to Mithat Koca, Antje Tessmer, Frank Kunitz, Torsten Bauer, Norbert Suttorp (Berlin); Santiago Ewig, Gernot Rohde (Bochum); Matthias Pletz (Hannover); Klaus Dalhoff, Meral Tosun, Petra Heyer (Lübeck); Manfred Schack, Judith Pischer, Adrian Gillisen (Leipzig); Tom Schaberg, Iris Hering (Rotenburg); Christian Schumann (Ulm), Torsten Illmann, Michael Wallner, Michael Weber (IT) Grit Barten, Ludmilla Gosmann, Michaelea Finsel (main office); and all study nurses.
Competing interests None.
Ethics approval The ethic committees of the participating universities of the competence network CAPNETZ approved the study protocol and all patients included in CAPNETZ gave informed consent to participation.
Provenance and peer review Not commissioned; externally peer reviewed.
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.