If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author: Dr. rer. medic. Muna E. Paier-Abuzahra, BSc, MA. Institute of General Practice and Evidence-based Health Services Research, Medical University of Graz, Auenbruggerplatz 2/9, 8036 Graz, Austria.
Institute of General Practice, Family Medicine and Preventive Medicine, Paracelsus Medical University Salzburg, Salzburg, AustriaInstitute for General Medicine and Evidence-based Health Services Research, Medical University of Graz, Graz, Austria
Institute of General Practice, Family Medicine and Preventive Medicine, Paracelsus Medical University Salzburg, Salzburg, AustriaInstitute of General Practice, College of Health Care Professions, Bolzano, Italy
Institute of General Practice, Family Medicine and Preventive Medicine, Paracelsus Medical University Salzburg, Salzburg, AustriaIWIMED (Institute for Worldwide Informationtransfer in MEDicine), Salzburg, Austria
Quality indicators to assess the quality of primary care have only been applied on a national or regional level in European countries, and there have been no comparisons between regions of different countries. In the interventional pre-post-study “Improvement of Quality by Benchmarking – IQuaB” (level of evidence: 3), we aimed to improve and compare quality of process care in 57 participating general practices in Salzburg, Austria, and South Tyrol, Italy.
Methods
The intervention consisted of self-audit, benchmarking and quality circles. Quality indicators for eight common chronic diseases (e. g., diabetes) were extracted from the electronic health records in 2012, 2013 and 2014. Based on 19 quality indicators, a supra-regional quality score was calculated and compared using Mann-Whitney U tests.
Results
A relatively weak baseline performance was identified in both regions. In all three assessments, the median quality score increased in both regions and was significantly higher in South Tyrol than in Salzburg. During the study period the median supra-regional quality score increased from 20.00 to 38.00 in the Salzburg sample and from 47.00 to 79.50 in the South Tyrolian sample. The differences between the two regions were significant at baseline and after intervention (2012: p = 0.015, 2014: p= 0.001).
Discussion
Despite data extraction challenges in Austria, we are convinced that our data highlight real differences in (processual) quality of care between the two regions.
Conclusions
The reasons underlying the persisting differences between the two regions may include: (1) different functions in electronic health records, (2) benchmarking as an integral part of the electronic health record, (3) gate-keeping system and use of registration lists, (4) state-supported quality initiatives.
Zusammenfassung
Hintergrund
Für Qualitätsmessungen in der Primärversorgung wurden bisher in europäischen Ländern Qualitätsindikatoren nur auf nationaler oder regionaler Ebene angewendet, jedoch keine Vergleiche zwischen Ländern durchgeführt. In der interventionellen Pre-Post-Studie „Improvement of quality by benchmarking – IQuaB” war es das Ziel, Prozessqualität in 57 teilnehmenden Allgemeinmedizinpraxen in Salzburg, Österreich und Südtirol, Italien zu vergleichen und zu verbessern.
Methoden
Die Intervention bestand aus Self-Audit, Benchmarking und Qualitätszirkel. Qualitätsindikatoren für acht häufige chronische Krankheiten (z. B. Diabetes) wurden in den Jahren 2012, 2013 und 2014 aus der Arztpraxissoftware extrahiert. Mithilfe von 19 Qualitätsindikatoren wurde ein überregionaler Qualitätsscore berechnet und mittels Mann-Whitney-U-Test verglichen.
Ergebnisse
In beiden Regionen wurde eine relativ schwache Performance zu Beginn der Studie festgestellt. Der Median des Qualitätsscores stieg in beiden Regionen während der Studie an und war zu allen Messzeitpunkten in Südtirol signifikant höher als in Salzburg. Der Median des Qualitätsscores erhöhte sich während der Studie von 20,00 auf 38,00 in Salzburg und von 47,00 auf 79,5 in Südtirol. Die Unterschiede zwischen den Regionen waren vor und nach der Intervention signifikant (2012: p = 0,015, 2014: p = 0,001).
Diskussion
Trotz der Herausforderungen bei der Datenextraktion in Österreich sind wir davon überzeugt, dass die erhobenen Daten tatsächliche Unterschiede in der (Prozess-)Qualität zwischen den beiden Regionen offenbaren.
Schlussfolgerungen
Wir ziehen vier mögliche Gründe für die bestehenden Unterschiede zwischen den Regionen in Erwägung: (1) unterschiedliche Funktionen in der Arztpraxissoftware, (2) Benchmarking als integraler Teil der Arztpraxissoftware, (3) Gatekeepingsystem und Registrierungslisten, (4) staatlich-unterstützte Qualitätsinitiativen.
Up to now, the use of quality indicators (QIs) to assess the quality of primary care has only been conducted on a national or regional level in European countries [
], and no comparisons between general practitioners (GPs) of bordering regions of different countries have been made. International QI comparisons are only available for healthcare systems [
Guideline adherence of antithrombotic treatment initiated by general practitioners in patients with nonvalvular atrial fibrillation: a Danish survey, Clin Cardiol.2013; 36: 427-432
Patient and physician related factors of adherence to evidence based guidelines in diabetes mellitus type 2, cardiovascular disease and prevention: a cross sectional study, BMC Fam Pract.
Except from an initiative called Health Search, where only 2% of Italian GPs take part, a structured nationwide system to evaluate quality of primary care is still lacking in Italy [
The data presented in this paper stem from the “Improvement of Quality by Benchmarking” (IQuaB) study which was launched as a quality improvement initiative in primary care in Salzburg (Austria) and South Tyrol (Italy). The aims of IQuaB were 1) to assess quality of care of patients with chronic conditions using a quality score and itemised QIs in GPs’ surgeries, 2) to improve quality by means of self-audits, benchmarking and quality circles, and 3) to compare quality of process care in general practices in Salzburg, Austria and South Tyrol, Italy and to evaluate the impact of the intervention on the quality of care in Austria and Italy. This paper focusses on the third aim, the comparison between the two regions.
The results of the longitudinal analysis for Salzburg, Austria and South Tyrol, Italy have been published elsewhere [9,10].
Participants and methods
The interventional pre-post- study was conducted between 10/2011 and 09/2014 in two study regions: South Tyrol (Italy) and two districts (Pinzgau and Pongau) in the province of Salzburg (Austria) (Level of Evidence: 3). These two regions were chosen because both regions are characterized by a rural (health care) structures. Moreover, South Tyrol seems to be an appropriate cooperation partner for Salzburg because of geographical proximity, common language and historical and cultural ties.
Recruitment of GPs
We aimed to recruit 60 GPs, 30 in each of the two regions. For the pre-post study, no sample size calculation was performed. In Salzburg, all practising GPs were considered eligible if they worked in a single-handed or group practice in primary care in Pinzgau or Pongau (n = 133 GPs). In South Tyrol, all GPs working in single-handed or group practices were considered eligible (n = 281 GPs). GPs in both regions were invited to take part by letter, email and telephone. Participating physicians received financial reimbursement.
Choice of QIs
The IQuaB study investigated the quality of eight common chronic conditions: diabetes mellitus type 2 (DM), hypertension (HT), coronary heart disease (CHD), cerebrovascular disease (CBVD), peripheral arterial disease (PAD), chronic heart failure (CHF), atrial fibrillation (AF) and chronic obstructive pulmonary disease (COPD). For each disease, several guidelines were screened [
Bundesärztekammer, Kassenärztliche Bundesvereinigung, Arbeitsgemeinschaft der Wissenschaftlichen Medizinischen Fachgesellschaften (AWMF). Nationale Versorgungsleitlinien http://www.leitlinien.de/nvl/ Last accessed 2011 Oct 10.
] were adapted consensually. In both regions, participating physicians were involved in the process of choice and adaptation.
The level of performance on a particular QI was defined as the percentage of the total number of patients with the diagnosis who fulfilled the respective criterion. For each individual QI, the respective documentation, diagnostic test or prescription was to have been performed within a specified period, but with a 3-month tolerance range (except for prescription intervals, as chronic conditions generally require uninterrupted treatment). Country-specific regulations (e.g. state-regulated prescription rules, package sizes) meant prescription intervals were defined differently in the two regions. We therefore retrieved all prescribed drugs in the previous five months in Italy and in the previous three months in Austria.
Collection of data
Data extraction was performed on 06-09/2012, 04-06/2013 and 01-04/2014. Practice data were retrieved (e.g. number of patients with diabetes and documentation of smoking behavior per GP), but no individual patient data. In Austria, the project staff extracted data manually in the GPs’ surgeries from five different electronic health records (EHRs) [
], whereas in Italy centralized data extraction was possible using standardised data extraction software (Millewin/MilleGPG®) because all GPs used the same EHR. For QI extraction, Italian GPs had to send their EHR data to a central server on a pre-specified study date. The local project staff then had access to the data via the central server. Diagnoses were searched for and recorded using International Statistical Classification of Diseases and Related Health Problems Version 9 (ICD-9), and medication was identified using Anatomical Therapeutic Chemical/Defined Daily Dose Classification (ATC-Codes) [
In Austrian general practices, diagnoses were not usually coded and were recorded in the EHR as string variables. GPs were therefore asked to standardise the terminology of diagnoses (e.g. DM2 instead of diabetes mellitus 2) to simplify the search, but to achieve a comprehensive assessment, project staff used several variations of diagnoses during the manual search. In addition, medications were not usually recorded using ATC-codes and were therefore searched for using brand and generic names as strings. In most cases, laboratory data, smoking behaviour and blood pressure had to be searched for using strings because they were not recorded in standardised data fields. The five EHR systems used by participating Austrian GPs required different search strategies, enabling differing degrees of data extraction. To maximise the standardisation of the data extraction process, variations of search terms for diagnoses and drug prescriptions were listed for the research personnel extracting the data [
]. Due to the challenging and the time consuming data extraction in Salzburg, we forewent a data extraction of possible confounding variables (e.g. age, gender).
Intervention
After each data extraction, all GPs received graphical and written information on their QI performance in percentages (self-audit), and anonymous results for their colleagues including median values for each region and defined quality standards [
]: Based on these analyses, written and graphical feedback reports were elaborated and sent to the participating GPs to inform them about their fulfilment of each QI in percentages. The research team visited every GP after each data collection to present and discuss the results and on demand in case of need for technical support. Also telephonic support was offered which was regularly utilised by 75% of the participating GPs.
In Austria, a written benchmarking report was provided by the project team, which was sent to the GPs by post. The research team did not present and discuss results in a personal meeting, but in the quality circles because the data collection was already very time consuming. Telephonic support received very little response. Regional quality circles were set up for participating GPs with the support of the project team and offering the opportunity to discuss results and differences between physicians and to elaborate possible strategies for improvement in care (e.g. creation of reminders in the EHR remembering the GP about diagnostic/therapeutic actions to perform).
Quality circles were conducted on average once or twice a year with a duration of approximately 2 hours. The research team guided all quality circles. Additionally, two quality circles (09/2013, 05/2014) were organised for the GPs to promote transnational networking and learning. Quality circle attendance was optional.
In South Tyrol 25 GPs (69.4%) and in Salzburg 15 GPs (71.4%) participated in the quality circles at least once.
Data analysis
Calculation of the quality score: We calculated a supra-regional quality score for summative cross-sectional analysis. For the supra-regional quality score, we considered all 19 QIs that were available for both regions throughout the study period. We assigned between zero and five points per QI (maximum 95 points in total). The point assignment for each indicator was based on the median for both regions at baseline: the median minus 10% was the minimum to achieve one point. For each more 5%, one further point was added. For simplification, percentages were rounded. E.g., the median for the QI at baseline “documentation of the BMI” was 34%. A GP who had 20% of patients with documented BMI (Body Mass Index) would have received 0 points. A GP who had 54% of patients with a documented BMI would have reached 5 points for this QI.
If the median fell below 20%, 10% was the minimum to achieve one point (e.g. recording of smoking behaviour in diabetes). For each GP, one individual quality score was calculated by adding the values of the individual QIs.
Using the individual quality scores, we calculated the median regional quality scores for Austria and Italy.
Statistical analyses were carried out using the software package IBM SPSS Statistics for Windows, Version 20.0 (Armonk: NY). Units of analyses were GPs. After proving data distribution using the Kolmogorov Smirnov test, we performed Mann-Whitney-U-tests (significance level 5%) to compare the supra-regional quality score and individual QIs from the two regions. As we had a rather explorative approach to generate hypotheses than a verifying approach, we did no correction for multiple testing.
Missing data: In Austria, one EHR (PCpo) could only extract prevalence rates. Therefore, the remaining quality indicators were recorded as missing values. For these three GPs, quality scores could not be calculated. Prescription data from GPs using the EHR software Ganymed and medXpert were excluded, because extraction was not possible. As the data extraction in Austria was performed manually using string terms for the search; values that could not be valid (e.g. percentages that exceeded 100%) were excluded.
Results
Participating GP surgeries and their EHRs
The number of GPs agreed to participate and took part at the baseline and follow-up data extractions is shown in the Flow Chart (Figure 1). Out of 133 GPs 27 GPs agreed and gave their informed consent in Salzburg, Austria. In Salzburg, six GPs had to be excluded because they withdrew, or patient data could not be made available due to the EHR. At baseline, 21 surgeries took part in the study. One GP withdrew because of retirement before the first follow up. Three Austrian GPs used an EHR that could only extract prevalence rates and not QIs, so quality scores could not be calculated. Data for the remaining 18 GPs were included in the Austrian sample. Out of 281 GPs 37 GPs agreed and gave their informed consent in South Tyrol, Italy. At baseline, 21 surgeries took part in the study. One GP withdrew because of retirement before the first follow up. One GP withdrew before baseline; 36 GPs took part at baseline and the follow ups. The overall participation rate of GPs for both regions was 15.5%. 13.8% took part in baseline data extraction.
The majority of GPs in both regions worked in single practices: in Italy 25% of GPs worked in group practices (9 GPs in 4 group practices), in Austria 10% of GPs worked in group practices (2 GPs in one group practice). In Austria 30.0% of GPs (6 of 20), in South Tyrol 27.8% (10 of 36 GPs) were female. The mean age of participating GPs in Austria was 52.9 (±9.2) years and 51.5 (±8.5) years in Italy. The prevalence for the target diseases did not differ in DM, HT, CHD, CBVD and CHF at baseline. We found significant differences for PAD prevalence (Salzburg 0.6; South Tyrol 0.3, U-Test: p = 0.011) and COPD (Salzburg: 1.8; South Tyrol: 0.7; U-Test: p< 0.001) at baseline.
The median supra-regional quality score at baseline was 20.0 (Q1-Q3: 9.0-42.0) in the Austrian sample and 47.0 (Q1-Q3: 19.0-70.8) in the Italian sample. At the first follow-up (2013), it had increased to 45.5 (Q1-Q3: 11.8-70.8) in Austria and 75.0 (Q1-Q3: 34.8-88.0) in Italy. At the second follow-up (2014), the median quality score had decreased to 38.0 (Q1-Q3: 17.0-64.5) in Austria, while it continued to rise and reached 79.5 (Q1-Q3: 51.8-88.8) in Italy. The supra-regional quality scores achieved by the Austrian GPs differed significantly from those of the Italian GPs on all three occasions (baseline: p = 0.015, follow-up 1: p = 0.007, follow-up 2: p = 0.001, Mann-Whitney-U-test) (Figure 2).
Figure 2Supra-regional quality score in Salzburg and South Tyrol.
Diabetes mellitus type 2. Higher percentages were reached in South Tyrol for all QIs except HbA1c measurements at baseline, but the difference between the regions only reached significance for the documentation of smoking behaviour. In both regions, QI performance rose over time with the exception of the documentation of the Body Mass Index (BMI) and the percentage of HbA1c measurements, which had decreased slightly at the second follow-up. South Tyrol was ahead of Salzburg in all three assessments (Table 1).
Table 1Median quality indicators in Salzburg, Austria and South Tyrol, Italy.
Quality indicator
2012
2013
2014
Austria
Italy
U-test
U-test
Austria
Italy
U-test
U-test
Austria
Italy
U-test
U-test
Median
Median
value
p-value
Median
Median
value
p-value
Median
Median
value
p-value
Diabetes mellitus type 2
Body Mass Index (recorded within last 15 months)
18.4
36.7
228.5
0.981
24.0
54.9
249,5
0.425
22.7
47.6
312.0
0.077
Documentation of smoking behaviour (smoker or non-smoker)
Hypertension. At baseline, the documentation of BMI was the same in both regions and had risen to the same extent at the first follow-up. A higher, but non-significant difference was reached at the second follow-up in South Tyrol, reflecting a fall in the percentage of BMI measurements to its baseline level in Salzburg. Documentation of smoking behaviour was significantly higher in South Tyrol than in Salzburg at all assessments. The percentage of creatinine measurements was higher in South Tyrol, but a significant difference was only reached at the second follow-up (Table 1).
Coronary Heart Disease. Smoking behaviour was recorded significantly more often in South Tyrol. GPs in South Tyrol also prescribed more beta-blockers, although the difference was no longer significant at the second follow-up due to a continuous rise in prescription rates in Salzburg (Table 1).
Cerebrovascular Disease. The documentation of BMI was higher in Salzburg at baseline and the first follow-up, but the percentage was the same at the second follow-up. The differences were not significant. Registration of smoking behaviour was higher in South Tyrol at all assessments, but the difference was only significant at the second follow-up (Table 1).
Peripheral Artery Disease. The percentage of BMI recordings was higher in Salzburg at baseline and first follow-up, but had risen sharply and was significantly higher in South Tyrol at the second follow-up. The registration of smoking behaviour was also higher in South Tyrol, but not significantly (Table 1).
Heart Failure. QI performance was higher in South Tyrol, except for the documentation of BMI at baseline and beta-blocker prescriptions at the second follow-up. However, the differences were not significant (Table 1).
COPD. GPs in South Tyrol had documented smoking behaviour for more patients than their counterparts in Salzburg. The number of patients with at least one recorded spirometry was higher in South Tyrol at baseline. At the first and second follow-ups, the number of patients with at least one recorded spirometry was higher in Salzburg (Table 1).
Discussion
Beside the fact that we found relatively weak QI performance in both regions, Austrian rates were significantly lower than Italian rates. The Italian quality score was significantly higher than the Austrian score at baseline, and remained significantly higher at the first and second follow-up. In fact, the differences in the quality score between Austria and Italy became even greater. This phenomenon was observed mainly in the QIs that assessed the documentation of smoking behaviour.
For validation, we compared our data with data from other studies. Our data comparisons to other studies [
] show a tendency to underestimation in our data in both regions: The measurement of QIs requires that diagnoses and services are documented and registered in the EHR, and that this documentation is preferably coded and attributable to a specified data field in the database. Underestimation is probable and reasonable in Austria as limited opportunities to extract data from the EHRs, as well as the limited transfer of data from specialists’ and hospitals’ EHRs (e.g. spirometry) may have led to incomplete documentation of some of the QIs; prescriptions (e.g. ACE-inhibitor or ARB prescriptions, beta-blocker prescriptions), laboratory tests (e.g. frequency of HbA1c measurements) and tests performed outside the general practice (e.g. spirometry) may not have been captured in full.
In our Italian sample, we found lower prescription rates than in representative Health Search data [
]. This finding implies either that our Italian results are not representative, or the participating South Tyrolian GPs’ quality of care differed from the other Health Search participants. We found an interesting phenomenon in Salzburg that some indicators (especially for registrations of the Body Mass Index and the smoking behaviour) had a strong increase from baseline to the first follow-up and decreased slightly afterward in the second follow-up. This might be an indication for the high motivation of the participating GPs in the beginning of the study that waned afterwards. In contrast to Salzburg, the QIs increased continuously in South Tyrol. We assume that the low-threshold access to the own data and the benchmark reports in the EHRs in South Tyrol influenced the motivation of the GPs and made the quality work in each practice easier.
Strengths and weaknesses
This study was the first to compare effects of self-audit, benchmarking and quality circles among GPs in these two regions from different countries. Guideline adherence in the treatment of patients with chronic conditions assessed using a quality score and itemised QIs persisted being higher among South Tyrolian GPs than GPs in Salzburg, but increased in both regions over time.
Low participation of GPs is probably the result of a lack of awareness and of scepticism regarding the improvement that can be achieved through structured work. Furthermore, GPs often have difficulty finding the time to take part in studies. Personal engagement (e.g. attendance of quality circle, reflecting benchmarking results and take action) could have had an impact on the results of each GP, but seems not to be very probable as the participation rate in the quality circles in Salzburg was higher than in South Tyrol. There were several limitations in our study that could have had an impact on the data quality: Our sample was small, hence probably not representative. In Salzburg/Austria, the project staff performed data extraction manually, which was technically challenging and time-consuming. Data should be standardised for age, gender and morbidity for comparison between regions. However, GPs from both regions are located in a rural setting with probably similar patient groups. We could not find significant differences in the prevalence of the target diseases except for PAD and COPD at the baseline data extraction. Moreover, we did not find significant differences in the single quality indicators for PAD and COPD. The difference in the prevalences in PAD and COPD between Salzburg and South Tyrol is an interesting finding as we found indications for both regions that COPD could be underdiagnosed [
]. Maybe the difference shows only a difference in the extent of underdiagnosing – an assumption that should be proved in each region. Moreover, we found differences in several QIs between Salzburg and South Tyrol. That affects primarily QIs on the documentation of smoking behaviour in the DM, HT, CHD and CBVD - but not COPD which is probably the most important indication to talk about smoking with the patient - and furthermore creatinine measurments in patients with HT, Body Mass Index documentation in patients with PAD and prescriptions of beta-blockers in patients with CHD.
Multiple testing of QIs could have led to a higher chance of significant results. However, our aspiration was to explore commonalities and discrepancies between these two regions to generate presumptions to be proved in further works.
In spite of these circumstances, we are convinced that our data highlight real differences in (processual) quality of care between the two regions. Firstly, we do not believe that the consistently large differences between Salzburg and South Tyrol can be solely explained by technical challenges in data extraction, and secondly, the increase in the difference in the measured quality of care over time can only be explained by external factors that go beyond limiting circumstances and the intervention.
Conclusions
We consider four possible causes for the differences in the measured quality of care between the two regions:
1.
GPs in Italy record their work better as a consequence of the development of an EHR customised to suit their needs.
2.
As benchmarking is an integral part of their EHRs, Italian GPs had the possibility to directly access benchmarking data at any time. After data extraction, Austrian GPs received benchmarking data by post. This may have influenced the usage of benchmarking data and may explain the increase in the differences between the two regions.
3.
State-supported quality initiatives existed in Italy but not in Austria. Quality of care in Italy may therefore have increased before our study began.
4.
The existence of a gatekeeping system and patient registration lists in Italy may have contributed to measured differences on the QIs. The Austrian population has direct access to all levels of the healthcare system, enabling patients to bypass their GPs and making it more difficult to measure quality of care in Austrian general practices. In Italy, patients have to visit their GPs before they can use other healthcare services, enabling Italian GPs to follow all patient pathways.
Differences in the basic conditions in the healthcare systems (EHRs, gatekeeping, registration lists, quality initiatives) may explain higher rates in QI in Italy. Adjustments in primary care in Austria could possibly facilitate GPs’ treatment in patients with chronic conditions e.g. by using registration lists, the use of codes and classifications for diagnoses and medications, adaptation of EHRs for GPs needs.
The identified differences between the two regions, Salzburg and South Tyrol, present an interesting subject for further comparative health systems research and should be investigated further. For instance, it is important to investigate whether the possibility to track patient pathways provided by health systems that employ gatekeeping enables them to provide higher quality care than health systems that do not. Moreover, it should be explored whether permanently available EHR-based benchmarking is more effective than paper-based benchmarking.
Ethic
As no individual patient data were collected or processed, the Ethics Committee of Salzburg granted an ethics waiver.
Acknowledgement
The study was financed by the European funding programme INTERREG IV Italy-Austria (project number 5222), the Institute for General Practice, Family Medicine and Preventive Medicine at the Paracelsus Medical University Salzburg (Austria) and the Province of Bolzano (South Tyrol). The funding parties played no role in the study design, in the collection, analysis, and interpretation of data, in writing of the report, or in the decision to submit the article for publication.
Conflict of interest
All authors declare that there is no conflict of interest.
Guideline adherence of antithrombotic treatment initiated by general practitioners in patients with nonvalvular atrial fibrillation: a Danish survey, Clin Cardiol.2013; 36: 427-432
Patient and physician related factors of adherence to evidence based guidelines in diabetes mellitus type 2, cardiovascular disease and prevention: a cross sectional study, BMC Fam Pract.
Bundesärztekammer, Kassenärztliche Bundesvereinigung, Arbeitsgemeinschaft der Wissenschaftlichen Medizinischen Fachgesellschaften (AWMF). Nationale Versorgungsleitlinien http://www.leitlinien.de/nvl/ Last accessed 2011 Oct 10.
00008-3&id=gr1.jpg){kind=link}
00008-3&id=gr2.jpg){kind=link}