Page 1augustus 8, 2002
Images of minimal clinical data – geographical variation of pathology in Belgian hospitals
Dr. Hilde Pincé
MD, Doctor in medical sciences, Master in scientific computer applications
Institute : University Hospital of Leuven, Staff member of general direction, Herestraat 49, 3000 Leuven, Belgium (= address for correspondence)
Tel. 32-16-34.49.99
Fax. 32-16-34.49.20
e-mail :
Summary
This paper presents a project with as main objective the exploration of the Belgian Minimal Clinical Data in an epidemiological context. The paper describes the methodology used to produce maps of the country, for selected medical domains. These maps visualise the geographical variation of pathology treated in the Belgian hospitals. In some cases the variation can be explained by differences in prevalence of the specific pathology. In other cases probably other factors influence the data. The paper presents and discusses a few examples of different elaborated themes.
In conclusion, the Minimal Clinical Data can be used for epidemiological purposes, if potential influencing factors are taken into account. They should be used together with other data sources as the general medical record and specific health surveys. In this way, they can help to complete the global view of the health status of the Belgian population. In addition, analyses done for this project can be taken as starting point for further investigations in the domains of registration audit, quality of care and evaluation of medical practice.
Objectives
This paper presents a project ‘MKG in Beeld’ (in French ‘RCM en images’), which could be translated as ‘Images of Minimal Clinical Data’.
The objectives of the project were the following. The first goal was an exploration of the possibilities of using the Belgian Minimal Clinical Data (MCD)1 for epidemiological purposes. Further, the result of the project should be a publication containing maps, presenting the Minimal Clinical Data to a broad public and not as an instrument to refine the hospital financing system. Further validation of the data, by analysing different specific medical domains, is a third objective.
Methods
From the beginning of the project it was obvious that, working with a hospital stay registration system, has some inherent limitations when used in an epidemiological context. Incidence or prevalence figures are not available. Indeed, the Belgian hospital registration system1 is based on care supplied in general hospitals. Morbidity present in the population, but not treated, or treated in another setting (consultations, psychiatric hospitals, …) is obviously missed. In addition, the registration unit is the hospital stay, and not the patient. Despite these limitations, the number of hospital stays for a specific pathology, generated by the population of a certain geographical area, will tell something about the prevalence of that pathology in that specific area. Without a doubt, other factors can influence the data, as we will see in the example of appendectomy below.
The methodology used to produce the maps is the same throughout the publication. The geographical level is the administrative “arrondissement”. This is similar to a ‘district’, this term will be used in the rest of the text. The basis to colour an area on a map is the city where the patient lives and not the place of treatment. Thus a patient living in Antwerp but treated in Leuven is counted for Antwerp and not for Leuven.
After the selection of hospital stays with a specific pathology, the data are standardised by age and sex. The methodology used here is the indirect standardisation2. The reference is the number of hospital stays for the specific pathology in the Belgian population. For each administrative district a SAR is calculated (SAR = standardised admission ratio) as
Observed number of stays * 100
Expected number of stays
The value of the SAR is the basis to colour a district on the map. There are 6 distinct classes. Districts in class 1, with a SAR value lower than 80, will be coloured white on the maps; districts in class 6, with a SAR greater than 120 will be coloured dark green.
Because a certain selection often results in a rather small amount of stays, a 95% confidence interval is calculated, to indicate whether the value of the SAR is significantly different from 100. If this is the case, the district gets a small yellow star on the map.
Results
The first result of the project is a publication in book form treating the national Minimal Clinical Data of the year 19963.
The publication is divided in several parts. Each part addresses another aggregation level of the data and has several themes: they are listed below.
Part I:1)Number of hospital stays; 2) Number of days in hospital; 3) Hospital-based outpatient care; 4) Hospital mortality
Part II:1)MDC 15 “New-borns and other neonates”; 2) MDC 24 “HIV-infections”; 3) MDC 25 “Multiple significant trauma”
Part III:1)APR-DRG 194 “Heart Failure”, 2) APR-DRG 225 “Appendectomy”
Part IV:1)Tuberculosis; 2) Malignant neoplasm of trachea, bronchus ad lung; 3) Diabetes mellitus; 4) Ischemic heart disease
Part V:1)Transurethral prostatectomy; 2) Hysterectomy
Each theme contains more or less the same components: a map of Belgium, a table presenting all the data used to produce the map, a sex and age distribution, and some additional relevant data available in the database. This last item differs for each specific theme.
Today, 10 of the 15 themes of version 1996 are updated for the registration year 1998. Probably the results for 1998 will be published on an Internet site, together with the results for 1996.
Two examples were chosen to illustrate the method. Both appear in the 1996 version3 and in the 1998 version.
Examples
This map visualises the hospital stays with a malignant neoplasm of trachea, bronchus and lung as principal diagnosis. It shows that in two areas, much more admissions are registered with this type of pathology than expected. In these areas, a lot of people have worked in coal mines for many years. This example shows that, despite the limitations of the methodology, a link between the class of the district and the prevalence of the pathology can be demonstrated.
Another example shows, however, that also other factors can influence the data.
The following map is created based on the selection of hospital stays assigned to the APR-DRG 225 “Appendectomy”. The DRG-level guarantees the selection of a homogenous group of hospital stays where the appendectomy will be, in most cases the only, and in all cases the most important procedure performed4.
For the registration year 1998, in Belgium, 14855 hospital stays were assigned to APR-DRG 225.
The map presenting the hospital stays assigned to APR-DRG 225 “Appendectomy” illustrates the existing variation within a small country as Belgium. Compared to the same map for 19963, a small levelling can be found. In 1996 there were even less districts in the more “neutral classes” 3 and 4 (14/43 while 19/43 in 1998). The global view however is the same.
The sex and age distribution shows that more than 80% of the appendectomies are performed for patients between 6 and 40 years. The procedure is somewhat more often performed on women (52%) than on men (48%). The top of the distribution for men is seen around the age of 10 years. For women, appendectomies are more often performed on young fertile women. Probably the differential diagnosis with
extra-uterine pregnancy or ovary problems is a greater incentive to explore the female abdomen. An American study5 suggests that,although the life time risk of appendicitis is higher for men than for women, the life time risk of appendectomy is almost twice as high for women than for men.
Selections on APR-DRG-level have the advantage that the distribution of the admissions over the four subclasses of severity of illness4 can produce additional interesting information.
The table above shows that more than 75% of all hospital stays in APR-DRG 225 are in the subclass ‘minor severity of illness’, and this is exactly as expected. We are talking about a young patient population. The principal diagnosis is in most cases a simple acute appendicitis. Only one of five admissions in this subclass has an additional secondary diagnosis. The mean length of stay for these admissions is 3,7 days. The figures for the other subclasses clearly illustrate the importance of secondary diagnoses for the determination of the severity of illness.
For most Belgian districts, the distribution of the admissions in APR-DRG 225, over the four subclasses of severity of illness, is more or less the same as for Belgium as a whole. Some districts however have a somewhat different distribution. The combination of data about
1)the distribution of the admissions over the four subclasses of severity of illness,
2)the mean length of stay per severity of illness subclass, and
3)the class of the district on the map,
allows in some cases to formulate hypotheses about under- or overregistration, and/or more or less selective therapeutic attitude of clinicians.
A few examples of these hypotheses are described below.
In most districts of the province West-Vlaanderen, the percentage appendectomies with a minor severity of illness is greater than the national reference. On the map, most districts of the province West-Vlaanderen are darkly coloured, which means that more appendectomies than expected are performed. The combination of this information could lead to the hypothesis that, in this area, clinicians tend to be less selective and quickly go over to an intervention. The result should be a rather short length of stay for the patients with a minor severity of illness. For example, in the districts Brugge and Ieper the mean length of stay for patients with a minor severity of illness is 3,4 and 2,9 days, respectively.
In the district Eeklo, there are rather few admissions with a minor severity of illness (63.6%) and a lot of admissions with a major severity of illness (12.1%). The conclusion should be that there are a lot of very sick appendectomy patients in Eeklo. The mean length of stay, however, of these patients with a major severity of illness is much shorter than the national mean length of stay for this subclass (6.1 days in Eeklo, whereas the national mean length of stay is 11.8 days). The mean length of stay for patients with a major severity of illness inEeklo comes close to the national mean length of stay for patients with a moderate severity of illness. The hypothesis of over-registration of secondary diagnoses can be put forth. One should be aware of the fact, however, that only about 100 admissions are taken into account here. On the other hand, the pattern found here is exactly the same as that found for the data of 1996, with one difference namely in 1998 a shift from more ‘moderate’ admissions to more ‘major’ admissions is found.
The example of APR-DRG 225 “Appendectomy’ has shown that the basic assumption that the number of hospital stays with an appendectomy reflects the incidence of acute appendicitis probably is not valid. Other factors such as a more or less therapeutically selective attitude of the clinicians can influence the data. The evaluation of the attitude of the clinicians is possible here. The reason is that in almost 80% of the cases, the patient is treated in the district he lives. Indeed, people would not prefer to go to a long distance reference centre for a basic pathology as appendicitis, which is, in addition, a typical emergency.
Conclusions
In conclusion, we will retake the objectives of the project and evaluate whether the different goals were attained.
The project explored the use of the Belgian Minimal Clinical Data for epidemiological purposes. The limitations were well described. They could only be, partly, overcome when the Minimal Clinical Data could use a unique, anonymous national patient number, which is not the case at the moment. For some medical domains it was possible to illustrate a relationship between the prevalence of the pathology and the number of admissions for that pathology (e.g. tuberculosis, HIV, respiratory neoplasm3). For other medical domains, it was shown that probably other factors influence the data (appendectomy, prostatectomy, hysterectomy3). One huge advantage in using the Minimal Clinical Data for such projects is the fact that they are available anyway. No extra data collection is needed. In Belgium, at this moment, the hospital data of almost ten years are accessible in an immense, exhaustive database.
The project definitely succeeded in the demonstration of existing geographical variation in the number of hospital stays with a specific pathology. Some differences could be an incentive for further investigation, for example in the domains of registration audit, quality of care and evaluation of medical practice. In conclusion, the Minimal Clinical Data can be used for epidemiological purposes, if potential influencing factors are taken into account. They should be used together with other data sources as the general medical record and specific health surveys performed in our country6. In this way, they can help to complete the global view of the health status of the Belgian population.
For the year 1996, the project result is a publication in book form3, which was distributed, for free, in the Belgian health care sector. This publication includes an extensive description of the history, contents and organisation of the registration of the Minimal Clinical Data. In each theme, relevant data available in the MCD are presented (mostly in graph, sometimes in tables) to further elaborate the topic. Examples of such ‘additional data’ are external causes of injury (E-codes – theme ‘Multipletrauma’), patient nationality (tuberculosis), length of stay, severity of illness, risk of mortality and effective deaths (selections on DRG-level), and of course principal and secondary diagnoses.
For the year 1998 the goal is a publication on a web site of the Ministry of Public Health, together with the data for 1996. In this way an even broader distribution of the results would be achieved.
The project included further validation of the data, internal as well as external. Internal validation was done by the check between ICD-9-CM procedure codes and ‘RIZV-codes’, both available in the Belgian MCD, for the surgical themes concerned. ‘RIZIV-codes’ are the codes for the national public health insurance on which reimbursement is based. In addition, the data of 1996 and 1998 could be compared, which allowed to describe trends as well. External validation was performed by checking data registered by other institutions, as for example the national institute for public health insurance, and specific scientific organisations7-8. No significant inconsistencies could be demonstrated, which resulted in an increased confidence in the validity of the Minimal Clinical Data.
References
- Richtlijnen voor de registratie van de Minimale Klinische Gegevens (M.K.G.). Nieuw concept. Ministerie van Sociale Zaken, Volksgezondheid en Leefmilieu, Bestuur van de gezondheidszorg, Bestuursdirectie gezondheidszorgbeleid. Juni 1999.
Directions d’enregistrement du Résumé Clinique Minimum (R.C.M.). Nouveau concept. Ministère des Affaires Sociales, de la Santé publique et de l’environnement. Administration des Soins de Santé. Direction de la politique des soins de santé. Juni 1999.
- Rothman, Kenneth, J., Modern Epidemiology, Little Brown, Boston (Mass), 1986, 358p.
- MKG 1996 IN BEELD. Geografische variatie van de pathologie in de Belgische ziekenhuizen. Ministerie van Sociale Zaken, Volksgezondheid en Leefmilieu, Bestuur van de Gezondheidszorg, Bestuursdirectie Gezondheidszorgbeleid.
RCM 1996 EN IMAGES. Variation geographique de la pathologie dans les hôpitaux belges. Ministère des Affaires Sociales, de la Santé publique et de l’environnement. Administration des Soins de Santé. Direction de la politique des soins de santé.
- APR-DRGs. All Patient Refined Diagnosis Related Groups. Definitions Manual Version 15.0. 3M Health Information Systems.
- Addiss DG, Shaffer N, Fowler BS, Tauxe RV. – The epidemiology of appendicitis and appendectomy in the United States. Am J Epidemiol 1990 Nov, 132(5): 910-925.
- Wetenschappelijk Instituut Volksgezondheid-Louis Pasteur, Afdeling Epidemiologie. Gezondheidsenquete door interview.
- Wetenschappelijk Instituut Volksgezondheid-Louis Pasteur, Afdeling Epidemiologie. AIDS in België: semesteriële rapporten.
8.Vlaamse vereniging voor respiratoire gezondheidszorg en tuberculosebestrijding (VRGT) VZW. Tuberculose-indidentieregister België 1998-1999.