Introduction

In December 1999, an Act of Parliament changed the German health system fundamentally [1]. The decision was made that nearly all inpatient treatments should be paid on an all-inclusive base, except for psychiatry and rehabilitation. For this change in health care payment a system of German Diagnosis Related Groups (G-DRG) was implemented and a refinement process was started. The first German version of this new system used the state of art Australian Refined Diagnosis Related Groups. Nowadays, there are 1,027 G-DRGs, i.e., codes, which categorize similar expensive hospitalizations [2]. The adaptation and development of this G-DRG system are still ongoing and conducted by an independent and special institute (called Institut für das Entgeltsystem im Krankenhaus-InEK). Foundation for this ongoing process of adjustment is a legally defined data set (so called §21 data set) which has to be delivered to this institute by each hospital, electronically every year.

This data set has, for every case, parameters like birth weight in case of newborn, duration of artificial respiration, duration of hospitalization, age, kind of discharge, of course treated diagnoses in ICD-10 and major procedures performed according to ICPM (international classification of procedures in medicine) in the modified versions for Germany. This information has to be collected and ciphered in hospital information systems by every responsible treating physician, according to the yearly updated coding rules. The accuracy of coding will be controlled systematically for validity, at least by health insurances and is of financial relevance because it triggers the kind of DRG and hence the payment of bill. Everyone can imagine that this new field of duty is not very popular but it takes a major part of the working hours of physicians. The reason for hospitalization has to be coded as main diagnosis. The main diagnosis categorizes into major diagnostic categories, which defines the first alphabetic character of the four-digit G-DRG code. The next two numeric digits define the disease group and the last alphabetic code particularize sometimes the magnitude. The so achieved final G-DRG code will be used to generate an algorithm using the ciphered data set with a special, annually certified software, called “grouper”.

Since 2004, the usage of this G-DRGs is mandatory for payment. But also in the years before 2004, the delivery to the InEK of a precisely defined data set (§21-data set) was stipulated.

On the one hand the enormous amount of collected and anonymized data in these §21-data sets contain sensitive information. Anyone who gain insight in a single hospital dataset knows not only the patient structure, but as well the financial foundation or major income of this specific hospital. On the other hand there is a request for some way of transparency for patients and other health care business partners. Hence some accumulated information is published freely, accessible on the homepage of the InEK institute: http://g-drg.de.

Of course the question arises if this information can be used for purposes other than economical questions. A method to guess treatment quality is already published [3]. In contrast to this complete data pool existing in Germany at present, the two or three existing regional registries for congenital malformations are covering only about 0.5% of the German population. Our research is to estimate the true prevalence of anorectal malformation in Germany, treatment for which is done by more than 80 pediatric surgeries [4].

Search in literature shows the prevalence rates for anorectal malformation ranges between 1:2,000 and 1:5,000. An overview about population prevalence studies can be found elsewhere [5, 10]. Exemplarily, results of further studies are presented in Table 1.

Table 1 Current ARM prevalence estimates
Full size table

But it must be considered that even in the largest collection of 33 European registries (EUROCAT), the estimated rates differ not only for region but also for observed time frame [11]: 4.0(EUROCAT 1980–1994), 3.1(EUROCAT 1987–1999) and 2.7(EUROCAT 1994–2003). In contrast to these, Orphanet reports a rate of 2.4 for anorectal malformation[12]. In EUROCAT, there are two German registries with very divergent rates (Mainz: 6.0 and Saxony–Anhalt: 2.2[10]).

The German self-help organization for anorectal malformation covers currently about 500–600 families [13]. Therefore, self-help wanted to know the total number of anorectal malformations requiring treatment in whole Germany.

Materials and methods

Even it could be regarded as an incidence, we use the term prevalence. This is in accordance with a suggestion to distinguish in birth defects between genetic prenatal incidence and surviving postnatal prevalence [10].

We present a new method to estimate the prevalence of rare congenital diseases using hospital discharge data published on http://g-drg.de. Since 2002, every year a MS® Excel or Access sheet for accompanying research is made available according to the German hospital law (§17b paragraph 8 hospital law) summarizing all delivered data sets according to §21 hospital paying law. These freely available datasets have anonymized and aggregated information and therefore restricted in some way. Inference on single hospital or regions and on special subgroups is not possible. But due to the manner of DRG system, some conclusions are possible. Main diagnosis of a hospitalization must be ciphered according to ICD-10. Hence physicians must mention, as main diagnosis of every ARM hospitalization, any of these ICD-10 codes: Q42 (congenital absence, atresia and stenosis of large intestine) or some subgroups of Q43 (other congenital malformations of intestine, e.g., cloaca, ectopic anus).

With this knowledge, multiple hospitalizations of the same patient cannot be excluded. But another information must be ciphered. According to the German modification of WHO ICPM (international classification of procedures in medicine) or ICHI (international classification of health interventions), 5-codes are relevant surgical interventions in ARM (Table 2). An initial ARM surgery must be coded with 5-495, 5-496 or 5-497. The other two codes describe other or further actions. These procedures together with a main diagnosis of Q42 or some subgroups of Q43 trigger in 2005, the DRG G11Z (pyloromyotomy or anoproctoplasty and reconstruction of anus and sphincter) or in the first month of life, the DRG P06C. If no procedure code is mentioned, then the DRG G71Z (other modest severe diseases of intestines) is assumed. Because of the DRG terms change every year and do not fit exactly to a disease, they are of economical but not epidemiological value. The trigger algorithm for each DRG is likewise available [14] and is implemented in the data collection software, called “grouper”.

Table 2 ARM procedures given in Germany
Full size table

Therefore, we use for yearly German prevalence estimation of ARM, the three primary intervention codes and only secondary main diagnoses. Our precondition or assumption is that an ARM surgery is mandatory and each initial ARM surgery occurs during the first year of life. To get a denominator, yearly birth rates can be collected from the German Federal Statistical Office homepage [15]. As further constrictive trick we ignore the little influence of time lag between birth and surgery. Thus, we can use numerator and denominator of known calendar year. Currently, complete data sets are available for 2002–2005. Considering the sampling error, the true value will be estimated appropriate to an exact 95% confidence interval [16, 17].

Results

In Table 2 we present the absolute frequency of anorectal surgeries in Germany. These numbers can be extracted from data sets published at http://g-drg.de. Bold numbers refer to initial ARM surgeries. These frequencies are used for prevalence estimation of ARM in Table 3.

Table 3 ARM prevalence estimation with G-DRG data information
Full size table

Summarizing the information of available data (see Table 3), in total 1,012 newborn patients required a lifesaving surgery in the first year of life between 2002 and 2005. Knowing that in the same time period 2,817,388 babies were born, we can estimate the national wide prevalence rate to be 3.59 (95%CI: 3.37; 3.82) per 10,000. Expressed in an other way, a child requiring a pull-through surgery for imperforated anus occurs in Germany for every 2,784 (95%CI: 2,621; 2,966) delivered child.

Further interesting information is that in 2005 the most specific procedure 5-495 was given by 71 hospitals, which means that nearly all German paediatric surgeries perform pull-through surgeries, on an average about three to four each year.

Discussion

Informations on DRG or ICD-10 diagnoses are less useful, because they are aggregated and hence not an identifying feature. Diagnoses group Q43 (other congenital malformations of intestine) include beside cloaca or ectopic anus and also Meckel’s diverticulum or Hirschsprungs disease, etc. But available data sets have no information about subgroups. Main diagnoses group Q42 (congenital absence, atresia and stenosis of large intestine) is more specific for imperforated anus and given about twice of the procedure rate, which we used for prevalence estimation. This is plausible, because only one pull-through procedure is necessary but at least two hospitalizations during first year of life.

The rate of other procedure codes, e.g., colostoma could be used as further information to narrow the true prevalence. But there are other diseases in which a colostoma is necessary and furthermore, there are seven subgroups of colostoma surgery in which they will be classified (5-460 to 5-466).

This estimation of ARM by means of economical data sets, which have been collected for development and research of G-DRG system, is successful due to unique surgical procedures for pull-through surgeries: 5-495 and in the broader sense 5-496 and 5-497. Another happenstance is that ARM cases are treated usually only once and that too during first year of life. These preconditions are at the same time limitations of the presented method.

According to presented data, it is evident that in 10–13%, a pull-through surgery seems to be performed between 1 and 5 years of age and in 2–4% during the age of 5–10 years. This may explain to some extent the slight variability in yearly prevalence rates. Considering only the first year of life may be accounted as a conservative prevalence estimation. Similarly in another prevalence study estimating birth defects using US hospital discharges, they found a tendency towards underestimation [18]. But the estimated prevalence of 3.6 per 10,000 newborns is comparable to existing prevalence studies [5–12].

Overestimation of prevalence is unlikely with the presented approach due to the fact that all §21- datasets are checked in an accurate controlling system. All health insurances ensure that no procedures are coded and hence have to be paid, which do not have a documentation that they have actual taken place.

Considering the requisite attributes [19] for surveillance systems: (1) timeliness, (2) completeness, (3) accuracy, (4) specificity, and (5) population-based ascertainment, these can be fulfilled mostly with the collected economic data sets of German diagnostic group system.

We know that the idea to use G-DRG data for additional purposes is not new [3], but we think that this rough prevalence estimation is a smart and useful approach, even if it has limitation of having only aggregated available data. But if the national institute (InEK) with “acta iure imperii” would analyse in-house the delivered §21 data sets under epidemiological perspective, the confidentiality could be maintained and more knowledge gained.

Conclusion

This kind of estimation cannot substitute an active surveillance system but can give a global prevalence estimates of treated babies born with malformed anus. For more specific epidemiological questions, singular instead of aggregated information is a must. Of course, there are limitations and presumptions. But due to specific surgical procedures under some conditions, yearly prevalence of some congenital diseases can be estimated for Germany. Further calculations for other epidemiological questions should be addressed.

We showed especially for anorectal malformations a rough prevalence estimation using freely available economic data of the G-DRG system. Interesting is the stability of case rates and hence prevalence of about one necessary pull-through procedure for every 2,500th–3,000th newborn. Due to its larger size it may be an useful supplement to smaller regional registries.