As an answer to our research question we found a strong correlation between age, gender, multimorbidity and health care utilization. These findings are not surprising and do not stand in contrast to comparable findings in the international scientific literature [1–4, 13]. Nonetheless, our study is the first approach to this phenomenon in our country with an international classification developed for primary care.
Generally, when addressing multimorbidity issues in order to compare the results of different studies, possible differences concerning the research question, the data sources and the definition of multimorbidity have to be taken into account. Fortin et al. collected prevalence estimations of multimorbidity in Europe, the Middle East, the United States, and Canada. Since research questions, information collection, and multimorbidity measures differed, major differences in the results were observed [14]. However, there is a broad consensus that multimorbidity and its high prevalence is an important issue in family practice that deserves more scientific research [15].
Van den Aker et al. concluded that multimorbidity, although it increases with age, is a frequent phenomenon among all ages [1]. This phenomenon was also observed within our study sample. 12.8% of the patients younger than 50 years featured 2 or more chronic conditions. Therefore, research into multimorbidity should not only focus on the elderly, who are especially at risk.
Multimorbidity as defined by routinely collected data in electronic patient records not only offers an epidemiological overview of morbidity patterns for the scientist, but can also help the GP to identify patients with an increased likelihood of needing more attention [16]. We observed a typical clustering of specific health problems (e.g. diabetes, hypercholesterinemia and hypertension, Table 3). These clusters can be easily identified by the GP on the basis of the EPR in order to apply an appropriate medical care and to initiate specific interventions (e.g. lifestyle modifications).
Limitations
Generally, a potential selection bias must be admitted since the GPs' participation is voluntary and by now mainly focuses on Southwest Germany. Moreover, the number of 17 practices is still too small to draw strong conclusions.
In order to assess morbidity, there are several detailed and validated morbidity indexes [17]. For example, the "Cumulative Illness Rating Scale" (CIRS) [18] index additionally regards the severity of each condition and was also validated for the use to quantify multimorbidity for primary care patients [19]. However, since we had no information of the condition severity within the CONTENT EPR, we could not calculate this index for our study and had to limit on disease counts. Moreover, it would have been challenging to analyse the influence of sociodemografic factors (e.g. education, profession, income) on multimorbidity. However, sociodemografic information was only available for a small fraction of the sample.
The definition of a specific chronic condition on the basis of ICPC codes is often ambiguous. For example, we defined Osteoarthrosis (OA) by inclusion of ICPC codes L89 (Osteoarthrosis of hip), L90 (Osteoarthrosis of knee) and L91 (Osteoarthrosis, other). L84 (Back syndrome without radiating pain) is not included in our selection but also includes OA of the back. However, L84 also includes diseases that are not related to OA (e.g. back strain). Moreover, L91 includes 'arthritis unspecified' and 'traumatic arthropathy' that are not directly related to OA. This general problem could be solved by using a more specific terminology level which would allow grouping of all osteoarthrosis (no matter the site) from all applicable ICPC-2 codes.
Strengths
As mentioned above, the CONTENT project is the first approach in Germany based on episodes of care and ICPC that facilitates detailed long term analyses of co- and multimorbidity.
Especially, the continuous registration of patients' presented symptoms is new in comparison to hitherto existing German EPRs. Thus, CONTENT data enable to analyse the correlation between presented symptoms and resulting diagnoses in consideration of existing comorbidities. Moreover, age, gender as well as seasonal and regional differences have to be taken into account. In the long run, for every ICPC symptom (SY) it will be possible to determine a list L of resulting diagnoses D
1,....., D
n
and corresponding probabilities P
1,....., P
n
taking into account the above mentioned constraints (A: age, G: gender, S: season, R: region, C
1,....., C
m
: existing comorbidities), as the following formal description shows:
This detailed model represents an extension of the model presented by Lamberts et al. [20].