Health care use and costs of adverse drug events emerging from outpatient treatment in Germany: A modelling approach

Previous studies for Germany have reported the percent of hospitalisations due to ADRs and the associated costs for the hospital admission [7, 8]. However, the burden of morbidity, mortality and costs of ADEs originating and occurring in ambulatory medical practice remains unknown. It was our objective to estimate morbidity and costs associated with adverse drug events occurring in the ambulatory setting in Germany. To estimate the frequency and costs of ADEs occurring in the context of outpatient drug treatment, the drug-related mortality and morbidity cost-of-illness model developed for the US was adapted to incorporate available German data. Sensitivity analyses were performed using rates of ADEs and healthcare utilisation found in the literature.

Parts of the drug-related morbidity and mortality cost-of-illness-model [15, 16] were used to simulate the morbidity and costs of ADEs in Germany. This model describes outcomes and costs associated with drug-related problems (DRPs) from the perspective of third party payers but it excludes illicit drug use and DRPs originating in institutional settings. In the model, drug related problems are divided into three mutually exclusive sequences of events: 1) treatment failure 2) new medical problem (NMP) 3) treatment failure and new medical problem (TF/NMP). The structure of the model is shown in Figure 1. The probability of these sequences of events, or that 'no DRP occurs' and the conditional probabilities of healthcare utilisation if a DRP developed were estimated by a panel of clinical experts for the USA [15, 16]. In the model, the healthcare utilisation endpoint is mostly mutually exclusive, i.e. if the patient is hospitalised, costs for previous consultations to the physician for the DRP are not considered and for long-term care, no prior hospitalisation or physician visit is included. This results in a systematic downward bias in the costing of these healthcare resources. In the case of death, the cost of prior hospitalisation is included as a cost of death. However, one can imagine that some patients with a severe ADE will die before reaching the hospital. Since the aim of this study is to estimate the frequency and costs of ADEs and the study used as a source of ADEs [8] did not document treatment failures, only the sequences of events for NMP and TF/NMP are included in the present calculations, the part of the model enclosed by a circle in Figure 1.

In the paper by Ernst et al. [16], the number of ADEs is calculated based on a cohort of patients making a total number of 734,493,000 physician visits. Since 90% of the German population is insured with social health insurance, healthcare facilities are used more frequently than in the USA. There were 7.4 physician visits per capita in Germany in 2006 compared to 3.8 in the USA according to OECD [17]. Similarly, prescription practices between Germany and the USA may differ considerably. However, it is conceivable that the likelihood of requiring healthcare services, such as hospitalisation, will be similar between the Germany and the USA if an ADE occurs. Thus, only the event probabilities after the node "drug prescribed" will be used for the present calculations. Our final probabilities are the result of multiplying the first node (NMP or TF/NMP) with the probability of resource utilisation. For example, in the paper by Ernst et al. [16], the final probability of "no therapy" for a "new medical problem" was 0.633 × 0.105 × 0.146 = 0.010. In our paper, we did not include the probability of drug therapy in the calculation (0.633), so that the probability of "No therapy" for a "new medical problem" is 0.0153. Our adapted probabilities are shown in Figure 1. As hospitalisations are included in the model as hospitalisations and death, the total number of hospitalisations is divided between these categories, 97.9% as hospitalisations and 2.1% as death. The probabilities of the types of resource utilisation (physician visit, additional treatment, emergency department visit, hospitalisation, long-term care and death) of the 2 model arms, as shown in Figure 1 were added together to obtain overall probabilities for the various types of resource utilisation for ADEs. Overall probabilities of resource use for ADEs were that 12% of medication users will require a physician visit (4% only consultation and 8% additional treatment), 2% an emergency department visit, 1% a hospitalisation, 0.3% long-term care and 0.02% will die.

To calculate the hospital costs of ADEs in the year 2007, the adverse drug reactions described in the Schneeweiss paper [8] were assigned possible ICD 10 codes (3 positions). Using data from the Federal Office of Statistics for patients over 20 years of age, the sum of hospital days reported for the possible ICD codes for each described adverse drug reaction was divided by the total number of patients with the ICD code to calculate an average length of stay. The calculated average length of stay was multiplied with the average cost per hospital day for general hospitals, 437€ [18]. The ICD codes of the adverse drug reactions, the number of patients per ICD code, number of hospital days per ICD code and the average cost per admission per adverse drug reaction is shown in Additional File 1. The final cost of hospital stay was calculated as the average cost of the described adverse drug reaction weighted according to its prevalence as reported by Schneeweiss [8]. The assignment of adverse drug reactions, their proportion in each drug class and the calculated weights and total costs are shown in Additional File 2. The final calculated cost of hospital stay (3,452.11€) was the weighted average cost of the described adverse drug reactions. Hospital admission costs are also included in the costs of death. The costs of other resource utilisation were determined as follows. The cost of an emergency department visit was assumed to equal the cost of one day in hospital (437€) [18]. The average cost of a prescribed medication for the German statutory health insurance system, which covers about 90% of the resident population, was 42.62 Euro in 2007 [19]. The cost of a physician visit was calculated based on the average contact value of a general practitioner and specialists of internal medicine with and without family practice responsibilities [20] in 1999 (16.20€ in 1999). The contact value was extrapolated to 2007 using price increases in physician reimbursement per case [21], which averaged 4.4% annually between 1999 and 2007. The value for 2007 was calculated to be 22.62 Euro per visit. Average costs for long-term care [22] in 2007 were calculated by dividing total expenditures for long term care in the year 2007 [23] (18.34 billion Euros) by the total number of recipients receiving benefits [24] (2.03 million). This average includes persons receiving benefits for the whole year or less, but it was assumed that the majority of persons would receive benefits for the whole year. This value was divided by 12 to obtain the monthly rate. The time points at which persons started to require long-term care for an ADE were assumed to be equally distributed across the year and among the persons requiring long-term care. Since expenditures were not reported according to age-groups, the overall average was used, which amounted to 4,890 Euro. Multiple visits to any healthcare facility were not accounted for since these data were not available for Germany. This cost calculation aimed to determine the excess costs of ADEs and thus the costs of initial treatment were not included, since they occur regardless of a ADE.

Since the only available data for Germany is the percent of hospitalisations related to adverse drug events [7, 8], the model was adapted to be based on this data. The relationship between the probabilities is used to calculate resource utilisation from the number of ADEs requiring hospitalisation (ie. 4 times as many patients require only a physician visit compared to those requiring hospitalisation, the calculated factors are shown in Table 1). The number of hospitalisations due to ADEs were calculated based on the Schneeweiss study [8], whereas the Dormann study [7] was used in the sensitivity analysis. Schneeweiss reported that 2.4% of all hospital admissions to medical wards over 30 months in 2 German cities (total of 3 hospitals) were drug-related leading to an estimate of the expected number of drug-related admissions of adults of 139,405 based on the published number of admissions to medical wards [25] (5,808,544 admissions in 2007) for the ages 20 and older. All other health resource use was calculated in relationship to the number of drug-related hospitalisations.

In order to test estimates based on the model, MEDLINE was searched for studies reporting the frequency of ADEs in adults in the ambulatory setting published after 1990. Studies including chemotherapeutic agents or only certain diseases or specific ADEs were excluded. The studies found in this search are listed and summarized in Additional File 3. Rates of healthcare resource use reported by the studies for outpatient services (e.g. general practitioner, clinic), use of the emergency department, and hospitalisations are shown in Table 2. The table shows the reported rates of healthcare use and adjusted rates of healthcare use according to the reported population taking medications (see also Additional File 3), since prescription rates may vary between countries.

The following sensitivity analyses were performed. To test the effect of a higher rate of drug-related hospitalisations, the complete model was calculated based on the rate reported by Dormann [7]. Other one-way sensitivity analyses examined upper and lower rates of specific resource use (physician visits, emergency department visits and hospitalisations) for ADEs, as reported in the studies found (Table 2). In these sensitivity analyses, the absolute numbers determined for base case healthcare utilisation (Schneeweiss study) were applied and only the numbers for the specific resource examined (e.g. physician visits) were changed. The rates of healthcare resource used from the literature review, adjusted according to the population treated with medications in the study (see Table 2) are multiplied with the population treated with medications in Germany. The estimate of the German population taking medications is based on the percent of health insurance company members, according to age, who had a medication prescribed in 2007 [26]. Percentages were adjusted for differences in age and sex-distribution between the health insurance population and the general population [27]. Overall, about 75% of the German population over 20 years of age (53.6 million of 66.2 million persons) take at least one prescribed medication over the year. All calculations were performed using Microsoft Excel.

The base case calculations using German data are shown in Table 1. The base case estimates that 2.14 million adults would develop ADEs during the course of ambulatory medical treatment in 2007. Only 12% of these patients require no extra healthcare services and 70% require only outpatient services such as a physician visit or additional outpatient treatment. In relation to the total German population taking medications, about 4.4% of the population ingesting medications in 2007 would be expected to have an ADE and about 3.8% would be expected to require health services. 3.0% would require ambulatory medical care (physician visit and additional treatment), 0.3% hospitalisation and 0.006% would die.

Expected resource use and cost estimates associated with ADEs are also reported in Table 1 (costs of resource utilisation and sources of costs are also shown in the table). According to the model, ADEs may have accounted for 816 million Euros of health care expenditures in Germany over 1 year, with 80% of expenditures for hospitalisation (also preceding death) and long-term care being generated by 8% of the predicted patients with ADEs. Ambulatory care, without the emergency department, accounts for 9% of all costs but is required by 70% of patients. Since 12% of patients with ADEs generate no costs, the average cost of an ADE requiring health services is 381 Euro if all persons with an ADE are considered and 434 Euro if only those seeking healthcare are considered.

The sensitivity analyses in which the German hospitalisation rate [7] or only the rates of physician visits, emergency department visits or hospitalisation individually are varied, are shown in Table 3. All rates are based on the rates reported in the literature (Table 2) adjusted according to the study population taking medications. The sensitivity analysis (Table 3) shows that the estimates based on the German hospitalisation rate reported by Schneeweiss [8] are much lower than the other estimates of the frequency and costs of ADEs. The calculation using Dormann's hospitalisation rate has a strong effect on both the number of ADEs and the costs postulated since the complete model is recalculated. Since the base case probabilities and resource costs are used, the cost per case is equal to that of Schneeweiss (381€). Dormann's percentage of hospitalisations due to ADRs is still lower than other values found in the literature. Variation of the number of physician visits has a large effect on the number of ADEs but a small effect on the total costs and leads to lower costs per case (upper limit: 117€ per case; lower limit: 358€ per case) than estimated by Schneeweiss. Variation of the number of hospitalisations has the greatest effect on the overall costs of ADEs. The upper limit of the number of hospitalisations estimates total costs which are 9 times higher than in the base case and costs per case which are almost 5 times higher than costs per case of the base case.