Study design

Results of qualitative interviews with FP and their medical assistants were used to help explain the quantitative results (triangulation design [23]). More detailed information on our study design, setting and research instruments are already published elsewhere [24].

This publication focuses on results of subanalysis of research parts (1) and (2) regarding the prescription of PIM by FPs.

Inclusion of study participants

All 33 FPs from academic teaching practices of the Department of General Practice of TU Dresden situated in Dresden were invited for participation in the study. After confirmation of participation by the FPs, all chronic patients of each participating practice were invited for participation if they met following inclusion criteria: a) at least two parallel chronic diseases and b) two prescribed long-term medications at least 6 months before 2012. Furthermore, all medical assistants of the participating FPs were asked to participate in qualitative interviews.

Data collection

Patients’ records [1] were analyzed from March 2013 to May 2014 by our project assistants at FPs’ practices using a semi-standardized entry mask in SPSS 22.0. Patient records of a randomized quarter of the year 2012 were retrospectively analyzed per patient, and information regarding morbidity, admissions/referrals to specialists and non-drug and drug prescriptions (including documented adverse events) were documented. Prescriptions by specialists were only included if FPs had documented them in the patient records based on the findings reports sent by the specialists or direct information by the patients.

We used these real-life case vignettes to get more authentic answers by FPs and as concrete starting points for reflection of the PIM-topic in general. One patient (>65 years) with the most prescription of benzodiazepine receptor agonists (so called Z-drugs) not corresponding to recommendations of PRISCUS per each FP was selected for the respective case vignette (see Table 1).

Analysis

Statistical analysis of data based on record analysis [1] were done using SPSS 22.0 to describe characteristics of the sample of patients > 65 years and especially of the subsample affected by PIM prescription. Screening all documented prescribed drugs, PIM were identified according to the German PRISCUS list. Differences in distributions were examined using chi²-test. T-test for unpaired samples was conducted to examine differences of means between different groups. A logistic regression was conducted to identify predictors for PIM-prescription.

The procedure of qualitative content analysis of interviews with FPs [2] is based on Mayring [25]. Two researchers analyzed the interview manuscripts by using MAXQDA® software to categorize contextual factors and individual reasons for PIM prescription from FPs’ point of view. The contents of the interviews were classified inductively to indicate pertinent categories. The derived categories were discussed and consented in the research team.

All data were pseudonymously recorded and analyzed. For integrating results, interview data (e.g. reported motives of prescription) were triangulated with data based on patient records’ analysis (e.g. relative frequencies of PIM prescription).

Description of participating family physicians

Out of 33 academic teaching physicians of the Department of General Practice of TU Dresden, working in Dresden, 7 (of which 5 were female) participated in this study. The age of the participants was distributed between 43 and 61 years, and professional experiences were reported between 7 and 22 years. Out of 7 FPs, 3 worked at group practices and 4 at single practices, At the FPs’ practices, between 2 and 6 medical assistants were employed.

Of all the 7 FPs’ patients a number of 2826 met the inclusion criteria (≥2 chronic conditions and ≥ 2 related long-term medication), and 1846 patients (65.3 %) gave their consent for participation in the study. The number of participating patients per FP varied between a minimum of 120 and a maximum of 455. 55 % were females, and the age distribution was between 23 and 100 years with a mean age of 69 (±11 SD) years. 69.4 % of the patients were > 65 years old who will be in focus of following analysis.

Based on the small sample of participating FPs, the results of the study are limited regarding representativeness of FPs and patients. To check for potential biases we compared our data with results of comparable studies (external validation, see below).

Description of the sample: elderly patients

Elderly patients [n = 1241, mean age: 76 (±6 SD), females: 56.6 %] were characterized in average by 8.3 [(±3.7 SD] documented chronic diagnosis (ICD 10) that include a total of 543 different ICD-Codes (shortened to first figure).

65.5 % of the elderly patients were given 4 and more drugs in long term medications (M: 6.3 ± 3.3 SD). The number of prescribed drugs correlated not very strongly with the number of chronic diagnosis (r = 0.34/p ≤ 0.001).

Regarding the interpretation of documented chronic diagnoses the limitation of validity based on missing or unspecific diagnoses needs to be considered. Patients’ records consist on information about chronic diagnoses documented primary for billing purposes. Although billed drugs and medical services need to be confirmed by documenting associated diagnoses, mismatches were found in our data. An exemplary analysis regarding congruence of prescriptions of thyroid medications and associated thyroid diagnosis showed insufficiently documented thyroid diagnosis in 26.8 % of patients with diagnosed or treated thyroid diseases [26].

Prescription of potentially inappropriate medication

Out of the 83 PIM in the German PRISCUS list (including inappropriate drugs and in some cases inadequately dosed drugs), 41 PIM were prescribed in our sample. Using the method of record analysis, our data include all of FPs’ documented drug prescriptions, incl. private prescriptions that are not covered by the statutory health insurance respectively over the counter medication. Prescriptions by specialists were only included if FP had been informed and documented it. Since our data are related to FP prescription, the real PIM-prescription in general will be underestimated.

To summarize, 23.9 % of elderly patients received at least one PIM prescription, which is comparable to other studies reporting PIM prescription rates between 22.5 and 28.4 % of elderly patients in the primary care setting [17, 18]. 21.5 % of patients in our study got at least one PIM as medication for chronic conditions and 3.1 % to care for acute symptoms. Sedatives and hypnotics were the most frequent (13.7 %) prescribed PIM-drugs in patients > 65 years (see Fig. 1).

A significantly higher number of female (F: 27.6 % vs. M: 19.1 %, χ ² = 12.0/p ≤ 0.001) and very old patients (≥80 years: 31.3 % vs. < 80 years: 21.6 %, χ ² = 11.7/p ≤ 0.001) received a PIM prescription. Patients with at least one PIM prescription were characterized by a significantly higher number of chronic diagnoses (M 9.3 ± 3.9 SD) and long-term medication (7.5 ± 3.7 SD) compared to patients without a PIM prescription (8.0 ± 3.6 SD/t-test: p ≤ 0.001 respectively 5.4 ± 3.0 SD, t-test: p ≤ 0.001).

Influencing factors on prescription of PIM

Conducting a logistic regression, we examined the influence of binary associated factors on prescription of PIM in a multivariate model. Including gender, age, documented diagnose of a mental disorder (ICD-10: F), number of diagnoses and number of long-term medication in the regression model, the strongest effect was detected for documented mental disorders (Exp(B) = 2.0/p ≤ 0.001), followed by female gender (Exp(B) = 1.6/p ≤ 0.01) and number of long-term medication (Exp(B) = 1.2/p ≤ 0.001) as predictors for the probability of a PIM prescription. Age and number of diagnoses were not predictive for PIM prescription. The three mentioned predictors explained just 14.9 % of the variance (Nagelkerkes R² = 0.149) of the dependent variable. However, there will be other influencing variables we did not include in the model. Our findings are comparable to other studies that found depression as one of the common mental disorders in elderly patients as main predictor (OR = 2.42) for PIM prescription, followed by the number of prescribed drugs (OR = 1.99) [18]. A positive correlation between the number of prescribed drugs and the risk of PIM in the elderly was also shown in the study of Schubert [19].

Knowledge and application of PRISCUS-list

The inherent problem of lack of time (the average duration of consultation at FPs’ practices is ca. 8 min per patient in Germany [27]) was mentioned by most of the FPs. This finding is comparable to other studies that also pointed out lack of time as a main barrier for appropriate prescribing [20, 21] or for implementing recommendations of guidelines in daily practice [28].

Several FPs were doubtful regarding the evidence of PRISCUS list since most of pharmacological studies exclude elderly patients, and therefore related evidence is missing. Other studies also showed partly existing skeptical attitudes of physicians regarding cited evidence in guidelines [29]. Two FPs criticized that the PRISCUS list does not recommend “practicable and pharmacologically based alternatives” [FP6]. For instance, regarding the prescription of sedatives/hypnotics, FPs reported great uncertainty (see below).

One FP argued that primary care physicians mostly act intuition- and experience-based. From his point of view, evidence-based practice that includes implementation of guidelines in daily practice would not be typical for FPs’ practice. Triangulating these interview data with related data of patient records’ analysis, we detected that this intuition- and experience-oriented FP had a visibly lower prescription rate of PIM while reporting the highest proportion of patients > 65 years. Nevertheless, this single example might not be representative for all FPs. However, intuition-based and evidence-based acting do not exclude each other in the dichotomous way of a “zero-sum relationship” [30].

Social desirability was assumed in some FPs who reported knowing PRISCUS. Asked for their concrete criticisms they first reported weakness of PRISCUS list in general, and, after repeating the question, they apologized for not remembering since they do not use the list. There were two FPs knowing nothing of PRISCUS and PIM at all. This is similar to results of other studies describing educational lacks in primary care as well as hospital physicians, regarding appropriate prescribing in elderly patients [20, 21].

Reported causes for prescription of potentially inappropriate medication

This finding is convenient to our previous result of the logistic regression (see above), showing that a documented mental disorder was the strongest predictor for prescription of PIM.

Most of the FPs described the importance of close cooperation with neurologists/psychiatrists regarding care for patients addicted to hypnotics, but several problems were pointed out. First, they reported patients’ difficulties to get access to a psychiatrist for “just” sleeping disorders and second, they reported communication lacks regarding prescriptions of medications and lacks in referral of clinical information. And third, some of the FPs pointed out an uncritical prescribing by psychiatrists without looking at the patient in their entirety including existing multimorbidity. But especially in the case of out-patient detoxification, FPs underlined the necessity of cooperation. To summarize, most of the FPs saw great challenges and chances in a better cooperation with psychiatrists to improve medical care for (elderly) patients addicted to hypnotics.