Impact of findings on practice statements

  • The new classification system GSASA V2 may serve as a helpful tool in daily practice to classify DRPs and clinical interventions undertaken by pharmacists.

  • Classification of DRPs together with according interventions enables demonstration of the performance/impact of clinical pharmacy services.

  • This classification system could be a helpful instrument to collect and quantify data on pharmaceutical interventions, thus enabling the merging of data for epidemiological studies.

Introduction

Drug-related problems (DRPs) are common in hospitalised patients. As defined by the Pharmaceutical Care Network Europe (PCNE), a DRP is an event or circumstance involving drug therapy that actually, or potentially, interferes with the desired health outcomes [1]. A drug-related problem can be a risk to the patient (potential problem) or cause harm (manifest problem) as an adverse drug event (ADE) or an adverse drug reaction (ADR). Multiple causes for DRPs are known such as medication error, poor documentation, failures in communication, inappropriate processes in the health care setting or the patient’s behaviour. A systematic review analysing DRPs in hospitals showed that problems associated with pharmacotherapy lead to a prolonged hospital stay and increased healthcare costs. Medication errors occurred in 5.7 % of all episodes of drug administrations, and 6.1 % of hospitalised patients experienced an ADE or ADR [2].

Increasingly, clinical pharmacists are involved in detecting and solving DRPs on a regular basis. Utilisation of a classification system would aid in the collection of DRPs and the assessment of pharmaceutical interventions; support continuity of care through the promotion of mutual information [3]; and, additionally, such data on pharmacists’ activities could be used for epidemiological studies.

In the literature several classification systems have been proposed. Most instruments, such as APS-Doc [4], DOCUMENT [5], and PI-Doc [6], were considered too time-consuming in practice. Another such system, the PCNE [12] classification system, was originally developed for a research and community pharmacy setting and has a strong focus on patient behaviour, therefore making it less appropriate for the hospital setting. Typical hospital medication errors such as application errors, incompatibilities, and incorrect transcription cannot be classified [3]. The large number of subcategories (n = 71) renders the tool very comprehensive, but hinders its application in a daily routine setting. Allenet et al. validated an instrument for the documentation of clinical pharmacists’ interventions (SFPC system), which proved to be suitable for daily practice [7]. However, this simple system lacks subcategories to document detailed information, and the cause of the DRPs is not assessed. Hence, validated, structured, and standardised classification systems for pharmaceutical interventions, which fulfil both requirements of comprehensive classification and simple use in daily clinical practice, are rare.

Validation confirms, through the provision of objective evidence, that the requirements for a specific use or application are fulfilled [8]. Validation of a classification system is necessary, not only to ensure that one code reflects a unique DRP, but to guarantee that this coding is understandable to user. The literature describes the following criteria for validating DRP classification systems: (1) appropriateness (is the classification content appropriate to the questions the application seeks to address?) (2) acceptability (is the classification acceptable to the users?) (3) feasibility (is the application easy to use?) (4) interpretability (how well can the classification codes be interpreted?) (5) reliability (does the classification generate results that are reproducible and internally consistent?) (6) validity (does the classification measure what it claims to measure?) (7) responsiveness (does the classification offer options to follow up interventions and monitor outcomes of interventions?) [9].

Up to now, there was no national consensus in Switzerland on how to demonstrate the clinical pharmacist activities to obtain data allowing epidemiological studies for research and political purposes. The working group on clinical pharmacy of the Swiss Society of Public Health Administration and Hospital Pharmacists (GSASA), comprising eight French- and German-speaking clinical pharmacists, recognised the need for the development of a new standardised and practical tool. To ease the recording of interventions in inpatients during daily practice, a tool was developed, which seeks to combine the advantages of existing systems such as SFPC (validated, practical, and based on hospital setting) and PCNE (validated, logical basic structure with the categories cause and intervention) systems. The classification system focused on interventions to enable a more objective assessment, and increased quality and reliability of data recording. We used the PCNE system, which is validated, well-established and internationally used, as a benchmark for our new intervention oriented classification system [3, 10]

Aim of the study

The aim of the study was to develop a classification system for drug-related problems and pharmaceutical interventions, and to validate this system using inpatients and against the PCNE classification system V6.2.

Ethical approval

According to the requirements of the Swiss federal law on human research this study is exempt from ethical approval.

Methods

Overview of development process

Figure 1 illustrates the process involved in developing the new GSASA classification system, which comprised four main steps. The topics were based on those of the PCNE classification system, while the structure followed that of the French classification system [7]. The first version (GSASA V1) of the classification system was developed by an expert panel of eight clinical pharmacists (GSASA working group on clinical pharmacy). After validation, a second version was developed (GSASA V2) which was revalidated.

Fig. 1
figure 1

Process of developing the classification system

Full size image

We defined a “pharmaceutical intervention” as a recommendation initiated by a pharmacist in response to a DRP occurring in an individual patient in any phase of the medication process. The intervention aims at optimising pharmacotherapy, in terms of efficacy, safety, economic, and humanistic aspects [11].

Step 1: Development of classification system GSASA V1

The GSASA working group (=expert panel) comprised four French and four German speaking clinical pharmacists (n = 8) from 8 different hospitals, whose professional experience in clinical pharmacy ranged from 3 to 14 years. Seven of them had previously used a DRP classification system. The first version, developed by the aforementioned GSASA working group, was based on the PCNE classification system for DRPs [12] and the instrument for documentation of clinical pharmacists’ interventions of the French Society of Clinical Pharmacy [7]. Any discrepancies were resolved by discussion.

Step 2: Validation of classification system GSASA V1

Version 1 was validated assessing appropriateness, interpretability, validity, feasibility, acceptability, and inter-rater reliability.

Appropriateness, interpretability, and validity

We measured appropriateness, interpretability, and validity of the classification systems by assessing the proportion of completely classified interventions. Classification was considered complete when all categories were filled out. At a 427-bed teaching hospital, six experienced clinical pharmacists used the GSASA V1 during a 6-week period to classify the interventions they performed themselves from their routine ward rounds (in geriatric ward, rehabilitation clinic, and orthopaedic ward). Additionally, they classified the same data with PCNE V6.2, and entered the classification codes into a Microsoft Excel sheet. For each DRP, only one choice per category was possible. Special attention was paid to the cases that could not be completely classified.

The pharmacists received training prior to data collection. Training mainly comprised classification of model cases according to standardised documentation forms of PCNE and GSASA, followed by plenum discussions. Validated model cases in a German translation were used [13]. The collected data were analysed by descriptive statistics.

Acceptability and feasibility

To evaluate acceptability and feasibility of both classification systems, an 8-item questionnaire, which has been used in an earlier study, was completed by the six pharmacists [13, 14]. The extent of their agreement or disagreement was assessed by a 5-point Likert scale (1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = strongly agree). Time spent for classification and the free text comments was then evaluated. A Mann–Whitney U test was used for statistical evaluation. The significance level was accepted at p < 0.05.

Inter-rater reliability

Three of the six senior clinical pharmacists assessed the reliability of the classification systems. Each had more than 5 years of professional experience in clinical pharmacy, and had worked with DRP classification systems before. They classified 10 model cases using GSASA V1 and PCNE V6.2. The model cases consisted of five validated model cases taken from the literature [15], and five model cases developed for the validation of PCNE V5.0 taken from the German translation. Drug names were only modified to suit the Swiss market. We randomised the order of model cases and classification systems, and each rater received the same instructions. For both classification systems, only one choice per category was possible to classify each detected problem.

For the four categories of both classification systems (detected problem, cause, intervention, outcome of intervention) Fleiss kappa was calculated using a Microsoft Excel template [16]. Resulting values were interpreted according to Landis and Koch [17] as ‘almost perfect’ (Fleiss’ κ 0.81–1.00), ‘substantial’ (0.61–0.80), ‘moderate’ (0.41–0.60), ‘fair’ (0.21–0.40), ‘slight’ (0.00–0.20), and ‘poor’ (<0.00). A kappa higher than 0.4 indicates that the system is reliable.

Step 3: Development of classification system GSASA V2

Revision of version 1

The GSASA working group reviewed the results of the validation of GSASA V1. Conclusions were drawn and discussed until consensus was reached.

Translation

The GSASA working group translated the German GSASA V2 into French during an open discussion. For the purpose of this paper, we additionally translated version 2 into English.

Step 4: Reliability of classification system GSASA V2

Inter-rater reliability

The GSASA working group assessed the inter-rater reliability of the German and French versions of GSASA V2 as described in step 2. They classified the same 10 model cases using the GSASA V2.

Results

Step 1: Development of classification system GSASA V1

The first version included 4 main categories and a total of 35 subcategories, i.e., detected problem (3 subcategories), cause of intervention (17 subcategories), intervention (10 subcategories), and outcome of intervention (5 subcategories).

Step 2: Validation of classification system GSASA V1

Appropriateness, interpretability, and validity

DRPs were collected from daily work on the wards during a 6-week period. We classified 115 DRPs with PCNE V6.2 and GSASA V1. The proportion of the classified cases and the categories involved are shown in Table 1. In both classification systems, the majority of the cases could be completely classified (PCNE 81.7 %, GSASA 80.9 %).

Table 1 Proportion of classified cases per system and per category
Full size table

Acceptability and feasibility

The six pharmacists completed an 8-item questionnaire on the usability of PCNE V6.2 and GSASA V1 using a 5-point Likert scale. Data was compared using Mann–Whitney U Test. The results of the questionnaire were not statistically significant. Table 2 shows the differences of the results for acceptability and feasibility of the two classification systems (questions 1–7).

Table 2 Users’ agreement on the classification systems adapted from AbuRuz et al. [14]
Full size table

Question 8 allowed the pharmacists to record their comments and suggestions. The subcategories ‘untreated indications’ and ‘documentation errors’ were missing in the category ‘problem’, ‘duplication’ and ‘insufficient effect of drug treatment/inappropriate drug’ in the category ‘cause’ and ‘recommendations of laboratory test’ in the category ‘intervention’.

Inter-rater reliability

Figure 2 illustrates the inter-rater reliability of the four classification categories, i.e., problem (GSASA V1 κ = 0.66, PCNE V6.2 κ = 0.32), cause (GSASA κ = 0.53, PCNE κ = 0.44), intervention (GSASA κ = 0.74, PCNE κ = 0.40), and outcome (GSASA κ = 0.63, PCNE κ = 0.52). The three pharmacists showed a fair agreement for the category ‘problem’ and a moderate agreement for the other categories of the PCNE classification system. In comparison, GSASA V1 reached a moderate agreement for the category ‘cause’ and a substantial agreement for the other categories.

Fig. 2
figure 2

κ-Coefficients of PCNE V6.2 and GSASA V1 classification systems for the four categories, based on standard cases (n = 10) classified by raters (n = 3)

Full size image

Step 3: Development of classification system GSASA V2

The results of the validation of GSASA V1 and the suggestions from the six users were discussed in the expert group, and resulted in the addition of one new category ‘type of problem’ and seven new subcategories, and in the modification of three subcategories. The subcategory ‘untreated indication’ was moved from the category ‘cause’ to ‘problem’. The major change concerned the category ‘detected problem’. To precisely describe the DRPs, we included two additional subcategories to this category, and introduced the new category ‘type of problem’ to differentiate potential and manifest DRPs. Table 2 describes the English version 2 and the modifications with respect to version 1. The resulting classification system GSASA V2 includes 5 categories with a total of 41 subcategories as follows: detected problem (5 subcategories), type of problem (potential/manifest) (2 subcategories), cause of intervention (18 subcategories), intervention (11 subcategories), and outcome of intervention (5 subcategories) (see Table 3).

Table 3 Description manual of the classification system GSASA V2 and illustrations with examples (bolded text category or subcategory added for version 2, italicized text subcategory modified)
Full size table

Only one choice per category is possible. Therefore, if a detected problem involved multiple interventions, each intervention required the use of a new form or line in the Excel sheet. An example to illustrate this classification is given in Fig. 3.

Fig. 3
figure 3

Example of a pharmaceutical intervention classified as a drug-related problem according to classification system GSASA V2

Full size image

Step 4: Reliability of classification system GSASA V2

Inter-rater reliability

The working group assessed the level of agreement of the version V2 in German and French (Table 4). They classified the same 10 cases used in step 2. Inter-rater reliability was moderate (κ = 0.52) for all categories.

Table 4 Level of agreement of the GSASA V2 among experts (n = 8), 10 standard cases
Full size table

Discussion

Our study showed that most (80.9 %) of the 115 pharmaceutical interventions could be documented with the first GSASA classification system V1 and a similar ratio of 81.7 % with the PCNE classification V6.2, our benchmark. Moreover, we found comparable inter-rater reliability and acceptability for the GSASA and PCNE systems. On the other hand, the comparative evaluation of the two systems revealed differences with respect to usability. Indeed, the category ‘intervention’ of the GSASA system allowed a more complete classification of the cases than the PCNE. This reveals that our system respected his original approach, which was focusing on recording the interventions.

The structure of the two systems could also explain these differences. The four main categories of GSASA V1 corresponded with the ones of PCNE V6.2. However, PCNE V6.2 contained a twofold larger choice of subcategories (n = 71) than GSASA V1 (n = 35) enabling the precise classification of most DRPs. Consequently, users could find the PCNE instrument to be more comprehensive than the GSASA system, knowing that, due to the small number of raters, the comparison of both tools showed no statistically significant results. In contrast, the GSASA system could be easier to use and more practical than the PCNE system. Time is an essential element for the acceptance of a classification system. In routine settings, application of the GSASA system in clinical practice demonstrated this tool to be less time-consuming than the PCNE system. This important factor should increase the chances of a successful and systematic use of the GSASA system. By addition or modification of several subcategories, the number of non-classifiable cases should decrease. In this way, the usefulness/comprehensiveness of the GSASA system could be enhanced without affecting its well-established practical use. In summary, the validation of the two existing systems showed an acceptable performance in enabling documentation and a better acceptability and feasibility of the GSASA system. The comments of the users provided helpful input for further improvement and the development of the classification system GSASA V2.

The goal of this development process was to create a classification system that permits the classification of DRPs detected and the recording of any pharmaceutical intervention. Van Mil et al. describe essential characteristics of classification systems [10]. Accurate classification of a detected problem should lead to only one choice per category. Therefore, the comprehensiveness of our instrument allows its systematic use and the consistency in the documentation of the interventions. Its detailed description manual, illustrated with practical examples, should enable homogenous data collection. In this way, the classification system would allow to collect and pool data from different sites, and by this generating a representative overview of clinical pharmacy activities within a given region. As a disadvantage, our instrument allows limited entry of details on individual cases. However, its open structure enables to enter additional and important information about the coded interventions.

The classification GSASA V1 reached good inter-rater reliability. Indeed, the four classification categories of GSASA V1 (κ = 0.64, which indicated a substantial agreement) was more reliable than the four categories of PCNE V6.2 (κ = 0.42, moderate agreement). Inter-rater reliability of GSASA V2 (κ = 0.52) was acceptable, although the κ-coefficients were lower than those calculated for the initial version. This decrease of the inter-rater agreement can be explained by the extension of the classification system from 4 to 5 categories. Additionally, the raters for the second version were more heterogeneous in terms of language, professional experience, and clinical background. Due to minor changes in GSASA V1 only inter-rater reliability was repeated when revalidating GSASA V2.

Average inter-rater agreement for GSASA V2 was moderate (κ = 0.52). This Kappa value was similar to that of the DOCUMENT [5] instrument (κ = 0.53), a recent validated system for classifying DRPs and clinical interventions in community pharmacy. Similarly, the APS-Doc system obtained a substantial agreement for the categories and a moderate agreement for the subcategories [4]. Considering that (a) the pharmacists involved in our study had only little experience with the GSASA system, (b) they had never used a description manual to aid in DRPs classification, and (c) that Kappa value higher than 0.4 indicates the internally acceptability and the good comprehensiveness of the classification system, these results fulfil the minimum requirement for an acceptable classification system. In the future, the use of the descriptive manual to assist with the classification should improve the Kappa score.

This study involved several limitations. As in most classification systems, subcategories are not mutually exclusive. The GSASA system shows similarities with the PCNE and SFPC systems, which it stemmed from. The validation and reliability of GSASA V1 were based on a small number of pharmacists (n = 6 and 3, respectively), so we cannot exclude a selection bias. Many raters were involved in the different stages in the development process. Therefore, we cannot ensure the generalisability of the system. We limited the validation of GSASA V2 on reliability as only minor changes were required in the first version. We considered most results of GSASA V1 validation as transferable to GSASA V2. To enable its implementation we tested the classification system in a limited number of users (n = 8). All were qualified clinical pharmacists, each classifying 10 cases. On-going projects aim to evaluate the implementation and the user’s satisfaction of GSASA V2 in daily practice and to analyse the pooled data retrieved from Swiss hospitals. In addition, we are currently adapting the system to also suit the community pharmacy setting and to support seamless documentation and transition from secondary to primary care.

Conclusion

The intervention oriented classification system GSASA V2 appeared to be valid and easy to use in daily clinical practice. The system is validated in terms of appropriateness, interpretability, validity, acceptability, feasibility, and reliability. The description manual assists in categorisation and hereby will increase the quality of data due to an appropriate use of the standardised classification system. Systematic use of the procedure will provide information on the performance of clinical pharmacy services on the whole. On-going epidemiological research aims to merge all interventions classified with the classification system GSASA V2 in Switzerland and to evaluate its implementation.