Investigating the Relative Significance of Drug-Related Problem Categories

The aim of the review was to investigate whether an assessment of clinical significance can be related to specific drug-related problems (DRPs) and hence may assist in prioritizing individual categories of DRP categorization systems. The literature search using Google Scholar was performed for the period 1990 to 2013 and comprised primary research studies of clinical pharmacy interventions including DRP and clinical significance assessments. Two reviewers assessed the titles, abstracts, and full-text papers individually, and inclusion was determined by consensus. A total of 27 unique publications were included in the review. They had been conducted in 14 different countries and reported a large range of DRPs (71–5948). Five existing DRP categorisation systems were frequently used, and two methods employed to assess clinical significance were frequently reported. The present review could not establish a consistent relation between the DRP categories and the level of clinical significance. However, the categories “ADR” and possibly “Drug interaction” were often associated with an assessed high clinical significance, albeit they were infrequently identified in the studies. Hence, clinical significance assessments do not seem to be useful in prioritizing individual DRPs in the DRP categorization systems. Consequently, it may be necessary to reconsider our current approach for evaluating DRPs.


Introduction
Drug-related problems (DRPs) are associated with increased morbidity and health care costs, and identifying and addressing DRPs are essential tasks within the health care system [1][2][3]. Various definitions of a DRP exist, and one of them is: "a circumstance related to the patient's use of a drug that actually or potentially prevents the patient from gaining the intended benefit of the drug" [4].
DRPs may be identified and solved as a part of conducting medication reviews, which in practice, are often performed by clinical pharmacists in collaboration with physicians [5]. When reporting process measures of medication reviews, DRP categorization systems are frequently used. Several DRP categorization systems exist [4,6]. Some have been developed in primary care, where patients are usually an active participant in the medication review, while other DRP categorization systems only target issues that can be identified based on patient charts. The quality of DRP categorization systems has been reviewed by Van Mill et al. [4]. The authors identified 14 DRP categorization systems, but only three had been assessed for their usability in practice and internal validity, and none of the categorization systems met the proposed criteria for an optimal system [4]. Some had integrated the suggested intervention as a part of the DRP categorization system, which has also been reported by other authors [7].
• described primary research • were published in English • described interventions delivered by clinical pharmacists Publications were excluded if they: • were not published as a research paper (e.g., reviews, books, congress abstracts, posters, reports, protocols) • did not present data on DRPs • described selected DRPs only (e.g., drug-drug interactions) • did not present a DRP-categorization system • did not present data on an assessment of clinical significance • presented data for a sub-study, where the original study had been included The search was performed for the period from 1990 to 2013 using Google Scholar (TRHN). Google Scholar was used to ensure inclusion of the largest possible number of papers, because several studies presenting DRPs were expected to have been published in non-indexed journals. A similar search in PubMed yielded less than five hits.

Assessment
All titles and publication types from the original search were reviewed independently by TRHN and LJK. Subsequently, the abstracts were reviewed by these two authors. Thereafter, full-text articles were reviewed independently by CO and LJK. Finally, CO and LJK extracted data from the studies independently. At every step, disagreements were resolved by consensus. The data extracted were details regarding the study, the intervention, DRP-categorisation, and the clinical significance assessment. Action on DRPs were differentiated between the acceptance rate and implementation rate, because the implementation of a suggested intervention is often dependent on an action by the physician, and even though a physician accepts a suggested intervention, the action of implementing it may be lacking. The clinical significance of DRPs related to medication reconciliation was excluded, since DRPs related to medication reconciliation were not part of the inclusion criteria.
Pharmacy 2017, 5, 31 3 of 18 physician, and even though a physician accepts a suggested intervention, the action of implementing it may be lacking. The clinical significance of DRPs related to medication reconciliation was excluded, since DRPs related to medication reconciliation were not part of the inclusion criteria.

Description of Studies
The included studies had been conducted in 14 countries in Europe, Asia, Australasia, Africa, and North America, and most frequently, in India and Australia, with six studies each ( Table 1). The majority of the studies were conducted at one hospital (n = 17), followed by community pharmacies (n = 5) ( Table 1). A single study included patients from hospitals in two countries (UK and Saudi Arabia). Patients included in the study ranged from 46 to 737, and only seven studies had included more than 200 patients ( Table 1). Most of the studies involved a medication review, either by itself of in combination with other pharmaceutical care activities ( Table 1). The description of the interventions varied considerably between the studies and it was not possible to establish the level of similarities among the interventions (i.e., level of patient involvement, nature of collaboration with physicians, data sources used (e.g., laboratory values, chart information), and follow up). The number of DRPs identified varied considerably , as well as the acceptance rate (47-97%) and the implementation rate (67-90%); however, most studies did not report a separate implementation rate (n = 20) ( Table 1).
Several of the studies used DRP categorization systems based on an existing one, but which were adjusted to the conditions of their study ( Table 2). Five existing DRP categorization systems were frequently referenced as inspiration: Modified versions of Hepler and Strand [67], Strand [68], Cipolle, Strand and Morley [69], PCNE [70], and DOCUMENT [66], and some studies had applied the DRP categorization systems directly (Table 2). However, some studies did not refer to any published DRP categorization system (n = 5). Overall, the number of categories used varied considerably among the studies, ranging from six to 17 categories. No category was used by all DRP categorization systems, but some of the categories were frequently used, e.g., "Untreated indication", "Improper drug selection", and "Adverse drug reaction" ( Table 2). Several DRP categorization systems used an "Other" category, irrespective of how many other categories the DRP categorization system consisted of. Consequently, a comparison of the frequencies of categorised DRPs in the studies was difficult.
Fewer categories were used to categorise clinical significance (Table 3). In general, three categories were used; "Major, Moderate, and Minor" or similar labels for three categories (Table 3). For the categorization systems where more categories were added (Extremely important, low, insignificant, and adverse significant), the three core categories covered the vast majority of the DRPs ( Table 3). The most frequently referenced categorization systems were Hatoum (n = 5) and DOCUMENT (n = 3); however, 10 studies did not provide any reference to the system used (Table 3) [11,66]. A variety of assessment methods were used to categorise the clinical significance of the DRPs; most used an expert panel with a consensus approach (n = 8); however, some studies did not describe the assessment process (n = 5). The assessment of DRPs suggested that interventions in the included studies were most frequently categorised as "Moderate significant", followed by "Minor" and "Major" (Table 3). When non-consensus methods were used for an assessment of the clinical significance, the correlation between raters was occasionally reported, and most of the studies found low correlations between the raters. Furthermore, the correlations of raters of the same profession were often low.
In addition, the number of DRPs, for which the clinical significance had been assessed, was in several studies lower than the number of DRPs identified in the study according to Table 1. Therefore, not all studies assessed all of the DRPs for clinical significance.
The category heading covers various labels, e.g., "Untreated indication" covers, among others, "An unfulfilled indication for drug treatment", "Need for additional drug", "Undertreated", "Need for additional therapy", "Untreated condition", "Requires drug but not receiving it", "Lack of drug therapy", etc. N refers to number of patients included. * +Categories regarding medication discrepancies,  [73], k: ASHP [74]. "X" indicates the presence of the category in the published categorisation system. More than one "X" indicates that more than one category of the published categorisation system falls into the category used in the current table.

Relations between Clinical Significance and DRP Categorisation
An assessment of the relationship between the clinical significance rating and categorization of DRPs was presented for six publications at various extents [44,50,51,61,62,66].
Bondesson et al. [44] presented a cross-table of the clinical significance categories according to Hatoum [11] and DRP-categories for the 127 suggested interventions. Most of the DRPs were rated as having a "significant" clinical significance, only the ADR category had higher clinical significance ratings, and a combined category of "Wrong dosage form/wrong drug" had a lower clinical significance rating [44].
A study by Elliott & Woodward [50] also published a cross-table of the clinical significance categories according to Standards Australia and DRP-categories of 113 DRPs. A high level of clinical significance was assigned to DRPs in the categories of "Untreated indication", "Medication management problem", "ADR", and "Drug interaction", while lower levels were assigned to, e.g., "Potentially unnecessary medication" and "Inappropriate medication choice" [50].
Granaas et al. [51] used a scoring system to categorise the clinical significance by Eadon [75], and of the 388 identified DRPs, 75 DRPs were selected for a clinical significance assessment to cover all DRP categories [51]. The DRP categories with the highest clinical significance scores were "Adding a medicine", "Drug interaction", and "Monitoring and counselling", while the lowest scores were assigned to "Cost-related" and "Generic substitution" [51]. ADR was not included in the DRP categorization system. Since the DRPs were non-randomly selected for the clinical significance assessment, the ratings may most likely not be generalised to the entire cohort.
In a study by Spinewine et al. [61], 700 of 1066 suggested interventions were categorised according to their clinical significance, based on van Mill and Hatoum [4,11]-366 suggested interventions were excluded, since they were assessed to have no clinical significance [61]. The categories with the highest clinical significance scores were "Change dose", "Add a new drug", and "Discontinue drug" [61].
Stafford et al. [62] selected the 316 DRPs with the highest clinical significance scores of 1038 DRPs using the DOCUMENT categorization system [66]. The DRP categories assigned the highest scores were "Toxicity or adverse reaction", "Drug selection", and "Untreated indications" [62].
Finally, Williams (2012) [66] also used the DOCUMENT categorization system to assess 5948 DRPs for clinical significance. As in the study published by Stafford et al. [62], the DRP categories assigned the highest scores were "Drug selection" and "Toxicity or adverse reaction", but in this study, "Over or underdose" was also frequently assigned a high score [66].

Discussion
Despite the difference in the categorization systems used for DRP categorization, as well as for clinical significance categorisations, it seemed like the category "ADR" and possibly "Drug interaction" were often associated with a high clinical significance (based on six heterogeneous studies). These two categories were infrequently identified in the studies, but when identified as a DRP, they seemed to be assessed as serious for the patient. No obvious pattern between the remaining DRP categories and the level of clinical significance could be established.

Relations between Individual DRP Categories and Clinical Significance
An evaluation of the studies, which compared clinical significance of individual DRP categories, showed no apparent relations between the DRP categories and the level of clinical significance in general. However, "ADR", which was one of the two categories possibly associated with a high clinical significance, has been established as a cause of hospitalization, which does support the relation with a high clinical significance [3].
Additionally, a large difference in the rating of clinical significance within individual DRP categories was observed. This may be explained by the difference in DRPs allocated to individual categories. For example, a high dose of a drug with a narrow therapeutic index may result in considerable damage to the patient compared to the prescription of a high dose of penicillin. Consequently, applying the clinical significance to individual DRP categories does not seem to provide valuable information for evaluation purposes.

DRP Categorization Systems
Despite a great variation in the type and number of categories of the DRP categorization systems, several papers referred to Hepler and Strand [67] regarding the choice of DRP-categories. However, the categories published by Strand et al. [68] are stated in the paper by Hepler and Strand [67], and it is likely that these categories form the basis of the majority of the DRP-categories subsequently used by other authors.
Most authors did not describe the method used to categorise DRPs. Van Mill et al. [4] reported that, in general, the validation of the categorization systems is poor. However, a practice study of a Danish DRP database showed that despite no formal training in the use of the database, the interrater reliability (Fleiss' kappa = 0.79) and reproducibility (Fleiss' kappa = 0.81) were high [76]. Other studies have also reported relatively high interrater reliability scores of different DRP categorization systems, with kappa values ranging from 60-75 [73,[77][78][79][80]. These findings suggest that irrespective of the type of system used, clinical pharmacists seem to agree on how to categorise DRPs; however, some cases of suboptimal medication treatment may be ambiguous and hence difficult to categorise [76]. Such cases will remain difficult to categorise irrespective of the number of categories available in the DRP categorization system [76].
No single category recurred in all of the DRP categorization systems; however, some were frequently used, including the category "Other", irrespective of the number of other categories. It is possible that this category is merely a "safe guard" to ensure that even rare and odd DRPs have a categorisation option.
The large variance in the number of categories available may be due to the type and purpose of the studies and practices locally, such as the need for the level of details, focus of medication review, patient involvement, etc. It is also possible that further categories will be added over time according to the development of medication reviews and appearance of new DRPs, e.g., related to computerised order entry systems [76,81]. Objectively, it does seems unnecessary that so many different DRP categorization systems exist with the aim of describing the findings of a clinical pharmacy service delivered internationally. International consensus on one DRP categorisation system may be impossible; however, it would improve the potential of comparing results internationally.
Finally, the clinical significance of using DRP categorization systems could be considered. For study evaluation purposes, DRP-categorisation systems are often considered as process measures to document a part of the activities delivered by the clinical pharmacist. DRP categorisation seems to be easy to apply in practice, but possibly quite time consuming. Since the ratings of clinical significance within individual DRP categories are inconsistent, the only value of applying a DRP categorization system seems to be of descriptive character. Whether the effort is worth the value should be considered for each individual study.

Clinical Significance Assessment
Irrespective of the number of categories available for the clinical significance categorization systems, the vast majority of the DRPs or suggested interventions were allocated into one of the following categories: "Minor", "Moderate", and "Major" (or similar). Whether a categorization system consisting of these three categories is optimal, will depend on the purpose of the study. Even though only a few DRPs or suggested interventions end up in the "Extremely important" and "Adverse significant" groups, these categories may serve as valuable sources for intervention purposes, i.e., to improve practice for physicians and clinical pharmacists, respectively. Some systems use a category of "No significance" [9,11], while others seem to exclude non-clinically significant DRPs. The DRPs either excluded from the evaluation or allocated to the "No significance" category may be a heterogeneous group ranging from DRPs of no clinical significance to DRPs of potential clinical significance, e.g., lack of patient understanding of medication regimen, and economical significance such as the choice of an expensive drug when an alternate inexpensive is available. The main aim of medication reviews is to optimise the medication treatment, but the cost of the drugs is often also assessed to help minimise costs for the society, as well as for the individual patient. In fact, the high cost of drugs may indirectly influence the clinical significance in the case of patients choosing to cease treatment with expensive drugs. These issues cannot be elucidated by the clinical significance categorization systems included in the current study-or, to our knowledge, any other assessment tool targeting the importance of DRPs.
The methods used to assess clinical significance included a consensus methodology. For example, expert panels may be a valid method to assess clinical significance, but using this methodology did not identify any obvious pattern between the assessments and DRP categorisation. When a consensus methodology was not used, the correlation between the raters was often low, even though only a few categories were available. This suggests that clinical significance may be related to the instruction of how to use the categorization system, but also the experience, background, and setting of the individual health professional rater. Clinical significance may even be rated differently according to the patient, who is the core person related to the medication treatment. For example, a patient might want to avoid a treatment due to side effects, irrespective of the benefits of the treatment. In addition, variation may also be a result of the variation of the type of intervention, patient population, setting, etc., of the included studies. Hence, using clinical significance categorization systems as assessment methods does not seem robust and is most likely highly dependent on the choice of individual raters. This is supported by a review by Vo et al., who found limited results for the validity and reliability of tools for assessing the potential significance of pharmacist interventions [82]. Consequently, using a clinical significance rating as a method to prioritise DRP categories does seem suboptimal. Indeed, it is possible that the low validity of DRP categorisation systems, as well as of clinical significance methods, may be a contributing factor to the fact that no apparent correlation was established.

Limitations
Generalisability may be questioned, since most studies were conducted at selected wards at one hospital and included less than 200 patients. However, the variation in the results of the number and types of DRPs and level of clinical significance may be explained by the heterogeneity of the studies, such as the type of intervention, training of clinical pharmacists, patient population, access to data sources, acceptance rates, etc.
Our literature search was challenged by the inconsistency of terminology used within the area of DRPs [4]." Drug related problems" (potential and actual) may be labelled differently like, e.g., "medication related problems", and "clinical significance" may be designated, e.g., "clinical significance". Additionally, some DRP categorization systems comprised categories of problems, while other systems used categories related to the interventions. This made a comparison of the categorization systems difficult.
As a part of presenting the data, categories were merged from the various DRP categorization systems, as well as for the clinical significance categorization systems. It is possible that the merged categories did not fully correlate with each other.

Conclusions
The current review could not establish a consistent relation between the DRP categories and the level of clinical significance. However, the categories "ADR" and possibly "Drug interaction" were often associated with a high clinical significance, albeit they were infrequently identified in the studies (based on six heterogeneous studies). Additionally, an assessment of clinical significance seemed to be a method of low validity. Hence, clinical significance ratings do not seem to be useful in prioritizing individual DRPs of the DRP categorization systems. Indeed, the value of applying DRP categorization systems to evaluate a study should be considered carefully in relation to the time spent performing the categorization. Consequently, it may be necessary to reconsider our current approach for evaluating DRPs.

Conflicts of Interest:
The authors declare no conflict of interest.