The DOCUMENT classifications allowed some insight into the frequency and types of DRPs that were identifiable by the various sources. However, the DOCUMENT analysis did not provide sufficient detail regarding the particular problems that were identifiable. Naturally, each source of DRPs used their own vocabulary to describe the DRPs they were identifying, making direct comparison of the specific DRPs each source identified challenging. To solve this problem we developed a set of descriptive classifications for each of the DRPs. This set of classifications provided a set of descriptive terms describing each DRP in sufficient detail to allow for comparison between DRP sources. Descriptive classifications typically included specific drugs, drug classes and/or disease states giving clinical context to aid comparison.
7.1 Methodology
All DRPs identified by each source were mapped to descriptive classifications which described the drug and/or the disease or other therapeutic problems in greater detail than DOCUMENT classifications. Descriptive classifications allowed direct and detailed comparison of the DRPs that were identified by pharmacists, MRM, MRX and prescribing criteria.
The similarities between sets of prescribing criteria, shown in Appendix 8, formed the basis of the list of classifications. The initial list of classifications was developed where at least two or more sets of prescribing criteria were in agreement concerning particular prescribing problems. Classifications described the PIM and often the associated diagnosis. An example is the classification: NSAIDs used with (risk of) renal failure.
DRPs identified by prescribing criteria, MRM, MRX and pharmacists were mapped to this initial list of classifications. The unmapped DRPs were then examined. Further classifications were developed where at least any two of MRM, MRX, pharmacists or prescribing criteria described the same DRP concept. Unmapped DRPs were then mapped to these additional classifications, where possible. Again, the remaining unmapped classifications were examined. Classification descriptions were broadened and included DRP classifications found in the DOCUMENT classification system.275 Examples include: therapeutic dose too
Descriptive classifications
commonality with other prescribing criteria, software or pharmacists were assigned unique classifications. An example pharmacist-only classification is: compliance – using too little medication.
The list of descriptive classifications are tabled in Appendix 9. Distinct classifications per source are shown, that is, where two or more DRPs from the same source mapped to just one classification, that classification was counted only once, so as to eliminate duplicated findings.
Several descriptive classifications were excluded from comparisons as they were exclusively associated with pharmacists and were considered clearly out of scope of analysis that was achievable by software, given the nature of the available data. Excluded classifications involved pharmacist-only compliance and not-classifiable DRPs, specifically:
Communication breakdown, Documentation insufficient, Compliance – Confusion about therapy, Compliance – using too little medication, Compliance – using too much medication, Cost of therapy concern, Difficulty using dosage form, Eligible for DVA funded DAA, Medication expired, Medication regimen complicated, Other DRP pharmacist.
Examples of DRPs mapped to the same classifications are presented in Table 19.
Table 19: Examples of DRPs mapped to the same classification
Classification DRP Source DRP Source
Hyperlipidaemia undertreated MRM: Patient has elevated triglycerides and is only takng a statin. Additional treatment, such as a fibrate, may be worth considering.
Pharmacist: Patient's cholesterol and triglycerides remain elevated despite Lipitor. This may be due to poor compliance or an inadequate dose. Glibenclamide prescribed STOPP: Glibenclamide or
chlorpropamide with type 2 diabetes mellitus (risk of prolonged
hypoglycaemia)
Beers12: Sulphonylureas long acting: glibenclamide, chlorpropamide
Heart failure and concurrent
verapamil or diltiazem STOPP: Use of diltiazem or verapamil with NYHA class III or IV heart failure (may worsen heart failure)
Beers12: Heart failure: NSAIDS, COX2, diltiazem, verapamil, pioglitazone, rosiglitazone
Descriptive classifications
Classification DRP Source DRP Source
Heart failure and concurrent verapamil or diltiazem
MRM: Heart failure with calcium channel blocker: Patient has a history of heart failure and is taking either verapamil or diltiazem. These agents can worsen signs and symptoms of systolic heart failure. Alternative agents should be considered if possible.
Pharmacist: Diltiazem may adversely affect patients with heart failure
Sedative long-acting or sedative
long-term MRX: Diazepam: Potentially inappropriate medications: Certain medications or medication classes should generally be avoided in older persons because they are either ineffective or they pose unnecessarily high risk for older persons and a safer alternative is available. Is there an indication for the medication?
Pharmacist: Patient has been taking diazepam and temazepam for several years, which increases the risk of adverse CNS effects
The list of classifications was validated by a second pharmacist (and supervisor) – Professor Gregory Peterson. The table of 141 classifications and the 28 groups into which they were categorised are shown in Appendix 9.
Descriptive classifications were compared by frequency and type between pharmacists, MRM, MRX and sets of prescribing criteria. Analysis of classifications was mainly descriptive, presenting classification frequencies found by one DRP source or another and by those found in common. Classifications were considered to be 'in common' if the same classifications could be identified by two DRP sources in the same patient.
The basic unit of analysis was the number of distinct classifications found in each case. The number of distinct classifications found in each case may differ from the number of actual DRPs found in each case. An example may be a patient using both atorvastatin and simvastatin, both statins. If statins are contraindicated due to risk of myopathy, each of these DRPs might be assigned the classification statin myopathy risk. These two DRPs were then collated into only one distinct classification, since the end goal was to determine if pharmacist or software was able to detect this central theme in this patient, with no extra credit being given for finding essentially the same problem multiple times.
Descriptive classifications
The Jaccard index, also known as the Jaccard similarity coefficient,280 was calculated on a per patient basis for classifications found by pharmacists and by software. The index was calculated as the number of classifications in common (set intersection) divided by the total number of classifications found by software and by pharmacists in the same patients (set union). The equation is shown in Figure 27. Potential values range from zero (no similarity) to one (complete similarity). The mean Jaccard index across all patients was calculated to determine how similar classifications were between pharmacists and MRM, pharmacists and MRX, and pharmacists and prescribing criteria.
7.2 Results
Descriptive classification results between MRM and pharmacists are shown for the larger 570 case cohort, and for the smaller 100 test case cohort which included MRX. The following subsections also show the classifications found by prescribing indicators. Bar charts highlight the unique classifications identified by software sources in dark-blue and unique classifications identified by pharmacists in light-blue. Beige highlights the classifications which were identified in common.