This study has defined 26 consensus CPQI, 16 patient outcomes and one PCB. To our knowledge this is the first study to define a PCB using a consensus-based approach. While Fernandes et al [19] defined a PCB in Canada, their study reached consensus on individual CPQI and then they bundled these together as a PCB, whereas this study reached consensus by giving respondents the opportunity to consider multiple PCB’s.
There were 26 individual CPQI which reached consensus. This represented 72% of all CPQI considered as part of this research. Lloyd et al [20] has previously defined a consolidated set of seven CPQI and proposed these as standard measures for hospital pharmacy practice in Australia. Each of the processes defined by Lloyd reached consensus in this study.
PCB9 (see Table 5) reached consensus. It is not unsurprising that the makeup of this agreed bundle contains these elements, as they align with the requirements of Australia’s National Safety & Quality Health Service Medication Safety Standard [35]. While the National Standards does not require clinical pharmacists to perform all activities, most are commonly performed by pharmacists during routine care. These activities also ranked as the top seven clinical pharmacy processes when participants ranked from most to least important in terms of patient outcomes.
Lloyd et al’s [20] list of indicators also included ‘percentage of attempted clinical interventions by pharmacists that were accepted by the clinician’ which was similar to the next best ranked process defined in this study which was “A Pharmacist makes a recommendation to optimise medicines which is accepted by the treating team”. It would be expected that as part of medication reconciliation and review processes, pharmacists would identify and resolve medication related problems and this is performed as part of pharmacist activities already defined in the PCB. Fernandes et al [19] also included a similar indicator when defining consensus CPQI in Canada. In this study, pharmacist recorded interventions and pharmacist interventions accepted by the treating team reached consensus as outcomes rather than CPQI process measures.
Interestingly, the CPQI’s related to provision of medicine stock/supply didn’t reach consensus. While pharmacists’ clinical activities have transitioned from the dispensary to hospital wards over a number of decades, if a medicine is not available for patient, there is a core responsibility for the pharmacy department to arrange supply for a patient. While it is highly likely that as part of a clinical pharmacist’s medication reconciliation process they identify the needs for medication supply, the link between medicine availability and patient outcomes is potentially being overlooked. This may not be seen as a ‘clinical’ task, but more of an administrative task, which is often these days delegated to pharmacy assistants.
One of the risks of defining and measuring a core set of CPQI is that it may lead to a delay in the uptake of emerging clinical activities to support patient care as that activity is not captured or it may occur at the expense of activities measured using CPQI. One such example is partnered pharmacist medication charting, which has demonstrated a reduction in medication errors and average length of stay when performed in general medicine units [36]. This activity was not considered as part of the CPQI process as it was not found in any of the articles during the literature search phase and it was not proposed by any of the DOPs as part of the multi-round surveys.
Similarly, post-discharge ambulatory care provision by pharmacists did not reach consensus. Despite expansion of services into ambulatory care settings in recent years [37], this CPQI did not reach consensus. There may be multiple reasons why it didn’t reach consensus including that historically clinical pharmacy services offered by hospital pharmacists has been inpatient focussed or that the measurement of such activities has been focussed on quantifying activity and service provision, rather than patient outcomes [38].
Outcomes
There were sixteen outcomes which reached consensus. Seven required a second survey round. Hospital acquired complications (HACs) was the highest ranked outcome measure, and interestingly ranked higher than the more specific medication related HACs outcome, which required two rounds to reach consensus.
There was variability in consensus on outcomes related to re-admission or representation. Readmission due to medication misadventure, within 10 days or within 30 days all reached consensus, however readmission rate and unplanned representations did not reach consensus. Perhaps the former three outcome measures were more specific and could be better attributable to clinical pharmacist activities rather than general readmission or unplanned representation rates.
Patient focussed outcomes such as self-reported confidence with managing medicines, experience measures and outcome measures all reached consensus, however they rated lower generally than other consensus outcomes. Quality of life did not reach consensus. This may reflect the participants of this multi-survey study, where patients themselves may be more focussed on these outcomes and less focussed on other historical hospital outcome indicators, which may have financial or administrative implications. Having said that, a number of historical outcome indicators which are often routinely measured by hospital administrators such as medication costs per weighted activity unit, hospital standardised mortality ratio and length of stay did not reach consensus.
Not unsurprisingly, only seven of the sixteen outcome measures which reached consensus are routinely measured by more than half of the survey respondents. Similarly, 10 out of the 26 CPQI were reported to be measured by 50% or more sites who responded to survey rounds. Interestingly, paper-based and digital sites monitored similar numbers of CPQI, however paper-based sites reported them more difficult to measure. Digital sites reported outcomes to be easier to monitor, however they monitored on average the same number as paper-based sites. This highlights an opportunity, as well challenges, to benchmark services across Queensland.
While standards of practice exist which inform key activities and suggest QI [39], these are for single isolated activities. Pharmacy managers and the workforce require standardised definitions to monitor overall service provision and measured as outcomes rather than process measures. Another risk with monitoring single process measures is that it may not adequately reflect the quality of service delivery. This was a theme which arose when Minard et al [40] conducted focus groups with clinical pharmacists in Canada. The linkage of delivery of a PCB to patient outcomes, which then considers patient and medication-related factors will assist to inform pharmacy managers of service delivery quality, as well as how best to target their services to patients at the highest risk of medication misadventure. Visibility of this data will enable the workforce to flex and perform activities/services according to patients at risk and enable rapid and ongoing monitoring. This could be achieved by defining each CPQI, outcome and PCB using routinely recorded data. Using an approach such as this will also be an enabler of rapid evaluation of new or altered models of care. For example, implementation of a new service or new clinical activity can be ‘added-on’ to the current PCB defined, and if routine reporting/monitoring is already in place, rapid evaluation of outcomes can be assessed.
Limitations
This research used DOPs across Queensland to define CPQI, PCB and relevant outcomes using a modified-Delphi consensus approach. One limitation of using this approach is that the participants are key to influence the findings. DOPs are well placed to understand current clinical pharmacy evidence and to inform how clinical pharmacy services should be delivered, hence the rationale for the choice of them as participants. However, there are many other stakeholders involved with medicines management, including patients, nurses, doctors, managers, executives and other multidisciplinary team members. The views of other stakeholders have not been explored during this research and this insight may provide different results.
Another limitation of the Delphi-approach is the way consensus is defined. We defined consensus as an average rating of greater than 4 on a 5-point LIKERT scale. While this approach means that on average the participants rated agreed or strongly agreed with the statement they were rating, if we had defined this differently, we may have reached a different outcome.
While there are limitations to the participants and the definition of consensus, the results are still useful to inform how to monitor processes as individual activities, bundled activities and link them to patient outcomes.