Development and consensus testing of quality indicators for geriatric pharmacotherapy in primary care using a modified Delphi study

Background Polypharmacy is associated with an increased risk of adverse drug events in older people. Although national guidance on geriatric pharmacotherapy exists in Japan, tools to routinely monitor the quality of care provided by community pharmacists are lacking. Aim To develop a set of quality indicators (QIs) to measure the quality of care provided by community pharmacists in improving geriatric pharmacotherapy in primary care in Japan, using a modified Delphi study. Method The development of QIs for the Japanese community pharmacy context followed a two–step process: national guidance review and consensus testing using a modified Delphi study. The latter involved two rounds of rating with a face–to–face meeting between the rounds. Ten experts in geriatric pharmacotherapy in primary care were recruited for the panel discussion. QIs were mapped to three key taxonomies and frameworks: the Anatomical Therapeutic Chemical (ATC) classification system, problems and causes of drug–related problems (DRPs) taxonomy and Donabedian’s framework. Results A total of 134 QIs for geriatric pharmacotherapy were developed. This QI set included 111 medicine specific indicators, covering medicines in 243 third–level ATC classifications. QIs were classified into the problem of treatment safety (80%) and causes of drug selection (38%) based on validated classification for DRPs. In Donabedian's framework, most QIs (82%) were process indicators. There were no structure indicators. Conclusion A set of 134 QIs for geriatric pharmacotherapy was rigorously developed. Measurement properties of these QIs will be evaluated for feasibility, applicability, room for improvement, sensitivity to change, predictive validity, acceptability and implementation issues in a subsequent study. Supplementary Information The online version contains supplementary material available at 10.1007/s11096-022-01375-x.


Introduction
The global population is aging and as a consequence, more people are living with multi-morbidity and the consequent polypharmacy [1][2][3]. Polypharmacy is associated with an increased risk of adverse drug reactions (ADRs) [4][5][6]. Thus, it makes sense to monitor older people taking multiple medicines in order to minimise the potential for medication-related harm.
Community pharmacists are in an ideal position to monitor the use of medicines for the people they serve as they are generally the last health care professionals individuals see before they initiate or continue to take their medicines. This monitoring role by community pharmacists in primary care directly aligns with a broader international trend towards the provision of professional pharmacy services, which came to prominence in the 1990s with the advent of "pharmaceutical care" [7]. Pharmaceutical care has been defined as 'the pharmacist's contribution to the care of individuals in order to optimise medicines use and improve health outcomes' [8], highlighting the need for routine monitoring of the use of medicines within healthcare systems [9].
In line with this international trend, the Japanese Geriatrics Society published their first guideline for geriatric pharmacotherapy in 2005 [10], similar to the American Geriatrics Society which has published Beers Criteria in the US since 1991 [11]. The Japanese guidelines were updated in 2015 [12,13] as were the Beers Criteria [14] and the Screening Tool of Older Person's Potentially Inappropriate Prescriptions [15,16]. These guidelines balance the potential benefits of using medicines with their associated risks. In 2018 the Ministry of Health, Labour and Welfare (MHLW) in Japan created guidance to reduce polypharmacy problems in collaboration with the Japanese Geriatrics Society [17]. The guidance is aimed at healthcare professionals including physicians, nurses and pharmacists, to ensure older people, particularly those aged 75 and above, use medicines in an optimal manner. Their scope was expanded in June 2019, specifically in relation to transitions of care between healthcare facilities and/or patients' homes [18]. Consequently, the role of community pharmacists in the supply of medicines and medication management for older people at the risk of drug-related problems (DRPs) is increasing.
Although the guidance is designed to support healthcare professionals, specific instruction on how to utilise the content and monitor the quality use of medicines in community pharmacy is lacking. A well-recognised mechanism to measure care quality is via the use of quality indicators (QIs) across structure, process, or outcome domains [19]. QIs are usually defined with a denominator (the number of target population being measured) and a numerator (the number who have received the specific service), and measured as a percentage of correct actions (recommended care) [20]. Calculated QI scores indicate the quality of care and are monitored over time or for a specific period [21].

Aim
The aim of this study was to develop QIs to measure the quality of care provided by community pharmacists in improving geriatric pharmacotherapy in primary care in Japan, using a modified Delphi study.

Ethics approval
Approval was obtained from the social university general incorporated foundation ethical committee in Japan (SU1814, SU1912). Informed consent was obtained from all panellists.

Method
The development of QIs for the Japanese community pharmacy context followed a two-step process [22][23][24]. Two preliminary sets of QIs were developed following a comprehensive review. This was followed by a modified Delphi study for each set of QIs to achieve consensus [25]. The policy guidance documents were: 'Guidance on Appropriate Medication for Elderly Patients (general) in 2018' [17] and 'Guidance on Appropriate Medication for Elderly Patients, particularly for the recuperation environment in 2019' [18] ( Fig. 1). Two separate sets of QIs were developed because the 2019 document was published whilst we were conducting a modified Delphi study of QIs based on the 2018 document. Hence two separate ethics approvals were obtained. The consensus data obtained from both documents were aggregated so that a comprehensive set of QIs for all therapeutic categories could be obtained. This study was reported in accordance with the consolidated criteria for reporting a Delphi study (CREDES) [26].

Preparation of a preliminary set of QIs
A preliminary set of QIs was developed from each of the two evidence-based documents for geriatric pharmacotherapy [17,18]. Additionally, a government document for remuneration of community pharmacy services was used to develop outcome indicators [27]. Principal researcher NS extracted recommendations from the aforementioned documents pertaining to quality use of medicines for older people and used this data to develop the preliminary sets of QIs [28][29][30]. A second researcher, KF, verified this process.
To undertake a comprehensive evaluation of the QIs, they were mapped to the following three key taxonomies and frameworks: (1) the Anatomical Therapeutic Chemical (ATC) classification system [31], (2) problems and causes of drug-related problems taxonomy (p-DRPs, c-DRPs, respectively) [32] and (3) Donabedian's framework [19]. First, QIs were categorized into medicine specific indicators or general indicators, depending on whether the definition of QIs mentioned specific medicines. For instance, a QI about 'laboratory monitoring of antidiabetics' was classified as medicine specific indicators whereas a QI about 'the assessment of swallowing function' was categorised as general indicators.

3
After this step, medicine specific indicators were classified according to the ATC code [31]. If QIs were related to more than one ATC code, they were mapped accordingly (e.g. for nonsteroidal anti-inflammatory drugs, ibuprofen is 'M02AA13: musculo-skeletal system' whereas acetylsalicylic acid is 'N02BA01: nervous system'). Secondly, QIs were mapped to the classification system for DRPs developed by the Pharmaceutical Care Network Europe (PCNE) to determine the types of interventions which community pharmacists may undertake to resolve the causes of DRPs (Table 1) [32]. For example, a QI 'QI-42 percentage of older patients taking warfarin who received an international normalised ratio monitoring' pertained to 'adverse drug event (possibly) occurring in p-DRPs (P2.1)' and 'no or inappropriate outcome monitoring in c-DRPs (C9.1)'. If QIs were related to more than one c-DRP, they were mapped accordingly. Lastly, QIs were also categorised into Donabedian's framework of structure, process or outcome to identify the care type [19]. All classification was undertaken independently by NS and then verified by KF. The final mapping was discussed with all members of the research team (NS, KF, KK and TC).

Consensus testing
A modified Delphi study, specifically the RAND/UCLA appropriateness method, was applied to combine evidence-based QIs with expert opinion [25]. It involved two rounds of rating with a face-to-face meeting between the rounds. For each modified Delphi study, a purposive selection of ten panellists with expertise in geriatric pharmacotherapy in primary care was recruited by e-mail or telephone (Supplementary Table 1   and November 2019 and January 2020 ( Fig. 1). All data at each stage were reviewed by the research team for feedback and editing before dissemination.

First round online survey
Panellists judged the face and content validity of each QI, using a 9-point scale (ranging from 1 "definitely not appropriate" to 9 "definitely appropriate") with an opportunity to provide suggestions or modifications via SurveyMon-key™. This study defined appropriateness as "whether care described in the QIs must be provided in principle" and "whether a high QI score would be interpreted as a high-quality care" [33]. A QI with a median score of 7-9, without disagreement (i.e. at least 3 panellists scored 1-3, and at least 3 panellists 7-9) was judged as "appropriate" (i.e. median score ≥ 7, agreement 80%) [25]. The result from the first round and any additional comments made by panellists were de-identified and sent to all panellists one week before the panel meeting.

Face-to-face panel meeting
The expert panel meeting was conducted after the first round survey. QIs which did not achieve consensus as "appropriate" were discussed. Panellists were also invited to comment on ways to improve the comprehensiveness and accuracy of QIs which had reached consensus and propose other QIs to cover any perceived gaps. QIs which were re-worded and new potential QIs were included in the second round surveys. Panellists who were not able to attend the meeting in person provided written comments which were discussed at the expert panel meeting. The discussion at the panel meetings was digitally audio recorded.
PCNE Pharmaceutical Care Network Europe, P problems, C causes, TDM therapeutic drug monitoring, ADR adverse drug reaction. a 5 QIs were allocated to more than one c-DRPs.
b Outcome indicators which pertained to financial related outcome indicators (QI-123 to 134) were considered as not applicable (N/A).

Second round online survey
After the meeting, all panellists completed the same 9-point scale for evaluating QIs, as was used in the first round. Agreement was assessed using the same criteria as in the first round. After the second round, data from the two studies were aggregated and reviewed by the research team to assess overall comprehensiveness. The final result was sent to panellists for confirmation.

Results
A set total of 134 QIs for geriatric pharmacotherapy in primary care was developed to assess the quality of care in community pharmacies (Fig. 1). The detailed description of QIs and results of consensus testing are provided in Table 2 and 3.

Preparation of a preliminary set of QIs
A preliminary set of 137 QIs was developed from the national geriatric pharmacotherapy guidance documents [17,18]. A further six potential QIs were developed based on a government detailing the remuneration of pharmacy services [27]. Hence, a preliminary set of 143 QIs was prepared for the consensus testing.

First round online survey
All panellists completed the survey. Of 143 preliminary QIs, 113 QIs were assessed as 'appropriate' and 30 QIs did not meet the threshold. No new QIs were proposed.

Face-to-face panel meeting
Eight panellists participated in each expert panel meeting, with comments in writing from two panellists who were absent (Supplementary Table 1). With agreement of all panellists who attended the meeting, 107 QIs were accepted without change or with minor rephrasing in response to the first round comments, or were combined with similar QIs. The remaining 34 QIs were discussed and modified to improve accuracy, and were included in the second round. Additionally, 7 QIs were newly proposed based on panellists' perspectives. QIs regarding 'assessment of influenza and pneumococcus vaccination status' were made in relation to a national immunisation programme [34][35][36]. QIs regarding 'topical drugs for pain and dry skin' were proposed to avoid oversupply since older people tend to store excess topical drugs [37,38]. The assessment of functional status such as 'swallowing function' and 'renal function' was added by panellists, saying that 'functional decline among older people should be monitored by community pharmacists as standard care'. Lastly, a QI on 'laboratory monitoring of antiepileptic medicines' was added.

Second round online survey
All panellists responded to the survey. Of 41 QIs, 30 QIs were assessed as 'appropriate'. Eleven QIs did not meet the threshold (e.g. 'use of angiotensin-converting enzyme inhibitors for hypertensive patients with recurrent aspiration pneumonia' and 'long-term stimulant laxative use').
After both modified Delphi studies, the research team reviewed all accepted QIs to improve comprehensiveness. Twelve QIs which were deemed to be similar were combined into six QIs. Furthermore, three QIs pertaining to financial related outcome indicators were added by the research team with the agreement of all panellists, via e-mail. This process was undertaken to ensure the currency of the QI set because the government remuneration system for pharmacy services (for the period 2020 to 2022) was revised during the course of the study (e.g. number of claims for community pharmacy services that pharmacists demonstrated correct inhaler technique and made a report to a prescriber) [39]. As a result, 134 QIs were developed. The final result was sent to all panellists and confirmed by them.

Characteristics of QIs
Key taxonomies and frameworks were used to understand the coverage of the QIs for medicines use in geriatric pharmacotherapy, acknowledging the multidimensional nature of responsible use of medicines in older persons. Of the 134 QIs developed in this study, the majority of QIs (n = 111, 83%) were medicine specific indicators. Some of them were allocated into more than one ATC code, resulting in 131 first-level ATC classifications (Fig. 2). The highest proportion of QIs pertained to nervous system (43%), followed by alimentary tract and metabolism (18%), cardiovascular system (14%) and respiratory system (12%). No QI pertained to dermatologicals, antineoplastic and immunomodulating agents, antiparasitic products, insecticides and repellents, sensory organs and various. Table 3 and Supplementary  Table 2 present the third level ATC codes. Of 134, 122 QIs could be mapped to problems and causes associated with DRPs at the primary and sub-domain levels of the PCNE taxonomy. The remaining 12 QIs which pertained to financial related outcome indicators were not mapped ( Table 1). The most common problems caused by DRPs at the sub-domain level were 'adverse drug event (possibly) occurring (81%)', followed by 'effect of drug treatment not optimal (10%)'. For the causes of DRPs, QIs commonly           (Table 3). No structure indicators were developed but it is noteworthy that the reporting structure indicators in Japan is mandatory (e.g. availability of pharmacy home visit services and pharmacy health promotion activities).

Discussion
This seminal study described the development and consensus testing of a set of 134 QIs for geriatric pharmacotherapy, designed to evaluate the quality of care provided by community pharmacists in Japan in primary care. This QI set can be used for routine monitoring of the care provided by community pharmacists in optimising geriatric pharmacotherapy in primary care. Furthermore, the use of QIs could encourage community pharmacists to keep additional clinical patient records, ensuring that all important decisions are documented in line with the expansion of professional pharmacy services [40,41]. Medicine specific indicators were widely distributed across 10 ATC categories at the first ATC level. The large number of ATC categories represented reflects the complexity of geriatric care, as this QI set targeted older people who have multiple conditions and medications. Most QIs could be mapped to medicines from the nervous system, alimentary tract and metabolism and the cardiovascular system, which aligns with QIs developed for other countries [42]. It follows that ATC categories with the greatest number of QIs would likely require significant opportunities for professional services from pharmacists. For example, when evaluating third level ATC codes, many QIs pertained to anti-dementia drugs (n = 15), blood glucose lowering drugs, excluding insulins (n = 11), antipsychotics (n = 10), other analgesics and antipyretics (n = 8) and, antidepressants (n = 8), all of which are known areas of importance for geriatric care in Japan [13] and other countries [42] (Supplementary Table 2). Likewise, there were no QIs for some ATC categories such as dermatologicals, indicating fewer opportunities for input from pharmacists.
Although most QIs were medicine specific indicators, there were 11 general indicators which were classified using the PCNE DRP taxonomy and Donabedian's framework. Some of them appeared to be unique to the Japanese context. As an example, QIs for counselling about influenza and pneumococcus vaccination (QI-116,117) were unique in Japan where the administration of vaccines by pharmacists is not currently endorsed. In contrast, in Australia [43], the UK [44], the USA [45] and Canada [46], pharmacist vaccination programs exist yet no vaccine-related QIs pertaining to pharmacist vaccination exist. These QIs point to the importance of checking patients' vaccine status and educating and/or reducing misconceptions about immunisation by community pharmacists. Furthermore, general indicators also included unique QIs such as 'social services taken' or 'patients' willingness to deprescribe', which differed from existing general indicators that mainly focused on logistic issues such as medication reconciliation [47,48]. That is, general indicators developed in this study might refer to the degree to which community pharmacists understand the patients' background and can use this information to provide a more person-centred approach to medication management.
A large proportion of QIs was mapped to the c-DRP taxonomy including drug selection and monitoring related QIs. This may be explained by the fact that pharmacists play a critical role in resolving DRPs in these areas, considering the scope of pharmacy practice [41,49,50]. The majority of preventable DRPs are attributed to these categories [42]. On the other hand, no QI was included in the c-DRP category of 'drug form'. Because a QI regarding 'evaluation of swallowing function' was developed, this QI could prompt pharmacists to find such a problem in patient counselling, if 1st level of ATC classification system necessary, with pharmacists suggesting recommendations about drug form. Since the PCNE classification for DRPs is well-recognised and internationally used in medication management research, mapping QIs to this taxonomy provides an opportunity to compare QI scores between countries where the same taxonomy is used. Thus the quality of care can be assessed from multiple perspectives by stratifying QI results using this taxonomy.
Most QIs developed in this study were process indicators. This is not surprising and aligns with other data which shows that more than 90% of existing QIs for responsible use of medicines were process indicators [42,51]. Indeed, since 2008 when the MHLW in Japan community pharmacy remuneration to generic-drug dispensing rates, performance-based payment models have expanded. Therefore, associations between process indicator scores and subsequent outcomes are expected to be evaluated.
Twelve of the 23 QIs which were mapped to outcomes pertained to financial outcomes (e.g. QI-123 follow-up services for those with diabetes), with the remainder aligning to medication appropriateness (e.g. QI-24 percentage of diabetic patients without sulfonylureas). It is noteworthy that whilst medication-related QIs which described adherence to the guideline may be considered as process indicators for physicians, they are considered outcome indicators for pharmacists. If pharmacists detect and reduce potentially inappropriate medications among older people by making a recommendation to a prescriber, adherence to the guideline could be improved. Indeed, a similar QI 'percentage of cardiovascular patients with concomitant statin use' has been used by community pharmacists in the Netherland as an outcome indicator [52].
We acknowledge that this study has some strengths and limitations. This study involved panellists with expertise in primary care and medication safety from different backgrounds including medical doctors and pharmacists. However, other healthcare professionals such as nurses were not included. Moreover, the QI characteristics were described using ATC classification system, but not all medicines are included in the ATC system (e.g. yokukansan: Japanese traditional medicine). In addition, the QI set was developed for the Japanese context as it was based on recommendations from national guidance similar to guideline recommendations. As geriatric pharmacotherapy guidelines can vary from country to country, these QIs may not be applicable to other countries. However, we believe that the concept and challenges for the appropriate use of medicines in older people are similar. Therefore, this QI set designed for Japanese pharmacies could also be of value to pharmacists in primary care in other countries.

Conclusion
A set of 134 QIs for monitoring and evaluating geriatric pharmacotherapy by community pharmacists in primary care was rigorously developed. This QIs set could provide specific data to inform strategies to optimise geriatric patient care by community pharmacists in Japan. The measurement properties of QIs will be further evaluated for feasibility, applicability, room for improvement, sensitivity to change, predictive validity, acceptability and implementation issues.