Development and validation of PART (Pharmacotherapy Assessment in Renal Transplant Patients) criteria to assess drug‐related problems in an outpatient renal transplant population: A cross‐sectional study

Abstract Kidney transplant recipients are at risk of pharmacological interactions and adverse drug reactions. Community pharmacists are uniquely poised to detect and intervene in cases of drug‐related problems. The aims of this study were to develop and validate a list of explicit criteria to be used by community pharmacists to assess drug‐related problems in kidney transplant patients, and to assess their frequency and their determinants. First, we used a modified RAND method where a panel of experts established the PART (Pharmacotherapy Assessment in Renal Transplant Patient) criteria. Then, we performed a cross‐sectional study in which we applied the PART criteria to 97 prevalent kidney transplant recipients followed at a single university‐affiliated center. The final list of PART criteria included 70 drug‐related problems and was reliable (kappa: 0.88). An average of 1.2 drug‐related problems per patient was detected when the PART criteria were applied, with 68% of patients having at least 1 problem. This figure was 1.4 per patient using the expert judgment of renal transplant pharmacists who had no access to the PART list. The total number of medications taken was the only factor associated with the number of drug‐related problems (β: 0.27 for an increase of five medications, 95% CI 0.005, 0.547). The PART criteria provide a novel tool for community pharmacists to systematically detect drug‐related problems in kidney transplant recipients.

of drug-related problems (β: 0.27 for an increase of five medications, 95% CI 0.005, 0.547). The PART criteria provide a novel tool for community pharmacists to systematically detect drug-related problems in kidney transplant recipients.

K E Y W O R D S
drug-related problems, kidney transplantation, quality improvement

| INTRODUCTION
Renal transplantation is the optimal treatment modality for patients with end-stage renal disease, improving both quality of life and life expectancy compared to dialysis. 1,2 However, transplantation entails significant changes in pharmacotherapy for patients, with complex medication regimen including multiple immunosuppressive agents that have pharmacokinetic and pharmacodynamic interactions with other drugs, frequent adverse events and a narrow therapeutic index. 3 After the early post-transplant period, patients are regularly returned to their referring center and may have limited access to expert transplant pharmacists who would rapidly intervene when problems occur with the medication regimen. 4 This is particularly relevant as transplant patients are at risk for serious adverse events such as graft rejection, opportunistic infections, hospitalization, and drug toxicity due to errors in their pharmacotherapy. 3 Furthermore, low adherence remains problematic in many cases, as 20% to 70% of kidney transplant recipients (KTR) are not totally adherent to their immunosuppressant therapy, increasing the risk of rejection. 5 Hence, the risk of drug-related problems (DRPs) is elevated in this patient population. A DRP is an event or circumstance involving drug therapy that actually or potentially interferes with desired health outcomes. 6,7 Community pharmacists are accessible for consultation and follow patients regularly. They are thus uniquely poised to improve care for KTR. Community pharmacists review the pharmacological profile of their patients with each medication renewal and can intervene quickly by notifying the transplant center, the primary care provider or the patient if a DRP arises. 8 However, given the relative scarcity of organ transplant recipients in the general population and the absence of extensive training on the topic in their curriculum, most community pharmacists lack the expertise necessary to detect DRPs in transplant patients accurately and rapidly.
Although decision tools to help evaluate pharmacotherapy in older adults and in patients with chronic kidney disease (CKD) have been created, [9][10][11] no such instrument exists for KTR. Therefore, the aims of this study were to develop and validate a list of explicit criteria to be used by community pharmacists to assess DRPs in KTRs, to determine the number and types of DRPs, and identify factors associated with the presence of DRPs in this patient population.
Although future studies will be needed to assess the use of our tool in community pharmacies, here we report the first step of a quality improvement process where the ultimate goal is to improve the quality of pharmacotherapy in KTR while providing community-based services. The PART criteria were developed following a modified RAND appropriateness method. 12 In the first phase, a list of potentially important DRPs for KTR was created through a review of the literature and consultation with transplant nephrologists and pharmacists.
A short summary of the available evidence and supporting references were provided for each DRP.

| Phase 2
A panel of experts including four transplant nephrologists from two different adult transplant centers in Montreal, four renal transplant pharmacists from four different transplant centers in Montreal and Ottawa, two community pharmacists and two family physicians from the region of Montreal was assembled to evaluate the appropriateness of each potential DRP from the initial list. First, in March 2016, an electronic survey including each DRP and a short bibliography was sent to the panel. The experts were asked to evaluate the clinical appropriateness of each DRP on a Likert scale ranging from 1 to 9 (1 being totally inappropriate and 9 being totally appropriate). Experts were allowed to propose amendments to existing DRPs and propose new DRPs if they so wished. The median score of each DRP was calculated and DRPs were classified as "appropriate" (median score of 7 to 9), "uncertain" (median score of 4 to 6) and "inappropriate" (median score of 1 to 3).
The experts then attended a 4-hour panel session at the Centre Hospitalier de l'Université de Montréal (CHUM) in May 2016 to determine the final PART criteria. Using data from the initial survey, DRPs that were classified as "inappropriate" were rejected and those classified as "appropriate" passed directly to the final round of discussion. DRPs classified as "uncertain" were included in a first round of panel discussion. The clinical relevance of each "uncertain" DRP was discussed and was voted on by show of hands, with at least seven votes required to pass on to the final round of discussion. All amended and new DRPs that were proposed in the survey were also discussed and rescored by each expert, using the same Likert scale.
New and amended DRPs were again classified according to their median scores, and those classified as "appropriate" passed on to the final round of discussion. In the final round, all DRPs that were retained in previous steps were discussed with the perspective of selecting those that are relevant to the practice of community pharmacists. Therefore, the community pharmacists were asked to examine whether they thought the detection and management of each DRP was possible in their everyday practice. If the pharmacists answered affirmatively, a short discussion took place, followed by a vote for the inclusion of each DRP in the final list. DRPs that obtained at least seven votes were included in the final PART criteria. To determine the prevalence of DRPs in our population of KTR and test the reliability and the validity of the PART criteria, we performed a cross-sectional study.

| Study design, patients, and setting
In this cross-sectional study, all KTR who had a functioning transplant for more than 1 year and were followed in a Canadian adult transplant center (CHUM) were recruited at the time of their regular follow-up visit in the outpatient clinic between February and May 2016.
Patients who were younger than 18, hospitalized, who had returned to dialysis, who had a nonrenal organ transplant, who could not provide consent or who did not speak English or French were excluded.
All patients who were eligible and agreed to participate in the project were interviewed by one of the investigators (SA, QD, LER, AM) using a locally developed questionnaire. The investigator also collected data from the patient's electronic health record and obtained a detailed medication list from the patient's community pharmacy. The study was approved by the local institutional review board (CE 15.314) and written consent was obtained for each patient.

| Measurements
A complete assessment of pharmacological therapy was performed, retrieving the name and dose of each current medication used through chart review, in person patient semi-directive interview and contact with the community pharmacy. We also collected data on age, sex, weight and height, history of previous renal transplantations, smoking and alcohol consumption, patient reported comorbidities, medication side effects, allergies, mode of medication dispensation, arterial blood pressure and capillary blood glucose values. Other laboratory results were obtained from the last available values in the year prior to the study visit in the patient's electronic health record, including serum creatinine, low-density lipoprotein cholesterol (LDL-C), total cholesterol, glycated hemoglobin, uric acid, total corrected calcium, serum phosphorus, and hemoglobin. Creatinine clearance was calculated using the Cockroft-Gault formula. 13 For each patient, all the collected information was gathered in a research file.

Prevalence of DRPs using the PART criteria
Once interviews were completed, the patient research files were denominalized and each file was evaluated by two independent pharmacy residents (SA, LER) who had not interviewed the patients and who had minimal experience in the field of renal transplantation.
Each patient's research file was reviewed using the PART criteria to assess the number of DRPs. After they had independently coded for the presence of each DRP, the two investigators met to reach a consensus if there were discrepancies regarding their evaluation. DRPs were deemed "applicable" if they were relevant patient's clinical situations and pharmacotherapies. If a DRP was not pertinent for a given patient, it was labeled as "not applicable". For instance, a DRP pertaining to a hypoglycemic agent in a nondiabetic patient or in a diabetic who is not using hypoglycemic agents would be labeled as "not applicable" for this patient and not be scored.

| Study size and statistical analyses
We wished to build a reliable tool with substantial to almost perfect agreement. Based on our previous work, 11 we expected a global kappa statistic of 85%. To obtain a 95% confidence interval whose full range remained in the substantial to almost perfect category (0.78-0.90), 90 patients needed to be recruited.
Continuous variables are reported as means and standard deviations (or median and interquartile ranges where appropriate) and categorical variables are summarized as proportions. The number of patients with at least one DRP in each category of DRPs is reported as a proportion and 95% confidence interval. We provide both data originating from the consensus of the two evaluators (using the PART criteria) and from the transplant pharmacist's implicit judgment (not using the PART criteria).
We assessed the inter-rater reliability of the PART criteria (whole list and each category separately) through kappa statistics, with their 95% confidence intervals. To identify patient or therapeutic characteristics associated with the number of DRPs per patient, we performed univariate and multivariate linear regression. Statistical analyses were performed using SAS version 9.3, Cary, NC, USA).    pharmacy. The characteristics pertaining to patient medications are presented in Table 2.  (Table 3). The only two categories that did not obtain a kappa above 0.8 were in the "drug interactions-immunosuppressant treatment and other" and "adjustment for renal function" categories (kappa statistics: 0.66 in both instances).

| Determinants of DRPs using the PART criteria
We sought to determine whether patient or therapy-related characteristics were associated with the number of DRPs. In multivariate analysis, the total number of medications taken was associated with the number of DRPs (β: 0.27 for an increase of five medications, 95% CI 0.005, 0.547). There were no statistically significant association between the other patient/medication characteristics and the number of DRPs (Table 5). We found no significant predictor of DRPs as identified by transplant pharmacists in regression analyses.

| DISCUSSION
We developed and validated a set of criteria that can be used by the community pharmacists to assess clinically relevant DRPs in KTR.   Adverse events of immunosuppressant treatment 12 9 Need for osteoporosis prophylaxis 4 3 nonadherence DRPs that we identified in our population. We will upgrade the PART criteria with DRPs for nonadherence to osteoporosis prophylaxis, abdominal and neurological side effects, as these DRPs were identified by the experienced transplant pharmacist only and could easily be evaluated and relevant for community pharmacists.
The unexperienced PART evaluators identified 29 PART-DRPs that had not been uncovered by the explicit judgment of experienced transplant pharmacists who used no checklist to perform their evaluation. These DRPs were found in all categories. This is consistent with the improved performance associated with using checklists, an observation that has been reported in the evaluation of pharmacotherapy 15 but also in other clinical contexts such as surgery 16 and the management of acute kidney injury. 17 The mean number of DRPs we identified in our cohort of prevalent kidney transplant recipients was less than half of the figures that our group previously reported in patients with chronic kidney disease (eGFR≤60 mL/min/1.73 m 2 ) who were followed in renal protection clinics in the same region, but not the same center. Although the PART criteria were different than the ones that were elaborated for patients with CKD, 11 we used a similar methodology to develop the transplant-specific tool. We hypothesize that these differences are due to closer follow-up and contact between kidney transplant recipients and their health providers when compared to patients with stage 3-4 CKD in native kidneys.
Our study has limitations. First, the prevalence of DRPs detected in our cohort can be underestimated because we recruited patients who are regularly seen at the transplant center outpatient clinic. In the province of Quebec, this is the standard of care as most kidney transplant recipients are followed by their transplant centers and are not returned to general nephrology units while they have a functioning graft. Second, in the context of the study, the PART criteria only took 6 minutes to fill. However, because all the data necessary to apply the criteria had already been collected, the process of evaluating a patient may take a longer time in community pharmacies. This may limit the clinical usefulness of a "paper-based" tool in the real-life setting.
As recent changes in the patients' pharmacologic profiles were not recorded at the time of data collection, it is possible that we overestimated the prevalence of DRPs associated with hypertension control-a patient deemed not controlled could become adequately controlled within a short subsequent period if recent modifications to the antihypertensive regimen were made. However, as neither the expert transplant pharmacists nor the PART evaluators had access to this information, we do not believe that this had an impact on our evaluation of the validity or the reliability of the PART criteria.
In conclusion, the list of PART criteria represents a tool that enables detection of clinically significant DRPs by community pharmacists in KTR. Although this tool cannot replace the implicit judgment of experienced transplant pharmacists, accessibility to the latter may be limited. Community pharmacists offer proximity services to patients and are uniquely poised to detect DRPs. The PART criteria can reliably aid community pharmacists to better evaluate the pharmacotherapy of renal transplant patients and intervene in order to ensure that they have proper follow up. We are now planning on developing and validating a clinical pharmacy software integrating PART criteria-based alarms at the level of community pharmacies to assess the acceptability and usefulness of this tool in the real-life setting.

ACKNOWLEDG EMENTS
We thank all members of the expert panel for the modified RAND