Improving medication safety and prescribing of higher-risk medications in individuals with chronic kidney disease: A validation study

Abstract

Background:

Chronic kidney disease (CKD) affects 1 in 10 Canadians. Medications cleared by the kidneys can be harmful if dosed improperly. Community pharmacists are well-positioned to optimize prescribing, but inconsistencies between medication resources can complicate dosing. This study developed and validated higher-risk medication toolkits, including decision support algorithms for community pharmacists managing people with CKD.

Methods:

Fifty-one toolkits and algorithms were developed by team experts using Lynn’s method (domain identification, item generation per domain, and instrument formation). Team experts followed by community pharmacists rated toolkit content and algorithm face validity using a 2-part questionnaire with Likert scales. Each toolkit was validated by 5 to 6 participants over 2 rounds. Content validity was computed using an item-level content validity index (I-CVI) and scale-level content validity index (S-CVI/Ave) per round. Face validity calculated percentages for level of agreement to 5 statements. Community pharmacist interviews were conducted after each round, data analyzed, and toolkit revisions were made between rounds.

Results:

Twenty-two team experts validated 51 toolkits in 2 rounds between August and September 2024. Toolkit I-CVI, S-CVI/Ave, and face validity per algorithm ranged from 0.5 to 1, 0.87 to 1, and 49% to 100%, respectively. Thirteen toolkits were excluded from the community pharmacist validation. In 2 additional rounds, 23 community pharmacists, with 13.7 ± 9.1 years of experience, validated 38 medication toolkits between October and December 2024. Toolkit I-CVI and S-CVI/Ave and face validity per algorithm ranged from 0.83 to 1 and from 0.87 to 1, which met the content validity threshold of 0.83 to 1 (P < 0.05) for at least 5 to 6 participants per round. Participants’ overall agreement for the face validity statements ranged from 75% to 100%, which was above the prespecified threshold of 70% for face validity consensus.

Conclusions:

Thirty-eight toolkits achieved high content and face validity. Future research will integrate them into a digital tool and assess their effectiveness and safety in community pharmacy practice in people with CKD.

Knowledge Into Practice.

• Inconsistent recommendations across drug resources complicate drug-dosing decisions in individuals with an estimated glomerular filtration rate below 30 mL/min, which can lead to harmful side effects or kidney damage. Validated drug-dosing decision support algorithms of these medications in community pharmacy practice are lacking.

• This validation study provides evidence and expert-informed toolkits of higher-risk medications to support community pharmacists in optimizing prescribing in this population.

• Validated medication toolkits and dosing algorithms may improve medication safety and support dose adjustment of higher-risk medications in advanced chronic kidney disease, with their impact to be evaluated in a community pharmacy implementation study using a digital health application in July 2025.

Mise En Pratique Des Connaissances.

• Des recommandations incohérentes entre les différentes sources d’information sur les médicaments compliquent les décisions relatives à la posologie chez les personnes dont le débit de filtration glomérulaire estimé est inférieur à 30 ml/min, ce qui peut entraîner des effets secondaires néfastes ou des lésions rénales. Il n’existe pas ou très peu d’algorithmes validés d’aide à la décision en matière de posologie de ces médicaments dans l’exercice de la pharmacie communautaire.

• Cette étude de validation fournit des preuves et des trousses d’outils élaborées par des experts sur les médicaments à haut risque afin d’aider les pharmaciens communautaires à optimiser la prescription au sein de cette population.

• Des trousses d’outils validées sur les médicaments et des algorithmes de posologie pourraient permettre d’accroître la sûreté des médicaments et de faciliter l’ajustement posologique des médicaments à haut risque pour les cas de néphropathie chronique avancée. Leur impact sera évalué dans le cadre d’une étude de mise en œuvre en pharmacie communautaire à l’aide d’une application de santé numérique en juillet 2025.

Introduction

Chronic kidney disease (CKD) affects 1 in 10 Canadians. ¹ It is categorized by abnormalities of kidney structure or a sustained (greater than 3 months) reduction of estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73 m² and/or urine albumin-to-creatinine ratio greater than or equal to 3 mg/mmol. ² It is commonly associated with multiple comorbidities, advanced age, and the use of multiple medications. ³ Since many drugs are eliminated by the kidneys, incorrect dosing can lead to drug accumulation.^4,5 As a result, individuals with CKD are at increased risk for adverse drug events, including further kidney damage.^6-9 A study published in 2024 identified that eGFR is a risk factor for serious adverse drug reactions, with the risk significantly higher in individuals with an eGFR <30 mL/min/1.73 m² compared with those with an eGFR ≥30 mL/min/1.73 m². ¹⁰

Although numerous medication resources are available to guide dosing, improper prescribing continues to result in negative health outcomes.^11-17 Variations in resources on drug dosing in those with CKD complicate adjustment decisions.^18,19 In a 2024 qualitative study interviewing community pharmacists, barriers to drug dosing included inconsistencies with resource recommendations, discrepancies with kidney function category cut-points, and the use of different equations for estimating kidney function. ²⁰ Further, pharmacists voiced the importance of having a concise, standardized list of higher-risk medications to assess, along with an evidence-based, expert-informed drug-dosing tool to aid in adjusting doses for individuals with CKD, especially those with low kidney function. ²⁰

Community pharmacists are in an ideal position to enhance safety and prescribing for individuals with CKD. The prevalence of CKD in Canadian primary care is higher in rural areas compared with urban settings. ²¹ In most Canadian jurisdictions, pharmacists are authorized to adjust medication doses or regimens. ²² Innovative approaches are needed to improve medication safety for those with CKD in a primary care setting. The aim of this study was to develop and validate medication toolkits, each with a decision support algorithm, for dose adjustment of higher-risk medications in CKD for community pharmacists.

Methods

This validation study was conducted in 2 phases. In phase 1, toolkits for higher-risk drugs were created to assist community pharmacists to adjust or avoid these medications in individuals with CKD. In phase 2, each medication toolkit that included a decision support algorithm was validated by team experts and community pharmacists. The study was approved by the hospital Research Ethics Board (study file No. 1030689).

Phase 1: Team development and validation of the medication toolkits and decision support algorithms

The development of the 51 medication toolkits and algorithms was informed by our qualitative study. ²⁰ To determine which medications to include for the toolkit and algorithm development, we conducted a scoping review and a modified Delphi process to identify relevant higher-risk or commonly used medications in Canadian community pharmacy practices that require dose adjustments for individuals with CKD. ²³ To identify the best evidence on safety risks associated with the selected medications, tertiary drug resources were reviewed, and a medical librarian developed search strategies for the Ovid MEDLINE and Embase databases up to June 2024. Each medication toolkit included 5 items: a decision support algorithm, a summary of potential harms, dosing, evidence tables, and references. The decision support algorithms were developed and revised using Lynn’s 3-step method (domain identification, item generation per domain, and instrument formation) ²⁴ by an expert panel of pharmacists, nephrologists, lived-experience patient partners, an implementation scientist, policy makers, and information system specialists. These experts identified 3 key domains for the decision support algorithms: kidney function assessment, dosing adaptations, and monitoring (including laboratory values such as serum creatinine, eGFR, drug levels, and physical assessment). For each domain, relevant evidence-based items were sourced from drug product monographs, drug databases, literature, and guidelines. The research team met virtually to develop the decision support algorithms. A sequential algorithm format, which follows a linear series of steps, was preferred by the team. ²⁵ Decision nodes were used, beginning with the initial node based on kidney function or eGFR in mL/min, which led to dosing recommendation and/or additional monitoring. All algorithms used eGFR, adjusted for an individual’s body surface area instead of Cockcroft-Gault equation for estimating creatinine clearance (C-G eCrCL), as it is the preferred method for making drug-dosing decisions.^2,26,27 Some algorithms were customized for specific indications, such as rivaroxaban algorithms for atrial fibrillation and venous thromboembolism. The first draft of the 51 medication toolkits, each including a decision support algorithm, underwent 2 rounds of validation and revision with team experts. Each toolkit was validated by 5 to 6 team experts per round. Consensus on the final toolkits was reached using the methods outlined below for community pharmacist content and face validation.

Phase 2: Community pharmacist validation of the medication toolkits and decision support algorithms

Community pharmacists were recruited using an advertisement sent by the provincial pharmacy association and via snowball sampling. Five to 6 participants per medication toolkit were needed to minimize the risk of chance agreement based on previous teamwork (M.B., J.W).^28-30 A minimum of 2 rounds of validation was planned, with revisions made by the research team between each round.

The validation of the medication toolkits followed the approach outlined by Lynn’s method. ²⁴ A 2-part questionnaire was created to validate the toolkit content (e.g., appropriateness, accuracy, completeness of content) and decision support algorithm face validity (e.g., clarity, comprehensibility, and appropriateness to the user) based on Feinstein’s concept of clinical sensibility ³¹ (Appendix 1, available online under Supplementary Materials). Each medication toolkit was divided into the following items: A (decision support algorithm), B (summary of potential harms), C (dosing table), D (evidence table), and E (references) to align with the 2-part questionnaire (Figure 1). Participants were sent the 2-part questionnaire and medication toolkits via email and given 10 to 14 days to complete their questionnaire. After the questionnaires were returned, a brief one-on-one virtual interview was conducted by a team member (J.W., K.H.) with each community pharmacist to review their ratings and comments.

Figure 1.

Medication toolkit example

Content validation

Participants were asked to independently rate each toolkit item (A to E) using a 4-point Likert scale (1 = strongly disagree [unacceptable, remove], 2 = disagree [unacceptable, major revisions needed], 3 = agree [acceptable with minor revisions], 4 = strongly agree [acceptable as is]). They were also encouraged to provide comments on areas requiring revisions. Any rating of 1 or 2 by 1 or more participants necessitated revisions before proceeding to the next round.

Face validation

Participants were asked to assess 5 face validity statements about each algorithm using a 5-point scale (1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = strongly agree). The statements covered aspects such as the clarity and understandability of the algorithm, the appropriateness of its language, its logical flow, and whether they would use or recommend it in their practice. A sixth statement asked if the medication should be included in the future development of an electronic drug-dosing and decision support kidney (eDoseCKD) tool tailored for community pharmacists. The ratings for the sixth statement were not included in the determination of algorithm face validity but informed which drugs to exclude. Participants were also asked to provide feedback for improving the algorithms. Any statement rated 1 to 3 by 1 or more participants required revisions.

Statistical analysis

The content validity index (CVI) was calculated using both the item-level CVI (I-CVI) and scale-level CVI (S-CVI/Ave) based on the average method for each round. ²⁴ The I-CVI represents the proportion of participants who rated an item (A to E) as 3 or 4. According to Lynn and others, for 6 participants in a round, at least 5 must rate the item 3 or 4 to establish content validity at P < 0.05 (I-CVI ≥ 0.83).^24,32-34 An item is considered invalid if 2 of 6 participants rate it 1 or 2 (I-CVI ≤ 0.67).^24,32-34 For a panel of 5 participants in a round, all must rate the item 3 or 4 to establish content validity at P < 0.05 (I-CVI of 1). ²⁴ The S-CVI/Ave is the average of all I-CVI scores for items (A to E) for each medication toolkit in each round. In addition, we calculated the overall S-CVI/Ave across 2 rounds. An S-CVI/Ave of ≥0.9 indicates excellent content validity for the medication toolkits.^32-34

For face validity, percentages were calculated for participants’ level of agreement to 5 face validity statements. Ratings of 4 or 5 were considered valid, while ratings of 1 to 3 were not. A consensus threshold was set at a minimum of 70% agreement, meaning at least 70% of participants must rate the statement as agree or strongly agree (4 or 5). This threshold is commonly used in the Delphi technique.^28-30,35-37

Medication toolkits and algorithms that did not meet both content and face validity were revised before the next round. The interview data were analyzed using qualitative descriptive analysis, as outlined by Sandelowski.^38,39 This method focuses on summarizing the content of the data (what participants said) rather than interpreting the underlying meaning (what participants meant). This approach, which involves minimal inference, is particularly effective for capturing pharmacists’ perspectives on the toolkit’s content and algorithm face validity, including usability and areas for improvement.

Results

Team experts’ validation of medication toolkits and decision support algorithms

A team of 22 experts in pharmacy, nephrology, implementation science, informatics, and patient partners with lived experience validated 51 medication toolkits, each with a decision support algorithm, through 2 rounds of content and face validation between August and September 2024. Thirty-four medications achieved content (I-CVI ≥ 0.83, P < 0.05 for at least 6 participants) and face validation (70% agreement or higher on 6 statements) in round 1 and 17 medications received validation after round 2. In rounds 1 and 2, the I-CVI for items A to E ranged from 0.5–1 and S-CVI/Ave for the medication toolkits ranged from 0.87–1 (Table 1 and Appendix 2, available online under Supplementary Materials). The overall S-CVI/Ave across 2 rounds for the medication toolkit was 0.9 to 1. The overall percentage of participants who agreed or strongly agreed to 5 face validity statements (average across 2 rounds) ranged from 49% to 100% (Table 2 and Appendix 3, available online under Supplementary Materials). Tool revisions were made between rounds, and the changes to the medication toolkits are available in Appendix 4, available online under Supplementary Materials. During face validation, team experts identified 13 medications to exclude from the community pharmacist validation. These medications either received a participant rating of ≤75% for inclusion in the future electronic community pharmacist tool (bezafibrate, febuxostat, famciclovir, nirmatrelvir/ritonavir, emtricitabine/tenofovir, norfloxacin, tolterodine, solifenacin, sotalol) or were excluded due to preference or complexity (escitalopram, bupropion, venlafaxine, duloxetine).

Table 1.

Team medication toolkit content validity scores for round 1 and round 2

Round(s)		Metformin	Glyburide	Saxagliptin	Sitagliptin	Bezafibrate	Fenofibrate	Rosuvastatin	Gabapentin		Pregabalin		Topiramate	Allopurinol	Colchicine	Febuxostat	NSAIDs	Morphine	Codeine	Tramadol	Apixaban		Edoxaban		Rivaroxaban
		1	1	1	1	1	1	1	1	2	1	2	1	1	1	1	1	1	1	1	1	2	1	2	1
Algorithm I-CVI	A	1	1	0.83	1	1	1	0.83	0.67	1	0.67	1	0.83	1	1	1	1	1	1	0.83	0.67	1	0.83	1	0.83
	B	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	0.83	1	1	1	1
	C	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	0.83	1	1
	D	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	0.83	1	1	1	1	1
	E	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
S-CVI/AVE		1	1	0.97	1	1	1	0.97	0.93	1	0.93	1	0.97	1	1	1	1	1	1	0.93	0.9	1	0.93	1	0.97
Overall *		1	1	0.97	1	1	1	0.97	0.96		0.96		0.97	1	1	1	1	1	1	0.93	0.95		0.96		0.97

I-CVI, item-level content validity index; NSAIDs, non-steroidal anti-inflammatory drugs; S-CVI/AVE, scale-level content validity index based on the average method; TDF, tenofovir disoproxil fumarate; TMP/SMX, trimethoprim/sulfamethoxazole.

^*Overall S-CVI/AVE across rounds 1 and/or 2. Item A, medication algorithm; item B, summary of potential harms; item C, dosing table; item D, evidence table; item E, references.

Table 2.

Team level of agreement with face validity statements per algorithm across 2 rounds

	Metformin	Glyburide	Saxagliptin	Sitagliptin	Bezafibrate	Fenofibrate	Rosuvastatin	Gabapentin	Pregabalin	Topiramate	Allopurinol	Colchicine	Febuxostat	NSAIDs	Morphine	Codeine	Tramadol
	Percentage of participants who agreed or strongly agreed
	n = 7	n = 7	n = 7	n = 7	n = 7	n = 7	n = 7	n = 12	n = 12	n = 6	n = 7	n = 7	n = 7	n = 8	n = 7	n = 7	n = 6
Q1	100	100	71	100	100	86	86	76	92	100	86	100	100	100	100	100	100
Q2	100	100	100	100	100	100	100	100	100	83	100	100	100	100	100	100	100
Q3	100	100	71	100	100	100	100	90	90	100	100	100	100	100	100	100	100
Q4	83	100	86	100	100	100	100	90	90	83	100	100	100	100	100	100	100
Q5	100	100	86	100	100	100	100	90	90	83	100	100	100	100	100	100	83
Q6	100	100	86	100	57	100	100	100	100	83	100	100	71	100	100	100	100
	Apixaban	Edoxaban	Rivaroxaban	Dabigatran	Dalteparin	Enoxaparin	Acyclovir	Valacyclovir	Famciclovir	Oseltamivir	Nirmatrelvir/ ritonavir	Emtricitabine/ TDF	Amoxicillin/ clavulanic acid	Ciprofloxacin	Levofloxacin	Norfloxacin	TMP/SMX
	Percentage of participants who agreed or strongly agreed
	n = 13	n = 13	n = 8	n = 13	n = 6	n = 6	n = 13	n = 8	n = 8	n = 8	n = 11	n = 11	n = 11	n = 8	n = 8	n = 6	n = 7
Q1	75	88	75	88	83	100	88	75	75	75	92	84	49	88	75	100	86
Q2	94	100	88	94	83	100	94	88	100	100	100	92	74	88	88	100	100
Q3	75	82	75	88	83	100	88	75	75	75	92	84	65	88	75	100	100
Q4	82	88	88	82	100	100	72	88	75	88	100	84	92	100	88	100	100
Q5	72	88	75	78	100	100	82	88	75	75	100	92	84	88	75	100	86
Q6	94	90	100	84	100	100	88	88	75	88	64	55	100	100	100	67	100
	Nitrofurantoin	Clarithromycin	Fluconazole	Bupropion	Escitalopram	Venlafaxine	Duloxetine	Lithium	Tolterodine	Solifenacin	Ranitidine	Famotidine	Methotrexate	Sotalol	Digoxin	Baclofen	Varenicline
	Percentage of participants who agreed or strongly agreed
	n = 7	n = 6	n = 11	n = 11	n = 11	n = 11	n = 11	n = 11	n = 7	n = 7	n = 7	n = 7	n = 11	n = 11	n = 7	n = 7	n = 7
Q1	100	100	84	92	50	67	57	92	100	100	100	100	74	82	100	100	100
Q2	86	100	82	84	75	75	84	100	100	100	100	100	82	82	100	100	100
Q3	100	100	90	92	67	75	75	100	100	100	100	100	100	82	86	100	100
Q4	86	83	100	100	82	84	100	90	83	83	100	100	84	82	100	100	100
Q5	86	83	100	100	74	84	82	82	100	100	100	100	75	65	100	100	100
Q6	86	100	100	92	100	100	100	90	71	71	100	100	84	75	100	100	100

n, number of team members; NSAIDs, non-steroidal anti-inflammatory drugs; Q1, the algorithm is clear and understandable; Q2, the algorithm uses appropriate language and wording; Q3, the algorithm flows in a logical manner; Q4, I would use this algorithm in my own practice; Q5, I would be confident recommending the use of this algorithm; Q6, should (medication) be included in electronic drug dosing decision support kidney (eDoseCKD) tool; TDF, tenofovir disoproxil fumarate; TMP/SMX, trimethoprim/sulfamethoxazole.

Community pharmacist validation of medication toolkits and decision support algorithms

Twenty-three community pharmacists participated from both urban (65%) and rural (35%) pharmacies across the province. Most (78%) participants were female. The mean ± standard deviation (SD) for participant age was 40 ± 8 years with 13.7 ± 9.1 years of experience (Table 3). Most of the participants (87%) had a bachelor of science degree in pharmacy as their highest academic qualification. The mean ± SD duration of the debrief interview was 25 ± 7 minutes.

Table 3.

Community pharmacists’ demographic and practice characteristics

Characteristic	Participants (N = 23)
Sex, female, n (%)	18 (78.3)
Age, years (mean ± SD)	40.17 ± 8.26
Years of practice (mean ± SD)	13.74 ± 9.14
Highest academic credential, n (%)
Bachelor of science in pharmacy	20 (86.96)
PharmD	2 (8.70)
Master’s degree	1 (4.35)
Location, n (%)
Urban	15 (65.22)
Rural	8 (34.78)
Area of practice, n (%)
Community pharmacy*	23 (100)
No. prescriptions per day (mean ± SD)	218.57 ± 107.33

PharmD, doctor of pharmacy; SD, standard deviation.

^*Includes 8 pharmacists practicing in both community pharmacy and community primary care settings.

Content and face validation

Two rounds of content and face validation were undertaken for 38 medication toolkits between October and December 2024. Thirty-three medications received content and face validation in round 1 and 5 medications in round 2 (saxagliptin, non-steroidal anti-inflammatory drugs [NSAIDs], digoxin, baclofen, varenicline). In round 1, there were 3 groups of 6 pharmacists (18 total), each reviewing 12 to 14 different medication toolkits (38 medications total), while round 2 had 1 group of 5 pharmacists reviewing 5 medication toolkits. Interviews were conducted between rounds with the 23 participants, and tool revisions were made after each round (Appendix 5, available online under Supplementary Materials). In both rounds, the I-CVI of each item (A to E) ranged from 0.83 to 1, which met the threshold for content validity for 5 to 6 participants (Table 4 and Appendix 6, available online under Supplementary Materials). The S-CVI/Ave for rounds 1 and 2 ranged from 0.87 to 1 and from 1 to 1, respectively. The overall S-CVI/Ave across the 2 rounds for the medication toolkits was ≥0.9, except for sitagliptin, which had a score of 0.87. Therefore, community pharmacists found the medication toolkits appropriate, accurate, and complete for drug-dosing adaptions in individuals with CKD.

Table 4.

Community pharmacist medication toolkit content validity scores in round 1 and round 2

Round(s)		Metformin	Glyburide	Saxagliptin		Sitagliptin	Fenofibrate	Rosuvastatin	Gabapentin	Pregabalin	Topiramate	Allopurinol	Colchicine	NSAIDs		Morphine	Codeine	Tramadol	Apixaban	Rivaroxaban	Edoxaban	Dabigatran	Dalteparin
		1	1	1	2	1	1	1	1	1	1	1	1	1	2	1	1	1	1	1	1	1	1
Algorithm I-CVI	A	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
	B	1	1	0.83	1	0.83	1	1	1	1	0.83	1	1	0.83	1	0.83	0.83	0.83	1	1	1	1	1
	C	1	1	0.83	1	0.83	1	1	1	1	1	0.83	0.83	1	1	1	1	0.83	1	1	1	1	1
	D	1	0.83	0.83	1	0.83	1	1	1	1	0.83	0.83	0.83	0.83	1	0.83	0.83	0.83	0.83	0.83	0.83	0.83	0.83
	E	0.83	0.83	1	1	0.83	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
S-CVI/AVE		0.97	0.93	0.9	1	0.87	1	1	1	1	0.93	0.93	0.93	0.93	1	0.93	0.93	0.9	0.97	0.97	0.97	0.97	0.97
Overall *		0.97	0.93	0.95		0.87	1	1	1	1	0.93	0.93	0.93	0.96		0.93	0.93	0.9	0.97	0.97	0.97	0.97	0.97

I-CVI, item-level content validity index; NSAIDs, non-steroidal anti-inflammatory drugs; S-CVI/AVE, scale-level content validity index based on the average method; TMP/SMX, trimethoprim/sulfamethoxazole.

^*Overall S-CVI/AVE across rounds 1 and/or 2. Item A, medication algorithm; item B, summary of potential harms; item C, dosing table; item D, evidence table; item E, references.

Five of the 38 medication algorithms did not achieve face validity in round 1 but were validated in round 2 (Table 5 and Appendix 7, available online under Supplementary Materials). The overall percentage of participants (average across 2 rounds) who agreed or strongly agreed with the 5 face validity statements for each medication algorithm ranged from 73% to 100% (Table 5). All algorithms were above the prespecified threshold for face validity consensus of 70% by round 2. Therefore, community pharmacists considered the algorithms clear, comprehensive, and appropriate for users. For statement 6 (across 2 rounds), should the algorithm be included in the future development of an electronic drug-dosing tool, all algorithms had greater than 75% agreement except for glyburide, saxagliptin, fenofibrate, topiramate, and nitrofurantoin.

Table 5.

Community pharmacist level of agreement with face validity statements per algorithm across 2 rounds

	Metformin	Glyburide	Saxagliptin	Sitagliptin	Fenofibrate	Rosuvastatin	Gabapentin	Pregabalin	Topiramate	Allopurinol	Colchicine	NSAIDs	Morphine	Codeine	Tramadol	Apixaban	Rivaroxaban	Edoxaban	Dabigatran
	Percentage of participants who agreed or strongly agreed
	n = 6	n = 6	n = 11	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 11	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6
Q1	100	100	100	100	100	100	100	100	100	100	100	92	83	100	100	100	100	100	100
Q2	100	100	100	100	100	100	100	100	100	100	100	100	83	100	100	100	100	100	100
Q3	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100
Q4	100	83	75	83	83	100	100	100	83	100	100	75	83	100	100	100	100	83	100
Q5	100	83	83	83	100	100	100	100	83	100	100	75	83	100	83	100	100	100	100
Q6	100	33	67	83	67	83	100	100	67	100	100	92	100	100	83	100	100	83	100
	Dalteparin	Enoxaparin	Acyclovir	Valacyclovir	Oseltamivir	Amoxicillin/ clavulanic acid	Ciprofloxacin	Levofloxacin	TMP/SMX	Clarithromycin	Nitrofurantoin	Fluconazole	Lithium	Ranitidine	Famotidine	Methotrexate	Digoxin	Baclofen	Varenicline
	Percentage of participants who agreed or strongly agreed
	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 6	n = 11	n = 11	n = 11
Q1	100	100	100	100	100	100	100	83	100	100	100	100	83	100	83	100	83	100	100
Q2	100	100	100	100	100	100	100	83	100	100	100	100	100	83	100	100	83	83	100
Q3	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100
Q4	100	83	100	100	100	100	100	100	100	100	100	100	83	83	83	100	83	100	92
Q5	100	100	100	100	100	100	100	83	100	100	100	100	83	83	83	83	73	92	83
Q6	100	83	83	100	83	100	100	100	100	100	67	83	100	83	83	100	100	100	73

n, number of team members; NSAIDs, non-steroidal anti-inflammatory drugs; Q1, the algorithm is clear and understandable; Q2, the algorithm uses appropriate language and wording; Q3, the algorithm flows in a logical manner; Q4, I would use this algorithm in my own practice; Q5, I would be confident recommending the use of this algorithm; Q6, should (medication) be included in electronic drug dosing decision support kidney (eDoseCKD) tool; TMP/SMX, trimethoprim/sulfamethoxazole.

Community pharmacist interview feedback

Consistent feedback emerged throughout the 2 rounds of content and face validation. Most comments focused on improving the formatting and wording of the medication toolkits and algorithms to enhance clarity and ease of use. For example, decision nodes for kidney function were revised to close-ended sentences, key words (e.g., kidney function, indications, route of administration) were bolded for emphasis, and additional details (e.g., monitoring, algorithm-specific indications for antivirals, digoxin, apixaban, rivaroxaban, and dalteparin, initial vs maintenance dosing, repositioning of safety images to align with highest risk eGFR categories <30 mL/min) were included in certain steps to reduce ambiguity and simplify workflow. During the team validation, the decision support algorithms were expanded to include all eGFR categories, not just those <30 mL/min. Where possible, it was recommended to use eGFR thresholds that aligned with CKD categories. Community pharmacists suggested making the decision support algorithms more comprehensive by adding usual and maximum doses, external links for calculating eGFR in mL/min (e.g., National Kidney Foundation eGFR calculator), lean body weight calculations for digoxin, and dosing recommendations for all NSAIDs, rather than a general class dosing statement. They also recommended grouping similar dosing recommendations together in the dosing tables to improve usability. In addition, they suggested adding footnotes to clarify that the decision support algorithm is not intended for dialysis patients and that eGFR in mL/min is preferred for kidney function assessment while the dosing table reflects the cited source recommendations, including both C-G eCrCL and eGFR. Information related to when not to use the decision support algorithm (e.g., when eGFR is not reliable in those with unstable kidney function or acute kidney injury) or when to refer or consult another prescriber (e.g., nephrology) was recommended to be included in future education and training.

The final versions of the 38 validated medication decision support algorithms are available in Appendix 8, available online under Supplementary Materials.

Discussion

We believe this is the first study to develop and validate medication toolkits, each with a decision support algorithm, to assist community pharmacists in making dosing adjustments for higher-risk medications in individuals with CKD. Content and face validation was achieved for 38 (75%) of the 51 developed medication toolkits. Although the toolkits, including decision support algorithms, have not yet been tested in the community pharmacy setting, their validation suggests potential for future application in guiding safer prescribing of higher-risk medications in those with CKD. Our previous qualitative study with community pharmacists highlighted the need for evidence-based drug-dosing algorithms, developed with experts, and featuring an up-to-date list of high-risk medications to adjust or avoid in community pharmacy practice. ²⁰

Many challenges arise in determining the correct dosing for individuals with CKD, often resulting in inappropriate prescribing that can lead to adverse drug reactions and poor health outcomes.^10-17 While drug resources are available to guide drug dosing, inconsistencies between these resources have been reported to complicate adjusting medications.^18-20 In addition, pharmacists have reported difficulties with dosing in CKD patients due to variations in kidney function estimation equations, differences in units, and cut-off points used in dosing recommendations across resources; as well, pharmacists may not know their patient has CKD and are still unable to order or access kidney function tests in many Canadian jurisdictions. ²⁰

Previous studies have focused on specific strategies to improve medication prescribing for individuals with CKD. The development and validation of the Pharmacotherapy Assessment in Chronic Renal Disease (PAIR) criteria focused on 50 clinically significant drug therapy problems that require intervention by a community pharmacist. ⁴⁰ While dosage-adjustment tables based on C-G eCrCl were provided for incorrect doses or contraindicated agents, the interventions also addressed treatment adherence and optimizing blood pressure control, among others. Building on the PAIR criteria, another study developed and validated a set of criteria to assess the severity of drug therapy problems in individuals with CKD, outlining specific interventions that community pharmacists should take with the patient and treating physician.^41,42 More recent studies have opted to provide a list of drugs or drug classes that require adjustment or avoidance in people with CKD, but these lists primarily serve as screening tools to identify patients who may benefit from a more detailed review. They generally do not include specific dosing recommendations.^15,43-45

Community pharmacists are well-positioned to optimize prescribing and mitigate harm from improperly prescribed medications. They can play a key role in drug stewardship in people with advanced CKD (see Box 1). Scope of practice continues to expand in Canada and our province allows community pharmacists to prescribe adaptations, such as modifying doses or regimens. ⁴⁶ Previous studies have reported the important role community pharmacists play in the detection of CKD and nephrotoxic drugs and the identification of drug therapy.^47-55 Our goal is to integrate the validated medication toolkits and decision support algorithms into an electronic drug-dosing and decision support kidney (eDoseCKD) tool to assist community pharmacists and evaluate its effectiveness. Few studies have evaluated kidney function-based drug-dosing clinical decision support systems outside of hospital settings, with many focusing on automatic alerts at the point of prescribing with limited dosing adjustment guidance.^56-62

Box 1.

Key concepts in drug stewardship in chronic kidney disease (CKD)

Encourage	Individuals to inform all prescribers they have CKD
Educate	Individuals that having CKD can affect medication dosing, including over-the-counter and natural health products
Explain	Medication benefits and possible risks so they can be identified and managed early
Assess	Kidney function for drug-dosing decisions
Convert	Laboratory estimated glomerular filtration rate (eGFR) to mL/min (in over- or underweight individuals)
Check	Two trusted drug resources and compare
Provide	Sick-day medication guidance to reduce the risk of acute kidney injury
THINK	Medications as kidney function declines

There are several potential limitations. First, while Lynn’s method is commonly employed for algorithm validation, a universally agreed-upon threshold for content validation does not exist. Second, the content experts from the research team who validated the medication toolkits and algorithms were also the developers, which could have introduced bias into their validation scores. However, the high I-CVI and S-CVI achieved across 2 rounds with the community pharmacists provide reassurance that the validation process was successfully completed. In addition, the decision support algorithms were designed to provide a static assessment when a new prescription or refill is made. Additional surveillance using community pharmacy software would be needed to allow for a more dynamic assessment. Fourth, the medication toolkits and algorithms currently do not include those on dialysis, but additional work is planned to address this population. Fifth, community pharmacists mentioned that they are not always aware if a patient has CKD. As a result, the future electronic tool and education and training will incorporate CKD screening criteria. Furthermore, not all comments could be included in the medication toolkits and algorithms but will be addressed in future education and training with the electronic tool. This includes topics such as understanding the nuances of eGFR in mL/min for drug dosing and assessing stable vs unstable kidney function. Finally, in our province, community pharmacists working in a community pharmacy primary care clinic are authorized to order blood work, while approval for blood work orders in community pharmacies is still pending. This may affect the use of the medication toolkits and algorithms in jurisdictions where pharmacists cannot order blood work.

This validation study has several strengths. First, a relevant list of medications was selected for the toolkits based on a recent scoping review and modified Delphi of Canadian pharmacists working in nephrology, geriatrics, and primary care. ²³ Many of the medications were high-volume or impact medications with significant risk for adverse events. Second, the development of the toolkits and algorithms was comprehensive and informed by evidence and team experts in pharmacy (including community pharmacists), nephrology, implementation science, informatics, professional associations, regulators, and patient partners with lived experience. Involving community pharmacists in the validation of these clinical algorithms will enhance their accuracy, practicality, and reliability, making them better suited to pharmacy practice and more likely to be adopted. Third, indication-based decision support algorithms were created for specific medications, such as antivirals, digoxin, and direct-acting oral anticoagulants, to improve the applicability of the algorithms. Fourth, the algorithms were developed with eGFR, adjusted for body surface area in mL/min to align with current recommendations.^2,26,27 Further, eGFR thresholds were aligned with CKD categories where feasible but also included the nuances of medication-specific eGFR cut-points. Lastly, this study included a balanced representation of urban and rural community pharmacists, each dispensing an average of about 200 prescriptions per day, which may support the adoption and scalability of the algorithms to other pharmacies.

Conclusion

Thirty-eight medication toolkits, each including a decision support algorithm, achieved content and face validity by community pharmacists. Community pharmacists are well-positioned to optimize prescribing of these higher-risk medications in individuals with CKD. Future research will focus on integrating the validated medication toolkits and algorithms into an electronic drug-dosing and decision support kidney (eDoseCKD) tool and assessing their effectiveness and safety in community pharmacy practice in people with CKD. ■