Development and Validation of a Knowledge, Attitude, and Practice Scale for Medication Safety in Patients with Chronic Kidney Disease: A Cross-Sectional Study

Purpose To develop a knowledge, attitude, and practice (KAP) scale for medication safety in patients with chronic kidney disease (CKD), test its reliability and validity, and provide a reliable assessment instrument for the self-medication of patients with CKD. Methods The initial scale was formulated based on the KAP theory together with a literature review, research group discussion and semi-structured interviews of patients with CKD. An expert panel consultation and pre-investigation were conducted to improve the scale. Next, 209 patients with CKD were enrolled to undertake the questionnaire linked to the primary scale. The item analysis, exploratory factor analysis (EFA), Cronbach’s alpha coecient and split-half reliability were used to evaluate the validity and reliability of the scale. and reliability coecient was 0.834 for the CKD, self-care; CPOE, computerized physician order entry system; CR, critical ratio; DI, drug interaction; DRP, drug-related problem; DWI, drug use without indication; EFA, exploratory factor analysis; ESRD, end-stage renal disease; ESRD-AQ,end-stage renal disease questionnaire; failure to receive drug; GMAS, general medication adherence scale; IDS, improper drug selection; IWD, indication without drug therapy; KAP, knowledge, attitude, and practice; KMO,Kaiser-Meyer-Olkin; MARS, medication MRP, Medication-related measure of OD, PIM, potentially inappropriate PSI, prescription RIM, renally inappropriate medication.


Introduction
Chronic kidney disease (CKD) is a huge problem in the public health system. The global prevalence of CKD is about 14.3%, and the overall prevalence rate of CKD in China is 10.8% [1] . Owing to increasing incidence of chronic comorbidity (5%~98%) [2,3] and the susceptibility to complications in patients with CKD [4] comorbidity, and polypharmacy (the numerical de nition of use of ve or more medications daily) [5] are very common in patients with CKD [6] . It is reported that 25% of patients with CKD have more than three diseases [7] . 59. 3% of them take multiple medications, and 11.4% are prescribed more than 10 kinds of drugs [8] . Previous studies have reported that in elderly patients with end-stage renal disease (ESRD), the rate of polypharmacy was 91%, and about 43% of these patients were found to have been prescribed more than 10 kinds of drugs [9] . Comorbidity and polypharmacy lead to many medication-related problems (MRPs) in patients with CKD, including indication without drug therapy (IWD), drug use without indication (DWI), improper drug selection (IDS), sub-clinical dose (UD), overdose (OD), adverse drug reaction (ADR), drug interaction (DI), failure to receive drug (FRD) owing to compliance issues, cost, or accessibility, and improper laboratory monitoring (Lab). 6 Previous studies have found that 17-80% of patients with CKD had low adherence and irregular medication behavior [10,11] . The overall incidence of ADRs in elderly patients with CKD is 3-10 times higher than that in the general population of elderly patients [12] . About 10-20% of patients with CKD have ADRs every year [13,14,15] . Moreover, of these patients, 45.2-77.0% had received renal inappropriate medications (RIMs) and 57. 6% had received potentially inappropriate medications (PIMs) [16,17,18] . Adverse drug events (ADEs) greatly increase the hospitalization rate, mortality rate and health care costs [6] .
Much research have explored strategies aimed at reducing the occurrence of ADEs. From the perspective of physicians, prescription safety indicators (PSIs) [19] , e-pharmacovigilance [20] , and computerized physician order entry systems (CPOEs) [21] were established and implemented to optimize prescriptions and identify ADEs objectively. From the perspective of patients, self-reported health literacy and a compliance scale should be developed specially [22] . Studies have informed the development of a disease-speci c framework for examining patient safety in relation to CKD [23] . However, there is no universally accepted guideline, instrument or self-report scale, and currently none of the adherence scales can be considered as a gold standard [22] . The most widely used is the Morisky medication adherence scale (MMAS) [24] . However, patients with CKD are often required to attend regular outpatient follow-up, and they need more medical care and physician intervention than patients with chronic comorbid conditions but no CKD [25] . It is necessary to develop a speci c and wide-ranging self-management instrument for medication safety in patients with CKD [26] .
Studies have shown that lack of knowledge and negative attitude may lead to wrong behavior and aggravate the occurrence of adverse events [27] . Therefore, we constructed the framework of a medication safety evaluation instrument based on knowledge, attitude and practice (KAP) theory, which is a progressive model aimed at changing health behavior based on knowledge and beliefs.It helps heathcare providers or patients to establish health beliefs and ultimately cultivate health behavior, which has great guiding signi cance for health education and promotion. At present, KAP theory has been applied to the prevention of the transmission of infectious diseases [28] and clinical diagnosis, treatment nursing [29] and the rehabilitation of chronic non-communicable diseases, such as hypertension [30] , coronary heart disease (CHD) [31] and diabetes, and their complications [32] . However, there are very few well-tested KAP scale concentrating on the medication characteristics of patients with CKD. A KAP survey instrument for CKD in a Swahili-speaking population measures knowledge, attitudes and practices pertaining to kidney disease, not medication safety [33] . A Chinese and western medication adherence scale in patients with CKD [34] included no special medication knowledge items relating to the characteristics of CKD, and involved a small proportion of behavioral aspects.Therefore, we have developed and validated the rst instrument evaluating knowledge, attitudes and practices for medication safety in patients with CKD.
In this study, we adopted an innovative approach to construct a scale incorporating knowledge of nephrotoxic medications, drug interaction, ADR and protection of renal function, participation, and attitude towards medication health education activities. This made it possible to implement targeted and tailored counseling and interventions to prompt persistent, correct, and safe self-medication for patients with CKD [35] , which could represent a pivotal step in reducing the occurrence of MRPs and ADEs in patients with CKD, and in delaying the deterioration of renal function.

Methods
The process used to develop the scale was composed of four major stages.

Stage 1.Establishment of the initial item pool
Firstly, we retrieved the relevant literature on the current situation regarding medication safety in patients with CKD and the development of evaluation instruments for medication literacy and the behavior of the general population or some other disease-speci c population. We searched PubMed, Embase, the WHO Global Health Library databases, China National Knowledge Infrastructure and WangFang data for studies published in English or Chinese language until October 1, 2020. Brie y, the following Medical Subject Headings terms were used: chronic kidney disease; end-stage kidney/renal disease; dialysis-dependent; medication safety; medication (non)compliance; medication (non)adherence; self-medication; medication literacy; adverse events; and over-the-counter medication. [22~24,30~33] Based on domestic and foreign studies, there are three main problems in medication safety of patients with CKD: 1. The current situation is grim. The lower medication adherence and the higher rate of RIMs and ADEs in patients with CKD is due to the common faulty cognitive perceptions about medication [36] , the irregular drug use behavior [37] and the lack of education activities about medication literacy. 2.Lack of speci c selfmanagement instrument for medication safety. The subject of the KAP survey instrument for CKD is about disease [33] . The MMAS is not comprehensive for patients with CKD [24] . 3.The localized measurement tools for CKD patients are awed and incomplete,such as the Chinese and western medication adherence scale in patients with CKD [34] included no nephrotoxic medication.
Based on the above questions, we conducted semi-structural interviews, with 10 patients at different stages of CKD, on the di culties and requirments of daily self-medication. The interview outline is about the characteristics, needs, and factors in uencing medication compliance in relation to their own experience (see Supplementary material 1). There were 6 (60%) males and 4 (40%) females with a mean age of (54.10±16.87) years. In addition, 2 (20%) participants had only a primary education and 2 (20%) participants had a university level education. The causes of kidney failure were hypertensive renal injury in 3 (30%) patients, diabetic nephropathy in 3 (30%) patients, chronic nephritis in 2 (20%) patients, IgA nephropathy in 1 (10%) patient, and unknown causes in 1 (10%) patient. There were 2 (20%) participants with stage 1-2 CKD, 4 (40%) participants with stage 3-5 CKD and 4 (40%) dialysis participants.
After the interviews, the records were sorted out in the rst time. Then the core meaning of the respondents' answers were selected to entry, written and concise, and similar entries were merged into the framework of KAP. Meanwhile, our research group-comprising one chief physician, one deputy chief physician, one attending physician, two nurses in charge, two clinical pharmacists, and one statistical expert with more than 10 years' professional work experience and three graduate students-discussed and integrated clinical treatment, nursing experience [38] , pharmaceutical knowledge [39] , statistical theory, and literature guides, to draft the initial item pool of the KAP scale for the medication safety of patients with CKD. Some of the items were formulated based on a series of strategies to promote patient safety proposed by the World Health Organization [40] , Chinese Hospital Association [41] , American Kidney Disease Education Plan [42] , German Medical Education Association [43] and other institutional platforms. Some items were extracted and adapted from the questionnaire "KAP investigation on behavior risk of drug safety of Chinese residents," which is the key project of science and technology communication innovation project of the national Medical Economic Information Network of the Chinese Pharmaceutical Association in 2018.
Thus, a primary medication safety scale-comprising three dimensions and 32 items-was constructed, incorporating the patients with CKD' knowledge of medication safety (13 items), behavior around selfmedication (11 items), and participation and attitude to health education activities in terms of medication safety (8 items). The Knowledge domain incorporated the methods and characteristics of the drugs patients used, causes and hazards of polypharmacy, identi cation and treatment of ADRs, side effects of healthcare products and Chinese patent medication, and medication safety and self-health management related to CKD diseases in daily life. The Practice domain incorporated the whole process of visit and self-medication management, which was formulated by analyzing the in uencing factors regarding medication safety in patients with CKD. For the Attitude and Belief part of the KAP framework, beliefs about medication have been investigated in most scales [30] . The health education plays an important role in acquiring knowledge and promoting healthy behavior [44] , but there is no scale involving attitudes towards health education activities. Therefore, instead of medication belief, this part of the framework was constructed with the dimensions of "participation in health education activities of medication safety" and innovative "attitudes towards health education activities of medication safety", which may be de ned as participation in different places and different forms of health education activities and the attitudes towards necessity of the above activities.
A response option of 5-point Likert scale for each item was used (totally ignorant, ignorant, general, partially aware, fully aware or never, seldom, sometimes, often, always or wholly unnecessary, not necessary, general, necessary, and very necessary), in which scores of 1, 2, 3, 4, and 5 were applied accordingly (see Supplementary material 2). There was no reverse item. The summed total score on this 32-item scale ranged from 32 to 160, with higher scores indicating higher levels of self-medication adherence.
Stage 2.Improvement of the primary scale Eight experts specialized in nephrology and pharmacy from four hospitals in Shanghai-including two chief physicians and pharmacy directors, two deputy chief physicians, and four nurses in charge, along with attending physicians with over 10 years of professional work experience-were invited to examine the clarity, feasibility, and scienti calness of the drafted items as well as each item's relevance and appropriateness to its associated construct. According to the feedback received from the expert consultation, item 16 was revised to "How often do you actively learn from the medical staff the name, indications, usage and dosage, adverse reactions, contraindications and precautions of the drugs used?" and item 17 was revised to "How often do you record the medication list and actively tell the doctor the medication history (including prescription drugs, over-the-counter drugs, and health products)?" Twenty patients with CKD were selected randomly from the Department of Nephrology, Shanghai Ninth People's Hospital, for pre-investigation with the revised initial drafted scale to test readability and feasibility.
They were asked to answer the questions below: Could you understand the items? Could you answer the scale by yourself? How did you comprehend the item content? Were the items ambiguous? Did it take too long in time? Patients' questions and suggestions on the scale were collected and discussed again by researchers. Item 13, which was di cult to understand in eight patients (40%), was revised to "Do you know how to protect your kidney in daily life?"

Stage 3.Clinical investigation
We targeted a minimum number of 160 participants to ensure a participant-to-item ratio (N/p) of at least 5:1 [45] . The overall prevalence rate of CKD in China is 10.8%. With a two-sided test (alpha = 5%, power = 90%) [46] and based on the assumed dropout rate(DR=20%) [47] , we estimated that a sample size of 209 patients should be enrolled in the study. By objective sampling, 209 patients with CKD (including dialysis patients) who visited the out patient department or ward of the Department of Nephrology of the 9th People's Hospital a liated to Shanghai Jiao Tong University School of medication from November 1, 2020, to April 30, 2021, were selected as the survey participants. A Questionnaire Star electronic questionnaire was distributed in the WeChat group of peritoneal dialysis patients. Patients who met the inclusion criteria were included based on the results of the online questionnaire. Paper questionnaires were distributed-in the outpatient, ward, and hemodialysis room of the nephrology department-to patients who met the inclusion criteria. Once the researcher had explained the purpose and signi cance of the investigation, each patient was able to complete the questionnaire without stress. For illiterate patients, face-to-face surveys were conducted by researchers. During the questionnaire investigation, participants' timely feedback on items or the entire questionnaire was taken into consideration and improvements were made accordingly.
Inclusion criteria were as follows (1) aged over 18 years; (2) diagnosed with chronic kidney disease [48] ; (3) taking more than two kinds of oral drugs daily (including prescribed drugs and over-the-counter drugs, excluding health products and herbal medications); (4) able to correctly understand and complete the questionnaire; and (5) were willing to participate in this study and sign the consent forms.
Exclusion criteria were as follows (1) unable to complete the questionnaire owing to hearing and/or communication disability, and (2) unable to take care of themselves owing to dementia or cognitive impairment, or severe diseases of other organs or systems.

Stage 4. Statistical analysis
The collected data were statistically analyzed using IBM SPSS 23.0 for item selection and tested for reliability and validity [49] .

Item analysis
The purpose of the item analysis was to test the appropriateness or reliability of the scale or individual items. Based on the following analysis criteria, the items that failed to meet the criteria were deleted. a. Critical ratio (CR): The total scores were listed in descending numerical order, among which 27% of the high total scores from the highest one were de ned as the high-score group and 27% of the low total scores were de ned as the low-score group. The difference in scores for each item in the two groups was tested using the independent t-test method. Items with scores that had no signi cant difference between the two extreme groups were deleted (CR <3 or P >0.05). b. Correlation coe cient method: Items with Pearson's correlation coe cient r <0.4 were deleted. c. Homogeneity test: including Cronbach's alpha(α), commonality and factor load. A reliability test-Cronbach's alpha(α)-was used to assess the internal consistency of the scale. If an item reduced the overall Cronbach's alpha, it could be deleted. If the scale contained more than two different domains, Cronbach's alpha was calculated with different domains as subscales. Items with commonality <0.20 indicated that the relationship between the items and common factors had low correlation, and were deleted. Items with a factor load <0.45 indicated a lower degree of homogeneity and could be deleted.

Validity test
The construct validity of this scale was tested by conducting exploratory factor analysis (EFA) to identify its constructs. Firstly, the Kaiser-Meyer-Olkin (KMO) >0.60 and Bartlett's test (χ 2 ) reached a signi cant level (P <0.05), indicating that there were common factors among the variables, which was suitable for EFA. The measure of sampling adequacy (MSA) <0.50 and homogeneity <0.2 after extraction could be used as exclusion criteria for factor analysis. Principal component analysis with Varimax rotation was employed to analyze the construct and factor structure of this scale and of each domain. The slope plot, components with eigenvalues >1.0, and Items with factor loadings >0.4 were identi ed to calculate the number of common factors. The factor structure would change after the item was deleted, so the factor analysis must be conducted again until a well-constructed scale is completed.

Reliability test
Internal consistency was tested using Cronbach's alpha values. A value of >0.80 in total scale, as well as a value of >0.70 in each dimension, was considered satisfactory internal reliability. Spearman-Brown split-half reliability was also calculated; a correlation coe cient ≥0.70 indicated good internal reliability. As additionally, the Pearson's correlation coe cient between each dimension and the whole scale was calculated. The Pearson's correlation coe cients were used to identify the direction and degree of association between the scale and its dimensions.

Results
Demographical characteristics A total of 213 eligible patients with CKD participated in this investigation. Of the 213 distributed questionnaires, 209 were collected back and checked for completion and validity. The response rate was 98.12%. There were 126 (60.29%) males and 83 (39.71%) females with a mean age of (60.50±12.97) years (

Item analysis
The critical ratio (CR), correlation coe cient method, and homogeneity test were used to test the appropriateness or reliability of the individual items. The results are shown in Table 2. Where there were more than three indicators, they would be deleted, and eventually three items were deleted, including Item 1, 6 and 8. Notes: *Represents the substandard items that can be deleted.
Maintaining an awareness of life kidney protection measures is a very important process in terms of managing chronic kidney disease successfully [50] , so Item 13 with one indicator was retained. The Cronbach's alpha value after deleting Item 20 was higher than the value of the subscale, that is 0.921, which reached the ideal reliability. Moreover, 0.926 is not considerably different from 0.921, so the deletion is insigni cant. After item analysis, the nal and formal medication safety scale for thepatients with CKD was accomplished and incorporated three domains and 29 items.

Validity analysis
The KMO test for the overall scale and its four domains indicated that the sample size for this scale was adequate with values of 0.869 for the scale, Bartlett's test: χ2 (chi square test value) was 3359.048 (DF=378, P=0.000), indicating the suitability of the correlation matrix to draw factors. The MSA of all the items was greater than 0.621, and the commonality was greater than 0.438, which was suitable for factor analysis.
Principal component analysis with Varimax rotation was employed to analyze the construct and factor structure of this scale and of each domain by conducting EFA. The results showed that ve components with eigenvalues >1.0 were identi ed. The slope plot gradually attened after the fth common factor (see Figure 1). Therefore, it is appropriate to reserve the ve common factors as the theoretical structure of the scale. The ve common factor eigenvalues were 7.55, 5.283, 2.986, 2.739, and 1.321, and the explanation rate of their cumulative variance was 68.548%.The rst four common factors obtained by EFA were the same as the structure of the prediction scale, indicating that the measurement scale had stable construct validity. Factor con guration containing fewer than three items could not be achieved. Therefore, the fth common factor, which only contained Item 20, was deleted. The panel agreed that Item 18 covered Item 20, and most patients indicated that a medication caregiver was not required except in those patients with vision disability or dementia in the investigation. Item 20, "Do you need your family members to implement medication monitoring, remind you of the time and dosage of medication, and nd out if you have any discomfort or abnormal feelings from your medications" was deleted, though the caregiver was found to be the protective factor in ADRs. 23 The factor analysis had to be conducted again after the item was deleted.
The second component matrix after the rotation axis showed that common factor 1 contained Items 3, 7,12, 5, 4, 10, 2, 9, 23, 19 and 13 (sorted by factor loading); common factor 2 contained Items 15,16,14,21,17,24,22,13,19 and 18; common factor 3 contained Items 30, 29, 31 and 32; and common factor 4 contained Items 25, 26, 27 and 28. The factor loadings of Items 13, 19, 18 and 23 were greater than 0.40 in two con gurations. Combining the design background of the items and the de nition of initial dimensions, Item 13 was included in common factor 1 and Items 19, 18 and 23 were included in common factor 2 following group discussion and a consultation among the experts.
According to the theoretical structure of the prediction scale and the characteristics of each item in each dimension, the extracted common factors were named "Knowledge about medication safety in CKD," "Medication safety behavior," "Participation of health education activities of medication safety," and "Attitude towards health education activities of medication safety." Finally, a formal scale was formed with 28 items and four dimensions (see Supplementary material 3). The eigenvalues of each dimension were 4.948, 2.982, 2.743, and 1. 276, respectively, and the cumulative variance contribution rate was 65.132% (see Table 3).

Reliability analysis
The Cronbach's alpha coe cient, along with the split-half reliability of the overall scale and of each domain, was measured. The Cronbach's alpha coe cient was 0.944 for the overall scale and ranged from 0.816 to 0.935 for four dimensions, indicating good internal consistency reliability for the overall scale and for each dimension. The Spearman-Brown split-half reliability coe cient was 0.834 for the total scale and ranged from 0.751 to 0.916 for four dimensions, indicating high internal consistencies between the halves across the scale and its dimensions (see Table 4). Table 4 The reliability coe cients of the total scale and among each dimension The Pearson's correlation coe cients between the dimensions and the total scale ranged from 0.417 to 0.917. The correlation coe cients among the dimensions ranged from 0.166 to 0.797 (see Table 5).
Therefore, good reliability and acceptable validity of this scale were established.

Discussion
This is the rst study to develop a self-reporting medication safety scale speci c for patients with CKD, which is also aimed at exploring the psychometric properties of medication safety in Chinese patients with CKD. There were several existing medication adherence scales and some of them focuse on special population, especially on hypertensive patients. For example, the medication adherence questionnaire (MAQ) with 4 items is the shortest scale and easiest to score but it didn't assess patient-self e cacy [51] ; the MMAS-8 scale was useful in measuring polypharmacy related non-adherence [52] . However, they had limitations, such as a yes/no response that did not differentiate between highly adherent and partially adherent patients for a given scenario and an increased cost of data collection because of licensure. In addition, medication adherence rating scale (MARS) was created for assessment of adherence in psychiatric patients [53] . The Hill-Bone compliance scale with 14 items in three subscales that assess medication adherence, sodium intake, and appointment keeping was developed to provide a simple method to determine patient-reported compliance levels. The nine adherence questions are worded speci cally in regard to high blood pressure medications and the generalizability across patient populations was limited.
However, there are a lack of scales that can be used in patients with CKD, especially with dialysis who require a complex therapeutic regime and standardized medication behavior. The chronic kidney disease self-care (CKDSC) scale [54] is a 16-item questionnaire with ve subscales that medication adherence, diet control, exercise, smoking behaviours and blood pressure monitoring. It was designed for CKD patients who have not yet undergone renal replacement therapy or kidney transplant. The end-stage renal disease questionnaire (ESRD-AQ) [55] is the only published scale currently available for assessing dialysis patient's adherence behavior. However, Internal consistency reliability has not been evaluated since the instrument's design does not possess homogeneous items. Thus, we considered it important to develop a KAP scale for medication safety in patients with CKD, inculding dialysis patients. There were several KAP scales for measuring knowledge, attitudes, and practices of the etiologies, diagnosis, and treatment of kidney disease of patients at risk for CKD, such as the CKD screening Index [56] and the Instrument of KAP associated with CKD [33,57] .
In total, 209 patients with CKD were purposively selected to conduct the questionnaire using the primary scale. It has been observed that about 15%~53% of patients with CKD are non-adherent to medical prescriptions [58,59] . Estimates of nonadherence in ESRD range from 22 to 74% [60] . Compared to a nondialysis population, poor adherence to complex therapies contributes to increased morbidity and mortality in maintenance dialysis patient population [61] . Therefore, this study recruited a larger proportion of dialysis patients.
The CR method and homogeneity test were used to select the items. The level of compliance in patients with CKD is generally higher than that in patients with common chronic comorbidities owing to the need for more frequent medical follow-ups and more guidance on the medical support available. Therefore, mastery of general knowledge about "name, function, indications, usage and dosage, and storage methods of drugs used" in patients with CKD is generally better than that in patients with common chronic comorbidities. Next, substandard items 1, 6, and 8 were deleted and 29 items were included to test the validity.
It is noteworthy that (1) (4) the validity of construction is based on the theory of logical analysis, and the correctness of the theory is tested according to the actual data and combined with the theory in "practical questionnaire statistical analysis [43] ," where it is stated that any scale should undergo a construction validity test or other validity analysis, such as content validity, expert validity, calibration related validity, and so on". Therefore, in this scale we tested the construction validity. Items with factor loadings greater than 0.4 on the same component, as well as non-salient loadings less than 0.4 on other components, were considered to contribute to a component. Given that entries with factor load greater than 0.4 in both common factors, priority can be given in the original dimension [43] , items 13, 19, 18, and 23 with factor loadings over 0.4 on two components were added to the original component. Finally, a scale with four dimensions and 28 items was created, which could account for a 65.132% variation in the entire scale. Among them, factor 2 "medication safety behavior" and factor 3 "participation in health education activities of medication safety" are both practices in the KAP theory.
The Pearson's correlation coe cients between dimensions and the total scale ranged from 0.417 to 0.917, indicating an average to strong association between the scale and its domains. The correlation coe cients among the dimensions ranged from 0.166 to 0.797, indicating that the four dimensions are not only relatively independent but also correlated with each other. Future studies could focus on conducting a path analysis among the four dimensions to determine how these four elements correlate or interact with each other.
The reliability results showed that Cronbach alpha coe cient of the total scale was 0.944, which is better than the result for the MMAS-8 scale [53] , a Chinese version of the general medication adherence scale in patients with chronic illness [37] , and a Chinese and western medication adherence scale in patients with CKD [34] -with a Cronbach's alpha of 0.83, 0.781, and 0.915, respectively. The split-half reliability coe cient of the total scale was 0.834, which is better than that for the CKD Screening Index [55] , with a split-half coe cient of 0.80. In addition, this scale demonstrated a high acceptability among patients with CKD, with a response rate of 98.12%. Therefore, this newly developed scale with greatly satis ed validity and reliability is scienti c and pragmatic and appropriate for assessing the medication safety of patients with CKD in China.
Owing to the effect of age, educational level, economic status, living environment, types of chronic comorbidity and drugs used, participation in medication safety education, and other factors, the selfmanagement level of medication safety among patients varies considerably [63] . This scale comprehensively investigates and quantitatively evaluates the mastery of medication knowledge, the accurateness of selfmedication behavior, and the participation and attitude towards education activities in patients with CKD. This may be helpful for health care professionals to manage Individually medication safety throughout the entire process of medication in CKD patients with chronic comorbidities, and identi e and reduce the occurrence of MRPs and ADEs.

Limitations And Prospects
Firstly, the scale only tests the construct validity. We can test the correlation content validity, expert validity and retest reliability to further improve the scale. Secondly, this study recruited a larger proportion of dialysis patients and was conducted in a single center, the result may be hence too heterogenous; we need to expand the sample size and conduct a multi-center survey in different regions and different populations.
Thirdly, owing to the self-reported data collection, there may have been some subjectivity and recall bias present in the process, thus the quality and accuracy of the data collected in this study might be compromised. The evaluation could be further improved by re ning the interview process and utilizing objective evaluation tools. In addition, this scale is just a preliminary and exploratory study and is suitable for preliminary screening. Finally, this study is a cross-sectional study. The correlation between the score of self-medication and clinical outcomes cannot be con rmed. Further studies are needed to determine whether this scale can accurately evaluate and prevent MRPs, PIMs, and ADEs, and improve the clinical outcomes of patients with CKD.

Conclusion
The KAP scale for medication safety in patients with CKD with greatly satis ed validity and reliability is scienti c and comprehensive, and has wide applicability and strong operability to assess medication safety in patients with CKD. Patients play an important role in medication safety [64] . The patient-centered medication safety evaluation of the scale can strengthen the cognition of patients or their caregivers on rational drug use, correct self-medication behavior, and promote medication safety. Also, the scale can be applied to the entire process of CKD management via a joint management team of general practitioners and specialists. General practitioners can then gather evaluation results and improve health records for the specialists, which can assist their team in making accurate clinical diagnoses and treatment plans and providing better medication nursing for patients and their families [65] . The general practitioners also could provide individualized medication guidance and health education for patients to improve their selfmanagement and self-e cacy [66] .

Declarations
Ethics approval and consent to participate The study was approved by the institutional ethics committee of Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of medication (approval no. sh9h-2020-t393-1) and has been performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments.An informed consent form was signed by each participant.

Consent for publication
Not applicable.

Data availability
All data generated or analyzed during this study are included in this published article and its supplementary information les (see Supplementary material 4).  Figure 1 slope plot