Development And Validation Of A Prediction Model For Medication Non-adherence In Patients With Chronic Kidney Disease

ObjectiveThe purpose of this study was to identify risk factors for medication non-adherence in patients with chronic kidney disease(CKD)based on the WHO five-dimensional framework of factors influencing medication adherence in patients with chronic diseases.A diagnostic model was developed and validated in order to predict the probability of medication adherence risk in patients with chronic kidney disease.In addition,the results of this study are expected to serve as an informative basis for subsequent clinical interventions and indirectly improve patient prognosis.Methods1.It was conducted by computer searches of national and international literature databases during the period from database creation to July 31,2021.The quality of the literature was evaluated based on the Newcastle-Ottawa Scale(NOS)and the Agency for Healthcare Research and Quality(AHRQ).According to the theoretical framework,the literature was reviewed to examine factors that influence medication adherence in patients with chronic kidney disease.2.From September 2021 to October 2022,patients with chronic kidney disease from tertiary hospitals in Guangzhou and Shaoyang were enrolled in the study.A multicenter cross-sectional questionnaire study method was used.The four-item Morisky Medication Adherence Scale was used to assess the medication adherence of the patients and predictors consisted of socio-demographic information,self-designed medication knowledge questionnaire,the 10-item Connor-Davidson Resilience Scale,the Beliefs about Medicine questionnaire,the Acceptance Illness Scale,and the Family Adaptation Partnership Growth and Resolve Index.Biochemical indicators of patients were also obtained from the electronic medical system.3.The least absolute shrinkage and selection operator regression was used to identify the factors,and Logistic regression was applied to construct the model after identifying the characteristic variables.The area under the receiver operating characteristic curve,Hosmer-Lemeshow test,and Decision Curve Analysis were used to calculate the discrimination,calibration,and clinical applicability of the model,respectively.R software was used to perform the statistical analysis.ResultsFollowing literature review,16 papers were screened,15 of which were included in this study after quality assessment,and 30 factors were evaluated.It was decided to include36 factors following extensive discussion and consultation with experts.A total of 1113patients with chronic kidney disease in 3 centers were examined in this study,with a response rate of 97.7%and a medication non-adherence rate of 64.3%.Whenλ=0.04,least absolute shrinkage and selection operator(LASSO)regression identified five characteristic variables,namely,educational level,occupational status,duration of CKD,medication beliefs,and illness acceptance.According to multivariable logistic regression analyses,the effect values of the five elements on the outcome variable were educational level(middle school vs.primary school and below,OR=1.23,95%CI=0.62-2.45;high school vs.primary school and below,OR=1.35,95%CI=0.69-2.61;university school vs.primary school and below,OR=2.70,95%CI=1.35-5.39),occupational status(in-service vs.retirement,OR=1.63,95%CI=1.08-2.45;unemployment vs.retirement,OR=1.44,95%CI=0.84-2.53),CKD duration(“1-5 years”vs.“<1 year”,OR=2.69,95%CI=1.44-5.03;“>5 years”vs.“<1 year”,OR=2.48,95%CI=1.32-4.66),medication beliefs(OR=0.89,95%CI=0.85-0.93),and illness acceptance(OR=0.94,95%CI=0.91-0.97).The receiver operator characteristic curves(ROC)indicated that the nomogram had good accuracy in the training set(C-index 0.72,95%CI:0.68-0.76),the internal validation set(0.79,0.73-0.85),the external validation set 1(0.90,0.85-0.96),the external validation set 2(0.81,0.72-0.90).As shown by the Hosmer-Lemeshow test,the predicted and actual probability were highly consistent(training set,~2=2.657,P=0.954;internal validation set,~2=12.126,P=0.146;Guangzhou external validation set,~2=8.439,P=0.392;Shaoyang external validation set,~2=4.503,P=0.809).The clinical decision curve(DCA)showed that when the threshold probability of medication non-adherence occurrence is between 0.08and 0.72,the net benefit level of the model was significantly greater than that of the"no intervention"or"full intervention"scenarios,which indicated that the model had good clinical applicability.ConclusionThere is a high prevalence of medication non-adherence among patients with chronic kidney disease.A predictive model based on five factors has been successfully developed and validated and could be incorporated into long-term medication management.A tool for online application has been developed that can be used for clinical implementation of targeted interventions.