To predict these criteria, three partial
least square regression models (PLSR) were constructed for each of the co-morbidities, risk factors and medicine classes. The accuracy and precision of the models were evaluated using the regression correlation coefficients (r2) and root mean square error of calibration (RMSEC) values respectively. The r2 value showed how close was the predicted value of the three MRP criteria (no MRP, potential MRP, definite MRP) find more to the actual criteria (r2 value = 1 in ideal case). The RMSEC values explained the error on the models; thus, the lower RMSEC value showed low error and higher precision of the model. In addition, the loadings of the model were evaluated for finding the major factors contributing to MRPs. The results showed that 31 out of 50 (62%) patients admitted with CVDs and/ or diabetes had 38 MRPs including 27 definite and 11 potential MRPs. The inter reviewer
learn more reliability showed very good level of agreement (kappa = 0.778). The 50 patients MRPs criteria (no MRP, potential and definite) were used to construct three PLSR models, along with independent variables including: 43 comorbidities (Model 1), 12 risk factors (Model 2) and 74 medicine classes (Model 3) respectively. The three models showed high precision with RMSEC values of 0.29, 0.42 and 0.78 respectively. However, the first two models showed higher accuracy than the third model with recovery values of 96.9%, 92.2% and 78.6% respectively. The first model loadings showed that the highest comorbidities
contributing to MRPs were diabetes type 2, hypertension and myocardial infarction. This was supported by the medicine classes; where the second PLSR model showed that the main medicine classes contributing to MRPs were medicines used in CVDs including statins, anticoagulants, ACEI, loop diuretics, potassium sparing diuretics, anti-angina and iron supplement. In addition, the third model (risk factors) showed the major contribution from patient non-adherence and poly-pharmacy. More than half of the patients admitted with Vitamin B12 CVDs and diabetes had MRPs. Evidently, these two types of diseases contributed more to MRPs when they were co-existing. In particular, medicines used in CVDs showed a major contribution to MRPs. The results of this study is consistent with the findings of a recent study made on patients with type 2 diabetes with hypertension which showed higher rate of MRPs than our study (Huri and Hoo, 2013). This emphasis the urgent emergence of a targeted prevention tool aimed at predicting patients at higher risk of developing MRPs among CVDs and diabetes patients. 1. Claydon-Platt K, Manias E, & Dunning T. Medication-related problems occurring in people with diabetes during an admission to an adult teaching hospital: A retrospective cohort study. Diabetes Research and Clinical Practice 2012; 5485:1–8. 2. Huri HZ and Hoo FW.