Risk Prediction Model Of CVD In Middle-aged And Elderly Patients With T2DM And Its Application Effect Evaluation

PurposeTo understand the morbidity trend and risk factors of CVD in middle-aged and elderly patients with T2DM,construct a CVD risk prediction model and scoring system suitable for middle-aged and elderly patients with T2DM,quantify the individual morbidity risk,guide the self-monitoring middle-aged and elderly patients with T2DM,and provide certain scientific bases for community graded health management.MethodsThis study was based on a retrospective cohort study of T2DM patients aged 45 years and older who had not developed CVD in CHARLS from 2011 to 2018.The outcome events were new CVD,including heart disease and stroke.The status of new CVD was obtained by a follow-up survey.The characteristics of CVD in T2DM patients were compared by T-test and rank sum test.The risk prediction model was constructed by Logistic regression and LASSO regression,respectively.SPSS22.0 software was used to carry out Logistic regression,single-factor Logistic regression was used to screen model variables,and progressive multi-factor Logistic regression was used to construct the prediction model.LASSO regression adopts R software(4.2.3),LASSO regression screens model variables,Irm function constructs the prediction model,and column graph establishes scoring system.The model was internally validated using the Bootstrap method,and externally tested in CHARLS T2DM patients>45 years old from 2015 to 2018 without CVD(excluding modeling data).ResultsIn this study,a total of 4148 middle-aged and elderly patients with T2DM were investigated from 2011 to 2018,and 996 were investigated at baseline in 2011,including 321 patients with CVD.The rate of T2DM patients with CVD was 32.23%.The observation starting point was 675 patients with T2DM without CVD in the 2011 survey,with an average follow-up of 6.72 years.In 2013,2015 and 2018,79 patients(13.72%),119 patients(22.20%)and 176 patients(36.29%)developed CVD,respectively.The incidence densities were 6.31 per 100 person-years,5.84 per 100 person-years and 5.67 per 100 person-years,respectively.A total of 198 patients had outcome events during follow-up,including gender(OR=1.642,95%CI=1.111-2.427,P=0.013),waist circumference(OR=2.267,95%CI=1.328-3.869,P=0.003),duration of T2DM(OR=1.644,95%CI=1.316-2.053,P<0.001),sleep duration per night(OR=0.656,95%CI=0.505-0.852,P=0.002),systolic blood pressure(OR=1.322,95%CI=1.092-1.599,P=0.004)was an independent factor for CVD in middle-aged and elderly patients with T2DM.The final variables included in the progressive forward multiple Logistic regression model were female(OR=1.624,95%CI=1.089-2.422,P=0.017),excessive waist circumference(OR=2.376,95%CI=1.384-4.080,P=0.002),duration of T2DM for 5-10 years(OR=1.724,95%CI=1.067-2.784,P=0.026),10～15 years(OR=3.019,95%CI=1.580-5.766,P=0.001),duration of T2DM≥15 years(OR=3.554,95%CI=1.471-8.589,P=0.005),sleep duration per night ≤5h(OR=2.372,95%CI=1.3914.046,P=0.002),systolic blood pressure 160～180mmHg(OR=3.199,95%CI=1.383-7.398,P=0.007),systolic blood pressure ≥180mmHg(OR=4.757,95%CI=1.516-14.926,P=0.008).The chi-square value of the H-L test was 8.457(P=0.294),and AUC was 0.692(95%CI=0.645-0.739).The chi-square value of the H-L test for the Bootstrap model was 12.676,P>0.100,and the AUC was 0.691(95%CI=0.6440.738).The chi-square value of the H-L test for external tests was 9.158,P>0.100,and AUC was 0.788(95%CI=0.727-0.849).The variables included in the LASSO regression model combined with the Bootstrap regression model were female(OR=1.624,95%CI=1.089-2.422,P=0.017),excessive waist circumference(OR=2.376,95%CI=1.384-4.080,P=0.002),duration of T2DM for 5-10 years(OR=1.724,95%CI=1.067-2.784,P=0.026),duration of T2DM for 10-15 years(OR=3.019,95%CI=1.580-5.766,P=0.001),duration of T2DM≥15 years(OR=3.554,95%CI=1.471-8.589,P=0.005),sleep duration ≤5h per night(OR=1.561,95%CI=1.000-2.440,P=0.050),systolic blood pressure 160～180mmHg(OR=2.794,95%CI=1.267-6.164,P=0.007),systolic blood pressure≥180mmHg(OR=4.155,95%CI=1.370-12.606,P=0.008).The model AUC was 0.692(95%CI=0.645-0.739),and the calibration curve was highly fitted.The AUC of external tests was 0.812(95%CI=0.755-0.869),and the calibration curve was highly fitting.The total values of the column chart ranged from 0 to 350 points,and the mean value was 93.12±57.27 points.When the maximum Yoden index is reached,the Cut-off fraction is≥105 points,the sensitivity is 78.28%,the specificity is 49.51%,the Yoden index is 0.28,the DOR is 5.52,the PPV49.84%,and the NPV is 78.06%.In the external test,the Cut-off score was still≥105 points when the Yoden index reached its maximum,the sensitivity was 74.32%,the specificity was 71.09%,the Yoden index was 0.45,the DOR was 29.72,the PPV was 39.29%,and the NPV was 91.67%.Select the best Cut-off score≥ 105 as a screening tool for primary prevention in the community.ConclusionsThe CVD risk prediction model and scoring system established by LASSO regression in this study are relatively reliable,which is expected to promote self-monitoring of middle-aged and elderly T2DM patients,help community medical workers to implement the classified intervention,and improve community graded health management.