Table 1.
Baseline characteristics of the development cohort and the two validation cohorts in Prostate Cancer data Base Sweden (PCBaSe) 5.
Table 2.
Duration of follow-up for mortality after the index date, number of deaths, and model discrimination in development and validation cohorts.
Fig 1.
Example of predictors selected in the prediction model.
ICD-10 codes for cardiovascular diseases are shown in the figure, specifically focusing on the ICD-10 code category I5 (heart failure and some heart disorders and diseases). The innermost circle represents ICD-10 codes with two characters and the outmost circle codes with five characters and grouped predictors. Each predictor variable (occurrence, frequency, recency, duration) derived for each ICD-10 code corresponds to a color-coded circle segment. A segment is colored if any coefficient within that group of predictors had a non-zero coefficient for that code, and grey if non-informative. I50 = heart failure, I51 = complications and ill-defined descriptions of heart disease, I500 = right ventricular failure, I501 = left ventricular failure, I509 = heart failure, unspecified, I510 = cardiac septal defect, acquired, I513 = intracardiac thrombosis, not elsewhere classified, I514 = myocarditis, unspecified, I516 = cardiovascular disease, unspecified, I517 = cardiomegaly, I519 = heart disease, unspecified, I5019 = left ventricular failure, unspecified, I5099 = heart failure, unspecified.
Fig 2.
Observed 10-year mortality risk compared to predicted risk based on the MDCI or the Charlson comorbidity index among 54,539 men without prostate cancer and 68,357 men with prostate cancer. In each cohort, the calibration plot was obtained by computing the predicted 10-year survival probability for each individual using a Cox proportional hazards model including the MDCI or Charlson comorbidity index, respectively as predictor, and a corresponding estimate of the baseline hazard function. The observed survival probability was computed as the average 10-year survival estimated by the Kaplan-Meier curve within the intervals of the predicted survival probabilities using the cutoffs 0%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90% and 100%. The predicted survival probability was similarly computed as the average of all individual survival probabilities among individuals within each interval. Probability intervals containing zero individuals are not shown.
Fig 3.
Survival of men in the validation cohort consisting of men without prostate cancer.
The validation cohort has been split in subgroups based on the Charlson comorbidity index (0, 1, 2, 3+) and the drug comorbidity index (quartiles of the DCI). The survival within each subgroup is then shown stratified in quartiles of the MDCI developed using 10 years of follow-up for mortality.
Fig 4.
1-year hazard ratios for men in the validation cohort consisting of men without prostate cancer.
Stratified by age, Charlson comorbidity index (CCI) and drug comorbidity index (DCI) for the multi-dimensional diagnosis-based comorbidity index (MDCI) developed using 10 years of follow-up for mortality. Not available (NA) indicates that the hazard ratio could not be estimated due to small sample size and/or few events in each stratum.