Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

chelseaxxlutz — Thu, 02 Nov 2017 18:10:43 GMT

Good afternoon,

I am trying to replicate the below methods, but I am lost on the SAS code/procedures needed to reproduce the bolded section. I have searched Google and chat rooms and other SAS sites for hours and have come up with half fixes every time or code that doesn't run. I am at a loss and now on a time crunch. Any assistance would be GREATLY appreciated. Thank you in advance.

"Two-thirds of the study participants (n ¼ 10,039) were randomly assigned to a derivation cohort for the development of risk prediction models with computer-generated random numbers, and the remaining one-third (n ¼ 5019) were reserved as an independent validation cohort. Among the patients allocated to the derivation cohort, new risk prediction models for all-cause and cardiovascular death were developed by using multivariable Cox proportional hazards models including the above-mentioned risk factors. In these models, survival analysis was censored at last date of follow-up or date of death. The performance of each risk prediction model was tested among either the patients assigned to the derivation or the validation cohort. The ability of each model to discriminate patients who died within 5 years from surviving patients was evaluated using the c-statistic for the survival analysis. Observed risk (based on the Kaplan-Meier product limit method) and predicted risk (based on the risk prediction model) of each outcome at 5 years were compared by ranking participants into deciles of predicted risk, and calibration of each model was evaluated in the derivation cohort using a modified Hosmer-Lemeshow c2-statistic with 9 degree of freedom. In addition, the risk prediction model for each outcome was translated into a simple score sheet in the same way as described in the Framingham Heart Study.22 All statistical analyses were performed with SAS 9.3 (SAS Institute, Cary, NC). Two-sided values of P < 0.05 were considered statistically significant."

Thank you!

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Reeza — Thu, 02 Nov 2017 19:36:18 GMT

PROC LIFETEST and/or PROC PHREG

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

chelseaxxlutz — Thu, 02 Nov 2017 19:45:25 GMT

Hi Reeza,

Thank you. I know the procedures for proc lifetest and phreg and have run them successfully. What I'm asking is how do I compare the two risks (observed and predicted) ranked into deciles of risk? Thank you!

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Reeza — Thu, 02 Nov 2017 19:47:17 GMT

I would assume at that point it's just a count of data, so wouldn't it be a Chi square test? The output in the paper should clarify that but I don't have time to read it.

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Reeza — Thu, 02 Nov 2017 19:48:26 GMT

I gotta run, but re-read your question. Once the risks are grouped, take a look at NPAR1WAY proc.

topic Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk in Statistical Procedures

Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk

Re: Uknown SAS procedure for comparing observed (Kaplan Meier) and predicted (Cox) risk