cstat_surv

Version 1.0.0 | 2026-04-08

cstat_surv calculates Harrell's C-statistic for Cox proportional hazards models and reports an infinitesimal-jackknife standard error with a confidence interval. It is meant for post-estimation discrimination checks when you want a survival-model analogue of AUC.

Requirements

• Stata 16 or later
• Survival-time data already declared with stset
• A Cox model fit with stcox in the current session

Installation

capture ado uninstall cstat_surv
net install cstat_surv, from("https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/cstat_surv") replace

Commands

Command	Description
cstat_surv	Calculate Harrell's C-statistic after stcox

How It Works

• Fit a Cox model with stcox.
• Run cstat_surv immediately afterward on the same estimation results.
• The command predicts fitted risk scores, compares all comparable survival pairs, and replaces e() with the C-statistic output.

Worked Examples

1. Basic workflow with built-in data

This is the shortest complete workflow and is runnable with Stata's built-in survival example data.

webuse drugtr, clear
stset studytime, failure(died)
stcox age drug
cstat_surv

2. Report a different confidence level

Use level() when you want a non-default interval around Harrell's C.

webuse drugtr, clear
stset studytime, failure(died)
stcox age drug
cstat_surv, level(90)

Important Behavior

• cstat_surv overwrites the active e() results with the C-statistic output, so rerun stcox if you need the original model results again.
• Weights from the original stcox model are not used in the pairwise C-statistic calculation.
• Delayed entry via _t0 is not accounted for in pair comparisons.
• Multi-record counting-process survival data are not supported.
• The calculation is pairwise, so very large datasets can take noticeably longer to run.

Key Stored Results

Result	Description
e(c)	Harrell's C-statistic
e(se)	Infinitesimal-jackknife standard error
e(ci_lo)	Lower confidence limit
e(ci_hi)	Upper confidence limit
e(somers_d)	Somers' D, equal to 2C - 1
e(N_comparable)	Number of comparable pairs

Version History

• 1.0.0 (2026-04-08): Initial Stata-Tools release

Author

Timothy P Copeland, Karolinska Institutet