Table 1.

Log hazard ratios from fitting an exponential model to predict risk of coronary heart disease: estimates from a data sample from The Health Improvement Network, followed by the average of estimates across 1000 simulations based on full data, complete records,‘baseline fully conditional specification (FCS) multiple imputation’, and missing values imputed using the two-fold FCS algorithm.

Variables		THIN cohort	Full data	Complete records	Baseline FCS imputation	Two-fold FCS
						DGM I	DGM II	DGM I	DGM II
Townsend	1	Reference
deprivation	2	0.1520	0.1503	0.1425	0.1497	0.1498	0.1588
score quintile	3	0.2377	0.2367	0.2431	0.2366	0.2366	0.2422
	4	0.2433	0.2400	0.2279	0.2391	0.2401	0.2535
	5	0.4034	0.4024	0.3935	0.4020	0.4023	0.4017
Weight (kg)		0.0019	0.0019	0.0015	0.0016	0.0019	0.0017
Systolic blood pressure (mmHg)		0.0048	0.0049	0.0051	0.0048	0.0051	0.0053
Anti-hypertensive drug treatment		0.2935	0.2868	0.2852	0.2897	0.2855	0.2915
Smoking status	Non-smoker	Reference
	Ex-smoker	0.0679	0.0692	0.0633	0.0672	0.0579	0.0567
	Current smoker	0.2386	0.2385	0.2307	0.2342	0.2325	0.2261
Age group (years)	40–44	− 1.2820	− 1.2872	− 1.3167	− 1.2869	− 1.2880	− 1.2890
	45–49	− 1.0632	− 1.0652	− 1.0892	− 1.0655	− 1.0662	− 1.0623
	50–54	− 0.6402	− 0.6392	− 0.6467	− 0.6398	− 0.6408	− 0.6330
	55–59	− 0.3589	− 0.3597	− 0.3700	− 0.3598	− 0.3605	− 0.3536
	60–64	− 0.2485	− 0.2473	− 0.2545	− 0.2480	− 0.2481	− 0.2423
	65–69	− 0.0396	− 0.0416	− 0.0470	− 0.0418	− 0.0409	− 0.0348
	70–74	Reference
	75–79	0.1108	0.1039	0.1116	0.1043	0.1057	0.1129
	80 +	0.1387	0.1421	0.1255	0.1383	0.1414	0.1368
Constant term		− 5.1993	− 5.2297	− 5.2550	− 5.2098	− 5.2552	− 5.2833

DGM, data generation mechanism; FCS, fully conditional specification; THIN, The Health Improvement Network.