Augmenting cancer registry data with health survey data with no cases in common: the relationship between pre-diagnosis health behaviour and post-diagnosis survival in oesophageal cancer

Table 1 The estimated proportions with each health behaviour, the phi coefficient between imputed values and the estimated excess matches for each analysis

Behaviour 5 years before diagnosis	N	Estimated proportion with behaviour, \( \hat{p_i} \)		Estimated phi coefficient, \( \hat{\rho} \) = φ		Estimated excess matches, \( n\hat{p_i}\left(1-\hat{p_i}\right)\hat{\rho} \)
Behaviour 5 years before diagnosis	N	Median	95% CI	Median	95% CI	Median	95% CI
Current smoking
overall	27,835	0.159	0.157,0.162	0.071	0.059,0.084	262.2	220.1312.2
ESCC	8914	0.166	0.162,0.170	0.077	0.061,0.097	94.8	74.5120.7
EAC	15,726	0.157	0.153,0.159	0.066	0.052,0.081	137.0	107.4169.5
Binge drinking
Overall	27,750	0.100	0.098, 0.102	0.060	0.049,0.077	150.5	121.5192.1
ESCC	8891	0.086	0.082,0.089	0.060	0.042,0.086	42.2	29.8,61.1
EAC	15,673	0.109	0.106,0.111	0.058	0.042,0.079	88.6	63.6120.3
Heavy drinking
Overall	27,749	0.048	0.047,0.050	0.011	0.002,0.025	14.3	2.7,32.0
ESCC	8888	0.046	0.043,0.049	0.015	−0.002,0.036	5.7	− 0.7,14.2
EAC	15,676	0.050	0.048,0.052	0.008	−0.004,0.028	6.0	−3.0,20.8
Physical activity
Overall	27,830	0.737	0.734,0.740	0.034	0.026,0.046	185.1	139.4247.4
ESCC	8912	0.716	0.709,0.721	0.036	0.016,0.056	64.7	29.6100.2
EAC	15,724	0.750	0.746,0.754	0.031	0.013,0.047	91.4	40.0,138.4
Obese
Overall	27,796	0.257	0.254,0.261	0.030	0.020,0.042	160.2	108.4226.8
ESCC	8898	0.262	0.255,0.268	0.045	0.024,0.061	77.0	41.4104.6
EAC	15,709	0.256	0.251,0.261	0.023	0.012,0.041	67.8	35.0,122.4
Current smoking with regular drinking
Overall	27,735	0.034	0.033,0.035	0.022	0.009,0.038	19.8	8.0,34.2
ESCC	8883	0.031	0.029,0.033	0.024	−0.000,0.049	6.2	−0.0,13.5
EAC	15,670	0.035	0.034,0.037	0.021	0.004,0.042	11.5	2.1,22.4

\( \hat{p_i} \) proportion of imputed values where the health behaviour is present
\( \hat{\rho} \) = φ the correlation between the pairs of imputed values (calculated as the phi coefficient)
\( n\hat{p_i}\left(1-\hat{p_i}\right)\hat{\rho} \)= the excess number of correct matches greater than would be expected through chance alone
Median median of 100 repetitions of the imputation algorithm,
95% CI = empirical 95% confidence interval created from the 2.5 and 97.5 percentiles obtained from 100 repetitions of the imputation algorithm,
N number of SEER oesophageal cancer cases receiving data from two donor records from the BRFSS health behaviour datasets
ESCC oesophageal squamous cell carcinoma,
EAC oesophageal adenocarcinoma

ISSN: 1471-2407