Adjusting for overdispersion in piecewise exponential regression models to estimate excess mortality rate in population-based research

Table 1 Piecewise exponential regression excess mortality models with and without correcting for overdispesion, n = 376,791 women diagnosed with breast cancer in England between 1997 and the end of 2005

	PEREM A		PEREM B (scaled SE)			PEREM C (Robust SE)			PEREM D (NBR)
Age at diagnosis	EMRR	SE	EMRR	SE	RLE (%)	EMRR	SE	RLE (%)	EMRR	SE	RLE (%)
50−59vs.<50	0.75	0.0107	0.75*	0.0493	21.3579	0.75*	0.0576	29.2222	0.75*	0.0380	12.6944
60-69 vs. <50	0.88	0.0130	0.88*	0.0600	21.3580	0.88*	0.0599	21.2823	0.87*	0.0486	14.0296
70-79 vs. <50	1.71	0.0235	1.71	0.1086	21.3578	1.71	0.1324	31.7953	1.65	0.1005	18.3181
≥80 vs. <50	3.39	0.0465	3.39	0.2150	21.3579	3.39	0.3159	46.1198	3.15	0.2222	22.8188
Quintiles of deprivation
Q2 vs. Q1	1.05	0.0153	1.05*	0.0705	21.3659	1.05*	0.0747	24.0197	1.05*	0.0626	16.8745
Q3 vs. Q1	1.16	0.0166	1.16	0.0767	21.3711	1.16*	0.0873	27.6612	1.15	0.0687	17.1404
Q4 vs. Q1	1.27	0.0182	1.27	0.0839	21.2723	1.27	0.0934	26.3313	1.27	0.0762	17.5240
Q5 vs. Q1	1.48	0.0218	1.48	0.1007	21.3249	1.48	0.1046	23.0039	1.47	0.0885	16.4928

EMRR Excess mortality rate ratio, NBR Negative binomial regression, PEREM Piecewise exponential regression excess mortality model, RLE Relative loss in efficiency, SE Standard error, *p-value >0.05

ISSN: 1471-2288