Consistency and precision of cancer reporting in a multiwave national panel survey

Table 1 Characteristics of the Cancer Survivor Sample

			Consistent vs. inconsistent response patterns¹
	All cancer reports		Consisten t		Inconsistent		Difference
N	978	100%	640	71.0%	262	29.0%
# waves reported (non-missing)							**
1	76	7.8%	N/A		N/A
2	78	8.0%	67	10.5%	11	4.2%
3	118	12.1%	78	12.2%	40	15.3%
4	706	72.2%	495	77.3%	211	80.5%
Respondent²							n.s.
Only self	497	50.8%	328	51.2%	146	55.7%
Only proxy	287	29.4%	187	29.2%	64	24.4%
Both self and proxy	194	19.8%	125	19.5%	52	19.8%
Primary cancer site (N = 628)³							*
Breast	127	20.2%	116	21.0%	10	18.9%
Colon	45	7.2%	43	7.8%	0	0.0%
Lung	18	2.9%	16	2.9%	1	1.9%
Lymphoma	20	3.2%	15	2.7%	2	3.8%
Melanoma	31	4.9%	27	4.9%	4	7.5%
Prostate	69	11.0%	65	11.8%	4	7.6%
Nonmelanoma skin	88	14.0%	83	15.0%	4	7.5%
Uterine	32	5.1%	28	5.0%	4	7.6%
Ovarian	15	2.4%	13	2.4%	1	1.9%
Cervical	50	8.0%	34	6.2%	10	18.9%
Other/DK/NA	133	22.2%	112	20.3%	13	24.5%
Limitations due to cancer							***
Some or a lot at any wave	272	27.8%	198	30.9%	48	18.3%
Timing of diagnosis (N = 964)
Mean year of diagnosis⁴	1995.3	(8.7)	1996.0	(8.9)	1993.3	(8.6)	***
Mean age at diagnosis⁴	49.4	(18.2)	52.1	(16.9)	43.1	(18.6)	***
Reports of diagnosis year⁵ (N = 504)							n.s.
No discrepancy	101	20.0%	80	20.6%	21	18.1%
Discrepancy of 1 year	151	30.0%	121	31.2%	30	25.9%
Discrepancy of 2-3 years	119	23.6%	89	22.9%	30	25.9%
Discrepancy of 4+ years	133	26.4%	98	25.3%	35	30.2%
Survivors' characteristics
Female	620	63.4%	386	60.3%	187	71.4%	**
Non-white	265	27.1%	153	23.9%	91	34.7%	**
Not married⁶	466	47.7%	279	43.6%	125	47.7%	n.s.
Rural residence⁶	152	15.5%	106	16.6%	35	13.4%	n.s.
Education⁷	12.7	(2.9)	12.9	(2.9)	12.4	(2.8)	**
Year of birth⁷	1945.8	(18.0)	1943.7	(16.7)	1950.1	(18.5)	***

* p < .05, **p < .01 ***p < .001
Note: The difference between consistent and inconsistent reports was evaluated using chi-squared tests and t-tests.
1 In/consistency is defined only for respondents with two or more nonmissing cancer reports.
2 All categories can include missing reports for some waves - i.e., 'only self' includes self or missing, but no proxy reports.
3 Collected only in 2005.
4 Calculated as the mean of all nonmissing timing reports.
5 Discrepancy is defined only for respondents with 2+ reports of diagnosis year. This constraint excludes 386 individuals who reported cancer occurrence more than once but the timing of diagnosis only once.
6 Defined as not married, rural, and south if respondents reported those categories at any of the 4 waves.
7 Defined as the mean of multiple reports per individual. There was little variation across waves in the reports.

ISSN: 1478-7954