Deaths of Longitudinal Study members
The quality of death data for England and Wales (E&W) is known to be extremely high. Death certificates are required by law before the burial or cremation of a body, and as a result, virtually all deaths occurring in England and Wales are registered.
Delays in certification can occur if an inquest is required or a person died while abroad. Deaths will only be missed when no body has been found.
Deaths are linked into the LS both through routine notification of deaths of traced LS members from the National Health Service Central Register (NHSCR) and through a search of registration data for deaths of people born on an LS date. The use of these two different methods to identify deaths of LS members greatly improves the linkage rates.
Methods
The LS includes linked census and events data for people born on one of four selected dates in a calendar year. As such, the LS samples 4/365 of people born in non-leap years and 4/366 of people born in leap years. This represents approximately 4/365.25 of the population resident in England and Wales in any given year. The number of expected deaths occurring to LS members in any calendar year will be given by FIGURE 1.
In census years, the above formula uses the estimated numbers of England and Wales deaths from the start of the year up to the day before census night (pre-census) and from the day after census night until the end of the year (post-census). This allows to derive one pre and one post-census estimate of expected LS deaths.
Results
Deaths of LS members occurring after the 2001 Census are shown in Table 2.1 (males) and Table 2.2 (females) together with the relevant sampling fractions and linkage rates
Table 2.1. Deaths of LS members by year of death, 2001-2005 (males)
| Year of death | LS | E&W | Sampling
fraction |
Exp. in LS | Linkage
rate |
| Part 2001 | 2,037 | 170,129 | 1.20 | 1,863 | 109.34 |
| 2002 | 3,064 | 253,144 | 1.21 | 2,772 | 110.53 |
| 2003 | 3,018 | 253,852 | 1.19 | 2,780 | 108.56 |
| 2004 | 2,959 | 244,130 | 1.21 | 2,674 | 110.66 |
| 2005 | 2,778 | 243,324 | 1.14 | 2,665 | 104.24 |
| Total | 13,856 | 1,164,579 | 1.19 | 12,754 | 108.64 |
Table 2.2. Deaths of LS members by year of death, 2001-2005 (females)
| Year of death | LS deaths | E&W deaths | Sampling
fraction |
Exp. in LS | Linkage
rate |
| Part 2001 | 2,176 | 187,328 | 1.16 | 2,052 | 106.04 |
| 2002 | 3,506 | 280,383 | 1.25 | 3,071 | 114.16 |
| 2003 | 3,446 | 284,402 | 1.21 | 3,115 | 110.63 |
| 2004 | 3,248 | 268,411 | 1.21 | 2,939 | 110.51 |
| 2005 | 3,236 | 269,368 | 1.20 | 2,950 | 109.69 |
| Total | 15,612 | 1,289,892 | 1.21 | 14,127 | 110.51 |
Looking at the sampling fractions and linkage rates, it appears deaths are slightly oversampled in the LS; this happens more for women than for men.
As shown in Tables 2.3 and 2.4. sampling fractions and linkage rates have been increasing since 1971 both for men and for women.
Table 2.3. Deaths of LS members by decade of death (males)
| Year of death | LS | E&W | Sampling
fraction |
Exp. In
LS |
Linkage
rate |
| 1971-1981 | 30,922 | 2,934,222 | 1.05 | 32,134 | 96.23 |
| 1981-1991 | 31,897 | 2,860,109 | 1.12 | 31,322 | 101.84 |
| 1991-2001 | 30,947 | 2,686,861 | 1.15 | 29,425 | 105.17 |
| 2001-2005 | 13,856 | 1,164,579 | 1.19 | 12,754 | 108.64 |
Table 2.4. Deaths of LS members by decade of death (females)
| Year of death | LS | E&W | Sampling
fraction |
Exp. In
LS |
Linkage
rate |
| 1971-1981 | 30,068 | 2,886,971 | 1.04 | 31,616 | 95.10 |
| 1981-1991 | 32,837 | 2,921,134 | 1.12 | 31,991 | 102.65 |
| 1991-2001 | 33,819 | 2,904,946 | 1.16 | 31,812 | 106.31 |
| 2001-2005 | 15,612 | 1,289,892 | 1.21 | 14,127 | 110.51 |
The dual method used to link deaths into the LS captures both those deaths where an LS date of birth is recorded on the death certificate and also those deaths of LS members who are registered with a non-LS date of birth at death. This latter group has been increasing over the years, causing more deaths to be linked to the LS than expected from the original sample design.
The increase in inconsistencies in the dates of births on the deaths certificates can be explained by considering that over time an increasing number of individuals enter the LS by mistakenly quoting an LS date as their date of birth at one of the censuses or at NHS registration. Inconsistencies in the dates of birth affect around 8% of all LS members and they are flagged by a set of ad hoc indicators.
Further breakdowns of sampling fractions and linkage rates for deaths of LS members since 1971 are available from the Downloads box.
