Response to Jones and Milne

We hoped that publication of these papers would spark public debate on such a complex and important matter. We noted that our conclusions were preliminary, not definitive, and called for further investigation. As such, we are grateful to those engaging in this debate.

Jones’ letter ¹ regarding further avenues for exploration is helpful. Unfortunately, in both papers, we were at the limit of permitted references, so others are in web appendices. Jones’ suggestions may provide pointers to future research into this complex occurrence, although any unknown infectious agent has struck with remarkable temporal coincidence to public funding cuts.

We thank Milne for drawing attention to an error in our first paper. During preparation of these complex papers, two figures were inadvertently switched. We regret that Figure 1 does not show age-standardised mortality rates (ASMR), but crude mortality, which is reflected in the accompanying text. The correct figure, using Office for National Statistics data, ² is shown below. However, this makes little practical difference to our inference as extrapolating forward the annual rates of decline between 1994 and 2011, and comparing observed values in 2015 with predicted values, the relative increases are almost indistinguishable, at 11% and 12% for ASMR and crude mortality, respectively.

Figure 1.

Age-standardised mortality rate and crude mortality rates, all persons, England and Wales 1994–2015.

Second, we should have clarified that the proportionate changes shown in Figure 2 use mid-year to mid-year data, the only source by single year of age available from Office for National Statistics when we wrote the paper. ³ Its advantage over the calendar year data, used by Milne and available subsequently, is that all deaths were among individuals counted in the denominator, but neither is ideal. This explains why his differ from ours, but again the implications are similar as both show substantial increases at every age.

We did note how different datasets and calculations give inconsistent results. For example, if the five-year average of weekly deaths is compared to the corresponding weeks of 2015, the extra deaths total 35,722. ⁴ If the mid-year numbers of deaths between 2014 and 2015 are compared, the excess in 2015 exceeds 52,000. ⁵ We standardised by single year of death to produce the most conservative estimate, but we had to use mid-year estimates, as noted above. Perhaps the most important message is that 15 months since the end of 2015 we still lack comprehensive published data on deaths by age/cause separately for Wales and England, to enable comparison of the effects of different policies, or by local authority.

As we noted in our paper, any reversal in mortality trends should be a warning that something may be seriously wrong. We contend that what Office for National Statistics describes as ‘the largest annual rise in deaths since the 1960s’ ⁶ warrants urgent review. Worryingly, the weekly pattern in early 2015 appears to be repeating in 2017, with the number of deaths at ages 75+ in week 4 31% higher than in the same week in 2014. We hope that our colleagues’ responses are the first of many further interrogations of the data, and we welcome such debate.