Were the entry-points for policy adequately chosen, given the overall targets?

Most of the policies addressed one or more determinants of health inequalities, but as is evident from the tables, only a few had direct relevance for achieving the life expectancy or infant mortality targets by the year 2010. A few policies related to underlying determinants of health inequalities, but effects on inequalities in life expectancy were unlikely either because any prevented deaths are relatively rare (as in the case of homelessness or fuel poverty) or because they will occur in a future far beyond 2010 (as in the case of child poverty, education and pathways to work). There were no policies addressing income inequality as such, or other important determinants of health inequalities such as working conditions and excessive alcohol consumption. While some policies directly addressed the main causes of adult death, they aimed at improving average access to treatment, not inequalities in access (as in the case of improving access to cancer or coronary heart disease services). Only a few policies in the fields of prevention and primary care (eg, road safety, smoking cessation, diet, flu vaccination, access to primary care) addressed entry-points which were directly relevant for achieving the life expectancy target in 2010. A similar story applies to the infant-mortality target.

That the match between targets and policies was far from optimal probably relates to the fact that the 1999 ‘Action Report’ and the 2003 ‘Program for Action’ were as much a product of scientific evidence as of political opportunity. The ‘Independent Inquiry into Inequalities in Health’ had resulted in a huge ‘shopping list’ from which policy-makers could freely choose.18 Many of the policies included in the programme were part of strategies which the government already pursued for other reasons, for example because it had decided to invest in better education or to upgrade the funding level of the National Health Service upgraded to the European Union average. There were also no explicit and quantified links between departmental commitments, headline indicators and overall targets, and the overall targets had been chosen without regard to the policies they would require. The ‘Program for Action’ therefore was only a loose attempt to link targets and policies.

Were the specific policies to address these entry-points adequately chosen?

Ideally, the 1999 ‘Action Report’ and the 2003 ‘Program for action’ should have been based on evidence of the effectiveness of specific policies, but at the time this scientific evidence base simply did not exist. This can easily be seen in the 1998 ‘Independent Inquiry’ which laid the basis for the English strategy to reduce health inequalities but did not list a single evaluation study which showed that the proposed interventions could actually reduce inequalities in health. All the evidence was from observational studies showing plausible entry-points for intervention at most.3

Although this has been commented on at the time, the lack of an adequate evidence base seems to have been ignored by most of those involved in the development of this strategy, probably because it was felt that this extraordinary window of opportunity should be used for action on health inequalities, and that no time could be lost on evaluating policies before they were implemented on a large scale.19 Instead, attempts were made to assess the effects of these policies while they were implemented, but as a result most of these evaluations had a weak, non-experimental design.

Collecting evidence on effectiveness of policies to reduce health inequalities is difficult, because these are mostly complex interventions and should be evaluated not only for their effectiveness but also for their differential effectiveness.20 It was only around 2005 that the need to collect evidence on differential effectiveness of policies clearly came into focus. Reviews appearing in the second half of the decade then unambiguously showed the almost complete lack of existing evidence. ‘Umbrella reviews’ of population tobacco control policies and policies on wider social determinants showed that where systematic reviews dealt with differential effects by socio-economic status, only a minority actually found some weak evidence for inequalities reduction.21 22

It is therefore impossible to say whether the policies were adequately chosen. A scientifically rigorous answer would be ‘no,’ but for those who accept the urgency argument, the answer could be ‘this would have to be determined during implementation.’

Were these policies successfully implemented on the scale needed for populationwide impacts?

Without any doubt, the English strategy was implemented on a large scale. As shown in table 2, many of the ‘departmental commitments’ had large budgets, and some (such as the commitments towards child poverty reduction, social housing quality improvement and expansion of the Sure Start scheme) had budgets in excess of £1 billion or even £10 billion (ie, €1.5 billion or €15 billion at the former exchange rate).10 This is by far the best resourced of all the Western European strategies to reduce health inequalities which started during the decade. Remarkably, these ‘departmental commitments’ were also mostly met (table 2). Nevertheless, when one starts looking in more detail, a number of problems come to light.

First, during the implementation of the 2003 ‘Program for Action,’ many changes were made to the original plans. To some extent, these reflected ‘learning by doing’: adapting the strategy to experiences in the field, and also to new insights which had emerged. For example, during the implementation of the strategy, a stronger emphasis was placed on a limited number of ‘spearhead areas’ (areas with a concentration of social and health deprivation), because it was felt that focusing on these areas was likely to generate better results. It then seemed logical to also rephrase the overall life expectancy target in terms of differences between these spearhead areas and the national average. However, the speed with which policies were changed and evaluation targets adapted made partnership working and policy learning practically impossible.14

Second, as mentioned above, the ‘departmental commitments’ were not based at all on an analysis of what would be needed to reduce health inequalities. Thanks to a ‘health inequalities intervention tool’ which became available in 2007, the scale at which policies need to be delivered gradually became clear. This revealed many examples of misalignment of ‘departmental commitments,’ ‘headline indicators’ and overall targets. For example, even if the headline indicator for the ‘major killers’ were met, the life expectancy target would not have been achieved.15

Do changes in intermediate outcomes suggest that the policies were effective?

That the ‘departmental commitments’ were insufficient to achieve the targets became quite clear in the 2007 ‘Status Report.’ This concluded that while almost all departmental commitments were wholly or substantially achieved (table 2), the headline indicators and outcome targets showed a more disappointing picture (table 3). Some of the headline indicators showed reduced inequalities (less child poverty and narrowing inequalities in housing quality, educational attainment and uptake of flu vaccinations), but others suggested stable or even increased relative inequalities between socio-economic groups (heart disease and cancer mortality, child road accidents, teenage pregnancies, smoking).11

That some of the ‘headline indicators’ showed narrowing inequalities is a major achievement of the strategy. Unfortunately, however, no favourable changes were seen for some of the main drivers of inequalities in life expectancy or infant mortality, such as smoking. At the population level, smoking went down both in manual and in non-manual groups, and at a similar absolute rate of decline. For example, among men in the manual group the prevalence of smoking declined from 34% in 2001 to 28% in 2008, while it declined from 22% to 16% among men in the non-manual group.23 As a result, while absolute inequalities in smoking remained stable, relative inequalities in smoking prevalence increased. Similarly, although absolute inequalities in mortality from cardiovascular disease declined, relative inequalities increased.12

Not included in the headline indicators, because it did not relate to government objectives, is income inequality. Inequalities in income and wealth in England have remained unchanged or even widened during 13 years of Labour government, although at a slower speed than during the previous decades.24

Mixed intermediate outcomes also emerge from the evaluation studies which have been conducted for a number of specific policies. Health Action Zones could loosely be evaluated but the authors concluded that these ‘made little impact in terms of measurable improvement in health outcomes during their short lifespan.’25 The evaluation of the Sure Start programme concluded that ‘there are some signs that Local Sure Start Programs may be effective,’ but no health outcomes indicative of health inequalities reduction were observed.26 Despite large-scale implementation of smoking cessation services in deprived areas, the evaluation study showed that they made only a tiny contribution to reducing inequalities in smoking.27

The latter study is particularly interesting because it showed how important the scale of implementation is for achieving populationwide impacts. Despite huge efforts, only 2.3% of smokers in the lowest socio-economic group were reached by the programme and made a quit attempt, against 1.3% of smokers in the highest socio-economic group. After 1 year, only 0.42% of all smokers in the lowest socio-economic group had successfully quit, against 0.17% of all smokers in the highest socio-economic group.27

Did final health outcomes respond, as specified in the overall targets?

In view of these intermediate outcomes, it is not surprising that the overall targets have not been met. This already became clear in the 2007 ‘Status Report,’11 and was confirmed by subsequent independent evaluations, including the 2009 ‘Tackling Health Inequalities: 10 Years On’ review. This pointed to ‘significant improvements in the health of the population over the last 10 years, an improvement almost wholly shared by disadvantaged groups and areas, as measured by life expectancy and infant mortality.’ However, ‘[h]ealth inequalities between different groups and areas and the whole population […] persist’ and ‘[t]he current data […] shows that the gap is no narrower than when the targets were first set.’12

Figure 1 shows the most recent update of progress towards the life-expectancy and infant-mortality targets. The life-expectancy target, specified in terms of the gap between ‘spearhead areas’ and the national average, does not show any sign of narrowing—if anything, the gap has grown. Life expectancy has gone up in spearhead areas, but the national average has improved at a higher speed.16 The picture is not better for the gap in life expectancy between occupational classes—absolute inequalities in life expectancy have not narrowed since before Labour came to power in 1997, and have even widened for women.14 Similar results are found for the gap in infant mortality17 (figure 1).

One may ask whether it is not too early to assess the strategy's effect on inequalities in life expectancy. The timeline linking determinants to infant mortality may be short, but the timeline linking determinants to life expectancy may stretch over decades. That is theoretically true, but it does not stop life expectancy from changing rapidly—as can be seen in figure 1, life expectancy in ‘spearhead areas’ increased by 2–3 years in the 10 years since the baseline for the targets was set. The ‘problem’ is that life expectancy in non-spearhead areas (or in higher socio-economic groups) has increased even more rapidly. It seems that the strategy has been unsuccessful in harnessing the underlying forces to help to reduce, instead of widen, the gap.

One may also ask what would have happened in the absence of the English strategy to reduce health inequalities: would the gap in life expectancy or infant mortality have widened even more? Possibly: several European countries have witnessed a widening of inequalities in life expectancy over the past decades,28 and the ‘spontaneous’ trend in England may therefore also have been towards a widening of the gap. As shown in table 3, the strategy has had some favourable effects which may have contributed to less widening. This attenuation is likely to have been modest in size, however, because an unambiguous narrowing of the gap for a headline indicator relevant for inequalities in life expectancy was only seen for flu vaccinations (tables 2 and 3).

Finally, one may ask whether these targets were not simply ‘inspirational’ goals, intended to provide a general policy direction, instead of success criteria for which those responsible for the strategy could be held accountable. From the documents published by and for the Department of Health, however, it is quite clear that they were intended to be success criteria but also have actually been used as such.15–17 Although the Department of Health at the time was warned that achieving these targets might be difficult or impossible,29 it is reasonable to ask whether it would have been possible to justify the expense if the targets had been less ambitious.

Remarkably, the 2010 Marmot review argued that the targets were not ambitious enough. It criticised the life expectancy and infant mortality targets because ‘the current target does not capture the social gradient in health or the more complex patterning of health associated with other groups (eg, ethnic groups)’ and ‘mortality [does] not reflect health status or other dimensions of well-being through the life course.’14 This may be true, but would it not have been even more impossible to create and implement a strategy that would have achieved a 10% reduction in health inequalities across the whole gradient, or a 10% reduction in inequalities in healthy life expectancy, a new measure advocated by the Marmot review?