Saturday, February 19, 2011

Update

This tumblr’s on hiatus for a month and a bit because I’m being buried with work. I hope to have a detailed post up in April summarising what the changes to the NHS are and why they’re bad, along with links to the evidence and the refutals.

I know some people have asked for a summary post they can copy and paste when arguing with people who don’t understand how destructive Lansley’s plans are so I’ll do that as well.

Saturday, February 5, 2011

Excellent article by Ben Goldacre

Ben Goldacre’s blog carries an excellent article today highlighting the evidence that debunks Lansley’s claims. Go and read it.

http://www.badscience.net/2011/02/andrew-lansley-and-his-imaginary-evidence/

Tuesday, February 1, 2011

Health and Social Care Bill 2011 Impact Assessment follow-up post

This is a follow-up to a previous post about the allegations that Lansley wants to pay private healthcare companies an additional 14%.

Juche Couture very kindly took the trouble to go through the impact assessment to find the evidence for this claim. Here’s what he found:

pg 42

B51. NHS services are provided by a range of provider types, including NHS organisations, private sector providers, social enterprises, and charities. Providers currently face different cost conditions purely because of their organisational type. These arbitrary cost advantages for particular provider types are referred to as ‘fair playing field distortions’.

B52. The existence of fair playing field distortions results in inefficiency if one type of provider is significantly advantaged or disadvantaged relative to another. Distortions can result in contracts not being awarded to the best provider following a competitive tendering process. Moreover, an otherwise efficient provider might be unable to compete under patient choice because distortions are enshrined in national tariffs. Tariffs are set on the basis of cost data provided by NHS organisations; if the costs of NHS organisations are significantly different from those of private and voluntary providers (e.g. they might be lower due to hidden subsidies, or higher due to a more complex case mix), then distortions to the fair playing field will result.

B55. A recent study of fair playing field distortions26 was able to quantify the impact of some of the distortions identified. The majority of the quantifiable distortions work in favour of NHS organisations; tax, capital and pensions distortions result in a private sector acute provider facing costs about £14 higher for every £100 of cost relative to an NHS acute provider. The pensions and cost of capital distortions are the most significant. (It should be noted that the extent of the distortion will vary by service depending on the input mix and capital employed. For example, the pensions distortion will be greater for higher paid staff, so a service that requires significant consultant input will have a higher pensions distortion than one which does not).

The report then went on to conclude:

pg. 44

B59. Evidence from the sources referred to above suggests that the following are needed: […]

Increased autonomy for NHS providers so that they can respond more flexibly to patient needs and compete for business with other healthcare providers

Strong policing of anti-competitive practices to restrain the market power of large incumbent providers and to give private and voluntary providers more opportunities to enter the market

Improved mechanisms for calculating efficient prices to avoid predatory pricing and cross- subsidy

A better understanding of how costs vary as a result of provider type and action to make the playing field fairer

pg. 52

B108. […] Once the net distortion facing different provider types is better understood, the tariff methodology could be developed in such a way as to move towards a fairer playing field by setting different prices for different providers in order to recognise different levels of implicit subsidies. (Note that to avoid compromising the fair playing field from a commissioner perspective, these charges need not be reflected in the charges faced by commissioners of care).

I suspect that’s where Spinwatch have pulled their 14% figure from and it doesn’t really support their claim. I think this serves as a nice reminder that statistical dishonesty crosses the political divide and that numbers should be treated with suspicion regardless of their origin.

A brief primer on the unreliability of cancer statistics

I originally wrote this for the Economist’s Blighty blog, who’ve kindly given me permission to republish it here.

DURING the run-up to the 2010 general election, David Cameron claimed during the televised debates that Britain’s death rate from cancer was worse than Bulgaria’s. A few weeks ago he reinforced his call for public sector reform by claiming that “our health outcomes lag behind the rest of Europe”. Comparing the NHS’s health outcomes to the rest of Europe or the Western world isn’t limited to Cameron and co: 72 point bold headlines decrying NHS failures in cancer care seem to be a staple of tabloid editors when they lack inspiration for the front page. Columnists across the political divide happily fall back onto criticising the NHS for not doing as well as the rest of the world whenever healthcare peaks in the political news-cycle. 

But much of what they say is nonsense. Last week John Appleby of the King’s Fund, a health think-tank, warned that comparing the survival rates the NHS achieves with those in other medical systems was “not straightforward”. His remarks were shrugged off by the coalition, but he is right, and the reasons he is right and interesting and illuminating. Health statistics are curious creatures, cancer statistics even more so. They’re deeply, heavily reliant on context to imbue them with meaning. Ripping them out of that context to use them in support of a political stance about the NHS robs them of meaning and therefore truth. 

For starters, cancer statistics live and die by the method of data collection. EUROCARE-4, the most recent publication of a Europe-wide cancer death registry, claims the UK has lower survival rates for the four most common cancers - lung, breast, colon and prostate - than the rest of Europe. But only 1% of German hospitals and 15% of French ones provided data to the EUROCARE registry. However nearly every British hospital submitted data thanks to the centralised registry run by the Department of Health. Suddenly the soundbite changes from “Britain worse than Europe for cancer survival” to “Britain worse than small proportion of Europe that bothers to submit data for cancer survival”. Bad sample sizes give bad data, which gives bad statistics when you don’t give their context.

Next, cancer statistics depend on how the cancer was diagnosed, which varies geographically. The World Health Organisation’s  World Health Statistics report 2009 says the UK has a cancer mortality of 147 per 100,000 people. The same report says Namibia has 91 cancer deaths per 100,000 people, Bangladesh has 107 per 100,000 and North Korea 95 cancer deaths per 100,000 people.  Surely the NHS, one of Britain’s proudest state institutions, can’t possibly be worse than, of all countries, North Korea?

Britain has the most equitable access to healthcare in the world, multiple different screening programs for various cancers and pre-cancerous conditions along with strong post-mortem requirements and a high skill-set among pathologist doing those post-mortems. North Korea and Bangladesh have none of those. Patients with cancer there may lack a doctor to diagnose them before dying at home at and being buried without a post-mortem providing the diagnosis. Inevitably, this means both countries look like they have better cancer survival rates than the UK, where cancers are found during screening programs, during routine consultations and at post mortem, which artificially inflates mortality rates in comparison to less developed countries. This leads to bad statistics when this background isn’t made clear.

Cancer deaths are affected by other diseases too. The average life expectancy in Namibia is 60, whereas life expectancy in Britain is 80. Cancer is mostly a disease of old age; for example the average age at diagnosis for breast cancer is 65. In Namibia people simply don’t live long enough for cancer to become a leading cause of death, often dying at a younger age of preventable infections and trauma due to lack of access to care. Inevitably this means they appear to be “better” at surviving cancer than the Brits—unless, of course, you give the context.

Next, cancer is not a universal disease. There is a huge difference in the survival rate for stomach cancer between Britain and Japan which, if presented alone, makes Britain look rather bad.  But stomach cancer has a much higher incidence in Japan (for poorly understood reasons), which means there is a strong nation-wide screening program to catch it early.  It’s a rarer cancer in the UK and a screening program wouldn’t be cost effective compared to screening for other cancers that are far more common. Many British patients with stomach cancer present too late to be cured compared with Japanese patients who are caught much earlier when screened. 

This runs us into the next rule: screening programs change everything for the better and for the worse. We’ve already seen how screening programs that exist in one country and not another can skew good international comparisons. America does cervical cancer screening every year, the UK does screening every 3 to 5 years, which means the Americans diagnose more cervical cancers. However not all cervical cancers found at screening will become fully cancerous and need treatment. There’s no way to tell through the screening program which are which, so many American women are having un-necessary treatment compared to Brits. When talking about cancer diagnosis and mortality, the details of screening programs must be given otherwise the statistics lose too much context to make sense.

Finally, cancer statistics are by definition out of date by the time they’re published. The EUROCARE-4 statistics involve patients diagnosed with cancer between 1995 and 1999, who were followed through to 2002 before the data was published. But in 2000 the NHS Cancer Plan was published, changing the uncoordinated mess that was British cancer care at the time into a formal, structured system with a greater emphasis on screening and on preventative programs to stop cancers from developing in the first place.  The results of this plan will take years to be observed because of the time lag between a cancer first developing and eventually being diagnosed, making it meaningless to use statistics from before the plan was implemented to criticise the current system. 

In an ideal world everything would be put into context and statistics would be used honestly but, until that happy day, it’s wise to reflect critically whenever people make glib claims about NHS under-performance on cancer.

Sunday, January 30, 2011

Statistics and time

This is going to be a unsourced post because I’m explaining concepts rather than hard figures so apologies in advance.

I think one of the reasons why Cameron and Lansley are getting away with abusing and whoring out statistics to support their case is a general ignorance about both the research process and about how statistics actually work. 

A common theme they’re both trotting out at the moment is outcome data. Specifically, how the UK spends the same or more on health and gets worse outcomes than the rest of Europe. “Outcome data” covers a lot of things, from number of people surviving cancer to the number of people dying of a heart attack; it’s basically numbers showing how well the healthcare system treats a particular disease.

The problem is that outcome data is very wooly and subject to a huge amount of context. The classic example I love using to explain the importance of context is that you’re more likely to survive cancer if you live in North Korea than if you live in the UK. It’s not because North Korea is better than the UK, it’s because you’re more likely to be diagnosed with cancer in the UK thanks to screening programs and easy access to healthcare. It’s easy to die of cancer in North Korea and have no-one realise that’s what killed you. It’s even easier to die of other causes in North Korea before you get old enough for cancer to kill you, because the biggest risk factor for developing cancer is old age.

The specific context to statistics that I want to discuss in this post is the importance of time.

Statistics are, by definition, out of date by the time you read them. Studies can either be prospective, where the researchers collect data as they go along, or retrospective, where the researchers go find data from the past. Both types mean the data is from the past by the time it’s published. On top of that, there’s a lengthy lag between finishing collecting your data and it finally being published - researchers spend a lot of time analysing the data, then it has to go through an institutional review process and then it has to go to a journal for approval. The journal will take weeks to months to make a decision and it’s not uncommon for papers to be approved subject to changes, which then increase the time before publication even more. 

The end result of this is that outcome data doesn’t tell you what’s happening right now, it tells you what was happening in the past. Thanks to the constant political meddling in the NHS, this means politicians are often using data from a time when the NHS was completely different to the way it is now. It’s like taking your car to a racetrack for a race then fitting a turbocharger to the car after the race. You then complain the car is too slow because you were beaten in the race. You’re completely ignoring the difference the turbocharger made to the car after the race was over, instead you’re obsessing about what happened before you made the car faster.

The biggest example of this in the NHS is ignoring the state the NHS was in after decades of being gutted by the Tories prior to 1997. Because the NHS was so badly affected by underfunding, it was starting at a huge disadvantage compared to the rest of Europe - you were more likely to die of a heart attack or cancer in the UK than in Europe because there was essentially no infrastructure to treat you. Since 1997 Labour poured money into the NHS to make up for the chronic underfunding during Tory governments, which resulted in a massive improvement in outcome data.

This graph, which I’ve posted before, is the perfect illustration of this concept. Although in 2007, the UK is worse than France, it’s a hell of a lot better than it was in 1979 and the trend can be extended to predict that we’ll overtake France in 2012.

Quoting outcome data for a specific point in time without clearly illustrating the concept and explaining the trend makes that data worthless.

Most of the outcome data Cameron uses relates to cancer, where the classic measurement is the 5-year survival rate: how likely you are to survive for 5 years after you’re diagnosed with cancer. As a result, you’re looking at data that’s subject to context over 5 years, which is a huge amount of time in the medical world. The scope for that data to be influenced by other factors is enormous, which makes it very hard to use it to accurately make claims about the NHS under-performing at this specific point in time. 

As an example, if a new drug was introduced 2 years ago, the effect it has on the 5-year survival rate won’t be apparent for at least another 3 years, possibly longer bearing in mind the long delay between collecting data and publishing it that I explained earlier. If you look at the survival rate right now then you’re actually commenting on the state of the NHS 5 years ago - it’ll be another 3 years before you can comment on how the NHS is at this point in time when it comes to treating that cancer.

I hope that makes it a little clearer why statistics aren’t just simple numbers that can be whored around without context. They’re so deeply sensitive to time that you need to be critical about what they purport to mean.

Anonymous asked: I'm an NHS manager (of sorts) can I submit management related stuff?

Absolutely, the tumblr needs as many different articles from as many different perspectives as possible. I’ll anonymise everything and if you take the care to submit anonymously then you’ll be free from repercussions.

Health and Social Care Bill 2011 Impact Assessment

I’ve been told the source for Lansley wanting to pay private healthcare companies an additional 14% is buried deep inside the Impact Assessment of the Health and Social Care Bill 2011.

It’s a huge document and I don’t have the time to read through it to confirm the accuracy of that number. If anyone else does, please do leave the reference in a comment or send it to me via the submission tab on this tumblr and I’ll republish it with full credit.

Left hand, right hand, elbow, arse

The Telegraph is reporting today that £400 million is being spent on mental health services to:

"Ministers want the doctors to stop simply prescribing pills to patients with a range of conditions including eating disorders, self harm, addictions, attention disorders and post-natal depression – a practice which has led to claims the NHS is turning Britain into a "Prozac nation"."

"In addition, ministers want the NHS to "cure" up to one million sufferers by the next general election, expected in May 2015, by which time they hope that more than 70,000 people currently out of work with mental health problems will be back in employment"

It’s a lovely proposal but it suffers from a case of the left hand not knowing (or worse, not caring) what the right hand is doing. Because the NHS’s funding is being cut*, the “unsexy” mental health services are often targeted first because they lack the visibility of popular illnesses like cancer and because they don’t produce nice statistics showing patients being cured.

One friend works for a mental health trust that has seen their budget cut by £20 million over 4 years. This is money that could be used to fund treatments like cognitive behaviour therapy, the most effective treatment for many mental illnesses. More importantly, it’s often the only alternative to sticking a patient on psychiatric medication. Without specifically ringfencing the mental health budget from cuts, the £400 million isn’t going to cover the lost funding.

Worse, by moving the social care budget away from councils and into the NHS and utilising private companies like ATOS to stop disabled people from receiving the funding they need to live an equal life, the chances of getting people with mental illnesses back into work is pretty slim. They’ve killed the access to work scheme that provided employers with the funding needed to make adaptations for ill and disabled workers, reducing their employability further.

I’d seriously encourage you to spend an evening reading through the archives of this blog. While it focuses on the effects of the cuts on disabled people, many of the stories apply to mentally ill people too. 

* the NHS needs a 3% budget increase every year to provide the same level of service because of the increasing burden of “diseaseogenic” conditions such as obesity. The Coalition hasn’t provided this increase, which has resulted in what is essentially a budget cut.

Why targets matter

Lansley likes to make a big fuss over scrapping NHS targets, claiming they actually worsen patient care by distorting clinical priorities. There’s some merit to the argument but they’re confined to fringe cases like the A&E 4-hour waiting time target. For the most part, targets are essential to make sure NHS trusts have something to work towards - they’re essentially a substitution for profits in any UHC system.

Last July the Coalition scrapped a target requiring patients to receive treatment within 18 weeks of an outpatient appointment. Now the number of people waiting more than 18 weeks has jumped by 25% . There’s also a 90% rise in the number of people waiting more than 6 weeks for diagnostic imaging (X-rays, CT scans, MRIs, ultrasounds etc).

This is not something where you can argue that clinical care hasn’t been compromised - waiting times leave patients suffering until they receive treatment, waiting times are unacceptable. Waiting time targets matter, scrapping them for political reasons is intolerable. 

European cancer statistics

This is just a placeholder post until I can post a long article on cancer statistics that I original wrote for another website. 

Comparing UK cancer deaths to the rest of Europe is a pretty common way of claiming that the UK is the “sick man of Europe” and that we need to spend more money on ultra-expensive chemotherapy drugs.

However, it’s a crap argument.

Most people use the EUROCARE-4 statistics when talking about European cancer deaths. It’s a cancer registry based in Italy that collects statistics voluntarily sent to it by participating hospitals. Using those statistics, the UK is worse than other European countries for a whole host of cancers, particularly the big 4 (breast, lung, prostate and colon).

The problem is that only 1% of the German population is included in the EUROCARE database and 15% of the French population. 100% of the British population is included thanks to the NHS’s strong centralisation of data collection. The results from EUROCARE-4 are complete nonsense because of the absurd difference in sample sizes between countries.

Anyone who argues the NHS is shit because of our poor cancer survival rate is just talking nonsense.

Case for NHS reform oversold.

I’d seriously urge everyone to read this article in the BMJ. It’s written by the King’s Fund and does a fantastic job explaining why a lot of the Coalition’s arguments for reforming the NHS are based on poor interpretation of statistics. It’s one of the reasons why I set up this Tumblr and that article does a much better job explaining it than I ever could.

I’ve put some short snippets below along with my own interpretations if you need a quick reference guide when arguing with people who won’t read articles you send them.

One of the common arguments for NHS reform is poor performance compared to other European countries. However this argument is based off single statistics isolated in time rather than the trend. Looking at the long term picture in the graph, we can see that the UK’s actually been doing better than the rest of Europe and will eventually exceed them. There’s lots of reasons but essentially:

1) Thrombolysis became more widespread in the 80s as a treatment for acute heart attacks. Angioplasty replaced it in the 90s and the massively increased funding for cardiac services means everyone should be near a centre that specialises in treating acute heart attacks, which is why the decline extends into the 00s.

2) NICE (which the Coalition are gutting) published very strong, evidence-based guidelines for what drugs to prescribe post-heart attack with the intent of preventing cardiac patients from having another one (“secondary prevention”). Centralised targets for getting patients onto these guidelines helped push the decline in the death rate.

3) Primary prevention programs and centralised targets for getting hypertensive patients onto treatment is preventing people from having heart attacks in the first place.

One of the importance reasons that comparison is pretty bad is that the French have a lower incidence of heart disease, the so-called “French Paradox”. So while the French have had access to all 3 things listed above, they just started out at a much lower incidence of heart disease in the first place.

The article also touches on cancer deaths at the end. I have a big long article written on that for another website - once they publish it, I’ll relink it here.

The public doesn’t support NHS reform

One of Lansley’s and Cameron’s favourite talking points is how the public supports their push for reforming the NHS.  

The Guardian reports on a YouGov poll debunking this claim. Selected quotes below:

"The YouGov survey found that only 27% of people back moves to allow profit-making companies to increase their role the NHS."  

"Overall, 50% of the 1,892 respondents opposed the policy. But hostility was more evident among Lib Dem voters, 56% of whom said they were against, with just 30% in favour. Conservative supporters were also split: while 46% backed the use of private firms, 32% were against."  

"The poll also reveals that half the public oppose the new GP consortiums that will emerge across England, using private management groups to help them with finance, planning and management. Lib Dem voters are even more hostile to this, with 57% against, while among Tory voters, 40% back it with 38% against."  

Please note the poll was commissioned by Unison, the public sector workers union, so, as with any externally-commissioned work, the results have the potential to be biased. Both YouGov and Unison have not yet released the raw data, including the questions asked.  

Personally I’m curious about this government statement in the article:  

"A Department of Health source said: ‘We want patients to be able to choose to be treated wherever they want, whether it’s an NHS hospital or in the voluntary or private sectors. More choice and more competition will lead to benefits for patients. Labour introduced this policy in 2007 and since then the number of patients choosing an independent provider has risen to more than 16,000 a month. But, unlike Labour, we won’t pay the private sector any more than we would pay the NHS.’"  

Spinwatch via Left Foot Forward claim that Lansley is proposing to pay private sector companies an additional 14% over what the NHS would be paid for the same service “to level the playing field”. They don’t source that claim, however.