Something Not Unlike Research

Data.gov, a casualty of budget cuts?

A few weeks ago, I praised Data.gov, the website that was supposed to be a clearinghouse for making federal data public. Now, according to Nathan Yau,

Last week, there were rumblings over the end of the Statistical Abstract, and I suggested that it was just a sign of changing technologies. I thought that Data.gov and similar sites were the natural progression. Here's the probably with that argument. Congress is planning on shutting down Data.gov and other transparency sites in the next few months.

So not only was I wrong, but now we're taking steps backward. The budget for these technologies will be cut from $37m to $2m. According to Federal News Radio, USASpending.gov and Apps.gov/now will remain online until July 30 and then go black soon after.

I have no idea whether this is a demand of the Tea Party, or a decision by the Obama administration. Perhaps the sites were badly run. I suppose that if they are willing to cut $1 billion from the National Institutes of Health, there is little reason to expect them to scruple tabout ending an initiative for data transparency. 

 

Hans Rosling on Data Literacy

I am a data analyst at heart. I know that sounds terrible -- has anyone ever confessed that they are a CPA at heart? However, discovering statistical empiricism was life changing. I had been an undergraduate humanities person, in love with the European novels of the 20th century. But I suffered terrible anxiety when trying to write. I wanted to write what I thought was true, but could not find a way to ground my thoughts in something outside myself. I'm not saying that it can't be done, just that I never mastered it. Trying to write was just torment. 

Then, after college, I took a programming course on a lark (FORTRAN 77!), and discovered that I loved it. The first time something that I wrote actually ran -- not so easy with punchcards and UNIVAC JCL -- was like the time a friend and I rebuilt an engine, and the internal combustion cycle actually caught. Mastering statistical computation transformed my writing: I learned how to build an empirical argument to transform data into evidence; data being the gasoline, that is, the energy that powered the argument to a conclusion.

I think some people now get this feeling; at least, that is how I understand the widespread popularity of first Edward Tufte, and now Hans Rosling. These men have shown how to create narratives using data, above all by constructing beautiful, expressive visualizations.

Here is Rosling about the importance of beauty in visualization to make data persuasive.

The problem is that companies have a strict separation between their IT department, where datasets are produced, and the design department, so hardly any presenters are proficient in both. Yet this is what we need. Getting people used to talking with animated data is, to my mind, a literacy project.

My emphasis -- this is exactly my point. We need to find ways to get everyone engaged in data. What Rosling is doing is finding ways to communicate historical process through graphical animation: go to the Gapminder website to see for yourself what this means. 

So this is my new project... to learn Processing, the visual animation language, and drive it from R. We'll see what we can do.

Civilian Casualties in Afghanistan

This is a difficult topic, on which I have no expertise. However there is an interesting discussion of the military's own data on this topic in Science, including a well-rendered graphic.  

Glucose Monitoring and Vaccines: How Health Policy Is Made

There was recently a meeting of the Washington Health Technology Assessment Committee received a lot of attention. Among other actions, they reviewed evidence on the effectiveness of glucose monitoring by adolescent diabetics, and recommended that it be covered by insurance.

Avik Roy commented on this event.  Avik, I am glad to read, is "a supporter, in principle, of comparative effectiveness research." He believes that it is unlikely to save as much money as some argue, a fear that I share. But what can Avik mean by his title,  "How Comparative Effectiveness Research Gets Corrupted by Politics"?  These words are strange, given that:

• He agrees with me that the board made the right decision.
• He doesn't identify any aspect of the board's decision or procedure that was politically motivated.
• He doesn't adduce any evidence of corruption, or even the appearance of conflict of interest.

What am I missing? As far as I can tell, the panel looked at the evidence, made the right choice, and moved on. Probably, reviewing the effectiveness of adolescent glucose monitoring wasted their time. That merited the attention of Forbes and the Wall Street Journal?

Avik's post includes the, to my knowledge, original suggestion that conflict-of-intert policies themselves corrupt decision making by excluding from decision-making anyone with a financial interest in a medicine or device being evaluated. I think I speak for the consensus of medical school professors that such policies are required given the history of manipulation of scientific evidence by the members of the health care industry. But I need to hear more about what Avik is advocating before commenting further.

If Avik is looking for an example of political intervention in clinical care, I call his attention to the U.S. House of Representatives, which is reported to be considering changes in the funding for the Centers for Disease Control that would reduce support for state and local vaccination programs. Here is a decision a) being considered by politicians, not scientists or clinicians; b) that is almost certainly motivated by economic rather than health considerations; and c) is very difficulty to justify on the grounds of comparative effectiveness. 

 

 

Procedure and Communication: Lessons from the Navy

Perhaps the key challenge in chronic care is how to get many organizations and individuals -- hospials, insurers, specialists, patients... -- to work together to accomplish a complex task. I want to recommend a book that captures these challenges in a wholly different context.

James Hornfischer's Neptune's Inferno is a superb popular history of the U.S. Navy during the Guadalcanal campaign. Although the role of the U.S. Marines has been celebrated, many more U.S. and Autralian sailors were casualties. These casualties occurred in part because the Navy was, in 1942, at a disadvantage to the Japanese in numbers of ships, in quality of training, and in some critical technologies. They were nevertheless able to achieve a strategic victory, in part because of more or less suicidal bravery, and in part because the Japanese eventually withdrew to protect naval assets that they knew they could not replace. 

What's most interesting, however, is watching the Navy function -- or fail to -- in the extraordinary time compression of sea battle at night. Surprise was everything, and just a minute of confusion could and did lead to catastrophe. What they needed was a common understanding, within and between ships, of how to fight, and how to communicate in battle. They didn't have these things, and came near to being wiped out. The book protrays a large, complex organization learning and adapting, rapidly and under extraodinary stress. We can learn things here.

What's wrong with patient portals?

Patient portals are the web pages that hospitals, health care plans, and other health care organizations have constructed for patients. You get a personal page that you can sign into that is supposed to provide you with, at least, access to your own health data. More importantly, the patient portal should be a tool for patients to engage in self-management of their illnesses, as prescribed by the celebrated Chronic Care Model. Here is the login page for the Cleveland Clinic, one of the more developed examples of a pediatric hospital patient portal. 

This is a great idea.

But as Chilmark describes, most of these these things don't work in practice. In an insightful post, he argues that most patient portals suffer from numerous ills including:

• Providing a patient only limited access to their records. There are no clear and consistent policies in place today as to what a healthcare organization is obligated to provide a patient access to.
• Do not support portability that allows the patient to export personal health information (PHI) to another site/repository that they can then control (e.g., Google Health, HealthVault or other PHR).
• Do not allow for patient entered data nor the ability for a patient to annotate records.
• Rarely support transactional processes such as online appointment scheduling, Rx refill requests, eVisits.
• Are just about as user friendly as a clinician’s EHR [Electronic Health Record].

The problem with patient portals is that they are not seen as an integral part of the care process. In fact, we would argue that the use of the terms PHR [Patient Health Record] and EHR create an artificial division – let’s just call it a CHR (Collaborative Health Record) and be done with it. 

It is perplexing that these tools should be so poorly implemented. Every business on the planet is developing tools to engage their customers; there's quite a lot known about this. Why, like everything else in information technology, do things work worse in medicine? Here are five obstacles I've encountered, in order of increasing importance.

1. Legal worries. Attorneys fret that patients may put inaccurate data into the record, and the hospital will be liable for mistakes carried out based on uncorrected data.
2. Security problems. EHRs are armored against hackers, and these fortifications impede the effort to give patients access.
3. EHR developers have incentives not to develop records that follow open engineering standards. If they did, there customers could more easily pick up their data and leave, or -- worse -- engineer their own applications without paying expensive fees to the developers' consultants.
4. The physician side of health care data is not yet well organized. Why is another post, but the problem is just overwhelming. This isn't the fault of the developers, and I question whether there was a golden age of paper records that made sense and facilitated straightforward information retrieval. Anyway, it's a mess. And now you want patients dumping information in there too? 
5. Health care providers can bear only so much patient interaction, and for good reasons. I recently developed an interactive voice response telephony application to improve communication between outpatients and specialists about medications with dangerous adverse effects. The technology was easy. The challenge was finding ways to filter the patient input through physician extenders so that the docs were not exposed to a firehose of patient communication. We had to rethink the care process and develop protocols for how various emergencies would be handled, making sure that the relevant information was channelled to the person qualified to handle it. This kind of thing needs to be done throughout the system; and it's slow and expensive.

 My guess is that the specialty outpatient surgery shops will solve these problems first. On the one hand, they have simplified their work processes quite a bit by focusing on a small number of procedures, and hence their work is protocol-driven. On the other hand, they compete aggressively on customer service.

This will be a step forward, but speciality outpatient surgery clinics will not and cannot manage the overall health care of a patient. Designing the patient portal that allow a patient to work with a primary care doc to really manage their care will take a long time.

Community healthcare: Hope in Vermont

From Health Affairs, an early and encouraging report about Vermont's effort to improve care and reduce costs through primary care medical homes. For better or worse, Vermont's effort is neither socialism nor even single payer. The key innovation is

Having community health teams work with primary care providers to assess patients’ needs, coordinate community-based support services, and provide multidisciplinary care for a general population. A web-based central health registry... capture[s] all patient data.

The idea is to have a team of physician extenders to help primary care providers in a community (a region of about 20,000 Vermonters). Each community health team is

...led by a registered nurse [and] includes one or more additional registered nurses, who work within physician practices. Their primary duties are to track patients who are overdue for appointments or tests, manage short-term care for high-needs patients, check that patients are filling prescriptions and taking their medications appropriately, and following up with patients on their personal health management goals. [A] behavioral health counselor also [helps] providers identify patients with untreated depression or substance abuse.

Briefly put, the team helps patients manage their illnesses. This is not a program to save cost by denying care; it's a program to improve care by seeing that prescribed care actually happens. These are tasks that primary care could do, in theory, but in practice a small partnership can't afford these staff. 

It's early days -- just over two years into the implementation -- but there are some data indicating that the program is improving the quality of care for chronically ill patients. Moreover,

...the St. Johnsbury pilot—the first one, launched in July 2008—found significant decreases from one year to the next in hospital admissions and emergency department visits per 1,000 patients, and their related per person per month costs: Inpatient use and per person per month costs decreased 21 percent and 22 percent, respectively. Emergency department use declined 31 percent, and associated costs per person per month fell 36 percent. And overall use and costs per person per month dropped 8.9 percent and 11.6 percent, respectively.

The opposition to comparative effectiveness research

The Wall Street Journal has an editorial today, criticizing a board in Washington State that planned to review the effectiveness of blood glucose monitoring for child and adolescent diabetic kids. The board would review the available data, and -- I can't get past the WSJ paywall -- I'm assuming, make recommendations about whether WA state Medicaid should cover this. Aaron Carroll responds here.

I'm not a physician, let alone an endocrinologist -- and, above all, I haven't looked at the data -- but my guess is that after a review they will continue to support insurance payments for glucose monitoring for adolescents. The editorial apparently notes the problem faced in child health research, that there are often too few experimental data on children. This is a problem, because for many reasons what is effective and safe for an adult may not be so for children.

I won't criticize the WSJ, whose words I can't read. Rather, I'll comment on how it has echoed across the web. Many blogs that describe themselves as conservative have echoed the editorials second sentence, "a government board may decide that modern medicine is too expensive for kids with diabetes". That is hyperbolic enough. But it is interesting to look at the comments on the editorial. There are people out there who think that researchers plan to murder them.

Curtailing aggressive monitoring for either Type 1 or 2 diabetes will definitely cost lives and will significantly lower quality of life for many diabetics. Children will die much earlier. This is a death panel.

These, in turn, are more restrained that the comments on the Free Republic,

this is a forewarning of what is coming with Obamacare. It looks as though Governor Gregoire and the Democrat legislature has already established state health care death panels.

DeathCare has nothing to do with health. It is about murdering MORE children!

I don't know what to think about this. I've lived in very conservative parts of the country, and am close to many red state people, including family and in-laws. I haven't run into anything quite like this before. Where do these emotions come from?

 

A critic of childhood obesity prevention

Paul Campos attacks Michelle Obama's Let's Move as a form of bullying and predjudice against fat people. Beyerstein answers him here.

I was tempted to dismiss Campos based on his inflammatory tone, and the weakness of many of his arguments. However, he has several sound criticisms of current prevention efforts.

1. We do not yet sufficiently understand the physiology of obesity.
2. We do not understand why obesity has increased around the world. Nor are we certain whether that increae is still continuing.
3. The key measure of obesity, body mass index, is deeply flawed. 
4. We do not yet have effective, non-invasive ways to reduce the prevalence of obesity.

I am persuaded that the extent of US obesity and the severity of the health risks that it poses are such that we have to be engaged in trying to prevent it. Succeeding will require answers to Campos's challenges.

Public use scientific data, from Google

I recently wrote about data.gov, the US government data clearing house. By supplying free data sets that anyone can analyze to develop their own evidence-based view on policy, data.gov is part of an information infrastructure for a better world.

So here is Google, beginning something even better.

 

Google is developing an application programming interface that will allow someone to use Google Earth to map geospatial data sets, and they putting relevant data sets on line. It isn't currently open to the public, but I sense that it will be, and I urge Google to make this happen.

Are we seeing momentum toward routine public access to scientific data?