A February Government Accountability Office report released today chides the Food and Drug Administration for failing to ensure the safety of common food additives. By law, manufacturers may assign additives a "generally recognized as safe," or GRAS, designation, a process that the FDA neither monitors nor verifies in most cases, according to the GAO.

Companies are supposed to make GRAS determinations based on the latest and greatest scientific evidence. Later, if new information turns up or if citizen petitions are filed with the FDA, additives can potentially be stripped of their GRAS designation.

One would expect that the FDA would keep tabs on GRAS substances. One would be wrong. Perversely, companies are not required to tell the FDA they have made a GRAS determination. (Many do notify the FDA voluntarily.) This makes it kind of difficult for the FDA to ensure food safety, according to GAO:

Once a GRAS substance has entered the marketplace, FDA would find it difficult to identify that substance as the potential source of a food safety problem, especially if FDA is unaware that the substance has been determined to be GRAS. Food products may contain numerous ingredients, including GRAS substances, making it difficult, if not impossible, for public health authorities to attribute a food safety problem to a specific GRAS substance. [Emphasis added.]

The report has many other specific criticisms of the FDA. It seems as though there are two overarching problems. First, regardless of whether the agency is aware that a GRAS determination has been made, the FDA tends to sit back and wait for new information to come to it. Second, when risk information does find its way to the FDA, the agency is slow to respond.

The report also delves into the issue of nanomaterials – tiny devices and products constructed of materials 100,000 times smaller than the width of a human hair. Although the FDA acknowledges that much remains unknown about nanomaterials (FDA has said it is not even comfortable defining nanotechnology), and even though some fear nanomaterials may have adverse effects on public health and the environment, manufacturers may designate nanomaterials as GRAS food additives. “Nevertheless, the decision to notify FDA of a GRAS substance, even one that contains engineered nanomaterials, is still voluntary,” the report says.

The GRAS problem is typical of the U.S.’s shoot-first-ask-questions-later approach to regulating. Safety is presumed, and regulation can only occur after evidence to the contrary reaches a critical mass.

GAO concluded the report by recommending several ways the FDA can sharpen oversight of GRAS determinations, including mandatory notification. “FDA generally agreed with the report’s findings and recommendations,” GAO said.

Image by Flickr user PunkJr, used under a Creative Commons license.

Earlier today, the Recovery Board released the list of Recovery Act recipients who did not file during the second reporting period. According to the Board, recipients of 1,036 awards failed to file during this quarter, which was from Oct. 1 through Dec. 31. That number represents a whopping 76 percent decline from the first reporting cycle, which saw 4,359 missing award reports, and is less than one percent of all the award reports. Equally good news is that of the 1,036 missing reports, only 389 were from "repeat offenders," or recipients who failed to file in both quarters.

The trend from the non-filer list echoes other data the Recovery Board also posts, such as the late filers and report corrections. According to the Board, the second reporting cycle saw half as many late reports, which are award reports filed after the filing deadline. This past cycle, 7.3 percent of recipients filed late, down from 14.9 percent of reporters in the first round of reporting, and of these late reports, a vast majority of them were not repeat offenders. Similarly, only 12.75 percent of award reports were changed after the fact (recipients can change their reports for several months after the filing deadline), as opposed to over 21 percent in the first round.

These data sets show what we've been assuming would happen: Recovery Act recipients are learning. As time passes, and recipients learn how the reporting system works and how they're supposed to file, the number of reporting errors are slowly decreasing. More recipients are reporting on time, fewer are forgetting to report (or are understanding that they have to report), and there are fewer mistakes to correct after the fact. And this progress is despite the fact that there are more award reports in the second round than the first.

This trend will probably continue over the coming cycles, although it will be interesting to see if it hits a floor at some point, i.e. if there is some baseline level of user error we just can't escape.

The next important statistic to look at will be the change in data quality between quarters. While we know recipients are learning how to file, what we don't know is if they are entering better quality information this time around. Are there fewer award amount errors? Fewer job counting errors? Late reports are bad, but flawed data is even worse.

Image by Flickr user ekilby used under a Creative Commons license.

Yesterday, a group of organizations, including OMB Watch, submitted concerns with the high-value datasets published on Data.gov in compliance with the requirements of the Open Government Directive (OGD) issued on Dec. 8. The OGD required that agencies submit at least three high value datasets within 45 days through Data.gov.   These groups outlined the major problems with the site and its implementation thus far.  I have summarized these issues below.

Yesterday, a group of organizations, including OMB Watch, submitted concerns with the high-value datasets published on Data.gov in compliance with the requirements of the Open Government Directive (OGD) issued on Dec. 8. The OGD required that agencies submit at least three high value datasets within 45 days through Data.gov.   These groups outlined the major problems with the site and its implementation thus far.  I have summarized these issues below.

Format & Usability

A major concern of the community is that releasing data in specific formats may make it more usable to coders and the tech-savvy, but not to the general public writ large.  If data is solely released in formats such as XML or CSV leaving the majority of the public unable to decipher these raw formats then Data.gov’s attempt to make the government more transparent has failed.    What good is information if we can’t read it?

The solution?  We proposed that the administration strike a balance between releasing data in machine readable formats and presenting the data to the public through web-based interfaces.  Web-based interface can be designed in a user-friendly way that aggregates the raw data for quick and convenient access.

Definition of High Value

The OGD mandated that agencies release three “high-value” datasets on Data.gov.  However, the Sunlight Foundation noted that only 16 of the 58 datasets posted by major agencies were previously unavailable.  The vast majority of released datasets were already online but not in machine readable format.  This meant that the administration only picked the low-hanging fruit in its first data release.  On the other hand, the data was now available in a central location and in a better format.

Still, this leaves open the question of how the government is defining “high value.”  According to the Open Government Directive, high value data is information that can be used to increase agency accountability and responsiveness; improve public knowledge of the agency and its operations; further the core mission of the agency; create economic opportunity; or respond to need and demand as identified through public consultation.  Yet, the agencies do not have to demonstrate how the releases they submit qualify under these categories.  To resolve this issue we expressed the need for such a requirement.  Other things we asked for:  notations that indicate which datasets are already available and unavailable as well as datasets that help hold agencies accountable for their policies and spending decisions.

Data Quality

This third issue mainly centers on the fact that some datasets could not be opened, were missing portions, or missing headers.  Missing headers, of course, means the data cannot be used even by coders.  Moreover, it was discovered that some of the datasets were being quietly removed from the site without public notification.

Here, we stressed the need for a better feedback mechanism than what exists on the site.  We made the point that there needs to be a system to report problems with specific datasets.  Further, we asked that all datasets on Data.gov be directly associated with their code sheets.
Ultimately, this is a great first step in showing the amount of data the agencies are capable of putting out in machine readable format.  Never before have we been able to access so much raw data in one place.  Despite the short deadline for this disclosure, several executive agencies released more than the required three datasets.  The implementation, however, needs to be improved through  creating public facing interfaces, requiring agencies to demonstrate the value of the data, and by providing a means of user feedback.

In order to improve Data.gov and the range of data included on the site, the administration is welcoming comments on its blog, Join the Dialogue. Additionally, Data.gov allows users to rate each dataset for ease of access, usefulness, data utility, and an overall ranking.

To read the group’s arguments and other points of contention in full, see the letter we sent here.  Feel free to give us your feedback.

Update (02/17/10): "Last-Minute Changes Will Improve Air Pollution Monitoring, EPA Says."

It’s looking more and more like the White House Office of Information and Regulatory Affairs (OIRA) overruled the EPA in a decision to place more air pollution monitors near major roadways.

On Monday, EPA finalized a rule which limits nitrogen dioxide exposure and sets up a monitoring network along heavily-polluted roads. The trigger for placement of a monitor is the population of a metro area. If the area has a population of 500,000 or more, it gets a monitor.

But originally, EPA indicated the threshold would be a population of 350,000. The change means 41 fewer monitors will be placed around the country. [I was wrong, see the update above.]

Based on documents in EPA’s rulemaking docket, we know that the change was made during the OIRA review of the draft of the NO2 final rule. We also know that at least one high-level EPA official was pushing for the lower threshold.

In an email dated Jan. 19, an EPA employee wrote, “We are willing to put forward an alternative threshold for the first tier of the near-road monitoring network for you to discuss with your management at your 2:30 meeting,” and then suggested the 500,000 population threshold.

What prompted this 11th hour policy change? Who knows. Semantically, “we are willing” suggests a certain level of coercion, but that’s mere speculation on my part.

The more interesting exchange came the next day when Lisa Heinzerling, EPA’s Associate Administrator for policy and an adviser to EPA Administrator Lisa Jackson, sent a follow-up email to OMB. It simply stated, in its entirety, “EPA does not support the alternative threshold described in the email below.” The email was in direct response to the EPA's employee's proposal, though the email itself was addressed to OIRA.

Despite Heinzerling’s clear protest, the threshold was raised. Sometime between Jan. 20 and Jan. 22, the day OIRA concluded its review of the NO2 rule, somebody changed his or her mind, or somebody was overruled.

We’re still investigating this, but if EPA was indeed overruled, it would be a clear-cut instance of political interference in an inherently scientific decision. Regulators need reliable data to enforce EPA’s new NO2 standard which, for the first time, targets short-term exposure to the pollutant. NO2 begins to wreak its havoc even in short bursts, causing respiratory illnesses “particularly in at-risk populations such as children, the elderly, and asthmatics,” EPA says. EPA describes the urban and near-roadway environments that will be closely monitored as “high-risk.” Given that information, EPA needs to go about its business monitoring pollution and rigorously enforcing the new standard. Instead, because of possible White House interference, the agency will have to operate with one hand tied behind its back.