October 09, 2013|Categories: Analytics and Research
With the government shutdown still in effect, the Centers for Disease Control and Prevention’s (CDC) weekly flu monitoring activities are still halted or scaled down. So, as promised last Friday, as long as the CDC staff is in furlough, our athenaResearch team will continue to regularly examine flu-related data from our cloud-based network and provide a report each week.
Fortunately, incidence of flu diagnoses has changed little from Week 39 to Week 40 (Figure 1) of 2013. In Week 40, the week from September 30 – October 6, about 4.4 in 10,000 patients (0.044%) seeing primary care physicians were diagnosed with the flu. That compares to 3.7 per 10,000 last week and 3.6 per 10,000 at this time last year.
We also saw no discernible signs of a significant outbreak per region (Table 1).
In Week 40, 11.8% of patients who visited a primary care physician received a flu vaccination, up from 9.5% last week (Week 39) and 10.6% in the same week of 2012. These figures only reflect visits to PCPs, and do not include patients receiving vaccinations at retail clinics, pharmacies, schools, the workplace, and other settings.
Given that there are no detectable signs of a flu outbreak in this week’s report, we’d like to take some time to provide a synopsis of our data methodology, in response to inquiries following last week’s announcement and initial flu report.
Our calculations are based on data from our claims database of more than 42 million patients, with the data representing procedures and diagnoses that health care providers submit to insurers. Each week, we calculate appropriate ratios for distinct patients visiting primary care providers.
The most common question we’ve been asked in recent days is whether our sample is representative of the United States. We have providers across 49 states, covering a wide variety of practice types and sizes. We are better represented in certain states and regions than others, but we do not statistically adjust our sample to better reflect a national geographic mix. Our patient demographics are well-represented across age groups and gender. (You can read more about our population in this post on our cloud-based insights.)
As a point of comparison, Figure 3 below shows the rate of flu diagnoses taken from our network, overlaid with the rates of influenza-like illnesses (ILI) published by the CDC during the same time. The rates we report differ from the rates reported by CDC for two principal reasons:
- We do not use the same measurement due to differences in data type. The CDC’s use of ILI is based on clinical observations (symptoms such as fever and chills), as reported by their network of providers. We use flu diagnoses that are recorded by providers in claims they submit to insurers. It is thus important to understand that our data consist of cases in which a clinician suspects the patient has influenza, but where the diagnosis has not necessarily been confirmed by laboratory testing.
- The CDC makes statistical adjustments to estimate true national rates, while we calculate unadjusted rates across our national network.
Despite these differences, we find that the patterns for the two measures follow similar trajectories. Since the shapes of the curves and the timing of changes are similar, we believe our numbers can be used to identify weeks where flu rates are increasing and accelerating.
Regardless of when funding for CDC staff is restored – and we hope this occurs very soon – we’ll continue to mine our data, report weekly, and solicit ideas for useful research and analyses.