ILI forecasts for 2017-2018-Epidemic Week 47

This is the fifth week of the CDC Flu Challenge.  Region 4 remained above its onset value (for three consecutive weeks) and Region 1 and the national have crossed their onset values in the past week. The %ILI values for Region 6 were adjusted downward, to just below the onset value of the region.  Currently, six regions (1, 2, 4, 5, 6 and 7) and the national are (mostly well) above their historic average %ILI values, and four regions (3, 8, 9 and 10) are tracking their average %ILI values closely.  As in the previous week our selected forecast for the national peak week (second from right panel in top row) comes from coupled spatial model with a force of infection that depends on the school vacation schedule and consequently shows a probability for either an early (EW 52) or later (EW 6-7) season peak. Our forecast for onset (second from left panel in top row) is based mostly on the observed data (EW 47) but includes also a historic NULL model to account for the possibility of a data update that will reduce the national %ILI to below its onset value of 2.2%. Our national forecast for the next four weeks is shown in the top left panel with the corresponding probabilities in the four panels of the bottom row. This forecast is based on a model of with a time-dependent (school vacation and specific humidity) force of infection.Picture1

Region 7 had a large increase in the previous week and has crossed its %ILI onset value (dashed grey horizontal line in top left panel below).  Our prediction for onset (second panel from left in top row) is therefore centered at the current epidemic week, but as in the national case we do allow for the possibility of ‘downward’ data update (and consequently a later onset week) by including the historic NULL model in our submitted forecast.   Similarly to the national case,  our prediction for this region season peak week is split between an early (EW 52) and a late (EWs 5-8) time.  Our forecast for the next four weeks (top left panel and bottom row) is based on school vacation and specific humidity time-dependent force of infection and a data augmentation procedure which uses %ILI data from previous seasons to inform the prior distribution used in the MCMC procedure.

Picture2

Advertisements

ILI forecasts for 2017-2018-Epidemic Week 46

This is the fourth week of the CDC Flu Challenge. Three of the regions: 1, 4 and 6 have crossed the onset value. Whereas five regions (1, 2, 4, 6, and 7) and the national are still well above their historic average %ILI values, four regions (3, 5, 9 and 10) are below their historic average %ILI values for this time of year.  Region 8 is following the average %ILI curve closely. Given this split between above and below average our selected forecast for the national peak week (second from right panel in top row) comes from a coupled spatial model with a force of infection that depends on the school vacation schedule and consequently shows a probability for either an early (EW 52) or later (EW 6-7) season peak. Our forecast for onset (second from left panel in top row) is based on a combination of a historic NULL model and our coupled spatial model (with dependence on specific humidity).  Our forecast for the next four weeks is shown in the top left panel with the corresponding probabilities in the four panels of the bottom row. This forecast is based on a model with a time-dependent (school vacation and specific humidity) force of infection.

Picture1

Region 6 had a large increase in the previous week and has crossed its %ILI onset value (dashed grey horizontal line in top left panel).  Our prediction for onset (second panel from left in top row) is therefore centered at the current epidemic week. As with the national case,  our prediction for the timing of this region’s peak is split between an early (EW 52) and a late (EWs 6-7) week.  Our forecast for the next four weeks (top left panel) is based on a data augmentation procedure that uses %ILI data from previous seasons to inform the prior distribution used in the MCMC procedure.

Picture2

ILI forecasts for 2017-2018 – Epidemic Week 45

This is the third week of the CDC Flu Challenge.  Most regions (1, 2 , 4, 5, 6 and 7) and the national are well above their historic %ILI values for this time of year.  Three regions (8, 9, and 10) are at their historic average value and only one region (Region 3) is below its historic value.  Our forecast for the nation was selected from a large number of predictions based on spatially-coupled mechanistic compartmental (Susceptible-Infectious-Recovered) models that allow the force of infection to depend on the school vacation schedule and/or on the specific humidity in a particular region. Our prediction for the next four weeks is shown in on the top left panel with the corresponding probabilities in the four panels of the bottom row. This forecast is based on a model of a fixed force of infection.  Onset is predicted based on a NULL model which uses historic values whereas peak week and intensity are selected from a force of infection model that depends on school vacation (and specific humidity) with the resulting typical double peak prediction for peak week: early (EW 52) or late (EW 6).

nation-EW45

Region 4 is the largest HHS region and its %ILI has just crossed the CDC onset value (dashed horizontal line in top left panel).  In agreement with this observation, our selected forecast predicts an early onset and an early peak week (EW 52).  The %ILI profiles shown in the top left panel are typical for our coupled S-I-R models with a time-dependent (school vacation and specific humidity) force of infection.

region4-EW45

ILI forecasts for 2017-2018 – Epidemic Week 44

We’re into the second week of the CDC Flu challenge. We’ve developed some new graphics to assess our forecasts. This week we’re showing our “histogram plots”. The two images below (PDF versions can be downloaded in the link below each image) show probabilities for (top) national and (bottom) Region 1 summaries of (1) onset; (2) peak week; (3) peak intensity; (4) 1-week forecast; (5) 2-week forecast; (6) 3-week forecast; and (7) 4-week forecast.

The first panel in each Figure shows the current data (up through Epidemic Week 44) in red, together with our forecast (dotted line) and some realizations to give a measure of the uncertainty in the current forecast. The details of the model variants used to make these forecasts will be explained in more detail in subsequent weeks.

SubHists_EW44-national

SubHists_EW44-national

SubHists_EW44-region-1

SubHists_EW44-region-1

The distributions are fairly wide at the moment, reflecting our uncertainty in many of the parameters. As the weeks go by, these will narrow.

Flu prediction for EW11

Flu prediction isn’t getting easier as the season (fails to) wind down. The incidence profiles (or %ILI technically) – for at least some of the regions – are still showing relatively complex variations.

Here we’ll focus on our two principal models, one incorporating both specific humidity and schools and the second, based on our recent PLoS – Computational Biology paper, which allows R(t) to have two values, with a change in R(t) fit by the model. (Additionally, we ran a relatively large ensemble of other model types; such as ones incorporating priors, but we’ll discuss those results another time).

Generally, both models do reasonably well in fitting to the data up to the present point in time. Typically, when there are strong changes in specific humidity or school holidays that occur at a crucial point in the profile, the SH+school model performs well, and better than the paper 2 model. For these reasons, if we believe that the historical variations in specific humidity are likely to hold out for the remainder of this season, these regions can also be predicted well forward in time. In most cases, these profiles suggest that either the season has peaked or will peak during the next week or two.

The paper 2 model on the other hand, is not constrained in the same way. The model is allowed to place variations in R(t) at points where large changes in incidence occur. But, this can mean that future extrapolation (“prediction”) may not be well constrained. If we consider region 2, for example, the paper 2 model results show that incidence will continue to climb through week 26, which is very unlikely (although if it does, we will certainly learn something from that). On the other hand, in some cases, such as region 7, the paper 2 model does a markedly better job at capturing the variability in %ILI up through the present time. Thus, if we use region 7’s paper 2 results, we will forecast a continued increase for the next 4 weeks (the four-week CDC prediction interval is marked by the grey shaded box in each panel). Both because the AICc scores are significantly lower for the paper 2 model, and because it seems to have captured the strong increase over the last six weeks, this is the model that will be used to make our four-week prediction for region 7. We are aware, however, that the most recent data point, while falling close to the paper 2 curve, may be significantly revised next week; but in which direction, we do not know.

We are keeping records of all predictions as well as the data provided by the CDC, which is adjusted from week to week, so that we can perform a retrospective analyses of our effort after the season is over. By mining the model predictions this way, we should be able to getter gauge our confidence in each model’s predictions, and how that confidence evolves through the course of the season.

Flu prediction for EW10

We’ve been quietly making predictions each week and submitting to the CDC’s Flu challenge. Here’s a brief update of some of the lessons learned.

First, the data are continually updated (at least the previous week, and occasionally  further back). These can be modest changes or significant, but they can’t be anticipated (both the direction of the change and magnitude vary even in the same region). This should be factored into the predictions, either in the form of weighting the most recent data less, or performing sensitivity studies with some representative confidence intervals.

Second, the profiles this year are generally considerably more complicated than previous recent years. This may be due, in part, to El Niño. If so, our reliance on historical values for specific humidity cannot accurately account for humidity variations.

Third, incorporation of priors into the forecast appears to produce more accurate forecasts, although not in all cases. This changes as we move through the season, so that initially, the prior forecasts are consistently better, but, by the end of the season, they will have little value, and solutions not using priors will perform better. (This week, three of the regions performed better without priors).

Fourth, we also incorporated a timing delay to account for the fact that there is a delay between the time that an individual becomes infectious and presents themselves to a clinic. The shape of the prediction was not measurably altered, but the likelihood was significantly improved, due to the shifting of the profile. A more robust approach would be to include an “Exposed” category to the compartmental model, which is currently being done.

Fifth, we found that school holidays is often an important driver of transmission, although because of the limited number of parameters allowed by the model, it can sometimes result in strong future predictions that may not occur. In particular, to account for the decrease over the Christmas period, the model introduces a significant decrease in R0. This same magnitude is then applied in late March during the spring break resulting in a dramatic decrease in %ILI (see Regions 4 and 7 in the accompanying Figure). Although this may yet occur, it is quite dramatic, and is more likely the result of the model over-fitting smaller fluctuations during the early rise portion of the season.

For almost all regions, we predict that the peak has already occurred. For any remaining ones, it will likely occur this week or next.

First prediction of CDC region-level influenza

We recently joined the CDC Flu Challenge competition and just submitted our predictions for Week 4 on Monday evening. These charts summarize our predictions for each of the ten US regions.  The data are shown by the solid black lines and the model predictions by the thick (smooth color) lines. Future predictions are shown as a dashed curve and the grey vertical bar marks our predictions for the next four weeks.  To make our prediction, we used information about school vacations, specific humidity, previous similar influenza seasons and the pattern if incidence so far. We anticipate that regions 1-6, and 9 have already peaked, while regions 7, 8, and 10 have yet to peak. We are eagerly awaiting Friday’s release of new data by the CDC.  Screen Shot 2016-02-11 at 11.44.26 AM.