Age Analysis

Age Analysis Update November 7, 2020

This post is an update to the series of posts made about one month ago analyzing the distribution of cases, hospitalizations, and deaths by age cohort. As I noted then, the state has made it quite difficult to perform this analysis for two primary reasons:

(1) Age-related information is only published weekly, but on a rolling two-week basis (the state stopped providing daily information on August 11). Because the age-related information covers two weeks, weekly figures must be estimated.

(2) The published totals  in the weekly reports do not align with those for the corresponding period in the daily reports.

To circumvent these issues, I developed a methodology outlined in a prior post ( to estimate the age distributions through time.  I’m using the same methodology to extend the analysis through November 7th for cases and deaths. 

I had been relying on hospitalization data published as part of the race / ethnicity daily reports to calibrate the weekly age-based hospitalization data – which tended to severely under count hospitalizations compared to the daily reports.  However, the race / ethnicity report was dropped on November 2nd.  I am working on an approach to estimate age-related hospitalizations, but won’t include it in this post.

As most everyone is aware, cases have increased dramatically since the beginning of October – almost tripling in the five weeks between October 3 and November 7.  Obviously, case incidence rates are significantly higher than previously, as Figure 1 indicates. 

As has been true since the end of July, case incidence rates have been higher for those under 60 compared to those over 60 – ranging from about 50% higher to more than double each week since the beginning of August. This case incidence rate differential appears to be generally holding, even as cases skyrocket.

This is evident from Figure 2, which shows the percentage of cases by age cohort over time. These percentages have held roughly steady since the end of July – with about 85% of cases in people under 60, and 15% of cases in people 60 and over. (This may seem very high for those under 60, but keep in mind that they represent 77% of the population in the state). In the past several weeks, there might be a slight trend to higher relative rates among those 60 and older, but it is difficult to determine whether this is just a normal fluctuation in the data or the start of a trend.

Unlike cases, which have increased significantly since October 3, deaths have increased modestly – up 37% to November 7th. This is positive news, but since deaths lag case diagnosis by at least several weeks, there is no guarantee this trend will continue. We can certainly hope that improved treatment protocols and the younger age of patients will keep deaths somewhat in check.

Figure 3 shows the death incidence rate by age cohort through November 7. Even though the death totals haven’t climbed as much as cases since October 3, the bulk of the increase is born by those 60 and older, as Figure 3 makes apparent. This is because over 90% of deaths are in the 60 and older population, and this percentage has been roughly the same for months.

Figure 4, which shows the percentage of deaths by age cohort, makes this clear. Although it is a bit difficult to see in the Figure, over the past five weeks the percentage of deaths in those 60 and older rose slightly – to 93% from about 91% in the five weeks prior.

4 replies on “Age Analysis Update November 7, 2020”

Why not use the raw data the state releases on a daily basis to track positives by age group? It shows cases by date reported, not assumed date of exposure, but still worthy or not?

What do you make of the clusters data the state includes now in its weekly report? It has data on about a third of all infections.

I’m not sure which data you are referring to. There is the Total Cases by Age Group on the first page of the dashboard, which is only updated once per week, just like the age data in the weekly report. (Although they supposedly cover the same time frame, the data in the daily and weekly reports don’t match up exactly. I’ve been sticking to the weekly report for consistency. It could be that once is by date reported, and the other by date occurred, but the state doesn’t make that clear). The other possibility is the Positive Tests by Age Group, a new chart on page 7. Unfortunately, positive tests don’t translate directly into cases so I’m not sure how I’d use it since the state doesn’t indicate how many tests are taken by age group.

In terms of the cluster data, it isn’t that useful if only one-third of all cases are in the data. I hadn’t really looked at it, but was aware to some degree of the low percentage of cases in the cluster data.

Thanks for pointing that out. I had looked at the Chapter 93 report when it first came out, but had decided there wasn’t much in it. But the age breakdown of tests and positive tests might be something to look into. It isn’t exactly what I’m looking for, since there is not a 1 to 1 relationship between cases and positive tests, but if I have the stamina to process these data (the fact that the state only publishes three days at a time means there are a lot of historical files to process), I might use it to try to look at test positivity rates and positive tests over time by age group (as opposed to cases by age group, which is what I’m looking at now.

Unfortunately, the death data by age group in the report is useless, since the state suppresses numbers under 5, which is almost always the case each day for younger age groups.

Leave a Reply