Introduction

Numerous mathematical models have been developed to project the ongoing COVID-19 pandemic in the United States (e.g. \cite{disagree}). The aim of this paper is to evaluate the accuracy of the predicted numbers of cumulative deaths made by the COVID-19 health service utilization forecasting team at the University of Washington's Institute for Health Metrics and Evaluation (IHME) in a non-peer-reviewed paper published on March 30, 2020 on the preprint server medRxiv. These predictions included the predicted number of deaths over the following four months for the United States as a whole, as well as for each of the 50 states and the District of Columbia.\cite{2020} My ability to assess these predictions' accuracy is limited by the fact that, whereas the predictions were intended to reflect all deaths over the next four months (which I take to mean by July 30, 2020), only slightly more than one month has passed since their paper was published at the time of writing. Thus, even if the IHME's predictions were all perfectly accurate, you would expect the number of deaths in the United States and each region therein to be, at the time of writing, significantly less than the corresponding predicted number of deaths. However, if the number of deaths has already, by May 5, matched or even exceeded the number predicted to occur by July 30, this would provide clear evidence that the predicted numbers of death were significantly too low.

Methods

I obtained the number of deaths predicted by the IHME over the four months (i.e. from March 30-July 30, 2020) for each state, as well as the corresponding predicted lower and upper bounds, from the aforementioned medRxiv paper(\cite{2020}, Table 1). I obtained the actual number of deaths in the United States for each state from the COVID Tracking Project's spreadsheet on May 5, 2020 at 4:18 PM.\cite{state} The time at which each state's data was last updated in that spreadsheet varied from state to state and is shown in table 1 below. As this table shows, each state's data was last updated between May 2 and May 4, 2020, inclusive. Nevertheless, for consistency, I will be referring to the number of actual deaths based on these data as being "as of May 5, 2020".
StateLast update (ET)Check time (ET)
AK05/04 0:0005/04 15:27
AL05/04 0:0005/04 16:24
AR05/03 15:4005/04 16:07
AS05/03 0:0005/04 15:35
AZ05/04 0:0005/04 15:35
CA05/04 14:0005/04 16:38
CO05/03 18:0005/04 15:17
CT05/03 16:0005/04 16:35
DC05/03 0:0005/04 16:14
DE05/03 18:0005/04 16:28
FL05/04 9:1205/04 15:25
GA05/04 15:2505/04 16:37
GU05/04 7:4005/04 15:36
HI05/03 18:0005/04 15:48
IA05/03 12:0005/04 14:51
ID05/02 19:0005/04 15:27
IL05/04 0:0005/04 16:08
IN05/03 23:5905/04 15:20
KS05/03 10:0005/04 14:55
KY05/03 17:0005/04 15:56
LA05/04 13:0005/04 16:28
MA05/04 12:0005/04 16:30
MD05/04 10:0005/04 15:12
ME05/03 11:5905/04 15:09
MI05/03 11:0005/04 16:35
MN05/04 0:0005/04 14:58
MO05/04 15:0005/04 15:39
MP05/04 0:0005/04 15:14
MS05/03 19:0005/04 15:58
MT05/04 0:0005/04 14:39
NC05/04 10:5005/04 16:31
ND05/04 13:0005/04 14:36
NE05/03 20:4505/04 14:35
NH05/03 9:0005/04 16:38
NJ05/04 13:0005/04 14:45
NM05/03 17:5805/04 16:13
NV05/04 11:0005/04 16:33
NY05/04 0:0005/04 16:32
OH05/04 14:0005/04 14:19
OK05/03 0:0005/04 14:11
OR05/04 11:0005/04 16:35
PA05/04 0:0005/04 15:21
PR05/04 6:0005/04 15:32
RI05/04 0:0005/04 15:24
SC05/03 15:3005/04 14:43
SD05/03 18:0005/04 15:35
TN05/04 15:0005/04 16:01
TX05/04 13:1505/04 15:34
UT05/03 15:0005/04 15:38
VA05/03 17:0005/04 14:55
VI05/03 20:3005/04 14:57
VT05/04 10:2005/04 16:40
WA05/03 2:5905/04 16:19
WI05/04 0:0005/04 15:56
WV05/04 10:0005/04 16:17
WY05/04 1:5605/04 16:02