NewsPhoenix Metro NewsCentral Phoenix News


Hidden data shows more details on Arizona coronavirus outbreak

Posted at 5:46 PM, Apr 08, 2020
and last updated 2020-04-09 09:21:50-04

Every day at 9 a.m. the Arizona Department of Health Services refreshes a dashboard full of statistics showing how COVID-19 is impacting the state. The state provides both COVID-19 case and testing numbers. They can be broken down by county, age group, gender and date of test collection. While the information provided on the dashboard is extensive, it lacks some key components that could provide additional context to Arizona’s COVID-19 outbreak, context that other states are already giving, such as more precise geographic areas, hospitalizations, ICU admissions, and people that have recovered from COVID-19.

Although the state has committed to transparency in data, there are some data points that the Arizona Department of Health Services is withholding from the public.

ABC 15 obtained the hard to find dataset that powers the state’s COVID-19 dashboard. After analyzing the dataset, it’s clear that while there was no privately identifiable information, it included entries on case records that indicated the outcome of the case, as well as if the case was admitted to the hospital. The dataset includes 2,456 Arizona COVID-19 cases dated from January 22, 2020 to April 5, 2020. At the time of this writing, there are 2,726 COVID-19 cases being reported on the AZDHS website.

The column that presumably records the outcome of the case had four categories; survived, died, unknown, or the entry was left blank. In all, 1,384 entries were left blank by the Department of Health Services. Nine hundred and ninety two entries were marked “survived”, 40.4% of the cases. 65 records were marked “died”, a number that matches the reporting from the state at the time the dataset was acquired.

We reached out to the AZDHS for an explanation of the state’s definition of survived, as well as entries that were left blank. A spokesperson for the agency told us that survived means that there was not a death. They also told us that local health departments were prioritizing filling out the fields of those that have died so blank were cases that are under investigation.

Case Outcome Number Percent
Blank 1384 56.4%
Survived 992 40.4%
Died 65 2.6%
Unknown 15 0.6%

The dataset also includes the date when the entry was first entered into the system. As many cases with a date earlier than March 30th have an outcome field, we aggregated the entries by date and plotted them along a chart. It shows that the gap has widened between the total number of cases and those marked survived as the graph gets closer to the present, a result that would be expected when recording possible recoveries.

There were three hundred and eleven entries that indicated that the case was admitted to the hospital. The vast majority, 82% of these were in Maricopa county, which is separately reporting hospital data outside of ADHS data. The next most hospitalizations in the dataset belonged to Pinal county, which only recently began adding hospitalizations to its separate reporting site. One thousand and three entries in the data were blank. An AZHDS spokesperson told us that “Unknown and blank for this field are for cases that are still under investigation or have not been filled out yet.”

In most of the counties outside of Maricopa, the hospitalization data may be skewed towards the most serious COVID-19 cases. Mohave county has six total hospitalizations, five of which were recorded to be in ICU. In Coconino county, which includes parts of the Navajo Nation, an area hit particularly hard by COVID-19 cases compared to its population, 5 of the 10 cases marked hospitalized were in ICU.

Arizona’s seniors make up the majority of hospitalized cases. 143 of the 311 cases, 46%, are 65 or older. Case numbers for ICU patients are similarly reflective of this trend at 50% of all cases.

Hospitalization Number Percent
Yes 311 12.7%
No 1132 46.1%
Unknown 10 0.4%
Blank 1003 40.8%

ICU Number Percent
Yes 108 34.73%
No 162 52.09%
Unknown 19 6.11%

County Hospitalization ICU Percent ICU
Cochise 3 2 67%
Coconino 10 5 50%
La Paz 1 0 0%
Maricopa 257 87 34%
Mohave 6 5 83%
Navajo 3 0 0%
Pima 7 1 14%
Pinal 16 5 31%
Yavapai 3 1 33%
Yuma 5 2 40%

Age Group Hospitalized Cases Percent
Less than 20 years 1 0%
20 - 44 years 58 19%
45 - 54 years 51 16%
55 - 64 years 58 19%
65 years and older 143 46%

Age Group in ICU Cases Percent
20 - 44 years 14 13.0%
45 - 54 years 12 11.1%
55 - 64 years 28 25.9%
65 years and older 54 50.0%

What we don’t know about the ADHS dataset is if and how it has changed over time. What we do know is that it will go through another change in the near future. Arizona’s Governor Doug Ducey signed an Executive Order that requires additional reporting requirements from hospitals. This includes:

  • Number of inpatient COVID-19 positive or patients with suspected COVID-19.
  • Number of ventilators in use by COVID-19 positive or patients with suspected COVID-19.
  • Number of ICU beds in use by COVID-19 positive or patients with suspected COVID-19.
  • Number of COVID-19 positive or patients with suspected COVID-19 discharged from the facility per day.
  • Number of COVID-19 positive or patients with suspected COVID-19 seen in the Emergency Department per day.
  • Number of intubations performed per day for respiratory distress.
  • Estimated number of N95 masks used per day.
  • Estimated number of surgical masks used per day.
  • Estimated number of face shields used per day.
  • Estimated number of surgical gowns used per day.

While the order does considerably expand what hospitals are reporting to ADHS, it is unclear if and how much additional information will be available on the public dashboard. The Governor and AZDHS has committed to publicly releasing “enhanced location data”, although they have not decided whether the release will include readily available geographic location data like zip codes, or will use medical referral districts.

What we do know, is that Arizona’s government has promised transparency in the COVID-19 reporting process. But it appears that, for whatever reason, some collected data that is not personally identifiable is being held back.