Memorial University of Newfoundland and Labrador, Canadian Network for Modelling Infectious Disease, Mathematics for Public Health, One Health Modelling Network for Emerging Infections


This report synthesizes information on COVID-19 cases, testing, hospitalizations and deaths, and models processes such as transmission, recovery and hospitalization to describe and forecast COVID-19 dynamics in Newfoundland and Labrador.

The data

These data are recorded from the NL data hub each time it is updated. On Fri May 13 hospital occupancy was 8 and the number of deaths over the last 7 days was 8.

Trends in reported cases may not reflect trends in the true number of cases when testing rates have changed.

To further understand the dynamics of COVID-19 cases in Newfoundland and Labrador, I consider two modelling approaches.


Estimating cases from hospital occupancy and deaths

I fit an epidemic model to reported hospital occupancy and death data. I do not fit to reported cases since trends in these data are sensitive to the number of tests performed which has changed over time. The model links together cases, hospitalizations and deaths. Most of the model structure is determined by epidemiological constants that are unlikely to differ between regions, but some parameters are specific to Newfoundland and Labrador (see here for full details)). The dots are data from the NL data hub, and the blue lines are the epidemiological model.

To further describe the epidemic model, here we explain why the BA.1 wave has two peaks in Newfoundland and Labrador.

This modelling approach has no reliance on provincial PCR testing data, and similar modelling approaches are used in other provinces and by the Public Health Agency of Canada.


Testing in Newfoundland and Labrador

After March 17, the reported number of PCR tests performed per day in provincial labs has decreased over time. This coincides with announced changes to the eligiblity for PCR testing.


Total active cases

Using an approach from the peer-reviewed literature, I estimate total active cases: both reported and unreported. Active cases are the total number of new cases reported over the last 7 days.

The number of tests performed daily has generally declined since March 17, however, this analysis suggests that even when accounting for the decreased rate of testing, the total number of active cases, including those unreported, has declined.

This second modelling approach is based on less reliable data, however the general trend in cases is similar to the first approach that was based on hospitalization and death data only.

Best estimates of true infection levels are determined from serology data (see also: COVID-19 Immunity Task Force).


The geometric mean method implies that the number of unreported cases per reported case, recently, is near 4. During media briefings, Dr. Fitzgerald had stated that screening of hospital admissions in January suggested total cases were 2-3 times greater than reported cases. On May 6, Dr. Fitzgerald stated that total cases are estimated as 4-6 times greater than those reported. In Ontario, it is estimated that 1 in 10 cases are reported (slide 3).


The local context

On March 14, 2022, Newfoundland and Labrador repealed the public health state of emergency order, lifting most public health measures, and introducing a new data reporting website. On January 7, 2022, and March 17, 2022, PCR eligibility changes were announced. There was a significant shift around late January when rapid antigen tests and testing programs first became widely available to the public. Positive rapid antigen test results can be reported via the provincial COVID assessment and reporting tool.

Modelling peaks in COVID-19 cases is subject to high uncertainty. Government officials stated that we could be near the Omicron variant peak on (April 5 and 13), noting that the peak in hospitalizations would lag the peak in cases. The peak number of people hospitalized with COVID-19 was 144 on March 21.

My modelling is consistent with the statements by Newfoundland and Labrador government officials. I find that hospitalizations for COVID-19 peaked around April 3. I estimate that COVID-19 cases peaked around March 17-21.


What data are used for this analysis?

Data after March 11 are manually tracked from the data hub. Data prior to March 11 were downloaded from the NL government website. The last 10 entries of data are:

##          date cases hosp.occ new.deaths
## 1  2022-05-04   140       14          3
## 2  2022-05-05   104       NA          0
## 3  2022-05-06   108       17          0
## 4  2022-05-07    47       NA          0
## 5  2022-05-08    46       NA          0
## 6  2022-05-09    58       14          1
## 7  2022-05-10    75       NA          0
## 8  2022-05-11    70        6          7
## 9  2022-05-12    53       NA          0
## 10 2022-05-13    55        8          0

The last 10 data entries from PHAC (excludes when 0 tests performed is reported) are:

##          date cases tests
## 1  2022-05-02    65   938
## 2  2022-05-03    44   265
## 3  2022-05-04   140   375
## 4  2022-05-05   104   477
## 5  2022-05-06   108   550
## 6  2022-05-07    47  1228
## 7  2022-05-09    58  1114
## 8  2022-05-10    75   362
## 9  2022-05-11    70  1114
## 10 2022-05-12     0   348


Useful resources

[1] Download the data used for this site here.

[2] Factors affecting hospitalization risk in NL.

[3] Estimated fraction of Omicron cases in NL that result in death.

[4] COVID-19 trends in NL might also be determined from wastewater surviellence, although generally the data reported here are from two weeks ago or more, and trends are difficult to assess from these data.

[5] Of COVID-19 deaths reported in Newfoundland and Labrador, 62% have occurred in hospital.

[6] On April 1, it was reported that 35-40\(\%\) of screened cases were BA.2. This estimate may be delayed relative to incidence owing to reporting lags.

Geometric mean method

I estimate the prevalence of unreported cases 7 days ago as approximately the geometric mean (\(n = 0.54\)) of the per capita reported new cases (14-day rolling average), and the proportion of tests that are positive (14-day rolling average; see Chui and Ndeffo-Mbah (2021) for details). The value of \(n\) is 0.54 (\(n=0.5\) is the geometric mean) and the lower and upper estimates are 0.46 and 0.67 (see Table 2 of Chui and Ndeffo-Mbah (2021))

The method of Chui and Ndeffo-Mbah (2021) is peer-reviewed and published. The estimates for unreported cases are validated by comparing to serological data for 50 states and Washington, DC, in the USA, and 15 countries (including Canada). Therefore, the equation used has been tested for its accuracy with an independent data source (blood tests to identify undetected cases).

A weakness of the approach is that the study was completed on December 31, 2020. Omicron overwhelmed testing in most jurisdictions and test eligbility has changed substantially from 2020. The formula to estimate unreported cases has not been validated across this type of testing eligibility.