Measles: GBD 2020

Todo

Need to update all below to GBD 2020

Disease Description

Measles is a highly contagious, serious disease caused by the measles virus (Measles morbillivirus). Symptoms usually develop 10-12 days after exposure to the virus, and last 7-10 days. Symptoms include fever, cough, runny nose, conjunctivitis, characteristic white spots inside the cheek (called Koplik’s spots), and a red, flat, blotchy skin rash that develops on average 14 days after exposure to the virus (range, 7-21 days) and lasts 5-6 days. Recovery from measles confers lifelong immunity. [WHO-Measles-2020], [CDC-Measles-2020], [Wikipedia-Measles-2020], [GBD-2017-YLD-Capstone-Appendix-1-Measles-2020]

Most measles-related deaths are caused by complications associated with the disease. The most serious complications include blindness, encephalitis, severe diarrhea, ear infections, and pneumonia. Serious complications are more common in children under the age of 5 or adults over the age of 30, especially those with vitamin A deficiency or those whose immune systems have been weakened by HIV/AIDS or other diseases [WHO-Measles-2020].

Measles is spread by coughing and sneezing, close personal contact, or direct contact with infected nasal or throat secretions. The virus remains active and contagious in the air or on infected surfaces for up to 2 hours. It can be transmitted by an infected person from 4 days prior to the onset of the rash to 4 days after the rash erupts [WHO-Measles-2020], [CDC-Measles-2020], [Wikipedia-Measles-2020].

Before the introduction of a measles vaccine in 1963 and widespread vaccination, major epidemics occurred approximately every 2–3 years, and measles caused an estimated 2.6 million deaths each year. Despite the availability of a vaccine, approximately 110,000 people died from measles in 2017, mostly children under the age of 5 years. However, due to accelerated immunization activities, global measles deaths have decreased 80% during the period 2000–2017, from an estimated 545,000 to 110,000, and measles vaccination prevented an estimated 21.1 million deaths during 2000–2017 [WHO-Measles-2020].

The ICD 10 codes for measles are B05-B05.9, Z24.4, and ICD 9 codes are 055-055.9, 484.0, V04.2, V73.2 [GBD-2017-YLD-Capstone-Appendix-1-Measles-2020].

Todo

Add data about global vaccine coverage and efficacy. Perhaps start with these references:

Also perhaps note this recent New York Times article (Oct 31, 2019):

Measles Makes Your Immune System’s Memory Forget Defenses Against Other Illnesses: New research shows the virus can have devastating effects on the immune system that persist much longer than the illness itself.

Second reference for this: https://www-ncbi-nlm-nih-gov.offcampus.lib.washington.edu/pmc/articles/PMC8590458/

Modeling Measles in GBD 2020

The non-fatal measles model “primarily leverages the relationship between direct reports of measles case notifications annually released by the World Health Organization (WHO) in the Joint Reporting Form (JRF), modelled estimates of measles-containing-vaccine (MCV) vaccination coverage proportions for doses 1 and 2, and supplementary immunisation campaign (SIA) coverage to produce global estimates of measles cases” [GBD-2019-Capstone-Appendix-Measles_2020] (p. 721). GBD assumes that 50% of measles cases are moderate (disability weight 0.051, 95% CI: 0.032, 0.074) and that 50% are severe (disability weight 0.133, 95% CI: 0.088, 0.19); symptoms of the measles sequelae in GBD include fever, aches/pain, and weakness. GBD assumes a universal 95% attack rate in the absence of vaccination. Measles prevalence was estimated by multiplying incident case estimations by an average case duration of ten days [GBD-2019-Capstone-Appendix-Measles_2020].

The GBD 2020 fatal measles estimates were modeled in one of two ways depending on available data quality for a given country. For data-rich countries, the Cause of Death Ensemble model (CODEm) was used, with measles-containing vaccination dose one, healthcare access and quality (HAQ) index, socio-demographic index (SDI), and mean years of education per capita as covariates (NOTE: the maternal care and immunization covariate was used in place of HAQ and SDI covariates in GBD 2017 and prior). For data-poor countries, a natural history model was used that built off of the GBD non-fatal estimates. The natural history model estimated a case fatality ratio via a model informed by data from a systematic literature search that was updated in 2019. “With the available measles CFR input data, we make location- and year-specific death estimates using a negative binomial model with Socio-demographic Index (SDI) as a country-level covariate, additionally accounting for three indicators (hospital-based or not; outbreak or not; and rural or urban/mixed) as study-level covariates, with country random effects” [GBD-2019-Capstone-Appendix-Measles_2020] (page 177).

Todo

Describe enough of the data sources and modeling process to verify that even though measles can lead to diarrhea or other causes that we include in our Vivarium models, we won’t be double counting mortality and morbidity from these causes. For example, a death caused by diarrheal dehydration due to measles should be counted in the GBD as a death due to measles, not as a death due to diarrheal diseases.

The relationship with vitamin A deficiency may also be important for our models.

Make sure to check on measles sequelae as well. Our models so far have not paid much attention to the nonfatal side, but it looks like some of the complications can persist well after someone recovers from measles, so maybe that’s important to think about.

GBD Hierarchy

Hierarchy Diagram:

Measles GBD hierarchy diagram

Cause Model Diagram

Simple SIR Measles cause model diagram

Model Assumptions and Limitations

This model is designed to be used for estimating DALYs due to measles that are averted from a country-level intervention (e.g. food fortification or supplementation given to a percentage of the population) that can reduce measles incidence as a downstream effect.

In particular, there are various uses for which this model is not suitable. For example:

1. The simple measles model described here does not explicitly incorporate vaccine coverage or efficacy, hence cannot be used to model the impact of a vaccination campaign.

2. This model uses country-level data, and cannot be used to model local measles outbreaks due to lack of vaccination in small communities.

Some of the assumptions made in this model are:

  1. There is no data available for population in recovered state in GBD. Since the early neonatal, late neonatal, and 1-5 month age groups are not modeled in GBD, we made the assumption that there are no individuals in the recovered state at the start of the 6-11 month age group (NOTE: prior GBD estimates started modeling at 28 days. This was changed in 2020 because children under 6 months often have immunity from maternal antibodies and so are assumed to not get measles.) We then calculated the proportion of the population in the recovered state in the 12-23 month and 2-4 year age group using GBD measles incidence and mortality rates. Note that we performed this calculation only for these 2 age groups because measles is most often of interest among the children under five population, but this assumption could be expanded to other older ages as well.

Todo

Edit above for age group changes once modeling approach has been decided

2. There is no data avaialable for remission rate in GBD. So a constant remission rate is calculated from average case duration assumption of 10 days [GBD-2019-Capstone-Appendix-Measles_2020].

Restrictions

Restriction Type

Value

Notes

Male only

False

Female only

False

YLL only

False

YLD only

False

YLL age group start

6-11 months

GBD age group id is 389

YLL age group end

50 to 54 years

GBD age group id is 15

YLD age group start

6-11 months

GBD age group id is 389

YLD age group end

50 to 54 years

GBD age group id is 15

Notably, our cause model as described accounts for the number of simulants that enter the recovered state by becoming infected with measles and recovering, but does not consider simulants who enter the recovered state by receiving the measles vaccine without becoming infected with measles. Therefore, we are underestimating the prevalence of the recovered state in our model. This strategy was employed so that the size of the susceptible and recovered populations at model initialization will remain stable as the simulation runs and simulants progress through the SIR cause model (this is an improvement from our previous assumption that there was zero prevalence of the recovered state upon model initialization, which caused an decrease in the susceptible population and an increase in the recovered population over time, which affected calibration of simulation incidence and mortality rates to GBD rates); see note below.

Note

A note on why the assumption of \(prevalence_R = 0\) at initialization used in the 2017 measles cause model document needed improvement:

\(prevalence_S = 1 - prevalence_\text{c341} - prevalence_R\)

\(incidence_I = incidence_\text{c341} / prevalence_S\)

Since vivarium calculates \(incidence_I\) once at the beginning of the simulation, if we assume that no simulants are initialized into the recovered state (\(prevalence_R = 0\)), then \(incidence_I\) will be scaled to the prevalence susceptible population that is artificially inflated by this assumption (as we can safely assume that the true prevalence of the recovered state is greater than zero). Then as simulants move from the susceptible state through the infected state and into the recovered state as the simulation progresses, the prevalence of the susceptible state will decrease as simulants accumulate in the recovered state. Because the prevalence of the susceptible population decreased over time in our simulation in this manner but the incidence rate did not increase, the overall incidence of measles in our simulation decreased over time. Therefore, we updated our assumption so that some number of simulants will be initialized into the recovered state in an attempt to avoid this issue.

Alternative model structures to consider include:

  • An SIS model. While this model does not accurately reflect measles disease dynamics, it allows for simple modeling of the expected annual rates of measles morbidity and mortality without consideration of disease-specific characteristics.

  • An SIR model that considers a non-susceptible state due to vaccine coverage. While more complicated, this model could be beneficial in modeling correlation with other risk factors such as vitamin A deficiency, in modeling differential impact by certain population subgroups, or in modeling vaccine coverage as an intervention. Notably, GBD does estimate measles vaccine coverage.

Todo

Describe more limitations and assumptions of the model as appropriate. For example,

  • There are 2 ways people can be in the “recovered” state - either they get measles and then recover, or they get vaccinated and move directly into the “recovered” state without ever having the disease. We should look into measles vaccination rates in the countries we’re interested in (Nigeria, India, Ethiopia) and compare this to the number of people who actually get measles. If the number of vaccinated people is much higher than the number who get the disease, then our assumption will have a smaller effect, because the few people who enter the recovered state in our model will be be a small proportion of the total number of people in the recovered state, and the GBD incidence rate is already accounting for people who are “recovered” by vaccination.

  • We should also look at the case fatality rate / excess mortality rate for measles, as this will also have an impact on the effect of this assumption, as well as on our assumption of a constant remission rate.

  • For our assumption of a constant remission rate (below), we should think about what the actual hazard function for remission should look like (we should be able to get some idea about this from the disease description), and estimate how replacing it with a constant rate will affect our results.

  • Also include about GBD’s assumption of 50% of measles cases as moderate and other 50% as severe.

Data Description

Definitions

State

State Name

Definition

S

Susceptible

Susceptible to measles

I

Infected

Infected with measles

R

Recovered

Recovered from measles
States Data

State

Measure

Value

Notes

S

prevalence

1 - prevalence_c341 - prevalence_R

S

excess mortality rate

0

S

disabilty weights

0

I

prevalence

prevalence_c341

I

excess mortality rate

\(\frac{\text{deaths\_c341}}{\text{population} \times \text{prevalence\_c341}}\)

I

disability weights

disability_weight_s117 \(\times\) prevalence_s117+ disability_weight_s118 \(\times\) prevalence_s118

GBD assumes 50% of measles cases as severe and other 50% as moderate [GBD-2017-YLD-Capstone-Appendix-1-Measles-2020].

R

prevalence

See below

R

excess mortality rate

0

R

disabilty weights

0

ALL

cause specific mortality rate

\(\frac{\text{deaths\_c341}}{\text{population}}\)

Prevalence of recovered state

Todo

Edit below equations and text for age group changes once modeling approach has been decided

We will use a age-group-specific prevalence of the recovered state for this cause model. We assume that the prevalence of the recovered state among the post-neonatal age group (age group ID 4), the first modeled age group for measles, is zero. For all other age groups modeled for measles, the prevalence of the recovered state relies on information from the preceeding age group, as detailed below.

For the post-neonatal age group:

\[0\]

For all other modeled age groups:

\[\frac{2 \cdot (prevalence_\text{R*} + incidence_\text{c341*} - CSMR_\text{c341*}) + incidence_\text{c341} - CSMR_\text{c341}}{2}\]

Where,

Transition Data

Parameter

Value

\(prevalence_\text{R*}\)

Prevalence of the recovered state in the preceeding age group

\(incidence_\text{c341*}\)

Measles incidence rate among the preceeding age group in the total population from GBD

\(CSMR_\text{c341*}\)

Meaasles cause-specific mortality rate among the preceeding age group

\(incidence_\text{c341}\)

Measles incidence rate among the age group of interest

\(CSMR_\text{c341}\)

Meaasles cause-specific mortality rate among the age group of interest

This approach makes the following assumptions:

  • There is no difference in all cause mortality rates between the population susceptible to measles and the population recovered from measles

  • The prevalence of the recovered state for a given age group is equal to the average between the upper and lower bound of that age group

  • There are no recovered cases of measles prior to an age of 6 months

Transition Data

Transition

Source

Sink

Value

Notes

i

S

I

\(\frac{\text{incidence\_rate\_c341}}{prevalence_S}\)

r

I

R

remission_rate_c341 \(= \frac{\text{365 person-days}}{\text{10 person-days} \times \text{1 year}}\) \(= \frac{\text{36.5}}{\text{year}}\)

GBD assumes average case duration as 10 days [GBD-2017-YLD-Capstone-Appendix-1-Measles-2020]. So constant remission rate is approximated to this calculation.
Data Sources

Measure

Sources

Decomp step

Description

Notes

prevalence_c341

como

step5

Prevalence of cause measles

deaths_c341

codcorrect

step5

Deaths from measles

population

demography

step4

Mid-year population for given country

incidence_rate_c341

como

step5

Incidence rate for measles

remission_rate_c341

YLD appendix

n/a

Remission rate for measles

GBD assumes average case duration as 10 days [GBD-2017-YLD-Capstone-Appendix-1-Measles-2020]. So constant remission rate is calculated from this assumption.

disability_weight_s{sid}

YLD appendix

n/a

Disability weights associated with each sequelae

prevalence_s{sid}

como

step5

Prevalence of each sequelae

Validation Criteria

Simulation results should replicate the GBD 2020 cause-specific mortality rate, incidence rate, and prevalence for all age/sex/location groups. Notably, these measures should be tracked over time in the simulation to ensure that simulation rates do not deviate from GBD rates as the simulation progresses.

References

[WHO-Measles-2020] (1,2,3,4)

Measles Fact Sheet. World Health Organization, 9 May 2019. Retrieved 13 Nov 2019. https://www.who.int/news-room/fact-sheets/detail/measles

[CDC-Measles-2020] (1,2)

Chapter 13: Measles. Epidemiology and Prevention of Vaccine-Preventable Diseases (The Pink Book, 13th Edition). Centers for Disease Control and Prevention, 2015. Retrieved 13 Nov 2019. https://www.cdc.gov/vaccines/pubs/pinkbook/meas.html

[Wikipedia-Measles-2020] (1,2)

Measles. From Wikipedia, the Free Encyclopedia. Retrieved 13 Nov 2019. https://en.wikipedia.org/wiki/Measles

Todo

update this cite to 2020