Comparing methods for early detection systems for seasonal and pandemic influenza

Research output: Contribution to conferenceAbstract

Abstract

The aim of this work is to investigate established methods for early detection of seasonal and pandemic influenza, then to develop a new method for routine use. We investigate daily data for influenza-like illness (ILI) General Practice (GP) consultations, collected from all 14 spatially located Health Boards (HBs) in Scotland. We use these real data to generalise their applicability to any number of geographic areas. The National Health Service (NHS) provided data on ILI consultation data from 2009 to 2015. Our work is extending the Weekly Cases Ratio (WCR) method to develop an automatic system to raise an alarm when there is a sudden increase in ILI cases. The WCR method uses two terms: 1) the value of WCR, defined in week w as WCR(w)=ILI GP consultation rate in week w divided by the corresponding rate in week w-1, and 2) the number of health boards N(HB) which report an increase in ILI cases in week w compared to week w-1 (i.e. WCR(w)>1). We initially used the above Scottish data, then extended this to more than 14 HBs, by simulating data with similar structure but from a different number of HBs. We used a constant rate of consultations to determine a joint null distribution for (WCR, N(HB)) to use in a hypothesis testing approach to detect a rise in ILI cases. Using more than 3 million simulations of each data set, we found a relationship in all cases between WCR and N(HB), then modelled the relationship between each of the mean (µ) and standard deviation (σ) of WCR with N(HB) in each dataset to find a general equation for use with any number of HBs. Further work will compare our approach with the Moving Epidemic and Cumulative Sum methods. Some results of this work will be presented.
LanguageEnglish
Number of pages1
Publication statusPublished - 4 Sep 2017
EventRoyal Statistical Society Conference 2017 - Glasgow
Duration: 4 Sep 20177 Sep 2017

Conference

ConferenceRoyal Statistical Society Conference 2017
Abbreviated titleRSS2017
CityGlasgow
Period4/09/177/09/17

Fingerprint

Influenza
Health
Mean deviation
Health Services
Cumulative Sum
Null Distribution
Hypothesis Testing
Rate Constant
Joint Distribution
Standard deviation
Generalise
Term
Testing

Keywords

  • early detection
  • influenza
  • weekly cases ratio (WCR)

Cite this

@conference{0ffc955245334642ba029fd10628c3d1,
title = "Comparing methods for early detection systems for seasonal and pandemic influenza",
abstract = "The aim of this work is to investigate established methods for early detection of seasonal and pandemic influenza, then to develop a new method for routine use. We investigate daily data for influenza-like illness (ILI) General Practice (GP) consultations, collected from all 14 spatially located Health Boards (HBs) in Scotland. We use these real data to generalise their applicability to any number of geographic areas. The National Health Service (NHS) provided data on ILI consultation data from 2009 to 2015. Our work is extending the Weekly Cases Ratio (WCR) method to develop an automatic system to raise an alarm when there is a sudden increase in ILI cases. The WCR method uses two terms: 1) the value of WCR, defined in week w as WCR(w)=ILI GP consultation rate in week w divided by the corresponding rate in week w-1, and 2) the number of health boards N(HB) which report an increase in ILI cases in week w compared to week w-1 (i.e. WCR(w)>1). We initially used the above Scottish data, then extended this to more than 14 HBs, by simulating data with similar structure but from a different number of HBs. We used a constant rate of consultations to determine a joint null distribution for (WCR, N(HB)) to use in a hypothesis testing approach to detect a rise in ILI cases. Using more than 3 million simulations of each data set, we found a relationship in all cases between WCR and N(HB), then modelled the relationship between each of the mean (µ) and standard deviation (σ) of WCR with N(HB) in each dataset to find a general equation for use with any number of HBs. Further work will compare our approach with the Moving Epidemic and Cumulative Sum methods. Some results of this work will be presented.",
keywords = "early detection, influenza, weekly cases ratio (WCR)",
author = "Muqrin Almuqrin and Chris Robertson and Alison Gray",
year = "2017",
month = "9",
day = "4",
language = "English",
note = "Royal Statistical Society Conference 2017, RSS2017 ; Conference date: 04-09-2017 Through 07-09-2017",

}

Almuqrin, M, Robertson, C & Gray, A 2017, 'Comparing methods for early detection systems for seasonal and pandemic influenza' Royal Statistical Society Conference 2017, Glasgow, 4/09/17 - 7/09/17, .

Comparing methods for early detection systems for seasonal and pandemic influenza. / Almuqrin, Muqrin; Robertson, Chris; Gray, Alison.

2017. Abstract from Royal Statistical Society Conference 2017, Glasgow, .

Research output: Contribution to conferenceAbstract

TY - CONF

T1 - Comparing methods for early detection systems for seasonal and pandemic influenza

AU - Almuqrin, Muqrin

AU - Robertson, Chris

AU - Gray, Alison

PY - 2017/9/4

Y1 - 2017/9/4

N2 - The aim of this work is to investigate established methods for early detection of seasonal and pandemic influenza, then to develop a new method for routine use. We investigate daily data for influenza-like illness (ILI) General Practice (GP) consultations, collected from all 14 spatially located Health Boards (HBs) in Scotland. We use these real data to generalise their applicability to any number of geographic areas. The National Health Service (NHS) provided data on ILI consultation data from 2009 to 2015. Our work is extending the Weekly Cases Ratio (WCR) method to develop an automatic system to raise an alarm when there is a sudden increase in ILI cases. The WCR method uses two terms: 1) the value of WCR, defined in week w as WCR(w)=ILI GP consultation rate in week w divided by the corresponding rate in week w-1, and 2) the number of health boards N(HB) which report an increase in ILI cases in week w compared to week w-1 (i.e. WCR(w)>1). We initially used the above Scottish data, then extended this to more than 14 HBs, by simulating data with similar structure but from a different number of HBs. We used a constant rate of consultations to determine a joint null distribution for (WCR, N(HB)) to use in a hypothesis testing approach to detect a rise in ILI cases. Using more than 3 million simulations of each data set, we found a relationship in all cases between WCR and N(HB), then modelled the relationship between each of the mean (µ) and standard deviation (σ) of WCR with N(HB) in each dataset to find a general equation for use with any number of HBs. Further work will compare our approach with the Moving Epidemic and Cumulative Sum methods. Some results of this work will be presented.

AB - The aim of this work is to investigate established methods for early detection of seasonal and pandemic influenza, then to develop a new method for routine use. We investigate daily data for influenza-like illness (ILI) General Practice (GP) consultations, collected from all 14 spatially located Health Boards (HBs) in Scotland. We use these real data to generalise their applicability to any number of geographic areas. The National Health Service (NHS) provided data on ILI consultation data from 2009 to 2015. Our work is extending the Weekly Cases Ratio (WCR) method to develop an automatic system to raise an alarm when there is a sudden increase in ILI cases. The WCR method uses two terms: 1) the value of WCR, defined in week w as WCR(w)=ILI GP consultation rate in week w divided by the corresponding rate in week w-1, and 2) the number of health boards N(HB) which report an increase in ILI cases in week w compared to week w-1 (i.e. WCR(w)>1). We initially used the above Scottish data, then extended this to more than 14 HBs, by simulating data with similar structure but from a different number of HBs. We used a constant rate of consultations to determine a joint null distribution for (WCR, N(HB)) to use in a hypothesis testing approach to detect a rise in ILI cases. Using more than 3 million simulations of each data set, we found a relationship in all cases between WCR and N(HB), then modelled the relationship between each of the mean (µ) and standard deviation (σ) of WCR with N(HB) in each dataset to find a general equation for use with any number of HBs. Further work will compare our approach with the Moving Epidemic and Cumulative Sum methods. Some results of this work will be presented.

KW - early detection

KW - influenza

KW - weekly cases ratio (WCR)

UR - http://www.rss.org.uk/RSS/Events/RSS_Conference/2017_Conference/RSS/Events/Conference/2017_conference.aspx?hkey=2a432b6b-6baf-4bc3-baa4-063221c13ab8

M3 - Abstract

ER -

Almuqrin M, Robertson C, Gray A. Comparing methods for early detection systems for seasonal and pandemic influenza. 2017. Abstract from Royal Statistical Society Conference 2017, Glasgow, .