A Proof Of Concept For A Syndromic Surveillance System Based On Routine Ambulance Records In The South-west Of England, For The Influenza Season 2016/17
Summary (2 min read)
Introduction
- The digitisation of ambulance healthcare records has created a large pre-hospital data source that to date is mostly untapped.
- South Western Ambulance Service NHS Foundation Trust introduced electronic patient care records in March 2015, making it possible to access and monitor all data recorded in near real-time.
- Temperature screening has been applied during outbreaks of infectious diseases, such as severe acute respiratory syndrome (SARS) (Samaan, Patel, Spencer, & Roberts, 2004; Syed, Sopwith, Regan, & Bellis, 2003).
- To evaluate a method adapted from Singh, Savill, Ferguson, Robertson and Woolhouse (2010) using case ratios (CRs) and its applicability as an early event detection (EED) system when applied to pre-hospital tympanic temperature readings.
Data extraction
- All ePCRs created between 1 January 2015 and 30 April 2017, with an incident postcode matching the county of Devon or Cornwall, were eligible for inclusion.
- The postcode, record creation date, tympanic temperature and age were requested and provided by the SWASFT Clinical Information and Records Office.
Temperature measurement in South Western Ambulance Service NHS Foundation Trust
- The most commonly used temperature probes within SWASFT are the Braun ThermoScan 7 IRT6520 and ThermoScan 5 IRT4520.
- A delay of one week was chosen because it includes the incubation time, meaning that secondary patients exposed to influenza should have developed pyrexia within one week (Lessler et al., 2009).
Influenza detection
- To establish whether seasonal influenza was detectable, weekly case numbers were compared with weekly sentinel influenza cases recorded by the ECDC in England.
- Sentinel surveillance data are based on a network of selected healthcare facilities, which select patients with symptoms suggesting influenza for laboratory confirmation.
Calculation of the modified case ratio CRd
- This value indicates the mean secondary infections caused by each infected host in a naïve population without immunity against the infectious agent.
- Methods exist to estimate R 0 from the progress of a disease outbreak, which rely on knowledge about the transmission characteristics of the infectious agent gathered from previous outbreaks (Althaus, 2014; Griffin, Garske, Ghani, & Clarke, 2011; Potapov, Merrill, Pybus, & Lewis, 2015).
- As this evaluation only focuses on abnormal temperature readings, the infection that could be responsible is not possible to determine and so cannot be compared directly to previous outbreaks.
- Here this method is applied to pyrexia cases as an unspecific substitute for infection.
Outbreak definition
- The outbreak definition is focused on the ascending slope, representing an increase in pyrexia case numbers.
- The different mean-CRd depending on window sizes.
- To establish the effect of different choices of d, the ascending area of pyrexia cases peak in 2016/2017 was used to calculate a sliding CR d with varying d for the ascending slope where pyrexia cases increased.
Improving accuracy
- The weekly data were smoothed using the EMA of 21 days (or three sample points) before the sliding CR 21 was calculated .
- Once again, the outbreak could not be detected using a threshold method as the number is below the baseline (565) as well as the mean (644.9).
Daily detection
- The peak was reached with 133 (18.5%) patients of 721 calls (fractions are caused by the smoothing process using the EMA).
- This value is 6.8% below the baseline (76.2) and within the standard deviation (23.5), which would not be detectable using a threshold method.
- This start of the seasonal increase of infections was detected earlier than influenza cases by the ECDC, which identified the start in week 46, 2016 (European Centre for Disease Prevention and Control, 2017).
Weekly detection
- 4(2) 22–30 indicator of infection allows the unspecific monitoring of infectious diseases within the community.
- The seasonal increase of fever cases was detected up to nine weeks before influenza cases were recorded by conventional methods employed by the ECDC.
- In the UK, the sentinel detection runs between October and March, thus it could not detect earlier cases.
Discussion
- Ambulance crews within SWASFT have collected data for 16% of the population in Devon and Cornwall within a year (Office for National Statistics, 2017).
- This reflects the fact that the elderly and the very young are more likely to require assistance by an ambulance.
- From these data, it was possible to establish that the pyrexia counts timely matched the seasonal influenza outbreak recorded by the ECDC.
- A proportion of cases will have been caused by other circulating infections.
Limitations
- The collected data could have included patients with multiple ambulance attendances a year, which cannot be accounted for, as no patient identifying data were extracted.
- An unknown proportion of pyrexia cases will be caused by other infections, although it can be expected that a large fraction was caused by the circulating seasonal influenza virus.
- Furthermore, the comparison data originated from different geographic populations (Devon and Cornwall vs. England) and were compared to confirmed influenza diagnoses.
- It still requires the user to define the value of d, which will normally require some knowledge about the transmission rate of the monitored infection.
Conclusion
- Data from ambulance service ePCRs correlate with the sentinel data collected by the ECDC, allowing these data to inform an EED system.
- The detection of events occurred earlier compared to the ECDC, but does not distinguish between infectious agents.
- The move to digital patient records makes it possible to monitor the large proportions of the population at high sample rates, and for several syndromes simultaneously, making it an ideal data source for an EED system.
Did you find this useful? Give us your feedback
Citations
237 citations
References
136 citations
"A Proof Of Concept For A Syndromic ..." refers background in this paper
...As the data included temperature readings outside the range of the device capabilities, records indicating physiological hypothermia (32◦ C) [14,15] and hyperpyrexia (42◦ C) [7] were excluded....
[...]
133 citations
109 citations
"A Proof Of Concept For A Syndromic ..." refers background in this paper
...However, this might be due to the fact that the sentinel surveillance runs between October and March and thus had not started when our system detected an increase in cases [33] To establish the performance of the method, the outbreak detected by PHE was considered the ground truth [31] which allowed to estimate a specificity of 99....
[...]
104 citations
"A Proof Of Concept For A Syndromic ..." refers background in this paper
...03◦ C of tympanic versus oral temperature [17]....
[...]
...A cohort study conducted by Obermeyer et al. (2017) [17] in 35,488 healthy patients confirmed the daily variation and found a normal range of 35.3-37.7° C....
[...]
...(2017) [17] in 35,488 healthy patients confirmed the daily variation and found a normal range of 35....
[...]
102 citations