ArticlePDF Available

Effects of an air pollution personal alert system on health service usage in a high-risk general population: A quasi-experimental study using linked data


Abstract and Figures

Background There is no evidence to date on whether an intervention alerting people to high levels of pollution is effective in reducing health service utilisation. We evaluated alert accuracy and the effect of a targeted personal air pollution alert system, airAware, on emergency hospital admissions, emergency department attendances, general practitioner contacts and prescribed medications. Methods Quasi-experimental study describing accuracy of alerts compared with pollution triggers; and comparing relative changes in healthcare utilisation in the intervention group to those who did not sign-up. Participants were people diagnosed with asthma, chronic obstructive pulmonary disease (COPD) or coronary heart disease, resident in an industrial area of south Wales and registered patients at 1 of 4 general practices. Longitudinal anonymised record linked data were modelled for participants and non-participants, adjusting for differences between groups. Results During the 2-year intervention period alerts were correctly issued on 208 of 248 occasions; sensitivity was 83.9% (95% CI 78.8% to 87.9%) and specificity 99.5% (95% CI 99.3% to 99.6%). The intervention was associated with a 4-fold increase in admissions for respiratory conditions (incidence rate ratio (IRR) 3.97; 95% CI 1.59 to 9.93) and a near doubling of emergency department attendance (IRR=1.89; 95% CI 1.34 to 2.68). Conclusions The intervention was associated with increased emergency admissions for respiratory conditions. While findings may be context specific, evidence from this evaluation questions the benefits of implementing near real-time personal pollution alert systems for high-risk individuals.
Content may be subject to copyright.
Effects of an air pollution personal alert system
on health service usage in a high-risk general
population: a quasi-experimental study using
linked data
R A Lyons,
S E Rodgers,
S Thomas,
R Bailey,
H Brunt,
D Thayer,
J Bidmead,
B A Evans,
P Harold,
M Hooper,
H Snooks
Additional material is
published online only. To view
please visit the journal online
Swansea University Medical
School, Swansea, UK
Cwm Taf Public Health Team,
Public Health Wales, Keir
Hardie University Health Park,
Merthyr Tydl, UK
Health Protection Team,
Public Health Wales, Cardiff,
Member of the Public,
Swansea, UK
Public Health England, Centre
for Radiation Chemical and
Environmental Hazards (Wales),
Metropolitan University,
Cardiff, UK
Neath Port Talbot County
Borough Council, Neath, UK
Correspondence to
Dr SE Rodgers, Swansea
University Medical School,
Swansea University, Data
Science Building, Floor 3,
Singleton Park,
Swansea SA2 8PP, UK;
Received 14 January 2016
Revised 28 April 2016
Accepted 6 May 2016
To cite: Lyons RA,
Rodgers SE, Thomas S, et al.
J Epidemiol Community
Health Published Online
First: [please include Day
Month Year] doi:10.1136/
Background There is no evidence to date on whether
an intervention alerting people to high levels of pollution
is effective in reducing health service utilisation. We
evaluated alert accuracy and the effect of a targeted
personal air pollution alert system, airAware, on
emergency hospital admissions, emergency department
attendances, general practitioner contacts and prescribed
Methods Quasi-experimental study describing accuracy
of alerts compared with pollution triggers; and
comparing relative changes in healthcare utilisation in
the intervention group to those who did not sign-up.
Participants were people diagnosed with asthma, chronic
obstructive pulmonary disease (COPD) or coronary heart
disease, resident in an industrial area of south Wales
and registered patients at 1 of 4 general practices.
Longitudinal anonymised record linked data were
modelled for participants and non-participants, adjusting
for differences between groups.
Results During the 2-year intervention period alerts
were correctly issued on 208 of 248 occasions;
sensitivity was 83.9% (95% CI 78.8% to 87.9%) and
specicity 99.5% (95% CI 99.3% to 99.6%). The
intervention was associated with a 4-fold increase in
admissions for respiratory conditions (incidence rate ratio
(IRR) 3.97; 95% CI 1.59 to 9.93) and a near doubling
of emergency department attendance (IRR=1.89; 95%
CI 1.34 to 2.68).
Conclusions The intervention was associated with
increased emergency admissions for respiratory
conditions. While ndings may be context specic,
evidence from this evaluation questions the benets of
implementing near real-time personal pollution alert
systems for high-risk individuals.
The UK Government has responded to the growing
evidence of adverse health effects from air pollu-
tion by specifying human health protection expos-
ure limits.
These recommend implementing
effective measures to reduce pollution levels and
minimise public exposure and consequent
Recommendations made by the
Committee on the Medical Aspects of Air Pollution
(COMEAP) in 2011 stimulated the development of
air quality alert systems providing important infor-
mation and advice to the public to minimise air
pollutant exposure.
These are expected to
reduce adverse effects of pollution and health
service utilisation through reduction of symptoms.
Air quality alert systems have operated in the UK
for a number of years, including the airText system
in London, airAlert in Sussex, Know and Respond
in Scotland and the Met Ofce has Healthy
Outlook. All operate on forecasts of air quality for
the following day.
We only found one published
quantitative evaluations of an air quality alert
system. This evaluation assessed changes in add-
itional hospital admissions estimated for a test
group of patients who would be likely to sign-up
for the alert service, modelled against the current
level of admissions for the populations of London
and Sussex, UK.
The airAware system was a novel development
because it integrated near real-time data rather than
forecasting. Its design facilitated early identication
of local air pollution problems and issued timely
warnings of air pollution episodes reecting levels
of particulate matter (measured as particulate
matter 10 μm or less in diameter (PM
)) at nearby
air quality monitoring stations. People who signed
up to the system received alerts from an independ-
ent contractor who operated the airAware system
on behalf of the multiagency Local Service Board
(who had responsibility to oversee the delivery of
the project). The contractor managed participant
sign-up and the issue of alerts.
The airAware system operated in a setting with a
well-established population data linkage system
capable of supporting the evaluation of individual-
level interventions.
We undertook a retrospect-
ive cohort study using individual linked data to
evaluate the effectiveness and impact of this novel
Our objectives were: (1) to evaluate the accuracy of
the airAware system in correctly issuing alerts in
terms of sensitivity, specicity, positive and negative
predictive values (NPVs); and (2) to evaluate the
effect of the system on subsequent health service
utilisation. We have reported this evalution follow-
ing STROBE guidelines.
Study design
We created an electronic cohort of individuals
invited to sign-up to the system from anonymised
primary care data to evaluate the impact of the air
Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222 1
Research report
JECH Online First, published on May 23, 2016 as 10.1136/jech-2016-207222
Copyright Article author (or their employer) 2016. Produced by BMJ Publishing Group Ltd under licence.
quality alerting system for patients with pollution-susceptible
diseases on their health outcomes, recorded in routinely col-
lected data. The invitation, patient information leaet and
patient registration welcome pack are shown in the online sup-
plementary material. We compared changes in primary and sec-
ondary care health service utilisation preintervention and
postintervention for an intervention group who signed up to the
alert system, to the control group who did not.
An Air Quality Management Area (AQMA) in Port Talbot, an
industrial urban area of south Wales, UK, adjacent to a steel-
works and motorway, with six air pollution monitors for PM
The airAware system sent alerts in near real-time by text, email
or pre-recorded voice. Alerts, automatically triggered by higher
pollution levels, advised of changed air quality and used advice
on self-care and healthy behaviour. Messages were based on
COMEAP Air Quality Index Health Advice
and agreed by the
project team (academics, public health practitioners, respiratory
physicians, general practitioners (GPs), technical staff and lay
members) and a service user group called SUCCESS (Service
Users with Chronic Conditions Encouraging Sensible Solutions)
to ensure they could be clearly understood by participants.
Alert messages and trigger criteria are shown in gure 1.
To avoid overalerting, messages were sent between 7:00 and
22:00 only. Once air quality returned to normal levels, an all-
clearmessage was sent. The number of messages per day was
limited to three; on any single day no more than three alerts
could be issued. Only one alert was issued for the day unless a
higher pollution trigger level was met during the daily alerting
Eligible participants were patients registered with one of four
local general practices, with asthma, chronic obstructive pul-
monary disease (COPD) or coronary heart disease (CHD). They
were included on the practice registers for these conditions as
required by the Quality and Outcomes Framework (QOF) of
the National Health Service (NHS) General Medical Services
contract. The list of General Practice Read codes dening these
conditions can be found in the the QOF.
Patients meeting the inclusion criteria were individually invited
by GP letter in May 2012 to sign-up to the airAware system. We
used a QOF-based algorithm to generate invitees, the QOF is an
incentive-based programme for GPs based in the UK to assess
their achievements, for example, in regularly treating diagnosed
The letter was accompanied by a pack that included
an information leaet (with online registration details), paper
registration sheet and prepaid addressed return envelope.
All documentation was produced in English and Welsh. The
leaet and pack were developed by the project team that included
service users. The invitation letter and information leaet, and
registration pack are shown in the online supplementary files.
To register, participants selected: preferred method of message
receipt (voice message to land-line, text to mobile phone, or
email); preferred language (English or Welsh); from which of the
six local monitors (identied by name and location) they wished
to receive alerts. Text message was the most popular method of
receiving the messages (43%), followed by voice message (31%)
and email (26%). Intervention and control group participants
were agged in an anonymised databank for follow-up (the
Secure Anonymised Information Linkage (SAIL) databank).
14 15
System performance
Data sources
Air pollution data were routinely obtained from six local
authority-operated monitors. Hourly air pollution data, mea-
sured at each monitoring station, were received by the system
contractor and used to forecast potential 24-hour episodes for
We assessed the proportion of alerts correctly issued when trig-
gers were breached, and the proportion of false alerts issued
when triggers were not breached.
Relative change in health service utilisation
Data sources
We used the SAIL Databank and associated linkage methods to
provide evaluation data.
15 16
The SAIL Databank contains anon-
ymised linked demographic and health records for the
Figure 1 Air quality bands, alert triggerthreshold criteria and health messages.
2 Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222
Research report
population of Wales, including those derived from the NHS
patient registration system, inpatient, outpatient, emergency
department (ED) attendance and general practice data. Privacy
protection is ensured by use of a split le, multiorganisation
encryption design that removes all names, addresses, dates of
birth and other identiers within the NHS data and replaces
them with unique non-identifying eight digit numbers for indi-
viduals and addresses before reassembling with the clinical
The Patient Episode Dataset for Wales (PEDW) is
held within SAIL and contains hospital admissions data
recorded in Welsh hospitals. The SAIL Databank also contains
data sets from the majority of GPs in Wales, including the four
general practices from which patient lists were derived for this
study. This is the rst time an air pollution intervention has
been evaluated using routine data linked at the individual level.
Our independent variable of interest was exposure to the
airAware intervention. We derived health outcome measures
from records using selected clinical READ codes related to
respiratory conditions (asthma and COPD) and CHD (collect-
ively referred to as relevant conditions). Outcome measures con-
sisted of counts for: hospital admissions (emergency and elective,
emergency only respiratory and CHD conditions combined and
separately); ED attendances (non-trauma); outpatient atten-
dances (non-trauma); primary care contacts (respiratory, CHD
and mental health combined and separately) and prescribed med-
ications (combined respiratory, CHD and mental health).
A hospital admission within PEDW was considered a relevant
admission when any episode within that admission contained a
primary diagnosis code for one of the relevant conditions,
recorded using the International Classication of Diseases, 10th
revision (ICD-10). Coding groups I20-I25, J42-J44 and J45
were used to identify records relating to CHD, COPD and
asthma, respectively.
Primary care contacts were approximated by measuring the
number of days on which an individual had any event recorded
in the GP data (it is not possible to measure whether a patient
visited the practice using the data provided to SAIL). Diagnoses,
symptoms and treatments associated with a contact were identi-
ed using the Read code system V.2. The Read codes used to
identify records related to the relevant conditions, as well as
common mental health disorders, can be found in an online
supplementary File.
We used the intervention groupssign-up date to the airAware
system as their intervention start date. We randomly assigned
pseudo sign-up dates to controls, matched to the distribution of
participant sign-up dates, to derive comparable observation time
periods for the intervention and control study arms. We col-
lected preintervention data for up to 1 year for primary care
outcomes and secondary care events (less frequent) for a 2-year
period prior to the start of the intervention.
Dates of migration in and out of the study area and the date
of death were both extracted from the databank for censoring
purposes. We followed up all participants until: they moved out
of the study area; they died; the end of the study period, which-
ever was earliest. We dened a participants observation period
end date of 30 April 2014 for primary care and 1 August 2014
for secondary care data.
Bias is possible in any evaluation of a non-randomised interven-
tion. We expected to observe self-selection bias in a voluntary
participation study such as this. We attempted to control for
bias by using a controlled before and afterapproach comparing
change in health service utilisation between intervention and
control arms. We adjusted for preintervention differences likely
to inuence the outcome of interest, such as age, gender, depriv-
ation index and smoking status (current and past).
Study size
This pragmatic evaluation of an existing service used all avail-
able routinely collected patient data to determine health out-
comes, and no formal power calculation was conducted.
Statistical methods
System performance
We assessed the validity of the intervention by considering
system sensitivity and specicity. In the context of this evalu-
ation, sensitivity indicates the proportion of alerts that were cor-
rectly issued when system alert triggers were met and specicity
is the proportion of alerts correctly not issued when system alert
triggers were not met. We also assessed the positive predictive
value (PPV), the proportion of all alerts sent that were true posi-
tives; and NPV, the proportion of all occasions when alerts were
not required and not sent. To calculate system validity, the total
assessment denominator was 13 140, dened as a 730-day
(2 years) test period with potential for a maximum of three
alerts per day from each of the six monitors.
Relative change in health service utilisation
We compared the preintervention characteristics of the interven-
tion group with those of the control group; differences were
tested for signicance using ORs and χ
tests. We compared
health outcomes of the two study groups during the preinter-
vention period with unadjusted and adjusted incidence rate
ratios (IRRs). We also compared outcomes between the two
groups in the postintervention period.
For our main analysis, we created separate, negative binomial
regression models for each health service utilisation measure.
We compared the total number of unique health utilisation
counts for each participant in the time period before the sign-up/
pseudo sign-up date, with the time period immediately after. The
estimated change in healthcare utilisation associated with the
intervention was derived from the interaction term between the
participation group and the time period in the regression model.
We tested confounding variables for signicant associations
with the dependent variables and possible interaction effects
were also explored. Random effects were included to account
for dependence between premeasures/postmeasures for the same
individuals and the natural logarithm of the number of observa-
tion days for each participant was added as an offset to account
for the different lengths of data collection periods. We exponen-
tiated coefcients of the regression model to produce adjusted
IRR, in order to aid interpretation.
System performance
Overall system validity is summarised for the entire pilot period
in table 1. The sensitivity of the system was 83.9% (95% CI
78.8% to 87.9%) corresponding to a total of 208 PM
correctly issued (187 yellow alerts, 17 amber alerts and 4 red
alerts) out of a total of 248 alerts that should have been issued.
Specicity was 99.5% (95% CI 99.3% to 99.6%) relating to the
proportion of alerts correctly not issued when system alert trig-
gers were not met. A total of 67 out of 275 alerts sent were
false positives, giving a PPV of 75.6% (95% CI 70.2% to
80.3%). The NPV was 99.7% (95% CI 99.6% to 99.8%))
Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222 3
Research report
relating to 12 825 true-negative events out of the 12 865 occa-
sions where an alert was not sent.
A total of 1395 patients attending the four practices met the
inclusion criteria, and 180 signed up for the intervention. Two
people, one in each of the intervention and control groups
could not be linked to the data in SAIL and were excluded from
analysis. Figure 2 shows the ow of patients in the study.
We took into account the varying observation periods for each
participant depending on their sign-up date. Recruitment took
place over more than 1 year but the majority of participants
(90.6%) signed up between May 2012 and July 2012, with the
remainder signing up before April 2013; 98.4% had 2 years of
data preintervention for secondary care outcomes; 89.8% had
1 year of data preintervention for primary care outcomes.
Descriptive data
Preintervention characteristics showed the modal age group was
6574 years for the intervention and control groups (table 2).
There was a higher proportion of older participants in the inter-
vention group compared with the control group, resulting in
statistically signicant differences in the age distributions
=46.3, p<0.001). There were statistically non-signicant dif-
ferences by gender, deprivation and smoking status.
Health outcome results
Preintervention, the intervention group had fewer hospital
admissions; all admissions, relevant emergency admissions,
respiratory emergency admissions and CHD-related emergency
admissions (gure 3). The intervention group had lower rates of
outpatient attendances, ED attendances, GP contacts for mental
health conditions, GP contacts for CHD conditions and slightly
lower rates of prescriptions. Numbers for each outcome and
rates both preintervention and postintervention are given in an
online appendix table. The rate of GP contacts for respiratory
conditions was higher among the intervention group compared
with the control group. GP contacts for all relevant conditions
were slightly higher for the intervention group but this was not
statistically signicant.
Intervention effect results
Variables listed in table 2 have been established as potential con-
founders and were used to adjust the nal model. The interven-
tion was associated with signicant increases in all relevant
emergency admissions, respiratory-related emergency admissions
and ED attendances (table 3).
Additional analyses
We completed analyses to investigate if excess admissions
occurred during periods of higher pollution by restricting obser-
vation periods to 7 days following PM
low band exceedance.
There were insufcient numbers (and in some cases no records)
of secondary care outcomes for analysis. Estimates of effects
were derived for GP outcomes, prescribed medications and out-
patient visits. There was no statistically signicant effect asso-
ciated with the intervention when looking at events immediately
following pollution episodes (results not shown).
Our objectives were to evaluate the accuracy of the airAware
system in correctly identifying pollution episodes and issuing
alerts; and evaluate the effect of the system on subsequent
health service utilisation.
Key results
The airAware air pollution alerting system missed a small pro-
portion of alerts and issued 67 false alerts among 275 alerts
issued. During the 2-year study period, system performance
Table 1 Validity of the alerts issued by the airAware system
Alert trigger met
Yes No Total
Alert issued
Yes 208 67 275
No 40 12 825 12 865
Total 248 12 892 13 140
Figure 2 Flow of patients through airAware system and evaluation. SAIL, Secure Anonymised Information Linkage.
4 Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222
Research report
sensitivity was 83.9% and specicity was 99.5%. Participants
with existing respiratory and CHD conditions who were in the
intervention arm had a greater relative change in health service
use following the start of the intervention compared with those
in the control arm of the study. The intervention group experi-
enced a doubling of emergency admissions for all relevant con-
ditions and a fourfold admissions increase for respiratory
conditions. These ndings are important given one intention of
these public health interventions is to reduce health service
Comparison with previous studies
This is the rst study of an air pollution alert system reporting
a statistically signicant adverse difference in health service util-
isation. This is noteworthy, given that several systems are in
operation but have been unable to detect such an association,
having been evaluated qualitatively, using self-report data, or
quantitatively assessing changes within the total population
using an ecological study design.
Our study has several
strengths. First, we used a search of disease registers in the
local GPs to identify all those meeting the high-risk guidelines
who would potentially benet from the alerts. Second, using a
data linkage system, we specically agged those who signed up
to the intervention with their precise date of intervention start.
Third, the use of retrospective and prospective data linkage is
novel, because it allowed us to extract detailed anonymised
healthcare utilisation for the control group in the same way as
for the intervention group with almost no loss to follow-up
and the avoidance of recall bias. Fourth, the use of six pollution
Table 2 Preintervention characteristics of the intervention and
control groups
group Total
014 7 (3.9%) 87 (7.2%) 94 (6.7%)
1524 5 (2.8%) 99 (8.2%) 104 (7.5%)
2534 9 (5.0%) 87 (7.2%) 96 (6.9%)
3544 13 (7.3%) 107 (8.8%) 120 (8.6%)
4554 14 (7.8%) 166 (13.7%) 180 (12.9%)
5564 37 (20.7%) 187 (15.4%) 224 (16.1%)
6574 60 (33.5%) 220 (18.1%) 280 (20.1%)
7584 26 (14.5%) 172 (14.2%) 198 (14.2%)
85+ 8 (4.5%) 89 (7.3%) 97 (7.0%)
Female 92 (51.4%) 583 (48.0%) 675 (48.5%)
Males 87 (48.6%) 631 (52.0%) 718 (51.5%)
WIMD quintile
Least deprived and next
least deprived
16 (8.9%) 95 (7.8%) 111 (8.0%)
Average deprivation 31 (17.3%) 214 (17.6%) 245 (17.6%)
Next most deprived 106 (59.2%) 742 (61.1%) 848 (60.9%)
Most deprived 26 (14.5%) 163 (13.4%) 189 (13.6%)
Current smoker
No 153 (85.5%) 1035 (85.3%) 1188 (85.3%)
Yes 26 (14.5%) 179 (14.7%) 205 (14.7%)
History of smoking
No 90 (50.3%) 721 (59.4%) 811 (58.2%)
Yes 89 (49.7%) 493 (40.6%) 582 (41.8%)
WIMD, Welsh Index of Multiple Deprivation.
Figure 3 Differences between groups
(unadjusted and adjusted IRR and
95% CIs) for preintervention and
postintervention groups. An IRR >1
indicates a higher rate in the
intervention group compared with the
control group. CHD, coronary heart
disease; GP, general practitioner; IRR,
incidence rate ratio; MH, mental
Table 3 The intervention effect (IRR and 95% CI)
Outcome measure
Intervention effect
IRR 95% CI
GP relevant contacts 1.04 0.98 to 1.11
GP respiratory contacts 1.04 0.96 to 1.13
GP CHD contacts 1.02 0.95 to 1.11
GP MH contacts 0.98 0.84 to 1.16
Prescribed medications 1.03 0.98 to 1.09
All admissions 0.82 0.58 to 1.14
Relevant emergency admissions 2.04 1.06 to 3.93
Respiratory emergency admissions 3.97 1.59 to 9.93
CHD emergency admissions 0.97 0.39 to 2.42
Outpatient attendances 1.01 0.83 to 1.25
Emergency attendances 1.89 1.34 to 2.68
Statistically significant IRRs are in bold.CHD, coronary heart disease; GP, general
practitioner; IRR, incidence rate ratio; MH, mental health.
Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222 5
Research report
monitors in a small area means that individual exposures and
breaches of limits were more accurately measured than across
entire cities.
This non-randomised observational study was conducted in the
specic context of exposure in a small area. Evaluations under-
taken in different settings may produce different results because
the potential for benets and harms from air alerting services
will depend on local pollution proles. It was not possible to
randomise people into control and intervention arms of the
study for this intervention. Participants were a self-selecting
sample with different characteristics from the control population
in terms of preintervention primary and secondary health ser-
vices utilisation. The intervention group were characterised by
their engagement in preventative health, as suggested from their
relatively higher rate of GP events (gure 3) preintervention,
resulting in a lower rate of emergency admissions compared
with the control group preintervention.
Although we made multiple design and analysis adjustments
to control for bias, it is likely that some residual confounding
remains. Our use of a controlled pre-post regression analysis of
health service utilisation, adjusting for socioeconomic character-
istics recorded in routine data, will not have captured the
impact of unmeasured confounders.
The increase in emergency hospital admissions for signing up to
the intervention was largely as a result of respiratory conditions.
We suggest that the message to follow your doctors usual
advicereceived by the intervention group, resulted in a heigh-
tened awareness in a group of people with chronic conditions
who are unwell frequently. Patients did not increase GP activity
but principally accessed EDs. Many of these patients will have
chronic limitations in respiratory function which, in the context
of reporting feeling unwell and being assessed by doctors
unfamiliar with their usual status, may have resulted in admis-
sion to hospital. We propose that the alerts caused worry to the
extent that it prompted intervention group participants to seek
medical care and this in itself constitutes harm.
While all AQMAs have features that are unique and hence the
potential trade off for benets and harms associated with air
pollution alerting services will not be the same in any two areas,
aspects of this study may be generalisable. It is likely that the
behaviour of people who have signed up to such services else-
where are similar to those in this study and the same conse-
quences of excess health service utilisation would arise.
Replication is an important component of science and we would
like to see this study repeated in another environment, particu-
larly if randomising participants to the intervention group
cannot be achieved.
This study of a near real-time air pollution alert was associated
with increased use of health services. Our ndings raise ques-
tions about the trade off between harms and benets of air pol-
lution alerting services. This intervention was associated with an
outcome that was in the opposite direction to expected from
the COMEAP guidelines. There is a growing evidence base
demonstrating some public health interventions are harmful.
Wider roll-out of such systems does not appear to be warranted
given the current evidence base.
Ethics and information governance
This study received approval from an independent Information
Governance Review Panel (IGRP), an independent body consist-
ing of membership from a range of government, regulatory and
professional agencies.
The cohort has also been assessed by
the Multi-centre Research Ethics Committee for Wales and
judged to be an anonymised research database which does not
require ethical review in line with National Research Ethics
Service guidance. In compliance with IGRP rulings, and in line
with the Data Protection Act 1998,
individual-level data and
the corresponding encrypted linking eld codes were not
removed from the SAIL databank and thus were not included in
exploratory or reference documents. Analyses were carried out
within the SAIL Gateway, which provides a secure remote access
service to the SAIL databank held at Swansea University
Medical School.
What is already known on this subject
It is well established that poor air quality causes substantial
excess mortality and morbidity and additional symptoms,
particularly for high-risk individuals with existing respiratory
and coronary heart disease.
Government policy promotes advising high-risk individuals of
imminent pollution incidents to encourage behaviour
modication to reduce the impact of pollution.
Evaluation of the effectiveness of air pollution alerting
systems has been limited to small scale qualitative and
self-report data and larger scale ecological study designs.
What this study adds
A robust quantitative evaluation using longitudinal linked
individual-level heath records to evaluate an air pollution
alert intervention.
Evidence for unanticipated harms through a statistically
signicant increase in health service utilisation in the
intervention group.
Acknowledgements The authors would like to thank Ms Jacqui McCarthy,
Dr Gwyneth Davies, Ms Joanne Davies and Dr Steven Rohman for their valuable
contributions to this evaluation.
Contributors RAL conceived and designed the study and contributed to drafting
the manuscript and interpretation of the ndings. SER contributed to study design
and analyses, interpreted the ndings and drafted the manuscript. ST designed the
study and wrote the funder report, on which this paper is based. RB conducted the
statistical analyses and contributed to drafting the manuscript. HB contributed to the
study design. DT completed data extraction in preparation for statistical analysis. JB,
BAE, PH and MH contributed to the interpretation of ndings. HS conceived and
designed the study and contributed to drafting the manuscript and interpretation of
the ndings. All authors were involved in manuscript revisions and have approved
the nal version.
Funding The airAware system and part of the evaluation were funded by the
European Social Fund. We acknowledge additional support from The Farr Institute
and the Thematic Research network for emergency and UNScheduled Trauma care
(TRUST). The Farr Institute is supported by a 10-funder consortium: Arthritis
Research UK, the British Heart Foundation, Cancer Research UK, the Economic and
Social Research Council, the Engineering and Physical Sciences Research Council, the
Medical Research Council, the National Institute of Health Research, the National
Institute for Social Care and Health Research (Welsh Assembly Government), the
6 Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222
Research report
Chief Scientist Ofce (Scottish Government Health Directorates), the Wellcome Trust,
(MRC Grant No: MR/K006525/1). TRUST was supported by the National Institute of
Social care and Health Research (Welsh Assembly Government) (20102015).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Open Access This is an Open Access article distributed in accordance with the
terms of the Creative Commons Attribution (CC BY 4.0) license, which permits
others to distribute, remix, adapt and build upon this work, for commercial use,
provided the original work is properly cited. See:
1 DEFRA. The Air Quality Strategy for England, Scotland, Wales and Northern Ireland
(Volume 1). 2007. Available at:
2 Seaton A, Godden D, MacNee W, et al. Particulate air pollution and acute health
effects. Lancet 1995;345:1768.
3 Dockery DW, Pope CA III, Xu X, et al. An association between air pollution and
mortality in six U.S. cities. N Engl J Med 1993;329:17539.
4 Pope CA III, Burnett RT, Thun MJ, et al. Lung cancer, cardiopulmonary mortality,
and long-term exposure to ne particulate air pollution. JAMA 2002;287:113241.
5 Ayres J. Review of the UK Air Quality Index. Committee on the Medical Effects of
Air Pollutants. 2011. Available at:
6 Shah AS, Langrish JP, Nair H, et al. Global association of air pollution and heart
failure: a systematic review and meta-analysis. Lancet 2013;382:103948.
7 Kelly FJ, Fuller GW, Walton HA, et al. Monitoring air pollution: use of early warning
systems for public health. Respirology 2012;17:719.
8 World Health Organisation. Review of evidence on health aspects of air pollution
REVIHAAP Project. 2013. Available at:
9 AirText. (accessed 3 Nov 2015).
10 airAlert: air quality early warning service. http://www. (accessed 3 Nov 2015)
11 Know & RespondScotland, the free air pollution alert messaging systemAir
Quality in Scotland. (accessed
3 Nov 2015)
12 Met Ofce. Healthy Outlook®helping patients with COPD this winter. | Met
Ofce News Blog.
healthy-outlook-helping-patients-with-copd-this-winter/ (accessed 3 Nov 2015)
13 Walton H, Baker T, Fuller G, et al.Air pollution alert services evidence development
strategyprediction of possible effectiveness and assessment of intervention study
feasibility. Vol 83. Kings College London, 2014.
14 Lyons RA, Jones KH, John G, et al. The SAIL databank: linking multiple health and
social care datasets. BMC Med Inform Decis Mak 2009;9:3.
15 Ford DV, Jones KH, Verplancke JP, et al. The SAIL Databank: building a national
architecture for e-health research and evaluation. BMC Health Serv Res
16 Lyons RA, Ford DV, Moore L, et al. Use of data linkage to measure the population
health effect of non-health-care interventions. Lancet 2014;383:151719.
17 GMS ContractQOF Business Rules 2014/15 Version 30.
sites3/page.cfm?orgid=480&pid=76672 (accessed 8 May 2015)
18 Maughan D, Ansell J. Protecting resources, promoting value: a doctors guide
to cutting waste in clinical care. Vol 60. Academy of Medical Royal Colleges, 2014.
19 Bonell C, Jamal F, Melendez-Torres GJ, et al.Dark logic: theorising the harmful
consequences of public health interventions. J Epidemiol Community Health
2015;69:958. doi:10.1136/jech-2014-204671
20 Data Protection Act 1998. 1998. Available at:
21 Jones KH, Ford DV, Jones C, et al. A case study of the Secure Anonymous
Information Linkage (SAIL) Gateway: a privacy-protecting remote access system for
health-related research and evaluation. J Biomed Inform 2014;50:196204.
Lyons RA, et al.J Epidemiol Community Health 2016;0:17. doi:10.1136/jech-2016-207222 7
Research report
... This review focuses on personal feedback. Apart from personal feedback, there is some literature on the effects of regional air pollution or heat alerts [18][19][20][21][22][23][24]. Some alerts are bound to specific events such as smoke alerts due to wildfire [25,26]. ...
... The number of measurements or participants varied largely, ranging from one participant [39] to 38267 noise measurements [41]. Most of the studies targeted air pollution (22), while fewer studies gave feedback about noise (9), or temperatures (2). ...
... Due to the fact that urban heat islands are threatening human health and comfort [3,4], individual temperature feedback also deserves more attention in the future. While heat alerts and air pollution alerts are spread via media such as newspapers, TV or the radio [18][19][20][21][22][23][24] we see great potential for personal sensors in this field. Not only can wearable sensors provide information on the temporal and spatial distribution of pollutants; feedback from wearable sensors can also give more precise information about individual exposure patterns, which can inform behaviour change such as choosing different routes or travel times in everyday outdoor mobility. ...
Full-text available
Feedback on personal exposure to air pollution, noise or extreme temperatures through wearable sensors or sensors installed at home or in the workplace can offer information that might motivate behaviours to mitigate exposure. As personal measurement devices are becoming increasingly accessible, it is important to evaluate the effects of such sensors on human perception and behaviour. We conducted a systematic literature research and identified 33 studies, analysing the effects of personal feedback on air pollution, noise and temperatures. Feedback was given through reports including different forms of visualization, in-person or over the telephone, or directly on the sensor or through a phone app. The exposure feedback led to behaviour changes particularly for noise and temperature feedback while findings on behaviour adaptation to avoid air pollution were mixed. Most studies reported increased awareness and knowledge from receiving exposure feedback. Many participants in studies on air pollution reported low levels of self-efficacy regarding exposure mitigation. For a better understanding of the effects of personal exposure feedback, more studies are required, particularly providing feedback from wearable sensors measuring outdoor air pollution, noise and temperature.
... In terms of public interventions, most of the studies have focused on the effects of PM reduction measures (Huang et al 2018, Han et al 2020, while only a few have assessed those of adaptation measures (Chen et al 2013, Zou et al 2019. In particular, in those cases in which the alert system was the subjectfor example, updates of Air Quality Index and Air Quality Health Index in regional newspapers, websites, and smartphone applications-the effect was analyzed by comparing people's behavior and/or the respiratory health outcomes before and after its introduction (Lyons et al 2016, Chen et al 2018, Mason et al 2019. ...
... To the best of our knowledge, this study is the first to analyze the relationship between a mobile-based alert system and its immediate health outcomes on a daily basis. Compared with existing studies, the current analytical results suggest that mobile-based alert systems perform better than other adaptive policy measures, such as news, smartphone applications, websites, and outdoor electronic boards, although the spatial and temporal context of each study differs, and the generalization is limited (Lyons et al 2016, Saberian et al 2017, Mason et al 2019. A part of the larger magnitude effect we derived might be due to the directness of the system by its very nature. ...
Full-text available
To reduce human exposure to particulate matter (PM), governments have enacted various preventive measures, to which a warning system is central. To the best of our knowledge, we are the first to assess the effectiveness of mobile-based warning systems on respiratory health outcomes, examining two types of ${\text{P}}{{\text{M}}_{2.5}}$ (particles less than $2.5\,\mu {\text{m}}$ in diameter) alerts via text messaging systems: Wireless Emergency Alert (WEA) and Air Quality Information Text (AIT) as employed in South Korea from January 2015 to October 2019. We used a generalized additive model to control the non-linear relationship between the ${\text{P}}{{\text{M}}_{2.5}}$ level and the number of hospital visits and admissions for four respiratory sicknesses—chronic obstructive pulmonary disease, respiratory tract infection, asthma, and pneumonia—while deciphering how such visits and admissions are reduced by the warning systems. Our results found that both systems reduced the number of new patients with the four sicknesses at a 5% statistical significance level. Of the two, WEA was found to be more effective than AIT. The former reduced the number of new patients by 16.4%, while the latter did so by 2.8%. WEA is for everyone with a cell phone connection. By sending simple and direct alerts to a broader range of people, WEA would help people to reduce the chance of short-term exposure to PM in general. The findings provide evidence with policy implications regarding air pollution adaptation.
... According to statistics, the Belgian Meuse Valley Fog disaster of 1930 injured thousands of residents and killed more than 60 people within a week [5]; the Donora Smog tragedy in 1948 sickened nearly 6000 people in 5 days [6]; and the Great Smog of London in 1952 killed more than 4000 people in 4 days [7]. At the same time, mankind was faced with worsening problems such as food shortage, energy crisis, and environmental pollution, intensifying "ecological crisis", slowing economic growth, and raising the local social unrest [8]. Such problems have forced humanity to re-examine its position in the ecosystem and search for a new path for long-term survival and development [9]. ...
Full-text available
Sustainable development (SD) has become a fundamental strategy to guide the world’s social and economic transformation. However, in the process of practice, there are still misinterpretations in regards to the theory of SD. Such misinterpretations are highlighted in the struggle between strong and weak sustainable development paths, and the confusion of the concept of intra-generational and inter-generational justice. In this paper, the literature survey method, induction method, and normative analysis were adopted to clarify the gradual evolution and improvement process of the concept and objective of SD, to strengthen the comprehensive understanding of the SD theory. Moreover, we also tried to bring in the situation and concepts of China. The results show that the theory of SD has gone through three periods: the embryonic period (before 1972), the molding period (1972–1987), and the developing period (1987–present). SD is gradually implemented into a global action from the initial fuzzy concept, including increasing practical wisdom. The goal of SD evolves from pursuing the single goal of sustainable use of natural resources to Millennium Development Goals (MDGs) and Sustainable Development Goals (SDGs). This paper argues that the theory of strong sustainability should be the accepted concept of SD. Culture, good governance, and life support systems are important factors in promoting SD.
... A first field of intervention concerns increasing communication and awareness. Several studies [31][32][33][34][35] have tried to assess the impact of timely and personalized communication highlighting the effectiveness of paying attention and changing personal habits in case of bad air quality. To this end, international bodies as well as research groups have tried to define specific indications in relation to the different air quality levels, while also considering synthetic indicators such as air quality indices (e.g., Air Parif, https://, ...
Full-text available
Reducing children’s exposure to air pollutants should be considered a primary goal, especially for the most vulnerable subjects. The goal of this study was to test the effectiveness of applying a protocol in the event of alert days, i.e., days with forecasted PM10 levels above the EU limit value (50 µg/m3). The test was conducted, before the onset of SARS-CoV-2 restrictions, in a classroom of a primary school in Parma (Italy)—a highly polluted area in Northern Italy. The protocol included indications for the frequency of opening windows and doors, as well as the activation of an air purifier. Teachers and students were asked to apply the protocol only in the event of alert days, while no indications were provided for non-alert days. A monitoring system measuring PM1, PM2.5, PM10, CO2, and NO2 was deployed in the classroom. Measurements of the same parameters were also performed outdoors near the school. The application of the protocol reduced the indoor/outdoor (I/O) ratio for all toxic pollutants. The reduction was also remarkable for PM10—the most critical air quality parameter in the study area (1.5 and 1.1 for non-alert and alert days, respectively). Indoor concentrations of PM10—especially during non-alert days—were often higher than outdoors, showing a major contribution from resuspension due to the movement of people and personal cloud. The protocol did not cause any increase in indoor CO2 levels. Our findings showed that the application of a ventilation protocol together with the contribution of an air purifier may represent an effective way to reduce children’s exposure to air pollution during severe air pollution episodes. Considering the onset of COVID-19 and the airborne transmission of pathogens, this protocol now has more meaningful implications for children’s welfare, and can be integrated with protocols designed as measures against the spread of SARS-CoV-2.
... 38 This finding may be explained by the ozone alert, which is adopted in many countries. 45,46 Starting from 2005, this ozone warning system advises children, senior citizens, and patients with respiratory diseases to limit their outdoor activities when the ozone level is considered high (Annual O 3 Alert Trend, www.airko 47 Except ozone, the studied air pollutants are largely produced from fossil fuel combustion. ...
Full-text available
Background Kawasaki disease (KD) is a systemic vasculitis of unknown etiology that primarily affects children under 5 years of age. Some researchers suggested a potential triggering effect of air pollution on KD, but the findings are inconsistent and limited by small sample size. We investigated the association between ambient air pollution and KD among the population of South Korea younger than 5 years using the National Health Insurance claim data between 2007 and 2019. Methods and Results We obtained the data regarding particulate matter ≤10 or 2.5 µm in diameter, nitrogen dioxide, sulfur dioxide, carbon monoxide, and ozone from 235 regulatory monitoring stations. Using a time‐stratified case‐crossover design, we performed conditional logistic regression to estimate odds ratios (OR) of KD according to interquartile range increases in each air pollutant concentration on the day of fever onset after adjusting for temperature and relative humidity. We identified 51 486 children treated for KD during the study period. An interquartile range increase (14.67 μg/m ³ ) of particulate matter ≤2.5 µm was positively associated with KD at lag 1 (OR, 1.016; 95% CI, 1.004–1.029). An interquartile range increase (2.79 ppb) of sulfur dioxide concentration was associated with KD at all lag days (OR, 1.018; 95% CI, 1.002–1.034 at lag 0; OR, 1.022; 95% CI, 1.005–1.038 at lag 1; OR, 1.017; 95% CI, 1.001–1.033 at lag 2). Results were qualitatively similar in the second scenario of different fever onset, 2‐pollutant model and sensitivity analyses. Conclusions In a KD‐focused national cohort of children, exposure to particulate matter ≤2.5 µm and sulfur dioxide was positively associated with the risk of KD. This finding supports the triggering role of ambient air pollution in the development of KD.
... The availability of information for the general population is nowadays essentially limited to periods of peak pollution, while the greatest health risk is linked with chronic exposure. As it is nearly always disseminated in an isolated manner, with associated recommendations, it ends up triggering anxiety [19], and the information can hence be counterproductive [20]. During periods of poor air quality, the transmitted information is often associated with the implementation of constraints, particularly in terms of mobility: restrictions on driving and traffic. ...
Background: More than 90% of the global population lives in areas exceeding World Health Organization air quality limits. More than four million people each year are thought to die early due to air pollution, and poor air quality is thought to reduce an average European's life expectancy by one year. Individuals may be able to reduce health risks through interventions such as masks, behavioural changes and use of air quality alerts. To date, evidence is lacking about the efficacy and safety of such interventions for the general population and people with long-term respiratory conditions. This topic, and the review question relating to supporting evidence to avoid or lessen the effects of air pollution, emerged directly from a group of people with chronic obstructive pulmonary disease (COPD) in South London, UK. Objectives: 1. To assess the efficacy, safety and acceptability of individual-level interventions that aim to help people with or without chronic respiratory conditions to reduce their exposure to outdoor air pollution. 2. To assess the efficacy, safety and acceptability of individual-level interventions that aim to help people with chronic respiratory conditions reduce the personal impact of outdoor air pollution and improve health outcomes. Search methods: We identified studies from the Cochrane Airways Trials Register, Cochrane Central Register of Controlled Trials, and other major databases. We did not restrict our searches by date, language or publication type and included a search of the grey literature (e.g. unpublished information). We conducted the most recent search on 16 October 2020. Selection criteria: We included randomised controlled trials (RCTs) and non-randomised studies (NRS) that included a comparison treatment arm, in adults and children that investigated the effectiveness of an individual-level intervention to reduce risks of outdoor air pollution. We included studies in healthy individuals and those in people with long-term respiratory conditions. We excluded studies which focused on non-respiratory long-term conditions, such as cardiovascular disease. We did not restrict eligibility of studies based on outcomes. Data collection and analysis: We used standard Cochrane methods. Two review authors independently selected trials for inclusion, extracted study characteristics and outcome data, and assessed risk of bias using the Cochrane Risk of Bias tool for RCTs and the Risk Of Bias In Non-randomised Studies - of Interventions (ROBINS-I) as appropriate. One review author entered data into the review; this was spot-checked by a second author. We planned to meta-analyse results from RCTs and NRS separately, using a random-effects model. This was not possible, so we presented evidence narratively. We assessed certainty of the evidence using the GRADE approach. Primary outcomes were: measures of air pollution exposure; exacerbation of respiratory conditions; hospital admissions; quality of life; and serious adverse events. Main results: We identified 11 studies (3372 participants) meeting our inclusion criteria (10 RCTs and one NRS). Participants' ages ranged from 18 to 74 years, and the duration of studies ranged from 24 hours to 104 weeks. Six cross-over studies recruited healthy adults and five parallel studies included either people with pre-existing conditions (three studies) or only pregnant women (two studies). Interventions included masks (e.g. an N95 mask designed to filter out airborne particles) (five studies), an alternative cycle route (one study), air quality alerts and education (five studies). Studies were set in Australia, China, Iran, the UK, and the USA. Due to the diversity of study designs, populations, interventions and outcomes, we did not perform any meta-analyses and instead summarised results narratively. We judged both RCTs and the NRS to be at risk of bias from lack of blinding and lack of clarity regarding selection methods. Many studies did not provide a prepublished protocol or trial registration. From five studies (184 participants), we found that masks or altered cycle routes may have little or no impact on physiological markers of air pollution exposure (e.g. blood pressure and heart rate variability), but we are very uncertain about this estimate using the GRADE approach. We found conflicting evidence regarding health care usage from three studies of air pollution alerts, with one non-randomised cross-over trial (35 participants) reporting an increase in emergency hospital attendances and admissions, but the other two randomised parallel trials (1553 participants) reporting little to no difference. We also gave the evidence for this outcome a very uncertain GRADE rating. None of our included trials reported respiratory exacerbations, quality of life or serious adverse events. Secondary outcomes were not well reported, but indicated inconsistent impacts of air quality alerts and education interventions on adherence, with some trials reporting improvements in the intervention groups and others reporting little or no difference. Symptoms were reported by three trials, with one randomised cross-over trial (15 participants) reporting a small increase in breathing difficulties associated with the mask intervention, one non-randomised cross-over trial (35 participants) reporting reduced throat and nasal irritation in the lower-pollution cycle route group (but no clear difference in other respiratory symptoms), and another randomised parallel trial (519 participants) reporting no clear difference in symptoms between those who received a smog warning and those who did not. Authors' conclusions: The lack of evidence and study diversity has limited the conclusions of this review. Using a mask or a lower-pollution cycle route may mitigate some of the physiological impacts from air pollution, but evidence was very uncertain. We found conflicting results for other outcomes, including health care usage, symptoms and adherence/behaviour change. We did not find evidence for adverse events. Funders should consider commissioning larger, longer studies, using high-quality and well-described methods, recruiting participants with pre-existing respiratory conditions. Studies should report outcomes of importance to people with respiratory conditions, such as exacerbations, hospital admissions, quality of life and adverse events.
Exposure to air pollutants may be associated with preterm birth (PB) through oxidative stress, metabolic detoxification, and immune system processes. However, no study has investigated the interactive effects of maternal air pollution and genetic polymorphisms in these pathways on risk of PB. The study included 126 PB and 310 term births. A total of 177 single nucleotide polymorphisms (SNPs) in oxidative stress, immune function, and metabolic detoxification-related genes were examined and analyzed. The China air quality index (AQI) was used as an overall estimation of ambient air pollutants. Among 177 SNPs, four SNPs (GPX4-rs376102, GLRX-rs889224, VEGFA-rs3025039, and IL1A-rs3783550) were found to have significant interactions with AQI on the risk of PB (Pinteraction were 0.001, 0.003, 0.03, and 0.04, respectively). After being stratified by the maternal genotypes in these four SNPs, 1.38 to 1.76 times of the risk of PB were observed as per interquartile range increase in maternal AQI among women who carried the GPX4-rs376102 AC/CC genotypes, the GLRX-rs889224 TT genotype, the VEGFA-rs3025039 CC genotype, or the IL1A-rs3783550 GT/TT genotypes. After adjustment for multiple comparisons, only GPX4-rs376102 and AQI interaction remained statistically significant (false discovery rate (FDR)=0.17). After additional stratification by preeclampsia (PE) status, a strongest association was observed in women who carried the GPX4-rs376102 AC/CC genotypes (OR, 2.26; 95% CI, 1.41-3.65, Pinteraction=0.0002, FDR=0.035) in the PE group. Our study provided the first evidence that association between maternal air pollution and PB risk may be modified by the genetic polymorphisms in oxidative stress and immune function genes. Future large studies are necessary to replicate and confirm the observed associations.
The Junction-less tunnel field-effect transistor (JLTFET) is a captivating device due to its excellent electrical properties and less variability in comparison to MOSFET at the nanometer regime. In this regard, we investigate a silicon-based pocket doped JLTFET in which an InAs pocket is inserted across the source-channel junction to enhance tunneling probability. In this respect, we have considered analog/RF and DC Figure of merit analysis for the conventional and pocket doped JLTFET (PD- JLTFET) in terms of an electric field, transfer characteristics, transconductance, parasitic capacitances, cut-off frequency, gain-bandwidth product, and maximum oscillation frequency. Additionally, we have examined the effect of spacer length variation across the junction between source and channel. The ATLAS device simulator is used for the simulations of the conventional JLTFET and PD-JLTFET. The proposed PD-JLTFET has shown a higher ION/IOFF ratio (~1013) and improved sub-threshold swing (~9.08 mV/decade). The notable characteristics demonstrated by PD-JLTFET make it an optimum device for low power and high switching application.
Full-text available
With the current expansion of data linkage research, the challenge is to find the balance between preserving the privacy of person-level data whilst making these data accessible for use to their full potential. We describe a privacy-protecting safe haven and secure remote access system, referred to as the Secure Anonymised Information Linkage (SAIL) Gateway. The Gateway provides data users with a familiar Windows interface and their usual toolsets to access approved anonymously-linked datasets for research and evaluation. We outline the principles and operating model of the Gateway, the features provided to users within the secure environment, and how we are approaching the challenges of making data safely accessible to increasing numbers of research users. The Gateway represents a powerful analytical environment and has been designed to be scalable and adaptable to meet the needs of the rapidly growing data linkage community.
Full-text available
Background Vast quantities of electronic data are collected about patients and service users as they pass through health service and other public sector organisations, and these data present enormous potential for research and policy evaluation. The Health Information Research Unit (HIRU) aims to realise the potential of electronically-held, person-based, routinely-collected data to conduct and support health-related studies. However, there are considerable challenges that must be addressed before such data can be used for these purposes, to ensure compliance with the legislation and guidelines generally known as Information Governance. Methods A set of objectives was identified to address the challenges and establish the Secure Anonymised Information Linkage (SAIL) system in accordance with Information Governance. These were to: 1) ensure data transportation is secure; 2) operate a reliable record matching technique to enable accurate record linkage across datasets; 3) anonymise and encrypt the data to prevent re-identification of individuals; 4) apply measures to address disclosure risk in data views created for researchers; 5) ensure data access is controlled and authorised; 6) establish methods for scrutinising proposals for data utilisation and approving output; and 7) gain external verification of compliance with Information Governance. Results The SAIL databank has been established and it operates on a DB2 platform (Data Warehouse Edition on AIX) running on an IBM 'P' series Supercomputer: Blue-C. The findings of an independent internal audit were favourable and concluded that the systems in place provide adequate assurance of compliance with Information Governance. This expanding databank already holds over 500 million anonymised and encrypted individual-level records from a range of sources relevant to health and well-being. This includes national datasets covering the whole of Wales (approximately 3 million population) and local provider-level datasets, with further growth in progress. The utility of the databank is demonstrated by increasing engagement in high quality research studies. Conclusion Through the pragmatic approach that has been adopted, we have been able to address the key challenges in establishing a national databank of anonymised person-based records, so that the data are available for research and evaluation whilst meeting the requirements of Information Governance.
Full-text available
Background Vast amounts of data are collected about patients and service users in the course of health and social care service delivery. Electronic data systems for patient records have the potential to revolutionise service delivery and research. But in order to achieve this, it is essential that the ability to link the data at the individual record level be retained whilst adhering to the principles of information governance. The SAIL (Secure Anonymised Information Linkage) databank has been established using disparate datasets, and over 500 million records from multiple health and social care service providers have been loaded to date, with further growth in progress. Methods Having established the infrastructure of the databank, the aim of this work was to develop and implement an accurate matching process to enable the assignment of a unique Anonymous Linking Field (ALF) to person-based records to make the databank ready for record-linkage research studies. An SQL-based matching algorithm (MACRAL, Matching Algorithm for Consistent Results in Anonymised Linkage) was developed for this purpose. Firstly the suitability of using a valid NHS number as the basis of a unique identifier was assessed using MACRAL. Secondly, MACRAL was applied in turn to match primary care, secondary care and social services datasets to the NHS Administrative Register (NHSAR), to assess the efficacy of this process, and the optimum matching technique. Results The validation of using the NHS number yielded specificity values > 99.8% and sensitivity values > 94.6% using probabilistic record linkage (PRL) at the 50% threshold, and error rates were < 0.2%. A range of techniques for matching datasets to the NHSAR were applied and the optimum technique resulted in sensitivity values of: 99.9% for a GP dataset from primary care, 99.3% for a PEDW dataset from secondary care and 95.2% for the PARIS database from social care. Conclusion With the infrastructure that has been put in place, the reliable matching process that has been developed enables an ALF to be consistently allocated to records in the databank. The SAIL databank represents a research-ready platform for record-linkage studies.
Although it might be assumed that most public health programmes involving social or behavioural rather than clinical interventions are unlikely to be iatrogenic, it is well established that they can sometimes cause serious harms. However, the assessment of adverse effects remains a neglected topic in evaluations of public health interventions. In this paper, we first argue for the importance of evaluations of public health interventions not only aiming to examine potential harms but also the mechanisms that might underlie these harms so that they might be avoided in the future. Second, we examine empirically whether protocols for the evaluation of public health interventions do examine harmful outcomes and underlying mechanisms and, if so, how. Third, we suggest a new process by which evaluators might develop 'dark logic models' to guide the evaluation of potential harms and underlying mechanisms, which includes: theorisation of agency-structure interactions; building comparative understanding across similar interventions via reciprocal and refutational translation; and consultation with local actors to identify how mechanisms might be derailed, leading to harmful consequences. We refer to the evaluation of a youth work intervention which unexpectedly appeared to increase the rate of teenage pregnancy it was aiming to reduce, and apply our proposed process retrospectively to see how this might have strengthened the evaluation. We conclude that the theorisation of dark logic models is critical to prevent replication of harms. It is not intended to replace but rather to inform empirical evaluation. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to
Acute exposure to air pollution has been linked to myocardial infarction, but its effect on heart failure is uncertain. We did a systematic review and meta-analysis to assess the association between air pollution and acute decompensated heart failure including hospitalisation and heart failure mortality. Five databases were searched for studies investigating the association between daily increases in gaseous (carbon monoxide, sulphur dioxide, nitrogen dioxide, ozone) and particulate (diameter <2·5 μm [PM2·5] or <10 μm [PM10]) air pollutants, and heart failure hospitalisations or heart failure mortality. We used a random-effects model to derive overall risk estimates per pollutant. Of 1146 identified articles, 195 were reviewed in-depth with 35 satisfying inclusion criteria. Heart failure hospitalisation or death was associated with increases in carbon monoxide (3·52% per 1 part per million; 95% CI 2·52-4·54), sulphur dioxide (2·36% per 10 parts per billion; 1·35-3·38), and nitrogen dioxide (1·70% per 10 parts per billion; 1·25-2·16), but not ozone (0·46% per 10 parts per billion; -0·10 to 1·02) concentrations. Increases in particulate matter concentration were associated with heart failure hospitalisation or death (PM2·5 2·12% per 10 μg/m(3), 95% CI 1·42-2·82; PM10 1·63% per 10 μg/m(3), 95% CI 1·20-2·07). Strongest associations were seen on the day of exposure, with more persistent effects for PM2·5. In the USA, we estimate that a mean reduction in PM2·5 of 3·9 μg/m(3) would prevent 7978 heart failure hospitalisations and save a third of a billion US dollars a year. Air pollution has a close temporal association with heart failure hospitalisation and heart failure mortality. Although more studies from developing nations are required, air pollution is a pervasive public health issue with major cardiovascular and health economic consequences, and it should remain a key target for global health policy. British Heart Foundation.
Research confirming the detrimental impact poor ambient air quality and episodes of abnormally high pollutants has on public health, plus differential susceptibility, calls for improved understanding of this complex topic among all walks of society. The public and particularly, vulnerable groups, should be aware of their quality of air, enabling action to be taken in the event of increased pollution. Policy makers must have a sound awareness of current air quality and future trends, to identify issues, guide policies and monitor their effectiveness. These attitudes are dependent upon air pollution monitoring, forecasting and reporting, serving all interested parties. Apart from the underlying national regulatory obligation a country has in reporting air quality information, data output serves several purposes. This review focuses on provision of real-time data and advanced warnings of potentially health-damaging events, in the form of national air quality indices and proactive alert services. Some of the challenges associated with designing these systems include technical issues associated with the complexity of air pollution and its science. These include inability to provide precise exposure concentrations or guidance on long-term/cumulative exposures or effects from pollutant combinations. Other issues relate to the degree to which people are aware and positively respond to these services. Looking to the future, mobile devices such as cellular phones, equipped with sensing applications have potential to provide dynamic, temporally and spatially precise exposure measures for the mass population. The ultimate aim should be to empower people to modify behaviour-for example, when to increase medication, the route/mode of transport taken to school or work or the appropriate time to pursue outdoor activities-in a way that protects their health as well as the quality of the air they breathe.