figures/calculations and conclusions

still valid). Meanwhile, we would

advise database researchers han-

dling datasets in excess of 65 535

rows of patient records to use later

versions of Microsoft Excel, and to

confirm their denominator data

through secondary, independent

analysis.

S. M. WhiteRoyal Sussex County HospitalBrighton, UKEmail: stuart.white@bsuh.nhs.ukI. K. MoppettUniversity of Nottingham, Queen’sMedical Centre CampusNottingham University HospitalsNHS TrustNottingham, UKR. GriffithsPeterborough and StamfordHospitals NHS TrustPeterborough, UK

Competing interests as given in refer-

ence [2]. We are grateful to Mr Tim

Bunning, Principal Consultant,

Crown Informatics Limited, for alert-

ing us to the possible error in the

numbers. Previously posted on the

Anaesthesia correspondence website:

www.anaesthesiacorrespondence.com

References1. Sessler DI. Big Data – and its contribu-

tions to peri-operative medicine. Anaes-thesia 2014; 69: 100–5.

2. White SM, Moppett IK, Griffiths R. Out-come by mode of anaesthesia for hipfracture surgery. An observational auditof 65 535 patients in a national dataset.Anaesthesia 2014; 69: 224–30.

3. Microsoft Excel specifications and limits.http://office.microsoft.com/en-gb/excel-help/excel-specifications-and-limits-HP005199291.aspx (accessed 06/02/2014).

4. Correction. Anaesthesia 2014; 69.

doi:10.1111/anae.12652

Big Data – ethicalperspectives

Large, linked databases (‘Big Data’

[1]) use pseudonymised personal

confidential data that have been

anonymised but retain a residual

risk of re-identification. Sessler out-

lines the potential benefits of these

databases in his editorial [1], with

regard to rapid improvements in

the quality and affordability of

healthcare. However, there are ethi-

cal aspects that must be considered

when balancing potential benefits

against potential harms.

Non-maleficence describes ‘an

obligation not to inflict harm on

others’ [2]. The potential for harm

in epidemiological database research

is not as immediately apparent as

for clinical trials, but individuals

can still be wronged when they

have not suffered direct harm per

se. There are two main ethical areas

of concern: firstly, that vulnerable

groups may become stigmatised;

and secondly that people are ‘trea-

ted’ as a means to an end, rather

than as an end in themselves.

Although the aim of collecting

data about vulnerable groups is to

improve care, information could be

used to discriminate against them, or

stigmatise them. Retained identifiers

such as postcodes can provide infor-

mation about disease burden linked

to geographical socio-economic

deprivation. In rare conditions, the

possibility of re-identification

becomes very real. Boyd [3] uses the

example of an ethnic minority with

an inheritable condition deemed

potentially harmful to the majority

population, who are coerced into

birth control. Public concerns often

centre around inadvertent disclosure

of information to third parties,

including employers, insurance com-

panies [4], drug companies or crimi-

nal organisations.

Personal autonomy is recogni-

sed and protected through the pro-

cess of providing consent, such that

“patients must be fully informed of,

and agree without coercion to, any

participation in medical research”

[5]. The Data Protection Act 1998

states that consent is not needed for

anonymised data, and it has been

argued that seeking formal consent

for data inclusion within health da-

tabases could cause harm. The

Academy of Medical Sciences has

suggested that any insistence on

formal consent could cause such

significant selection, recruitment

and participation bias that it would

skew the data to such an extent as

to make it not worth collecting [6].

It would also use precious resources

such as time and money, and pro-

hibit large studies. In addition to

this, a traditional informed consent

process, where a researcher asks for

permission to do something to a

participant at a single point in time,

is likely to be unfit for purpose –

the initial researcher is unlikely to

know the nature of research to be

carried out in the future, and future

research would require re-consent,

and therefore the necessity for

patient contact identifiers to be held

simultaneously on any database.

Non-consensual use even of

pseudonymised data risks damaging

trust between patients, healthcare

professionals and government

bodies. NHS England has proposed

390 © 2014 The Association of Anaesthetists of Great Britain and Ireland

Anaesthesia 2014, 69, 387–398 Correspondence

an ‘opt out’ option, whereby all

patients using NHS services are

provided with information about

how their data will be used for

health services research and their

consent assumed unless they specify

their refusal [7]. However, public

awareness and engagement are

essential, and the recent decision by

NHS England to delay introduction

of their ‘care.data’ program indi-

cates that there is considerable work

still to be done in this area [8].

The aims of ‘Big Data’ are

impressive, and the potential benefit

is significant. Reversible anonymisa-

tion in a secure environment, sup-

ported by sanctions and regulations,

helps to protect individual privacy.

Coupled with ongoing engagement

with the public and ultimately posi-

tive public approval and trust, this

maintains individual autonomy. By

training researchers and governors

to high ethical standards, and foster-

ing the virtuous character traits,

maleficent breaches of confidentiality

should be minimised. Finally, by em-

phasising the principles of solidarity,

altruism and unselfishness, with an

acceptance that there will be some

trade-offs in the pursuit of these

goals, we should seek to use the

opportunities that ‘Big Data’ offers

to research and clinical practice.

A. DochertyRoyal Infirmary of EdinburghEdinburgh, UKEmail: abdocherty79@gmail.com

No external funding and no com-

peting interests declared. Previously

posted on the Anaesthesia corre-

spondence website: www.anaesthesia

correspondence.com.

References1. Sessler DI. Big Data – and its contribu-

tions to peri-operative medicine. Anaes-thesia 2014; 69: 100–5.

2. Beauchamp TL, Childress JF. Principles ofBiomedical Ethics, 6th edn. New York:Oxford University Press, 2008.

3. Boyd KM. Ethnicity and the ethics ofdata linkage. BMC Public Health 2007;7: 318.

4. Scottish Government. Public acceptabil-ity of Data Sharing Between the Public,Private and Third Sectors for ResearchPurposes, 2013. http://www.scotland.gov.uk/Publications/2013/10/1304/8(accessed 20/02/2014).

5. Cassell J, Young A. Why we should not seekindividual informed consent for participa-tion in health services research. Journal ofMedical Ethics 2002; 28: 313–7.

6. Academy of Medical Sciences. Personaldata for public good: using health infor-mation in medical research, 2002.http://issuu.com/acmedsci/docs/per-sonal (accessed 20/02/2014).

7. NHS England. Better information meansbetter care. www.england.nhs.uk/our-work/tsd/care-data/ (accessed 20/02/2014).

8. Triggle N. Care.data: How did it go sowrong? 19 Feb 2014. www.bbc.co.uk/news/health-26259101 (accessed 20/02/2014).

doi:10.1111/anae.12656

Co-amoxiclavdiscolouration

Further to the experience of Kears-

ley et al. [1], whilst reconstituting

1.2 g amoxicillin sodium/clavulanic

acid (co-amoxiclav, Wockhardt UK

Ltd, Wrexham, UK) in water (Fan-

nin, Leopardstown, Dublin), I noted

an unusually dark colour of the

resultant solution (Fig. 1). Close

inspection of the co-amoxiclav vial

revealed a crack in the glass around

the neck. The solution was dis-

carded and not administered.

Co-amoxiclav changes colour

slowly after reconstitution, due to

changes in the pH of the solution

[2]. Presumably, oxidation of the vial

contents altered their pH, resulting

in early discolouration. After recon-

stitution in 20 ml water, the manu-

facturer recommends intravenous

injection within 20 minutes [3].

This incident serves to remind

anaesthetists to check for defective

drugs packaging, which could result

in chemical/physical degradation

and subsequently ineffective drug

action, or in potential microbial

contamination of that drug.

J. S. DawsonNottingham University HospitalsNHS TrustNottingham, UKEmail: james.dawson@nuh.nhs.uk

Figure 1 Discoloured co-amoxiclav (top) drawn up from cracked vial, com-pared with normal co-amoxiclav (bottom).

© 2014 The Association of Anaesthetists of Great Britain and Ireland 391

Correspondence Anaesthesia 2014, 69, 387–398