Transcript Document

Spreadsheets In Clinical
Medicine
Grenville J. Croll & Raymond J. Butler
EuSpRIG 2006, Cambridge, UK
What We Know about Spreadsheets








Spreadsheets are error-prone
Spreadsheets are not tested
Errors are difficult to find
Users are overconfident
Users are inconsistent
Users interpret information differently
Backup and Archiving are overlooked
Spreadsheets are pervasive
What We Know about Medicine

Medical error is a major killer
98,000 deaths per year in US
30,000 in UK
7th largest cause of death
Rate of non-fatal injuries is not known

Spreadsheets are used in Clinical Medicine




Spreadsheets in Clinical Medicine

Initial Google Searches
Spreadsheet + Oncology, Anesthesia, Cardiac etc
 very many hits


Search of PubMed Database



>800 references, many clinical
Discovered www.medal.org - Institute of
Algorithmic Medicine - 8,000 spreadsheets
Numerous specialist sites
Spreadsheets in Clinical Medicine






Widely used for decision support
Applied in diagnosis, dosage, prescription,
case management, classification, statistical
planning & analysis
Home-grown by practitioners
Used by medical device manufacturers
Described in books & papers for re-keying
Many downloadable, modifiable templates
Controls on Software Development

Drug development & trials


Onerous FDA software development
requirements
Medical Devices
Medical Device Regulations
 Apply to Medical Devices intended for sale


Clinical Medicine
No regulation found
 Medics free to develop & deploy for own use

Some Initial testing

Investigated the following models
Pediatric Anesthesia
 Apache II – Risk of Cardiac Problems in patients
undergoing non-cardiac surgery
 Brodaty – to support a decision to assess an elderly
patient for masked depression


Investigation restricted to spreadsheet
mechanics only as we are not medically
qualified


But 15 pages of notes
Used HMRC SpACE methodology
Pediatric Anesthesia
“…. here are two guides to help you in your
anesthesia practice. First is a Microsoft Excel
spreadsheet titled ‘Pediatric Anesthesia
Worksheet’. Use it to calculate medications
and other parameters for pediatric patients”.
Pediatric Anesthesia
Pediatric Anesthesia
Apache II - Chronic Health
Apache II - Chronic Health
Cardiac Risks
Problems Found







Constants in formulas
Complex nested IF’s
Formulas with no dependents
Poor or no embedded documentation
Little/No data validation
Some Mixed units
Weak cell protection
Embedded constants




Same risk as in finance/tax etc domains
What happens if medical practice changes?
What happens if manufacturers reformulate
drugs to require different dosages?
Table Lookups preferable
Nested IF’s

e.g. Atropine dose calculation (L7):
=IF(E19*0.02>0.6,0.6,IF(E19*0.02<0.1,0.1,E19*0
.02))
 E19 is “Bodyweight”
 The rest are embedded constants for drug
dosage




Hard to follow
Hard to update
Table Lookup preferable
Other Issues

Documentation


None/almost none embedded
Data Validation
None or inadequate
 Forms tools, data validation essential


Units
Milligrams / Micrograms appear in same
column
 NB Medication errors cause 7,000 annual deaths
in US

Pediatric Anesthesia Documentation
Inherent Risks

Distribution channels
Large scale web enabled distribution
 via Global Medical Literature





Secondary “viral distribution”
Inability to recall source
Spurious “stamp of approval”
Spreadsheet may be changed somewhere in
the chain
Conclusions





Initial indications are that same risks exist
No specific clinical good practice for
spreadsheet development & deployment
No evidence that this well studied problem
is known in the clinical domain
Opportunity for dialogue and advancement
Obvious reasons for increased caution when
one is involved in medical procedures!
Disclaimers
“The authors have exerted every effort to
ensure that the drug dosages set forth are in
accordance with current recommendations
at the time of publication. The user is
urged to check the drug's package insert for
any changes in indications and dosages as
well as for warnings and precautions. The
responsibility is ultimately that of the
prescribing clinician”.
“Delegation of responsibility is no
barrier to the repeated perpetration
of grave errors”
G.J.Croll & R.J. Butler, July 2006