Rule-based microbiology
reporting
Raymond Lin, MDNational University Hospital
Singapore
Objectives
� For medical microbiologists (pathologists), lab directors and supervisors� Enhancing clinical and operational benefit� Discuss some clinical applications and issues
� For IT managers, lab managers, LIS developers� Appreciate the scope and function of clinical
microbiology; what IT solutions are required� For all
� General implementation issues� Factors influencing choice of LIS
� For medical microbiologists (pathologists), lab directors and supervisors� Enhancing clinical and operational benefit� Discuss some clinical applications and issues
� For IT managers, lab managers, LIS developers� Appreciate the scope and function of clinical
microbiology; what IT solutions are required� For all
� General implementation issues� Factors influencing choice of LIS
Why I am passionate about deploying rule-based reporting in microbiology
Rule-based reporting as a central feature of a microbiology laboratory
My experience
Life before and after rules
Experiences with various LIS
Before LIS
Manually transcribe results onto request form
Write comments; write rules for MTs to follow
Log into record book and flip pages to traceEarly LIS (translate practice)
Key results into computer
Maintain paper worksheet
Can take up more time than before!Modern LIS (transform practice)
All work captured on-line; trackable
Interfaces; rules
Increased efficiency, more info, fewer errors
Practice of clinical microbiology
introduction
�Hospital and healthcare cluster�Microbiology department� IT solution and architecture
�Hospital and healthcare cluster�Microbiology department� IT solution and architecture
The National University Hospital (NUH), Singapore
�900 bed referral hospital with all medical and surgical specialties
�Part of National Healthcare Group (NHG) cluster of institutions� National University Hospital� Tan Tock Seng Hospital� Alexandra Hospital� Institute of Mental Health� Polyclinics
�900 bed referral hospital with all medical and surgical specialties
�Part of National Healthcare Group (NHG) cluster of institutions� National University Hospital� Tan Tock Seng Hospital� Alexandra Hospital� Institute of Mental Health� Polyclinics
The Microbiology Department
�Bacteriology�Automated blood culture system� Identification and susceptibility testing
�Mostly using Vitek2 system�Backed by manual systems, API, disc
diffusion, E test�Virus IF, isolation�Parasitology, Mycology, Serology�Mycobacteriology�Molecular tests (PCR) in molecular lab
�Bacteriology�Automated blood culture system� Identification and susceptibility testing
�Mostly using Vitek2 system�Backed by manual systems, API, disc
diffusion, E test�Virus IF, isolation�Parasitology, Mycology, Serology�Mycobacteriology�Molecular tests (PCR) in molecular lab
The microbiology department
�2 medical microbiologists (MD)�Scientific officer�20 medical technologists�Reception; attendants
�2 medical microbiologists (MD)�Scientific officer�20 medical technologists�Reception; attendants
Roles of staff�MDs
�Plan and approve test approach, interpretation and report format
�Sign off significant reports (clinical validation)�Liaise with attending physicians: advice on
treatment, specimens and tests� Infection control
�MTs�Perform tests (technical validation); QC�Sign off negative and “simple” results
�MDs�Plan and approve test approach, interpretation
and report format�Sign off significant reports (clinical validation)�Liaise with attending physicians: advice on
treatment, specimens and tests� Infection control
�MTs�Perform tests (technical validation); QC�Sign off negative and “simple” results
Objectives of microbiology service
�Rapid and reliable results�Meaningful interpretation�Additional comments to help in
interpretation and treatment
�Rapid and reliable results�Meaningful interpretation�Additional comments to help in
interpretation and treatment
IT in the NHG cluster
�Common IT infrastructure and managed by central IT team
�Laboratory has its own LIS team
�Common IT infrastructure and managed by central IT team
�Laboratory has its own LIS team
NHG Cluster LIS Implementation
Overview
NHG LIS Cluster Configurations
• Comprises 4 NHG Institutions and 9 Polyclinics� Alexandra Hospital� National University Hospital� Tan Tock Seng Hospital� Institute of Mental Health� NHG Diagnostics
• Information Flow� Between Institutions via LIS to LIS connection� Connectivity with External Systems
via Socket connection e.g. Nauticus SAP (Billing & ADT), EMRs.
NHG LIS Cluster ConfigurationsLIS to LIS connection (L2L)
• 5 instances of Database for each of 5 Institutions
• Each Institution receives ADT, transmits results to EMR, transmits billing to Nauticus SAP and manages external billing separately, but workflow is harmonized for all Laboratories and standardizedacross cluster.
CLOVERLEAF (ADT)
CLOVERLEAF (EMR)
TTSH NHGDAH NUH IMHL2L L2L L2L L2L
Physical Setup – Server Diagram
• Primary Data Centre (PDC) located at Bedok• Secondary Date Centre (SDC) located at Jurong
Storage Area Network Storage Area NetworkOracle Oracle
The nature of microbiology testing
�Many qualitative observations�Many ways to report the same thing� Interpretation and advice
�Both verbal and on the report� Interpretation dependent on
�Site, organism, antibiotic, underlying condition
�Many qualitative observations�Many ways to report the same thing� Interpretation and advice
�Both verbal and on the report� Interpretation dependent on
�Site, organism, antibiotic, underlying condition
�Flexible and based on judgment�Every 2 microbiologists will have a
different opinion on best way to interpret and report
�May have a series of interim reports spread over several days
�Serve related functions� Infection control & epidemiology�Mandatory reporting
�Flexible and based on judgment�Every 2 microbiologists will have a
different opinion on best way to interpret and report
�May have a series of interim reports spread over several days
�Serve related functions� Infection control & epidemiology�Mandatory reporting
The nature of microbiology testing
Therefore automated interpretation must be
Flexible, customized to patient and organism parameters, yet consistent
Allow for exceptions - manual review and comment
The nature of microbiology testing
But microbiologists traditionally
less machine-savvy
less accustomed to intense focus on efficient process (compared to core lab)
less familiar with idea of IT-assisted work processes
And yet …
Microbio MTs enter the highest number of key strokes per specimen
The nature of microbiology testing
MTs on the keyboard
Requisition (order entry): site, specimen description, antibiotics, added requests
On-line worksheet: media, kits, biochems, observations
Toggle screens for microscopy, identification, antibiotic susceptibility, comment box
Auto-validation possible only for limited types (e.g. negative blood cultures)
Series of interim reportsReduce key strokes = efficiency
Rules help achieve it
Types of rules in microbio LIS
�Operational rules�Tag nosocomial pathogens�Tag mandatory pathogens for reporting�Automated ordering of media�Auto-negatives (blood culture)�Duplicate specimens�Organism-matched antibiotic panel
�Operational rules�Tag nosocomial pathogens�Tag mandatory pathogens for reporting�Automated ordering of media�Auto-negatives (blood culture)�Duplicate specimens�Organism-matched antibiotic panel
Types of rules in microbio LIS
�Expert rules�Reflect professional practice�Some follow guidelines in CLSI document�Most devised to suit local practice�Additional interpretation not contained in CLSI
document�Reflect professional knowledge, opinion and
practice
�Expert rules�Reflect professional practice�Some follow guidelines in CLSI document�Most devised to suit local practice�Additional interpretation not contained in CLSI
document�Reflect professional knowledge, opinion and
practice
Objectives of expert rules in microbiology
� Increase efficiency�Consistency in reporting�Reduce error�Opportunity for more clinically useful
guidance
� Increase efficiency�Consistency in reporting�Reduce error�Opportunity for more clinically useful
guidance
Design of expert rules
�Who?�MDs must be involved�Supervisor must be engaged and excited�MT : choose a team and point person�LIS team : advantageous to know some
microbiology
�Who?�MDs must be involved�Supervisor must be engaged and excited�MT : choose a team and point person�LIS team : advantageous to know some
microbiology
Caution: if LIS staff left to “implement”system, he/she will just use existing practice to translate into new system; lack of MD/ supervisor involvement = unable to take advantage of potential of LIS to transform practice
LIS is more than just an electronic version of traditional paper reports
A good LIS will introduce a whole new way of doing things
A large part of the improvement depends on the ability to set up rules
Design of expert rules
A good LIS - expert rules must be quick and easy to set up and monitor
It should be accessible for an MT to maintain without having to go through extensive training
Design of expert rules
Design of expert rules
�What is asked of MDs & supervisors�Must realize potential of rules�Focused on clinical impact - know what you
want�Understand capabilities of LIS�Vision to achieve more with rules�Understand logic in system: if … then …else�Be alert to pitfalls and exceptions�Some good habits in designing rules
�What is asked of MDs & supervisors�Must realize potential of rules�Focused on clinical impact - know what you
want�Understand capabilities of LIS�Vision to achieve more with rules�Understand logic in system: if … then …else�Be alert to pitfalls and exceptions�Some good habits in designing rules
Design of expert rules
�MT and LIS team�Translate instructions from supervisor to
laboratory rules�Explain capabilities and limitations�Make rules simple and avoid conflict�Maintain database of rules�Assign codes - should be easy to maintain
�MT and LIS team�Translate instructions from supervisor to
laboratory rules�Explain capabilities and limitations�Make rules simple and avoid conflict�Maintain database of rules�Assign codes - should be easy to maintain
Design of expert rules
�Coded text should have implicit logic�E.g. STano2, STspn (better than ano2ST,
spnST)�Very hard to undo system of naming
�Coded text should have implicit logic�E.g. STano2, STspn (better than ano2ST,
spnST)�Very hard to undo system of naming
Design of expert rules
� IF … THEN … ELSE�Designing platform should be accessible,
easy to use
� IF … THEN … ELSE�Designing platform should be accessible,
easy to use
Testing of expert rules
�Test environment vs live environment�Conflicts not always predictable�Should be able to troubleshoot readily�Understand hierarchy of rules
�Some LIS have a priority listing�Should monitor post-implementation
�Test environment vs live environment�Conflicts not always predictable�Should be able to troubleshoot readily�Understand hierarchy of rules
�Some LIS have a priority listing�Should monitor post-implementation
Maintenance of expert rules
�Tabulated, document-controlled�Easy for alternative staff to take over�Control at MT or LIS team level
�Tabulated, document-controlled�Easy for alternative staff to take over�Control at MT or LIS team level
Adding new expert rules
�Approval and documentation process�Testing each time�Should also test “downstream” system if
LIS interfaced to a separate EMR for viewing
�Approval and documentation process�Testing each time�Should also test “downstream” system if
LIS interfaced to a separate EMR for viewing
Types of expert rules
�Antibiotics�CAR (conditional antibiotic reporting)�Suppress (non-printable), conditional�Consolidate (whether applicable)�Additional comments
�Suggestion for treatment�Statistics�References�Qualifiers
�Antibiotics�CAR (conditional antibiotic reporting)�Suppress (non-printable), conditional�Consolidate (whether applicable)�Additional comments
�Suggestion for treatment�Statistics�References�Qualifiers
Types of expert rules
�Organism and site�Comments on significance�Release of appropriate antibiotic results
�Organism and site�Comments on significance�Release of appropriate antibiotic results
Rules for LIS (Microbiology)
IF THENSource Organism Antibiotic Result= MIC= Antibiotic Suppressed Result StatusAll All gentamicin S amikacin Y
All Staphylococcus aureus cloxacillin Svancomycin, ciprofloxacin Y
Urine All
erythromycin, clindamycin, tetracycline, chloramphenicol Y
Not urine Allnitrofurantoin, nalidixic acid Y
All
Enterococcus sp., Enterococcus faecium, Enterococcus faecalis
erythromycin, clindamycin, tetracycline, chloramphenicol, ciprofloxacin Y
All
Aeromonas sp., Aeromonas hydrophila, Aeromonas sobria, Aeromonas caviae
imipenem, meropenem, ceftriaxone, ceftazidime, aztreonam, cefepime Y
Burkholderia pseudomallei ceftriaxone R
Pseudomonas aeruginosa ceftazidime S
imipenem, meropenem, aztreonam Y
Blood All ESBL positiveamoxy/clav, amp/sulb Y
All
Enterobacter cloacae, Enterobacter aerogenes, Serratia marcescens, Morganella morganii, Citrobacter freundii
ampicillin,cefazolin,cefuroxime,amp/sulb/amox/clav R
Example of table of expert rules - CAR
IF THEN
Source Organism Antibiotic Result= MIC= Add coded textALL Aeromonas hydrophila M1AER
Aeromonas caviaeAeromonas sobriaAeromonas sp.
ALL Enterobacter cloacae Ceftriaxone S M1AMC
Enterobacter aerogenes Ceftriaxone SMorganella morganii Ceftriaxone SSerratia marcescens Ceftriaxone SCitrobacter freundii Ceftriaxone S
ALL
Bacteroides fragilis, Bacteroides sp., Prevotella sp. M1ANA
Peptostreptococcus sp. M1ANAClostridium sp. M1ANA
STOOL Campylobacter jejuni M1CAMCampylobacter sp. M1CAM
BLOOD Enterococcus faecalis Ampicillin S M1ENCEnterococcus faecium Ampicillin S M1ENC
BLOOD Enterococcus faecalis Genta-HC S M1SYNEnterococcus faecium Genta-HC S M1SYN
Expert rules with auto-insert comment
code text when used
new
V. parahaemolyticus - diarrhoea associated with this organism is usually self-limiting and does not require specific antibiotic treatment. For Stool Vibrio parahaemolyticus
new
The pathogenic role of this organism for gastroenteritis is debatable. In the absence of symptoms, no treatment is necessary. Symptoms when they occur are usually self-limiting.
for Stool Aeromonas sp./ A. hydrophila/ A. sobria/ A. caviae
"for Stool blastocystis (parasite bench)
M1AMC (revise)
(old) Organism produces cephalosporinase (AmpC phenotype).Consider using non-cephalosporin antibiotics. Regard as resistant to Cefuroxime, Augmentin and Unasyn.
(new) Organism produces cephalosporinase (AmpC).Consider using non-cephalosporin antibiotics, or use cefepime. Regard as resistant to Cefuroxime, Augmentin and Unasyn. as per current rules
Maintaining list of coded text
Examples of expert rules - CAR
IF source=all organism = all
AND antibiotic gentamicin=S
THEN suppress amikacin
ELSE
Plain language: “whenever gentamicin is S, suppress amikacinresults”
Example 1 – Suppress reporting Criteria : Do not report Amikacin when is sensitive to Gentamicin Rule Script : if (Result.Antibiotic(GN1)=S) then { if (KnownResult.Antibiotic(AN1)) then { DoNotPrint.Antibiotic(AN1) } else { Nothing } } else { Nothing}
Examples of expert rules - CAR
Examples of expert rules - CAR
IF organism = Pseudomonas aeruginosa
AND ceftazidime = S
THEN suppress imipenem, meropenemand aztreonam
IF site=blood
AND ESBL= positive
THEN suppress amoxy-clavulanate, amp-sulbactam
Examples of expert rules - CAR
IF site = all
AND organism = Enterobacter cloacae, E. aerogenes, Serratia marcescens, Morganellamorgannii, Citrobacter freundii
THEN result= R for {cefazolin, cefuroxime, amp/sulbactam, amoxy-clav}
Examples of expert rules - CAR
Example 2 – Set to Resistant Criteria : Set Ampicillin and Cefuroxime to Resistant when Proteus vulgaris is isolated. Rule Script : if (Identification=PRVUL) then {SetRes.Antibiotic(AM1, R) : SetRes.Antibiotic(CXM1, R)} else { Nothing }
Examples of expert rules - CAR
IF organism = Enterobacter cloacae, E. aerogenes, Serratia marcescens, Morganellamorgannii, Citrobacter freundii
THEN insert comment M1AMC
Expert rules – insert comment
M1AMC: “Organism produces cephalosporinase (AmpC). Consider using non-cephalosporin antibiotics; or use cefepime”
Based on ordered test : Galactomannan test
LIS code : ASG4 Comment description : The galactomannan test has an overall sensitivity of 0.71 and specificity of 0.89 for invasive aspergillosis (Pfeiffer 2006). Repeat testing for low values of 0.5-0.8 is recommended (Maertens 2004). Predictive values also depend on patient type. Penicillin-containing antibiotics, cereals and milk products may cause false-positive readings.
Expert rules – Auto Insert
Comment description : This test has the following accuracy in the detection of activepulmonary tuberculosis. These standards are applicable to sputumand not to other specimens. Smear +ve patients Smear -ve patientsSensitivity 87% - 100% 64% - 71% Specificity >99% 99%
M. tuberculosis: nucleic acid amplification test
Expert rules –
Organism reported but no treatment recommended
Vibrio parahaemlyticus
V. parahaemolyticus - diarrhoea associated with this organism is usually self-limiting and does not require specific antibiotic treatment.
Expert rules –
Organism reported of uncertain significance
Stool with numerous blastocystis
The pathogenic role of this organism for gastroenteritis is debatable. In the absence of symptoms, no treatment is necessary. Symptoms when they occur are usually self-limiting.
Expert rules –
Example 3 – Auto Tag comment Criteria : Auto Append comment “Antibiotic treatment required only for severe disease.” (M1STO) for stool cultures when any of the following organism is isolated :
• Salmonella spp • Salmonella enteritidis • Salmonella Gp A/B/C/D/E • Campylobacter jejuni • Campylobacter spp
Rule Script : if (Protocol = MSA) then {if ((((Identification = SALSP) OR (Identification = SALEN)) OR ((Identification = SALGA) OR (Identification = SALGB))) OR ((((Identification = SALGC) OR (Identification = SALGD)) OR ((Identification = SALGE) OR (Identification = CAMJE))) OR (Identification = CAMSP))) then { AppendConclusion([M1STO]) } else { Nothing } } else { Nothing }
Expert rules –
Preliminary report at 24h: no growth of stool pathogens
Pending further cultures
(2-4% may be positive after another one to two days)
Use lab statistics to inform physicians
Expert rules –
Other LIS functions to complement expert rules
Expert rules work in the context of other LIS features to ensure optimal LIS utilization
Benefits
�More productive�For MD/ supervisor especially!�120 sign offs a day
�2 hours (old) @ 1 min per case�20 minutes (new) @ 10 seconds per case
�Concentrate on exceptions and consultation rather than “getting through”
�More productive�For MD/ supervisor especially!�120 sign offs a day
�2 hours (old) @ 1 min per case�20 minutes (new) @ 10 seconds per case
�Concentrate on exceptions and consultation rather than “getting through”
Benefits
�Standardized - consistent, less error�MTs don’t have to “remember” or “look up”�Refocus MTs training and knowledge
requirements
�Standardized - consistent, less error�MTs don’t have to “remember” or “look up”�Refocus MTs training and knowledge
requirements
Benefits
�More information for physician�Antibiotic comments, qualifiers, suggestions�More customized interpretation based on
conditional parameters� Include statistics that matter; sensitivity,
specificity� Include references - basis of
recommendations�Cautionary comments when non-standard
tests are done
�More information for physician�Antibiotic comments, qualifiers, suggestions�More customized interpretation based on
conditional parameters� Include statistics that matter; sensitivity,
specificity� Include references - basis of
recommendations�Cautionary comments when non-standard
tests are done
enhancements and application of rules not yet implemented
Real-time alerts
Alerts on threshold
�Alerts based on counting information selected on predefined criteria. These alerts are triggered when a defined threshold is reached: e.g. Alert when the number of requests not yet reviewed has reached 20
�These alerts are based on a counter where the user will set the entity to count.
Alerts on threshold
�Alerts based on counting information selected on predefined criteria. These alerts are triggered when a defined threshold is reached: e.g. Alert when the number of requests not yet reviewed has reached 20
�These alerts are based on a counter where the user will set the entity to count.
Single alerts
�Alert based on the contents of one request
�No need to reach a threshold to alert. Alert is immediate when the criteria are matched: e.g. blood culture and Burkholderiapseudomallei
�These alerts are based on a pre-defined rule linked to the relevant test.
Single alerts
�Alert based on the contents of one request
�No need to reach a threshold to alert. Alert is immediate when the criteria are matched: e.g. blood culture and Burkholderiapseudomallei
�These alerts are based on a pre-defined rule linked to the relevant test.
Real-time alerts
Single alerts - Examples
� Examples of single alerts:
� New MRSA isolated ( combination of identification and particular antibiotype)
Immediate alert to infection control service for appropriate action
� Alert on delta check on antibiotic result
Could indicate organism mutation/adaptation. Infection control service must be notified.
� Examples of single alerts:
� New MRSA isolated ( combination of identification and particular antibiotype)
Immediate alert to infection control service for appropriate action
� Alert on delta check on antibiotic result
Could indicate organism mutation/adaptation. Infection control service must be notified.
Conclusion
� Microbiology lab heavily reliant on LIS� Good LIS will have rules to facilitate functions
of reporting and epidemiology� Expert rules allow significant enhancement of
the professional value-add of the lab� Expert rules ensure consistency and increase
efficiency� Supervisor/ MD must be involved to maximize
the potential of the LIS to change microbiology and clinical practice
� Future enhancements: more customization, real-time decision support based on rules
� Microbiology lab heavily reliant on LIS� Good LIS will have rules to facilitate functions
of reporting and epidemiology� Expert rules allow significant enhancement of
the professional value-add of the lab� Expert rules ensure consistency and increase
efficiency� Supervisor/ MD must be involved to maximize
the potential of the LIS to change microbiology and clinical practice
� Future enhancements: more customization, real-time decision support based on rules