Using Data to Reduce Error, Standardize
Practice and Improve Patient Outcomes
James H. Nichols, Ph.D., DABCC, FACB
Associate Professor of Pathology
Tufts University School of Medicine
Director, Clinical Chemistry
Baystate Health System
Springfield, Massachusetts
[email protected]
Baystate Health System
Baystate Health System
•
•
Baystate Medical Center - tertiary care
572 beds, 3rd largest acute care in NE
• 40,000 discharges/200,000 inpatient days (4.7 mean LOS)
• 600,000 ambulatory visits
•
•
•
•
•
•
Western Campus of Tufts School of Medicine
Franklin and Mary Lane Hospitals
Over 40 Ambulatory Care Practices (1 million visits)
Home nursing and assisted care (156,000 visits)
Reference Lab (BRL) - 4 million tests/year
Clin Chemistry - Core 1 Roche TLA (2500/day)
Medical Errors
• Institute of Medicine of the National Academies
•
•
report 1999
Medical errors kill 44,000 - 98,000 patients in US
hospitals each year.
“Number one problem facing health care” Lucien
Leape, Harvard Professor of Public Health
Medical Errors
•
•
•
•
2002 Commonwealth Fund report estimated that 22.8
million people have experienced a medical error,
personally or through at least one family member
Reinforces the 1999 IOM report, “To Err is Human”
Annual costs estimated at $17 – 29 billion
US Agency for Healthcare Research and Quality
(AHRQ) estimate medical errors are the 8th leading
cause of death in the US – higher than:
• Motor Vehicle Accidents (43,458)
• Cancer (42,297)
• AIDS (16,516)
Laboratory Errors
• Typically think patient, tube or aliquot mix-up.
• Other, more insidious errors to consider
• Overutilization of testing – “fishing”
• Inappropriate use of testing – method selection or test
for symptoms, screening vs management
• Misunderstanding – wrong test, assume ‘test is a test’
• Delays – ordering, receipt of result, clinical action
Laboratory Errors
•
•
•
A minireview of the literature found the majority of
errors occur in the pre and post analytical phases.
Bonini P, Plebani M, Ceriotti F, Rubboli F. Clin Chem 2002;48:691-698.
Many mistakes are referred to as lab error, but actually
due to poor communication, actions by others involved in
the testing process, or poorly designed processes outside
the lab’s control.
Medical errors occur in prevention, diagnosis and drug
treatment occur. Among errors in diagnosis; 50% were
failure to use indicated tests, 32% were failure to act on
results of tests, and 55% involved avoidable delay in
diagnosis. Leape LL, Brennan TA, Laird N, et al. N Eng J Med 1991;324:377-84.
Man
A creature made near the end
of the week when God was
tired.
Mark Twain
Medical Errors
•
•
•
•
•
The Person
Easier to blame a person than an institution for errors.
In aviation, 90% of quality lapses are judged to be blameless.
The System
• Active failures due to personal interaction with system
• Latent conditions, weaknesses in system due to design flaws or
heirarchical decisions
Need to engineer systems that prevent dangerous errors and are
able to tolerate errors and contain their effects
Reason J. BMJ 2000;320:768-770.
Automation
•
•
•
•
•
Collects raw data and processes to information (trends)
Reduces practice variability (device prompts)
Consolidates operator interactions (barcoding)
Assists decision-making (internal checks for QC pass,
expiration dates, operator ID)
When linked to information management and data
algorithms can warn of possible errors (delta checks,
device flags like inadequate sample, analyzer
interferences)
Improvement
Data
Inquiry
Change
•
•
•
•
•
•
•
Hemolysis in the ED
Coagulation specimens must be rejected if hemolyzed
and recollected
Inpatient rates of hemolysis are typically <1%
ED had rates approaching 20% or more
Related to implementation of a flexible catheter and
practice of collecting blood through lines
Manufacturer even distributed a customer warning
against collecting blood through this catheter
Yet, ED unwilling to change practice – customer
satisfaction issue and comfort level of IV lines
Number of redraws and delays of ED patients led to
elimination of practice.
25
20
Implement Practice Change
ED
Inpatient
Outpatient
15
10
5
9/1/2003
7/1/2003
5/1/2003
3/1/2003
1/1/2003
0
11/1/2002
Percent Hemolyzed Specimens
Phlebotomy Hemolysis Rates
Middleware
•
•
•
•
•
Data server sits between an analyzer and LIS/HIS
POCT servers are a form of Middleware
Allows data processing before sending results LIS, also
functions as data repository for report searches
Common current uses – autoverification, insertion of
data flags for H/I/L indices
More sophisticated functions are limited only by
imagination of the lab
Clinical Alarms
•
•
•
•
•
•
Critical pathway ordering practices and variant
ordering practices
Hct level and POCT glucose testing
Medication (propofol) and potential test interference
(i-Stat)
Insulin dose, individual response and prediction of
future dose
Disease/medication (high blood pressure/loop diuretics)
vs predicted lab result (low K) vs questionable lab
results (high K)
Medical devices (flexible catheters) and potential for
hemolysis and laboratory interference
POCT Error Management
• POCT – diagnostic testing conducted close to
•
•
•
•
the site where clinical care is delivered
POCT error rates are not known in literature
POCT conducted by nursing but managed by lab
Requires considerable interdisciplinary
communication to deliver effectively
POCT QI can be a tool to uncovering ongoing
errors and addressing system weaknesses
Reducing Errors through Automation
•
•
•
•
Newer POCT devices have data management
Prompts operator to perform testing same way every
time
Lock-outs act as internal “fail-safes” to prevent a
patient result if QC fails, not performed or operator is
not certified for testing.
Feb 2004 CLIAC meeting discussion of possible
changes to CLIA waived category suggested that
waived tests have
• fail-safe or failure alert mechanisms whenever possible
• include QC materials with kits
• specimens requiring significant manipulation not be waived
Medical Errors
•
•
•
•
•
The Person
Easier to blame a person than an institution for errors.
In aviation, 90% of quality lapses are judged to be blameless.
The System
• Active failures due to personal interaction with system
• Latent conditions, weaknesses in system due to design flaws or
heirarchical decisions
Need to engineer systems that prevent dangerous errors and are
able to tolerate errors and contain their effects
Reason J. BMJ 2000;320:768-770.
Patient Identification Errors
• POCT results are transmitted to the POCT
•
•
•
manager when devices are downloaded
The data manager orders and results the test in
the LIS
If the test does not match an active patient
account the data manager holds the result for
resolution
Compliance problems as test cannot be billed,
and some results transmitted to incorrect patient
record and inappropriate medical management
Failure Mode and Error Analysis
• FMEA identifies an error
• Outlines possible steps that could lead to the
•
•
•
error.
Identifies the reasoning behind the various
pathways, why they exist and ways that paths
can be improved.
Establishes quantitative monitors and the means
of measuring improvement.
FMEA improves motivation by seeking route
causes of errors rather than placing blame.
ICU FMEA
•
•
•
•
Incidence of patient ID errors in our ICU led to an
administrative demand for improved compliance or loss
of privileges (3 strike rule)
Conducted FMEA analysis
ID errors due to multiple issues:
• Long number entry (9 digits), transposition of numbers
• Some devices can’t accept leading zeros
• Patient wristbands are not legible (clin engineering)
• Need for patient care, share operator IDs (retraining)
Barcoding seen as optimum solution
Barcoding
•
In practice, one of the more challenging projects to
implement in an institution:
• Devices only read specific barcode languages
• Wristbands vary in durability
• Ink isn’t permanent (thermal vs inkjet)
• Devices don’t require barcode entry!
• Try to engineer around manual entry by adding special
characters or digits to ID
• These work-arounds lengthen the barcode and increase read
failure if barcode not flat on wrist.
• How to print? Wristbands only or labels that an operator can
stick onto device or paper towel? What about neonates?
Barcoding
•
•
•
•
During implementation, operators continued to
manually enter patient IDs due to the scanner failing on
the 1st attempt
An investigation was conducted into why scanners fail
i-Stat scanners failed more frequently than glucose
Operator interaction with the POCT device was the
primary determinant in scanner failure
Scanner Angle
Scanner Distance
Scanner Depth of Field
Scanner Depth of Field
P=0.014
P=0.0007
P=NS, 0.378
P=0.048
Barcoding
•
•
Barcode acceptance and difficulties in implementation
lead to <100% effectiveness:
• Manual entry
• Barcoding patient with the wrong account or patient ID
• Patients with multiple wristbands
• Scanning the wrong barcode (lot number instead of patient)
From the AACC listserv, those successful institutions
communicate the value of barcoding and have operators
who have acknowledged the advantages and implement
strategies to enhance success
Communication
•
•
•
•
How best to reach clinicians?
Errors are a system weakness and require an
interdisciplinary system fix, one person is not
responsible.
Utilize available resources:
• Hospital Quality Improvement Teams
• Peer-Reviewed Literature
• Practice Guidelines
Learn to speak ‘clinicalese’ – Use Clinical Protocols
Portland Protocol
•
•
•
Examined glucose levels and surgical complications in
1,585 cardiac surgery patients with diabetes (990
preprotocol and 595 postprotocol)
Implemented protocol of postoperative intravenous
insulin to maintain glucose <200 mg/dL.
Intensive monitoring and insulin therapy on
hospitalized inpatients lowers blood glucose levels in
the first 2 postoperative days with concomitant
decrease in proportion of patients with deep wound
infections (2.4% vs 1.5%, p<0.02)
Zerr KJ et al. Ann Thorac Surg 1997;63:356-61.
Portland Protocol
•
ACC/AHA Guidelines for CABG Surgery
“Another patient characteristic that has been associated
with postoperative mediastinitis is the presence of
diabetes, especially in patients requiring insulin. In
addition to the microvascular changes seen in diabetic
patients, elevated blood glucose levels may impair
wound healing. The use of a strict protocol aimed at
maintaining blood glucose levels 200 mg/dL by the
continuous, intravenous infusion of insulin has been
shown to significantly reduce the incidence of deep
sternal wound infection in diabetic patients.”
Eagle KA, Guyton RA. JACC 1999;34:1262-1347.
Portland Protocol
Blood Glucose Insulin Unit/hr
<125
0
125-175
1
175-225
2
>225
3
•
•
•
•
q1hr until glucose 125-175 with <15 mg/dL change and insulin
rate unchanged x4 hrs. Then q2hr.
Weaning vasopressors (Adrenalin) check q30min until stable
Stop q2hr testing on POD #3
Test q2hr during the night on telemetry if glucose <200
Portland Protocol Operational Issues
• Which method to utilize? [TAT, Accuracy]
•
• Glucose meter – glucose oxidase
• Blood Gas glucose – glucose oxidase
• Core laboratory glucose - hexokinase
Preferred sample? [Method, Line Contamination]
• Whole blood or plasma
• Fingerstick, line draw or venipuncture
Unmodified direct-reading biosensor result
”relative molality” of glucose
in plasma or whole blood
(not recommended)
1.18
0.94
Concentration of
glucose in plasma
(recommended)
1.11
Concentration of
glucose in whole blood
(not recommended)
Fig. 1. Conversion factors for different quantities of glucose.
Meter Performance Criteria
ADA ‘87
ADA ‘94
Agence du Médicament
(95% of data)
CSA
FDA
(95% of data)
ISO
(95% of data)
IMSS
NCCLS (C30A)
TNO
All Levels
All Levels
< 100 mg/dL
 100 mg/dL
< 45 mg/dL
 90 mg/dL
< 100 mg/dL
 100 mg/dL
< 100 mg/dL
 100 mg/dL
< 60 mg/dL
 60 mg/dL
< 100 mg/dL
 100 mg/dL
< 117 mg/dL
 117 mg/dL
± 15%
± 5%
± 20 mg/dL
± 20% (CV <7.5%)
± 25% (CV<12.5%)
± 15% (CV <7.5%)
± 20 mg/dL
± 20%
± 20 mg/dL
± 20%
± 25%
± 20%
< 15 mg/dL
± 20%
± 20 mg/dL
± 15% mg/dL (CV <10%)
•
•
•
•
Quality Specification Modeling
Monte Carlo simulation to generate random “true” and
“measured” glucose based on mathematical model of meters
having defined imprecision and bias. (N=10,000 - 20,000 pairs)
Analytical error
Insulin dose errors
5%
8 - 23%
10%
16 - 45%
2x or greater insulin dosage errors >5% of time when analytic
error exceeded 10 - 15%
Total error < 1 - 2 % required to provide intended insulin >95%
of time.
Boyd JC. Bruns DE. Quality specifications for glucose meters: Assessment by
simulation modeling of errors in insulin dose. Clin Chem 2001;47:209-214
Portland Protocol
• Glucose meters may or may not be applicable
•
•
for tight management, as can vary by +/-20% in
the 100–200 mg/dL range.
Blood gas and some analyzers perform better
than glucose meters, may be more appropriate in
these cases.
Should be a clinical not a laboratory decision,
role of laboratory to inform not dictate method
I-Stat 1: Cartridge versus PCx Strip Glucose Comparison
Which Glucose is Appropriate for My Patient?
(Package Insert Recommendations)
i-Stat Cartridge
Hematocrit ranges
15 – 75% Hct
Sample Type
Fresh whole blood or lithium
heparin within 30 mins
Manually heparinized syringes
can overfill heparin or syringes
not completed filled with blood
can result in wrong results due
to heparin conc
Collection Considerations
PCx
20 – 70% Hct, falsely
increased or decreased values
outside Hct range
Fresh capillary preferred or
arterial/venous lithium heparin
Same issues
Tubes need to be well mixed
Reportable Range
Accuracy and Precision
Neonates
Patient Limitations
IV line collections can
contaminate and dilute
specimen. Recommend
discarding 3 to 6 x length of
catheter before collection
20 – 700 mg/dL
20 – 600 mg/dL
More accurate and precise
Less accurate and precise
~3% coefficient of variation
(CV) across reportable range,
or 55+/-3mg/dL and 215+/14mg/dL (compare to core lab
glucose with 1% CV, or 94+/2mg/dL and 283+/-4 mg/dL)
Acceptable within accuracy
limits
Glucose Oxidase Method
~5-19% coefficient of
variation, depending on
concentration, or 49+/-18 and
283+/-30 mg/dL
pH variations from 7.4 can
alter by 1 mg/dL or more for
every 0.1 increment.
Dehydrated or severely
hypotensive patients in shock,
hyperglycemic/hyperosmolar
states (with or without ketosis)
may give erroneous results
Low pO2 can decrease results
Bromide can decrease results
Thyocyanate can decrease
results
Hydroxyurea can give
Caution advised <50 mg/dL
Glucose Oxidase Method
LABORATORY FACT SHEET I-STAT 1 GLUCOSE TESTING
Representing the next step in the
evolution of point-of-care testing,
the i-STAT 1 Analyzer uses one
platform to perform glucose strip
testing along with all i-STAT test
cartridges.
Summary
Point of care testing is a mode of testing in which the analysis is performed at the site
where health care is provided. The I-STAT Point-of-Care system is an instrument,
designed to offer a rapid broad range of laboratory tests in a variety of panel
configurations, which include: glucose, blood gases, potassium, ionized calcium,
hematocrit, chloride, pH, sodium, bicarbonate, BUN, etc. and various calculated
parameters. The analyzer utilizes single-use multisensor test cartridges and Precision PCx
glucose strips. The glucose testing can be performed as a single test, using the PCx
glucose test strip or in a panel, using the test cartridge. Quantitative measurement of
glucose can be done on venous, capillary, arterial, or neonatal whole blood.
Operational considerations
Depending on the patient condition the clinical staff will need to decide on which side of ISTAT 1 to obtain a glucose, the PCx test strip or I-STAT cartridge. The following table
and considerations may be helpful.
Which glucose test is right for my patient?
PCx test strip
I-STAT cartridge
1)Use for routine glucose monitoring and
insulin management
1)Use only in following circumstances:
a) in combination with blood gases or
electrolytes on same cartridge
b) to monitor hypoglycemia where a more
accurate assessment of glucose is
required
c) neonates
d) in presence of PCx interference
2)Less accurate and precise
2)More accurate and precise
~5-19% coefficient of variation, depending on
concentration, or 49+/-18 and 283+/-30 mg/dL
~3% coefficient of variation (CV) across reportable
range, or 55+/-3mg/dL and 215+/-14mg/dL
(compare to core lab glucose with 1% CV,
or 94+/-2mg/dL and 283+/-4 mg/dL)
3)Less expensive
$0.75/test
3)More expensive
$5.00/test
Clinical Protocols
•
•
•
Clinical protocols provide a pathway of care to manage
patients with specific disorders in the most effective
manner for optimum patient outcome.
Incorporating laboratory testing into clinical protocols
standardizes practice, reduces practice variability,
ensures appropriate ordering of tests and can assist the
interpretation of test results.
Clinical protocols are a good means of communicating
with clinicians and providing reminders or important
components of decision-making
2004 National Patient Safety Goals JCAHO
•
•
•
•
•
•
•
Improve the accuracy of patient identification
Improve the effectiveness of communication among
caregivers
Improve the safety of using high-alert medications
Eliminate wrong-site, wrong patient, wrong-procedure
surgery
Improve the safety of using infusion pumps.
Improve the effectiveness of clinical alarm systems.
Reduce the risk of healthcare-acquired infections.
NACB Laboratory Medicine Practice
Guidelines – Evidence Based
Practice for POCT
• Clinicians, staff and laboratorians need guidance
•
•
to apply POCT in the most effective manner for
patient benefit.
This guidance should be based on a concurrence
of the scientific evidence to date.
This need for evidence-based practice was the
concept behind the NACB Laboratory Medicine
Practice Guidelines for POCT
Evidence-Based Practice for POCT
•
•
•
•
•
POCT is an increasingly popular means of delivering laboratory
testing.
When used appropriately, POCT can improve patient outcome by
providing a faster result and therapeutic intervention.
However, when over-utilized or incorrectly performed, POCT presents
a patient risk and potential for increased cost of healthcare.
This LMPG will systematically review the existing evidence relating
POCT to patient outcome, grade the literature, and make
recommendations regarding the optimal utilization of POCT devices
in patient care.
Develop liaisons with appropriate professional, clinical organizations:
ACB, ADA, ACOG, CAP, etc.
Evidence-Based Practice for POCT
•
•
•
•
•
•
•
•
•
•
•
•
•
Focus Group Chairs
Cardiac – Robert H. Christenson, Ph.D.
Diabetes – Christopher Price, Ph.D.
Reproduction – Ann M. Gronowski, Ph.D.
Infectious Disease – Robert Sautter, Ph.D.
Coagulation – Marcia Zucker, Ph.D.
Parathyroid – Lori J. Sokoll, Ph.D.
Drugs – Ian Watson, Ph.D.
Bilirubin Screening – Steven Kazmierczak , Ph.D.
Critical Care – Greg Shipp, Ph.D.
Renal – William A. Clarke, Ph.D.
Occult Blood – Kent Lewandrowski, M.D.
pH – James Nichols, Ph.D.
Introductory Comments – Ellis Jacobs, Ph.D.
Evidence Based Practice for POCT
pH Guidelines I
•
•
Does the use of pH paper for assisting the placement of
nasogastric tubes, compared to clinical judgment (air,
pressure) improve the placement of tubes on inpatient,
endoscopy, home care and nursing home patients?
We recommend the use of pH testing to assist in the placement of
nasogastric tubes. The choice of measuring pH with an
intragastric electrode or testing tube aspirates with a pH meter or
pH paper will depend on consideration of the clinical limitations
of each method, and there is conflicting evidence over which
method is better. (Class II – prospective comparative trials and
expert opinion)
Evidence Based Practice for POCT
pH Guidelines I
•
•
•
Assuring correct NG or NI tube placement:
• Measure length of tube
• Direct visualization of oropharynx
• Auscultation of stomach by air insufflation
• Aspiration of gastric contents
• Irrigation of tube with 10 to 50 mL water
• Direct palpation of tube within stomach during intraabdominal procedures
Gold Standard - Abdominal roentgenogram to confirm
position
pH may be faster, safer and more economical
Evidence Based Practice for POCT
pH Guidelines I
•
•
•
•
Gastric contents more acidic
Neuman – pH < 4 can reduce need for x-rays (PPV
100%, Sens 100%, Spec 88% for N = 46 patients and 78
placements.) pH>4 not useful – respiratory or duodenal.
Acid suppressors increase gastric pH and 6.0 may be a
better cutoff (81% pH 1 – 4, 88% intestinal >6.0,
pulmonary >6.5). Confounds aspirate pH 4 – 6.
pH of gastric fluid may replace 85-95% of x-ray cases.
Significant decrease radiation exposure
Evidence Based Practice for POCT
pH Guidelines I
•
•
•
•
•
•
Method to determine pH controversial
Continuous monitor or pH tipped NG tube preferred for those
patients that are equipped, but expensive.
Question whether pH probes are measuring gastric contents or cell
surface pH
Aspirate pH may not generate sufficient volume, may differ from
intragastric pH, as antacid, drug salts, protein and bile may
interfere with some methods.
pH meter more accurate than pH paper, but paper simpler (0.5 –
1.0 increments), cheaper, easier to use and quality assure, and can
be performed bedside.
X-ray confirmation still the “gold standard” and recommended in
indeterminate cases.
Question Five: Can gastroccult
testing of gastric fluid from a
nasogastric tube be used to detect
gastrointestinal bleeding in highrisk intensive care unit patients
receiving antacid prophylaxis?
Recommendation Five:
• We cannot currently recommend for or against
•
the use of gastroccult to detect gastric bleeding
in intensive care unit patients receiving antacid
prophylaxis.
Grade of Evidence: III – small study, clinical
evidence
Gastroccult Tests
• FOBT should not be used to measure occult
•
blood in gastric fluid because of interferences
from low pH, certain medications and metal
ions.
The presence of occult blood in gastric fluid can
be useful to detect stress ulcer syndrome, so
specific gastroccult tests are utilized.
Bleeding in ICU Patients
•
•
•
A small study with 41 patients showed that 13/14
patients with positive gastroccult tests had a source of
upper GI bleeding as seen by upper endoscopy.
Study suggest gastroccult testing may aid in detecting
occult bleeding in critically ill patients.
However, patients with negative gastroccult tests did
not undergo upper endoscopy which may have
documented false negative results.
Baystate Gastroccult Testing
• Discontinued without incident
• Approached Chief of GI and Division of
•
•
•
•
Healthcare Quality with clinical utility.
Researched literature
Developed recommendation and justification
Draft letter to medical staff reviewed by select
clinicians
General announcement and test removal
Gastroccult Discontinuation
•
•
•
•
•
No peer-reviewed literature indicating improved outcomes based
on Gastroccult
Use of test after NG tube placement leads to positive results
solely due to trauma of tube insertion
Overt bleeding is a medical concern and doesn’t require test to
detect
pH is medically useful, pH paper is a better alternative because
it’s easier to QC, already available on units and lower cost
Elimination would reduce hospital burden of training and POCT
documentation on nursing staff and reduce risk of developer
mixup with hemoccult.
Gastroccult Cost Savings
•
•
•
•
•
Reagent: (12,000 tests/year)
• Cards
• Developer
Labor
• Nursing (5 min/test, 45K= 125d)
• Competency (1100 x 15 min)
• Lab oversight (4hr x 8 units x 12 mo)
Total Annual Savings Estimate
Total billed previous year
Cost estimate for pH replacement
$21,000
$ 5,000
$22,000
$ 6,000
$ 8,500
$62,500
12
$ 250
•
•
•
•
•
Summary
Medical errors are a significant problem and the laboratory
should be aware of the many opportunities to reduce errors
Interdisciplinary teams and positive attitudes are important
factors in achieving successful outcomes and changes to practice
Need to engineer systems (not people) that prevent dangerous
errors and are able to tolerate errors and contain their effects
Automation, information management and communication are
effective strategies to reduce errors.
The next challenge for laboratorians is to better integrate the data
we have at hand and condense the literature into standard
practice pathways that assist clinicians in appropriate decisionmaking for optimal patient care
Descargar

POCT for COMACC, Atlanta 97