Improving the Accuracy and Timeliness of the Medical Problem List

scarfpocketΤεχνίτη Νοημοσύνη και Ρομποτική

24 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

107 εμφανίσεις

Problem Statement

In

a

clinical

environment

where

patient

data

is

managed

electronically,

the

problem
-
oriented

approach

to

medical

documentation

can

contribute

to

the

delivery

of

consistent,

high

quality

care
.

The

key

to

this

model

is

the

basic

medical

Problem

List
.

The

Problem

List

serves

the

dual

function

of

providing

a

succinct

summary

of

the

patient’s

aggregate

medical

condition

as

well

as

a

focus

around

which

the

continuing

process

of

documenting

medical

data,

decisions,

and

the

plan

of

care

can

evolve
.


Maintaining

the

medical

problem

list

as

a

component

of

the

EHR

resolves

a

key

challenge

by

providing

this

resource

in

multiple

care

settings

simultaneously
.

However,

in

order

to

guarantee

the

accuracy

and

timeliness

of

the

Problem

List,

approaches

unique

to

the

electronic

healthcare

environment

are

required
.

Here

we

describe

three

efforts

which

we

believe

will

enhance

the

value

of

the

Problem

List

and

help

maximize

the

ability

of

this

document

to

influence

care
.

Introduction

In

recent

years,

events

have

focused

national

attention

on

the

quality

of

medical

care
.

Not

only

are

there

continuing

efforts

to

define

and

promote

quality

medicine,

but

also

the

causes

of

medical

errors

are

receiving

intense

scrutiny
.

A

substantial

subset

of

the

errors

seen

in

modern

care

settings

relate

to

failings

in

the

diagnostic

process
.

These

failings

represent

not

only

errors

in

assigning

the

proper

diagnosis

to

patients

but

also

reflect

a

failure

to

communicate

known

diagnoses

among

the

multiple

healthcare

professionals

who

participate

in

the

delivery

of

care

to

an

individual

patient
.


Improving the Accuracy and Timeliness
of the Medical Problem List

Peter Haug, M.D.

1, 2
, Stephane Meystre, M.D., Ph.D.

1
, Kathryn Gibb Kuttler,
Ph.D.
2
, Jauhuei Lin, M.D., Ph.D.
1, 2

University of Utah
1

and Intermountain Healthcare
2
, Salt Lake City, Utah

Results


NLP

System
:

A

significant

improvement

in

the

completeness

of

the

problem

list

was

noted

in

a

randomized

trial

of

the

technology
.


Data

Preparation

Framework
:

The

DPF

appears

capable

of

dramatically

reducing

the

amount

of

time

necessary

to

assemble

the

data

and

apply

the

training

algorithms

required

to

build

disease
-
detection

modules


Structured

Daily

Note
:

Incorporating

reporting

of

new

problems

into

the

automated

intensive

care

note

increased

the

frequency

with

which

these

problems

are

appropriately

documented
.


Conclusion

The

three

approaches

described

will

provide

us

a

starting

point

in

our

efforts

to

improve

the

completeness

and

timeliness

of

the

electronic

medical

Problem

List
.

Our

goal

is

to

first

guarantee

that

the

list

is

complete

and

timely
.

Second,

we

would

like

to

use

the

information

captured

there

to

support

key

clinical

processes

including

ordering,

clinical

documentation,

and

the

delivery

of

interventions

in

a

way

consistent

with

each

patient's

overall

clinical

status
.

Acknowledgements


Deseret

Foundation

Research

Grant
:

BNs

in

Problems,

2002

(Co
-
PI’s
:

Peter

Haug,

MD

and

Jauhuei

Lin,

MD)


Deseret

Foundation

Research

Grant
:

Problem

List

Management

Using

Natural

Language

Understanding,

2003

(Co
-
PI’s
:

Peter

Haug,

MD

and

Stephane

Meystre,

MD)

Contact Information

Peter Haug MD

Peter.Haug@imail.org

The

use

of

natural

language

processing

(NLP)

to

support

the

completeness

of

the

Problem

List

(Figure

1
)
.

In

analyzing

electronic

Problem

Lists,

we

observed

a

low

level

of

completeness

and

timeliness

in

this

document
.

In

response,

we

developed

a

NLP

system

to

automatically

search

for

a

group

of

common

diagnoses

in

narrative

documents

from

the

medical

record
.

When

a

diagnosis

was

identified,

it

was

compared

with

existing

diagnoses

recorded

in

the

problem

list
.

If

absent,

it

was

proposed

as

an

addition

to

this

list

the

next

time

a

clinician

accessed

the

Problem

List

Management

Application

(PLMA)
.

Deriving

suggested

problems

using

machine

learning

techniques

(Figure

2
)
.

In

the

past,

we

have

used

machine

learning

techniques

to

develop

diagnostic

systems

for

single

diseases

of

interest
.

Developing

these

models

has

proven

to

be

labor

intensive
.

Recently,

we

have

tested

a

Data

Preparation

Framework

(DPF)

that

greatly

reduces

the

effort

required

to

train

individual

diagnostic

modules

and

that

optimizes

the

data

for

machine

learning
.

We

anticipate

that

this

approach

will

allow

us

to

generate

hundreds

of

modules

These

modules

will

inspect

each

patient's

clinical

data

and,

upon

concluding

that

an

undocumented

disease

is

present,

will

add

it

to

the

list

of

proposed

problems

for

adjudication

by

the

physicians

in

the

PLMA
.

Three Approaches to Support the Medical Problem List

Figure 1

Figure 2

Figure 3

Embedding

problem

documentation

in

a

structured

daily

note

(Figure

3
)
.

In

an

effort

to

incorporate

problem

list

management

into

the

daily

documentation

effort,

we

developed

an

electronic,

intensive

care

note
.

This

application

expedites

documentation
-
by
-
organ
-
system

while

encouraging

the

recording

of

those

problems

associated

with

each

system

reviewed
.