Analyzing Performance Improvement Flashcards
data aggregation
pool data in 1 place, collect performance indicators, helps see big picture
benchmarking
comparison of data, common expectations/standards
internal benchmarking
org performance against itself over time, stricter
external benchmarking
compare one org to a group or org collecting data on the same measures in the same way- same scale
types of data
qualitative: nominal, ordinal
quantitative : discrete, continuous
nominal data
categorical data, values assigned to name specific category
ex: gender, inpt/outpt
bar/pie charts
ordinal data
ranked data
compare evaluation of various characteristics and value relative to each other
ex: pain scale, stages of disease
how respondents feel about issue
bar/pie charts
discrete data
numerical values represents whole numbers
ex: # of children in family
bar graphs
continuous data
assumes infinite number of possible values, decimal values
ex: weight, BP, temp
histogram
sampling
sufficient # of observations can be predictive of overall configuration of data
not efficient to collect every single occurrence - too frequent
ex:
30 pts pop –>use all cases
pop greater than 596 –> use 120 cases
key performance indicators
types of data most important for org, balanced look
clinical quality –> adverse events, mortality
financial viability –> net revenues, growth
customer loyalty –> staff loyalty, pt satisfaction
operational effectiveness –> staff efficiency, readmission rate,
data sources
MR, admin database, pt surveys, adverse drug rxn reports, incident reports, performance evaluations, infection surveillance, JC surveys
when data aggregated, what happens
data loses context and usefulness
ex: average of test is high, but there could be individuals issues overlooked
used as report card, won’t pinpoint what needs to be fixed
bundled data
overall report, lose context, begin to pinpoint issue
unbundled data
break down data to show the differences, and pinpoint where action needs to be
Qualitative QI tools
brainstorming
set priorities
maintain direction
determine problem causes
clarify process
present ideas in useful form
quantitative QI tools
measure performance
collect, display data
monitor performance
present performance measurement in useful form
fishbone/ishikawa diagram
breaks down problem, does not determine root cause, gives options
5 whys
asking multiple times will identify root cause, the more whys, the harder it gets to answer
run chart
view performance trends over time
ex: breast pts in imaging center
reveal interesting structure present
averages can help avoid overreacting
bar graph
compares relative size of various data categories
histogram
graphical representation that is used to describe single set of continuous data- bell curve
scatter diagram
graphical representation used to determine relationships between quantitative variables of interest
ex: correlation
pie charts
relationship of each part to the whole
pareto chart
80-20 rule (focus on 20%, improve by 80%)
prioritize problem solving on vital few
assumes quantity = importance (not always true)
prevent shifting problem to other causes
ongoing measurement of progress
radar chart
display importance categories of performance
defines full performances for each category
shows gaps between current and full performance
where do dashboards pull data from
EHR
variation
how it currently works vs how it should work
process inputs
ppl, method, machine, measurement
source of variation, always changing, reflected in output
types of statistics
descriptive, inferential
why is it better to use median than mode
outliers can skew the mean
numerical methods under descriptive analytics
measures of central location - mean, median, mode
measures of variability - range, SD, interquartile range
control chart
determine unstable/stable process
contains median of upper control, low control (3 SD away from mean)
turn bell curve on its side
distinguish special and common causes of variation
predictive analytics
make future prediction about key performance measures
AI
SD
spread of values, how far off from mean
bell curve
1 SD- 68%
2 SD- 96%
3 SD- 99.7
tall- tight process
wide- unstable process
absolute frequency
number of times that a score or value occurs int he data set, denoted by f1
relative frequency
percentage of time that the characteristic score appears in data set
considered a special cause when
one or two points outside of UCL or LCL
or
two out three successive values are on same side of centerline/3 SD away from mean
