Coincidence Analysis (CNA): A method to identify conditions - - PowerPoint PPT Presentation

Coincidence Analysis (CNA):

A method to identify conditions influencing implementation

Deborah Cragun, PhD, MS University of South Florida College of Public Health dcragun@health.usf.edu Alanna Kulchak Rahm, PhD, MS Geisinger Health Systems

Overview

• 1. Why and when to use CNA
• 2. What CNA can and cannot do
• 3. Step-by-step example (hypothetical data)

Path Model

Inferential Statistics: Path Modeling

• Requires large sample sizes
• Random sampling (gold standard)
• Quantitative data
• Independent variable
• r

probability of dependent variable (holding other variables constant) Configurational Comparative Methods: Coincidence Analysis

• Small to large sample sizes
• Purposive sampling
• Quantitative or qualitative data
• Combinations of one or more

factors (conditions) may be needed for an outcome

Fundamental Dif ifferences

Home Fire

Outcome

Faulty electrical AND nearby couch

AND other key conditions also make a difference

Conditions

Inferential Statistics: Path Modeling

• Requires large sample sizes
• Random sampling (gold standard)
• Quantitative data
• Independent variable
• r

probability of dependent variable (holding other variables constant) Configurational Comparative Methods: Coincidence Analysis

• Small to large sample sizes
• Purposive sampling
• Quantitative or qualitative data
• Combination of several factors

may be needed for an outcome

• Uncover multiple independent

paths to the same outcome

Fundamental Dif ifferences

Home Fire

More than

• ne cause

#1 #2

Together flame source AND nearby fuel

are minimally necessary and sufficient to start fire

Inability to detect OR put out

fire then leads to house fire

Home Fire CNA identifies ONLY conditions that make a difference among observed cases Presence of oxygen is necessary but not a difference maker Detect & put out Fuel Source

An it iterative process

Determine research question Select cases Conduct CNA Collect data

Apply theory and empirical knowledge of cases

Research Question

Implementation Strategies Outcome

Successful Implementation

Contextual Factors

Data

Qualitative - interview transcripts Quantitative - scale measures

Select cases

Conduct CNA

Determine research question & design

• 1. Select conditions

and calibrate scores

• 2. Evaluate data-

truth table

• 3. Run analysis
• 4. Interpret results

Select cases Collect data

Step 1a: Select Conditions

Implementation Strategies

• Collaborative formed

held multiple planning meetings

• Information

accessed from LSSN website

Contextual Factors

• Crisp-set (0 or 1) - presence or absence of condition
• 1 =full member
• 0 =non-member
• Fuzzy set (value between 0 and 1) - degree of each condition
• .75= mostly a member of the set
• .25=mostly outside set membership
• Multi-value (0, 1, 2, …) finite number of values
• 0= no hospitals
• 1= some hospitals
• 2= all hospitals

Step 1b: Calibrate Set Membership Scores

Home Fire

House fire=0 (hf)

House fire=1 (HF)

Fuel = 0 (f) Detect & put out =1 (D) Detect & put out =0 (d) Fuel =1 (F) Source=1 (S) Source = 0 (s)

Step 2: Evaluate Data -Truth Table

• Ensure diversity
• Truth table can be created in the

cna package for R

• Shows configurations (patterns)
• f conditions and outcomes

Q A N E P C I S 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Configur

• ation

Conditions Outcome Cases Contextual factors Strategies

High quality evidence (Q) Negative attitude of key person (A) Strong communi- cation networks (N) Strong leadership engage- ment (E) Peer pressure (P) Collaborative group holds multiple meetings (C) Informa- tion from LSSN website (I) Successful implement- tation (S) Hospitals (N=30)

c1 1 1 1 1 1 1

LU, UR, SU, OW, NW, AR, AI

c2 1 1 1 1 1

GL, UG, SO, SG, AG

c3 1 1 1 1 1

GR, TG

c4 1 1 1 1 1 1 1

UH

c5 1 1 1 1 1 1

BE

c6 1 1 1 1 1 1

SH

c7 1 1 1 1 1

BL

c8 1 1 1 1

TI

c9 1 1 1

VS

c10 1 1 1 1

FR, EU

c11 1 1 1 1 1

JU

c12 1 1 1

VD, NE, GE, PP

c13 1 1 1

BS, KP, GP

Step 3: Run analysis in cna package for R

• Default coverage and consistency threshold = 1 (can lower)
• Specify “causal ordering” (if known)

Communication networks = N Leadership engagement = E Negative attitude = A

• Collaborative multiple

meetings = C

• Information accessed

from LSSN website = I

Successful Implementation = S

Peer pressure = P Evidence quality= Q

Solutions

Atomic solution formulas:

• Outcome C:

solution consistency coverage N + a*e <-> C 1 0.947 Outcome S: solution consistency coverage C + E <-> S 1 1.000 C+ a <-> S 1 0.957 N+ a + E <-> S 1 0.957

Complex solution formulas:

• utcome solution

consistency coverage C,S (N + a*e <-> C)*(C + E <-> S) 1 0.947 C,S (N + a*e <-> C)*(C + a <-> S) 1 0.947 C,S (N + a*e <-> C)*(N+ a +E<-> S) 1 0.947

Key presence absence

Step 4: Interpret solutions

Complex solution formulas:

• utcome solution

C,S (N + a*e <-> C)*(C + E <-> S) C,S (N + a*e <-> C)*(a + C <-> S) C,S (N + a*e <-> C)*(a+ N +E<-> S) Collaborative & Multiple Meetings (C) Successful Implementation (S) Strong Communication Networks

• r

+

• r

+

Model ambiguity

N Negative Attitude – Key Person Strong Leadership Engagement a*e C E

Step 4: Interpret solutions

Complex solution formulas:

• utcome solution

C,S (N + a*e <-> C)*(C + E <-> S) C,S (N + a*e <-> C)*(C + a <-> S) C,S (N + a*e <-> C)*(N+ a +E<-> S) Collaborative & Multiple Meetings Successful Implementation Strong Communication Networks

• r
• r

Strong Leadership Engagement Negative Attitude – Key Person

Model ambiguity

Key presence absence

Step 4: Interpret solutions

Complex solution formulas:

• utcome solution

C,S (N + a*e <-> C)*(C + E <-> S) C,S (N + a*e <-> C)*(C + a <-> S) C,S (N + a*e <-> C)*(N + a +E<-> S) Successful Implementation Strong Communication Networks

• r

Collaborative & Multiple Meetings

• r
• r

Model ambiguity

Strong Leadership Engagement Negative Attitude – Key Person Key presence absence

Step 4: Interpret model fit

Complex solution formulas:

• utcome solution

consistency coverage C,S (N + a*e <-> C)*(C + E <-> S) 1 0.947 C,S (N + a*e <-> C)*(C + a <-> S) 1 0.947 C,S (N + a*e <-> C)*(N+ a +E<-> S) 1 0.947

Atomic solution formulas:

• Outcome C:

solution consistency coverage N + a*e <-> C 1 0.947 Outcome S: solution consistency coverage C + E <-> S 1 1.000 C + a <-> S 1 0.957 N + a + E <-> S 1 0.957

Thank You! Questions will be held until the end of the session

Determine research question & design

• 1. Select conditions

and calibrate scores

• 2. Evaluate data
• truth table
• 3. Run analysis
• 4. Interpret

results

Select cases Collect data

Configur- ation

Conditions

Outcome Cases Contextual factors Strategies

Negative attitude of key person (A) Strong communi- cation networks (N) Strong leadership engage- ment (E) Collaborative group holds multiple meetings (C) Successful implement- tation (S) Hospitals (N=30)

c1 1 1 1 1

LU, UR, SU, OW, NW, AR, AI

c2 1 1 1

GL, UG, SO, SG, AG

c3 1 1 1 1

GR, TG

c4 1 1 1 1

UH

c5 1 1 1 1 1

BE

c6 1 1 1 1

SH

c7 1 1 1

BL

c8 1 1

TI

c9 1 1

VS

c10 1 1

FR, EU

c11 1 1 1

JU

c12 1

VD, NE, GE, PP

c13 1

BS, KP, GP