Post-test probability depends on:

Pre-test probability and updating factor

Refusing to change DDx once formed

Formal method for calculating posterior probability?

Likelihood ratios and Bayes’ theorem

If FN are more costly, recommended cutoff?

Lower cutoff → more sensitivity

How ROC curves help with cutoffs?

Show trade-off between Se and Sp

Why might a diagnosis not be reconsidered after relapse?

Clinical thinking must adapt beyond textbooks

Optimal cutoff aims to minimize:

False positives and false negatives

Purpose of applying Bayes’ theorem:

Convert conditional probabilities to post-test values

In linear reasoning, diagnoses are:

Weighted by consistency with findings

Particular population’s prevalence

When are LRs especially useful?

Continuous or ordinal result tests

Next step in estimating naive probability:

Estimate number of diagnoses per species

Two probabilities to multiply in diagnosis:

Sign probability × regional prevalence

Bayes’ theorem helps calculate:

Convert conditional probabilities into clinical use

Differential list ordering is based on:

Textbook compatibility + population probability

To calculate likelihood of a condition, you should:

Multiply observed sign probability by regional prevalence

What is an “optimal cutoff”?

Balance false positives and false negatives

Three considerations for DDx ordering:

Signalment, signs, and local probability

Post-test probability in Bayes’ theorem represents:

Probability after adjusting pre-test probability

studbud

Epidemiology: Diagnostic Test II & Clinical Reasoning

200 termsStudBud Administration

Imported from external platform

Flashcards200 cardsFPractice quizPHost liveL

Questions 200

Answers

1
Post-test probability depends on:
Pre-test prob + specificity
Pop size and prevalence
Pre-test probability and updating factor
Sensitivity and specificity alone
2
When ordering DDx list, what two factors are used?
Owner preference & cost
Textbook match only
Most severe first
Presentation match & frequency
3
How does Youden index treat FN and FP?
Equal weight
No weight
Higher weight to FN
Higher weight to FP
4
Youden index (J) if Se = 1.000, Sp = 0.000:
1
0.5
-1
0
5
Purpose of using LRs in diagnostics:
Estimate post-test probability
Measure accuracy
Assess disease prevalence
Determine cutoff value
6
Formula for updating factor in Bayes’ theorem:
P(test+│disease) × P(disease)
P(test+│disease) / P(test+)
P(disease│test+) / P(test+)
P(test−│disease) / P(test−)
7
"Value rigidity" is:
Quickly changing DDx list
Refusing to change DDx once formed
Preferring common diseases
Focusing on new diagnoses
8
Why prioritize common conditions first?
They’re more frequent
Easier to diagnose
Textbooks only cover common ones
Rare ones aren’t serious
9
Formal method for calculating posterior probability?
Use result alone
Sens/Spec analysis
Likelihood ratios and Bayes’ theorem
Addition and subtraction
10
What happens if cutoff is too high?
↑ Spec, ↑ FN
↑ Sens, ↑ FN
↓ Se and Sp
Perfect balance
11
If FN are more costly, recommended cutoff?
Higher cutoff → more specificity
Equal Se and Sp
No cutoff needed
Lower cutoff → more sensitivity
12
Lower specificity affects LR+ by:
Increasing LR
Decreasing LR
Increasing false negatives
No effect
13
Main focus of PVP?
Detect rare diseases
Predict presence of disease
Predict absence of disease
Test sensitivity across populations
14
How ROC curves help with cutoffs?
Show trade-off between Se and Sp
Eliminate false positives
Maximize positives
Plot total test results
15
In the algorithmic model, clinicians:
Assign weights
Work without structure
Follow a flowchart
Ignore probabilities
16
Why might a diagnosis not be reconsidered after relapse?
It was ruled out
All relapses show treatment success
Clinical thinking must adapt beyond textbooks
Impossible to diagnose relapses
17
In high prevalence populations, PVP:
Increases
Is unaffected by prevalence
Stays the same
Decreases
18
Each segment’s area in AUC represents:
ΔSens × Avg Spec
Sum of true positives
Avg false positives
ΔSpec × Avg Sens
19
First step to estimate naive probability from records:
Calculate treatment cost
Find severe cases
Count total patient visits
Focus on individual histories
20
Optimal cutoff aims to minimize:
Specificity
False positives and false negatives
Test cost
Sensitivity
21
Purpose of applying Bayes’ theorem:
Convert conditional probabilities to post-test values
Statistical modeling
Find optimal test cutoff
Calculate test accuracy
22
In linear reasoning, diagnoses are:
Never revised
Based on intuition
Randomly chosen
Weighted by consistency with findings
23
If Se = 1.000, Sp = 0.000, conclusion?
Identifies negatives only
Perfectly accurate
No ability to ID negatives
High specificity, low sensitivity
24
PVN is specific for:
General population prevalence
Particular population’s prevalence
Test cost
Test sensitivity
25
Bayesian update model suggests DDx list should:
Include untreatable diseases
Evolve with new information
Prioritize rare diseases
Stay the same throughout
26
In clinical diagnosis, we start with:
A confirmed diagnosis
A set of clinical findings
The most common disease
One test result
27
AUC value of 1.0 indicates:
Random chance
Low specificity
High false positive rate
Perfect performance
28
When are LRs especially useful?
Continuous or ordinal result tests
Tests with high specificity
When prevalence is unknown
Tests with binary outcomes
29
Next step in estimating naive probability:
Focus on 1 diagnosis only
Count only common diseases
Ignore number of diagnoses
Estimate number of diagnoses per species
30
If cutoff = 0, what is Sp?
1
0
Undefined
0.5
31
Two probabilities to multiply in diagnosis:
Preference and cost
Treatment rate + sign severity
Age and breed
Sign probability × regional prevalence
32
Importance of "horses before zebras"?
Uncommon signs don’t matter
Rare ones ruled out first
Textbooks focus on zebras
Common diseases more likely
33
Best description of a quantitative test?
Is qualitative
Continuous measurable data
Gives only +/– results
Only measures specificity
34
Bayes’ theorem helps calculate:
Sensitivity and specificity
Test accuracy
False-positive rate
Convert conditional probabilities into clinical use
35
Differential list ordering is based on:
Vet’s preference
Textbook compatibility + population probability
Random disease selection
Owner request/cost
36
When prevalence = 50%, the Youden index minimizes:
Total errors
Sensitivity
Specificity
False positives
37
Naïve probabilities don’t consider:
Climate
Treatment options
Clinical factors
Owner’s preferences
38
To calculate likelihood of a condition, you should:
Multiply observed sign probability by regional prevalence
Ignore regional prevalence
Only use signs
Use experience only
39
A key challenge in diagnosis:
Textbooks cover all cases
Animals share signs
Diagnosis is always visible
One disease per animal
40
A diagnosis likely in population may be:
Irrelevant to practice
Incorrect at individual level
Always confirmed by tests
Always correct for individual
41
Optimal cutoff outcome?
100% Se and Sp
All animals classified correctly
No confirmatory tests needed
Balanced Se and Sp
42
How can AUCs be compared between tests?
Statistically compare AUCs
Observe false negatives
Measure specificity at each cutoff
Average of sensitivities
43
What does PVP measure?
Sensitivity of the test
Specificity of the test
Test-positive truly affected
Test-negative truly unaffected
44
What is an “optimal cutoff”?
Maximize Se and Sp equally
Cheapest cutoff
Lowest positive threshold
Balance false positives and false negatives
45
Three considerations for DDx ordering:
Signalment, signs, and local probability
Owner’s concern, cost, age
Signs, treatment, age
Only common diseases
46
Skipping the full exam can:
Eliminate DDx
Simplify diagnosis
Improve accuracy
Risk missing findings
47
The true disease status of animals is usually:
Easily determined
Uncertain in most cases
Irrelevant to clinical diagnosis
Known with certainty
48
Post-test probability in Bayes’ theorem represents:
False positive likelihood
Probability before testing
Probability after adjusting pre-test probability
Test accuracy
49
If you hear hoofbeats, expect to find:
A zebra
A horse
Any exotic animal
A rare species
50
Hypothesis-finding is learned through:
Textbooks only
Experience
Lab tests only
Guesswork
51
ROC curve was developed by which engineers?
Electrical and radar engineers
Veterinarians
Data scientists
Biochemists
52
Clinician’s informal role in probability revision?
Use statistical software
Weigh evidence to decide best action
Calculate LR mathematically
Use math to determine accuracy
53
Decreasing the cutoff results in:
Sens ↓, Spec ↑
Both ↓
Both ↑
Sens ↑, Spec ↓
54
Conditions that can’t be diagnosed/treated:
Most important to diagnose
Never include
Exclude from list
Less important unless severe
55
In forming a DDx list, you should consider:
Rare diseases first
Most common diseases first
Diseases matching one sign
Owner preferences
56
Formula for LR– (negative result):
Se ÷ (1 – Sp)
(1 – Sp) ÷ Se
(1 – Se) ÷ Sp
Se ÷ Sp
57
A disease stays on DDx if:
Owner requests it
Has effective treatment
Has never been seen
Is not common locally
58
Predictive values are especially important in populations where:
Disease is hard to diagnose clinically
Disease is common and prevalence is high
Disease is rare and prevalence is low
Test is invasive
59
Exception to excluding untreatable conditions:
If rare and hard to test
If slight chance of treatment
If severe and clients need awareness
If client insists
60
PVP increases when:
Disease is rare
High specificity, low sensitivity
Disease prevalence is high
Low sensitivity, high specificity
61
Why should a full physical exam always be done?
To avoid missing key signs
All animals share symptoms
Replace diagnostic tests
To confirm a single disease
62
For confirmatory testing, which cutoff approach?
High sensitivity, low FP
High sensitivity, low cutoff
High specificity, high cutoff
Balanced Se and Sp
63
Frequency of a condition matters because:
Rare ones first
Common ones are more likely
No need for testing
Prevents misdiagnosis
64
Poorly selected cutoff can result in:
High specificity
Unclear results
Accurate diagnosis
Bad clinical decisions
65
Experienced diagnosticians think in:
Probabilities
Certainty
Simple rules
Random guesses
66
Cognitive reasoning styles:
Don’t affect diagnosis
Are irrelevant
Same for all vets
Vary among individuals
67
Role of prior probability in probability revision?
Ignored during post-test calc
Estimates likelihood before testing
Only applies to screening
Calculated post-test
68
If cutoff = 0, what is Se?
0
1
0.5
Undefined
69
“Naïve probabilities” refer to:
Condition likelihood in practice area
Disease severity
Treatment success rate
Prognosis
70
Ordering a DDx list should consider:
Joint probability of signalment, signs, and region
Most expensive test
Random disease selection
Only most likely from textbooks
71
Bayes' theorem calculates:
Disease prevalence
Prior probability
False negatives
Post-test probability
72
Posterior probability represents:
Test result was incorrect
Updated disease probability after result
Test accuracy
Initial disease likelihood
73
Clinicians base decisions on prior knowledge from:
Owner observations only
Populations of similar animals
Gut feeling
The specific patient only
74
How does PVN differ from specificity?
PVN only for symptomatic animals
PVN unaffected by prevalence, specificity depends on it
PVN = predicts disease absence, specificity = overall accuracy
PVN = test accuracy for all diseases
75
A priority when forming DDx list:
Can’t-miss diseases
Mild consequences
Only rare diseases
Non-life-threatening
76
Diagnostic reasoning is an exercise in:
Clinical epidemiology
Clinical trials
Population statistics
Lab testing
77
Model most similar to a flowchart:
Bayesian
Algorithmic
None
Linear
78
Overall AUC is determined by:
Mean specificity
Number of true positives
Avg sensitivity at cutoffs
Sum of all segment areas
79
What is probability revision in diagnostics?
Calculate test accuracy
Convert prior to posterior probability
Ignore clinical history
Use result w/o prior info
80
Adjusting a DDx list should be:
Focused on rare diseases
Hypothesis-driven
Random
Based only on labs
81
LR– if Se = 0.90 and Sp = 0.95:
0.05
0.95
18
0.11
82
What does the AUC represent in an ROC curve?
Number of true positives
Specificity
Overall diagnostic performance
Sensitivity
83
Hemorrhage in granulocytic ehrlichiosis is due to:
Thrombocytopenia/clotting failure
Excessive platelets
More clotting factors
Hypercoagulation
84
Diagnoses are either correct or incorrect because:
Clinical judgment is always right
Every test is accurate
Probability = 1 or 0
All diseases have clear markers
85
LR close to zero indicates:
Test is unreliable
No diagnostic value
Useful for ruling in disease
Useful for ruling out disease
86
What happens to sensitivity when cutoff is decreased?
Stays the same
Decreases
Becomes irrelevant
Increases
87
History-taking and physical exams should be:
Focused on one system
Hypothesis-driven but complete
Done after treatment
Skipped for tests
88
What happens when you increase test cutoff?
Sens ↑, Spec ↓
Sens ↓, Spec ↑
Both Sens & Spec ↑
Both Sens & Spec ↓
89
In low prevalence, a positive test is:
Less likely to show disease
More likely to show disease
More likely to be false positive
Unaffected by specificity
90
ROC curve y-axis:
Prevalence
Sensitivity
False negative rate
Specificity
91
Bayesian model updates diagnosis probabilities:
Ignores new findings
As new info is gained
After all tests
Only at start
92
Each animal does or does not have a disease, so probability is:
Between 0 and 1
Always 1
Always 0
Either 0 or 1
93
Best way to learn DDx:
Memorizing diseases
Relying on textbooks
Ignoring probability
Watching skilled clinicians
94
Experienced clinicians form hypotheses:
After all tests
Early in exam/history
Only after reading
After treatment ends
95
P(disease│test positive) in Bayes’ theorem represents:
Probability of disease given a positive result
Probability of a false positive
Overall disease probability
Probability of test given disease
96
A dog with granulocytic ehrlichiosis may show:
Cyanosis and bradycardia
PU/PD
Reluctance to move/neck stiffness
Increased appetite
97
Why is clinical diagnosis different from textbook cases?
Always follows the book
Real-world may not match books
Cases never follow patterns
Textbooks are rare condition-based
98
For screening, which cutoff priority?
High specificity and higher cutoff
Balanced Se and Sp
High sensitivity
High specificity
99
What does PVN measure?
Sensitivity of the test
Probability a positive result is true
Probability that test-negative means no disease
Test usefulness across all populations
100
The most likely diagnosis is the one with:
Easiest treatment option
Rarest disease
Highest probability in similar population
Most severe signs
101
In diagnosis, which is more likely?
Common signs of rare condition
Rare signs of common condition
Rare diseases more frequent
All conditions present similarly
102
When changing the cutoff, what happens to specificity?
Unaffected by cutoff changes
Decreases
Increases
Increases with lower cutoff
103
Probability represented by PVP?
Detects disease when present
Positive = truly affected
Negative truly unaffected
False positives occur
104
A definitive diagnosis should be used only when:
Condition is common
No other diagnosis is possible
Symptoms match textbook
Lab tests confirm it
105
“Book” descriptions help determine:
Probability of signs matching a condition
Prognosis
Owner’s concerns
Treatment options
106
Why choose a cutoff in diagnostics?
Estimate disease from signs
Define positive or negative status
Assess specificity
Determine test cost
107
Electronic records estimate:
Treatment success
Naïve probabilities for conditions
Most severe cases
Most common age
108
What does J = 0 mean?
No classification ability
Perfect test
Highly accurate test
Detects positives only
109
Higher Se increases LR+ because:
LR decreases
Test is less accurate
Test is less useful
Helps rule in disease
110
Primary goal of probability revision in decisions?
Update disease probability using new evidence
Calculate test accuracy
Determine pre-test probability
Identify confirmatory test
111
Key feature of Bayesian model:
Probabilities change with new data
Uses a decision tree
Uses numerical scores
Diagnoses are fixed
112
Typical lectures present diseases by:
Focusing on rare cases
Starting with a diagnosis
Listing symptoms randomly
Describing cases
113
Updating factor in Bayes’ theorem:
Determines prevalence
Adjusts prior probability using test info
Calculates specificity
Converts test to sensitivity
114
A common presentation of a rare condition is:
More likely
Less likely
Equally likely
Basis for every diagnosis
115
NOT a common sign of granulocytic ehrlichiosis:
Weight gain
Petechiae/ecchymoses
Fever
General stiffness
116
Disease from uncontrolled cell growth:
Neoplasia
Trauma
Inflammation
Degeneration
117
Formula for Youden index?
J = Se + Sp
J = Se + Sp - 1
J = Se - Sp
J = Se - 1 + Sp
118
Purpose of trapezoidal integration in AUC?
Calculate sensitivity
Approximate AUC when ROC is nonparametric
Measure specificity
Improve Se accuracy
119
AUC value of 0.5 indicates:
High Se and Sp
Perfect test performance
Random chance; test is useless
Low false results
120
The “horses vs. zebras” analogy means:
Consider rare diseases first
Use textbook patterns
Focus on common conditions first
Ignore rare diseases
121
Hypotheses in clinical diagnosis are:
Fixed and final
Fluid and update with info
Never revised
Based only on labs
122
LR > 1 indicates:
Test not informative
No diagnostic value
More likely true positive
More likely false negative
123
"Zebras" should:
Stay on list until outcome
Be excluded
Be ignored
Be prioritized
124
Large positive LR indicates:
Test has low sensitivity
Useful for ruling in disease
Useful for ruling out disease
Uninformative test
125
Formal probability estimates require:
Random chart pulls
Only interviews
Experience only
Published data and local info
126
LR value of 1 indicates:
Disease is certain
Uninformative test result
Very informative test
Perfect test
127
ROC curve x-axis:
Specificity
1 - Specificity
Disease prevalence
Sensitivity
128
When diagnosing, cutoff choice should consider:
Test cost
Disease prevalence and predictive values
Sensitivity only
Sample ease
129
Human minds have a limit on considered hypotheses:
—
TRUE
FALSE
—
130
In low disease prevalence, a negative test result is:
Unaffected by sensitivity
More likely to indicate true absence
Less likely to be true
More likely to be a false positive
131
LR+ if Se = 0.90 and Sp = 0.95:
0.95
0.11
18
1
132
No benefit in diagnosing a condition that:
Can be treated right away
Is easily tested
Cannot be treated
Is common
133
Consequence of too high cutoff?
↑ Spec, ↑ FN
Better generalization
↑ Sens, ↓ FP
↑ False positives
134
Why not quickly discard severe conditions from a DDx list?
Treatable later
Always most likely
Usually harmless
May require rapid action
135
ROC curves are useful for tests that:
Have fixed cutoff
Are qualitative
Are quantitative with multiple cutoffs
Low Se, high Sp
136
How would test perform with Se = 1.000, Sp = 0.000?
Never miss a true positive, mislabels healthy
Perfect disease and no disease detection
Many false negatives
High Se and Sp
137
Category for injuries from external force:
Trauma
Degeneration
Neoplasia
Malformation
138
First step in diagnosis:
Consider system + diagnosis categories
Most expensive test
Only rare diseases
Owner’s preferences
139
ROC curve graphically represents:
Sensitivity at cutoffs
Overall accuracy
Specificity at cutoffs
False positive/negative counts
140
What must be considered in building a DDx list?
One disease only
Age only
Most expensive test
Regional condition likelihood
141
When should regional disease probability be considered?
When ordering DDx list
Only for rare diseases
During treatment
When ignoring signs
142
How does higher Sp affect LR–?
No effect on LR–
Decreases LR–
Increases LR–
Affects only LR+
143
AUC indicates what when calculated?
Number of true negatives
Affected subject has higher result than unaffected
Detects unaffected individuals
Number of false positives
144
Historical significance of “ROC”?
Coined by biologists
Radar effectiveness in WWII
Origin of diagnostic tests
Disease prevalence method
145
Formula for LR+ (positive result):
Se ÷ Sp
(1 – Se) ÷ Sp
(1 – Sp) ÷ Se
Se ÷ (1 – Sp)
146
How is PVP related to prevalence?
PVP decreases as prevalence increases
Higher in low prevalence populations
PVP increases as prevalence decreases
PVP is unaffected by prevalence
147
Malformations involve:
Degeneration
Trauma
Structural birth defects
Cancer growth
148
How is AUC typically calculated for a parametric ROC?
Calculus (integration)
Trapezoidal approximation
Rectangular approximation
Nonparametric methods
149
Why is PVN important in diagnostics?
Measures test sensitivity
Measures overall accuracy
Determines if test-negative is truly unaffected
Shows likelihood of false positives
150
Effect of decreasing cutoff?
No change in performance
More sensitive, less specific
More specific, less sensitive
Not applicable
151
What do likelihood ratios estimate?
Test accuracy
Post-test probability from prior odds
Disease prevalence
Test specificity
152
What happens if cutoff is too low?
↑ Sens, ↓ FP
↑ Spec, ↓ FN
↓ Se and Sp
↑ Sens, ↑ FP
153
NOT a method of probability revision?
Clinical intuition
Bayes’ theorem
Sensitivity and specificity
Likelihood ratios
154
A hypothesis-driven approach means:
Consider only treatable diseases
Never change the list
Ignore probabilities
Update DDx list with new info
155
Signalment includes:
Cost of treatment
Disease severity
Breed, age, and sex
Prognosis
156
A large LR means:
Strong evidence of disease
Not useful clinically
Low specificity
Likely a false positive
157
NOT a major inflammation category:
Metabolic
Parasitic/protozoal
Fungal
Autoimmune
158
PVN is higher in populations with:
Unaffected by prevalence
Equal prevalence
Low disease prevalence
High disease prevalence
159
When disease prevalence is high, PVN:
Stays the same
Decreases
Increases
Is unaffected by prevalence
160
Astute clinical diagnosticians focus on:
Guesswork over data
Absolutes
Probabilities
Personal experience only
161
Even zebras should remain on DDx list:
If client wants
If mild
Until recovery or death
If treatable
162
Signalment and history should:
Have no impact
Be skipped
Guide diagnoses and raise questions
Wait until imaging
163
Compatibility of clinical signs with condition presentations is:
Only for rare diseases
Not useful for prognosis
Irrelevant
A key factor in ordering DDx list
164
Physical exams can focus on specific regions based on:
Expensive tests
Owner’s concerns only
Initial hypotheses
Random chance
165
Probabilities for differential diagnoses should be based on:
Clinical experience + scientific data
Clinical trials only
Textbooks only
Client opinion
166
Formula for Youden index (J)?
J = Se + Sp
J = Se + Sp - 1
J = Se - Sp
J = (Se*Sp)/(1-Se+1-Sp)
167
Informal approach to probability revision?
Weigh evidence intellectually
Use nomograms and math
Use only prior probability
Calculate post-test probability
168
A DDx list should be:
Focused on one system
Fixed after creation
Unchanged after exam
Re-ordered with new data
169
Changing test cutoff is like:
Using single test to rule out
Increasing test cost
Interpreting tests in series/parallel
Using only qualitative results
170
How is LR calculated?
Se ÷ Sp
Disease prob ÷ no-disease prob
Se × Sp
Test result ÷ false positives
171
J = 1.0 means:
Low specificity test
Perfect test with no errors
Random chance
High sensitivity, many false negatives
172
Clinical experience or literature can determine:
Regional prevalence of conditions
Animal age most at risk
Best rare disease treatment
Test costs
173
PVP is specific for:
Rare disease detection
General population prevalence
Test cost-effectiveness
Specific population prevalence
174
Granulocytic ehrlichiosis is commonly associated with:
GI symptoms
Acute lameness or stiffness
Renal failure
Neurologic signs only
175
Which model assigns positive/negative weights to findings?
Linear
Bayesian
Algorithmic
None
176
Differential diagnosis lists should be:
Narrow only
Comprehensive, even if partially pursued
Based only on rare diseases
Dropped after diagnosis
177
Probability represented by PVN?
Chance of false positives
Disease present given a positive result
Disease negative given a negative result
Affected will test positive
178
What influences the probability of diagnosis?
Owner expectations
Source population
Cost of treatment
Test availability
179
Youden index assumes:
No cost to FN or FP
FP cost more
False negatives cost more
FN and FP are equally costly
180
Youden index (J) of 0 indicates:
Perfect test performance
Low sensitivity with high specificity
Equal sensitivity and specificity
Random chance; no diagnostic value
181
Youden index is used to assess:
Number of false positives
Accuracy of test cutoff
Equipment reliability
Sensitivity of the test
182
Final step in naive probability calculation:
Check outcomes first
Divide diagnoses by presentations
Add all species together
Multiply diagnoses by population
183
When is PVN higher?
Test detects rare diseases
High sensitivity but low specificity
High disease prevalence
Low disease prevalence
184
ROC curve term introduced during:
Civil War
WWII
Cold War
WWI
185
Range of likelihood ratios (LR):
–1 to 1
0 to 100
0 to ∞
0 to 1
186
Compatibility with presentation can be treated as:
Probability between 0 and 1
Irrelevant
Always 0
Always 1
187
Main focus of PVN?
How well a negative test predicts disease absence
Overall sensitivity
Probability of false positives
Accuracy of positive results
188
The “perfect test” would have:
Fixed sensitivity/specificity
Unchangeable cutoff
No overlap of positive/negative distributions
Overlapping distributions
189
Why is PVP important in diagnostics?
Measures general accuracy
Shows if positive = truly affected
Measures likelihood of false negatives
Indicates test sensitivity
190
Purpose of ROC curve in testing?
Measure test cost
Evaluate true positives
Assess specificity
Find optimal cutoff
191
How does PVP differ from sensitivity?
Sens = more accurate in low prevalence
PVP = test result count
PVP = same as sensitivity
PVP = presence in a specific population, Sens = detection
192
Why do clinicians create a differential list?
Ensure all diseases are treated
True diagnosis can’t be directly observed
To confirm diagnosis immediately
Avoid diagnostic tests
193
Degenerative diseases cause:
Progressive function/structure decline
Rapid tissue healing
Bacterial infections
Congenital defects
194
What does a specificity of 0.000 imply?
Identifies both equally
Always identifies healthy individuals correctly
Perfectly identifies healthy individuals
Never identifies healthy individuals correctly
195
“Joint probability” in diagnosis means:
Use only local prevalence
Considering signalment, signs, region together
Ignoring signs
Focus only on severity
196
Bacterial vs viral inflammation difference:
Viruses always more severe
Bacteria respond to antibiotics
Viruses need antifungals
Bacteria don’t cause inflammation
197
Informal probability estimates are:
Based on publications
From client opinion
Done “in your head”
From computers
198
Region where granulocytic ehrlichiosis is common:
Northeast US
Rocky Mountains
South-central US
Pacific Northwest
199
LR < 1 indicates:
Test is uninformative
More likely to rule out disease
More likely true negative
More likely false positive
200
Hematologic finding in granulocytic ehrlichiosis:
Leukocytosis
Anemia and thrombocytopenia
Polycythemia
Lymphocytosis

Epidemiology: Diagnostic Test II & Clinical Reasoning Flashcards | StudBud