Making a Diagnosis in Primary Care

1
Making a Diagnosis in Primary Care

• Prevalence, Possibilities and Likelihood

2
Making a Diagnosis in Primary Care

• How do we normally achieve it?
• 1) Pattern recognition
• E.g. Parkinsons, ganglion, Downs etc.
• intuitive, reflexive, experiential
• multiple senses
• simple and occasional

3
Making a diagnosis in primary care

• 2) Algorithm
• Logical, useful for multiple users
• Triage
• Deals with a single symptom in detail
• NICE, PCTs and other agencies love them
• Narrow usefulness

4
Making a diagnosis in primary care

• 3) Complete History and Physical Exam
• Thorough
• Inclusive
• Exhaustive/ exhausting
• 2 stage - all info. Then sort by priority
• Essential to know if you are a novice
• use selectively if you are experienced.?

5
Making a diagnosis in primary care

• Try this -
• a) we have a 56 year old man here to see you,
  doctor , with chest pain and shortness of
  breath.
• or
• b) We have just sutured the wrists of a 19-yr-
  old student.
• What are your thoughts ?

6
Making a diagnosis in primary care

• 4) Hypothetico- deductive
• Early clues formulation probability list
• Most usual
• 28 seconds on average !(11-55 range)
• Correct hypothesis on average _at_ 6 mins
• On average 5.5 hypotheses per patient
• Prove/ disprove
• experience helps not in the technique but the
  accuracy

7
Making a diagnosis in primary care

• How can we improve?
• 1) Practice our observation and listening skills
  (hence all the work on consulting skills)
• 2) Develop our dynamic understanding of both
  health and disease processes (lifelong business
  feeding on innate curiosity but needing focus and
  energy.
• 3) Know how to balance and interpret data in our
  context

8
Making a diagnosis in primary care

• Problems...
• Rules of thumb (heuristics) can be very
  contextual - we judge the odds by prior
  experience all the time - correct experience?
• Broadened range of possibilities in unsorted
  community
• Unsynthesised jumble of symptoms
  (under-reporting, telescoping, recall,
  fabrication, interpretation of terms

9
Making a diagnosis in primary care

• Problems (cont.).
• Tendency to prefer the organic hypothesis
• Investigations all have physical, psychological
  and financial costs
• We have been trained by 2ndary care specialists
  who can give well-meaning guidelines irrelevant
  to the primary care context

10
Making a diagnosis in primary care

• .. The main problem we are talking about is
• Prevalence
• ( Prior/ Pre-test odds)

11
Prevalence Probability

• Bear with us - this may be painful but it will
  be rewarding!

12
Prevalence Probability

• Have a go - guestimate-
• 1)55yrs, male, hypertensive, exertional central
  chest pain radiates to arm and jaw resolves with
  rest.
• Probability of coronary disease?

13
Prevalence Probability

• 2)35yrs, male, no risk factors, heartburn for
  years, now 6/52 non-exertional lower sternal
  squeezing pain for 6 weeks, radiates to back,
  worse when lying down.
• Probability of coronary disease?

14
Prevalence Probability

• 3)45yrs, male, 20/day smoker, no other risk
  factors,3/52 precordial/substernal pain, usually
  sharp fleeting, occasionally heavy weight,
  inconsistent relation to exercise. Slight
  costo-chondral tenderness does not replicate the
  pain.
• Probability of coronary disease?

15
Prevalence Probability Patient (1) (pretest
odds 90)
16
Prevalence Probability Patient (2) (pretest
odds 5)
17
Prevalence Probability Patient (3) (pretest
odds 50)
18
Prevalence Probability

• Looking at these three examples we can see the
  gain in certainty (or probability) achieved in
  each case with a positive and negative result
  varies wildly with the prevalence/ pretest odds.
• The nearer we are to pre-test certainty the less
  valuable the test

19
Prevalence Probability

• i.e. A tests value in positively or negatively
  predicting is profoundly affected by disease
  prevalence/ pre-test odds.
• We must therefore be fully aware of prevalence
  in our context and the tests power before
  ordering it

20
Prevalence Probability

• Likelihood ratio expresses the odds that a given
  result would be expected in a patient with the
  disease
• ( The proportion of those with the disease who
  test positive (sensitivity) divided by the
  proportion of those without the disease who
  tested positive (1- specificity))
• Sensitivity()
• 100 - Specificity()

21
Prevalence Probability

• Because likelihood ratios of a test are
  calculated vertically (like sensitivity and
  specificity) they are not so affected by
  prevalence
• They can also be set for the strength of a result
  (e.g. how much ST depression)
• Pretest x Likelihood Post-test
• odds ratio odds

22
Prevalence Probability

• Nomograms for likelihood ratios prevent the need
  for converting probabilities to odds ratios and
  back again for post test probability, and can be
  a real help clinically.

23

• To determine post-test probability draw a line
  through pre-test probability and likelihood ratio