School for Primary Care ResearchIncreasing the evidence base for primary

care practice

Carl Heneghan

Diagnostic Reasoning forEvidence-Based Practice.

Background

The strategies clinicians use to arrive at a diagnosis makes only asmall contribution to current research and the medical curricula.

Debate still exists about our understanding of the complex strategiesused in diagnostic reasoning.

Determining diagnostic strategies is particularly pertinent wherepatients present with an array of conditions and possible diagnoses.

Our aim was to set out - for a forthcoming series of diagnosticarticles - the strategies and methods that are employed in routineclinical consultations.

What we did

1st phase: focus group of GPs and researchers in primary care toidentify several strategies - based on consensus opinion and thepublished literature - used in diagnosis.

Strategies mapped in a consecutive series of 100 patients by 1 GP(CH).

2nd phase: From phase 1 a model incorporating 3 stages used indiagnosis formulated.

Six GPs recorded data using a revised data collection sheets for 50clinical consultations.

Final focus session using a consensus development approach.

Spot diagnosis Self-label Presentingcomplaint

Pattern0%

10

%2

0%

30

%4

0%

50

%6

0%

70

%8

0%

90

%1

00

%

Figure 2: initiation of the diagnosis

% strategy used by GPS and overall mean

Diagnostic stages & Strategies

Self-labelling

Tonsillitis – had it before

I have a chest infectiondoctor

Groin strain

I have asthma

I have a uti – just like lasttime

Spot diagnosis Self-label Presentingcomplaint

Pattern0%

10

%2

0%

30

%4

0%

50

%6

0%

70

%8

0%

90

%1

00

%

Figure 2: initiation of the diagnosis

% strategy used by GPS and overall mean

Diagnostic stages & Strategies

Prediction rules: which ones do you use in

practice?

Would be useful to have them all in easily accessibleformat.

I find the pictures very helpful, they help me rememberwhere to look for tenderness.

If you have a prediction rule, the way you take the historychanges... you go for the 3-4 things.

I would like to use them as guidelines for teachingstudents.

Clinical prediction guides project

Balla J, Glasziou P, Heneghan C.

Pattern recognition fit

49 yr old with severe flare up of Ulcerative colitis

Intravenous hydrocortisone, aminosalicylates, and ciclosporintreatment was started, and after seven days oral prednisolone

was introduced.

Is the use of a specific but probably imperfectsymptom, sign or diagnostic test to rule in or out adiagnosis.

Used in less than 10% of cases. Onepossible reason for this is that a certainamount of informal probabilisticreasoning was not recognized during theclinical consultation

(informal refers to reasoning not based on numerical understanding, but relieson decision making based on qualitative information, e.g. the disease is more orless likely).

Probabilistic reasoning:

Murtagh’s Stepwiserefinement

Probability Pattern CP rule0%

10

%2

0%

30

%4

0%

50

%6

0%

70

%8

0%

90

%1

00

%

Figure 3: Refinement of the diagnosis

%strategyusedbyGPSandoverallmean

Refinement strategies

% strategy used by GPs (mean)

Spot diagnosis Self-label Presentingcomplaint

Pattern0%

10

%2

0%

30

%4

0%

50

%6

0%

70

%8

0%

90

%1

00

%

Figure 2: initiation of the diagnosis

% strategy used by GPS and overall mean

Diagnostic stages & Strategies

Strategies employed at the final diagnostic stage,

known diagnosis

ordering of further tests

test of treatment

test of time

In some cases the final

diagnosis cannot be

given a label

Known

Diagnosis

Furthertestsordered

Test oftreatment

Test oftime

No Label

10

%2

0%

30

%4

0%

50

%6

0%

70

%8

0%

90

%1

00

%

Figure 4: Defining the final diagnoses

The subsequent evolution of clinical features overtime is reviewed to see if the clinical features resolveor a particular diagnosis later becomes more obvious.

Thompson et al. Clinical recognition of meningococcal disease in children andadolescents. Lancet. 2006 Feb 4;367(9508):397-403

Test ofTime

TimeZero

VagueSigns

present

VagueSymptoms

present

IndeterminateInvestigations

Classical symptomsand signs present

Investigationspositive

BMJ diagnostic reasoning series

Diagnostic Reasoning|: Heneghan C, Glasziou P, Thompson M, Balla J,Rose P, Scott C, Lasserson D, Perera R.

Restricted Rule Out: Thompson M, Harnden A and Del Mar C Test of Treatment: Glasziou P, Rose P, Heneghan C, Balla J. Clinical Prediction Rules Fahey T Dan Mayer D Test of Time: Nick Summerton. Probabilistic Reasoning: Doust J. Self diagnosis: Spot” Diagnosis: a rapid but unreliable process: Glasziou P,

Heneghan C, Del Mar CB.