Seminar on

1
Seminar on Intelligent data analysis and data
mining Application in medicine

• Research on poisonings in children public health
  perspective for the development of clinical
  algorithms
• by
• Dr Sergio Pièche

2
Developing clinical algorithms in public health

• The problem
• The target
• Principles
• Research

3
Developing clinical algorithms in public health
The problem

• Injuries
• Mortality causing deaths
• Morbidity burden of the condition
• Age group at risk
• Costs hospital and primary health care
• Likely impact of interventions

4
Developing clinical algorithms in public health
The target

• Health providers at primary health care level
• Health background doctors, medical assistants,
  nurses, other health workers
• Type of facility equipment, supply, access to
  referral facility

5
Developing clinical algorithms in public health
Principles

• Safe and effective guidelines
• Sensitive and specific clinical signs
• Minimum number of clinical signs
• Requiring simple skills to be used
• Standard and simple assess-classify-treat system
• Possible to teach and learn
• Minimum number of essential drugs
• Best care possible for severe cases

6
Clinical algorithm
ASSESSMENT signs CLASSIFICATION for action TREATMENT the action
Danger signs SEVERE Referral pre-referral treatment
Other signs MODERATE Treatment (follow-up needed)
Other signs or no signs MILD Home care
7
Developing clinical algorithms in public health
Research

• Hydrocarbon poisoning
• Organophosphate poisoning

8
Developing clinical algorithms in public health
Research on poisoning prospective study

• Clinical predictors of severity of accidental
  poisoning from hydrocarbons and organophosphates
  in children below 5 years old

9
Developing clinical algorithms in public health
Research on poisoning aim

• to develop an algorithm for the outpatient
  management of children with hydrocarbon and
  organophosphate poisonings at primary health care
  facilities in developing countries.

10
Developing clinical algorithms in public health
Research steps

• Derivation of clinical decision rule (factors
  with predictive power)
• Prospective validation of the algorithm in
  different settings
• Provider performance analysis
• Impact

11
Developing clinical algorithms in public health
Research approach

• Identification and standard definition of signs
  and symptoms
• Gold standards for diagnoses
• Definition of outcomes
• Observer variability and bias
• Procedures (protocol and instruments training,
  supervision)

12
Developing clinical algorithms in public health
Research methodology - 1

• Enrolment
• Children 2 to 59 months old
• History unintentional exposure to hydrocarbons
  or organophosphates
• Acute exposure
• Seen within 48 hours of exposure to poison
• New cases

13
Developing clinical algorithms in public health
Research methodology - 2

• Procedures
• All children admitted for at least 48 hours
  post-exposure irrespective of severity (written
  consent and free admission)
• Examined by study physician investigations upon
  admission
• Followed up at 6, 12, 24, 48 hours post-exposure
• No delay or interference with quality care

14
Follow-up
Post-exposure Post-exposure Post-exposure Post-exposure Post-exposure Post-exposure
OPD/ER 6 hours Follow-up 12 hours Follow-up 24 hours Follow-up 48 hours Follow-up Discharge / death
Cl. exam. (Lab tests X-ray) Intermediate outcomes Intermediate outcomes Intermediate outcomes Intermediate outcomes Final outcome

15
E.g. Hydrocarbon poisoning
Respiratory signs cough, fast breathing,etc Vomiting Chemical pneumonitis Outcome
Respiratory signs cough, fast breathing,etc Vomiting Bacterial pneumonia Outcome
Severity
16
Developing clinical algorithms in public health
Research methodology - 3

• Sample size
• To detect the overall association and prediction
  of common symptoms and signs with poisoning
  severity and outcome
• To account in the analysis for stratification of
  cases in sub-groups based on time of exposure to
  poison

17
Key questions

• Which common clinical signs and symptoms best
  predict poisoning severity and outcome?
• How long is the safe clinical observation period
  before sending home a child who has been exposed
  to hydrocarbons or organophosphates?

18
Developing clinical algorithms in public health
Research Analysis

• Chi-square statistics or Fisher exact test, risk
  differences, risk ratios, odds ratios
• Multivariate logistics regression - incl.
  stepwise techniques
• Data mining techniques to be considered
• Sensitivity, specificity, predictive accuracy

19
Data analysis

• The challenge!