EPIDEMIOLOGICAL RESEARCH METHODS (PUB HLTH7106HO) 2020
ASSIGNMENT 2
This document contains
1. Assignment question – the study results you are asked to interpret
2. Good Answer Guide
3. Answer Template
-------------------------------------------------------------------------------------------------------------------------
1. Assignment question - the study results you are asked to interpret
Research setting: A longitudinal cohort study involving South Australian pregnant women and their children until 8 years of age.
Research question: What is the effect of preeclampsia on children’s academic achievement at 8 years of age?
Exposure: Children born from mothers who had a diagnosis of preeclampsia.
Preeclampsia was defined as gestational hypertension or postpartum hypertension with proteinuria (24-h urinary protein 300 mg or spot urine protein: creatinine ratio ≥30 mg/mmol creatinine or urine dipstick protein ≥++) or any multisystem complication of preeclampsia. These are standard diagnostic criteria that are used in clinical practice and are defined in the Perinatal Practice Guidelines for South Australia. The reference group comprised all children born from mothers who did not have a diagnosis of preeclampsia.
Outcome: Poor academic achievement.
Children’s poor academic achievement was assessed using the Year-3 National Assessment Program – Literacy and Numeracy (Y-3 NAPLAN). The Y-3 NAPLAN is completed by students when they are approximately 8 years of age and covers five areas of academic performance: reading, writing, grammar, spelling and numeracy. A child is considered to have a poor academic achievement on each Y-3 NAPLAN domain if they score at or below the national minimum standard. In this study, we were interested in the numeracy domain only.
The analysis involved univariable and multivariable regressions. Log-binomial regression models* were used to generate Risk Ratios (RR) that examined the association between preeclampsia and child poor academic achievement on the numeracy domain of Y-3 NAPLAN at 8 years of age.
* Log binomial models generate RRs rather than odds ratios from logistic regression. They are prefe
ed to logistic regression because RRs are generally easier to interpret and are more appropriate when the outcome is not rare (i.e., >10% prevalence).
1
Table 1: Characteristics of the study sample (n=47,092)
Preeclampsia
n = 2,100
No preeclampsia
n = 44,992
Maternal characteristics collected at baseline
Socioeconomic disadvantage (recorded at birth)*
4.57 ± 2.81
4.89 ± 2.85
Maternal ethnicity
Aboriginal
Non-Aboriginal
5%
95%
4%
96%
Maternal education
Year 12 or more
65%
79%
Less than Year 12
35%
21%
Maternal occupation class
Management or Professional
Othe
15%
85%
27%
73%
Maternal age at pregnancy
≤39 years of age
≥40 years of age
91%
9%
96%
4%
Number of previous live births
None
1 or more
33%
67%
39%
61%
Smoked during pregnancy
Smoke
Non-smoke
22%
78%
12%
88%
Attended the recommended number of antenatal visits
Yes
No
90%
10%
94%
6%
Child characteristics at age 8 years
Poor academic achievement on NAPLAN Y-3 numeracy
Poor academic achievement
No poor academic achievement
30%
70%
22%
88%
Values are mean ± SD, or %
*The Index of Relative Socio-economic Disadvantage is a general socio-economic index that summarises different data such as educational qualifications, unemployment, occupational skill, crowding and disability. Higher scores indicate more advantaged areas. Higher scores are better.
Table 2: Unadjusted and adjusted risk ratio (RR) of the association between preeclampsia and child having poor academic achievement on Y-3 NAPLAN numeracy at 8 years of age.
Unadjusted
N=47,092
Adjusted (model 1)
N=43,658
Adjusted (model 2)
N=41,657
RR
(95% CI)
P
RR
(95% CI)
P
RR
(95% CI)
P
Preeclampsia
XXXXXXXXXX, 1.30)
<0.001
XXXXXXXXXX, 1.25)
0.002
XXXXXXXXXX, 1.12)
0.033
Socioeconomic disadvantage
XXXXXXXXXX, 0.93)
0.033
XXXXXXXXXX, 0.99)
<0.001
Maternal ethnicity
- Non-Aboriginal
- Aboriginal
Ref
XXXXXXXXXX, 2.12)
0.001
Ref
XXXXXXXXXX, 1.61)
0.021
Maternal education
- Year 12 or more
Ref
Ref
- Less than Year 12
XXXXXXXXXX, 3.00)
<0.001
XXXXXXXXXX, 2.90)
<0.001
Maternal occupation
- Othe
- Management or Professional
Ref
XXXXXXXXXX, 0.70)
0.003
Ref
XXXXXXXXXX, 0.73)
0.002
Maternal age
- ≤ 39 years
- ≥ 40 years
Ref
XXXXXXXXXX, 1.32)
0.005
Number of previous live births
- 1 or more
- None
Ref
XXXXXXXXXX, 1.10)
0.051
Maternal smoking in pregnancy
- Non-smoke
- Smoke
Ref
XXXXXXXXXX, 1.52)
0.007
Attended the recommended number of antenatal visits
- Yes
- No
Ref
XXXXXXXXXX, 1.29)
0.089
A
eviations: CI, confidence interval; ref, reference category; RR, risk ratio.
*Model 1 is adjusted for maternal ethnicity, maternal occupation and socioeconomic disadvantage.
**Model 2 adjusted for variables in model 1 plus maternal age, number of previous live births, maternal smoking in pregnancy, attended the recommended number of antenatal visits.
2. The Good Answer Guide to Assignment 2
This assignment tests your ability to interpret the results tables shown in the assignment. They are typical of tables that might accompany a published research article.
Your task is to demonstrate that you can interpret these results appropriately. Do not include references. Your assignment MUST be presented using the template above.
Question
A Good Answer Would
1) Interpret the Results section of the paper. (10 points)
Refer to Tables 1 and 2. This should be no more than 500 words or about 4 normal paragraphs.
SELECT particular results from Table 1 to describe in words to
ing to the reader’s attention. NOT write a sentence describing every row in the Table 1.
DESCRIBE in words the unadjusted and adjusted results in Table 2.
NOT describe results as “significant” or “not significant”.
2) Interpret the results for the number of previous live births. (5 points)
This should be no more than 3 sentences or about 60 words.
CONSIDER whether the confounder effect estimate can be interpreted in the same way as the exposure effect estimate.
CONSIDER the causal assumptions used to select covariates for inclusion in the model.
3) Study limitations. (6 points)
This should focus on any aspects of the study that need to be considered in making strong causal inferences about the results. This section should be no more than 350 words or 2-3 normal paragraphs.
CONSIDER potential bias of the estimated effects – confounding, selection and information. This consideration would be SPECIFIC describing, for example, how information bias could be generated by poor recall.
CONSIDER the major assumptions for treating observational data like an RCT – consistency, exchangeability, positivity, and adequacy of the statistical approach including missing data, statistical model choice. Does their analysis match their implied DAG?
4) Comparison with existing literature (5 points)
Use the information provided to place the cu
ent findings in the context of another study on this topic. This should be no more than 3 sentences or about 60 words.
SUGGEST reasons why the cu
ent results are, or are not consistent with the previous study, such as measurement protocols, analytical differences.
5) Conclusion section of an abstract for the paper. (4 points)
Write the sentences that would appear in the conclusion of an abstract for the paper. This should be no more than 60 words or 2-3 sentences.
WRITE one to three sentences providing the reader with a CLEAR, SPECIFIC, and CORRECT interpretation of the main findings.
TAKE every opportunity to be SPECIFIC
NOT say poor academic achievement was just higher or lower but HOW MUCH higher or lower.
3. Answer Template – only submit this template
Name: ___________________________
Student number: a __ __ __ __ __ __ __
Date: __ __ / __ __ / 2018
1. Results section of the paper (10 points). Discuss Tables 1 and 2. Interpret the unadjusted and adjusted effects. This should be no more than 500 words or about 4 normal paragraphs.
Write your answer here.
Number of words = ____
2. Interpret the results for the number of previous live births (5 points). This section should be no more than 60 words or 2-3 sentences.
Write your answer here.
Number of words = ____
3. Study limitations (6 points). This should focus on any aspects of the study that need to be considered in making causal inferences about the results. Consider the three big threats to causal inference from observational data – confounding, selection and information bias. This section should be no more than 350 words or 2-3 normal paragraphs.
Write your answer here.
Number of words = ____
4. Compare your results with the existing literature (5 points).
Place the results of the cu
ent study in the context of the other study on this topic. The only other study of this topic was an observational cohort of 1000 pregnancies. They used mothers self-report of preeclampsia and adjusted for age and sex of child. They estimated the OR for the association between preeclampsia and poor academic achievement (Y3 NAPLAN numeracy domain) to be XXXXXXXXXX% CI XXXXXXXXXXThis should be no more than 60 words or 2-3 sentences.
Write your answer here.
Number of words = ____
5. Conclusion section of an abstract for the paper (4 points).
Write the conclusion of an abstract for the paper. This should be no more than 60 words or 2-3 sentences.
Write your answer here.
Number of words = ____
Page 2 of 2