Short answer

PBHL20003 2022T1 Assessment 3 sample responses.docx

PBHL20003 Social Epidemiology and Statistics Term 1 2022

Notes on Assessment

Question	Marks
1	3
2	4
3	2
4	5
5	2
6	4
7	5
Total	25

Question 1 [3 marks]

The two figures below are taken from the Australian Burden of Disease Study (2018 data)*. The first figure show DALY# rate by life stage and remoteness. The second shows DALYs by disease groups and remoteness areas.

*Australian Institute of Health and Welfare 2021. Australian Burden of Disease Study: impact and causes of illness and death in Australia 2018. Australian Burden of Disease Study series no. 23. Cat. no. BOD 29. Canberra: AIHW. This report provides estimates of the burden of disease analysis for the Australian population in 2018, using the disability-adjusted life years (DALY) measure.

# Definition of DALY: One disability adjusted life year (or 1 DALY) represents 1 year of healthy life lost, either through premature death (‘years of life lost’ or YLL) or from living with an illness or injury (‘years lived with disability’ or YLD).

General comments

This question is testing your ability to read and interpret data and trends related to morbidity, burden of disease and socio-economic factors, and contextualise from a public health perspective.

The reference for the Figure and Table was given. Reviewing the discussion related to the Figure and Table in the Report (ie the section ‘Burden of disease by remoteness areas’ commencing page 111) would help give context and background (and thus guide your responses to the questions 1a-1c. The comments below are general – you may have other insights in your submission.

a) What conclusions do you draw from Figure 8.3?

The burden of DALYS (years of healthy life lost) are higher in remote areas compared to urban regions, and this increases disproportionately with age.

(Refer to the Report!) “Each remoteness area showed a similar pattern of increasing rates of burden in older age groups with Remote and very remote areas having the highest rates across all age groups (Figure 8.3). Inner regional and Outer regional areas experienced similar burden rates for all age groups.”

You would consider these findings and make your own conclusions, paraphrasing to show your understanding of what was being stated. You would reference the Figure. Many of you focused on the increasing DALY according to age group: however, this is expected – the burden of disease will always increase with age, no matter the population. They key concern here (indicated by the chart title!) is the differential increase in burden according to remoteness area, and how this compares between regions as well as age groups.

b) What conclusions do you draw from Table 8.6?

This table reports a lot of interesting information related to burden of disease according to disease group, across levels of remoteness. Look at the rate ratios and rate differences between regions in particular.

Its main message is …’ For most disease groups, total burden rates increased with increasing remoteness. Table 8.6 compares rates in the least remote areas (Major cities) with the most remote areas (Remote and very remote) to show the impact of remoteness for each disease group. For most disease groups, the burden rate was greater in Remote and very remote areas than in Major cities (represented as rate ratios greater than 1).’ Delve into the data findings. Which disease group has the greatest relative difference in disease burden? (kidney and urinary). Which disease group had the greatest absolute difference according to location? (cardiovascular).

c) (From a public health perspective), what factors may be impacting on the burden of disease as shown in the above two figures?

(This is where you could refer to other findings of the Report related to DALY burden and burden of disease by remoteness area, as well as what you have learnt in this Unit about socio-ecological factors shaping health).

Question 2 [4 marks]

A study was conducted of the association of high birthweight (macrosomia) with social determinants of health (SDH) using a specially created Index. Macrosomia is defined as a birthweight of 4500 grams or more at birth. A composite Antenatal Health Index algorithm was calculated, comprising SEIFA*, access to transport facilities, access to antenatal care and extreme weather events impact to create the SDH Index. The SDH Index was classified as 1-3, with 1=low SEIFA, poorer transport access, poorer access to antenatal care, and significant impact from extreme weather events; and 3 = high SEIFA, excellent transport access, good access to antenatal care, and less impact from extreme weather events. The results are in the Excel workbook.

* Socio-Economic Indexes for Areas (SEIFA)

a) Compute the mean (Formulas > Insert Function > AVERAGE) and median (MEDIAN function) for each Antenatal Health Index category.

Before you do the calculations, break down the question, reword if necessary, so you understand what the research is exploring (what is the exposure? What is the outcome?)

Maternal antenatal health index (MAHI)	MAHI 1	MAHI 2	MAHI3
Mean	3462	3297	3037
Median	3300	3000	2850
Quartile 1	2450	2600	2600
Quartile 3	4500	4100	3400
Interquartile range (IQR)	2050	1500	800

b) Examine the boxplots for the three Index categories. Compare the data. Discuss with reference to outliers, IQR and measures of central tendency:

(Requires review of previous weeks’ lectures and activities related to statistics and interpreting data). Remember, a boxplot gives a standardised way of displaying the distribution of data based on a number summary (“minimum”, first quartile (Q1), median, third quartile (Q3), IQR and “maximum”). It can tell you about your outliers and what their values are. It can also tell you if your data is symmetrical, how tightly your data is grouped, and if and how your data is skewed.

Review the data in the Boxplots (Box and Whiskers chart in Excel). Review the definitions for the boxplot components.

Remember, the IQR (the middle “box”) represents the middle 50% of scores for the group. The range of scores from lower to upper quartile is referred to as the inter-quartile range. The middle 50% of scores fall within the inter-quartile range. Looking at the three boxplots:

Box plot 1 MAHI1 in particular is comparatively tall, suggesting a wide range of birthweight values. This decreases steadily across MAHI2 and MAHI3.

The birthweight median also steadily decreases from MAHI1 to MAHI3.

The median and mean birthweight values differ within each MAHI category, suggesting the distributions are skewed (you can see this visually in the whiskers, which are longer in the upper limits). It has an symmetrical distribution.

c) What are these data telling us?

The data suggest that there is an inverse relationship between MAHI and high birthweight (macrosomia): as the MAHI Index goes up, there is less likelihood that the baby will have macrosomia (in this study population).

Nb macrosomic babies are not termed ‘giants’!

d) Are there any considerations or potential limitations that you can identify in making inferences about these data findings? Why?

The study samples overall meet the central limit theorem (CLT) which states that the distribution of sample means approximates a normal distribution as the sample size gets larger, regardless of the population's distribution. Sample sizes equal to or greater than 30 are often considered sufficient for the CLT to hold. However, the MAHI1 group has 29 participants, not 30. And, generally, a larger study sample will minimise risk of error and bias.

The Excel boxplot does not show outliers: however, if you look at the table data you can see a large gap between the two highest values.

There are other factors related to potential error and bias to consider: eg recruitment: how were the participants selected? Were there other potential factors impacting? Was there no significant difference between the three groups of participants apart from the MAHI?

Regardless of these potential limitations, in this study population using the MAHI Index as defined, we can say that there appears to be a strong inverse association between and risk of macrosomia in this population.

·

Question 3 [2 marks]

In a case-control study of osteoporosis, female outpatients over 60 years of age on long-term steroid use attending a high-risk falls clinic were compared to outpatients not on long-term steroid use. Osteoporosis was no more common in women on long-term steroid use than amongst the controls.

Conclusion: Long-term steroid use is not associated with an increased risk of osteoporosis in women.

Do you agree with the study findings? Why? a) What issues can you identify in the way women were recruited to this study?

· The main issue is that the Control group participants were attending a high-risk falls clinic: thus, there may be other contributing factors that lead to similar rates of osteoporosis in the Control group (even if they do not have a history of long-term steroid use) as the Cases.

· In a case-control study, the Cases have the disease or condition under study, and the Controls do not. The Cases’ history of exposure or other characteristics (in this case, long-term steroid use) prior to onset of the disease (osteoporosis), is ascertained. From the above description, it’s not clear if the Control group were also taking long-term steroids. The study design is potentially flawed.

· Other comments that may impact findings: We don’t know how many participants were recruited to the study.

· It’s not stated whether or not the Control group (‘outpatients not on long-term steroid use’) were female only or all genders.

· The age group of the Control group was not stated.

b) Suggest an alternate method of recruitment to address these issues.

A more robust method of sampling (and study design) would be to:

· Draw the sample from a broader target population base

· [And, review inclusion criteria: in a case-control study, participant characteristics should not be statistically different to each other except in the condition being studied).

· A prospective cohort study would typically give a stronger level of evidence by looking at participants who have long-term steroid use (exposure) over time (outcome), reducing possible bias.

· Why would an RCT randomised control trial not be appropriate? (Hint: think about the exposure)

·

Question 4 [5 marks]

A group of researchers examined whether there was an increased risk of agricultural workers experiencing CVA (cardiovascular accident, or stroke) associated with adherence to the Australian standard Workplace health and safety (WHS) principles. The results are reported in the Excel workbook.

a) Calculate the relative risk for CVA according to whether the agricultural workers’ workplace adhered to Aust standard Workplace health and safety (WHS) principles [Excel]. Interpret the findings below in 1-2 sentences.

Relative Risk (RR) = Incidence in exposed/Incidence in unexposed =Ie/Iu = (a/a+b)/(c/c+d) (don’t forget brackets! Order of Operations in maths. And don’t round up the incidence values ie should be 0.020231214/0.003391599, not 0.02/0.00339.

There was a relative risk of 5.97. RR >1 indicates an association between the exposure and the outcome. [RR= 1 indicates no association; RR<1 indicates protective association]

So, a RR of 5.97 suggests that a significant association between NO adherence to Aust standard Workplace health and safety (WHS) principles and CVA. Agricultural workers in workplaces with NO adherence to Aust WHS principles are nearly 6 times more likely to experience CVAs than agricultural workers compared to where there is workplace adherence to WHS principles.

b) What is the (absolute) risk difference in CVAs in this population according to whether the workplace adhered to Aust standard Workplace health and safety (WHS) principles or not? [Excel] Interpret your answer below in 1-2 sentences.

Absolute risk difference (=excess risk) = Incidence in exposed – incidence in unexposed =- 0.0168 (or 16.84 per 1000 person-years of followup)

For every 1,000 person-years there are 16.8 extra CVAs for agricultural workers in this study where there was no adherence to WHS principles compared to where there was adherence.

c) Now calculate the proportion of CVAs in this population that would not have occurred where the workplace adhered to Aust standard Workplace health and safety (WHS) principles (attributable fraction) [Excel]. Give an interpretation of your answer below (1-2 sentences).

Attributable Fraction AF (or Attributable Risk or Attributable proportion) = ((ie-iu)/ie)

The attributable fraction is 0.832358093 or 83.2%. In this study population, 83.2% of CVAs would not have occurred if adherence to Aust standard Workplace health and safety (WHS) principles had been followed. This suggests that not adhering to WHS principles for agricultural workers in this population is associated with significantly higher risk of CVAs.

d) What is the overall incidence of CVAs in this study population? [Excel; answer below in words]

total CVAs/Total study population = B5/D5 = 0.003978446 (or 3.98 per 1000)

3.98 per 1000 population 0.4%

Question 5 [2 marks]

How can the concepts of relative risk and attributable risk help shape health policy and public health programs? Illustrate with an example.

Relative risk (RR) compares cumulative incidence between populations exposed and not exposed to a given (often risk, may be protective) factor. Relative risk tells us how much greater the incidence of a certain condition is for one population as compared to another one. This value helps us to determine whether or not there is likely to be an association between a certain risk/protective factor and the condition of interest, and therefore whether intervention targeted at a specific risk/ protective factor is likely to have an effect on a population’s health outcomes.

Attributable risk (AR) shows the difference in the incidence between exposed and unexposed groups. Where RR allows to understand whether there is an association between an exposure and disease/ protective risk, AR allows to see how the proportion of disease cases could change if the considered exposure factor is eliminated. Attributable risk provides an even more practical, or tangible interpretation of risk values; what does this data mean in terms of impact of action or intervention? It provides a measure of the magnitude of the public health impact of an issue by telling us how many cases of the condition of interest could be avoided if exposure to the target risk factor was eliminated. Since decisions need to be made about how to distribute finite resources, measures of attributable risk help us to decide which target may have the biggest impact, therefore what is the most effective and efficient way to use scarce resources.

RR and AR help shape health policy and public health programs by identifying preventable risk factors and assessing the impact of mitigating them. They can also be used to assess the effectiveness of certain treatments in populations.

Example 1: An exposure to a specific chemical at work is associated with a large relative risk of a blood cancer. AR can show that incidence in the population can be reduced substantially if this particular exposure is eliminated. Accordingly, policies could be implemented that oblige manufacturers to impose measures reducing or eliminating workers’ contact with the said chemical.

Example 2:Consider a data set that compares smoking behaviours and the rate of lung cancer in a population. It’s been repeatedly shown over the years that there is an association between these factors, so we can assume that the relative risk calculations would provide us with a value greater than 1 (suggesting that exposure to smoking increases the risk of developing lung cancer). This result would tell public health professionals that smoking is a behaviour of interest that can be addressed if we want to decrease the rate of lung cancer in the community. Calculating the attributable risk would then help to inform decisions about policy planning by telling us how much of an effect we can expect from targeting this behaviour – ie. is the expected improvement in health outcomes worth the resource input into the program? How many cases of lung cancer can be avoided by reducing exposure to cigarette smoke? Or is there a better target behaviour/risk, that may provide a larger decrease in rates of lung cancer that can be the basis of intervention instead?

Question 6 [4 marks]

Results of a screening test to detect the likelihood of lung cancer in the detectable pre-clinical phase (DPCP) through Low-dose CT screening for workers using high temperature insulating wools have recently been published. You are Director of Public Health advising the government whether or not you recommend an implementation of this test at the state level. The results of the published data are in Excel.

a) Calculate the sensitivity and specificity of the screening test in Excel, and write your answers below:

(Refer to the lecture and activities on Screening. Create a 2x2 table. Screening test positive is entered on the top row and the disease positive (in this case, lung cancer) is placed in the left hand column). So, the true positive for those who tested positive in the screening test is 28.

	Lung cancer	No lung cancer	Total
Test positive	28 CELL A	29,540 CELL B	29,568
Test negative	30 CELL C	35,675 CELL D	35,705
Total	58	65,215	65,273

Screening test	Formula
Sensitivity	a/(a + c)
Specificity	d/b + d
False Positive Rate	1-specificity
False Negative Rate	1-sensitivity
Prevalence of lung cancer	total cases/ total population
Positive Predictive Value	a/(a+b)
Negative Predictive Value	d/(c+d)

See https://sphweb.bumc.bu.edu/otlt/mph-modules/ep/ep713_screening/EP713_Screening5.html#headingtaglink_5 for more info

Sensitivity of the test is defined as ability of the test to identify a true positive = 0.4828 = 48.28%. The percentage of people who had a positive screening test and confirmed lung cancer ‘true positives’ is 48.3%.

Specificity of the test is defined as the ability of the test to identify a true negative = 0 0.5470 = 54.70%. The percentage of people who had a negative screening test and confirmed NO lung cancer ‘true negatives’ is 54.7%.

b) Calculate the false positive and false negative rates in Excel, and write your answers below:

False Positive Rate = 45.30% (1-specificity)

False Negative Rate = 51.72% (1-sensitivity)

(not ‘Fake’ Positive rate!)

c) Would you recommend the government adopt the implementation of this screening test? Why? Discuss your reasons in 2-3 sentences, with reference to your results:

I would not recommend that the government adopt this screening test. Both sensitivity and specificity are low around 50%, which implies that their ability to identify true presence or absence of the condition (in this case, lung cancer) is not much better than chance.

Also of concern is the false positive rate of 45%. This implies that 45% of the tests that come back positive are in those who do not in fact have lung cancer. This test, therefore, does not provide us with useful information, and is not an efficient or effective way to screen for cancer. One practical implication of these results would be that almost all screened individuals would be sent for further testing and investigation, when a significant number of diagnostic tests would show negative. This would be a significant waste of resources, not to mention the unnecessary stress and concern for the patients.

Generally speaking, a good test is the one that has high sensitivity and specificity. Power et al. (2013) stated a rule of thumb that a good test would have a combined sensitivity + specificity value of at least 1.5. Our test has the value of 1.02 which is much lower. Further, the test seems to have a rather poor positive predicting value of just 0.09% (Trevethan 2017). Therefore, I would not recommend an implementation of this test.

References

Power M, Fell G, & Wright M 2013, ‘Principles for high-quality, high-value testing’, BMJ Evidence-Based Medicine, vol. 18, pp. 5-10. Trevethan R 2017, ‘Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice’, Frontiers in Public Health, vol. 5, pp. 307-315.

d) What is the prevalence of lung cancer in the screened population? [calculate in Excel and write below]

Prevalence = 0.00089 or about 89 per 100,000 population (or 8.9 per 10,000 or 0.89 per 1,000 population ie 0.09% of study population)

Question 7 [5 marks]

Download the following research paper from the Moodle assessment page: BRYSON, H., MENSAH, F., PRICE, A., GOLD, L., MUDIYANSELAGE, S. B., KENNY, B., DAKIN, P., BRUCE, T., NOBLE, K., KEMP, L. & GOLDFELD, S. 2021. Clinical, financial and social impacts of COVID-19 and their associations with mental health for mothers and children experiencing adversity in Australia. PLOS ONE, 16, e0257357.

Read the paper carefully and then answer all questions below:

From the paper, Background/Intro: WHAT IS KNOWN: “Research examining the impacts of the COVID-19 pandemic and public health restrictions suggests mental health difficulties amongst parents and children have increased” AND “Australia’s lockdown measures led to widespread job and income loss.” AND “The nexus between financial hardship, such as parent unemployment and loss of income, and poorer child and family mental health is well researched”. What is NOT KNOWN/NOVEL: “… lockdown has affected families experiencing adversity; including both risk and protective factors that have emerged”

a) What type of study design is it?

Prospective cohort study

“This was an Australian cohort study drawing on data collected within the ‘right@home’ randomised controlled trial of nurse home visiting.” “The strengths of this study are the unique, prospective cohort of mothers with young children, who were recruited during pregnancy for their experience of adversity.”

b) What is the PICO that informs the research question?

P (population) = Women enrolled into the right@home trial, who were less than 37 weeks gestation into their pregnancy at the time of recruitment, attending the antenatal clinics of 10 public maternity hospitals in metropolitan and regional areas of Victoria and Tasmania, Australia, between 30 April 2013 to 29 August 2014; and self-reported two or more of 10 antenatal adversity risk factors (see p5 for full details)

I (intervention/ prognostic factor) = exposures = predictors of outcome. COVID-19 impacts including Clinical exposure, Financial circumstances, Family stress and resilience (Table 1)

C (comparison) = none

O (outcome) = Maternal and child mental health (Table 4)

c) What are the methods?

Methods (apart from research question and study design) includes RECRUITMENT, DATA COLLECTION, ANALYSIS.

RECRUITMENT: The study was nested within the ‘right@home’ trial with families’ experiences of the COVID-19 pandemic and restrictions as well as mothers’ and children’s mental health either a) assessed as part of the 6-year assessment or b) those who had completed the 6-year assessment before 6 May 2020 (ie children aged 6.1–7.2 years)

DATA COLLECTION: Participants were invited to complete a stand-alone COVID-19 survey.Mothers reported COVID-19 impacts, their own mental health (Depression, Anxiety, Stress Scales shortform) and their child’s mental health (CoRonavIruS Health and Impact Survey subscale).

ANALYSIS: Associations between COVID-19 impacts and mental health were examined using regression models controlling for pre-COVID-19 characteristics.

(From Abstract) “Participants were mothers recruited during pregnancy (2013–14) across two Australian states (Victoria and Tasmania) for the ‘right@home’ trial*. A COVID-19 survey was conducted from May-December 2020, when children were 5.9–7.2 years old. Mothers reported COVID-19 impacts, their own mental health (Depression, Anxiety, Stress Scales shortform) and their child’s mental health (CoRonavIruS Health and Impact Survey subscale). Associations between COVID-19 impacts and mental health were examined using regression models controlling for pre-COVID-19 characteristics.”

* the right@home trial was an Australian cohort study drawing on data collected within the ‘right@home’ randomised controlled trial of nurse home visiting. Ie this study of impact of COVID-19 on maternal and child mental health was based on another study.

d) What are the main findings? (Summarise in 2-3 sentences)

“Of the 406 women enrolled in the extended 5- to 8-year follow-up [of the right@home study], 319 (79%) completed the COVID-19 data collection either as part of their routine 6-year follow-up (n = 123) or as a stand-alone survey (n = 196)”

“The majority of families were living in Victoria (n = 210, 66%) or Tasmania (n = 97, 30%) at the time of the survey. Pre-COVID-19 characteristics showed mothers were experiencing high levels of adversity in the year before the pandemic: 53% were not in paid employment, 41% received their main source of household income from a benefit or pension, 21% had not completed high school or any further education and 27% were not living with another adult”.

“…high proportions of mothers reported negative financial and social impacts including job and income loss, worries about becoming infected, stress related to changes in contact with family and friends, and difficulties managing usual duties in addition to children’s at-home learning. Many mothers also reported positive social impacts such as their family finding good ways of coping and supporting others in the community. Of these, both financial impacts and greater family stress were associated with poorer maternal and child mental health, while greater family resilience was associated with better mental health.”

Comments: some of you gave regression coefficients (β) and z-scores. While this is not incorrect, we haven’t gone through full statistical tests including regression: more relevant at this level is that you understand whether or not statistical test results were significant, and their implications.

e) What are the conclusion and implications? (3-4 sentences)

The impact of COVID-19 lockdowns in this vulnerable population of mothers and children exacerbated inequities arising from adversity and poorer mental health. A stronger investment in income support and universal and equitable access to family health services is critical.

“The financial and social impacts of Australia’s public health restrictions have substantially affected families experiencing adversity, and their mental health. This cohort of mothers and children were already at disproportionate risk of poor mental health prior to the pandemic, with now potentially worsening inequities as we have seen globally; even without the impact of the virus itself. Unless the financial and social consequences of lockdown are addressed, the inequities arising from adversity are likely to be exacerbated by this crisis. To recover from COVID-19, the economic and healthcare needs of women and children living in adversity must be prioritised. Policy investment in income support and universal and equitable access to family health services are critical.”

PBHL20003 assessment 3 T1 2022 Page 2