Hypothesis and study design

.bg-main4.column[.vmiddle.content[
.font1.amber[Aly Lamuri]  
.font1[Indonesia Medical Education and Research Institute]
]]

---

---

.bg-main1.column[.vmiddle.content[
- .amber[Variables: independent and dependent]
- Formulating a hypothesis
- Types of study design
- Assessing hypothesis in published papers
]]

---

<img src="https://i.pinimg.com/736x/16/47/a8/1647a81b28650ac3cf9c40a20819a3a4.jpg" width="80%">
]

---

# What are...?

- .font2[Cases]
  - Observable .amber[subjects]
  - Can be anything
--

- .font2[Variables]
  - .pink[Characteristics] in observed cases
  - Quantifiable

.amber[Ten schizophrenic patients] participated in a study on brain activity. 
--
They underwent computer-based tests on .pink[psychotic symptoms] and fMRI
examination to observe the .pink[functional connectivity].

A group of botanists observed .amber[different species of molds] under
conditioned environments. 
--
They set up different .pink[humidity levels] and
.pink[temperature] to observe .pink[cellular viability].

The Ministry of Finance investigated an urban population. They considered age
and ethnicity as a predictor of socioeconomic mobility.

---

- Unaffected by other variables
- In experiment: under your direct control
- Design-wise: A presumed .amber[cause]
]

- Affected by other variables
- In experiment: under your indirect control
- Design-wise: A presumed .amber[effect]
]

---

# Grouping the cases

.font2[
- Use categorical .amber[independent] variable
- Separate each observation
- Differentiate *within* and *between* subjects
]

## Within?

- Comparing the .amber[**same**] subject
- When we test the subject multiple times
- Temporal differences of `\(t_0\)` and `\(t_1\)`
- Using different measures

## Between?

- Comparing .amber[**different**] subjects
- When we use a test on multiple groups

---

# That's neat! 
--
But why?

# Any catch?

.font2[
- We need to control .amber[confounding variables]
- Reduce .amber[bias]
- Minimize .amber[random error]
]

---

# Confounding variable

.font2[
- Any variable: categoric or numeric
- Causal association with either in- / dependent variables
- .pink[**Reduces**] internal validity!
]

<div id="htmlwidget-64cdbed306c86dc3cd43" style="width:90%;height:30%;" class="grViz html-widget"></div>
<script type="application/json" data-for="htmlwidget-64cdbed306c86dc3cd43">{"x":{"diagram":"\n  digraph confounding {\n    graph [overlap = true, fontsize = 10, bgcolor = \"#ffffff00\"]\n    node [shape = box, fontname = Helvetica, color = lightseagreen, style = filled]\n    {rank = same; Independent; Dependent}\n    node [shape = circle, fontname = Helvetica, color = lightcoral, style = filled]\n    Confounding\n    # Set edges\n    Independent->Dependent [color = lightseagreen, style = bold]\n    edge [color = lightcoral, arrowhead = diamond, style = bold]\n    Confounding->{Independent; Dependent}\n  }\n","config":{"engine":"dot","options":null}},"evals":[],"jsHooks":[]}</script>

---

.column[.vmiddle.content[
<img src="https://i.ytimg.com/vi/0xKklLplngs/maxresdefault.jpg" width="100%">
]]

???

Bias is a systematic error in research
- Selection bias: the proportion of selected subjects does not represent the population
  - Undercoverage
  - Overcoverage
- Information bias: the information does not represent the truth:
  - Recall bias
  - Interviewer bias
  - Response bias (vaguely interpreted questions)
  - Confirmation bias (only confirm what conforms)
- Detection bias: only detecting outcome in a group with certain characteristics
- Performance bias: a tendency to give different treatment to favour an outcome
- Attrition bias: due to drop out and lost to follow-up
- Reporting bias: only report significant findings and neglect insignificant ones

---

# Handling bias

.font2[
- Random sampling
- Random assignment
- Blinding:
  - Participant
  - Investigator
  - Mediating parties
]

---

# Systematic error

.font2[
- Both confounding variables and bias contribute to .pink[systematic errors]
- We have a hunch how they will affect our data
- So we can prevent them
- Often have a mathematical solution
- Example: offset and scale factor errors
]

---

# Random error

.font2[
- A.K.A: unsystematic error, .pink[system noise], random variation
- Unpredictable and unreplicable
- Thus, no mathematical solution
]

# Handling random error

---

.bg-main1.column[.vmiddle.content[
- Variables: independent and dependent
- .amber[Formulating a hypothesis]
- Types of study design
- Assessing hypothesis in published papers
]]

---

# Hypothesis: definition

.font2[
- Assumptions about the population parameter
- Stated as the *null* `\(H_0\)` and *alternative* `\(H_a\)` hypothesis
- Requires a formal procedure to reject `\(H_0\)`
]

## Null hypothesis

## Alternative hypothesis

---

# Hypothesis: example

---

```r
set.seed(1)
coin <- sample(c("H", "T"), 10, replace=TRUE, prob=rep(1/2, 2)) %T>% print()
```

```
##  [1] "T" "T" "H" "H" "T" "H" "H" "H" "H" "T"
```

.font2[
`\(H_0:\)` The probability of having the head is `\(P = 0.5\)`  
`\(H_a:\)` The probability of having the head is `\(P \neq 0.5\)`
]

---

```r
binom.test(x={sum(coin=="H")}, n=length(coin), p=0.5) %>%
	broom::tidy() %>% knitr::kable() %>% kable_minimal()
```

<table class=" lightable-minimal" style='font-family: "Trebuchet MS", verdana, sans-serif; margin-left: auto; margin-right: auto;'>
 <thead>
  <tr>
   <th style="text-align:right;"> estimate </th>
   <th style="text-align:right;"> statistic </th>
   <th style="text-align:right;"> p.value </th>
   <th style="text-align:right;"> parameter </th>
   <th style="text-align:right;"> conf.low </th>
   <th style="text-align:right;"> conf.high </th>
   <th style="text-align:left;"> method </th>
   <th style="text-align:left;"> alternative </th>
  </tr>
 </thead>
<tbody>
  <tr>
   <td style="text-align:right;"> 0.6 </td>
   <td style="text-align:right;"> 6 </td>
   <td style="text-align:right;"> 0.754 </td>
   <td style="text-align:right;"> 10 </td>
   <td style="text-align:right;"> 0.262 </td>
   <td style="text-align:right;"> 0.878 </td>
   <td style="text-align:left;"> Exact binomial test </td>
   <td style="text-align:left;"> two.sided </td>
  </tr>
</tbody>
</table>

.font2[
Seeing the p-value > 0.05, we can .amber[formally] state the failure on rejecting `\(H_0\)`
]

---

---

# Hypothesis: formal test

.font2[
- Does our data correspond to the `\(H_0\)`?
- Sample data are .amber[consistent] with `\(H_0 \to\)` do .pink[not reject] `\(H_0\)`
- Sample data are .amber[inconsistent] with `\(H_0 \to\)` .pink[reject] `\(H_0\)`
]

---

# Type of test

.font2[
- Two-tailed: `\(H_0 \neq H_a\)`
- Right-tailed: `\(H_0 > H_a\)`
- Left-tailed: `\(H_0 < H_a\)`
]

---

# Rejecting hypothesis

---

# Hypothesis: interpretation

.font2[
- We start from the assumption that `\(H_0\)` is true
- Collect evidence to show otherwise
- We then .amber[reject] or .amber[fail to reject] the `\(H_0\)` based on p-value
- When testing the hypothesis we may happen to see .pink[statistical errors]
]

# .pink[Statistical error]

.font2[
- Type I: False Reject Rate `\(\alpha \to\)` false positive
- Type II: False Accept Rate `\(\beta \to\)` false negative
]

---

---

---

# Hypothesis: p-value

.font2[
- Reflect the probability of making a type I error
- Lower p-value `\(\to\)` lower chance of `\(H_0\)` occurred by chance
- Lower p-value `\(\to\)` higher significance
- Common cut-off: .amber[0.05]
]

---

.bg-main1.column[.vmiddle.content[
- Variables: independent and dependent
- Formulating a hypothesis
- .amber[Types of study design]
- Assessing hypothesis in published papers
]]

---

# Study design

.font2[
- Cross-sectional
- Cohort
- Case-control
- Randomized controlled trial
- Systematic review
]

---

# Cross-sectional

.font2[
- A .amber[snapshot] of outcome and associated characteristics
- No intervention, inferencing .amber[existing differences]
- Relatively .amber[inexpensive and quick] to conduct
- .pink[Static,] reveal no temporal context
- .pink[Unable] to establish .pink[causality]
- Different time frame may lead to .pink[different result]
]

---

# Cohort

.font2[
- .amber[Longitudinal] study to see an effect overtime
- .amber[Prospective / retrospective:] observing the future or looking back in time
- Starting from a .amber[potential cause]
- Involving at least .amber[two groups] of interest
- Cannot completely control the .pink[confounding variables]
- No randomization `\(\to\)` lower .pink[external validity]
- .pink[Long] completion period
]

---

# Case-control

.font2[
- .amber[Longitudinal] study to see an effect overtime
- .amber[Retrospective] in nature
- Starting from an .amber[outcome of interest]
- Involving at least .amber[two groups]
- Usually employed to investigate .amber[rare conditions]
- Assessing multiple .amber[risk factors]
- Difficult to find a .pink[suitable] control group
- Inadequate to establish a .pink[diagnostic study]
- Should carefully address .pink[confounding variables]
]

---

# Randomized controlled trial

.font2[
- Randomization to control bias `\(\to\)` higher .amber[external validity]
- Akin to an experimental study, can effectively employ .amber[blinding methods]
- Statistically .amber[efficient]
- .amber[Clearly] identified population
- .pink[Expensive] to conduct
- Risk to have a .pink[volunteer bias]
- Loss to follow up `\(\to\)` .pink[attrition bias]
]

---

# Systematic review

.font2[
- Concatenate findings from .amber[multiple studies]
- Critical appraisal .amber[removes redundancies] and .amber[addresses inconsistencies]
- Delineate where knowledge is lacking to .amber[guide] future research
- Does not establish .pink[novelty]
- .pink[Variation] among published articles is a challenge to overcome
- .pink[Potential bias] for including unreviewed articles
]

# Meta-analysis

.font2[
- Takes systematic review to another level :)
- Analyzing .amber[previous] analysis (thus the name: meta-analysis)
]

---

.bg-main1.column[.vmiddle.content[
- Variables: independent and dependent
- Formulating a hypothesis
- Types of study design
- .amber[Assessing hypothesis in published papers]
]]

---

Source: [10.1371/journal.pone.0165816](https://doi.org/10.1371/journal.pone.0165816)

---

# Reading scientific articles

.font2[
- Look for its underlying .amber[structure]
- Abstract
  - I: Introduction
  - M: Methods
  - R: Results
  - D: Discussion
- Introduction
  - .amber[S:] Situation
  - .amber[P:] Problem
  - .amber[Q:] Question
  - .pink[R: Resolution]
]

???

- By knowing the structure, you can extract information more efficiently
- When reading the abstract, focus on introduction and methods first
- Then shift into conclusions, since it contains the summary
- Finally, you may want to read the results and discussion as well
- Introduction usually elaborates the background and hypothesis
- You can look for the study design in methods
- Discussion part will help you understand current limitation and further suggestions

---

# Abstract - Introduction

.font2[
.pink[Current studies] of depression among people .amber[living with HIV] focus on
describing its .pink[point prevalence]. Given the .pink[fluctuating nature] of
depression and its profound impacts on clinical and quality-of-life outcomes,
this study .amber[**aimed**] to examine the .amber[**prevalence**], recurrence
and incidence of current depressive symptoms and its underlying
.amber[**catalysts** longitudinally] and systematically among these
individuals.
]

---

# Abstract - Methods

.font2[
We conducted a .amber[**prospective cohort**] study between October 1, 2007 and
December 31, 2012 using longitudinal linked data sources. Current depressive
symptoms was .amber[identified] using the Centre for Epidemiologic Studies
.pink[Depression Scale] or the Kessler Psychological .pink[Distress Scale],
first at .amber[**baseline**] and again during .amber[**follow-up**]
interviews. Multivariable regressions were used to characterize the three
outcomes.
]

---

# Abstract - Conclusion

.font2[
.amber[Depressive] symptoms are prevalent and .amber[**likely to recur**] among
people living with HIV. Our results support the direction of Ontario’s HIV/AIDS
Strategy to 2026, which addresses medical concerns associated with HIV (such as
depression) and the .pink[**social drivers**] of health in order to enhance the
overall well-being of people living with or at risk of HIV. Our findings
.pink[reinforce] the importance of providing effective .amber[mental health care]
and demonstrate the need for long-term .amber[support] and routine
management of depression, particularly for individuals at high risk.
]

---

# Abstract - Results

.font2[
Of the 3,816 HIV-positive participants, the .amber[point prevalence] of depressive
symptoms was estimated at .amber[28%]. Of the 957 participants who were identified with
depressive symptoms at baseline and who had at least two years of follow-up,
.amber[43% had a recurrent episode]. The .amber[cumulative incidence] among 1,745 previously
depressive symptoms free participants (at or prior to baseline) was .amber[14%]. During
the five-year follow-up, our multivariable models showed that participants with
greater risk of .pink[recurrent cases] were more likely to feel .amber[worried] about their
housing situation. Participants at .pink[risk of developing] incident cases were also
likely to be .amber[younger, gay or bisexual], and .amber[unable to afford] housing-related
expenses.
]

---

# Introduction - Situation

(par. 1)

.amber[Depression] affects up to half of people living with HIV, a prevalence that is
.amber[two to four times higher] than that found in the general population [1]. Over
50% of people living with HIV and depression do not receive treatment for their
depression [2–9], and this failure to treat contributes to significant .amber[negative]
clinical and quality-of-life .amber[outcomes] [10–14].

]

---

# Introduction - Situation

Growing evidence supports a bi-directional .amber[relationship] between .amber[HIV] and
.amber[depression] involving a number of .pink[biological, psychosocial and social] .amber[factors]
[1,14–16]. The persistent viral presence in the central nervous system may
release toxic viral proteins that induce depression-like symptoms [17,18];
people living with HIV may possess a negative self-image or experience stigma
[1,15,19–21]; and people living with HIV are more likely to struggle with
stressors such as financial insecurity and unstable housing [22–25]. Recent
reviews also suggest that people who suffer from severe mental illnesses
(including depression) and/or co-occurring substance use disorder are more
likely to engage in risky sexual behaviour, thereby elevating their risk of HIV
acquisition [26–34].

]

---

# Introduction - Problem

(par. 3)

To date, most studies about the prevalence of depression among people living
with HIV have used cross-sectional designs [1,15]. Six studies have documented
the incidence [35–38] and persistence (or recurrence) [39,40] of depression
over time among people living with HIV. In Canada, information describing the
epidemiology of depression among people living with HIV is scarce. In Canada,
information describing the epidemiology of depression among people living with
HIV is scarce. There have been two small convenience sample studies describing
the prevalence of depression among people living with HIV. Williams et al.
(2005), employing a small convenience sample of 297 individuals, described the
prevalence of depressive symptoms at 54% among people living with HIV based on
a self-report screening instrument [41]. Logie, James, Tharao, and Loutfy
(2013), employing a sample of 173 Africa, Caribbean, and Black women, described
the prevalence of depressive symptoms as 64% [42]. .amber[Thus, the epidemiology of this condition is not yet well documented in Canada.]

---

# Introduction - Question

(par. 4)

Given the .amber[fluctuating nature of depression over the life span and its profound impacts on clinical and quality-of-life outcomes,] our study aimed to examine the prevalence, recurrence, and incidence of current depressive symptoms longitudinally and systematically among people living with HIV. We also characterized these three outcomes by HIV-positive participants’ socio-demographic characteristics, housing and neighbourhood conditions, substance-use behaviours and health status over a five-year follow-up period.  Understanding change in the burden of depressive symptoms and the underlying catalysts of the condition from a longitudinal perspective would be important to program planners, policy-makers, and health care providers when planning and implementing effective mental-health programs and interventions for people living with HIV.

---

# Introduction - Resolution

(par. 4)

Given the fluctuating nature of depression over the life span and its profound impacts on clinical and quality-of-life outcomes, our study .amber[aimed to examine the prevalence, recurrence, and incidence] of current depressive symptoms longitudinally and systematically among people living with HIV. We also .pink[characterized] these three .pink[outcomes] by HIV-positive participants’ .amber[socio-demographic characteristics, housing and neighbourhood conditions, substance-use behaviours and health status over a five-year follow-up period.] Understanding change in the burden of depressive symptoms and the underlying catalysts of the condition from a longitudinal perspective would be important to program planners, policy-makers, and health care providers when planning and implementing effective mental-health programs and interventions for people living with HIV.

---

# Extracted information

.font2[
- Case: people living with HIV
- Variables
  - .amber[Independent]: depressive symptoms prevalence, recurrence, incidence
  - .amber[Dependent]: younger, gay or bisexual, unable to afford expenses
  - Potential .amber[confounding]: worrying their household situation
- Study design: cohort prospective
- Hypothesis
  - `\(H_0\)`: Depressive symptoms are not likely to recur
  - `\(H_a\)`: Depressive symptoms are likely to recur
]

---

.amber.font5[Query?]

---

# Quick assignment

- Read two articles we briefly discussed in the first lecture
- Identify their underlying structures (abstract and introduction)
- Succintly describe the case, variables and study design
- Determine the hypotheses and formulate them as `\(H_0\)` and `\(H_a\)`
- Make a presentation on the abstract, introduction and extracted information
]

.font2[
Articles:
- [10.1371/journal.pone.0233137](https://doi.org/10.1371/journal.pone.0233137)
- [10.1097/acm.0000000000000134](https://doi.org/10.1097/acm.0000000000000134)
]