From Birth to Breakdown: Postpartum Illnesses and Their Link to Schizophrenia Onset
Advisor: Rob Biemesderfer
Internship Location: Iyer Clinic
Onsite Mentor: Dr. Ravi R. Iyer
Schizophrenia is a psychological disorder whose origins remain largely elusive. Research has revealed that various genetic and environmental factors contribute to its development, but it is hard to pinpoint exactly what triggers its onset. As research on schizophrenia has become more prominent in recent years, focus has also shifted to possible psychological comorbidities in relation to schizophrenia. The comorbidity of interest that this project will examine is postpartum disorders and schizophrenia. Through reviewing existing medical and psychiatric research on postpartum disorders and their neurobiological links to Schizophrenia, and analyzing case studies and clinical data that document instances of postpartum-related psychotic breaks leading to Schizophrenia diagnoses, my research aims to uncover how postpartum mental illnesses function as potential triggering events for the onset of schizophrenia.
Project Blog
Week 0: Introduction!
Hi! My name is Yalini Kathamuthu, and my senior project regards the connection (if there is one!) between postpartum illnesses (i.e, postpartum depression and postpartum psychosis) and the onset of Schizophrenia in women. Of the two disorders, schizophrenia is more widely known by the general public, but its etiology is still largely elusive, which is why I want to investigate it.
The idea to connect schizophrenia with postpartum disorders first occurred to me when recalling a close family member’s postpartum experience. She had experienced mental health challenges during and after her pregnancy and was eventually given a diagnosis of bipolar disorder that later evolved into schizophrenia. This anecdote, alongside the passion for abnormal psychology that I've been fostering since middle school, influenced me to choose the topic I did for my senior project. I’m so excited to dive deeper into it for the next 3 months!
As for my internship, I am still ironing out the details, but it will likely be at a clinic near me called The Iyer Clinic. I’ve connected with one of the Doctors there and am looking forward to gaining some hands-on experience with patient care.
There is also a psychiatric office right next to the clinic, where I might be able to gain insight into my desired field of psychiatry, so I’m also super excited to have that opportunity! That’s a rundown of everything I’ve got planned for the next few weeks. I will (hopefully!) have updates about my internship for my first post.
Till next time,
Yalini
Week 1: Learning Approaches to Meta-Analysis
Hello everyone! In this blog, I’ll discuss my week one progress, where I focused on determining a starting point for my meta-analysis. For the past month, I’ve been re-solidifying my statistics knowledge that I previously acquired in AP Psychology and precalculus through various free online courses and video resources, such as Khan Academy and Crash Course statistics, and I recently watched LearnMetaAnalysis’s video workshop called “How to do your first meta-analysis from start to finish.” Albeit daunting at first with its 3-hour runtime, this video simplified the process for me and made me much more confident in my project overall.
I have started documenting my initial steps (mainly my literature search) in a scoping document and have imported papers from PubMed, APA PsycINFO, and Europe PMC into Zotero, where I’ll eventually put them into Excel or Google Sheets and screen them! Before I can do that, however, I will have to determine my exclusion criteria, which I hope to get done next week, alongside sorting through my whopping 5000+ studies (I plan to narrow it down to ~1000 because some of these studies don't align fully with my original research proposal).
I have chosen three questions to guide my meta-analysis:
What is the association between postpartum psychiatric illness and the subsequent onset of schizophrenia
What factors moderate the relationship between postpartum illness and schizophrenia onset (e.g., illness type, genetic predisposition, time to onset, parity, or sociodemographic variables)?
What is the potential impact of publication bias on the estimated association between postpartum psychiatric illness and schizophrenia onset in this sample of studies?
And will be looking for criteria within each study that will (hopefully) allow me to answer at least one of these!
I have also downloaded R and am looking into coding and possibly AI tools that can help me with the screening process, which is exciting, as I am looking forward to stepping out of my comfort zone and trying out new things throughout my process. One challenge I am navigating through is remembering to document EVERY step of my process (in alignment with PRISMA standards), as I almost forgot to record how many duplicate citations in Zotero I had to merge. Luckily, I had to redo it anyway as I made an error while importing, so I didn’t lose much time!
Moments like these remind me I am still learning every step of the way, but that's what makes it all the more rewarding. Hope to have more updates about my process in the next post!
Till Next time,
Yalini Kathamuthu
Week 2: First Days at the Iyer Clinic
Hello everyone! This week, I have exciting updates about my internship, so I’ll spend the majority of the blog post talking about some takeaways from there. But before that, I’ll talk a little more about the Iyer Clinic. The Iyer Clinic started in November 1997, and opened another satellite office in 2017. My internship is at the original location, where I both help with patient admission and shadow Dr. Iyer. The clinic specializes in ambulatory primary care, with all doctors there specializing in internal medicine. They treat a wide array of illnesses, from diabetes to mental health issues.
I started at the clinic on Thursday, where I quickly met the staff, who were all super kind and welcoming, and immediately got to work. My duties included taking vitals for incoming patients, recording them, and entering them into the computer system. I was nervous at first, but I quickly got the hang of it and gained experience interacting with patients. My background as a certified nursing assistant (CNA), for which I completed certification in the summer before my junior year, definitely equipped me well to do this, as I would’ve been a lot less confident in my ability to take vitals without it. I really enjoy putting my skills to use in a clinical setting; the transition was a little awkward and hectic due to the sheer amount of patients that we had (it was an extremely busy day), but it was very rewarding when my shift ended.
During my time there, I also met two other (college) students who were there for similar reasons, and it was great hearing about their premed experiences and learning skills from them! I’ve learned a lot from observing and seeing how Dr. Iyer treats patients and prescribes medication for them, such as the balance between being stern and understanding, and just how crucial time management and compartmentalization are in a clinical setting. For the privacy of patient information (PI) reasons, I can’t discuss too many details of what I did in the clinic, but in future blogs, I will aim to share at least 2-3 things I’ve learned each week, whether it be a new medicine, a biological pathway, or something as simple as a skill from being at a new workplace.
In terms of my meta-analysis, I am (still!) learning about setting it up, but I am ready to transition into the screening process. I have already screened a few manually, but am looking into AI tools that were recommended in LearnMetaAnalysis’s video workshop called “How to do your first meta-analysis from start to finish,” which has been my guide on this journey. I have begun brainstorming exclusion criteria. My two ideas so far are: studies should be excluded if they do not include a clinical or validated diagnostic measure of either postpartum illness or schizophrenia spectrum disorder, or they do not report or control for pre-existing psychotic disorders diagnosed prior to pregnancy. Most of my time this week was spent preparing for the transition to my internship, so, regrettably, I did not get as much work done on my project as I would have liked, but that's one of my goals for next week! I aim to balance my internship (which is sometimes a 9-5) and my project even on days when I may be exhausted. I'm looking forward to another week of more learning as I get settled into my internship at the clinic!
Week 3: Scrubs and Screening!
Hello everyone! For this week’s updates, I’ll first start with new experiences and things I’ve learned at the Iyer Clinic, then transition into progress on my meta-analysis. Some new things I’ve learned to do at the clinic include requisition forms, physicals, and additional patient system updates (beyond just vitals).
The Iyer Clinic also does lab work, as it is connected to a larger medical chain group in Virginia that has two laboratories, and requisition forms are essentially orders for lab work to be done. I’ve mainly done requisitions for physical/annual appointments, which include the basics (i.e., CBC w/differential) and other, more specific labs (if needed) based on what Dr. Iyer requests. Physicals are another, more complicated thing that I have started to take on (with the help of a nurse). They are complicated since many questions regarding medication, family history, smoking/drinking habits, etc., must be asked and are more high-stakes than regular walk-ins or appointments, but they have made me more comfortable (in general) on calling up patients. On a similar note, I have also learned how to input this extra information into the system and have also learned more about filing patient documents. Overall, this week at the clinic has been extremely productive, albeit tiring. Interacting with patients, nurses, and doctors has been rewarding, and every day I learn more about healthcare, which makes me even more excited to eventually explore more of the field in the future!
For my meta-analysis and to clarify something from my last post: I’ve opted to use AI review systems more as a secondary review (potentially for later phases of screening?), partially because ASReview won't work on my MacBook, but also because I want more control over the studies that make it through Phase 1 of screening. In terms of the actual screening process, I’ve finished phase 1 of screening (and have included a snippet below), based on my inclusion criteria:
Must contain enough information to calculate an effect size
Must be published in English
Must be publicly available
Must examine the relationship between Schizophrenia and postpartum disorders
I ran into a roadblock because I decided not to include any of the studies I collected from Europe PMC, as it was very disorganized and contained mostly duplicates from PubMed, and way too many studies for me to do manually. This led me to have to delete my citation collection on Zotero and redo it, which also led me to redo my literature search for documentation purposes, but all is well now! I plan on incorporating more open-source LLMs, possibly in later screening phases, as I discovered their usage from a meta-analysis YouTube video I stumbled upon. I hope they can also serve as pseudo-second readers, as interrater reliability is optimal for meta-analyses, and, since I am a one-person team, that is a limitation of my study. I hope to have more updates regarding my journey navigating Docker, Python, and LLMs for next week, so stay tuned!
Week 4: A Pivot in Research Design
Hello everyone! For this week’s updates, I’ll mainly be focusing on my independent research outside of my internship, since nothing new has really happened at the clinic this week. I still haven’t gotten deep into LLMs since I'm still looking for resources to guide me; I’ll likely reach out to my father for resources since he has a lot of experience in IT and AI. In terms of my screening process, I am in the process of phase 1.5 of screening/rescreening. The reason I’m doing this second phase is that I ended up with around 20 studies after my last phase of screening, which proves to be a little too small a sample size for my meta-analysis. I’ve chosen to broaden my research question from the association between postpartum (PP) disorders and schizophrenia onset to the association between postpartum disorders and schizophrenia (i.e, both directions). So, if schizophrenia diagnoses lead to a heightened risk of PP disorders, AND vice versa.
As a result, the research questions informing my meta-analysis are:
What is the association between postpartum psychiatric illness and the subsequent onset of schizophrenia? Vice versa?
What factors moderate the relationship between postpartum illness and schizophrenia onset (e.g., illness type, genetic predisposition, time to onset, parity, or sociodemographic variables)?
What is the potential impact of publication bias on the estimated association between postpartum psychiatric illness and schizophrenia onset in this sample of studies?
I have also made some adjustments to the formatting of the sheet I’m using to screen my studies. In my first phase, I am primarily doing title and abstract screening to see if it fits the inclusion criteria of my study (which I’ve discussed in previous posts) or meets any of the exclusion criteria (and thus cannot be included). I’m using a format of 1s and 0s, where 0 = not included and 1 = included in the study. I’ve also color-coded them since my last post, so 0s are red and more noticeable. I am also using the “column stats” feature to keep track of how many studies are included and excluded, which can also be seen in the following snippet!
Unfortunately, this week was a bit slower than I would’ve liked due to the rescreening, but I hope to bear more news about my journey with LLMs next week.
Till next time,
Yalini Kathamuthu
Week 5: ASReview and More Screening
Hello everyone! For this week’s updates, I’ll again mainly be focusing on my independent research outside of my internship, since Dr. Iyer has been out this week and I’ve been taking a break as a result. No worries, though, since I’ll be returning next week and bringing more updates. As a follow-up from last week, I reached out to my father and finally got ASReview to work on my computer, but, as a result, I’ve had to redo screening (again) with harsher parameters. I started screening really early, though, so as a bonus, I still have enough leeway to handle small setbacks like these. Screening again did take me around 2 workdays, so I didn't get to the data extraction (as I had hoped I’d get to), but using an AI-assisted data extraction program on R (an open-source programming language and environment specializing in statistical computing, data analysis, and graphics) should streamline the process.
Of course, as with using any AI tools, I’ll be double/triple-checking everything to ensure the data is correct, but it serves to help make the extraction marginally faster. This whole process has been just learning as I go on, so mistakes are bound to happen, but I am doing my best at logging everything and tracking all the changes I make so my study will fit meta-analysis standards. With ASReview, I still screen everything by hand, but it is much easier to separate texts since they have a really intricate tagging system, which allows me to annotate for studies I want to go back and reference mine or studies that I can't use for my analysis, but would still be good to include in my literature review. It also keeps track of the intervals between each study marked “relevant” and records data during the screening process (all of which I’ve attached below).
I do have a small sample of studies (mainly because my topic is not heavily explored yet), so I do have some concerns about that, but to combat that, I am trying to gain access to more databases and go through the references of some other meta-analyses and systematic reviews. I have high hopes going into data extraction and am looking to have at least one model in the works for next week to share. Below is a summary of all the exclusion criteria and inclusion criteria I used for (hopefully) my last time screening.
Inclusion Criteria:
In order for studies to be included in this analysis, they must:
Must contain enough information to calculate an effect size
Must be published in English
Must be publicly available
Must examine the relationship between Schizophrenia and postpartum disorders
Exclusion Criteria:
Studies were excluded from the analysis if:
Did not fit the inclusion criteria
They do not include a clinical or validated diagnostic measure of either postpartum illness or schizophrenia spectrum disorder (e.g., studies relying solely on self-report without clinical confirmation, or that use non-standardized definitions of postpartum illness such as general "postnatal distress")
The focus was on the effects of a medication used to treat both disorders (or neither of the disorders at all—i.e., wrong focus)
The population used in the study was not human (i.e, mice, etc)
Duplicate/same dataset as another included study
Study type: Case report, editorial, qualitative only, review/meta-analysis
Till next time,
Yalini Kathamuthu
Week 6: New Experiences at the Clinic
Hello everyone! For this week’s updates, I’ll mainly be talking about some new experiences I’ve garnered from the clinic this last week. Recently, the main nurse at the clinic no longer works here, and as a result, I now take all the patients back, ask for and enter all their medical/social history, and write some notes and forms for them, since these were formerly the nurse’s duty. It is great for building more experience in a healthcare setting, but, on a more vulnerable note, it is definitely taxing, leaving me a little more burnt out after long 8-4 shifts. However, another (more positive) development is that I have experienced many more procedures at the clinic, one of which is the pre-op process.
The pre-op process includes EKGs and labs, so I got to watch that process. Before letting the patient go, we also have to ask the doctor about the EKG (after printing it) to ensure the patient is healthy enough for the operation. Another new experience was taking oral temperature. Due to allergies, there has been an influx of sick symptoms, so I’ve been having to do vitals with temperature. The thermometer wasn’t working for me initially, so I had to practice on myself a few times.
Another test I’ve watched is urinalysis. Sometimes in labs, the patients are required to give a urine sample, and we just do a dipstick test and compare the results with the packaging to determine if there are any underlying issues. I’ve also seen some injections being performed, and have, in general, become more familiar with the clinic, staff, and patients! I now have my own (general) ID for the EHR database, so that’s also been a new development.
I've also learned a few things about the administrative side of things, especially regarding different insurances and how they differ in what tests they will cover and how Medicare falls into all of this. For example, sometimes it is advised that patients take specific tests separately (on different appointment days) so that insurance has a higher chance of covering them.
Overall, from learning about how specific health insurances work to operating new medical equipment, I have garnered many new experiences this week at the clinic! I will be back next week for some in-depth updates from my meta-analysis, alongside my reflection on the process.
Till next time,
Yalini Kathamuthu
Week 7: The Merits of Meta-Analysis and Common Misconceptions
Hello everyone! For this week’s updates, I’ll be focusing on a topic my project advisor recommended: the nitty-gritty of meta-analysis (as well as my experience with this process that, at surface-level, seems “easy”). For my meta-analysis, I have put a final close to my data collection process (after many weeks of doubt and rescreenings) and have extracted some of my data. For screening, I have officially completed the PRISMA flow diagram (attached below).
The whole purpose of the flow diagram is to transparently map the flow of information through different phases of a systematic review/meta-analysis, while outlining the number of records identified, screened, included, and excluded. It also serves to ensure the meta-analysis is authentic and replicable.
Now, for the crux of my blog, I want to address the misconception that meta-analyses are easy** and shed light on the merits of meta-analysis in research (think pros and cons, alongside the significance of meta-analysis in various fields of research). Albeit a simple concept on the surface (collecting papers, narrowing down papers, and typing code into R for analysis), there is so much more complexity to it.
Meta-analyses, by nature, are very thorough since they systematically collect and statistically combine results from dozens (and sometimes upwards of hundreds or thousands) of independent studies on the same topic. It’s such a diverse tool, however, and its purpose can’t truly be defined in a few broad terms without it being an oversimplification. But the essence of meta-analyses is to filter out the noise of individual study limitations, such as small sample sizes or methodological specificities, to reveal more reliable patterns in the data. By synthesizing data from multiple sources, the results of meta-analyses are also highly regarded in their given fields since they avoid cherry-picking one-off results. In fact, they can even reveal publication biases within a specific topic.
To me, as a high school student with no access to many paid resources, one of the largest merits or pros of meta-analysis is that they are very accessible. With time and many free resources available online, a meta-analysis can be done by anyone, albeit with some limitations in the application of results. However, with a process so thorough and widely used by funded researchers, there are many hurdles I’ve had to navigate. Although accessible, meta-analysis is by no means easy.
One of the biggest challenges of meta-analysis is getting started. In the literature search and screening process, for more niche topics (such as Schizophrenia and postpartum disorders), where incidence is already rare and undocumented, there might be a distinct lack of papers to even perform a meta-analysis with. Also, while navigating free tools, many small mistakes can lead to having to redo the whole screening process (i.e., not documenting each phase of your literature search) and complicate replicability. The full-text screening phase can also be very time-consuming, as it includes reading full papers and determining if they have enough data to actually include in the analysis. In my case, this led me to narrow down a full 30 papers to only 9, and I had to backtrack and reference-mine even more studies to get back to a stable number of 20 studies (~10 for each directionality, which is the standard for more niche, unexplored comorbidities such as postpartum disorders and schizophrenia).
Finally is data extraction and analysis (which I'm still navigating). I’ve made a coding form that I have to fix prompts for to ensure that the extraction program that I’m using can properly identify regions in the study that have relevant information. However, that is not the end of it. After, I’ll have to navigate moderator variables and determine what effect size metric I’ll use (which is another realm of confusion).
Although difficult, meta-analyses are highly regarded in their given fields and can highlight patterns that have significant implications. As I am still figuring it out, I have no doubt my research direction might slightly shift or change along the way, depending on what is accessible to me, but I hope to create a final product that can reveal something meaningful in the field of thought disorders and postpartum care.
Till next time,
Yalini Kathamuthu
**To clarify, I’m not claiming that many think meta-analyses are easy. In fact, to many researchers, meta-analyses are generally not considered easy, and there are a few interesting papers about the subject, such as Berman & Parker’s 2002 paper “Meta-analysis: neither quick nor easy.” But rather, I am addressing the fact that it's a process that seems simple on the surface, especially with the advancement of AI tools.
Week 8: Days at the Clinic and Data Extraction
Hello everyone! For this week’s updates, I’ll be focusing on some developments at the clinic and the long-awaited data extraction portion of my meta-analysis. The aforementioned developments at the clinic are much more lighthearted and involve getting closer to the staff rather than any developments in patient care. Dr. Iyer, two other staff members (Mrs. Lay and Mrs. Claudia), a training medical student (Dr. P), and I went out for lunch in celebration of the birthdays of Dr. P and me (which were relatively close together). It was something I was not expecting at all, and it was really sweet! Mrs. Lay has been my main support throughout the clinic, helping me behind the scenes, so this was a great bonding opportunity and made me feel a lot more included in the clinic’s work setting. In the clinic itself, I’ve gotten a lot closer to the phlebotomist and the other side’s nurse, so it’s kind of bittersweet to know I’ll be leaving soon.
In terms of data extraction, I’ve officially collected all my effect sizes (in the form of hazard ratios/HR) and variances that I’ll need for analysis and plugging into the statistical program R. All my values can be found below (for both directionalities).
The reason I chose to use the hazard ratio instead of alternatives like the risk ratio (RR) or the odds ratio (OR) is that my research questions work well with time-to-event analysis, and since Schizophrenia and postpartum disorders are relatively rare, the OR, RR, and HR are roughly equivalent due to the rare disease assumption, which is commonly used in epidemiological studies and meta-analyses dealing with rare diseases. I am still keeping OR as a backup since some of the studies have nuances that make it so converting to HR is slightly problematic.
Now, what's left is to perform analysis, which I’ll report back on next week!
Week 9: Data, Data, and more Data
Hello everyone! For this week’s updates, I’ll just be sharing some of the data I’ve collected and plots I’ve created, as well as talking about heterogeneity and what the implications of it are for my specific study. Firstly, heterogeneity in meta-analyses refers to the variability in study outcomes beyond what is expected from sampling error alone. It signifies that studies differ in populations, interventions, or designs, affecting how results are combined. My studies are from all around the globe and have slightly different study designs, so it would makes sense that my heterogeneity would be extremely high (the ideal value for heterogeneity/I2 is 50-75%). Last week, I ended with a heterogeneity of around 99%, which was way too high, causing my HR to not really be significant. After reflecting, I found the heterogeneity to be largely affected by a few Mendelian Randomization (MR) studies that I decided to include in my sample.
MR studies are epidemiological analyses that use genetic variants as instrumental variables to determine if a specific risk factor (exposure) causes a health outcome. Essentially, they don't use real cases to measure occurrence, but rather assign occurrence (at conception) to model what a population would be like. In the end, I decided not to include them in my final HR value calculation (which I believe is statistically justified, and I will explain more about this in my paper) because they overstate the occurrence of schizophrenia (when compared to actual cases in a real population).
To illustrate some of my findings for the Schizophrenia (SZ) to Postpartum Disorders (PPD) directionality, which ultimately produced an HR ≈ 1.29–1.71, depending on what studies were included after leave-out sensitivity analysis. This indicates that schizophrenia precipitates a 29-71% higher risk of postpartum disorders. This result may seem lackluster because of the huge range, but the reason for this is due to one study in particular, Hippman 2025, which had a large HR value of 41.31 and accounts for most of the heterogeneity because it is an outlier, and most of the other studies are in agreement with their HR values. However, the result does not reach conventional significance (p-val < .0001 in any model, primarily due to small k and one highly influential outlier whose inclusion substantially changes the estimate. The 95% prediction interval also ranged from HR = 0.49 to HR = 5.99, indicating that while the average effect was positive, considerable variability exists in the true effect across settings, consistent with the high between-study heterogeneity observed (I² = 72.46%).
To sum it up, there is evidence of a positive association between SZ and PPD (in that directionality), but the application is much more nuanced. Due to the nature of schizophrenia and how different contexts apply diagnoses and treat postpartum disorders, the effect of SZ on postpartum illness is genuinely heterogeneous across real-world contexts, meaning there is not one universal risk ratio that can be assumed for everyone. I’ll talk more about the final write-up and the other directionality in the next blog post, since this one is already quite lengthy.
Till next time,
Yalini Kathamuthu