TL;DR
This study leverages internet search queries to identify potential under-reported adverse events following colonoscopies, particularly respiratory symptoms like cough and pneumonia, which appear to peak about a week after the procedure. This suggests a need to investigate these late-onset symptoms as a potential public health concern, especially with the increasing use of anesthesia during colonoscopies.
What Problem Does This Paper Address?
Colonoscopies are a cornerstone of colorectal cancer screening in the US, generally considered safe with rare severe complications. However, milder, yet clinically significant, adverse events that don't necessitate hospital visits are difficult to track using traditional methods like medical records or insurance claims. These milder events, such as respiratory issues, might be under-reported. The authors aim to use a novel approach—analyzing internet search queries—to capture these less severe, self-reported symptoms and understand their timing relative to colonoscopies. This is particularly relevant given the rise in anesthesia-assisted colonoscopies, which may increase the risk of respiratory complications like aspiration pneumonia.
Methods
The researchers utilized data from Bing search queries made by users in the United States over a 12-month period. They filtered queries to identify those containing "colonoscopy" and any queries made by users who searched for the term. A key step was identifying queries that allowed for the estimation of the colonoscopy date (e.g., "colonoscopy in X days," "colonoscopy X days ago"). For users with an identifiable colonoscopy date, the study then analyzed subsequent search queries for medical symptoms.
To quantify symptom reporting, they developed a "Symptom Ratio Time Series" (SRTS). This metric normalizes symptom query frequency relative to the colonoscopy date, accounting for baseline query rates and general symptom querying behavior. They focused on symptoms with statistically significant SRTS values over at least three days, and specifically examined those queried by at least 75 individuals.
Datasets
The primary dataset used was anonymized Bing search query data from users in the United States between October 1, 2017, and September 30, 2018. This dataset included query text, timestamp, and an anonymized user identifier. User age and gender were available for a subset of users (23%).
Key Insights and Ideas
- Internet Search as a Proxy for Self-Reported Health: The study demonstrates the potential of using large-scale internet search data to identify and track self-reported health concerns that might otherwise go unnoticed in traditional healthcare data.
- Late-Onset Respiratory Symptoms: A significant finding is the observed increase in queries related to respiratory symptoms (cough, pneumonia) approximately one week after colonoscopy. This suggests these might be under-reported, late-onset adverse effects.
- Timing is Crucial: The analysis highlights that while most colonoscopy-related queries occur before the procedure (often concerning diet), a substantial portion (28%) occur after, indicating patient concern about post-procedural well-being.
- Anesthesia's Role: The timing of these respiratory symptoms, coupled with the increasing use of anesthesia during colonoscopies, raises a flag for potential aspiration-related complications, even if mild.
Key Results
- Out of 641,223 users mentioning "colonoscopy," 7,013 (1.1%) had queries allowing for the estimation of their procedure date.
- Of these, 28% of users queried symptoms after their colonoscopy.
- A notable finding was a statistically significant increase in queries for respiratory symptoms (cough, pneumonia) around 7-10 days post-colonoscopy. Specifically, 2.6% of users queried respiratory symptoms, with cough at 1.2% and pneumonia at 0.6%.
- Other symptoms like bloating, nausea, and stomach pain tended to cluster in the first few days post-procedure.
Limitations and Future Work
The authors acknowledge several limitations:
- Veracity of Queries: They could not confirm the accuracy of self-reported colonoscopy dates or symptom queries via medical records.
- Hypothetical Queries: Users might be querying on behalf of others or hypothetically.
- Selection Bias: Only a small fraction of users had identifiable colonoscopy dates, and an even smaller fraction queried symptoms afterward. These users may differ significantly from those who did not query symptoms.
- No Control Group: The study did not include a control group of individuals who did not undergo colonoscopy, making it harder to definitively attribute all observed symptom increases solely to the procedure.
Future work should focus on:
- Measuring the incidence and risk factors for cough and other respiratory symptoms following colonoscopy.
- Recording the type of sedation used during procedures.
- Employing prospective questionnaires to gather data.
- Including a non-colonoscopy control group to better isolate procedure-specific effects.
- Investigating symptoms that may not reach the threshold for hospitalization.
Summary updated