Home - News ( United States | United Kingdom | Italy | Germany ) - Football scores

Showing content from https://cmu-delphi.github.io/delphi-epidata/api/covidcast-signals/google-symptoms.html below:

Google Symptom Search Trends | Delphi Epidata API

• Source name: google-symptoms
• Earliest issue available: Aug 20, 2017
• Number of data revisions since May 19, 2020: 1
• Date of last change: February 28, 2025
• Available for: county, MSA, HRR, state, HHS, nation (see geography coding docs)
• Time type: day (see date format docs)
• License: To download or use the data, you must agree to the Google Terms of Service

This data source is based on the COVID-19 Search Trends symptoms dataset. We use this data to estimate the volume of web searches related to COVID-19 and H5N1 highly-pathogenic avian influenza (HPAI).

The resulting daily dataset for each location shows the average relative frequency of searches for sets of specific symptoms. The signals are measured in arbitrary units that are normalized for overall search users in the location and scaled by the maximum value of the normalized popularity within a location across a specific time range. Larger numbers represent increased relative popularity of symptom-related searches.

Values are comparable across signals in the same location, but NOT between locations or between geographic region types. For example, within a state, we can compare s01_smoothed_search and s02_smoothed_search. However, we cannot compare s01_smoothed_search between states, or between a state and a county.

Between May 13 2024 and August 6 2024, signal values were much lower compared to previous time periods due to a data outage.

• s01: Cough, Phlegm, Sputum, Upper respiratory tract infection
• s02: Nasal congestion, Post nasal drip, Rhinorrhea, Sinusitis, Rhinitis, Common cold
• s03: Fever, Hyperthermia, Chills, Shivering, Low grade fever
• s04: Shortness of breath, Wheeze, Croup, Pneumonia, Asthma, Crackles, Acute bronchitis, Bronchitis
• s05: Anosmia, Dysgeusia, Ageusia
• s06: Laryngitis, Sore throat, Throat irritation
• s07: Conjunctivitis, Red eye, Epiphora, Eye pain, Rheum
• scontrol: Type 2 diabetes, Urinary tract infection, Hair loss, Candidiasis, Weight gain

Symptoms sets s01-s06 are designed track a variety of COVID-19 systems. They are positively correlated with COVID-19 cases, especially in the period when the Omicron variant was dominant. Symptom set s07 is designed to track novel eye-related symptoms of H5N1. Note that symptoms in scontrol are not COVID-19 or H5N1 related. This symptom set can be used as a negative control.

Until January 20, 2022, we had separate signals for symptoms Anosmia, Ageusia, and their sum.

1. Overview
   1. Symptom sets
2. Estimation
3. Geographical Aggregation
4. Lag and Backfill
5. Limitations
6. Source and Licensing

Each signal is the average of the values of search trends for each symptom in the symptom set. For example, s05_raw_search is the average of the search trend values of anosmia, ageusia, and dysgeusia. Note that this is different from the union of anosmia, ageusia, and dysgeusia related searches divided by 3, because the data volume for each symptom is calculated based on search queries. A single search query can be mapped to more than one symptom. Currently, Google does not provide intersection/union data. Users should be careful when considering such signals.

For each symptom set: when search trends for all symptoms are missing, the signal is reported as missing. When search trends are available for at least one of the symptoms, we fill the missing trends for other symptoms with 0 and compute the average. We use this approach because the missing observations in the Google Symptoms search trends dataset do not occur randomly; they represent low popularity and are censored for quality and/or privacy reasons. The same approach is used for smoothed signals. A 7 day moving average is used, and missing raw signals are filled with 0 as long as there is at least one day available within the 7 day window.

The state-level and county-level raw_search signals for each symptoms set are the average of its individual symptoms search trends, taken directly from the COVID-19 Search Trends symptoms dataset.

We aggregate county and state data to other geographic levels using population-weighted averaging.

For aggregation purposes only, we assign a value of 0 to source regions that have no data provided due to quality or privacy issues for a certain day (see Limitations for details). We do not report aggregated regions if none of their source regions have data. Because of this censoring behavior, the resulting data for aggregated regions does not fully match the actual search volume for these regions (which is not provided to us).

Google does not currently update the search data daily, but usually twice a week. Each update will usually extend the coverage to within three days of the day of the update. As a result the delay can range from 3 to 10 days or even more. We check for updates every day and provide the most up-to-date data.

Between May 13 2024 and August 6 2024, signal values were 25%-50% lower compared to previous time periods. This affected all google-symptoms signals and symptom sets. The drop does not reflect actual search term popularity during the affected period. The apparent decrease in search volume was caused by an outage in the data pipeline on the source side. The data was unfortunately not recoverable and the dip can not be repaired, but data outside the listed time period is unaffected.

When daily volume in a region does not meet quality or privacy thresholds, set by Google, no daily value is reported. Weekly data may be available from Google in these cases, but we do not yet support weekly data.

Google uses differential privacy, which adds artificial noise to the raw datasets to avoid identifying any individual persons without affecting the quality of results.

Google normalizes and scales time series values to determine the relative popularity of symptoms in searches within each geographical region individually. This means that Delphi’s computed symptom set popularity values are NOT comparable between geographic regions or region types, but are comparable within the same location.

Standard errors and sample sizes are not available for this data source.

More details about the limitations of this dataset are available in Google’s Search Trends symptoms dataset documentation.

This dataset is based on Google’s COVID-19 Search Trends symptoms dataset, which is licensed under Google’s Terms of Service.

To learn more about the source data, how it is generated and its limitations, read Google’s Search Trends symptoms dataset documentation.

RetroSearch is an open source project built by @garambo | Open a GitHub Issue

Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo

HTML: 3.2 | Encoding: UTF-8 | Version: 0.7.4