Where does America search for health?

Google Trends lets us see what people search for at a very local level. This project tracks searches for nine common health issues from cancer to diabetes in every US metro area from 2004 to 2017, and compares them with the CDC's leading causes of death to understand how search data reflects life for millions of Americans.

PythonPandasPostgreSQLFlask APIGoogle TrendsCDCChart.js
210Metro areas
48States + DC
2004–201714 years
9 + 10Health topics · CDC causes of death

Interactive map

Where is interest highest?

Pick a health topic and a year. Darker states search more for it; each dot is one metro area (a city and its surrounding region) at its real location bigger dot, more interest. Scores run from 0 to 100 (see how to read them).

Search interest by state

Highest metro areas

    Results

    What the data shows

    Each finding in plain words first, with the statistics behind it underneath. Everything is calculated live in your browser from the search data and CDC surveys.

      How to read the scores. A Google Trends score is not a count of searches. For each topic and each year separately, Google gives 100 to the metro area where that topic made up the biggest share of searches, and scores every other place relative to it (0 = too few searches to report). So scores tell you where a topic is searched most not how many people searched, which topic is more popular, or whether searching grew.
      New to statistics? Key terms in plain words
      Google Trends score (0–100)
      Not a number of searches. For each topic and year, Google gives 100 to the place where that topic made up the largest share of all searches, and scores everywhere else relative to it. 0 means too few searches to report.
      Metro area
      One of Google's 210 US media markets a city plus its surrounding region (Nielsen DMA). Some span two or three states.
      CDC prevalence
      The share of adults who report having a condition in the CDC's yearly telephone survey (BRFSS).
      Age-adjusted death rate
      Deaths per 100,000 people, recalculated as if every year had the same age mix so an ageing population doesn't distort the trend.
      Correlation (r)
      How closely two things rise and fall together, from −1 to +1. 0 means no link; around 0.3 is a moderate link and 0.5 or more a strong one. A correlation does not prove that one thing causes the other.
      p-value
      How likely a link this strong would appear by pure chance if there were really none. Below 0.05 is called statistically significant unlikely to be a fluke.
      95% confidence interval (CI)
      The range the true value very probably lies in. A narrow range means a precise estimate.
      Controlling for / adjusted
      Taking another factor out of the comparison, so the link that remains is not just that factor in disguise.
      Mean vs median
      The mean is the ordinary average. The median is the middle value when everything is sorted, so a few extreme values don't pull it.
      Partial correlation
      A correlation after controlling for a third factor here, how much a state searches about health in general.
      Principal component analysis (PCA)
      A way to boil many related measures down to a few underlying patterns. Here nine search topics reduce to two main patterns.
      Percentile
      The share of places this one beats. 50 is typical; 90 means more interest than 90% of metro areas.

      State profile

      How does a state compare?

      Average interest in each condition for one state's metro areas, against the average of all metro areas. Tip: click a state on the map.

      Go deeper

      Explore the full analysis

      • Analytics

        Search interest validated against CDC disease prevalence, the north–south structure of health searching, co-occurring conditions, stability over time and mortality trends.

      • City comparison

        Profile any of the 210 metro areas: percentile ranks, change over time and the metros that search most alike.

      • Data explorer

        Filter, sort and export all 2,940 metro × year records and the CDC table.

      • API

        Twelve JSON endpoints the original Flask API, now always available as static files.

      Data quality

      Corrections made in this version

      An audit of the original database found errors that affected the map and charts. All fixes are scripted in tools/clean_data.py.

      Wrong statesFixed

      18 metro areas spanning two or more states were filed under the last state in their name e.g. "Tallahassee FL-Thomasville GA" under Georgia so Washington DC, Rhode Island and Vermont were missing.

      Wrong coordinatesFixed

      23 metros sat outside their own state after geocoding by bare city name: Victoria TX was placed in Australia, Lima OH in Peru, Albany GA in New York. All 210 now sit inside their state on the map.

      CDC typoFixed

      The 2017 cancer death rate read 52.5 instead of the CDC's 152.5 per 100,000 which showed cancer deaths falling by two-thirds in one year.

      Invalid comparisonsRevised

      Summing scores across conditions and years made the first version conclude "cancer is most searched" and "+120% growth". Google scales each separately, so these are revised on the Analytics page.

      How it was done

      Methodology & architecture

      Research question: How have the most searched diseases changed over the years in the US and how does that compare with the real-life leading causes of death?

      1. Extract

        Google Trends health searches by US metro area, 2004–2017 (Kaggle / Google News Lab) and CDC leading causes of death (CDC FastStats).

      2. Transform

        Reshaped and cleaned in Python/Jupyter (Health_Analysis.ipynb); geocoded each metro area.

      3. Load

        Loaded into PostgreSQL (loadData.ipynb) and served through a Flask REST API now pre-built as static JSON so it's always available.

      4. Visualise

        Interactive maps and charts; every statistic on this site is recomputed in your browser from the data.

      Team Eagle

      The team

      Built as a group project in the University of Toronto Data Analytics & Visualization program.

      Ermias Gaga

      Ermias Gaga

      Data scientist & researcher · project lead

      Amanda Qianyue Ma

      Amanda Qianyue Ma

      Economics & analytics

      Amos Johnson

      Amos Johnson

      Data journalism & storytelling

      Adedamola ('Damola) Atekoja

      Adedamola ('Damola) Atekoja

      Chartered accountant & data analyst

      Maria Lorena

      Maria Lorena

      Project manager & data analyst