Obesity was 33.7% in neighbourhoods with the fewest fast-food visits and 44.4% in those with the most. The phones did not show what each patient ate

Counting restaurants near a home does not tell us whether anybody visits them. A Houston study took a different approach: it used anonymised mobile-device movement data to estimate what share of all food-related visits in each census tract went to fast-food establishments, then linked that neighbourhood measure to electronic health records.

The analysis included 1,033,824 adults in the Houston Methodist Cardiovascular Disease Learning Health System Registry. Each patient was assigned the fast-food visitation share of the census tract where they lived. Obesity was defined as a body mass index of at least 30 kg/m², using the first available height-and-weight measurement in the record.

The gradient was sizeable. Obesity prevalence was 33.7% among residents of tracts in the lowest quarter of fast-food visitation share and 44.4% in the highest quarter — a difference of 10.7 percentage points. In a multilevel model, the highest-quarter group had 42% greater odds of obesity before adjustment and 40% greater odds after adjustment for demographics, neighbourhood social vulnerability and smoking. The fully adjusted odds ratio was 1.40, with a 95% confidence interval from 1.32 to 1.49.

The researchers also compared their neighbourhood measure with two area-level estimates. Fast-food visitation share had a modest positive correlation with the proportion of registry patients who had obesity (r = 0.28), and a weaker correlation with the US Centers for Disease Control and Prevention's PLACES estimate (r = 0.17). Both correlations were statistically significant, but neither is close to a one-to-one relationship.

What the study cannot tell us matters. The movement data describe the combined visiting pattern of a neighbourhood, not a named patient's journeys, purchases, portion sizes or diet. Someone linked to a high-visit tract may never eat fast food. The medical record and mobility snapshot were analysed cross-sectionally, so the work cannot establish whether the food environment contributed to weight gain, whether communities with more obesity developed different restaurant patterns, or whether other unmeasured factors shaped both.

The registry is also a health-system population rather than a random sample of Houston, and one BMI measurement cannot show a person's weight trajectory. Statistical adjustment reduces some differences between neighbourhoods but does not erase income, work schedules, transport, food prices, marketing, safety, access to supermarkets or other possible explanations.

The practical value is therefore at community level. Actual visits may add information that a simple map of nearby outlets misses, and that can help planners decide where to investigate affordable food access or support healthier defaults. It is not a score for judging residents, and it does not turn an area-level association into personal nutrition advice. For an individual, repeatable meals, health needs, medicines, budget and culture remain more useful starting points than a postcode statistic.