In Pakistan, which carried an estimated tuberculosis incidence rate of 266 per 100,000 population in 2024 and ranked fifth globally for overall TB burden, knowing how far people actually travel to reach care matters enormously. A new study published in BMC Global and Public Health set out to measure geographic accessibility to TB service points across the country using two very different kinds of data: in-person exit interviews with visitors at clinics and drug vendors, and aggregated mobile phone movement records from one of Pakistan’s largest telecom providers. The comparison revealed striking gaps between what the two sources capture, with important implications for planning TB case-finding campaigns.
The first were public-sector TB facilities, where diagnosis and treatment are free and usually provided at the same location. The second were private-sector vendors distributing rifampicin, an anti-TB drug used in almost all first-line treatment regimens for drug-sensitive TB. In Pakistan’s sprawling private sector, which includes general practitioners, private hospitals, clinics, pharmacies and even shops, some vendors dispense rifampicin without evidence of a confirmed TB diagnosis and charge variable fees.
For the public facilities, the team drew on the National Tuberculosis Program’s master facility list as a sampling frame, classifying each facility by province or subnational area, urbanicity, and socioeconomic class derived from night-time light data, road density, house sizes, property values and survey data. They aimed to sample roughly 100 public facilities and 100 private vendors, randomly selecting five facilities per province-urbanicity-socioeconomic classification where permissions allowed. In total, 110 private vendors were randomly selected from IQVIA’s list of 27,830 rifampicin-distributing vendors for 2022, aggregated into 1,949 geographic clusters of three to ten vendors for confidentiality. Exit interviews were conducted in October and November 2023, with visitors asked to point out their home location on a tablet map so that straight-line travel distances could be calculated.
The headline finding on distance was modest: across all 200 TB service points visited, the mean of the median distance travelled was 6.77 kilometres, while the median of the median travel distance was just 2.66 kilometres. Median distances varied by province, with larger values in Balochistan at 3.67 kilometres and Punjab at 3.26 kilometres, and the single facility sampled in the Federally Administered Tribal Areas showing the largest median at 20.96 kilometres. Non-Karachi Sindh had the lowest median travel distance at 1.48 kilometres. Distance travelled to public TB facilities was slightly higher than to private vendors, at 2.88 versus 2.48 kilometres, which the authors expected given the broader distribution of vendors.
The core analytical measure was a geographic accessibility rate for each service point: the number of visits from a given concentric distance band, divided by the number of people residing in that band, calculated for rings of 0 to 5, 5 to 10, 10 to 20, 20 to 30 and 30 to 60 kilometres around each point. For the interview data, the denominator came from a gridded population dataset; for the telecom data, it came from the active Telenor subscriber database. The telecom mobility records reflected six months of aggregate movements, from January to June 2023, of Telenor subscribers transiting from up to 60 kilometres away into the area nearest each service point cluster. Both sets of rates were log-transformed for regression analysis.
The decay curves from both sources told a consistent qualitative story: accessibility falls steeply with distance. But the magnitudes differed by orders of magnitude. Interview-based accessibility rates per 100,000 population were 4.78 visits in the 0 to 5 kilometre band, falling to 1.19, then 0.17, 0.10 and finally 0.03 visits in the 30 to 60 kilometre band. Telecom-based rates per 100,000 were vastly higher throughout: 4,282.54 visits in the nearest band, declining to 211.55 in the farthest. The authors attribute this to a fundamental difference in what each source measures. Telecom data count all mobile subscribers moving into a commercial area, whatever their purpose, while interviews counted only individuals visiting each specific service point on a single day.
Statistical modelling sharpened the picture. The team applied Gaussian generalized linear models separately to public facilities and private vendors, and a third binomial model to test whether telecom-based accessibility could predict interview-based accessibility at public facilities. A one-unit increase in the log rate of telecom mobility-based accessibility was significantly associated with an increase of 0.10096 in the log rate of interview-based accessibility after adjusting for geographic and demographic factors. In other words, the two sources were related, but the interview data showed a much steeper decay curve, with lower relative accessibility at each successively distal band. The authors suggest this may mean interviewed people under-report longer travel distances, and it clearly shows that mobility data capture movement of large numbers of people into commercial areas rather than toward specific buildings.
Province mattered. Among public TB facilities, accessibility was significantly lower in non-Karachi Sindh than in Balochistan, the reference category, even after adjusting for data source. Among private vendors, Balochistan again showed the highest accessibility, significantly higher than Karachi. The authors caution that province-level interpretation is difficult given the enormous internal variation in urbanicity, socioeconomic class and service density. Notably, after adjusting for data source, urban public TB facilities showed significantly lower accessibility than rural ones, a counterintuitive result the authors attribute to their catchment definition: a rural facility serving as the sole option within 60 kilometres may be visited by a larger proportion of its catchment population. They warn this finding should not inform urban TB control planning, since it reflects utilization patterns rather than true geographic proximity.
Service point density showed divergent patterns. Among public facilities, accessibility was similar in high and low density areas and lower, though not significantly, in medium density areas. Among private vendors the pattern reversed, with significantly higher accessibility in medium density areas compared to high density ones. No significant differences in accessibility were observed by socioeconomic class or urbanicity in the unadjusted public-sector models. The authors also note limitations: multicollinearity among regressors was low to moderate, with variance inflation factors between 1.5 and 2.7, so coefficients should not be read as independent effects; regression analyses were not specified at the time of data collection, so power was not calculated for them; the 60 kilometre cut-off excluded some longer-distance visitors; and Telenor, while one of the largest providers, does not represent all mobile users.
The practical payoff lies in geoprioritization of TB case-finding services, where early and efficient screening and diagnosis determine efficacy. Telecom mobility data, the authors argue, are not a representative measure of service point accessibility, but they do quantify how many people travel from distal to proximal areas around a facility and could, in theory, be targeted by a screening campaign or public health event in its vicinity. They may also help estimate how accessible a prospective facility is to the daytime commuting population, assuming a visit can be made en route. Since low geographic access has been shown to worsen TB treatment outcomes, and since TB outcomes and accessibility both decline with distance from service points, the study concludes that closing treatment coverage gaps will require bridging long travel distances through community health interventions, sputum transport networks and additional diagnostic services, with modelled travel times contextualized against the method of measurement used.
Subject of Research: Geographic accessibility of tuberculosis service points in Pakistan measured with exit interviews and telecom mobility data
Article Title: Evaluating tuberculosis service accessibility in Pakistan by analyzing population movement patterns from telecom mobility and survey data
Article References: Mergenthaler, C., Ifejube, O. J., Stephens, P., Syed, R., Ng, Y. T., Saif, U., Butt, B., Cobelens, F., & Rood, E. (2026). Evaluating tuberculosis service accessibility in Pakistan by analyzing population movement patterns from telecom mobility and survey data. BMC Global and Public Health, 4(1), Article 97. https://doi.org/10.1186/s44263-026-00329-z
Image Credits: AI Generated
DOI: 10.1186/s44263-026-00329-z
Keywords: tuberculosis, Pakistan, geographic accessibility, telecom mobility data, exit interviews, health services research, travel distance, spatial analysis, TB case-finding, healthcare planning, public health, mobile phone data
News Source: Ophelia Keating. (October 29, 2026). TB Care in Pakistan: Telecom Data and Exit Interviews Reveal Different Travel Patterns. Scienmag.
