A probabilistic sampling method (PSM) for estimating geographic distance to health services when only the region of residence is known

Kirsten Beyer, Audrey F. Saftlas, Anne Wallis, Corinne Peek‐Asa, Alexander M. RushInternational Journal of Health Geographics 2011 · DOI

Abstract

The PSM has a number of benefits over traditional geocoding approaches. This methodology improves the precision of estimates of geographic access to services when complete residential address information is unavailable and, by computing the expected distribution of possible distances for any respondent and associated distance confidence limits, sensitivity analyses on distance access measures are possible. Faulty or imprecise distance measures may compromise decisions about service location and misdirect scarce resources.

Introduction

In public health research, measuring the distance between patients and healthcare facilities is critical for identifying barriers to care. However, privacy regulations often restrict researchers to coarse geographic data, such as ZIP codes or counties, rather than precise street addresses. This lack of granularity forces a reliance on proxy locations, which can lead to significant errors in distance estimation and misinformed resource allocation.

The Limitations of Geometric Proxies

The standard approach involves using the centroid (geometric center) of a ZIP Code Tabulation Area (ZCTA) or a ZIP code population center as a proxy for a resident's home. These "quick and dirty" solutions suffer from several flaws:

  1. Geometric Bias: People cluster in neighborhoods; they do not live at the geometric center of a polygon.
  2. Zero-Distance Errors: If a clinic is near the centroid, the calculated distance may be zero, which is physically impossible for the vast majority of residents.
  3. Lack of Uncertainty: These point estimates provide no measure of variance or confidence.

The Probabilistic Sampling Method (PSM)

To address these issues, this study introduces the Probabilistic Sampling Method (PSM). Instead of a single point, the PSM uses Monte Carlo simulations to generate a distribution of 30 possible residential locations within a ZCTA.

Crucially, these locations are selected using "probability proportional to size" (PPS) sampling. By leveraging Census block data, the algorithm weights potential locations based on the population density of the participant's specific demographic group (e.g., women of reproductive age). For each sampled point, the actual road-network distance to the nearest health facility is calculated.

Comparison of travel routes from traditional geographic proxies versus sampled locations

Key Results

The study validated the PSM using data from 1,467 women in Iowa seeking intimate partner violence (IPV) services. The PSM was compared against traditional proxies, revealing systematic biases in the older methods:

  • ZCTA Centroid: Generally overestimated the most likely travel distance.
  • ZIP Code Population Center: Generally underestimated the distance.

As shown in the comparative analysis, the PSM median provides a more stable estimate, while traditional proxies fluctuate wildly depending on the shape and population distribution of the ZCTA.

Comparison of distance estimates: ZCTA Centroid and ZIP Code Center vs. Probabilistic Sampling Median

Beyond point estimates, the PSM is the only method capable of producing confidence intervals (e.g., 93% confidence limits). This allows researchers to perform sensitivity analyses, acknowledging that distance estimates in sprawling rural areas are inherently less certain than those in dense urban centers.

Conclusion

The PSM offers a statistically robust alternative to traditional geocoding proxies. By accounting for population distribution and providing measures of uncertainty, it ensures that decisions regarding service location and resource allocation are based on precise, demographic-aware geographic data. The implementation of this method in R makes it a practical tool for public health researchers facing data privacy constraints.

Summary updated

Abstract

The PSM has a number of benefits over traditional geocoding approaches. This methodology improves the precision of estimates of geographic access to services when complete residential address information is unavailable and, by computing the expected distribution of possible distances for any respondent and associated distance confidence limits, sensitivity analyses on distance access measures are possible. Faulty or imprecise distance measures may compromise decisions about service location and misdirect scarce resources.
BibTeX
@article{beyer2011probabilistic,
  author = {Kirsten Beyer and Audrey F. Saftlas and Anne Wallis and Corinne Peek‐Asa and Alexander M. Rush},
  title = {{A probabilistic sampling method (PSM) for estimating geographic distance to health services when only the region of residence is known}},
  journal = {International Journal of Health Geographics},
  year = {2011},
  doi = {10.1186/1476-072x-10-4},
}
Figure