Evaluating rural health outcomes: A methodological approach using population-level data

The sustainability of rural surgical and obstetrical facilities depends on their efficacy and quality of care, which are difficult to measure in a rural context. In an evaluation of rural practice, it is often the case that the only comparators are larger referral facilities, for which facility-level comparisons are difficult due to differences in population demographics, acuity of patients, and services offered. This publication outlines these limitations and highlights a best-practice approach to making facility-level comparisons using population-level data, risk stratification, tests of noninferiority, and Firth logistic regression analysis. This includes an investigation of minimum sample-size requirements through Monte Carlo power analysis in the context of low-acuity rural surgical care.

Monte Carlo power analysis was used to estimate the minimum sample size required to achieve a power of 0.8 for both logistic regression and Firth logistic regression models that compare the proportion of surgical adverse events against facility type, among other confounders. We provide guidelines for the implementation of a recommended methodology that uses risk stratification, Firth penalized logistic regression, and tests of noninferiority.

We illustrate limitations in facility-level comparison of surgical quality among patients undergoing one of four index procedures including hernia repair, colonoscopy, appendectomy, and cesarean delivery. We identified minimum sample sizes for comparison of each index procedure that fluctuate depending on the level of risk stratification used.

The availability of administrative data can provide an adequate sample size to allow for facility-level comparisons in surgical quality, at the rural level and elsewhere. When they are made appropriately, these comparisons can be used to evaluate the efficacy of general practitioners and nurse practitioners in performing low-acuity procedures.