Breaking barriers in health surveillance: privacy-preserving techniques for tracking vaccine hesitancy and disease outbreaks in Pakistan.

Résumé

BACKGROUND: Precise measurement of sensitive public health outcomes is often limited by underreporting, stigma, and social desirability bias. These challenges can affect estimates of COVID-19 infection and vaccine hesitancy, leading to incomplete evidence for surveillance and public health decision-making. This study examined the use of privacy-preserving stratified randomized response models (SRRMs) to improve the estimation of sensitive health information in a population-based setting.METHODS: We conducted a cross-sectional quantitative survey in Punjab, Pakistan, during June-August 2021. A total of 1,200 participants were recruited using simple random sampling with replacement within a stratified design, with equal allocation to urban ( = 600) and rural ( = 600) populations. Privacy-preserving SRRM-I and SRRM-II procedures were applied to estimate underreported outbreak cases and vaccine hesitancy while reducing response bias related to confidentiality concerns. Estimates were compared with directly reported responses, and precision was assessed using the percentage relative efficiency (PRE).RESULTS: The empirical estimates reflect the June-August 2021 survey period in Punjab, Pakistan. Estimated outbreak prevalence was higher than directly reported prevalence in both urban and rural populations, indicating underreporting of infection. In urban areas, directly reported COVID-19 cases (10.5%) were lower than privacy-preserving estimates obtained using SRRM-I and SRRM-II (15.3% and 17.4%). In rural areas, directly reported cases (13.7%) were also lower than the corresponding estimates (16.7% and 19.5%). Vaccine hesitancy estimates were also higher under the privacy-preserving procedures (26.3% reported vs. 27.4% and 27.2%), although the difference was considerably smaller than for outbreak cases. All PRE values exceeded 100, indicating improved efficiency relative to the benchmark models. These findings suggest that conventional self-reporting underestimates sensitive public health outcomes, particularly where disclosure concerns are present.CONCLUSION: Privacy-preserving SRRM frameworks can improve the estimation of sensitive public health outcomes, including outbreak underreporting and vaccine hesitancy. The advantage of the two-stage design was most pronounced for the more sensitive outcome, indicating that the truthful-reporting parameter should be matched to the perceived sensitivity of the outcome under study. In settings where respondents are reluctant to disclose health-related information, such methods can strengthen surveillance data quality and support more reliable public health planning and policy decisions.