Independent variable
Our main independent variable was the status of pairing assistance (PA) of a county. We assigned a dichotomous variable for all people that denoted whether they are in a PA county or not. If a county hospital sent more than 30% (>30%) of its practicing physicians to paired-up PHC facilities within the medical alliance, this country was coded as having PA. If a count-level hospital report that there is no establishment of a medical alliance or send 30% or less (≤30%) of its practicing physicians to its medical alliance, this county was coded as having no pairing assistance (NPA). The medical alliance policy was gradually implemented in different regions, so some counties were still in the process of building alliance when the NHSS survey started in 2018. The distribution of the PA proportion was highly skewed, and thus we used a considerable proportion of doctors to define PA county that represented the county has built a relatively more stable PA mechanism within their medical alliances. It should be noted that the medical alliance was launched between 2013 and 2018, but due to data constraints, the definition of PA was generated from the NHSS 2018 wave.