Objective: The aim of this study was to investigate inequalities in primary healthcare visits by using the Gini coefficients of primary healthcare visits (GPVs) as derived from large-scale and region-wide electronic records and to explore estimation and utility of GPV. Design: The study used a cross-sectional study design. It first extracted a random sample of 7.09 million primary care records from 1 October 2019 to 31 December 2021. Then it developed logarithmic models of GPVs using different months of records and performed descriptive and generalised linear mixed-effects regression analyses of the GPVs for all-cause diseases. The study also produced topographic maps of the GPVs for nine selected diseases. Setting: All primary healthcare facilities within the 105 townships or communities randomly selected from Anhui province, China. Participants: All practising doctors working with and all patients presenting to the above sampled facilities during the 27-month study period. Results: The overall rate of primary healthcare visits during the 27-month period was estimated at 147.78%, with intra-regional variation coefficients by all-cause and system-specific disease ranging from 49.02% to 68.96%, and women were more likely than men to seek primary healthcare. The observed GPVs for all-cause and system-specific diseases all fitted very well with logarithmic equations and the goodness of fit increased rapidly when the months of EMRs were extended for the first few months, being over 78.92%, 91.17% and 94.78% for the first 3 months, 6 months and 12 months, respectively. These logarithmic models predicted at least high disparity (GPV>0.4) for all the system-specific diseases when the time period reached 6 months to 8 years. The observed GPVs for system-specific diseases, as estimated using the 27-month electronic medical records, ranged from 0.341 for skin-immune disease to 0.514 for cardiovascular disease. While the observed GPVs for all causes witnessed: great inter-region variations, with the highest GPV being 4.38 times the lowest; a general decreasing trend over the 9-quarter period, being reduced by 18.48% on average; and atypical J-shaped trajectories along age groups for both sexes. The multivariate modelling revealed statistically significant associations between the all-cause-GPV and 10 out of 14 commonly available community-level variables studied. Conclusions: GPVs can be accurately estimated using a limited number of months of EMRs and, guided by the relevant framework, analysis of GPVs can unveil useful clues in addressing unequal primary healthcare utilisation.
Antimicrobial resistance caused by inappropriate antibiotic use has become a global public health crisis. The majority of antibiotics are prescribed at primary care settings which often lack sufficient capacity and surveillance. This study aimed at identifying and testing pragmatic measures and models derived from routinely collected electronic records of primary care encounters to inform continuous antibiotics stewardship. We first extracted a total of 7.097 million records of primary care visits over 25 months from Anhui, China, through stratified random cluster sampling and a minimum set of data about related communities. We then identified pragmatic measures and models for examining antibiotic prescribing at primary care settings through repeated cycles of measure/model identification and relevance analysis. The ‘identification’ used multidisciplinary group meetings while the analysis adopted hybrid methodologies, including descriptive analysis, random forest classification modeling, and data visualization. The study revealed that: (a) antibiotic prescribing rates ranged from 36.82 to 86.25% for the 7 categories of diagnoses studied, including respiratory diseases (RD), digestive diseases (DD), urogenital diseases (UD), skin infections (SI), injuries (IJ), eye infections (EI), and oral and dental diseases (OD); (b) although overall antibiotic prescribing decreased from 67.80 to 47.11% over the study period, the proportion of broad-spectrum added up to 78.96%; (c) the top 10% and 20% clinicians prescribed 59.9% and 80.0% of all the antibiotic prescriptions; (d) 50.8% of the antibiotic recipients received 2 or more antibiotic prescriptions within the 25-months; (e) the AUC of models of antibiotic prescribing ranged from 0.92 to 0.97 for the 7-category- diagnoses, in which the patient, clinician and spatiotemporal variables contributed 0.08 ~ 0.27, 0.30 ~ 0.70 and 0.16 ~ 0.62 respectively. The measures and models derived out of routinely collected electronic records of primary healthcare encounters in this study are both feasible and useful.