Key findings
This study reveals useful findings on antibiotic use and microbiological test results for patients presenting RTIs symptoms at rural and township care settings in rural Anhui, China. It documents an antibiotic prescription rate as high as 87.8% and that 35.5% of these prescriptions contained two or more kinds of antibiotics. The most commonly prescribed antibiotics were levofloxacin, amoxicillin and erythromycin. Nearly one third (30.8%) of the specimens were isolated with pathogenic bacteria or conditional pathogens with K. pneumonia, H. influenza, H. parainfluenzae, P. aeruginosa and S. aureus being the top five bacteria strains. K. pneumoniae witnessed the highest resistance rate to ampicillin followed by S.pneumoniae to Clindamycin, H. influenzae to Trimethoprim/Sulfaisoxazole and H. parainfluenzae to Ampicillin. The study also found, via logistic regression modeling, that bacteria detection was associated with age, season, days since infection onset, runny nose, cough with green sputum and cough with white sputum; while antibiotic use was only linked to days since infection onset and sore throat.
Implications in context of other research and for policy
The above study findings have important implications for antibiotics stewardship. The very high rate of antibiotics prescriptions contradicts a common belief among policymakers in China that excessive antibiotics use is being brought under control as a result of the nationwide Special Antibiotics Use Rectification program (initiated in 2011) and the New Health System Reforms [19–21]. These initiatives focus on antibiotics use at secondary and tertiary hospitals. Given that about 57% of China’s vast population lives in rural and township areas and over 70% of antibiotics prescriptions occur at settings in these areas[22, 23], there is a clear need for added attention on excessive antibiotics use at these settings and communities. Narrow (rather than broad) spectrum antibiotics are recommended whenever applicable. However, the most commonly prescribed antibiotics observed in this study belong to broad spectrum antibiotics and around one third of the prescriptions contained two or more types of antibiotics. This may be attributed mainly to medical uncertainty though decisions on which specific antibiotics to use depends on a variety of factors including availability, price, sensitivity, adverse effects and other characteristics of the antibiotics under concern[24–26]. Rural and township healthcare doctors in China work in a difficult situation in which microbiological tests are unavailable and it is hard to tell the pathogen and its sensitivity to specific antibiotics from clinical symptoms/history. So they tend to view broad spectrum or combined antibiotics as a safer strategy than narrow spectrum antibiotics since the former have greater chance of hitting the actual pathogen [27].
The study sheds new light on relations between clinical symptoms/history, microbiological tests and antibiotics use at rural and township care settings. Our multiple logistic regression modeling of bacteria detection indicates that patients who get RTIs in spring and who present runny nose and cough with green/white sputum, are older, and report a longer duration since onset of the infection, are more likely to test positive for bacterial pathogens. Given that antibiotics work only for patients with bacterial infections, the multiple logistic regression modeling of antibiotics use should result in similar associated factors as that of the aforementioned modeling of bacteria detection. However, our modeling found that only two variables, sore throat and days since onset of infection, were associated with antibiotics use. Of these two variables, one (sore throat) was not associated with bacteria detection and the other (days since onset of infection) showed an inverse correlation. These differences suggest that the doctors’ decision on antibiotics use were not driven mainly by considerations for controlling bacteria pathogens but for assuring the patients or relieving their sense of urgency [28]. Sore throat develops rapidly after onset of infection and is one of the most unpleasant symptoms of RTIs. Attendees with sore throat as a complaint may indicate that: a) their infection is at its early stage; and b) their symptoms are severe or intolerable enough to have driven them to visit the clinic so early. On the contrary, attendees with longer the time-lag between onset of infection and visit to the clinic may have less severe/ intolerable symptoms. Although patients with previous visits to other clinics may have longer time-lag, they account for only a small proportion [29].
The study also reveals a number of clues for leveraging microbiological tests, in rural and township areas, in tackling antibiotics use and resistance issues in China. First, the study has tested the feasibility of conducting microbiological tests for rural and township care attendees in resource-poor rural China. As specified in our separate protocol paper, the testing proceeded by collecting specimens at the rural and township care settings and sending the specimens to a tertiary hospital with a microbiological lab via existing transportation services. We collected and tested 1068 specimens out of 1073 RTI patients. Our overall rate of bacteria detection was 30.8%. This is compatible with published results for similar population groups [30]. These suggest that the testing is acceptable to both patients and physicians and test results are relatively reliable. Second, the study highlights the need for incorporating rural and township cares settings into China’s national antibiotics use and resistance surveillance systems. Our study indicates that antibiotics prescription rate at rural and township care settings are much higher than that from current national surveillance system [31]; while antibiotics resistance rate among rural and township care attendees is substantially lower than that among patients of hospitals forming the national antibiotics resistance surveillance network [31]. Third, the study uncovers various possibilities in using microbiological tests to inform antibiotics prescription by rural and township care givers. For example, the specific strains of bacteria identified and their relative frequencies/compositions by different patient groups may be used to inform more tailored selection of specific types of antibiotics. The findings that antibiotic resistance rates are substantially lower among rural and township care attendees than patients of higher level hospitals and that the list of guideline-recommended antibiotics are all found with enough sensitivity (e.g., a resistance rate of lower than 15% as found in our study) may be used to convince rural and township practitioners that their guideline is valid and their appeal for authorization of more advanced antibiotics is not justified. More importantly, introduction of some monitoring indicators linking microbiological tests with clinical prescriptions may help in promoting a reorientation of decisions on antibiotics use from preventing “rare cases” or coping with patients’ perceived urgency to controlling bacteria pathogens [32]. One of such indicators may be rate of antibiotics prescriptions with microbiological evidence, if it is linked to a policy that requires physicians not to prescribe antibiotics unless the patients are tested with pathogen bacteria.
Strengths and limitations of the study
This study has both strengths and limitations. It is the first study that collected data from healthcare providers and users via a non-participative observation whilst most of the existent research on antibiotics use in China uses data from medical records or reports by medical care givers who may be incentivized to omit recording overuse or misuse of antibiotics so as to meet relevant policy requirements [33]. It is also the first study that performed both microbiological testing and clinical data collection at rural and township care settings and thus enables cross-linking between data from different source. However, the study suffers from limited number of patients and site clinics. It involved only four counties and one village clinic and one township health center from each of the counties. Due to lack of enough number of specific bacteria isolates, the multiple regression models used only collective results (e.g., detection of any bacteria, 7-bacteria, 6-bacteria). In addition, the non-participative observation may also have intervened, to some extent, the routine encounters between the patients and doctors and the prescription behaviors being observed though we had arranged a two-week preparation for each site clinic to allow the field researchers to build trust with the doctors.