Assessment Of Dots Treatment For Rural Pulmonary Tuberculosis Patients Based On Bayesian Network

Tuberculosis (TB) is a serious chronic communicable disease caused by Mycobacterium tuberculosis (M.TB). The results of the fifth National TB Epidemic Survey in2010show that the number of new onset TB patients is about1.3million per year, accounting for14.3%of the world’s, which remains the second severest in the world, and the number of new onset MDR-TB patients is about120thousands per year. Most of the TB patients are young male, live in poor rural areas in middle and west parts of China. It is estimated that80%of the new-onset TB patients live in rural areas, and the TB epidemic condition is most serious in the rural area.In1994, World Health Organization (WHO) developed the DOTS strategy, since then, the DOTS strategy is considered to be the best method in TB control with lowest costs and best effects. No matter in the developing countries or in the developed, the coverage rate of DOTS strategy is high. In China, the coverage rate in county unit has reached100%in late2005, with the discovery rate of new smear-positive cases of79%and the cure rate of91%. However, the TB prevention is still an urgent task in China, and the incidence is still high compared to other countries in the world. The WHO recommends bacterial clearance of a patient as the primary indicator to evaluate the curative effect of pulmonary tuberculosis. But in china, the challenge of TB prevention is the shortage of instruments, health practitioner in rural area, so smear-positive rate is low and sputum culture is hard in remote rural area. Furthermore, high quality of sputum specimen could not provided by patients. For that, the effect of DOTS strategy by bacterial assessment is not practical in rural area. So we combined imaging examination of pulmonary tuberculosis as the diagnostic standard in China. Of all imaging examinations, tuberculosis X-ray examination is a routine method, which is a indispensable standard criteria for the diagnosis and treatment of tuberculosis patients. Due to the characteristics of sputum culture in China, X-ray examination is of great value.In this study, the subjects are selected from the new tuberculosis patients in Linyi City in Shandong province who have received DOTS treatment. To get the curative effect, the results of X-ray examination are in three levels:the imaging prognosis, risk factors of X-ray changing and individualized imaging prognosis of TB patients. The aim of this study is to explore the feasibility of using X-ray examination to assess the effect of the DOTS strategy, to analyze influencing factors of curative effects, and then using Bayesian networks to propose individualized curative effect assessment after DOTS, and to provide new ideas on estimation of TB curative effects.Results:1.1037TB patients are included in this study. The sex ratio is2.95:1. There are818patients in15-64age group, accounting for78.88%.949live in rural area, accounting for91.51%. The results of rank sum test show that gender, age and occupation are statistically different in X-ray changing.2. The rank sum test, gender, age, and patient outcomes in the shadow of the lesion has a relationship, female patient outcomes better than male patient outcome, older patients less effective.3.For curative plan,888take2HRZE/4HR (85.63%) and144take2H3R3Z3E3/4H3R3plan (13.89%), which are recommended by WHO, and only5patients take other curative plans. Curative plan is not statistically different in X-ray changing.4. The results of logistic regression analysis showed that:1) the different treatment schedule had no significant effect to the treatment effect(p>0.05), and the cavity form of the patients also had no effect on the treatment effect of the lesion (P>0.05);2) delayed treatment was a risk factor for the treatment effect in a univariate logistic regression analysis, while this effect disappeared after adjusting other factors in a multivirate logistic regression analysis;3) in progressive stage, the result of univariate logistic regression analysis showed that lesions forms in all six lung field were risk factors for the treatment effect (P<0.05), with OR values greater than1, while in the adjusted multivirate logistic regression analysis, the lesions forms of all the lung field except the left middle and left lower lung field were risk factors for the treatment effect4) in the stable stage, the result of univariate logistic regression analysis showed that except the lesion form of left upper lung field, the lesion forms of the remaining five lung field were all risk factors for the treatment effect of the lesion (P<0.05), while in the adjusted multiple logistic regression analysis, the lesion form of all the lung field except the right upper filed had no significant effect to the treatment effect.5) age was a risk factor for the outcome of the lesion(p<0.05)5. We conducted a Bayesian Network with the tuberculosis lesion cavities shadow, age, treatment delay time, and treatment strategies of the patients, which has a good performance with AUC equals to0.88.6.Through the network causal reasoning we found that:(1) people in progressive stage who had lung shadows in the upper right and left middle lung field had better therapeutic effect than those had lung shadows in other lung fields (RR>1), while people in progressive stage who had lung shadows in the right middle, low right, upper left, left lower lung field had poor therapeutic effect than those had lung shadows in other lung fields (RR<1);(2) the improvement rate of patients with six lung fields’lesions in progressive stage was sorted by the improvement rate, right upper lung field was84.14%,82.97%for the right middle lung field, right lower lung field was80.17%,82.22%for the upper left lung field, left middle lung field was83.06%,81.56%for the left lower lung field;(3) according to the occurrence and non-occurrence of the shadow in stabilization period, there was a significant difference in treatment effect, the improvement rates of patients with shadows were67%or less, a minimum of59.01%, while the improvement rates of patients without shadows were more than83%. up to83.34%. patients occurred lung shadows in the stabilization period had poor outcome, compared with those didn’t occur lung shadows in the stabilization period (RR<1);(4) the improvement rates of patients with six lung fields’lesions in stabilization period was sorted by the improvement rate, right upper lung field was59.01%,59.68%for the right middle lung field, right lower lung field was60.25%,66.15%for the upper left lung field, left middle lung field was63.52%,60.87%for the left lower lung field;(5) patients occurred lung shadows in the stabilization period had poor outcome, compared with those occurred lung shadows in progressive stage;(6) the improvement rate of patients with six lung fields’pulmonary cavity was sorted by the improvement rate, right upper lung field was94.16%,91.13%for the right middle lung field, right lower lung field was80.73%,89.4%for the upper left lung field, left middle lung field was80.21%,94.80%for the left lower lung field;7. Through network explaining away:(1) The conditional probability of shadow improving is higher in patients who only have lesion shadow in progressive stage in each lung field, all of which are above75%; patients with lesions shadow in the stable stage have a lower probability of improvement, all below67%; the cavity in lung field have little effect to improvement rate of the shadow, all the improvement rate of shadow is above78%;(2) To rank the lesion form and cavity shadow in each lung field by the conditional probability:Discovery of the path is "the lesions shadow in progressive stage in the right upper lung fieldâ†'the lesion outcome "is best, and the path is "cavityâ†'there’s lesions shadow in the progressive stage in the right upper lung fieldâ†'the lesion outcomes" ranks in the second place, the path is "lesions shadow in the progressive stage in the right upper lung fielâ†'lesion outcome "ranks in the third place.Conclusions:1. According to the sputum characteristics of Chinese rural tuberculosis population, the proposed assessment strategy using X-ray is effective and feasible. X-ray examination has a great value to sputum bacteriological inspection as a supplementary method.2. The proportion of male and female tuberculosis patients is unequal in rural areas, and most patients are older than15-year old. Some factors such as gender, age, occupation has an significant impact on the outcome of the lesion. The longer you delay treatment, lesion imaging outcome the worse. The unity of the WHO recommended initial smear-positive TB patients chemotherapy scheme2HRZE/4HR and2H3R3Z3E3/4H3R3have the same curative effect.3.Tuberculosis patients in different Shadow area, scope, the shadow of the disease tuberculosis lesions have different curative effect. Lung disease is focal on shadow, especially the upper right lung field and the upper left lung field. Patchy shadow shadow is the most common lesions. Occurred in patients with less empty shadow, empty curative effect to cure also have effect, hollow parts had a great influence on curative effect. Lung lesion shadows, empty treatment curative effect are unfavourable factors for patients4. The shadow form of lung field lesions, treatment delay time, and age are risk factors of treatment and outcome for lesions.5.We constructed the bayesian network using the tuberculosis lesions’hole shadow in eachlung field, age, clinic treatment delays and treatment methods. Through the causal reasoning of the network we could find:(1) the treatment effect was better if the hole shadow in advanced disease periodoccurred in the upper right lung and left lung field, and if all of the six lung fields had shadow in the state of advanced disease period, the outcome was better if the hole shadow located in the middleupper lung field.(2) the outcome of the hole lesions in stable disease periodwasless effective, and it showed no obviousrelationships with the range of the lesion locations(3) the therapeutic effect in the advanced lung lesion was significantly better than that in stable lung lesions in the same lung field. Through the supportive reasoning of the network we could find that:(1) the lesion outcome had an obvious relationship with the type of lesion shadow, and the outcome of the lesion in stable disease period was much worse than that in advanced disease period (2)the outcome of patients with empty hole was less effective (3)in the stable period, the right upper lung field and the right middle lung field were the key illness locations, and the occurrence of lung lesions in both locations had the worst outcome. Innovations:1. For rural China TB sputum, crowd in X-ray as evaluation standard, puts forward some ideas of evaluating the effect of China’s rural tuberculosis populations in DOTS.2. To clarify the influence factors of the lesion outcome.3. Established the x-ray lesions of tuberculosis treatment of Bayesian Network, and clarified the role of these factors on the chemotherapy of pulmonary tuberculosis from the overall level. Based on the network causal inference, the effect of lung field type, lesion location, lesion range, the cavity location and range on cavity treatment could be acquired, which has a good significance in evaluating the curative effect of patients with pulmonary tuberculosis and their medicine therapy.Limitations:1. In this study, does not take the medication normalization factor into consideration.2. We only used the static Bayesian method in this study without considering the results of the X-ray lag time.3. The patients living habits, family economic income situation were not taken into account in the design of questionnaire and analysis because of time constraints.