Study Of A Risk-Adjustment Model Of Hemodialysis Associated Infection Indicators Based On Transparency

[Backgrounds] Transparency performs a useful function in regulating global actors and promoting accountability at a transnational level in numerous fields. In our country, health-care associated infection surveillance has become a tool for health-care associated infection control and quality of care improvement. However, the results of surveillance only provide internal feedback of health-care associated infection indicators so far. The actions of several countries show that enforcing the public reporting of health-care associated infection indicators can better improve the quality of care. Howerer, the existing patients characteristics in those institutions and regions make the original health-care associated infection indicators often cannot reflect the real quality of care. In the respect of hemodialysis-associated infection, study for hemodialysis associated infection of domestic, as well as international countries only stay in the study of risk factors and no validated and optimal risk adjustment model for hemodialysis associated infection is currently available.[Objectives] This study is aimed to explore the risk adjustment factors for hemodialysis associated infection and to derive a risk adjustment model for hemodialysis associated infection on the basis of objective patient related risk factors easily acquired. It will provide a high evidence based method for hemodialysis associated infection prevention and control, and also for the proposition of transparent regulatory policy in order to achieve patient safety and quality of care improvement, guide the patients’selection, and reduce the medical expenses. It will also rich the strategy for hemodialysis associated infection control and transparent regulation. [Methods] (1) The theoretical basis of this study is "Transparency Selection-Change Two Pathway Model" and "Patient-Medical-Environment Factors-Based Risk Adjustment Model"(2) Literature research and expert opinion were used to theoretical research for risk adjustment factors and methods of hemodialysis associated infection rate, and to build the key risk factors set of hemodialysis associated infection rate. The hemodialysis associated infection rate was the outcome indicator. A cross-sectional study was used to data collection and verification of risk factors, and then identifed the key risk factors. In addition, this method was used to verify and revaluate the risk adjustment model for hemodialysis associated infection. An optimal risk adjustment model for hemodialysis associated infection was expected to establish.(3) The validation method for key risk adjustment factors:Verified them by calculating the reliability statistics of logistic regression model and using the verification sample. In addition, artificial neural network analysis was used to verify the importance of each key risk adjustment factor including the factor closely related to hemodialysis. Thus, the key risk adjustment factors set was finally determined by those multiple verification methods.(4) The validation method for risk adjustment model:For the first one, computing the Hosmer Lemeshow goodness-of-fit statistics of the model; Seceondly, layered the data by the expected infection rate calculated using the risk adjustment model derived, compared the actual infection rates and expected infection rate of each layers, and then caculated the sensitivity, specificity, the general accuracy and area under the ROC curve; Thirdly, using the new data (verification sample); Fourthly, artificial neural network analysis was used to verify the risk adjustment model derived.(5) Excel17.0 was used to organize and clean the data; Data Analysis was performed using STATA 12.0. Artificial neural network analysis was performed using SPSS 17.0 (The ratio of training sample, test sample and support sample was 7:2:1).[Results] (1) Current status of hemodialysis associated infection:A total of 403 cases of hemodialysis patients were identified from January 2014-March 2015, among them,76 hemodialysis patients were infected. The derivation sample included 305 hemodialysis patients and 63 of them got the hemodialysis associated infection, the infection rate was 20.66%, patients were more men (60.66%), and the average age was 47.4+15.8 years. The verification sample included 98 hemodialysis patients and 13 of them got the hemodialysis associated infection, the infection rate was 13.27%, and the average age was 45.4+15.5 years. The top three hemodialysis associated infection types were respiratory tract infection (59.2%), blood vessel related infection (25%) and urinary tract infections respectively (10.5%).(2)Risk factors of hemodialysis associated infection:Through univariate analysis, patient basic characteristic factors" receiving hemodialysis in more than one hospital"(X2)、clinical treatment fators which must be accepted for patients with basic diseases" length of stay(X1), blood transfusions(X3), using hormone before infection(X4)" and blood routine index at early hospitalization" serum calcium level" have significant correlation with hemodialysis associated infection (The P value of them were 0.01、0.00、0.01、0.00 and 0.02 respectively; Through stepwise logistic regression analysis, the factor of serum calcium level was eliminated for its P value was above 0.05 and the final logistic regression was bulid:The predicted infection probability of the model:(3)The verification of risk adjustment model:In the logistic regression analysis, the P value of Hosmer-Lemeshow goodness-of-fit test was 0.82. After layering the data by the expected infection rate, the P value of goodness-of-fit test was 0.7, the actual infection rate in the top layer was 13.4 times of that of the lowest layer, achieving at least 10 times as much as the ideal value. Sensitivity, specificity and the general accuracy were 66.67%、 70.25%、 69.51% respectively. The area under the ROC curve was 0.76. All of these showed the model had good reliability. False positive rate and false negative rate of verification sample were 10.6 (9/85) and 15.4%(2/13) respectively. In the artificial neural network analysis, length of stay, blood transfusions, albumin level, serum calcium level and using hormone before infection were the top 5 important factors for hemodialysis associated infection. The area under the ROC curve of the prediction model was 0.93. The sensitivity, specificity and the general accuracy of the training sample were 61.7%,98.3% and 90.4% respectively.[Conclusions and suggestions] (1) Risk adjustment model for public reporting hemodialysis associated infection rate can be derived by four risk adjustment factors:receiving hemodialysis in more than one hospital, length of stay, blood transfusions and using hormone before infection. It may provide a high evidence based method to create and innovative transparent regulation policy for health-care associated infection. (2) Recommendations:Firstly, strengthening the system construction for hemodialysis associated infection based on transparency to promote the health-care associated infection regulation level; Secondly, putting forward methodological researches for health-care associated infection based on transparency to provide scientific hemodialysis associated infection regulation methods; Thirdly, strengthening the hospitalization management for hemodialysis patients, including controlling the length of stay, rational using of hormone, strengthening blood transfusion management, strictly controlling the patients transfer, to improve the level of health-care associated infection control; Fourthly, promoting the education and training for medical personnels and hemodialysis patients on hemodialysis associated infection to control hemodialysis associated infection from the aspect of knowledge and awareness.