278. Developing a Logistic Regression Model to Aid Clinicians Evaluate Outpatients and Predict Odds of Hospital Transfer in a Nicaraguan Pediatric Population: Comparison of Epidemiological Models to Predict Hospitalization with a Focus on Antimicrobial Stewardship.
Session: Poster Abstract Session: Pediatric Antimicrobial and Diagnostic Stewardship
Thursday, October 4, 2018
Room: S Poster Hall
  • ID Week Poster_Edit_AV_9.27.pdf (729.0 kB)
  • Background: Upper Respiratory Infections (URI) represent a significant disease burden to children worldwide. Clinicians must rely on clinical acumen and evidence-based medicine to responsibly prescribe antimicrobials to curb the rise of antimicrobial resistant pathogens. We propose a model to help clinicians predict the odds of hospital transfer upon initial evaluation of pediatric patients presenting with URI in a low to middle income setting.

    Methods: We performed a prospective cohort study of 2311 children aged 3 months – 15 years enrolled in an outpatient government health clinic in Managua, Nicaragua over a 5-year period. Symptoms, exam findings, laboratory studies, diagnoses, and data on antimicrobial use were collected. Primary outcome was hospital transfer. Using forward-selection logistic regression, we constructed a model of the risk factors and exam findings most likely to predict hospital transfer. WHO criteria were used to risk-stratify pneumonia cases. We examined the frequency and type of antimicrobials used. We then applied Hay et al’s STARWAVe model to examine its utility in our population.

    Results: Of the 2,311 children that participated in the cohort between 2011 and 2015, 2155 children (93%) experienced one or more URI. Those children experienced a total 18,826 URI episodes. 5383 (28.6%) of URI cases received antibiotics. 332 URI cases were transferred to the hospital, of which 167 (50.3%) were given antibiotics. Age < 2 years, male sex, having 4 or more symptoms, vomiting, poor appetite, diagnosis of “flu-like illness,” wheezing, subcostal retractions, rhonchi and fever were all independently associated with hospital transfer (P<0.05). STARWAVe had fair predictive value (AUC 0.6709) but our model had better predictive value (AUC 0.7011). 90% of all pneumonia cases were properly managed by WHO criteria.

    Conclusion: We defined a set of clinical criteria that predict hospital transfer in a low and middle-income community setting. We also examined the fit of a validated predictive model developed in a high-income setting and found that this model performed reasonably well in our setting. Overall, most pneumonia cases were treated effectively by WHO criteria indicating that local physicians were properly prescribing antimicrobials.

    Amit Vahia, M.D., M.P.H.1,2, Guillermina Kuan, M.D.3, Sergio Ojeda, M.D.4, Luis Nery Sanchez, M.D.4, Angel Balmaseda, M.D.3, Eva Harris, PhD5 and Aubree Gordon, Ph.D2, (1)Infectious Diseases, Henry Ford Health System, Detroit, MI, (2)School of Public Health - Epidemiology, University of Michigan, Ann Arbor, MI, (3)Ministry of Health, Managua, Nicaragua, (4)Sustainable Sciences Institute, Managua, Nicaragua, (5)University of California, Berkeley, Berkeley, CA


    A. Vahia, None

    G. Kuan, None

    S. Ojeda, None

    L. N. Sanchez, None

    A. Balmaseda, None

    E. Harris, None

    A. Gordon, None

    Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 3rd with the exception of research findings presented at the IDWeek press conferences.