Community Validation of Influenzalike Illness as a Predictor

Community Validation of Influenza-like Illness as a Predictor of Influenza Jonathan L. Temte, MD/Ph. D & Alexis Eastman, MS-2 University of Wisconsin School of Medicine and Public Health Peter A. Shult, Ph. D, Carol J. Kirk & Mary Wedig Wisconsin State Laboratory of Hygiene Madison, Wisconsin

Influenza-like Illness § Definition § Fever of 100 o. F (37. 8 o. C) or higher § Cough and/or Sore Throat § Not due to any other illness § Utility § Simple and elegant § Clinically relevant § Easily ascertained

ILI uses § Community surveillance of influenza § Clinical identification of influenza infection § High PPV from research protocols § Adults § Children

Factors Affecting Symptoms Host Factors Age Viral Factors Viral Strain Immune Status Viral Subtype Underlying Disease

Purpose of Study § Review the contents of a large database § Surveillance data emerging from a partnership between a public health laboratory and primary care clinicians § Symptoms and virus identification § Validate ILI for influenza infection § Community—not research—perspective

The Surveillance Database § Partnership of WSLH and UW-DFM since 1994 § Major modification of symptom check off in 1997 § Opportunistic sampling with “fee-exempt” virus culture § physicians obtain specimens, record demographic and symptom data, sample is transported to WSLH by courier. § Standard culture methods with isolation rate = 45% § Limited, de-identified data used § 1997 -2007 § IRB approved § 3, 796 episodes of acute respiratory illness care available

Preferential Collection from Children and Young Adults Range: 0 – 103 years 55. 6% female

Predictors Working definition of ILI § F+Cor. ST § F = Fever on symptom checklist § No requirement for level or documentation § Cor. ST = Cough and/or Sore Throat § s. F+Cor. ST (includes seasonality) § December through March § Period with > 90% of influenza cases

Outcomes influenza isolation § Paradigm 1: “clinical primary care” § Influenza (+) vs. all other specimens § Influenza = 1230 § Non-influenza + no virus isolated = 2566 § Paradigm 2: “ideal virus capture” § Influenza (+) vs. non-influenza virus (+) § Influenza = 1230 § Non-influenza = 523

Distribution of outcomes Reference Season population included Yes All ARI specimens No Criteria used Influenza (+) Influenza (-) s. F+Cor. ST (+) 1020 1034 s. F+Cor. ST ( -) 210 1532 s. F+Cor. ST (+) 1082 1529 s. F+Cor. ST ( -) 148 1037

Distribution of outcomes Reference Season population included Yes Virus (+) specimens No Criteria used Influenza (+) Influenza (-) s. F+Cor. ST (+) 1020 188 s. F+Cor. ST ( -) 210 335 s. F+Cor. ST (+) 1082 302 s. F+Cor. ST ( -) 148 221

Performance Characteristics Criteria Reference F+Cor. ST All ARI OR flu 4. 96 F+Cor. ST Virus (+) 5. 25 0. 88 0. 42 0. 78 0. 60 s. F+Cor. ST All ARI 7. 20 0. 83 0. 60 0. 50 0. 88 s. F+Cor. ST Virus (+) 8. 66 0. 83 0. 64 0. 84 0. 62 population Sens Spec PPV NPV 0. 88 0. 40 0. 41 0. 88

Additional Fiddling assessing effects of age § Concentrate on seasonal data § Clinician informed by surveillance § Concentrate on virus (+) specimens § Symptomatic patient § Early in illness § Collection technique good § Concentrate on age categories § § 0 -4 5 -24 25 -64 65+

Effects of age (reference age = 25 -64 years) Binary logistic regression via Minitab – Release 13. 1 Factor Odds Ratio 95% CI s. F+Cor. ST 7. 55* 5. 81 – 9. 80 0 -4 years 0. 10* 0. 07 – 0. 14 5 -24 years 1. 21 0. 90 – 1. 65 25 -64 years reference 65+ years 1. 67 0. 86 – 3. 25 * P<0. 001

What about little kids? the percent of ILI cases due to: Virus Influenza Adenovirus Parainfluenza Rhinovirus RSV Herpes simplex Enterovirus 0 -4 years 34. 8 6. 6 14. 4 1. 7 37. 0 1. 1 2. 9 5+ years 84. 7 3. 2 3. 3 3. 7 1. 4 2. 0 0. 8

Conclusions § ILI (F+Cor. ST) performs well § § Public health tool for surveillance Early detection of influenza High sensitivity ( 0. 88) Limited by low specificity (0. 40) § but fined tuned by virological methods § ILI (s. F+Cor. ST) performs well § § Clinician tool for diagnosis of influenza Informed by public health surveillance High PPV (0. 84); moderate NPV Excluding young children raises PPV to 0. 90

Conclusions § Influenza is the primary cause of ILI in patients age 5+ years § Many viruses can cause ILI in children 0 -4 years of age. ILI should not be used for diagnosis alone in this group. § ILI for predicting influenza infection has been validated in a primary care, community-based population

Final Words If influenza is in the community and your patient is over 4 years old Is it influenza? F+Cor. ST “Of Course”

Acknowledgements § Wisconsin Primary Care Clinicians § UW-DFM residency clinics § Numerous private physicians § UW-DFM Summer Student Research and Clinical Assistantship Program § Ms. Alexis Eastman § Wisconsin State Laboratory of Hygiene

Additional Material

Basic Characteristics of Surveillance System § Mean age of patient = 26. 6 years § Range [ 0 to 103 years] § Sex § Female = 55. 6% § Male = 44. 4% § Time between illness onset and collection § Mean = 3. 86 days § Median = 2 days § Rate of virus isolation = 44. 6%

Specimens Collected during “Respiratory Virus” Season

Collection Day (Monday through Thursday Preferred)

Most Specimens Collected at Optimal Time

Percent of Specimens with Positive Virus Isolation