Estimating the Undiagnosed Fraction: Jeanette Birnbaum 10/7/2014SPRC PHC1 presenting on behalf of Martina Morris, Ian Fellows and Matt Golden Disseminating.

SPRC PHC 1

Estimating the Undiagnosed Fraction:

Jeanette Birnbaum

10/7/2014

presenting on behalf ofMartina Morris, Ian Fellows and Matt Golden

Disseminating the “Testing History” Method

SPRC PHC 2

Overview

10/7/2014

• Summary and assumptions• KC & WA state analyses

The Method

• Prototype “Rshiny” web app

Dissemination

• Refining the method• Working with stakeholders

Your Feedback

SPRC PHC 3

The Testing History Method

10/7/2014

SPRC PHC 4

Approach: Backcalculation

• New diagnoses today…• …are based on infections that occurred in the

past

10/7/2014

Time

10 diagnoses observed

2010 2011 2012 2013 2014

1 infection

1 infection

2 infections

3 infections

3 infections

SPRC PHC 5

Time

10 diagnoses observed

2010 2011 2012 2013 2014

1 infection

1 infection

2 infections

3 infections

3 infections

Time from Infection to Diagnosis (TID)

10/7/2014

TID = 4 yearsTID = 0 years

SPRC PHC 6

Backcalculation’s 3 Components

10/7/2014

Sum of ( *

)

Cases diagnosed at T

Incidence at X yrs ago

TID probability for X yrs ago

=OBSERVED

NEED AN ESTIMATE

BACKCALCULATE

SPRC PHC 7

TID from Testing History

“Have you ever had a

last negative test?”

Yes Use date provided

No Assign age 16 or age-18 years

Missing 2 data inclusion options

10/7/2014

Infection window = [Date of last negative test -> Date of diagnosis]

1) Get date of last negative test

2) Determine time of infection within the infection window

SPRC PHC 8

Impact of TID Assumptions on TID

10/7/2014

UNIFORM

# infections

Last neg test Diagnosis

AT LAST NEG TEST

# infections

Last neg test Diagnosis

OBSERVED ONLY

Density of last negative tests

18 years0 years

18 years

% undiagnosed“BASE CASE”

18 years

% undiagnosed“WORST CASE (OBS)”

SPRC PHC 910/7/2014

INCLUDING MISSINGDensity of last negative tests

18 years0 years

AT LAST NEG TEST

# infections

Last neg test Diagnosis 18 years

% undiagnosed

“WORST CASE (MISS)”

UNIFORM# infections

Last neg test Diagnosis

AT LAST NEG TEST# infections

Last neg test Diagnosis

OBSERVED ONLY

18 years

% undiagnosed

18 years

% undiagnosed

“BASE CASE”

“WORST CASE (OBS)”

Density of last negative tests

18 years0 years

SPRC PHC 10

Applications to KC and WA State

10/7/2014

Morris/Fellows/Golden SPRC PHC 11

KC MSM: Incidence ~ Constant

11/4/2013

15% missing testing history

SPRC PHC 12

KC MSM: Undiagnosed Varies by Race

10/7/2014

SPRC PHC 13

WA: Incidence is Declining

10/7/2014

MSM

Hetero

35% missing

60% missing

SPRC PHC10/7/2014 14

White40%

missing

Black49%

missing

Hispanic48%

missing

SPRC PHC 15

WA: Undiagnosed Counts by Group

10/7/2014

Hetero vs MSM Race

SPRC PHC 16

Dissemination to Public Health Departments

10/7/2014

SPRC PHC 17

Web Portal

• Goals:– Provide self-guided access to model for local use– Minimal learning curve– Minimal pre-processing of data– Outputs presented in most policy-relevant form

• Mechanism– “Rshiny” interface

10/7/2014

SPRC PHC 18

Protoype Rshiny App

https://hivbackcalc.shinyapps.io/HIVBackCalc_App/

10/7/2014

SPRC PHC 19

Your Feedback

10/7/2014

SPRC PHC 20

Issues

• Data privacy requirements– Local, protected upload, unprotected upload

• Detail in outputs– Race, mode, sex subgroup results?– Impact of missing data (suggestions for better

assumptions?)• Training for users

10/7/2014

SPRC PHC 21

Thank you!

10/7/2014

SPRC PHC 22

Impact of TID Assumptions

10/7/2014

DATA INCLUSION TIME OF INFECTION TID DISTRIBUTION

2310/7/2014 SPRC PHC

DATA INCLUSION TIME OF INFECTION TID DISTRIBUTION