Presentation on theme: "Estimating the Global Health Impact of Improved Diagnostic Tools Jeffrey Wasserman Federico Girosi Emmett Keeler November 2006, April 2007."— Presentation transcript:
Estimating the Global Health Impact of Improved Diagnostic Tools Jeffrey Wasserman Federico Girosi Emmett Keeler November 2006, April 2007
2 3/2008 Introduction to Diagnostic Tools Project for BMGF Better diagnosis might reduce the burden of disease Modeling the Benefits of New Diagnostics Basic approach Key issues and decisions Findings Benefits of New Diagnostics for Tuberculosis Selected Findings from Other Diseases Outline
3 3/2008 The Burden of Disease Remains Large in the Developing World 01,000,0002,000,0003,000,000 Malaria Child diarrhea Child respiratory Tuberculosis HIV/AIDS * World Health Organization, 2003 (HIV) and 2002 (all others) Annual Deaths Worldwide*
4 3/2008 Diseases Are Concentrated in Some Regions Latin America Africa Asia
5 3/2008 Many Factors Might Contribute to Improved Health Outcomes in the Developing World Good Health LifestyleNutrition Public Health Programs Clinical Care Diagnosis Treatment
6 3/2008 How Better Diagnostics Could Help More accurate tests could help target therapy to those who need it and eliminate wasteful treatments Earlier diagnosis could allow therapy to start sooner, reducing impact on the individual and reducing spread of the disease Simpler, easy-to-use tests could help increase access to diagnostics to those who currently have no care
7 3/2008 But Barriers Exist to Using Many Current Diagnostic Tools in Resource-Poor Countries Tests often require advanced infrastructure High-performing tests are expensive Some existing tests are inadequate and slow Cultural and political considerations may impede acceptance Needs may vary across countries
8 3/2008 We were asked two Key Questions What are the global health benefits of better clinical diagnostic tools? What performance specifications and infrastructure requirements does a test need to achieve the estimated benefits?
9 3/2008 Introduction Better diagnosis might reduce the burden of disease Modeling the Benefits of New Diagnostics Basic approach Key issues and decisions Findings Benefits of New Diagnostics for Tuberculosis Selected Findings from Other Diseases Outline
10 3/2008 We Worked with the Gates Foundation to Create a Global Health Diagnostics Forum Malaria Tuberculosis Acute Lower RespiratoryInfections HIV and Sexually- Transmitted Diseases DiarrhealDiseases Forum Working Groups The Forum’s Role Provide expertise on relevant diseases, diagnostic needs, emerging technologies Identify key intervention points for each disease Refine the approach for assessment Serve as conduits to the broader scientific community
11 3/2008 Status quo Model for calculating benefits of new Test Population characteristics Population characteristics Health status Health status Access to diagnostics and treatment Access to diagnostics and treatment Health outcomes With new diagnostic test Health outcomes Difference in outcomes = Gains from improved diagnostic tests
12 3/2008 Access to a New Diagnostic Depends on the Level of Infrastructure Available Advanced/ Moderate Hospitals and urban clinics Hospitals and urban clinics Electricity, clean water, well-equipped laboratories, trained clinicians Electricity, clean water, well-equipped laboratories, trained clinicians Minimal Health clinics (Africa), rural clinics (Asia, Latin America) Health clinics (Africa), rural clinics (Asia, Latin America) No reliable electricity or clean water, no laboratory, minimal expertise No reliable electricity or clean water, no laboratory, minimal expertise None Village or community Village or community No electricity, clean water, physical infrastructure, or trained staff No electricity, clean water, physical infrastructure, or trained staff
13 3/2008 Characteristics of potential new tests Accuracy = sensitivity, specificity Access Time to diagnosis ( related to loss to follow-up) Cost of equipment and operation Other (specimen type, multiplex diseases….)
14 3/2008 Introduction to Diagnostic Tools Project for BMGF Better diagnosis might reduce the burden of disease Global Health Diagnostics Forum Modeling the Benefits of New Diagnostics Basic approach Key issues and decisions Findings Benefits of New Diagnostics for Tuberculosis Selected Findings from Other Diseases Outline
15 3/2008 Key modeling issues and decisions Get intervention points from experts or models? Decision trees are good for studying diagnostic tests Static vs. Epidemic Models, future trends What about costs, especially of over-treatment? Potential vs probable access, diffusion Where can we get data to populate the models?
16 3/2008 Static or dynamic models For TB and HIV, state of the art is dynamic models: people flow between states according to diff. equations, researchers calculate long-run policy impacts. states: no disease, treated early dis., untreat dis.,… decision trees give some parameters for models An alternative is static decision trees: calculate costs and effects in one year, with incidence assumed unaffected by new DX tool. long run effects ~ proportional to one year effects our main interest is comparative performance transmission handled by multipliers Static model can be used for all five diseases Simpler method allows us to do project with limited time and money.
Susceptible people Each case infects k n-1 others S n-1 SnSn The difference equations S n = Function(S n-1, other “n-1” conditions) show how the system evolves over the generations. Epidemics If k = 2 Let S n be the number of cases in generation n
18 3/2008 A dynamic model: Diseases in Equilibrium (TB) Active TB TB Death cures Healthy Latent TB Each active case “infects” about 20 people on average. - they then have latent TB, but a few progress Some active cases are cured, some die A generation is ~ 4 years from activation to resolution If Disease is in equilibrium, what % p of latent cases progress to be active each generation? What happens if a higher % progresses? New diagnostic tool affects the transitions from active TB. 20 p?
19 3/2008 What are the costs of over-treatment? BMGF said not to consider costs, but… If no cost of treatment, no need for Dx, just treat all equivocal patients. What are the costs? Side-effects of treatment Possible super-germs? Opportunity costs
20 3/2008 Methods to calculate C = cost*of over-treatment Ask panel of experts: identifying one more person with TB justifies the costs of tagging N more people who don’t have TB as having it? Use opportunity costs of wasted money, assuming we can save a life by spending $x on another program. Girosi: Use treatment guidelines to bound C If we don’t treat when p(dis) < Pmin then C > lower bound L If we treat whenever p(dis)>Pmax then C < upper bound U * This cost is called the Harm in Hunink.
21 3/2008 Girosi’s Method to calculate C = cost of over-treatment (2) Girosi: Often, recommended treatment involves a test and then only the positives get treated. for ARI it’s a clinical judgment with TPR =.9, TNR =.7, the benefit of treatment = =.1 death averted, and prior p =.015. costs of treating everyone = H(1-p)~H costs of treating no one = Bp costs of following clin judgment: (1-.9)pB + (1-.7) (1-p)H If the last is best.1pB +.3H .1pB and.3H.1pB and.3H<.9pBm so pB/7
22 3/2008 Potential vs Actual access Actual: people currently being tested for disease X in Level L facilities Potential: people who could get to a facility of level L with a certain amount of time and effort We used potential: with better tests and treatment, demand would rise thinking about the post diffusion future Fit potential access using geographic data from a few countries, and data on actual access to TB clinics in a regression and estimate for all countries of interest.
23 3/2008 Data Problems Models need incidence of disease of interest, incidence of seeking treatment for symptoms, outcomes of treated and untreated cases, accuracy of status quo tests Future test characteristics can be whatever we choose.
24 3/2008 Introduction Better diagnosis might reduce the burden of disease Modeling the Benefits of New Diagnostics Basic approach Key issues and decisions Findings Benefits of New Diagnostics for Bacterial Lower RI Selected Findings from Other Diseases Outline
25 3/2008 Respiratory Infections Are the Leading Cause of Childhood Mortality Respiratory infections contribute to the deaths of more than 2 million children each year, mostly in Africa and Southeast Asia Most of these 2 million children die of bacterial pneumonia, which is treatable with antiobiotics In developing countries, the main form of diagnosis is clinical assessment A large number of children are not being diagnosed, while others are being treated unnecessarily, leading to wasted treatments and antibiotic resistance
26 3/2008 Child with symptoms Access to clinical diagnosis Self-treat No care Status Quo New Test Access to new test No access to new test We Modeled the Status Quo and Access to a New Test for Bacterial Pneumonia in Children Under 5 Disease Yes Disease No Test + Treat Test – No Treat Survives Dies Disease Yes Disease No
27 3/2008 We Traded Off Test Performance and Infrastructure Requirements Advanced infrastructure Good Performance Lives saved by new test for bacterial pneumonia* Perfect Performance Minimal infrastructure *Results assume access to treatment
28 3/2008 Easier Tests Produce Large Health Benefits, Even with Less Than Perfect Performance Advanced infrastructure Good Performance 261,000142, , ,000 Significant gains, potentially achievable Perfect Performance Minimal infrastructure *Results assume access to treatment Lives saved by new test for bacterial pneumonia*
29 3/2008 Our Recommendations for a New Diagnostic for Bacterial Lower Respiratory Infections Requires minimal infrastructure Requires minimal infrastructure Preferred samples types include saliva, urine, or dried blood spot Preferred samples types include saliva, urine, or dried blood spot Results should be available within 2 hours or less Results should be available within 2 hours or less
30 3/2008 Much of the Benefit of the New Diagnostic Would Be Due to Reductions in Overtreatment Developing World Lives Saved (1000s) Reduction in overtreatment Reduction in disease burden Benefits of New Test for Bacterial Pneumonia With C =.001 !
31 3/2008 Most Lives Saved from Reducing Disease Burden Accrue to Africa, While Other Regions Benefit from Reducing Overtreatment Developing World AfricaAsia Latin America Lives Saved (1000s) Reduction in overtreatment Reduction in disease burden Reduction in overtreatment Reduction in disease burden Benefits of New Test for Bacterial Pneumonia Benefits of New Test by Region
32 3/2008 All Disease Models Show a Significant Benefit from New Diagnostics 0500,0001,000,0001,500,0002,000,000 TB Children under age 5 (Africa) Children under age 5 Pregnant women (Africa) Adults with persistent cough Lives Saved Annually from New Forum-Recommended Tests Additional benefits were found for other diseases studied Malaria Syphilis Bacterial pneumonia
33 3/2008 Study Findings Highlight Other Themes Tests that are more accessible have greater benefits Access can be more important than test performance For many diseases, a more accurate test is not needed For example, current rapid tests for malaria could lead to significant benefits if made more widely available Current diagnostics do not pay enough attention to harm of overtreatment Tests are typically better at identifying people with disease than identifying those without Reducing overtreatment provides a public health benefit