ACT Demand Forecasting: Current and Future Issues

1
ACT Demand Forecasting Current and Future Issues

• Prashant Yadav

THE ARTEMISININ ENTERPRISE CONFERENCE Meeting
the Malaria Treatment Challenge Effective
introduction of new technologies for a
sustainable supply of ACTs October 8-10,
2008 University of York
2
Acknowledgements

• This work was done as a part of the Forecasting
  Task Force of the Procurement and Supply
  Management Working Group of the Roll Back Malaria
  Partnership

It has gained tremendously based on discussions
and comments from the following. Any errors,
omissions and inaccuracies are the authors sole
responsibility

• Jaques Pilloy

Artepal

• Inder Singh, Megan O'Brien, Justin Cohen, Oliver
  Sabot

CHAI

• Wouter Deelder

Dalberg

• Girindre Beeharry

Gates Foundation

• Steen Stottrup, Andrew Freeman, Jean-Paul Moatti

GFATM

• Renia Coghlan

MMV

• Jan van Erps, Philippe Verstraete

RBM

• All PSM WG Members and Co-chairs

RBM PSM WG

• Philippe Duneton, Lorenzo Witherspoon

UNITAID

• This project was done by May Ongola, a graduate
  student at the MIT-Zaragoza International
  Logistics Program, as a part of her thesis under
  the MIT-Zaragoza Africa Health and Humanitarian
  Systems Scholarship Program
• Earlier versions of this work were improved with
  comments and critique from Kara Hanson, Catherine
  Goodman and Ramanan Laxminarayan

3
Demand Uncertainty in ACT Supply Chain Ishikawa
Analysis
Demand uncertainty for API and raw materials

• Lead-time 18-22 months
• Minimum forecasting horizon 2 years (assuming no
  artemisinin inventory)

4
Demand Uncertainty in ACT Supply Chain Ishikawa
Analysis
Private sector Demand?
Demand uncertainty for API and raw materials
5
Vicious cycles of supply and demand uncertainty
Demand uncertainty for API and raw materials
Supply constraints

• Asymmetric information about supply capacity
  creates further supply uncertainty

6
Bottom-up approach to forecasting private-sector
ACT demand
(1)
(2)
(3)
(4)

• Uptake modeling
• Use standard product uptake curves to determine
  time-phased uptake
• Use CHAI and pilot studies to calibrate uptake
  curve
• Other judgmental adjustments

• Malaria Cases, Incidence and Patient Population
• Adjust for complicated cases and pregnant
  mothers
• Use case estimates from WHO/RBM MERG/Malaria
  Indicator Survey estimates

• Fraction seeking AM in private-sector
• Use data from existing published studies
• Use MoH/National malaria survey data wherever
  private-sector share is recorded

• Affordability and Willingness-to-Pay
• Use data from small sample existing studies on
  WTP for anti-malarials
• Use WB income data , HHS Household Expenditure
  to compute affordability at different price points

7
High Risk Areas
Low Risk Areas
Less ITN, IRS Coverage Adjustment
Less ITN, IRS Coverage Adjustment
Net population at risk (L)
Net Population at risk (H)
Children Under 5
Adults
Children between 5-14 years
Adults
Children Under 5
Children between 5-14 years
Apply Incidence Rates by age group
Apply Incidence Rates by age group
Net Cases Children under 5
Net Cases Children 5-14 years
Net Cases Adults
Net Cases Children under 5
Net Cases Children 5-14 years
Net Cases Adults
Adjust based on Country Treatment Guidelines
Net Market Size
Private Sector (based on usage)
Public Sector (based on usage)
8
Price elasticity of Anti-malarials?

• US price elasticities compiled from various
  studies

Adult Toothbrushes -0.65
Tomato Sauce -1.25
Baking Mix Ingredients -1.0
Dry Dog Food Dry Cat Food -1.9
Decongestants -0.4
Cough/Cold Medicine -0.9
Frozen Breakfast -1.1
Niche Products -0.6
Cooking Oil -2.08
Coffee -1.5
Price Elastic/Sensitive
Price Inelastic/Insensitive
-1
Canned Dog Food -2.0
Baby Food -1.2
Internal Analgesic -0.9
Dry Soup -0.7
Mouthwash -0.5
Canned Tomatoes -1.6
Mens Fragrance -0.3
Cat Litter -1.8
Mustard -1.3
Household Cleaners -0.8
Shampoo -1.1
9
Private-Sector Willingness-to-Pay Estimation
Data Source Income World Bank (Expenditure
level) / Global Fund (Reference products and
Price)
10
Private-Sector Demand and Price Curves
Sample to Illustrate Methodology Only
11
Problems with WTP estimation
12
Estimating market size
BTP5
13
Estimating market size-2
Sample to Illustrate Methodology Only
Total market size Earlier studies 311 Million
Treatments (WHO) 400 Million Treatments
(various) This analysis 380 Million Treatments
14
Modeling ACT uptake in the private market
Asymptotic uptake in 3 years
y2
y1
D share of ACTs
Need country specific uptake curves
15
Bottom-up forecasts

• Private sector anti-malarial market size

Overall ACT demand 217 M treatments in the 3rd
year after retail prices fall below 0.15 60
asymptotic coverage assumed
16
Problems with bottom-up approach

• Number of case estimates or incidence estimates
  vary a lot
• No standardized instrument to measure share of
  private-sector treatment
• Private-sector share of treatment seeking not
  available by demographic sub-groups
• Small sample willingness-to-pay studies not
  representative of all countries/regions
• Low statistical rigor in estimating coefficients
  of the uptake curve in the absence of a full size
  rigorous experiment
• Does not exploit public sector procurement plan
  data

17
Top down forecasting
GFATM PMIWB (from GFATM proposals)
? Current estimates are based on a single global
uptake curve
160 M
?
Private sector first line buyers
MoH or PR
Public Sector
Private Sector
IMS Presentation by Peter Stephens 26/2/07
18
Next steps
Interviews with countries to estimate planned
volumes for this flow
GFATM PMIWB (from GFATM proposals)
Current estimates are based on a single global
uptake curve 20-40M
160 M
?
Private sector first-line buyers
MoH or PR
Public Sector
Private Sector
Leverage country specific knowledge of PSM WG or
other technical partners to re-calibrate country
specific/cluster specific uptake curves.
Interview a sample of private first-line buyers
in each country to re-calibrate country
specific/cluster specific uptake curves
IMS Presentation by Peter Stephens 26/2/07
19
Updates on the forecasts to follow soon
20
Appendix
21
The aggregate level sales curve
22
The country level sales curves