Standard max-min system
Step 2: Forecasting
Forecasting, the second step in the quantifcation process, uses the data collected during the preparation step to estimate the quantity of each product that will be dispensed or used during each year of the quantifcation. Tese quantities are the basis for calculating the total commodity requirements in the supply planning step. Te forecasting step in a quantifcation exercise is a four-part process (see table 6-2):
Table 6-2: Forecasting Process
fOreCasTing PrOCess
Part 1 Organize, analyze, and adjust the data.
Part 2 build and obtain consensus on the forecasting assumptions.
Part 3 Calculate the forecasted consumption for each product.
Part 4 Compare and reconcile results of different forecasts.
During preparation for a quantifcation exercise, team members begin to collect program background information; and as many types of data, from as many sources, as possible. Now, you can begin to evaluate and organize these data. Te four primary types of data are demographic, morbidity, services, and consumption (see table 6-3 for examples of these data).
Demographic data are data on population characteristics, growth, and trends. Tey are not usually recommended for forecasting health commodity needs for procurement purposes, unless they are combined with other sources
of data. However, when forecasting for contraceptives, you can usually fnd reliable demographic data in the Demographic and Health Surveys, or national census data; you can use this data for forecasting.
Morbidity data are data on estimated incidence or prevalence rates of specifc diseases, or health
conditions, occurring within a defned population group. Tese data can be extrapolated to defne total estimated need and then refned to determine specifc targets, or percentage of total need, to be reached.
Because forecasts using morbidity data tend to overestimate commodity needs, you should compare them to forecasts using consumption and services data. Morbidity data is not used to forecast for preventative services, such as family planning and pregnancy prevention.
Services data include number of services provided, number of service visits at which products are dispensed, tests conducted, episodes of a disease or health condition treated, or number of patients on continuous treatment during the last 12-month period (when data is available or can be estimated).
Consumption data are data on the actual quantities of health commodities. Consumption data includes actual dispensed-to-user data, or data on the numbers of commodities that were actually given to clients. Issues data can also be used as a proxy for consumption data; issues data are data on the number of commodities transferred from one level of the supply chain to another. For example, if the district warehouse issued 400 cycles of pills to its health centers in this reporting period, 400 cycles is the issues data that could be used to estimate actual consumption.
Table 6-3: Types and Sources of Data for Forecasting Product Consumption
TyPe Of daTa sOurCes Of daTa CHaLLenges in daTa QuaLiTy Program
background information
Program progress and evaluation reports, policy and strategic planning documents, technical reports, and workplans that specify the timing of training and expansion of services
May be outdated and not refect current policies, strategies, or context.
Demographic • demographic Health survey (dHs), national census data, Population reference bureau
• data on population growth and trends
• data on population characteristics, e.g.,
geographical distribution, age, gender, occupation
Tends to be outdated (1–4 years old or more).
data may not refect the same time period and, therefore, cannot be easily aligned.
Morbidity • epidemiological surveillance data or research data on incidence and prevalence of disease or health conditions in a given population
• expressed as a ratio or percentage of a defned population (denominator) with a specifc disease or health condition (numerator)
data from epidemiological studies may be outdated (1–2 years).
if data are specifc to a particular population group, you will need to extrapolate to estimate incidence or prevalence in the general population.
Services • HMis reports, program M&e reports, facility surveys of service records, daily registers
• reported number of services provided, e.g., number of cases of disease or health condition treated, number of HiV tests conducted, number of children immunized
data may be unavailable, outdated, incomplete, or unreliable for the past 12 months.
Consumption • LMis reports, facility surveys of stock records and consumption records
• reported quantities of products dispensed to patients/clients or quantities of products used
data may be unavailable, outdated, incomplete, or unreliable for the past 12 months.
Program
targets • national policy and strategic planning documents
• national annual program targets or service coverage rates set as goals for the program
Program targets may be politically motivated for advocacy purposes and not based on realistic program capacity.
Remember that data collection activities initiated during the preparation step will continue throughout the forecasting and supply planning portions of the quantifcation exercise, as well.
Forecasting Part 1: Organize, analyze, and adjust data
After you collect the available data, you need to assess its quality. To estimate the data that would have been reported, you should adjust for incomplete, outdated, or unreliable consumption and services data, as described in chapter 3. If the program experienced a stockout, you may want to adjust the reported consumption data to account for that. Be sure to document your methodology for making any data adjustments, noting any adjustments made for stockouts, for percentage of facilities reporting, or for outdated data. Table 6-4 describes an example of the assessment of data quality for a quantifcation in Tanzania.
Table 6-4: Data Quality Analysis for ARV Drug Quantifcation in Tanzania TyPe Of daTa daTa QuaLiTy Of daTa nOTes Demographic/
morbidity Total population (40,454,000) HiV prevalence rate (6.1%)
1 year old not used for the forecast because, given program capacity, calculated quantity would have been unrealistic.
Services Total number of patients on antiretroviral therapy (arT) (102,769 adults)
facility reporting rate
is 67% includes the cumulative number of patients that ever started arT since October 2004, when services began. does not account for any patients that discontinued treatment.
number of patients on arT by regimen (e.g., 44,190 adults on aZT + 3TC + nVP)
Collected at 9 facilities and from individual partners supporting facilities
newly revised arT patient registers collect the number of patients on arT, by regimen; but data are not being reported or aggregated at central level.
Consumption Quantities of arV drugs issued to facilities over the past 12 months (e.g., 650,000 bottles of d4T/3TC/nVP)
Consumption data not available. Central-level issues data used as proxy for consumption.
not used for the forecast because central-level data does not represent actual consumption.
Central-level stock on hand (e.g., 700,000 bottles of d4T/3TC/nVP on hand)
facility-level stock on
hand not available. used later during the supply planning step.
Program targets
national program targets for 2011 and 2012 (e.g., target number of arT patients for 2009 is 400,000)
not based on existing patients or historical scale-up rates.
not used for the forecast.
Note that some types of data will require conversions. Because consumption data are collected as quantities of commodities, you will not need to convert them later. However, if you use demographic, morbidity, or service data, after you analyze the trends and factors that you expect will infuence demand and establish the agreed-upon numbers for the previous years, then you must convert all data into numbers of commodities.
For example, you receive services data as number of visits; however, you receive morbidity data as number of cases. For each type of data, the quantifcation team will need to translate that data into number of commodities, using the appropriate conversion factor (see table 6-5).
Table 6-5: Conversion of Data into Product Quantities
TyPe Of daTa COnVersiOn faCTOr fOreCasTed COnsuMPTiOn
Consumption estimated quantity of product
to be dispensed/used X =
Quantities of product services
(family planning) services (HiV and aids,Tb,malaria,essential medicines, labs)
estimated # of visits or users
estimated # of patients, # of episodes of disease, or health condition, # of lab tests
X
X
dispensing protocol (contraceptives) sTgs, testing algorithm, lab procedure
=
=
demographic (family planning) demographic/
morbidity
estimated # of users
estimated # of patients, # of episodes of disease or health condition, # of lab tests
X
X
CyP factor
sTgs, testing algorithm, lab procedure
=
= Program targets Targeted # of users, # of
patients, # of episodes of disease or health condition, # of lab tests
X
CyP factor, sTgs, testing algorithm,
lab procedure =
Sample assumptions for quantifcation in Zambia
during Zambia’s national quantifcation of public sector contraceptives for 2010–2012, the forecasting team made the following assumptions:
• The method mix for oral contraceptives was assumed to be 90 percent combined orals and 10 percent progesterone-only orals.
• use of long-term contraceptives was expected to increase due to promotion of such methods by the ministry of health and training of more health workers in the insertion of iuds and implants.
• as a result of the quantifcation:
— Consumption of pills was reduced and added to implants.
— use of lactational amenorrhea (LaM) and injectables were reduced and iuds increased.
Forecasting Part 2: Build and obtain consensus on forecasting assumptions In many cases, you will fnd that data are incomplete, outdated, unreliable, or unavailable. Terefore, to develop the forecast, you need to make assumptions about program performance, targets, and future demand.
Even if data are of very high quality, you will still need to make some assumptions about—
• expected uptake in services
• compliance with recommended treatment guidelines
• impact of changing program policies and strategies on supply and demand
• service capacity
• provider behavior
• client access to services
• seasonality
• geographic variations in disease incidence and prevalence
• other factors that might afect demand.
You will need to discuss these assumptions and build consensus among key informants, including program managers, policymakers, procurement ofcers, providers (e.g., clinicians, pharmacists, nurses) and technical experts.
Forecasting Part 3: Calculate the forecasted consumption for each product Regardless of the data the quantifcation team uses for the exercise, they must document the sources of the historical data, actual data collected, data quality issues, and any data adjustments.
Ten, for each product—
• Estimate the future consumption of each product—the number of units of each product needed for each month and the year of the quantifcation period. Base this estimate on a review and analysis of historical trends in consumption and assumptions about program plans, targets, and any changes in product selection, STGs, or other policies and strategies that are expected to afect future demand.
OR
• Estimate the future types and number of services that will be provided; number of episodes of a disease or health condition that will be treated; or number of patients that will be treated, based on historical data. Using table 6-5, convert services, episodes, or cases into actual quantities of products.
You can use a number of diferent methods when you estimate the future consumption of products. For example, when using historical data from facilities, a historic trend can be determined by either the—
• average percentage increase/decrease from one reporting period to the next OR
• average absolute number of increase/decrease from one reporting period to the next.
You can then project these trends forward—monthly, quarterly, or annually—to calculate the future number of products, episodes, or patients.
If you use demographic/morbidity or services data, you must convert the number of patients or episodes into the number of products. Tis is done after you establish the forecast number of patients or episodes.
Forecasting Part 4: Compare and reconcile results of different forecasts
If availability and quality of data permits, the quantifcation team can use diferent types of data to conduct multiple forecasts. Te forecasting steps must be repeated for each of these data types. Use at least two types of data and prepare separate forecasts, if possible. Compare the fnal forecast consumption quantities from each forecast and consider the implications of the diferent forecasts for the program, including service capacity, storage and distribution capacity, funding availability; and other issues that could afect demand, supply, and use of the commodities. You can either select one of the fnal forecast fgures; or reconcile the forecasts by adjusting, weighing, or averaging the diferent forecast quantities. How you reconcile the forecasts will depend on your confdence in the data you used and the strength of your assumptions. Remember that you should use as many diferent types of data when you do forecasting;
this helps improve forecast accuracy, validate forecast results, and foster ownership of the quantifcation process and results. After you reach the fnal forecasted quantity, you can go to the supply planning step.