TIA was first developed in the late 1970s to answer an especially difficult and important question in future research. This method was introduced and developed by Theodore J. Gordon in 1994, as an appropriate method to mix the quantitative and qualitative methods and as an applied tool in scenario-writing. Quantitative forecasting methods, from time series techniques to econometric methods, assume that the forces that existed in the past and had an impact on the data will continue to work in the future with the same approach, and do not envisage a place for future events that can affect the continuation of the past trends or deviate the trends from their routine path, or they do not consider a significant impact for them. As a result, the use of such quantitative methods that ignore the impact of possible future events will lead to surprise-free results and predictions that will seem unlikely to be realized (11).
TIA is a method and approach of forecasting in which extrapolation is corrected for a time series so that the impacts of available trends of other fields, which are interacted with the issue understudy as well as the impact of unprecedented events can be considered. In this regard, a set of influential trends in other fields and potential events that can occur in the future are prepared and considered by taking into account the probability of their occurrence and the magnitude of their impacts so that their impacts will be applied to the results of quantitative methods and more consistency between forecast and reality or believable results will be made. In exploring the predictable events, a wide range should be considered, which includes the technological, social, economic, environmental, and political issues as well as the issues focused on the values in the society.
Assume a case study that, considering the importance of the HIV/AIDS program and the time period of 2030 in achieving the comprehensive goals of international organizations such as UNAIDS, experts are seeking to estimate the positive cases of HIV in the coming years to not only update their knowledge and achieve the considered indicators but also assess the diagnostic and treatment facilities required for the management of future disease-related programs. Therefore, assessment of the required facilities in proportionate to data and the existing trend extrapolation can be helpful and used for forecasting; however, it should be noted that various factors can provide the chance for the emergence and increase of the disease, and ignoring them can affect the forecasting for the future. The socioeconomic challenges such as addiction, unemployment, divorce, increasing the age of marriage, and the emergence of new sexual behaviors along with a wide range of effective political, cultural, and environmental factors not only have a mutual correlation but also will create a predisposing ground for making the change in the trend of sexually transmitted diseases and HIV/AIDS in the future. Therefore, given regular data collection and disease-related data including demographic characteristics and epidemiologic information, mortality rates, incidence and prevalence of the disease, etc., these data can be used and predicted in quantitative analysis methods. On the other hand, due to the special importance of the topic, they have always been considered by policymakers and experts. Given these two features, TIA was implemented and used as an appropriate model to estimate data and trends in the future.
1. Implementation phases
In developing a new method in this technique, the advantages of both quantitative and qualitative methods were used, and after determining the subject, the method was implemented in three phases using different tools. Fig 1 shows a general schematic of the model implementation phases.
1.1 Quantitative phase
The first step is to extrapolate the data by using historical data and the related curve fitting. In this step, the future values are predicted by using the previous data and methods such as time series, econometric techniques, trend lines, etc., which are based on a range of observations related to a variable and arranged by time. The main goal of quantitative forecasting methods, such as time series, is to forecast the future, and finally, temporal patterns and short, periodic, or long-term sudden changes in existing trends can be obtained after describing and explaining the data and presenting characteristics of the data such as the upward or downward trend of the data (12, 13). Time series in healthcare services are used to predict the outbreak of diseases, the number of referring patients, the number of personnel needed in different healthcare departments, etc. Most quantitative forecasting methods first determine the general shape of a curve to fit a set of historical data. Then, a curve fitting algorithm is used to select a basic trend line that matches the historical data as much as possible and the sum of the squared deviation is the minimum. Then, the curve is extrapolated using the obtained algorithm to predict the data during the period under investigation. There are different forms of a curve for the fitness of the data, including linear, polynomial, geometric, exponential, and logarithmic trend curves, each of which can well match the historical data and create different extrapolations. Selecting an appropriate form of the general form is of great importance and the important note is that selecting the initial trend line plays an important role in the final forecasting.
Using different statistical programs such as Curve expert, STATA, and R, and applying different methods, it is possible to select the most suitable statistical models, make predictions or draw the best trend lines that match the existing data as much as possible, and achieve a better and more accurate extrapolation o data. It should be noted that the trend impact analysis method improves the preliminary quantitative forecasting and the more we try to select a more suitable initial model in the input of the TIA method, the results obtained from the correction of the trends will match reality more (8). Fig 2 shows different trend lines available in Iran to estimate the data concerning the new cases of AIDS during 1997-2021.
Given the charts and study of the trend of existing data in identifying new cases of the disease, four trend lines have been drawn for data forecasting, and extrapolation can be performed by different mathematical and statistical functions. For instance, the polynomial trend line chart has the lowest number of squared errors (S=121.5) and the highest likelihood ratio (R2=0.71) among the assumed models. This chart can indicate the dynamic nature of new data of patients in different periods and seems good for interpretation of the data. As the fig shows, the data initially faced an increasing slope, and after a few years, there was a decreasing trend in the rate of identifying patients, and it is predicted that the trend will be increasing in the coming years. The data for the following years can be estimated and extrapolated by using this quadratic polynomial equation.
The technique of time series and technical models such as Box Jenkins, which are also known as ARIMA (Auto Regressive Integrated Moving Average) models, can be used to achieve a suitable model for forecasting and extrapolation of data during the appropriate period in the future. Other models such as simple regression, multiple regression, moving averages, and many unknown models that are suitable for time series can be derived from these models. In this modeling method, in addition to the trend factor, seasonal, periodic, and random changes are also considered. Fig 3 shows the trend of detected cases of HIV/AIDs during 1997-2021 and Forecasting using the Box Jenkins model.
1.2 Qualitative phase
In this step, a list of influential events in the future and the desired period is prepared, which will affect quantitative forecasting and deviate the trends if occurred. This step is performed with the presence of relevant specialists and experts in accordance with each field. For instance, in the case study of HIV, which was used in the present research, the experts included infectious disease specialists, epidemiology specialists, future research researchers, expert managers in Iran, regional and international organizations, field experts and therapists, and other experts in this field, including psychiatrists, sociologists, and virologists, etc. The diversity of experts and participants enables us to present a wide range of views, which is not possible for an organization as a trustee or a researcher.
This phase can be implemented by all qualitative tools and methods, including consensus, Delphi, brainstorming, and considering the existing conditions. Judgment and imagination are important elements of the second stage of TIA. In this stage, the surprise-free view toward the future of trends should be changed and important and unexpected events should be considered. First, a list of probable events is provided; the events should be acceptable, believable, and potentially influential, which can result in deviation of surprise-free extrapolation trends. Various methods such as environmental scanning, trend analysis, analysis of inventions and discoveries, observations, etc. can be used to provide the list of events through a comprehensive review of various information sources and following the task objectives, consensus with consultants, or using other sources of qualitative studies in related fields. Events can include technological, political, social, economic, and cultural issues as well as other changes influential on the trend.
Regarding the selected events, it is required to extract several indicators from experts' opinions. First, estimates of the probability of occurrence of each event are made as a function of time, which includes the determination of time, from the occurrence of the influential event to the following times (Fig 4):
A: the trend begins to be affected.
B: the impact on the trend will reach its maximum level.
C: the impact will reach the final rate or a steady state.
Each of the three determined times and their related values are considered completely independent, and various states can be observed. For instance, the maximum impact can be positive and the impact of the steady state is negative, or the impact of the steady state is none, i.e., the impact is momentary and temporary. Sometimes, the maximum impact may be the same as the impact of the steady state, and the impact rate will continue until the end and many other states. The impact rate in two times the above points is of great importance; when the impact reaches the maximum level, it can be positive or negative, and when the impact reaches the steady state.
The next step in this phase is to determine and estimate two key indicators. 1. the magnitude or impact rate of each event on existing trends and forecasting models, and 2. the probability of occurrence of each event in different years within the desired period. This step is also performed by using experts' opinions and gathering and convergence their opinions. This step is an important part of the process and is based on correcting the extrapolated trends, according to the unexpected events likely to happen in the future. After determining these indicators using mathematical methods and the formulas introduced in the next phase, we will consider the impact of these events on the initial forecast.
For instance, in the study of probable predicted events that affect the occurrence of AIDS/HIV and estimates of the occurrence and their impact on quantitative forecasting trends, some issues can be mentioned, such as discovery and access to an effective vaccine for the prevention of disease based on mRNA technology or the exposure and spread of new behavioral patterns in socio-cultural issues or the outbreak of sexual behavior that affects the spread and incidence of the disease (chem sex, etc.) or even the possibility of a military war between Iran and Israel or allies. All these issues can have significant effects on identification of the new cases of the disease. To study the probability of these events in 2030, the opinion of experts and specialists should be asked for the probability of occurrence, the highest impact, and the time needed to reach s steady state for each of the new cases.
1.3 Output: The final phase (correcting the trends and identifying the logic of scenarios)
The simplest approach to the probable and predicted events is to consider them independent. Another approach is to assume a combined state, i.e., if the occurrence of an event affects the probability of the other, the cross-impact analysis (CIA) can be considered a complementary method. To calculate the impact, the product of the probability of occurrence of the event in a specific year or the time unit of the study is multiplied by the intensity of its effect (the greatest impact) in the baseline of that year according to the forecasts from quantitative models. Then, to calculate the corrected estimate value, the index impact rate is added to the baseline in that year according to the quantitative forecasting per time unit of the study. For instance, the value of the initial trend line obtained in quantitative studies to identify new cases of HIV in 2030 was 1500 cases. The impact rate of the discovery and access to an effective vaccine to prevent the disease, with the estimated probability of 50% and the maximum effect of -70% is calculated as:
Impact rate=1500* (50%*-70%) = -535
In other words, the occurrence of the discovery of the effective vaccine will have a reducing effect on the predicted quantitative values in the desired year, and the calculated index (impact rate) is added to the trending baseline of that year, which was obtained from the quantitative analyses, to calculate the corrected estimate value:
corrected estimate value= 1500+(-535) =965
If there are several effective events in any desired period, the algebraic sum calculated for all effects will be applied to the baseline. Finally, the impact and probability of occurrence of events on the existing trends, which can include all possible situations for the highest or the lowest effect, can be considered. For instance, some of events and variables investigated in the mentioned case study assuming affecting the incidence rate of HIV/AIDS, were used to corrected the values of trends obtained from quantitative extrapolations. Among those events, the discovery and access to an effective vaccine, the occurrence of emerging or re-emerging epidemics, the emergence of sexual behavior new patterns affecting the spread of the disease and so on can be mentioned.
A solution is to calculate the standard deviation from the mean for the obtained index, which is calculated in the positive and negative intervals of this confidence interval, the upper and lower limits, and its median is considered as the corrected trend line resulting from the TIA approach. Another solution is to calculate all the high probabilities of a positive effect on the existing process, which will be considered as the upper limit, and to calculate all the low probabilities of a negative effect, or in other words, the opposite effect in the process, which will be considered as the lower limit, and the mean will be the corrected trend line of TIA (11). Fig 5 shows the trend impact analysis approach output to estimate the new cases identification of HIV/AIDS up to 2030.
There is also another solution, i.e., the use of the Monte Carlo approach in which experts provide estimates of the occurrence of certain events in different periods in the future. The sum of these probabilities can reduce uncertainties of decision-making. In a Monte Carlo simulation, random values of a large number of loads are generated for all uncertain parameters in an equation that forecast future values, and for each uncertain parameter, a probability distribution is defined from which random values are generated, allowing the calculation of upper and lower confidence limits for the predicted values (10).
In other words, in the future, the upper limit, lower limit, and the average of these effects as the corrected trend line of TIA will happen and consequently, the corrected trends will be prepared, and the related scenarios will be processed, interpreted, and narrated. Indeed, the trend impact analysis approach can be considered a way to quantify the scenarios. In other words, probable events, their interpretations, and description of the new trends obtained from the upper and lower limits, which are drawn in the form of trend lines, can be considered as a scenario that happens with different sequences of events, and their overall effects will determine the shape of the future curve.
The scenarios can be used as a tool for learning. Given the in-depth and extensive information about the influential forces on a certain topic, it leads to a deeper understanding of environmental changes. They can also be used to make better decisions at present and resolve the problems and inadequacies that prevent us from reaching the desired future(14).