Health Manpower Forecasting: A systematic Review for Models and Approaches

Background: An appropriate planning requires forecasting the demand and supply of health manpower in an accurate manner. Objectives: The paper objective was to review health manpower forecasting approaches and methods and propose the improvement features for effectiveness of health manpower planning. Data sources: A systematic review was conducted for studies published in English from 1970-2014 using Pub Med, Science Direct, Pro Quest, and Google Scholar databases. Study eligibility criteria: Review articles, qualitative studies, questionnaire driven surveys, retrospective and prospective studies describing or applying various types of forecasting approaches and methods in health manpower forecasting were included in the review. Data collection and study appraisal: Two reviewers (SR and MA) independently appraised the studies to identify relevant ones to be included in the review. Results: A total of 128 studies were included in the review. As a result, two main categories of approaches (conceptual and analytical) for health manpower forecasting were identified with their strengths and weaknesses. Conclusions: Future forecasting methods should benefit from advantages of current approaches and overcome some of their main limitations. Applying a method which incorporates a wide range of factors affecting supply and demand, facilitates the collection of good quality data, monitors changing trends over time, considers the uncertainty of health systems and has a dynamic approach is proposed as an improvement roadmap for future


Introduction
Every health care systems are dependent on human resources 1 . To have a well functioning health care system, it is inevitable to have adequate number with proper performance of health manpower 2 . Imbalance in manpower demand and supply increases health care costs through provider induced demand 3,4 while insufficient supply may cause health needs of the population remain unmet or provide with lower quality 4,5 . Therefore policy makers are to make certain about balance between supply of and demand for manpower 12 . First by estimating current workforce supply, the adequacy of supply will be assessed. This gap analysis identifies proportion of population not covered by health manpower services 6 . Then the workforce requirements will be forecasted to define an optimal size for manpower. After defining an appropriate number of health manpower needed to meet requirements for health services, there would be necessary to develop some policies to keep supply and demand of manpower in a balance 9,10 . There are several forecasting methods with their own strengths and weaknesses 2,8, . Most of them do not have the ability to capture all key factors affecting demand and supply of health manpower. They also lack the ability to consider the dynamic interactions among full range of variables and to make possible scenarios in response to what-if questions 27,33,45,46,49 .
To propose a single comprehensive approach incorporating strength point of other methods, it is required to conduct a systematic review on existing approaches, their advantages and limitations. In this way, challenges of such models would be analyzed to determine possible roadmaps for improvement. Almost always, this kind of study has been ignored in the literature and there is rarely a well appraised description of forecasting methods in which both conceptual and analytical approaches are examined together.
This paper aims to provide a comprehensive overview toward human resource forecasting approaches and consequently propose the features that potentially improve the effectiveness of such an approach.

Methods
The study method was structured based on PRISMA statement 50 .

Eligibility criteria:
1) Types of studies: Reports, books, review articles, qualitative and quantitative studies were included in our study.

2) Study subjects:
We included in the review those articles or reports describing the conceptual and analytical approaches in health manpower forecasting also those applying various forecasting methods to estimate health manpower supply and demand both in developed and developing countries.

3) Types of participants:
We included in the review all health care professionals such as: doctors, dentists, nurses, dieticians and pharmacists.

Search strategy
We chose a time period from January 1, 1970 to December 30, 2014 for searching process of the study. Similar search terms including "forecasting approaches OR forecasting methods OR predict OR project" AND "demand OR supply" AND "health manpower OR health workforce OR health human resources OR health care professionals" were used for all databases. The search was restricted to articles and reports with full text accessibility published in English during the years 1970-2014.

Study selection
Data base searching led to 2816 documents. After removing duplicates, two reviewers (SR and MA) independently appraised the publications according to the title and abstracts. After this level of filtering, studies were checked to meet the eligibility criteria.
In reference to the inclusion criteria of "study type", "study subject" and "types of participants", we included review studies describing conceptual approaches and analytical methods in health manpower forecasting. Original articles applying various forecasting methods to estimate health manpower supply and demand were also considered to be eligible for inclusion in the review (figure 1). When there were more than one publication with similar method and presented data, authors decided to include only the most recent publication. Detailed information about searching process and final selection of publications is given in bellow table.

Data collection
As no systematic review with a similar objective of this study was available to provide an applicable extraction form, authors designed an extraction data sheet based on study questions. This form had been used to collect data on studies references, countries and designs, type of forecasting approaches whether discussed or applied with their strengths and weaknesses (table2). For each study, two authors (SR and MA) completed the sheet independently. Any disagreement was solved through consensus.

Results
In an initial search restricted to language, publication time and full text accessibility 2816 publications were found. Screening documents based on their titles and abstracts revealed 1433 duplicates or non-relevant studies. After reviewing the full texts of remaining studies 1051 of them were excluded because of non consistency with inclusion criteria. Among which 570 did not meet our criteria for study subject, 15 were not compatible with inclusion criteria of study design, 350 studied forecasting approaches among non health workforce and 116 provided no data on strengths and weaknesses of forecasting approaches. 211 documents were also removed due to similarity in method and data provided. Considering the reference lists of remaining studies led to 7 additional publications that met our inclusion criteria and were not the duplication of other studies. Documents' flow in a searching process is depicted in figure 1. As a whole, 128 studies were included in the review both from developed and developing countries. More than 50% of them were published in or after 2000.
Two main categories of forecasting approaches in health manpower planning were identified [51]. The first category is conceptual approaches which clarify different affecting factors involving in health manpower planning. Each of the approaches relies on specific factors to forecast health manpower supply and demand.

Conceptual approaches
Main conceptual approaches for health manpower forecasting were identified from the review.

Manpower to population ratio Definition
This is a simple method to forecast future requirements for health manpower by calculating the number of staff needed to serve future population in a way they are currently being served 35 . This ratio is often identified by governmental policy makers, professionals in expert panels or technical agencies like World Health Organization [19] . There are three main assumptions regarding to this approach: current number, skill mixture and of distribution of manpower are adequate; productivity, age and sex pattern of manpower will keep on unchanged in the future; number and demographic specification of manpower alters according to current observed trends over time [39,52] .

Strong sides
This approach is easy to understand and use by simply depending on existing data bases. Applicability of the method in a national context besides its comparability among different regions and countries are other main advantages [4,19,24,26,45,53 ] .

Weak sides
The defined ratio in this approach is supposed to be appropriate and efficient in a way that future estimation should be based on it. This faulty assumption will put a great obstacle in a way of manpower planning improvement [13,26] . Synonymous of supply to need and considering manpower as a homogeneous category are other faulty assumptions regarding to this approach. In fact, differences in productivity, type of activities manpower perform and population dissimilarities are neither reflected [19] . Explaining health sector functions and ways of increasing manpower productivity are features that are mostly ignored in this approach [17,25,26,36] . Disability to demonstrate distributional imbalances and taking little account to age distribution of the population, the epidemiological features and socio economic conditions are other weakness points [19,26,31] . Overall, the real capability of health system and actual needs for services are not taken in to account in this approach [4,17,19,28,39] .

Demand based approach (Requirement model or utilization based approach) Definition
This approach calculates utilization rates of health services among different population categories and then adjusts the calculated rates to future characteristics of the population in order to find out service demands. Finally service demand will be converted to required health manpower 25,26,54 . The approach is recommended to be used in countries with a dominant private sector and little imbalances of services to different segments of the population 26,55 . There are three related assumptions: existing demand for health care services is suitable and will be met by present number, skill mixture and manpower distribution; age and sex pattern of manpower will keep on unchanged in the future; number and demographic specification of manpower alters according to current observed trends after a while 4, 39 .

Strong sides
Availability of data and the assumption that providers currently deliver services in accordance with societal values are among the advantages can be mentioned for [16,22] . The approach can capture future forces of increasing number and ageing population on amount of health care services demanded 4,19,24,39,45,53,56 .

Weak sides
The main problem of this approach is neglecting the consequences of errors arising from its faulty assumptions. It is notable that utilization of services is dependent to supply, therefore any incorrect estimation of service requirements will lead to changes in the number of health manpower per capita population [45,53] . There is also an interaction between demand, need and supply which is ignored in this approach 45 . Second issue is that demand for services is not a proper representation for health care services need; somewhat it is a measure for perceived need 8,10,17,29 . The third invalid assumption is that demand based approach considers current utilization rate efficient which can be regarded as a basis for future demand estimation 28 . As a result, such approach may assign current injustice and insufficiencies in to the future 18,19,21,35 . The fourth issue is that professional, economic, political, technological and socio cultural factors are rarely well intended 27,33 . Another disadvantage is the assumption that there is no excess use in the utilization pattern of the population and people use only the amount of health care services which is needed to maintain or maximize their health condition 22 . Finally the approach assumes that there is no room for health improvement of population and the health status is optimal and should be maintained in a same way in the future 22 .

Need based (Epidemiological) approach Definition
This approach defines and predicts health deficiencies of a population and level of appropriate services needed to make best standard of health status possible 4,34,39 . This method has three main assumptions: all health care needs are possible and supposed to be met 53 , allocating resources for the purpose of meeting all health needs of the population is a cost effective approach regardless of other needs existing in different sectors of the country, methods of dealing with needs are clarified and can be recognized without facing any difficulty, and finally health care resources are always employed in reference to actual amount of need 39, 53 .

Strong sides
When health needs are defined, this approach can identify a normative standard for manpower supply to meet service requirements of a population 25,26 . The method is more practical for planning services with defined standards of need for specific population characteristics 24,40 . It is also useful in comparing burden of illness among different population of an area 24 . Change in health care delivery and work styles among providers are considered in this approach 28,40,41,52 . Unlike utilization based approach, needs based planning assumes that current number and distribution of manpower are not necessarily optimal 30,38 . Method also includes unmet needs in the estimation process and is useful in planning preventive and national public health programs. The approach is mostly suitable for planning in countries with a high public awareness and sound planning capacity.

Weak sides
These standards are mainly based on expert panels and technical opinion of a group of professionals and do not rely on objective techniques for identifying health service requirement {Weiner, 1987 #36;Roberfroid, 2009 #23;Solutions., 2013 #35}. The approach does not reflect the capability and eagerness of population to pay amount of money needed for health services 46,57 . Although the approach seems rational but it is most difficult to performed. Defining health needs is very challenging. Measurement of needs is also a major difficulty. The problem may be even more critical when there are no established norms yet 19,24,39,40 . There is a fact that health needs change dramatically over time and there is a necessity to monitor these changes over time 10,19 . Needs based approach often neglects the effects of new developments such as technological advances on their estimates 10 . The approach also ignores the efficiency in resource allocation between health care and other activities. Therefore considering a connection between principles of needs based approach and economic values such as opportunity costs of resources have a great importance 53 . Benchmarking Definition This approach uses health manpower to population ratio as a proxy for need 13,18,23,26 and relies on identifying a service standard in a health system which is supposed to perform appropriately and provide affordable and sustainable health care services. Then the standard would be applied to a comparable population 22,40,43,44 . The approach assumes that current condition continues in to future to estimate health manpower requirements 39 . The second assumption confirms that defined ratio is the finest and most valid possible in providing suitable health care services for a new population 46, 58 . It is also believed that desirable health for a population can be met with equal number of manpower after modifying for demographic characteristics of a population [24,39] .

Strong sides
This approach is simple to perform and able to compare manpower resources between regions with same health care system characteristics 16,22,23,39,40 . Weak sides Such models are mostly criticized because of the faulty assumption that health condition of a particular population is desirable for other areas and can be met after some adjustments in demographic characteristics. Considering the defined ratio as the finest and most valid possible is also a doubtful assumption that endangers the utility of the approach. Similar to manpower to population ratio, this approach also neglects changing in productivity, age and sex pattern of providers over time. In fact benchmarking is not a suitable approach for regions with infrastructure differences and does not take in to account the geographic distribution of population [14,24,31,47] . Service Target Approach Definition Service target approach links manpower provision to specific health care services that are mentioned as an objective to be produced for a specific number of population or for a defined region [2,26,58] . It assumes that needs are the same among populations and there is no heterogeneity in productivity and practice style of providers 19,26 .

Strong sides
This approach is politically appealing and most suitable to be used at a local level 32,45 . The advantages of this method is easy separation of health services into component parts and simplification of cost estimates also providing an active approach towards health services improvement. Being flexible and requisite of modest data are another advantages that could be mentioned 19 .

Weak sides
Standard setting in this approach is based on desires rather than reality and this is one of the shortcomings of the approach as errors of judgment would occur consequently. In fact, targets are not clearly confirmed in all situations and may only be a sign of experts' opinions; therefore suffer from subjective manner [19] . Ignoring the heterogeneity among populations and providers is also a notable weak point regarding to this approach 19,26 .

Mixed approach Definition
Since different forecasting approaches have their own strengths and limitations and vary in assumptions, mixed models are introduced to be useful in estimating health human manpower in a most effective way. In this approach an efficient number of required manpower is determined by integrating several factors of health human resource conceptual approaches and creating a single comprehensive model. In reality, however mix (combined) models are used in projections 32,39 .

Strong sides
Mixed approach considers a variety of economic, social, political and technological factors influencing health manpower forecasting. The dynamic nature of population health needs and providers' characteristics is taken in to account by the model. To estimate future requirements for health manpower the interaction between supply and demand factors is fully considered. Model can be applied in both national and local levels and provides an opportunity to investigate different health policies before putting them in to action 6,10,27,34,49,53,59 .

Weak sides
Demanding varied sources of data and somehow complexity of the approach are the main limitations of the method 33,53 .

Different mixed approaches
The most common mix approach relates supply and demand based parameters in a way to build a supply to health care utilization ratio 15,20,42 . Another mix model is effective demand based approach in which economic values harmonize the epidemiological concerns related to need based approach 53, 60 . As there are clear possibilities for resource tradeoffs, it is not wise to ignore financial resource constraints in forecasting. This approach ensures that human resources are organized efficiently 29 . A conceptual model "dynamic system based framework" was introduced as a mix model by O' Brien, et al considering:1) health levels related to population (needs based factors), 2) health care utilization rates (utilization factors), and 3) economic, social, environmental and political factors affecting health expenses (effective demand based) 4 . This model integrates two main methodological approaches of needs and demand based to introduce a comprehensive model 32, 53. Effective infrastructure approach is another mixed approach which is based on needs based parameters added to infrastructure variables 4 . In this approach, physical facilities of a health system should be clarified to make use of workforce according to available infrastructure 4,21 . Integrating Markov population model with principles related to benchmarking approach is another mix approach which was used for projecting endocrinologists in USA 37, 39 .

Analytical approaches
There are some analytical methods used in manpower forecasting. The methods are useful in decision making and supporting plans and policies related to manpower system 35,39 . As conceptual methods are different, analytical approaches are also varied. A desired outcome expected from analytical methods is to provide practical and easy to understand information for managers, planners and policy makers. These approaches must have the capacity for longitudinal analyses. They must take in to account the inherent uncertainties in manpower supply and demand variables and use high-quality databases to achieve viable data also capable for scenario planning using a variety of techniques 46 . Analytical methods in health manpower forecasting are categorized in to some main groups: extrapolative methods including trend analysis, explanatory variable methods such as econometric models and regression analysis, operations research methods including Markov model, linear and non linear programming, queuing theory, simulation and system dynamic and finally workload indicators of staffing needs developed by WHO 10,32,35,46,59 .

Trend Models Definition
These models correlate historical data trends to project future requirements for manpower based on different scenarios and altering key assumptions 46 . Trend models also called supply projection models consider current amount of health care services offered by health manpower. The model defines need as a number of manpower necessary to add current supply to maintain or attain a pre-determined level of service 4,39 . Consequently, manpower requirements are not calculated on the basis of population health needs 35,[61][62][63][64] . This approach assumes that current size, skill mix and distribution of manpower are adequate and their productivity will remain constant in the future 4, 53 . Another assumption is that the number of manpower and their demographic characteristics changes over time in a current observed trend 29,41,53,65-67 . In such models it is assumed that entry, re entry and attrition rates of manpower from current supply would remain the same in the future 68 .

Strong sides
The advantage of this approach is that forecasting is easily possible by using historical data and retrospective trend analysis 62,69 .

Weak sides
The shortcoming is that the forecast only reflects the historic trends of manpower supply not the actual number existing in the future. The model incorrectly assumes that past trends in health manpower supply or demand continue in to the future 39, 62, 70-72 . However relying on historical trend and considering the growth rate dependent on the start and end points of time intervals is fully problematic and the assumption of a long-term trend is also impractical 73 .

Econometric Models Definition
This statistical technique suggests that the number of health manpower required is a consequent of population's demand for health services which is related to economic factors such as health service utilization, access to services, consumer health preferences, income or set of market variables 35,65 .

Strong sides
Econometric approach including regression models is useful in examining the relationships between stock, wages, demand and budgets. These analyses can be used for forecasting population demand for manpower considering future growth in national income of a country 35, 51.

Weak sides
Model often ignores political/socio/economic variables, changing health system and the relative outcomes 32 . Similar to trend models, econometric approach does not consider changes in assumptions and data flows. In fact the model is not comprehensive and able to capture full range of variables and dynamic interactions among them. Another limitation is that model is not self regulating to allow for output changes as a result of manipulation in some variables or changing their power of importance 46, 74-76 .

Artificial Neural Networks Definition
This method allows for nonlinear connections between input and output variables and for learning patterns in data [77][78][79] . Artificial neural networks are consisted of a range of variables connected to each other in different layers in a way that one layer obtain input from the former layer and passes the output to the following one 80 .

Strong sides
Neural networks are able to simulate system behavior and capture many features of real situation as networks when input and output variables are clearly identified. They can also provide high precise results comparing to regression models and do not necessitate modelers to have special knowledge or understanding of modeling 80 . The method can solve some of the limitations related to former models such as linear regression by estimating nonlinear relations as well 81 . The accuracy of these models can be evaluated by mean squared error 80 .

Weak sides
One of the shortcomings of the model is that it needs a lot of data to be run correctly and its accuracy is largely dependent on data quality. Second, it is difficult to define analytical relationship between input and output variables of the model. Likewise, as the model finds any relationship between each set of independent and dependant variables there is a risk of finding unreal associations 82 . Finally the model is not able to automatically adopt itself to environmental changes; thus in response to any alteration the network must be restructured 80,83,84 .

Markov Models Definition
Estimating the supply of manpower in a region using Markov and semi Markov methods includes flows of manpower in supply forecasts. In this type of method, fluctuations in labor levels are projected using historical transition rates 85 . These fluctuations are projected by analyzing historical trends in staff retention or movement, and keep it in to the future with the same trends of the past 86 .

Strong sides
This method has commonly been used in studies to forecast recruitment requirements of manpower in a local level, but it is also possible to aggregate this information across a nation and provide an overall estimation. Such models increase the precision of forecasting, and are appealing because they can be analytically solved 53 . They are able to give a variety of parameter values, also use confidence intervals and test statistical significance for different estimates 45 . Ability to simulate the consequence of different policy changes is another characteristic can be mentioned for 39, 45, [86][87][88][89] .

Weak sides
The main weakness of these models is that they assume current behavior and health care delivery patterns will be constant in to the future and project future supply in compliance with current number and age profile of manpower 88 . Trends are considered to keep on developing according to the past. The model also requires major generalization of a problem to make it fit for the projection purpose 53 . Uncertainties in inflow and outflow parameters are often ignored and calculation of statistical confidence intervals is impossible. Inability to capture dynamic variables and the interaction between them is also a shortcoming of the model 14, 89, 90.

Linear programming Definition
This method analyzes fluctuations in labor levels using an objective function as well as organizational and environmental constraints 39, 91 . Optimal number of manpower recruitment is identified in a way that total cost of planning including recruitment, promotion, overstaffing, wastage and retention are minimal 92 .

Strong sides
This method helps to use available resources in a best manner and is appropriate in problems with objective and constraint functions in a linear equation form.

Weak sides
Linearity of objective and constraint functions as one of the model assumptions is not true in real life conditions. As model variables and parameters increase, the number of equations and difficulty to perform the models would raise dramatically. Uncertainty is one of the factors neglected in this model. It also lacks the possibility to perform scenario making and ignores political, technological and socio cultural features in the modeling process [91][92][93][94] . Failure to incorporate as much as variables related to manpower supply and demand also the correlation among them is a remarkable shortcoming of the model.

Queuing Theory Definition
Queuing theory is a mathematical model to facilitate service providing in an organized manner. To apply this method, it is necessary to know the distribution of arrival time, service time and the queue discipline 73,92,95 . The situation with characteristics of a Poisson arrival process, an exponential distribution of service times, a first come first served queue discipline and a service arrangement of multiple services acting in parallel is suitable for applying queuing theory 73,96,97 .

Strong sides
Queuing theory can be used in manpower planning and is able to identify how many staff is required to give adequate services to a population. Comparing to simulation, this analytical tool is simpler and needs much less data to answer health manpower planning dilemmas. The results of the model can be solved analytically and are applicable at different scales including individual units/departments, healthcare facilities and healthcare systems in district level 53 .

Weak sides
One of the shortcomings of the model is that it uses an average of variables instead of real numbers 73 . It also supposes that system operates in a steady state, times spent for service delivery and service rate are known and the latter is greater than arrival rate also time spent for services is stated by "negative exponential probability distribution" 73 . In fact the method fails to consider the dynamic nature of health system. The other weakness is that the method is not appropriate in complex cases related to national levels. Model also fails to consider full range of variables to reflect the complexity of real life situation 73,98,99 .

Nonlinear programming Definition
Nonlinear programming is the course of action in terms of various constraints and a set of variables along with an objective function to be maximized or minimized where they are not in a linear form 100,101 .

Strong sides
Most of the situations in real world are consistent with nonlinear programming. The model has the ability to capture the complexity and nonlinear interactions between system components which is regarded as its main strength point 92,100,[102][103][104] .

Weak sides
Nonlinear programming solutions are often local optimums 92,100 . Neglecting the effect of wide range of variables, dynamic interactions among them and uncertainty inherent to the variables are among the main shortcomings of this method 92, 100, 102-104 .

Simulation Definition
Simulation is a method of designing model for a real world phenomenon with the intention of understanding its behavior or assessing the effect of different strategies on it without having to manipulate real world systems. This method models the real system through mathematical relations between variables and their likelihood distributions.

Strong sides
The model provides the opportunity for modelers to understand the system behavior or evaluate different strategies without having to manipulate real world systems. The ability to evaluate system behavior under various scenarios is its important usage in policy making in response to different situations. The method is easier to perform, requires less simplifying assumptions and provides more reasonable reproduction of system than mathematical methods 2,64,65,105 . Facilitating sensitivity analysis over the model variables also involving stakeholders in the process of modeling are among other features of simulation 106 . There are two related approaches used to assess uncertainty in health system forecasts: "deterministic sensitivity analysis and stochastic simulation". Stochastic simulation generates more realistic and meaningful forecasts than deterministic sensitivity analysis. Where there is no systematic data for some features of the problem, stochastic modeling can be useful to take uncertainty in to account by means of random variables 4, 39, 107-109.

Weak sides
The necessity to use more detailed data makes this method costly to be applied 2, 64, 65 . Simulation is not a precise method able to achieve accurate solutions or single determined answers. Instead, it generates multiple system's responses to different running conditions 110 . As there is no standardized approach for simulation, models generated by different modelers may vary extensively 105 . As the number of parameters increases, model faces with difficulty to find the best possible answer. To ensure the accuracy of the model, validation methods are required 111 .

System Dynamic Modeling Definition
System dynamics (SD) initiated by Forrester (1961) is a modeling method consisted of causal loops and stock-flow diagrams depicting interactions among variables and the structure of a system flow 112,113 . Stock variables represent the current situation of system; while flow variables explain changes constantly occur in the stock [112][113][114] . Method is used to identify the organization, performance and behavior of complex systems.

Strong sides
The method enables us to simulate real systems with complicated structure and understand their feedback interrelationships and dynamic nature of variables 113,115 . Interaction among system components causes the dynamic behavior of systems over time which is able to be considered in SD modeling. Method allows for capturing dynamic impacts related to time delays and feedback relations of a system 113 . In fact, managers profit from the interactive environment created by system dynamics 113,114,[116][117][118][119] . Also ability of the model to depict effects of different policies makes it a useful method for better policy making in different decision areas including manpower planning 113,118 . By reviewing manpower projections and altering assumptions when appropriate, possible surpluses or shortages in future manpower can be recognized 113,116 . Experiencing long term effects in a short time by testing what-if scenarios can help modelers to design better strategies for system development. The model is useful for forecasting purposes in medium to long term periods of time 113 . The main advantage of the method is its capability of being operated in a fast manner which is suitable for decision makers to effectively plan for the future in a proper time and minimum risks 114 . Furthermore, qualitative data can be modeled in this approach 114,120 .

Weak sides
To test the method validity, each parameter should have an equivalent in the real world. Obtaining information on these parameters, variables and potential interactions among them is not a simple process. Additionally, drawing doubtful and hypothesized feedback loops might generate a model that does not characterize real world behavior. Model is also in danger of becoming complicated as a result of growing in number of variables or considering multiple interactions 113,121 .

Workload Indicators of staffing needs (WISN) Definition
This method has been introduced in 1998 by World Health Organization for determining activity norms for each category of health professionals and converting them into relative workloads in order to obtain a method for defining health manpower requirements [35,122,123] . By dividing the actual number of manpower to the required level, WISN can be calculated 35,123,124 .

Strong sides
The method is a helpful management guide to make valid decisions about employment, distribution and development of health sector manpower 124 . The method is a combination of both expert judgment and measurement of activity standards to define workload norms for each category of health manpower 124 . Deciding on manpower allocation through geographical areas and their functional deployment to achieve an optimal number and staffing pattern are other uses of the method 123 . Weak sides WISN is estimated by means of annual data available for work activities done in a health care facility. Thus, accuracy of the method is dependent on the accuracy of annual data 124 . When there is a poor information system in a facility, results of the method will be incorrect. It is mostly neglected that low level of workload in a health facility might be the result of lacking materials 124 . Shortages of supplies, poor services, inconvenient location of facilities, short hours and their potential effects are usually ignored in this method 123,[125][126][127] . Considering a wide range of variables related to manpower supply and demand, their correlation and dynamic nature are not possible in this method 123,[125][126][127] . more effective method in health manpower forecasting Most of the health manpower forecasting methods do not regard entire dynamic variables playing important role in manpower forecasting and their causal relations. They also lack the capability of making scenario to help policy makers in effective decision making 128 . Modern forecasting approaches need to meet the key features found in the literature in order to develop a dynamic and comprehensive method necessary for today complex health care systems 32,46,129 . Some of the key features are: taking in to account full range of dynamic variables affecting manpower supply and demand, considering their inherent uncertainty and complex interactions, having a system level perspective with a dynamic approach and ability to scenario making useful for policy making purposes 128 . Our study recommends using a dynamic model which meets following features 4,39,128 . Incorporating all relative factors with dominant effect on manpower supply and demand, monitoring trends of these factors over time, considering complex interactions among these factors, applying what if scenarios helpful for policy making purposes. Providing high quality and longitudinally data around the key factors are the essentials in making a valid and reliable manpower forecast 8,29,32,37,[128][129][130][131][132][133] . Considering the impact of emerging technologies, socio cultural evolutions, economic changes, political incentives, demographic and epidemiological trends on health system capacity to meet these requirements are the examples emphasizing the importance of monitoring trends in affecting factors over time 32,46,130,134 . Taking in to account multifaceted interactions among variables influencing manpower supply and demand is another important attribute. Illuminating all possible connections between variables disclose the complex nature of manpower modeling 113,114,116,119 .
Checking what-if scenarios is another important factor which should be regarded. Sensitivity analysis allows policy makers and planners to make different evaluation of required services and consequently plan for the number of manpower required 32,113,114,116,117,120 .
A system dynamic modeling approach takes in to account a complete number of dynamic variables in manpower planning and consider both quantitative and qualitative aspects. The model is useful in better policy making for manpower supply and demand by taking in to account a combination of various perspectives 32 . Ability of the model in responding what if scenarios can help planners and policy makers to decide more effectively in manpower planning 112-114, 116, 117, 119, 120 . Discussion This paper reviews different forecasting approaches and methods for health manpower and demonstrates the issues that develop the effectiveness of health manpower forecasting. There are several conceptual approaches to forecast health manpower supply and demand [2,8, . Most of them are supply based approaches including manpower-population ratio which is helpful for estimating manpower requirement in a national level 13,17,19,25,26,28,31,36 . Service utilization models are used to measure health manpower requirement in private sector. Similar to manpower-population ratio, demand based approach covers main regional imbalances. Both approaches also disregard the effects of manpower planning on health system outcomes 32,39 ; they also ignore the complexity and dynamic nature of health system and do not consider population health needs in manpower planning 2, 8, 14, 16-18, 21, 22, 25-27, 29, 33, 35, 45 . Needs based approach seems to be reasonable because instead of considering the current level of health manpower to be necessarily adequate; it focuses on population health needs 16,19,24,30,34,40,41,46,48. In health service target approach, a degree of health manpower supply is defined to respond targeted health services in a health system for a particular population. This approach is politically accepted and makes manpower planning possible in local levels 19,26,45,55,58 . There are a number of limitations regarding to above mentioned approaches. One of the main shortcomings is ignoring the health system's dynamic nature and comprehensiveness of several factors which have important effect on health manpower planning 27,32,33,39,45,128 . The last approach is benchmarking which relies on homogenous health system indicators to determine health manpower requirement. Despite of the approach simplicity, it suffers from similar important deficits 23,43,44,47,58 . In fact none of the five approaches captures a comprehensive set of factors influencing health manpower demand and supply including economic, socio-cultural, political, technological, disease incidence, service utilization and other affecting factors. The dynamic relations among all variables cannot be shown in above approaches and making what-if scenarios to respond different situations is also impossible 27,33,45,46,49 . Our study proposes a comprehensive health manpower model which includes as many of influencing factors as possible such as demographic factors, economic, social, political and technological factors, service utilization, health needs etc with their causal effects 8,10,29,32,37,128,[131][132][133] . A road map for developing an effective model in manpower supply and demand forecasting is supposed to be the one that considers as many variables involving in manpower planning as possible, effective monitoring of their trend over time, their inherent uncertainty and complex interactions among them along with the ability to forecast supply and demand for a long period of time and for a large number of population 112-114, 116, 117, 119, 120 .

Conclusion
Lack of adequate attention to this key process, will decrease the quality of health services patients receive and increase the burden of expenses enforced to health system [135] . There are a number of forecasting approaches and methods with their own strengths and weaknesses to forecast health manpower requirement. System dynamic modeling has the ability to consider a complete set of dynamic factors in human resource planning and regard both quantitative and qualitative dimensions. Model is also useful in stronger manpower planning for supply and demand and making reliable policies for their staffing, allocation, education and attrition objectives 113,114,120,136 . Ethicalclearance