Modeling the annual health costs per capita of
I.K.A. insurants
by
Mertika Katerina,
s1629115
Thesis Supervisors:
Prof. dr. R.H. Koning
Dr. C. Praagman
DEPARTMENT OF ECONOMETRICS
FACULTY OF ECONOMICS & BUSINESS
UNIVERSITY OF GRONINGEN,
THE NETHERLANDS
Abstract
In this specific study, the relationship between the general indicator of inflation of Greece and the special inflation of health care costs of the biggest public insurance fund of the country, IKA (Social Insurance Institute), will be examined. We will also examine if other factors, such as the average age of the insurants and the usage frequency of the health Services affect the relationship mentioned above.
Contents
1. Introduction
5
1.1 Health costs background ………. 5
1.2 Aim of the study ………. 6
1.3 Research questions ………. 7
1.4 Structure ………. 8
2. Theoretical background 9
2.1 Definition of social insurance ………. 9
2.1.1 Social insurance systems ……… 10
2.1.2 Historic evolution of social insurance ……… 11
2.2 Social insurance in Greece ……….……… 11
2.2.1 Social insurance types ……… 12
2.3 Analysis of IKA benefits ……….... 13
2.3.1 Direct costs ………. 14
2.3.2 Indirect costs ……….. 15
2.4 Inflation in Greece ………. 17
2.5 Demographic trends ………... 18
3. Methodology and Research Process
21
3.1 Data analysis 21 3.1.1 Data description ………. 21 3.1.2 Average age ………... 23 3.1.3 Inflation ………..………. 24 3.1.4 Direct costs ………. 27 3.1.5 Direct costs ………. 28
3.1.6 Health benefits use ……….. 30
3.2 Theory 32
3.2.1 Regression analysis ……… 32
3.2.2 Dickey-Fuller’s test for stationarity ……… 36
3.2.3 Autoregression analysis ………. 36
3.2.2 Model’s validity ……….………. 37
3.3 Health costs modeling 39
3.3.2 Direct benefits per capita ………... 39
3.3.2.1 Total direct costs……….. 39
3.3.2.2 Analytical total direct costs……….. 41
3.3.3 Indirect benefits per capita ………. 44
3.3.3.1 Total direct costs ……….. 44
3.3.3.2 Analytical total direct costs……….. 45
3.3.4 Arima models ……….……… 50 3.3.4.1 Days of hospitalization ……… 50 3.3.4.2 Count of drugs ……….… 51 3.3.4.2 Visits .……..……… 52 3.3.4.2 Checks …….……… 54
4. Forecasts 55
4.1 Direct costs………...………... 55 4.1.a Illness ………... 55 4.1.b Birth …...………... 55 4.2 Indirect costs………...……… 56 4.2.a Doctors ………..………. . 57 4.2.b Hospital .……… 58 4.2.c Drugs ……… 595. Conclusions
60
6. Appendix
61
a. Average age ……… 61 b. Inflation …….……… 62 c. Direct costs ………. 63 d. Indirect costs ………... 65e. Use of health benefits ………. 66
1. Introduction
1.1 Health costs background
Over the last decades, the cost for health has dramatically increased at most of the countries worldwide. This fact makes the governments worry, as they are asked to re-assign the designation of the social schemes.
The rate of the increase on the health expenses has been greater than this of the total economic magnification. According to the OECD’s report, in 2003 the OECD countries spent in average 8,8% of their GDP on health, against 7,1% in 1990 and a little over 5% in 1970.
The increase on health expenses comes from many factors. Generally, the OECD countries having a higher GDP per head tend to spend more money per head in health. The population ageing equally contributes to the increase of health expenses. The percentage of the population aged 65 years old or more has been increased in all of the OECD countries and is expected to increase further in future. As the needs of the population groups of older age for medical care and lasting care tend to expand, the population ageing is expected to increase the expenses on these areas.
Some other factors that come upon the cost of health are the use of new technologies in the sector of prevention, diagnosis and treatment and the advancement in the science of pharmacology. The research cost is included in the final prices of these products. (Health at
a glance: OECD Indicators-2005 Edition, p.3-4)
1.2 Aim of the study
In Greece, social insurance is widely developed. Public funds cover the majority of the country’s population. Institute of Social Insurance (IKA) is the biggest of them. Its insurants overcome half of the country’s population.
The aim of this study is to determine an econometric model to analyze the cost per head of health expenses for every insurant of IKA. The claims for benefits of IKA are equal for every insurant. and expenses are independent of personal income.
The increases on the costs of benefits are related to the increases on the prices of services inside the country. For that reason, an attempt is made to find a relation to the general consumer price index, which is the annual inflation of the country. We will examine if the costs of dispensations to the insurants increase on the same rate as the costs of concrete benefits (e.g. the hospital care) and also if the costs of concrete benefits increase in parallel. Next we will examine whether there is a correlation between the annual increase of cost of health benefits and the increase on the average age of Greeks during the years 1957-2006. Unfortunately, there are no data kept at the Actuarial department of IKA, relating the costs to the insurants’ age and thus, we cannot examine the correlation degree between age and diseases.
1.3 Research questions
The determination of the factors that contribute to the increases on the cost of health is very useful to IKA. By turning to account the results of this study, IKA could find a help in drawing its policy for the next years. For example, it has to decide whether and to which degree it is going to increase the contributions or to track any particularities to specific health sections that need to be controlled more carefully, on the aspect of their management.
In principle, the main question that will be investigated in this study is the following:
Can we predict the health costs of IKA in the upcoming 20 years using predictions of the General Index of Consumer Prices of Greece and predictions of demographic developments?
We will address it responding the following subquestions:
• What is the influence of both (i) the rate of CPI and (ii) the mean age of insurants, to IKA primary health costs
o for Benefits: Hospital, Doctors and Drugs,
o for Dispensations: absence from work because of illness, birth and accidents.
• Concerning the demographics: hospitalization days, frequency of visits to doctors, diagnostic health checks, and consumption of drugs
o we will model their long term changes and
1.4 Structure
Chapter 2 first analyzes the estate of social insurance worldwide and afterwards specializes on the way it operates in Greece and more specifically on IKA. The basic dispensations of IKA are analyzed. The basic factors that determine the cost of the benefits are introduced: the inflation rates and the demographic factors of the country.
In chapter 3, an empirical analysis of the data is carried out. At first, some statistical values are calculated and afterwards a reference is made to regression analysis, the method that will be used in our study. Finally, the regression and the autoregression methods are applied to our data.
In chapter 4, the econometric model derived in the third chapter, is applied to possible future alternative models with different prices of the inflation.
2. Theoretical background
2.1 Definition of social insurance
Social insurance is the insurance that aims to deal with the risks that workers run during their work and which may lead to a decrease of the workers earning capacity, may create additional needs which cannot be satisfied by his regular income or may lead to an increase of his cost of living. (Petros Kiohos, Social insurance in Greece, 1993, p. 19)
Social insurance covers the entire working population and also people that haven’t entered the production process yet.
2.1.1 Social insurance systems
The classic Social insurance systems are based on 2 types:
The one type relies on the fact that the workers pay the insurance organization of social welfare by their payment of the insurance contribution and that they count on these contributions in order to satisfy any future needs. They have then some similarities to the private insurance companies. In this case the state is not involved at all.
The other type of insurance coverage is provided by insurance entities that have a degree of independence and usually have the form of juristic persons of public law under the state supervision. The insurance coverage, in such a case is provided by the incomes of the national budget and by the contributions of the workers and the employers. This system ensures dispensations and satisfies economic needs to the entire working population. (Petros Kiohos, Social insurance in Greece, 1993, p.18-19)
In Greece, second system is in valid. The money for the social insurance funds is collected in the following way:
works gives to IKA 800 euros per month for his health coverage and also for his future pension. In Greece the state does not give any pensions to the citizens.
b) By the government which gives approximately 1% of the Gross National Product to the Social Insurance funds
The applied system is the distributive financing model (pay as you go). That means that the payment of the dispensations to the payees is provided by the incomes that arise from the workers at the same time. The common use of the distributive system relies on the raise of funds from the various programs and their redistribution, whenever it’s necessary.
The insurants by social insurance foundations are basically classified to 2 categories: 1. Direct insurants: Persons that have the right of insurance coverage, due to the work
that they do.
2.1.2 Historic evolution of social insurance
The estate of social insurance is very old; in the city of Athens, relief for the poor had already been established since the 6th century (BC). The first organizations of reciprocal
help of the free workers, especially of the artisans, appeared in ancient Rome.
The first completed social insurance system appears in Germany, where by an act of 1883 the compulsory insurance of the employees against accidents was established. Life as well as disability was added into the insurance scheme after six years.
New Zealand was the first country that has been applying since 1939 a system of social insurance with pecuniary dispensations that cover the cases of age, disability, unemployment and death.
Generally, from 1885 to 1935 various systems of social insurance were developed in the whole of Europe, which had clearly been influenced by the German system. The Second World War contributed to the development in all of the countries of a larger desire of security against all the social risks which can possibly hurt a person. The reason is that people had more health problems and need to retire early because of the war. They needed extra help and care.
2.2 The social insurance in Greece
In Greece, the first stature about the constitution of an entity of social insurance was the decree of 15/12/1836 according to which the Natal Retirement Fund (NAT) was constituted and which started to run in the year 1861.
Furthermore, by the law 6234 of 1934 TEBE was constituted, the insurance fund for practitioners and tradesmen.
In Greece, the most important step regarding the social insurance was the bill about the Social Insurance, as by the law 6298/10/10/1934 the Social Insurance Institute (IKA) was constituted, which started to run as an insurance organization in 1937. The law allowed the social insurance of the employees who didn’t belong to any other Fund until this date. From 1951 the fund started to give benefits to the insurants. Today, IKA covers about the 50% of the country’s habitants.
A watershed was the promulgation of the law 4190 of 1961 about the “Agricultural Insurance Organization” which was a new entity that included the country’s entire agricultural population. (P. Kiohos, Social insurance in Greece, p. 24)
2.2.1 Social Insurance types
a. Mandatory insurance
b. Optional insurance
Optional insurance relies merely on the interested person’s will. At the optional insurance, the insurant must pay the entire amount of contribution to the social insurance entity.
c. Auxiliary insurance
Auxiliary is called the type of insurance provided by an entity or a branch that has been especially constituted in order to cover, in some insurant groups, fringe, periodic or superannuation benefits, apart from those that are given by the main insurance entity, to which they belong. In that case, the insurant pays additional contributions, according to the professional category where he belongs. (P. Kiohos, Social insurance in Greece, 1993, p.
28) IKA provides all three types of social insurance.
2.3 Analysis of IKA benefits
The payments of IKA are discriminated to:
1) Money dispensations, which are the pensions and the various benefits (sickness, accident, maternity, funeral benefit etc)
2) Health benefits, which include medical, pharmaceutical, hospital, dental care etc.
2.3.1 Dispensations (Direct costs)
1. Pensions
The dispensation of a pension is provided at the below circumstances: a) Because of ageing
b) In case of disability c) In case of death
2. Illness or accident dispensation
The insurant of IKA have the right to receive a sickness benefit, if they fulfil the requirements below:
1) They are not able work, due to sickness
2) The sickness or the accident wasn’t caused by the insurant’s fault 3) They are not in retirement and do not serve in the army
4) They have worked for at least 100 days, during the last calendar year.
The sickness or accident benefit is provided to the payee from the 4th day of his absence. The case of industrial accident makes an exception, where it is provided since his 1st day of absence. The benefit can be provided until one year of absence, totally. The amount of the compensation is calculated according to the wage scale that corresponds to the insurant’s job.
3. Childbirth dispensation
IKA pays a one-time sum for the natural or premature birth or for a still-born of a gestation period of 6 months to the directly insurants, to those that retire on a pension and to the wives of the directly insurants and the pensioners, since they are their family members and as a consequence deserve sickness benefits by the IKA.
The benefit is given in case that the birth has taken place into a hospital with an admission of IKA.
4. Funeral dispensation
2.3.2 Benefits (Indirect costs)
a. Medical care
The medical care provided by the IKA includes the necessary medical care that the insurant and their family members need. All of them (directly and indirectly insured) have the following rights:
• To visit the doctors at the private practice of IKA. • To visit the family doctor.
• To visit the rural doctor, in the rural regions.
• To call the doctor of IKA home, in case of emergency. • To visit the Health Centres.
• To have general and specific checks. • To call the first aids in case of emergency.
• To take part to programs of Preventive Drugs (vaccines, check-ups etc.)
b. Pharmaceutical use
The drug prescriptions written by the doctors of IKA aim to the insurants’ convalescence and their full recovery. The insurants pay for the drugs a contribution, with some exemptions to it.
c. Hospital care
maladjusted children, but only to the ones that have an agreement with IKA. In case of emergency or even in a case of necessary hospitalization, the insurant has the right of admission to any hospital or nursing home even to those that do not have an agreement with IKA. The expenses will be covered according to the state fixed policy for prices. In case that the disease cannot be diagnosed or healed in Greece, IKA takes over the entire expenses of the insurant’s hospitalization abroad. It also covers a percentage of the occupancy and food expenses of the patient and his companion.
d. Additional care
For the treatment of special health problems, IKA provides to its insurants, pensioners and to their family members additional care as: pacemakers, plastic vassal grafts, inhalation devices, speaking devices, artificial limbs, eyeglasses etc. The insurant’s contribution is usually the 25% of the purchase price.
REST SCHEMES:
2.4 Inflation
Inflation is the percentage change of the general level of the prices in an economy, during a specific period of time. The inflation may be positive or negative, for example in Japan during the last decade.
Let’s note that the inflation is the increase of the general price level. It doesn’t exist when prices are stable, independently if they are high or not. In an economy, when we calculate its inflation rate, in fact we study the percentage change of the level of prices, not for the entire goods or services used, but for certain goods or services.
According to OECD, price stability is one of the primary objectives of the European Central Bank (ECB), with the inflation rate used as a prime indicator of the management of monetary policy within the euro area. The ECB has defined price stability as an annual increase in the harmonised index of consumer prices (HICP) for the euro area of close to but below 2% (over the medium term).
The rises on Health Services and also on drugs don’t follow the same rate to the Consumer Price Index (CPI). In the United States the health inflation rate in 2005 was 6.9%, double compared to the general inflation rate. In Greece, the health inflation in 2006 was 6.2% while the general inflation rate was 3.2%. Generally, in most of the European countries, the health inflation rate is about 2 degrees higher than the general inflation rate.
This creates a problem for the social insurance, at the moment that the inflation rate is the factor that determines the raises on salaries. The incomes of the social insurance are a percentage deduced by the salaries. Therefore, the incomes of the insurance funds increase according to the country’s inflation rate, while their costs, the expenses on the health benefits, increase much faster.
2.5 Demographic trends
Certain demographic factors have a great influence on the function of the social insurance. In details, the most important are the following:
1) The ratio of the part of the population older than 65 years to the whole population 2) The expected lifetime or the age of death
3) The rate of men and women in the population
The above factors play a special role on the increase of the cost of benefits, not only of the pensions but also of the health benefits, as the needs for medical care are totally connected to the age. (Nektarios Miltiadis, Social insurance in Greece. p. 27)
More analytically:
According to the National Statistic Service in Greece in 1957, people aged over 65 years old were 627.408, a percentage of just 8% while in 2006 they were 2.060.584, a percentage of 18,5%. Prediction is still more ominous for the future. It is estimated that in 2050 people in Greece aged over 65 years old are going to be 3.390.998, which is the 31,45% of the total population of the country.
The expected lifetime has increased in all of the European Union countries, including Greece. In Greece in 1929 a man of 65 years old was expected to live for 11.94 more years. A woman aged 65 years old was expected to live for 13.89 years. In 2005, that expected lifetime has reached 16,4 years for a man and 18.33 years for a woman. Predictions for 2050 are 19.6 for men and 22.3 for women.
true in Greece. In 1957, the number of women was 4.146.445, which is 51% of the population, while in 2006 the percentage decreased to 50.5% of the population. The prediction of the National Statistical Service for 2050 is that there will be a 50 – 50 proportion of men and women.
The population’s ageing, apart from increasing the cost of health benefits, also results in a decrease of the incomes of the insurance funds. The contribution of a retired person is
proportionally quite lower with respect to an employed person.
Supposing that the population remains stable in number, when the proportion of people older than 65 years old increases, then the proportion of people ageing from 15 to 64 years old, who are able to work, decreases. Evidentially, we observe that in our country, the percentage of the population ageing from 15 to 64 years old, in 2000 was 68% of the total population, in 2006 it decreased at 67.15% and in 2050 is estimated to be 56.41%.
Since the incomes of the insurance system result from the contributions on the salary of the working population, a result from the above is a relative reduction of the income of public funds.
Additional to that is:
a) The low employment rate b) The high unemployment rate and c) The reduction of births.
The unemployment rate is estimated by dividing the number of people ageing 15-64 years old that don’t work but ask for a job by the total number of people ageing 15-64 years old. This rate was particularly high in Greece in the past. In 2000 it was 11.3%; in 2005 it was 9.3%. Great attempts are being made in order to decrease it at 7%, which is the future prediction.
3. Methodology and research process
3.1 Data analysis
3.1.1 Data description
There is a fifty-year information database, from 1957 to 2006. The information comes from the Statistical and Actuarial Department of IKA and from the National Statistical Service.
IKA data
IKA keeps much information that concerns the costs of the insurants. We refer to the information that we use in our study with the respective information from the variables that are inserted in SPSS and in S-PLUS.
We focus on the health costs per insurant, because all these years the total number of the insurants has not remained stable, thus, if we study the total costs of the institute, the figures will not be comparable. It should be mentioned that the number of the insurants in 1957 was approximately 1,300,000, whereas that number in 2006 was about 5,500,000. The National Statistical Service provides us with the following information:
1. Age of Greek people categorized in age groups and in gender.
2. Inflation information. 2005 is considered as the base year. The inflation is centralized and in percentage.
Analytically:
YEAR: the report year
AGE: the average age of the insurants
INFLATION: the consumer price index (CPI)
Indirect cost per capita
INDIRECT: the cost for benefits per insurant
DRUGS: the corresponding per insurant annual medication care cost HOSPITAL: the corresponding per insurant annual hospital care cost
Direct cost per capita
DIRECT: the cost for money dispensations per insurant
ILLNESS: the proportional per insurant annual cost because of income loss due to illness ACCIDENT: the corresponding per insurant annual cost because of income loss due to accident
BIRTH: the proportional per insurant annual cost because of income loss due to birth FUNERAL: the money paid to the relatives of the decedent for funeral expenses OTHER: other benefits given to certain categories of the insurants
MANAGE: administrative costs per year and per insurant
Use of health benefits
VISITS: average number of visits to the doctor per insurant per year
CHECKS: average number of diagnostic health checks per insurant per year DRUGSCOUNT: average number of drugs per insurant per year
DAYS: average number of days per year spent in the hospital per insurant
Acceptances and modifications
Because IKA has not kept records of the age of the insurants and the protected members, we consider the average age of Greeks as the average age of the insurants in the respective year.
whatsoever. For these variables, we will study the annual percentage change of their values. We will act in the same way for the several individual costs such as drugs costs and pregnancy dispensations.
As far as the ages are concerned, since the figures which we will study are the average age of the insurants, and additionally they are about time series, we will examine the age changes during the years, but this time as absolute numbers and not as percentage changes. Regarding the variables that have to do with frequencies, we will examine their regular values.
3.1.2 Average age
With regard to ages, we observe a constant increase of the average during all the fifty years, except 2 years – in 1971 when there was 0.14 years decrease and in 1986 when there was 0.2 years decrease. The biggest average increase appeared in 1970 and it was 1.13 years. Since 1997, the annual increase is no more than 0.2 years per year.
Overall, the average age of Greeks from 1957 to 2006 has increased by 9.92 years. In 1957 the average age was 30.95 years and in 2006 it was 40.87 (Appendix).
Possible causes
Higher level studies hold many young people out of work and this might be one of the factors that imply the significant increase of the average age of the Greek insurants. Usage of technology is dynamically propagated and special qualifications are obtained with long sustained education rather than work experience...
The increase of life average of the population is another reason of this situation. As a consequence of the better living circumstances, the death age has increased considerably. Thus, bigger percentages of elderly people in the population increase the average age.
3.1.3 Inflation
We observe very significant differences between the inflation of prices and generally the inflation of health care costs of IKA and especially in direct and indirect costs.
General inflation rate is 9.43 +8.09, the inflation rate of indirect health care costs is 13.62+10.08, the inflation rate of direct health care costs is 15.83+14.98 and inflation rate of total health care cost is 14.03+9.26.
In 1962 we observe the lowest general inflation rate (0.37%), but generally, during the years 1957-1972 the general inflation varies in low levels (under 5%). This happens again after 1998. In the meantime, the values vary significantly and peak in 1974, when the general inflation rate was 26.85%. In the years 1980 and 1981, the values are also high (24.88 and 24.46 respectively).
We observe that though the lowest values of the 3 inflation rates that are compared, present a remarkable time gap among them, their higher values appear successively (1974-1975-1976). Perhaps this means that the increase in the General Index of Prices involves the next year increase of direct health care costs and later the increase of indirect costs
During most of the years, the general inflation rate is much lower than the other two. There is an exception regarding five specific years for the inflation of direct funds (1973, 1986, 1988, 1993 and 1994) and for 11 years for the inflation of indirect funds (1960, 1969, 1976, 1979, 1981-1984, 1986-1987 and 1998). Perhaps this happens because the indirect costs are more easily controlled by the institute, in contrast with the direct costs which are related to external circumstances (for example wages or policy).
In addition, in all distributions we observe positive skewness (Pearson’s 1.2, 1.72, 0.91 for direct and indirect cost and general inflation respectively)
The inflation of the total cost is less variable than the other two. Its lowest value -3.98 appears in the year 1968 and the highest 40.54 in 1975, exactly in the same years with the inflation of indirect cost. In average is 14.05% and shows positive (Pearson’s 0.67).
Possible causes
The biggest leaps in prices are observed in three periods: 1974-1976, 1980-1981 and 1990-1993. During these years in Greece, there were always changes in the government. In these periods the dictatorship has fallen (1974), the socialistic government ruled the country (1980) or the Democratic Party won the elections (1990). The changes of the inflation rate appear in the year before the elections and are caused by reasons that the government wanted to give in order to get re-elected, or in the year after the elections that constitutes the implementation of pre-election promises (Appendix)
3.1.4 Direct costs
In the two primary categories of costs for benefits, the biggest increases of values appeared simultaneously in 1976, with illness (99.52%) and birth (61.14%), i.e. two years after the highest value of the general inflation (26.85%) in 1974.
In addition, in 1977 and 1978 we had big increases as well. The lowest values appeared for illness in 1986 (-7.56) and for birth in 1961 (1.14). The first is slightly unreasonable. The second is reasonable since; in any case, the inflation was much lower in the years 1960-1962.
Figure 3.1.5 Timeplot of analytical direct costs
Possible causes
During the years 1974-1978, the State came out of a period of dictatorship, thus, generally, the financial condition of the country changed. This justifies the high values during these years.
3.1.5 Indirect costs
Analytically, the inflation of drugs proceeds with mean being 16.33%, the inflation of hospitals follows with mean being 14.02% and finally the inflation of doctors with 12.58%. In addition the inflation of doctors displays the most significant stability with st. deviation being 9.78 against 16.7 that the other two figures display. Furthermore, the median is very close to the mean, with value 12.75%.
The lowest value of the inflation of doctors is observed in the year 1968 (-7.07%) as well as in the next years 1969 (-4.10%) and 1969 (-0.73). Let’s remember that 1968 was the year that the general inflation had its lowest value (0.39).
The highest value of the inflation of doctors appears in 1978.
Drugs and hospitals have totally different time periods max and min. Drugs demonstrate very big increase in 1975 (103.97%) to fall in the next year, i.e. in 1976 by 16.65 %. Of course now, we have to remember that a year earlier, in 1974, the general inflation had the highest value (26.85%).
Hospital costs per capita increased by 102.04% in 1993, whereas their lowest value was displayed in 2006 with negative inflation (-7.37%).
Figure 3.1.4 timeplot of analytical indirect costs
Possible causes
As we observe in every type of costs, extreme values appeared during the quinquennium 1974-1978 as well as in 1993. In these years, as it was mentioned before, we had intense political changes in our country.
3.1.6 Health benefits use
The annual consumption of drugs per person demonstrates a steadily upward course. The lowest value occurs in 1960 with drugs 4.09 and the highest in 2000 with drugs consumption 9.14. The mean is 7.16 drugs with standard deviation being 1.48. 50% of the registrations are more than 7.47 drugs per year.
On the contrary, hospitalization days demonstrate a decreasing course during the years 1957-2004. The highest value was 1.82 days per person in 1965 and the lowest 0.92 in 2001. The mean is 1.4 days per person with small st. deviation 0.328. Generally there are small variations and big concentration between the values 1.35 (mean) and 1.40 (median). (Appendix)
Figure 3.1.6 timeplot of annual visits and checks Figure 3.1.7 timeplot of drugs and hospital days
Visits to doctors demonstrate an unexpected performance. While they steadily increased from 1957 to 1975, when the highest frequency in visits per person occurred (6.66), then they gradually decrease. In 2001 they decreased to 4.02. The mean is 5.38, which is very close to the median 5.66.
Possible causes
The consumption of drugs gives direct solutions to health problems. Since the short-term illnesses, such as virus infections, headaches and psychological illnesses, increase because of the contemporary lifestyle, people consume more drugs than they did in the past.
The need for hospitalization decreased drastically as a result of using rapid methods of surgery (Laser). In this case, the patient does not need hospitalization. Naturally, the expenses of the Institute for the operations are still big, so the fewer hospitalization days do not necessarily lead to lower costs.
Visits to doctors decrease steadily. The reason might be the preference of the insurants to specific doctors who do not work at the Institute or the time deficiency of the insurants to go to the doctors of the Institute due to the vast number of obligations and the incapability of programming their visits. Presumably this is because there is a plethora of doctors who are not appointed by the Institute and are in private practice which are preferred by the insurants.
On the other hand, the diagnostical health checks have almost been sextupled. This probably shows a tendency of our society for precautionary control. In addition, because of the progress of science and drugs, new laboratory examinations are constantly added for the benefit of the patients in order to have more accurate diagnoses.
3.2 Theory
3.2.1 Regression analysis
Definition
Multiple regression analysis is a statistical technique that can be used to analyze the relationship between a single dependent variable and several independent variables. The objective of multiple regression analysis is to use the independent variables whose values are known to predict the single dependent value selected by the researcher. Each independent variable is weighted by the regression analysis procedure to ensure prediction from the set of independent variables. (Hair-Black-Babin-Anderson-Tatham, Multivariate
Data Analysis p.169-268)
Weights denote the relative contribution of the independent variables to the overall prediction
The general form of a regression model for k independent variables to a dependent variable Y is
Assumptions of Multiple regression
Assumption 1: Existence: For each specific combination of values of the independent variables , Y is a univariate random variable with a certain probability distribution having finite mean and variance.
Assumption 2: Independence: The Y observations are statistically independent of one another.
Assumption 3: Error is normally distributed with zero mean and fixed variance (homoscedasticity) and independent to any variable
Assumption 4: The independent variables are statistically uncorrelated. (Kleinbaum,Kupper, Nizam, Applied Regression Analysis & Multivariate Methods)
Least Square Approach
In general, the least squares method chooses as the best-fitting model the one that minimizes the sum of squares of the distances between the observed responses and those predicted by the fitted model. That is, the better fit, the smaller the deviations of observed from predicted values. Thus, if we let
,
Where the matrix of the independent variables (I is a n-sized vector of ones), and T is for transpose.
This minimal sum of squares is generally called the residual sum of squares.
Anova table
The anova table provides an overall summary of a multiple regression analysis. The particular form of ANOVA table may vary depending on how the contributions of the independent variables are to be considered. A simple form reflects the contribution that all independent variables considered collectively make to prediction. The following three terms are estimated:
is the total sum of squares and represents the total variability in the Y observations before accounting for the joint effect of using the independent variables
is the residuals sum of squares, which represents the amount of y variation left unexplained after the independent variables have been used in the regression equation to predict y.
is the regression sum of squares and measures the reduction in variation (or the variation explained) due to the independent variables in the regression equation. We thus have the familiar partition
or = +
Test of fit of predictions
The R2 provides a quantitative measure of how well the fitted model predicts the dependent
variable. The formula for R2 is
R2 =
The quantity R2 lies between 0 and 1. If the value is 1, we say that the fit of the model is perfect.
Test of overall fit
To test how the entire set of independent variables contributes significantly to the prediction of Y, we will perform an F-test.
The null hypothesis for this test may be generally stated as Ho : ‘All independent variables
considered together do not explain a significant amount of the variation in Y’. Equivalently, we may state the null hypothesis as Ho :β1=β2=…=βκ=0. Under this last
version of Ho, the full model is reduced to a model that contains only the intercept term β0.
To perform the test, we use the mean-square quantities provided by an ANOVA table. (S-plus or SPSS output). We calculate the F-statistic
F=
The computed value of F can then be compared with the critical point , where α is the preselected significance level. We would reject Ho if the computed F exceeded the
3.2.2 Dickey-Fuller’s t-test for stationarity.
Regression of a non-stationary variable on one other variable that is also non-stationary produces an apparently significant relationship even there is not any physical relationship between them. This is a common effect of time dependent variables and the way to define if it’s stationary or not we apply a simple AR(1) process:
where is a time (t) dependent variable and are Gaussian random errors of zero mean on constant variance. is stationary if and non-stationary if . For practical reasons the testing regression is given by . Using a t-test known as Dickey-Fuller’s t-test, if (null hypothesis) then is non-stationary and if
(alternatively) then is stationary.
3.2.3 ARIMA
ARIMA (autoregressive integrated moving average) models are extensively used for time series prediction. This is done by taking into account only the past values of the time process of interest, i.e. any internal structural form or influential status among the process and the independent variables it is being implicated are not in use.
In principle, the time process has to be stationary and if not, we apply a D-F t-test for stationarity on the first differences . If is not stationary too, then we
on until is stationary. The ARIMA( ) model is then as it follows
Where and are the autoregressive and moving average coefficients respectively.
AKAIKE criterion
Akaike’s Information criterion (AIC) is the most commonly used to select a model in the best way. It is given approximately by
AIC= -2 ln(max.likelihood) + 2 p
where p denotes the number of independent parameters estimated in the model.
Thus the AIC essentially chooses the model with the best fit as measured by the likelihood function subject to a penalty term that increases with the number of parameters fitted in the model. This should prevent over fitting. Ignoring arbitrary constants, the first (likelihood) term is usually approximated by Nln(S/N), where S denotes the residual (fit) sum of squares, and N denotes the number of observations.
The best model is the one that minimizes the criterion AKAIKE. (Chris Chatfield.
3.2.4 Model’s validity
A linear model is valid if and only if all of the hypotheses concerning errors are satisfied, i.e. they are normally distributed, they are independent each other and they have fixed variance. Of course errors are unknown, and testing of each hypothesis separately is not feasible. Instead, we test only the hypothesis that the studentized residuals are normally distributed using a non-parametric one sample Kolmogorov-Smirnov test (K-S test). Studentized residuals are defined as where are the
residuals, and is the ith – diagonal element of array .
3.3 Health costs modelling
3.3.1 Dickey-Fuller’s t-test for stationarity
Given that a linear regression method among non stationary variables returns totally unreliable estimates, we performed Dickey Fuller tests on all series. Their relative changes were found to be stationary (see 3.2.2); γ- coefficients for the relative change of variables of interest are showed in table 3.3.1.
Table 3.3.1
Variable γ-Coefficient Std. Error t p
Relative change of Illness -0.332 0.107 -3.107 .003
Relative change of Birth -0.146 0.078 -1.668 .048
Relative change of Doctors -0.231 0.084 -2.737 .009
Relative change of Hospitals -0.431 0.120 -3.595 .001
Relative change of Drugs -0.568 0.133 -4.410 <.001
Relative change of Days -1.161 0.148 -7.861 <.001
Relative change of Drugs count
-0.625 0.137 -4.544 <.001
Relative change of Visits -0.891 1.148 -6.021 <.001
Relative change of Checks -0.6 0.148 -4.544 <.001
Relative change of Direct -0.262 0.099 -2.644 .011
Relative change of Indirect -0.242 0.095 -2.54 .014
3.3.2 Direct costs
3.3.2.1 Total direct costs
Model 1 was finally found valid (although limited above level of 5%) according to a K-S test (Z=1.342, p=.054) and the normal Q-Q plot that follows (Figure 3.3.2.1).
Model 1
Coefficients
Dependent Relative change of Inflation of Direct Benefits
Independent Coefficient Std. Error t p
An extension to Model 1 was that in which the mean age of insurants was included as a second dependent variable. While we found a negative influence, i.e. a 1.13% of the RCDB was leaded by a 1% increase of the mean age of insurants, this was not found significant (p=0.26) and the overall implication increased only slightly (adjusted R2 =0.146). At last, coefficient of RCGI was found about steady (0.766, p=.004) compared to that of Model 1.
3.3.2.2 Analytical Direct Costs
Illness-Birth
Following the same steps as for direct benefits, relative changes of illness (RCI) and birth (RCB) were both found stationary with respect to a D-F test (Table 3.3.1), and linear related to RCGI. In details, in both Models 3 and 4, a 1% increases of the RCGI explains about a 0.75% significant increase of RCI (0.733%, p=.032) and RCB (0.778%, p<.001). R2 values were found 0.0995 and 0.251 for Models 2 and 3 respectively, showing that the variability of RCI is quite less implied due to variability of RCGI rather than the variability of RCB to the variability of RCGI.
Finally, both Models were found valid (Model 3, limited above level of 5%) according to a K-S test (for Model 2, Z=1.31, p=.053; for Model 3, Z=1.149, p=.143) and normal Q-Q plots of studentized residuals (Figure 3.3.2.2.a, Model 2; Figure 3.3.2.2.b, Model 3).
Mean age of insurants, as in Model 1 does not improve significantly the implication of either RCI or RCB.
Model 2
Coefficients
Dependent Relative change of Illness
Independent Coefficient Std. Error T p
Constant 9.246 4.144 2.231 .03
Relative change of General inflation
Model Summary R2 Adjusted R2 0.095 0.075 ANOVA Variability Sources Sum of Squares d.f.. Mean Square F p Regression 1689.378 1 1689.378 4.912 0.032 Residual 16163.772 47 343.910 Total 17853.15 48 Model 3 Coefficients
Dependent Relative change of Birth
Independent Coefficient Std. Error T p
Figure 3.3.2.2.a: Model 2
3.3.3 Indirect Costs
3.3.3.1 Total Indirect Cost
Relative change of total indirect cost (RCTI) increases significantly with RCGI. In details, a 1% increase of RCGI implies a 0.662% increase (Model 4, p<.001) of RCTI and the assigned R2 value of that implication is 0.272. Model’s implication was found valid according to a K-S test (Z=.849, p=.467) and a normal Q-Q plot (Figure 3.3.3.1)
Model 4
Coefficients
Dependent Relative change of Total Indirect costs
Independent Coefficient Std. Error T p
Figure 3.3.3.1: Model 4.
3.3.3.2 Analytical Indirect Costs
a. Doctors
Model 5
Coefficients
Dependent Relative change of Doctors
Independent Coefficient Std. Error T p
Constant 3.651 1.958 1.865 .069 Relative change of General inflation Relative change of Checks .748 .404 .135 .172 5.554 2.357 <.001 .023 Model Summary R2 Adjusted R2 0.423 0.396 ANOVA Variability Sources Sum of Squares d.f.. Mean Square F p Regression 1702.57 2 851.29 16.105 <.001 Residual 2325.76 44 52.86 Total 4028.33 46
b. Hospitals
Relative change of hospitals (RCH) was also implied by a multivariable model (Model 6) where RCGI and relative change of days (RCDA) were the independent variables. ANOVA table shows that Model 6 is totally significant (p=.001) and value of R2 shows that the variability of RCH is by 26.7% due to changes of RCGI and RCDA. Both variables were found significant and effectuate about the same on the RCH. Finally, A K-S test showed that the Model 6 is valid (Z=1.048, p=.222) and the normal Q-Q plot of studentized residuals confirms that (Figure 3.3.3.2.b).
Model 6
Coefficients
Dependent Relative change of Hospital
Independent Coefficient Std. Error t p
Figure 3.3.3.2.b: Model 6
c. Drugs
Relative change of drugs (RCDG) was implied by a multivariable model (Model 7) where RCGI and relative change of drugs count (RGDC) were the independent variables. Model is totally significant (ANOVA table, p=.004) and value of R2 shows that the variability of
RCDG is by 22.0% due to changes of RCGI and RCDG. Both variables were found significant but the RCDC effectuates more than twice (1.463) the RCGI (0.663) to the dependent variable RGDG. Finally, A K-S test showed that the Model 7 is valid (Z=1.008,
Model 7
Coefficients
Dependent Relative change of Drugs
Independent Coefficient Std. Error t P
3.4 ARIMA Models
3.4.1 Days of hospitalization
The best ARIMA model for Days is that with 11 AR-lags, since we could not find any significant lag-coefficient for k<11. For k=11, coefficient assigned to lag 5 is significant (p=.044) which implies a repetitive relationship every 5 years. We preferred not to increase k more than 11 because political history in our country has marked unpredictable changes (i.e. governments of different political parties are elected and different health-care systems are applied) every decade. Prediction curve (figure 3.4.1) shows a temporal increase in days but thereafter a systematic decrease follows.
Days, (k=11, d=1, q=0)
Coefficients
Y Days
Independent Coefficient Std. Error t P
Constant 1.006 2.982 .338 .738 Zt-1 -.275 .172 -1.594 .120 Zt-2 -.117 .181 -.647 .522 Zt-3 .143 .183 .780 .441 Zt-4 .223 .186 1.198 .239 Zt-5 .404 .193 2.088 .044 Zt-6 .356 .197 1.813 .079 Zt-7 .029 .193 .152 .880 Zt-8 -5.075E-5 .190 <.001 >.999 Zt-9 -.177 .189 -.936 .356 Zt-10 -.123 .189 -.651 .520 Zt-11 -.108 .182 -.594 .557
AIC index for selection ARIMA model
Figure 3.4.1 Prediction of days
3.4.2 Drugs count
The same repetitive relationship of 5 years (p=0.043) is implied for drugs count although a model of k = 8 (AIC = -0.71) was, in principle, the one that enhanced this repetition. Prediction curve (figure 3.4.2) shows a fairly constant increase after 2005. This is surprisingly not expected since biotechnology produces more effective drug products and therefore we might expect one drug to substitute more. Presumably, an increased use of drugs might be due to a corresponding increase of modern syndromes – as depression it is - related to work pressure, human relationships, financial insecurity etc.
Drugs count, (k=8, d=1, q=0)
Coefficients
Y Drugs count
Independent Coefficient Std. Error t p
Zt-7 -0.018 0.178 -0.099 0.921
Zt-8 0.196 0.174 1.130 0.266
AIC index for selection ARIMA model
K 1 2 3 4 5 6 7 8 9 10 AIC -0.9 -0.86 -0.81 -0.81 -0.80 -0.78 -0.71 -0.71 -0.67 -0.62 AR -lag Significant at .05 - - - - - - - 5 5 - .
Figure 3.4.2 Prediction of drugscount
3.4.3 Visits
Visits, (k=5, d=1, q=0)
Coefficients
Y Visits
Independent Coefficient Std. Error t p
Constant 5.285 3.652 1.447 0.156 Zt-1 0.038 0.142 0.268 0.790 Zt-2 0.051 0.141 0.361 0.720 Zt-3 0.188 0.139 1.351 0.184 Zt-4 -0.161 0.141 -1.141 0.261 Zt-5 -0.466 0.143 -3.251 0.002
AIC index for selection ARIMA model
K 1 2 3 4 5 6 7 8 9 10 AIC -1.24 -1.19 -1.16 -1.13 -1.16 -1.13 -1.109 -1.05 -1.02 -0.97 AR -lag Significant at .05 - - - - 5 5 5 5 5 5
3.4.4 Checks
A short –term repetition of 2 years is implied for checks (k=2, AIC = -1.65, p = 0.006). A linear increase is showed from the prediction curve (figure 3.4.4) which might be explained due to better diagnostic methods and precautionary control.
Checks, (k=2, d=1, q=0)
Coefficients
Y Checks
Independent Coefficient Std. Error t p
Constant -15.247 5.603 -2.721 0.009
Zt-1 0.008 0.167 0.046 0.963
Zt-2 0.483 0.169 2.861 0.006
AIC index for selection ARIMA model
K 1 2 3 4 5 6 7 8 9 10 AIC -1.61 -1.65 -1.61 -1.56 -1.52 -1.50 -1.45 -1.42 -1.38 -1.36 AR -lag Significant at .05 - 2 2 2 2 2 2 2 2 2
4 Forecasts
We applied three hypothetical prices of the Consumer Price Index (CPI) to our model. The predictions concern the next 20 years 2007-2026.
First, we examined the figures that came up considering the CPI to be 3% (which is the goal of our government). Next, we assumed that the CPI is 4% for the next 20 years (approximately as it is this year at the country). In the third and most utopian model, we regarded the CPI to be 10% for all the next years to see what will happen if Greece will not be able to handle inflation, something usual in the past.
The predictions for the hospitalization days per insurant, the quantity of drugs that will be consumed within a year, the visits to the doctor and the laboratory examinations per capita came up using the ARIMA method, as it was analyzed in the previous chapters.
4.1 Direct costs
C.P.I. (%) 3 4 10
ANNUAL
CHANGE (%) 11.51 12.21 16.41
DIFFERENCE 8.51 8.21 6.41
Regarding the direct health care costs (dispensations), the increase rate of the costs for dispensations to insured will continue to be high and specifically for 4% annual inflation rate, i.e. the inflation rate will be 12.21%. We can observe that the bigger the CPI gets, the smaller the difference between CPI and Direct cost inflation becomes.
The analytical direct costs take the following prices
4.1.a Absence from work due to illness
CPI (%) 3 4 10
ANNUAL
CHANGE (%) 11.45 12.18 16.58
Damages for work absence due to illness show equal changes with the total direct costs. They are much higher than the CPI. Further survey should be done to see if this is due to more days of absence or to higher dispensations. This should become because of bigger income of the workers. Since this is a compensation that is given only to people who still have a job and not to retired, it is connected with real salaries in the market.
4.1.b Absence from work due to birth
CPI (%) 3 4 10
ANNUAL
CHANGE (%) 12.96 13.73 18.40
DIFFERENCE 9.96 9.73 8.40
The money given to young women due to absence from their job because of birth shows highest increase percentage than any other figure under study. Again it has to be investigated if this is due to a higher number of births per year or to higher dispensations due to higher income of these women.
4.2 Indirect costs
C.P.I. (%) 3 4 10
ANNUAL
CHANGE (%) 9.51 10.17 14.14
DIFFERENCE 6.51 6.17 4.14
Regarding the indirect health care costs (the benefits for insured), we observe that is it much higher than the C.P.I. Comparing with direct cost it is about 2 % less. This is because IKA can control some of the indirect costs. For example the salaries of the doctors change according to CPI. Again, the difference between CPI and Indirect costs inflation is getting smaller in high prices of CPI (e.g. 10%).
4.2.a Doctors
Year visits visits cha% checks checks cha % CPI 3% Dif. CPI 4% Dif. CPI 10% Dif.
2006 4.07 5.71 2007 3.9 -4.18 6.02 5.43 8.09 5.09 8.84 4.84 13.32 3.32 2008 3.66 -6.15 6.44 6.98 8.71 5.71 9.46 5.46 13.95 3.95 2009 3.51 -4.10 6.76 4.97 7.90 4.90 8.65 4.65 13.14 3.14 2010 3.38 -3.70 7.14 5.62 8.17 5.17 8.91 4.91 13.40 3.40 2011 3.19 -5.62 7.48 4.76 7.82 4.82 8.57 4.57 13.05 3.05 2012 3.11 -2.51 7.85 4.95 7.89 4.89 8.64 4.64 13.13 3.13 2013 3.05 -1.93 8.2 4.46 7.70 4.70 8.44 4.44 12.93 2.93 2014 2.94 -3.61 8.56 4.39 7.67 4.67 8.42 4.42 12.90 2.90 2015 2.85 -3.06 8.93 4.32 7.64 4.64 8.39 4.39 12.88 2.88 2016 2.76 -3.16 9.31 4.26 7.61 4.61 8.36 4.36 12.85 2.85 2017 2.6 -5.80 9.69 4.08 7.54 4.54 8.29 4.29 12.78 2.78 2018 2.44 -6.15 10.08 4.02 7.52 4.52 8.27 4.27 12.76 2.76 2019 2.3 -5.74 10.47 3.87 7.46 4.46 8.21 4.21 12.69 2.69 2020 2.13 -7.39 10.88 3.92 7.48 4.48 8.23 4.23 12.71 2.71 2021 1.97 -7.51 11.29 3.77 7.42 4.42 8.17 4.17 12.65 2.65 2022 1.83 -7.11 11.71 3.72 7.40 4.40 8.15 4.15 12.63 2.63 2023 1.69 -7.65 12.13 3.59 7.34 4.34 8.09 4.09 12.58 2.58 2024 1.54 -8.88 12.57 3.63 7.36 4.36 8.11 4.11 12.60 2.60 2025 1.41 -8.44 13.01 3.50 7.31 4.31 8.06 4.06 12.55 2.55 2026 1.26 -10.64 13.46 3.46 7.29 4.29 8.04 4.04 12.53 2.53
Money spent for medical care and for laboratory examinations is also expected to show an almost stable annual increase which is almost 4% more than CPI for small prices. It is slightly minimised as the time goes on.
4.2.b Hospital
year days days cha% CPI 3% Dif. CPI 4% Dif. CPI 10% Dif
2006 1.04 2007 1.07 2.88 11.91 8.91 12.92 8.92 18.99 8.99 2008 1.10 2.80 11.82 8.82 12.83 8.83 18.90 8.90 2009 1.11 0.91 9.60 6.60 10.61 6.61 16.68 6.68 2010 1.11 0.00 8.54 5.54 9.55 5.55 15.62 5.62 2011 1.10 -0.90 7.48 4.48 8.49 4.49 14.56 4.56 2012 1.10 0.00 8.54 5.54 9.55 5.55 15.62 5.62 2013 1.09 -0.91 7.47 4.47 8.49 4.49 14.55 4.55 2014 1.09 0.00 8.54 5.54 9.55 5.55 15.62 5.62 2015 1.08 -0.92 7.46 4.46 8.48 4.48 14.54 4.54 2016 1.06 -1.85 6.37 3.37 7.38 3.38 13.45 3.45 2017 1.03 -2.83 5.23 2.23 6.24 2.24 12.30 2.30 2018 1.01 -1.94 6.27 3.27 7.28 3.28 13.34 3.34 2019 1.00 -0.99 7.38 4.38 8.39 4.39 14.46 4.46 2020 0.98 -2.00 6.20 3.20 7.21 3.21 13.28 3.28 2021 0.95 -3.06 4.96 1.96 5.97 1.97 12.03 2.03 2022 0.92 -3.16 4.84 1.84 5.85 1.85 11.92 1.92 2023 0.90 -2.17 5.99 1.99 7.01 3.01 13.07 3.07 2024 0.87 -3.33 4.64 1.64 5.65 1.65 11.72 1.72 2025 0.85 -2.30 5.85 2.85 6.86 2.86 12.93 2.93 2026 0.82 -3.53 4.41 1.41 5.42 1.42 11.49 1.49
4.2.c Drugs
Year count count cha% CPI 3% Dif. CPI 4% Dif. CPI 10% Dif.
2006 9.80 2007 9.77 -0.31 9.50 6.50 10.16 6.16 14.14 4.14 2008 9.89 1.23 11.75 8.75 12.41 8.41 16.39 6.39 2009 9.84 -0.51 9.21 6.21 9.87 5.87 13.85 3.85 2010 9.73 -1.12 8.31 5.31 8.98 4.98 12.95 2.95 2011 9.75 0.21 10.25 7.25 10.91 6.91 14.89 4.89 2012 9.78 0.31 10.40 7.40 11.06 7.06 15.04 5.04 2013 9.84 0.61 10.85 7.85 11.51 7.51 15.49 5.49 2014 10.02 1.83 12.63 9.63 13.29 9.29 17.27 7.27 2015 10.13 1.10 11.56 8.56 12.22 8.22 16.20 6.20 2016 10.22 0.89 11.25 8.25 11.91 7.91 15.89 5.89 2017 10.3 0.78 11.09 8.09 11.76 7.76 15.74 5.74 2018 10.31 0.10 10.09 7.09 10.75 6.75 14.73 4.73 2019 10.3 -0.10 9.81 6.81 10.47 6.47 14.45 4.45 2020 10.32 0.19 10.23 7.23 10.90 6.9 14.87 4.87 2021 10.32 0.00 9.95 6.95 10.61 6.61 14.59 4.59 2022 10.36 0.39 10.52 7.52 11.18 7.18 15.16 5.16 2023 10.43 0.68 10.94 7.94 11.60 7.6 15.58 5.58 2024 10.49 0.58 10.79 7.79 11.45 7.45 15.43 5.43 2025 10.55 0.57 10.79 7.79 11.45 7.45 15.43 5.43 2026 10.61 0.57 10.78 7.78 11.44 7.44 15.42 5.42
5. Conclusions
By examining the sequence of years 1957-2006, it came up that the inflation of IKA health cost benefits per insurant demonstrated 3.9% average annual increase more than the annual Consumer Price Index of Greece. On the other hand, the IKA health benefits dispensations per insurant was on average 6.1 % higher than the CPI in the country. It is predicted from this study that the health care costs for the insurants will continue to increase in the next 20 years (2007-2026). Particularly, the increase rate will be much bigger than the rate of CPI of Greece. The bigger the rate of CPI of Greece is, the smaller the difference between the rate of CPI and the inflation of health care costs will be. The mean age of the insured people will not influence the health costs.
The increase rate of IKA dispensations cost will be always bigger (at about 2% more) than the increase rate of IKA benefits cost. It seems that IKA can more easily control benefits than dispensations. The dispensations cost because of work absence due to birth will lead the cost for dispensations. Regarding the benefits cost, the drugs expenses will demonstrate the greatest annual increase.
6. Appendix
Table 3.1.1 Average age
YEAR Average Age of insurants
Yearly change of Average Age
YEAR Average Age of insurants Yearly change of Average Age 1957 30,94 - 1982 35,842 0,16 1958 31,14 0,20 1983 36,012 0,17 1959 31,33 0,19 1984 36,206 0,19 1960 31,531 0,20 1985 36,439 0,23 1961 31,722 0,19 1986 36,415 -0,02 1962 31,922 0,20 1987 36,642 0,23 1963 32,112 0,19 1988 36,883 0,24 1964 32,288 0,18 1989 37,135 0,25 1965 32,471 0,18 1990 37,385 0,25 1966 32,738 0,27 1991 37,575 0,19 1967 32,782 0,04 1992 37,750 0,17 1968 32,892 0,11 1993 37,921 0,17 1969 33,044 0,15 1994 38,087 0,17 1970 34,178 1,13 1995 38,255 0,17 1971 34,043 -0,14 1996 38,440 0,18 1972 34,216 0,17 1997 38,644 0,20 1973 34,430 0,21 1998 38,853 0,21 1974 34,657 0,23 1999 39,065 0.21 1975 34,818 0,16 2000 39,299 0,23 1976 34,951 0,13 2001 39,588 0,29 1977 35,082 0,13 2002 39,910 0,32 1978 35,206 0,12 2003 40,191 0,28 1979 35,333 0,13 2004 40,437 0,25 1980 35,501 0,17 2005 40,716 0,28 1981 35,683 0,18 2006 40,870 0,15
Table 3.1.3 Inflation rates and percentage annual changes
Table 3.1.4A Analytical direct costs and percentage annual changes Year Direct cost Direct cost Relative change (%)
Illness Illness Relative Change (%)
2005 21,00 0,31 72,12 0,10 26,54 0,05
2006* 16,10 -0,23 74,00 0,03 28,47 0,07
2007* 16,92 0,05 81,40 0,10 30,19 0,06
Table 3.1.5 analytical indirect costs and percentage annual changes
1998 245.09 -0.74 52.38 12.70 73.52 -12.70 74.91 6.23 1999 290.19 18.40 57.41 9.59 84.00 14.25 98.69 31.74 2000 310.99 7.17 58.82 2.46 106.46 26.74 101.38 2.73 2001 343.60 10.48 62.02 5.43 125.27 17.67 113.01 11.46 2002 383.50 11.61 70.11 13.05 147.42 17.68 123.08 8.92 2003 448.10 16.84 74.63 6.44 180.53 22.46 155.62 26.44 2004 530.49 18.39 82.45 10.48 219.38 21.52 177.06 13.78 2005 579.52 9.24 81.88 -0.69 255.18 16.32 184.80 4.37 2006 610.89 5.41 106.96 30.64 269.21 5.50 171.18 -7.37
Table 3.1.6 count of use of health benefits
