Diversity & Equality in Health and Care Open Access

Health Insurances: Gender Rating and Regional Effect on Insurance Price

¹Department of Arts and Sciences, Middle East Technical University, Turkey
^*Correspondence: Bereketoglu Abdullah Burkan, Department of Arts and Sciences, Middle East Technical University, Turkey, Email:

Received: 01-Aug-2022, Manuscript No. IPDEHC-22-14081; Editor assigned: 03-Aug-2022, Pre QC No. IPDEHC-22-14081 (PQ); Reviewed: 17-Aug-2022, QC No. IPDEHC-22-14081; Revised: 22-Aug-2022, Manuscript No. IPDEHC-22-14081 (R); Published: 29-Aug-2022, DOI: 10.21767/2049-5478.19.7.33

Introduction

After the industrial revolution, with the increase of big cities and fast population growth, civilizations passed on a new gen- eration of humane issues to be upheld some of these particular issues are held by private entrepreneurs with the rise of the United States. These entrepreneurs started insurance compa- nies for certain humane matters that the government should provide by following specific government regulations and procedures that protect the consumer from the abuse of the private company. Later on, in the EU/EEA, new, more humane insurance policies and systems models were initiated. Insur- ance, specifically health insurance, after the mid-20^th century, initialized with its new model for newly emerging economic ar- eas. Countries such as Switzerland had a 100 year programmed system that continuously updated to the new age of the 21^st century, and it is suggested in the Orlu et al. to be adopted by Turkey due to this research to be concluded, which is gender based insurance policy. Furthermore, another interesting pric- ing procedure is held in the US for people living in different states and even in different counties of a specific state, whether it is urban or rural. Moreover, whether there is more than one insurance company that is giving service to the area is also effective in the current holding system in the US [1,2].

American Center for Disease Control and Prevention Center (CDC) suggests that particular disability and risk factors can be used like the following price policy and coverage procedure for insurances, and they can be named; Alcohol usage, illicit drug usage, body measurements (height, size, etc.,) mandatory diet, certain disabilities, physical or mental function issues, exer- cise or physical activity capacity, obesity/BMI, smoking. Even to date, many private and social health insurance companies still use other parameters, such as gender or age, to calculate the insurance price per year. Even today, they include different premiums for different age groups that are specially tailored to the consumer group of the specific age groups. On the con- trary, in recent years, gender rating based pricing is started to diminish by banning done by Obama legislation and European Commission, but still, many countries use both gender and age based and other factors to calculate the insurance price. For the US, before the ACA, which stands for the Affordable Care Act, Montana was the first to prohibit the usage of gender as an insurance price factor. Intrinsically ACA, initiated and immediately put into effect in early 2010, was expected to give equal pricing and premiums for all but did not give equal pricing for all; it was missing age, income, and, most importantly, region.

Affordable Care Act started a chain reaction in the health in- surance pricing with the new regulations it came with, such that giving more potential to see the effects of the other not forbidden pricing indexes, one of which is region based pricing. Besides that, ACA did not prevent private insurance companies from switching from age to some other feature to measure their pricing for regions because it did not forbid the usage of age. With that, companies in the US started to give more im- portance to the region and age, even though, in recent years, there were some movements against age and income based health insurance pricing. Furthermore, companies after the ACA even started to give numbers as ratings to the states and their counties for pricing and give different pricing for rural and urban areas in specific states according to their new policies.

In this project, the total and direct gender effect also total and direct region effect on the price will be excerpted to understand the importance and whether it is needed to be a significant factor in health insurance pricing policies. For the analysis, a data set called Medical Cost Personal Dataset Health Insurance Cost Prediction (Insurance) data by Amy Aguirre is used with ~1300 samples. Moreover, in the analysis, the dataset used, is the Medical Cost Personal Dataset Health Insurance Cost.

Prediction (Insurance) data, may not be data that is taken from the United States due to its regular descriptions of the features in the metadata given by Amy Aguirre are not in English but rather Spanish also, since the Affordable Care Act, gender is no more used in the pricing or held in the data storage for the pricing reasons, so it is assumed that the data is taken from a country that has not started to use regulation such as in the US case [3-8].

Review

In the literature, it is seen that age and gender play a signifi- cant role in the determination of the prices of health insurance. Still, many contrary regulations started to be integrated into the governmental systems against the latter, gender, playing a role in the health insurance prices and premiums. A private institution with the closing gap in premiums between the gov- ernment issued health insurance and the gender gap is now over in the EU faces a drastic decline in people who buy and continue private insurance coverage. Also, with the increasing costs, many exits the entire healthcare insurance system [9,10].

This price gap mentioned in the system is according to one of the insurance companies run in India named, Gender Rating, and still in use in India [11]. Turkey also still has this gender gap, but women after a certain age risk breast cancer and other re- productive system problems [1]. By Merzel, it is also discussed that in low income zones such as Central Harlem, NYC, people with low income most of the time cannot afford private cover- age, but females have the opportunity to have it covered when they are in the workforce, with that being said, females do not get affected by low income or socioeconomic factors. On the other hand, male counterparts are heavily influenced by few- er available opportunities even in the workforce due to com- panies not offering them insurance making them less covered within the year 2000 [12]. Literature states that certain states in the United States of America and after Obama care now have better coverage for people from different socioeconomic back- grounds [5]. It is stated in the 2008 Los Angeles Times article that being female increases the insurance price rates drastically [13]. It is still for every one in 10 women who affect women between 19 to 64, consisting of nearly 98 million people in the United States [14]. In Taiwan, elderly or mid aged women have different cancer insurance policies with lower claim rates for dread diseases which show that the system is biased towards gender/age [15]. In the IFFCO-TOKIO factors article, it is stated that 10 factors affect the health insurance premiums in most insurance policies. These are the recent day’s age, medical his- tory, occupation, policy duration, BMI, smoking, and location; as stated in Orlu, prices are determined by gender. Still, the premiums are not as given in the IFFCO-TOKIO article [1,16]. Lastly, Huang and Salm mentioned that the ban on gender based pricing increased the number of people who buy health insurance; even though the new prices are slightly lower, but not drastically.

Another important factor to add that is not discussed priorly in the review is the region effect that is causing the current price disparity between whom can access how affordable health insurance and how much coverage from the premiums they can get at the end result in the United States [2]. According to Wengle, in the US, living in urban or rural areas highly differ- entiates the prices and the premiums people can access in the healthcare industry for insurance. For lower competition envi- ronments, insurance companies tend to use their monopolistic power to increase costs. Hence the price is higher for the areas which have fewer health insurance companies [7]. The Afford- able Care Act, or in abbreviation ACA, was initiated in the first quarter of 2010 by Barack Obama to eliminate unequal access to healthcare, gender, smoking, and various other factors used in the pricing of healthcare insurance [6]. However, the region was not put in the ACA, which led companies to use the region as one of their main factors in determining new prices in differ- ent regions [2].

Methodology

In the project, it is planned to use the following approach to attack the problem of whether gender also the region individually have a tremendous effect on the prices of health insurance and also to see whether they are individually crucial indicators that should be used in the determination of price by making a directed acyclic graph of dependencies and pulling out the gen- der as the independent factor to affect the cost of insurance.

In the Medical Cost Personal Dataset Health Insurance Cost Prediction (Insurance) data [8], there exist precisely 1338 data with a slightly imbalanced gender ratio of consumers given with a balance of 51:49 (Male: Female) [8]. Furthermore, ac- cording to the CIA World Factbook, the world has a 1.01:1.00 male to female ratio, which gives the fact that the imbalance is rather insignificant when compared with the total population, so the dataset has no bias toward gender in the light of the total population information is known [17]. The features of the data set can be provided as age, sex (gender), BMI, children count of the person, whether the person smokes or not, their region, and the price for the total of 6 features given [8].

As mentioned above, the dataset that will be used has six fea- tures and one outcome: Health insurance cost. Health insur- ance prices are affected by several measures, two of which are the region and gender of the individual who is going to pur- chase the insurance. These two features that affect the price have an inferential domain in their system, as seen in Figure 1. In some regions of a country, one can assume by their prior knowledge that in some regions there exist many elderly and in some others, many young people are living for various rea- sons, such as university location. Also, regions can also affect the population if small regions are picked. Technical (STEM-fo- cused) universities and the regions near the university will have gender differences due to the lack of women in STEM fields worldwide [18]. Dataset also shows that age can also indicate gender, and it is because of natural death age imbalance of genders for people who live in equal ecosystems [19]. There- fore age also has an effect on gender, and that leads to differ- ent pricing for insurance [19]. Even though the life expectancy difference is significant, it is not true to directly jump into a conclusion it is for every case for men and women living in [20]. Life expectancy measures are given for people who live simi- lar livelihoods and end up dead at the end through a process rather than measuring for the richest women/men compared to the poorest women/men [20]. One can also infer the knowl- edge of whether someone is smoking or they are in older ages of their life from their children count. Furthermore, one can also infer that if someone is obese, which means higher BMI, they might quit smoking recently, or vice versa is true for the lower BMI measures. Since smoking prevalence is higher by 5 times in adult males, we can also indicate that sex (gender) can also be inferred from the smoking factor of the person. Fur- thermore, in the end Figure 1 then gives us the end scientific model in such a way that we can find the total and direct ef- fects of the region and gender by the abovementioned infor- mation and more.

Figure 1: Directed Acyclic Graph of Medical Cost Personal Data- set-Health Insurance Cost Prediction (Insurance) data for the model to be used in the project.

Bayesian analysis of the data for the total and direct effects of region and sex will be conducted with the regular PyMC3 Markov Chain Monte Carlo algorithm that is used by the sam- pler. Furthermore, there will also be basic statistical analysis to see whether these values are correlated and give measures corresponding to the results. The PyMC3 package of Python gives us four chains of MCMC samples for our desired statis- tical model. The model will be created by taking the Figure 1 directed acyclic graph as the reference to build the model for direct and total effects of the desired features. For the prior statistical analysis, it is seen that before actually working on the dataset now, some correlations between various features should be stratified to find the direct effect of gender and re- gion. Also, find the correlation between male female differenc- es and smoking, smoking, and BMI. It is also will be noted that children can also be one of the factors to be taken into account in the correlation of BMI for gender regardless of male or fe- male for the beginning; the correlation limit in this experiment will be determined by the famous economic principle called Pareto Principle which indicates that 80% of the cases will be driven by 20% of the indicators, so if the correlation is below 20% correlation will be neglected and reported in the analysis. Later, samples that are created by the MCMC sampling algo- rithm will be analyzed with diagnostic tools to measure if the sample is reliable. These measures are R-hat, trace-plots, sev- eral effective samples, and various other tests, such as WAIC and PSIS, will be used. In the end, we compare our 4 models to see which one gives better results for inferential analysis and prediction on a different basis.

The Bayesian multiple linear regression and hierarchical analy- sis methods will be used in the model. The hierarchical meth- od can be used to analyze the different regions in the system for model creation. For more intricate systems for some steps other than multiple linear regressions, splines can be used after the behavior of the data in regular statistical analyses is determined. If the charge is assumed to behave logarithmical- ly, exponentially, or in a polynomial shape, a different system rather than linear regression without separating data into dif- ferent sections would be more desirable. This behavior will be reported in the study if any such behavior exists in the analyses in the pre-process. If the behavior is linear for the features for the charge, then the multiple linear regression method is to be followed. Else non-linear approaches such as splines are to be followed.

The charge predictions for gender and regions are to be visu- alized with summary statistics, and the highest density interval for 91% interval is given. 91 is picked because it is below 10% from tails with a slight difference. For the last step, it is to be concluded with the difference between males and females also with the regional differences to be simulated, and the differen- tiation statistics as the main result to be reported after specif- ic interventions to make the gender and region independent from each other charge independently from the rest.

Results

Gender Based Effect Model

Here stratification of the smoker parameter and age parame- ters in Figure 1 to see the effect of the gender parameter on the health insurance prices is done. Results show that female or male does not significantly differ in pricing; however, prices change drastically for whether the individual is a smoker or not (Table 1) (Figure 2).

Table 1: Part A-Gender effect on Pricing

Figure 2: From left to right: Density for Male~Female, Rank Bars for Genders (top-female, bottom-male), Rank Bars for Age on Gender model, Rank Bars for Smokers on Gender model.

Region Based Effect Model

Here for the region effect model, we stratified BMI and age parameters to understand the region’s effect directly and analyze whether it has a significant effect on the change in health insurance prices. This model’s results show that BMI does not significantly affect changing prices; however, it adds variance to the model. Moreover, with the increasing age, it started to be seen that prices started to have an increasing trend when compared with the younger ages, which corresponds to the first indexes of the beta_A parameter (Table 2) (Figure 3). In Table 2, Figure 3 one can see the results mentioned above in the region effect model.

Table 2: Part B-Region effect on Pricing

Figure 3: Density for four different regions, Rank Bars for Regions, Rank Bars for BMI on Region model, Rank Bars for Age on Region model.

Region and Gender based Gaussian Mixture Ef- fect Model

Here, a different method is used to see the effect of gender and region fully, combining both of them in a multivariate Cholesky Covariance model. This model suggests that region two has significantly higher health insurance prices when compared to the other three. Also, the first region is comparably lower than the other two in pricing. This may be due to various reasons that will be talked about in the conclusion section (Table 3) (Figure 4).

Table 3: Part C-Region and Gender mixed effect on Pricing

Figure 4: Rank Bars for Regions ~ Gender mixed effect.

Discussion

Cholesky Covariance and Correlation

Here it is worked with the model structure of the mixed effect model. Furthermore, it is seen that from the Cholesky param- eter that by Table 4 we can state that the same index Cholesky covariances are effective and gives great parametrization result (Figures 5 and 6).

Table 4: Part C-Region and Gender mixed Cholesky Coefficient

Figure 5: Mixed Effect Model.

Figure 6: Cholesky Correlation.

Conclusion

Here in the proposed research, to conclude the results and lit- erature based prior beliefs. It should be stated that gender has, according to the models given above, a lower potential to have a direct effect on the pricing, positively or negatively, compared to smoking behavior, older ages, region, and comparably equal to the effect of overweight pricing based on the projections of the models built and analyzed. Even though literature suggests that gender plays a significant role, the literature researchers never gave any background information about whether gender, in reality, plays a significant role rather than playing a role due to systemic bias in certain areas. It is also evident that some Western countries have started to implement new regulations and insurance policies that disregard the effect of gender and prohibit its usage. The data used may get affected by these new policies, therefore, can prove the fact that gender does not play a statistically significant role in health insurance pricing poli- cies; however, thought as it is relevant and, in the past, used as a pricing index, which was as seen from the results a biased policy.

Further Work

As to study, the further studies can be concluded with higher computation power model bigger matrices with multivariate normal or more complex gaussian mixture models to analyze all the effects in a more generalized behavior, and also dive into the data used, since the USA and some other European countries started to implement new health insurance policies that disregard gender as a pricing parameter, it can be further proved that gender is indeed not necessary to be put into con- sideration to give higher prices for one’s self.

Acknowledgement

None.

Declarations

Conflict of interest on behalf of all authors, the corresponding author states that there is no conflict of interest.

REFERENCES

1. Orlu C, Köse A (2018) Türkiye Ve Isviçre Genel Sağlik Sigortasi Sistemlerinin Kiyaslanmasi Ve Türkiye Için Yeni Bir Model Önerisi. Finansal Araştırmalar ve Çalışmalar Dergisi 10(19):303-329.

   [Crossref] [Google Scholar]

2. Health Care Is Local: Impact of income and geography on NASHP.

   [Crossref] [Google Scholar]

3. CDC (2016) Disability and risk factors. Centers for disease control and prevention.

   [Crossref] [Google Scholar]

4. Yamamoto DH (2013) Health care costs from birth to death. Age-Curve-Study.

   [Crossref] [Google Scholar]

5. Fontinelle A (2022) Gender and insurance costs. Investopedia.

   [Crossref] [Google Scholar]

6. DCD (2021) What is the affordable care act? HHS.gov.

   [Google Scholar]

7. Wengle E (2019) Are Marketplace premiums higher in rural than in urban areas?

   [Google Scholar]

8. Central Intelligence Agency. (n.d.). The World Factbook.
9. Future Generali How gender impacts health insurance premium.

   [Crossref] [Google Scholar]

10. Gender differences in health care access indicators in an urban, low-income community (2000) Am J Public Health 90(6): 909-916.

    [Crossref] [Google Scholar]

11. Lazarus D (2008) Los Angeles Times: Gender can cost you in individual health.

    [Crossref] [Google Scholar]

12. KFF (2021) Women's health insurance coverage.

    [Crossref] [Google Scholar]

13. Li CS, Hung CJ, Peng SC, Ho YL (2021) Impact of gender and age on claim rates of Dread Disease and cancer insurance policies in Taiwan. Int J of Environ Res Public Health 19(1): 216.

    [Crossref] [Google Scholar] [Pubmed]

14. Factors that affect your health insurance premium costs. Motor, health, travel, and home insurance online.
15. Huang S, Salm M (2019) The effect of a ban on gender‐based pricing on risk selection in the German health insurance market. Health Economics 29(1): 3-17.

    [Crossref] [Google Scholar]

16. Women and girls in Science Technology, Engineering, and Mathematics.
17. Robert H, Shmerling MD (2020) Why men often die earlier than women.

    [Crossref] [Google Scholar]

18. Crimmins EM, Shim H, Zhang YS, Kim JK (2019) Differences between men and women in mortality and the health dimensions of the morbidity process. Clin Chem 65(1): 135-145.

    [Crossref] [Google Scholar]

19. Men smoke 5 times more than women.
20. Chernew M, Cutler DM, Keenan PS (2005) Increasing health insurance costs and the decline in insurance coverage. Health Serv Res 40(4): 1021-1039.

    [Crossref] [Google Scholar] [PubMed]