Inequality in Health: The Correlation between Poverty and Injury—A Comprehensive Analysis Based on Income Level in Taiwan: A Cross-Sectional Study

Is income still an obstacle that influences health in Taiwan, the National Health Insurance system was instituted in 1995? After collecting injured inpatient data from the health insurance information of nearly the whole population, we categorized the cases as either low-income or nonlow-income and tried to determine the correlation between poverty and injury. Chi-square tests, Fisher’s exact tests, an independent-samples t-test, and percentages were used to identify differences in demographics, causes for hospitalization, and other hospital care variables. Between 1998 and 2015, there were 74,337 inpatients with low-income injuries, which represented 1.6% of all inpatients with injury events. The hospitalization mortality rate for the low-income group was 1.9 times higher than that of the nonlow-income group. Furthermore, the average length of hospital stay (9.9 days), average medical expenses (1681 USD), and mortality rate (3.6%) values for the low-income inpatients were higher than those of the nonlow-income group (7.6 days, 1573 USD, and 2.1%, respectively). Among the injury causes, the percentages of “fall,” “suicide,” and “homicide” incidences were higher for the low-income group than for the nonlow-income group. These findings support our hypothesis that there is a correlation between poverty and injury level, which results in health inequality. Achieving healthcare equality may require collaboration between the government and private and nonprofit organizations to increase the awareness of this phenomenon.

3 increase awareness of this phenomenon.

Background
In the past several decades, many countries have experienced different levels of economic growth under free commerce, leading to a growing income gap between the wealthy and the poor. Poverty heavily influences individuals' ability to obtain health care, which can lead to increased health problems in disadvantaged social groups [1,2,3].
In 1995, Taiwan instituted National Health Insurance (NHI), which requires mandatory participation for all citizens from birth and provides a basic level of medical care for all citizens. NHI greatly increased medical care accessibility, reduced the financial obstacles to obtaining medical services, prevented financial hardship caused by the cost of medical care, and encouraged those without money to seek medical care [4].
Previous research revealed that while low-income individuals comprise 0.7% of the total population, their inpatient care claims represent 4.8% of national inpatient costs.
Additionally, although only 4.3% of the low-income population receives social welfare assistance, their NHI expenses account for 14.4% of the total NHI costs [5]. Other research showed that hospitalization rates for low-income individuals (14.1%) was 8.5% higher than that of non-low-income individuals, indicating that the low-income population utilized more medical resources than the national average [6].
However, previous research has focused on either low-income families without a control group or on overall medical care expenses rather than medical care costs associated solely with accident injuries. Therefore, the purpose of our research was to compare characteristics of "low-income" and "non-low-income" accident injury inpatients and evaluate correlations between income level and health inequality.
Materials And Methods 4 Data source Implemented in 1995, NHI currently covers 99% of all Taiwan citizens. The Health and Welfare Data Science Center, Ministry of Health and Welfare (HWDC, MOHW) collects all emergency room and hospitalization data. Further, the law requires medical facilities to submit claims for emergency room and hospitalization expenses on a monthly basis.
Therefore, the HWDC is the most authoritative data source for medical and healthcare related research [7]. We used the original inpatient and outpatient medical claims data collected between 1998 and 2015.

Variable definitions
Variables include: Low-Income (Yes, No), Gender (Male, Female), Age (1-4, 5-14, 15-24, 25-44, 45-64, and ≥ 65 years), Charlson Comorbidity Index (CCI), Intentionality of Injury (ICD-9-CM E-Code, E800-E949 unintentional, E950-E979 intentional, and E980-E989 unspecific and unable to determine), Cause of Injury (ICD-9-CM E-Code, E800-E848 transport injuries, E850-E869 poisoning, E870-E879 medical malpractice, E880-E888 falls, E890-E899 burns, E900-E909 natural and environmental factors, E910 drowning, E911-E915 suffocation, E916-E920 crushing, cutting and piercing, E921-E949 others unintentional, E950-E959 suicide, E960-E979 homicide, E980-E989 undetermined), Surgical Operation (Yes, No), Level of Care (medical center, regional hospital, local hospital), Hospitalization Area (northern, central, southern, eastern, outer islands), Length of Hospital Stay (day), Medical Expense (USD), and Prognosis (survival, mortality). The low-income qualification was stipulated by Article 4 of the Public Assistance Act [8] of Taiwan with the following conditions: 1) individuals must submit an application and be approved by their local municipality authority, 2) the average monthly income per person in the household must fall below the poverty line, and 3) the total household assets must not exceed the specific amount set by the central and municipality authorities in the year 5 the application is submitted. The poverty line is based on the standard published by the central department of budget, accounting and statistics, and is defined by the central and municipality authorities as 60% of the median personal expenditure in the household's local area in the past year.
The CCI [9] selects the first five diagnostic codes (ICD-9-CM N-Code), weighs them according to scoring criteria defined by Charlson, and calculates the total score. Higher scores indicate more complications or a more severe diagnosis. Additionally, the "prognosis" for the deceased includes deaths in the hospital and voluntary discharge for the terminally-ill.

Statistical analysis
IBM SPSS Statistics 20.0 was used to conduct all statistical analyses. Statistical significance was set at p < 0.05. Univariate statistics and multivariate logistic regression were used to compare mortality rates during hospitalization between the low-income and the non-low-income groups, with survival as the dependent variable (survived, deceased), and demographics, hospitalization cause, and other hospitalization care measures as independent variables. We then compared the mortality rate during hospitalization between the low-income and the non-low-income groups.

Results
Data from 4,647,058 accident injury inpatients between 1998 and 2015 were collected.
Patient characteristics are summarized in Table 1. In all, 74,337 inpatients (1.6%) were low-income and 4,572,721 (98.4%) were non-low-income. The male-to-female ratio for the low-income group (1.74) was significantly higher than that of the non-low-income group (1.41). The highest hospitalization rates occurred in patients aged 65 years and older both of the low-income group and non-low-income group. Hospitalization rates were also 6 significantly different between the two groups for the 5-14 and 25-44 age groups. Lowincome inpatients scored higher on the CCI measure than non-low-income inpatients (0.6 and 0.5, respectively), indicating that the number or severity of injury complications was much higher for the low-income inpatients.   More low-income inpatients sought treatment in eastern Taiwan, tended to stay longer in the hospital and incurred higher medical expenses than non-low-income inpatients. Lowincome inpatients also had a higher mortality rate than that of non-low-income inpatients for both unintentional and intentional injuries (3.1% vs 1.7% and 4.4% vs 3.1%, respectively).  Table 4 shows the distribution of treatment outcomes for non-fatal and fatal injuries by group. Low-income inpatients were less likely to receive surgery for non-fatal injuries (41.6% and 51.4%, respectively). Low-income inpatients were also less likely to receive medical care in major hospital centers (21.3% vs 29.1%) and were more likely to receive medical attention in regional hospitals (44.8% vs 41.6%). Moreover, low-income inpatients tended to stay longer in the hospital and incur higher medical expenses (9.9 days vs 7.6 days and USD $1,681.5 vs USD $1,573.9, respectively). The comparative results for fatal injuries between the low-income group and the non-low-income group were consistent with those in non-fatal injuries. 4 10 Table 4 Fatal and non-fatal injury hospitalization related variables by income level between 1998-2015 in

Discussion
Health inequality Between 1998 and 2015, an average of 237,877 citizens were classified as low-income [10], and their accident injury hospitalization rate was 1.74 per 100. In comparison, an average of 22,611,119 were classified as non-low-income [11], and their accident injury hospitalization rate was 1.12 in 100. That the accident injury hospitalization rate for the 13 low-income group was twice as high as that of the non-low-income group implies poverty and injury are correlated. In addition, low-income inpatients had more complicated injuries than non-low-income inpatients (2.3 vs 1.9) and the hospitalization mortality rate for lowincome inpatients was 1.888 times higher than that of non-low-income inpatients, showing that health inequality exists between the low-income and non-low-income groups. The high hospitalization mortality rate for low-income inpatients is primarily driven by intentional injuries as it was 2.014 times higher for non-low-income inpatients.
Many researchers have attributed the more complicated and severe injuries in low-income individuals to their lower socioeconomic status, which may compel them to accept highrisk entry-level jobs that require heavy labor. Consequently, they are more susceptible to injuries, which may develop into severe and chronic conditions if left untreated. This is especially true for those who must work to support their families regardless of illness or injuries. These individuals have a higher risk of harm due to poor physical condition [12,13].
Serious injuries require more comprehensive care at an advanced medical facility. Our results showed that a significantly lower proportion of the low-income inpatients received treatment in medical centers compared to non-low-income inpatients, providing additional evidence of health inequality. Figure 1

illustrates differences between low-income inpatients and non-low-income inpatients in hospitalization payment processes under
Taiwan's NHI program, where the cost of a doctor visit includes medical expenses plus a registration fee-a processing or administration fee set by the medical institution that typically corresponds to the level of care. Therefore, the registration fees at medical centers are higher compared to those at regional and local hospitals. Also, while Taiwan's NHI program covers most medical expenses, inpatients are still required to pay a small portion of the medical expenses as a co-payment. To reduce health inequity, low-income inpatients are exempt from the co-payment. However, the NHI program does not cover the registration fees. Therefore, unless the local government or hospital social welfare measures (SWM) provide relief for low-income inpatients, they have to pay the registration fee. In addition, postoperative patients may also be required to pay living costs and caregiver expenses during hospitalization, and/or pay for other medical equipment or prescription drug expenses that are not covered by NHI. In general, non-low-income inpatients have private health insurance to help pay those costs; however, low-income inpatients can only rely on minimum support from the government's SWM since they cannot afford private health insurance. Low-income inpatients' hospitalization costs also tend to be higher than those of non-low-income inpatients because they are hospitalized longer and are more seriously injured (Table S1). Thus, low-income inpatients are more reluctant to receive surgery despite having higher CCI scores. Moreover, the out-of-pocket medical expenses are generally proportional to the level of care received; the higher the level of care, the greater the expense. In Taiwan, the highest level of medical certification accreditation is the medical center. Consequently, low-income inpatients generally lean toward regional or local hospitals because of lower expenses. However, when taking the severity of the injury into consideration, low-income inpatients are more willing to receive care in medical centers for fatal injuries as opposed to non-fatal injuries. We also observed a significant gender gap in the low-income group. The Ministry of Health and Welfare of Taiwan has reported that among low-income individuals, the number of single-person households is higher for men than for women [14]. Typically, men in Taiwan are the main source of income because of the patriarchal nature of Chinese families.
Therefore, low income men are less likely to marry and more likely to accept jobs with a poor working environment, making them more vulnerable to injuries. In addition to the gender gap, the low-income and non-low-income groups differed significantly in the percentage of individuals in the 5 to 14 age group (9.3% vs 5.1%). Our findings were consistent with previous research reporting that adolescents from lower socioeconomic families were more likely to have serious injuries requiring hospitalization [15].

Cause of injury
With unintentional, non-fatal injuries, our results showed that hospitalization rates related to "medical malpractice" and "falls" were higher for the low-income group than the nonlow-income group.
A multinational retrospective study on the global burden of disease (GBD) showed that, as a result of medical adverse events, disability-adjusted years (DALYs) equal 23 million globally every year, for which two-thirds come from low-income and mid-income countries [16]. Other research on these same countries also showed more medical accidents, lower patient safety, and lower medical care quality for the low-income group [17].
We further observed that falling injuries are more prevalent among low-income inpatients in the middle and older age groups (more than 45 years old), as shown in Table S2. Past research on fall risk in older adults found that low income is a contributing factor [18].
Other factors include socioeconomic status (low education, solitary living, and lack of care), living environment (inconvenient floorplan and insufficient light), and physical condition (poor vision, chronic illness, and aging) [18,19,20]. Accordingly, the lowincome group may be more susceptible to falling and hospitalization because of poor living conditions, more living hazards, and lack of safety protection equipment [19].
Intentional injuries showed a significantly higher rate of hospitalization for "suicide" and "homicide" for low-income inpatients in both fatal and non-fatal injuries. Additionally, previous research has shown that unemployment and specific occupations were also associated with suicide and suicidal behavior [21,22,23,24]. According to the lowincome family study conducted by the Ministry of the Interior of Taiwan, 62% of lowincome family members have suffered catastrophic illnesses in the past [12]. In addition, 47% of the breadwinners in these families are out of the workforce.
Previous research also found that poverty increases the risk of mental illness and suicide [25,26]. Low-income individuals generally view themselves as a financial minority, and tend to feel powerless, helpless, and repressed when facing competition, factors that may be associated with contemplating suicide. Therefore, low-income individuals have a much higher risk of repeating suicide attempts that result in hospitalization [27]. Some studies also indicate that low-income and severe illness exacerbate the risk of suicide during hospitalization. Patients with both severe illness and low-income may suffer multiple complications that require long-term care. Without sufficient resources, these patients may become depressed and consider suicide as an escape and a relief for their families [28].
Furthermore, one comprehensive analysis that collected criminal data from 169 countries found a positive correlation between income inequality and homicide/injury. This is especially prevalent in low-income and mid-income countries [29]. One could use Durkheim's Anomie theory [30]  Therefore, it is no surprise that our study found motorcyclist injuries to have the highest hospital admission rates for both the low-income and non-low-income groups. As shown in Table S3, motorcyclist injuries were more frequent in the low-income group than the nonlow-income group. However, injuries suffered by the driver of a motor vehicle showed a more significant difference in hospital admission rates between the two groups compared to motorcyclist injuries. In general, low-income families are less likely to own a private passenger car because of costs or access barriers; therefore, they are less vulnerable to overall transport injuries.
In summary, falling and transport injuries are the most common causes of injury in the low-income group. Typically, inpatients with injuries or older adults who are at an increased risk of falling often need to use mobility aids such as wheelchairs, canes, and walkers. However, these mobility aids are not covered by Taiwan's NHI, as per Article 51 of the National Health Insurance Act, and low-income inpatients are effectively denied access. The health inequality between low-income inpatients and non-low-income inpatients still exists despite the implementation of Taiwan's NHI. Government agencies should take actions and eliminate the health inequity for low-income patients.

Limitations
The data from HWDC did not provide any information on immigration status (natives, new immigrants), marital status, education level, or occupation. The low-income group reported in this research was sorted by those who qualified for insurance under the Public Assistance Act and not by actual income. The health inequality in the most vulnerable group is likely to be higher than those in a low-middle income group (income quintiles or deciles). Therefore, potential non-differential misclassification bias may exist and may have resulted in findings that favor the null hypothesis [34]. Therefore, this study may underestimate the differences between the low-income and the non-low-income groups.

Conclusions
The low-income group had (1) a higher accident injury hospitalization rate, (2) more severe injuries, and (3) a higher hospitalization mortality rate than the non-low-income group, supporting our hypothesis that health care inequality exists and is correlated with income despite mandatory, nation-wide, affordable medical insurance. Therefore, the government should collaborate with private and nonprofit organizations to create a more comprehensive system, including employment assistance, job fairs, public education, etc., to gradually eliminate health care inequality for low-income citizens. In addition, transport and falls injuries accounted for the majority of the injury cases in low-income inpatients.
Therefore, mobility aids should be covered by the NHI program for low-income citizens.

Consent for publication
Not applicable

Availability of data and materials
The data that support the findings of this study are available from the Health and Welfare Data Science Center, Ministry of Health and Welfare (HWDC, MOHW) but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of HWDC.
data interpretation, performed the statistical analysis, and drafted the manuscript. W-C Chien, C-H Chung, L Pai, C-S Tsai, and C-S Lin participated in the design of the study and data interpretation. Y-C Lin wrote the paper.

Figures
24 Figure 1 Comparison of medical costs between low-income and non-low-income inpatients under the NHI program in Taiwan.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download. TableS1-S5.doc