Risk factors associated with glaucoma status: a case study of Abesim Health Center (2019-2021)

Introduction

Glaucoma is a global public health problem since it is the second leading cause of irreversible blindness worldwide. It is asymptomatic in the early stages and is mostly detected in the late stages. According to the World Glaucoma Association, it is a chronic, progressive degenerative disease of the optic nerve that results in characteristic visual field damage [1, 2].

There is currently no clear and accepted explanation for the causes of glaucoma [3], but the condition is usually triggered by a buildup of high intraocular pressure, which damages the optic nerve and disrupts the transmission of images to the brain. However, there are general symptoms of glaucoma, which vary depending on the type of glaucoma and the severity of the condition. Unfortunately, glaucoma is incurable. But it is possible to slow or stop the progression of the disease with medication, and if ignored, it will eventually lead to blindness. By 2040, glaucoma is expected to affect 111.8 million people worldwide, mostly in Asia and Africa [2, 4]. There are 285 million people with visual impairment worldwide, of whom 249 million experience limited vision and the rest are blind (WHO, 2010). Currently, at least 2.2 billion people globally have near or far vision impairment [5].

About 1% of people over 50 years of age worldwide have glaucoma, and the risk increases with age [6]. Glaucoma is six times more common among African Americans or dark-skinned people than among Caucasians [4]. According to a 2013 study [7], primary open-angle glaucoma (POAG) is more common and progresses faster among Sub-Saharan Africans than among Caucasians. Compared with people of other continental ancestry, African ancestry increases the likelihood of POAG fivefold, resulting in a more severe course of the disease and a higher risk of blindness [8]. Adults in Africa have the highest prevalence of glaucoma, followed by Japan and Latin American countries [9].

A recent nationwide blindness study conducted in Ghana shows that over 19% of blindness in Ghana are due to glaucoma [10]. C. Liu et al. [11] showed that of the 976 patients diagnosed with glaucoma for the first time between January 2021 and October 2022 at the Northern Community Eye Hospital in Tamale, Ghana, more than a third of the 588 people who met the inclusion criteria had POAG. Numerous risk factors, including intraocular pressure (IOP) [12], aging [13], race [14], and a family history of glaucoma [15] (which is considered the first major risk factor for developing glaucoma) are often associated with the disease. The risk of developing glaucoma may also increase if in the presence of certain medical conditions, such as a family history of the disease [16], [17], hypertension [18], diabetes [19] ​​or certain lifestyle habits such as smoking [20].

The second highest prevalence of glaucoma in the world has been reported in Ghana [21]. It has affected northern ethnic groups more than the southern ones, yet little attention has been paid to this silent thief of sight. D.B. Kumah et al. [22] investigated the awareness and knowledge of glaucoma among undergraduate students at Kwame Nkrumah University of Science and Technology, Kumasi, Ghana, using a descriptive cross-sectional study based on convenience sampling and purposive sampling approaches. Their results showed a statistically significant correlation (p=0.047) between the participants’ knowledge of glaucoma and the college they belonged to in the institution. The students’ knowledge of glaucoma was limited and participation in vision tests or screening was also relatively low [22]. Another study found no significant difference in the incidence between the sexes, which may be due to the fact that the proportion of women affected by the disease was underreported [23].

Although the study [8] was limited to the age group of those above 40 years, the results, however, showed that West Africa had a higher prevalence of glaucoma vs. East Africa, Southern Africa, and individuals who were not of African descent. The causes of blindness in glaucoma patients were determined by M.T. Asemu et al. [24] using a parametric accelerated failure time (AFT) model and Kaplan-Meier survival analysis. According to their multivariate Weibull analysis, the probability of glaucoma patients to survive was significantly dependent on age, time of diagnosis, and glaucoma stage. Their results showed that the Weibull regression model was the most appropriate parametric AFT model to determine the important variables associated with glaucoma patients.

A study conducted by K.B. Otabil et al. in Sunyani, Ghana [25], demonstrated that children had a low prevalence of glaucoma (1.42%), while in adult men and women it was 9.25% and 8.77%, respectively. This implies that the risk of glaucoma increases with age, thereby making the adult population more susceptible to the disease explaining a high incidence rate in their productive age group. Another study conducted in Tema, Ghana, in the same year (2013) showed that the glaucoma prevalence was 3.7% and 14.6%, respectively, among people in the age groups of 40–49 years and 80+ years [25]. The study not only looked at the risk factors associated with glaucoma cases but also conducted statistical tests to analyze the data, draw conclusions and make decisions based on empirical evidence.

Numerous studies conducted in other parts of the world have confirmed diabetes and high blood pressure, along with increasing age, family history of glaucoma, African race, IOP, etc., among the consistent risk factors for developing glaucoma [12-15, 26]. Yet, glaucoma has been under-researched and under-funded despite being a sight-threatening disease. It has been ignored in the list of priority health interventions by the Ministry of Health in Ghana. Therefore, we are inspired to contribute to the body of research by performing a statistical analysis of the risk factors associated with glaucoma based on the patients who attended Abesim Health Centre, Sunyani, Ghana.

Material and Methods

Design of the study

In this section, we discuss the target population, coverage of the study region and data collection methods. In addition, we discuss several statistical methods and tools used to analyze these data. The scope of this study was limited to the demographic and medical characteristics of patients with and without glaucoma who visited Abesim Health Center from 2019 to 2021. This health center was built by the government in the early 1980s to provide basic health services to residents of the area and beyond. About half of the 900 outpatient cases recorded at the facility constituted self-reported vision problems. Patients who visited the center between January 2019 and December 2021 for general health and vision, aged 6 to 101 years, were included in the study. Secondary data on patients with and without glaucoma who visited the clinic during the study period were obtained from their medical records at the health center. Age, gender, marital status, employment status, religion, and place of residence were the demographic factors that were noted. Glaucoma, whether present or not, family history of glaucoma, diabetes, hypertension, and previous eye surgery were the medical variables collected from the patient records. During the study period, 537 patient records in Abesim Health Center were used using a simple random sampling approach. Of the 537 medical records, 250 belonged to patients with glaucoma and 287 represented glaucoma-free patients. For our study, we only included records containing all the data for the variables of interest.

Inclusion criteria for our study were as follows: patients who visited the center between January 2019 and December 2021 for general health and vision problems; age ranging 6 to 101 years, old and new patient medical records.

Exclusion criteria were as follows: patients under 6 years of age and incomplete patient medical records.

Patients

Our descriptive statistics show that of the total of 537 patients used in this study, glaucoma was present among 250 (47%) of them and absent in 287 (53%) individuals. All categorical variables are presented as frequency and percentage (Table 1). Continuous variables are reported as mean and standard deviation (SD).

Table 1. Frequency distribution of categorical variables used in the study

Table 1 shows that male patients prevailed (304; 56.61%) over female patients (233; 43.39%). Rural and urban areas were represented by 398 (74.12%) and 139 (25.88%) patients, respectively. More married individuals (58.47%) were in our sample than single patients (31.66%). More than half of the sample was represented by Christians (375; 69.83%) and 296 patients (55.12%) were employed. The share of unemployed patients was slightly smaller (241; 44.88%). Most of the patients had no previous eye surgery (485; 90.32%). Hypertension was characteristic of 269 (55.12%) patients, while diabetes was present in 287 (53.45%) individuals.

Table 2 demonstrates that most of the patients included in this study were between the ages of 51-60 years, followed by the patients aged 61-70 years. The age group 100+ years had the smallest representation.

Table 2. Age distribution of patients

The minimum and maximum ages of included patients were 6 and 101 years, respectively. The mean age was 53.1±19.2 years (Table 3).

Table 3. Descriptive statistics for age

Outcome variables of interest

The dependent variable (glaucoma) was denoted as categorical and assigned two possible levels (present or absent) with corresponding numerical codes of 1 and 0, respectively. Age, gender, location, marital status, employment status, religious affiliation, family history of glaucoma, diabetes, hypertension, and previous eye surgery were considered independent variables.

Gender was also a two-level categorical variable coded as 1 for males and 2 for females, and age was a continuous variable. Rural and urban areas were classified as level 1 and level 2. The categorical variable of marital status had four levels, with codes of 1 for single, 2 for married, 3 for divorced, and 4 for widowed individuals. The three levels of religious affiliation were categorical and were coded as 1 for Christians, 2 for Muslims, and 3 for all other religions. Employment status, previous eye surgery, diabetes, hypertension, and family history of glaucoma were all two-level categorical variables with 1 for ‘Yes’ and 2 for ‘No’.

Statistical analyses

For all variables included in this study, we calculated descriptive statistics. Mean and standard deviation (SD) are presented for quantitative variables, while frequency and percentage are presented for qualitative variables. After checking the assumptions of the model, categorical variables were tested using the chi-square test to see if they were associated with the dependent variable (glaucoma), and logistic regression was used to identify the risk factors of patients with or without glaucoma. Excel was used to collect data from the department of records in the health center, and R-Studio was employed to analyze the data for this study.

Based on various assumptions, dependent variable should be represented by binary independent observations, no multicollinearity should exist among explanatory variables, there should be no outliers. By testing the linear relationship between the explanatory variables and the logit of the response variable, we obtain the logistic regression model (1) of A.K. Iddrisu et al. [28]. This statistical model was executed using R-studio software [27, 28].

Based on observations, the dependent variable (glaucoma) has only two unique outcomes: 0 (absence) and 1 (presence) of glaucoma. The patient observations/measurements used for this study were independent of each other. Thus, the assumption of independent observations has been met. As can be seen in Table 4, the variance inflation factor (VIF) [29, 30] calculated for the independent variables (in order to determine whether they correlate with each other) ranges from 1.01 to 1.23, which is low. Since the VIF values ​​of all independent variables are less than 10, the assumption of multicollinearity among the explanatory variables has been met.

Table 4. Variance inflation factor (VIF) of independent variables

We used Cook’s distance for each of the observations to check for the presence of outliers [31]. The first 15 highlighted observations in the dataset are outliers (Figure 1). Regardless of the presence of outliers in the dataset, they have no effect. And since the value of Cook’s distance is less than 1, this assumption is met. We assumed a linear relationship between the log odds and age, the continuous variable used in this study.

Figure 1. Plot of Cook’s distance.

Results

We performed the chi-square test to determine the association between dependent variable and independent categorical variables using a significance level of 5% [32]. Both univariate and multiple logistic regression were performed with all variables used in this study to examine the performance of the variables both individually and when controlling for other predictor variables.

This test aimed at identifying risk factors that accurately predict the status of the outcome variable (glaucoma status). The risk factors associated with glaucoma status based on our chi-square test results presented in Table 5 indicate that gender (chi-square=19.003, p<0.0001), marital status (chi-square=11.181, p=0.0108), religious affiliation (chi-square=35.671, p=0.0001), family history of glaucoma (chi-square=9.0801, p=0.0026), previous eye surgery (chi-square=9.0637, p=0.0026) and diabetes (chi-square=11.8970, p=0.0006) exhibit statistically significant associations with the outcome variable (glaucoma status). Contrariwise, the locality type (chi-square=0.0017, p=0.9667), employment status (chi-square=0.6676, p=0.4139) and hypertension (chi-square=0.6676, p=0.4139), did not show statistically significant relationships with glaucoma; hence, they were not considered for further analysis.

Table 5. Chi-square test of association between outcome variables and risk factors

An independent t-test was performed to determine the glaucoma status (presence or absence) vs. age (Table 6). An alpha level of 0.05 was used. We found no statistical difference (t=-1.6468, df=534.6300, p=0.1002) between the scores of glaucoma absence status (M=51.8397) and glaucoma presence status (M=54.5360). The magnitude of difference in means with 95% CI (-5.9127; 0.5201) was not significant. Thus, the null hypothesis stating that the means of the two groups are similar was supported.

Table 6. Results of t-test of glaucoma status and age

We then performed univariate logistic regression to assess the effect of individual factors on glaucoma and multiple logistic regression (MLR) to assess the effect of factors on glaucoma when other factors are controlled.

Table 7 shows that marital status (married) was marginally significant for glaucoma. Gender, marital status (widowed), religious affiliation (Muslim), family history of glaucoma, diabetes and previous eye surgery were also significant for glaucoma. There is a decrease in the log odds of glaucoma status in women vs. men (Est. 0.7898; CI: 1.1431, -0.4413; p<0.0001). The log odds for married (Est. -0.3202; CI: -0.6959, -0.0536; p=0.0937) and widowed (Est. -0.8346; CI: -1.5261, -0.1758; p=0.0149) were lower compared with single. Muslims exhibited lower log odds of glaucoma status than Christians (Est. -1.1674; CI: -1.5772, -0.7714; p<0.0001). Those without a family history of glaucoma had lower incidences of glaucoma vs. those with a family history of glaucoma (approx. -0.5586, CI -0.9144, -0.2055, p-value 0.0020). Glaucoma was lower among those without diabetes (Est. -0.6192; CI: -1.5781, -0.2760; p=0.0004). Individuals without previous eye surgery demonstrated lower incidences of glaucoma vs. those with previous eye surgery (Est. -0.9499; CI: -1.5781, -0.3576; p=0.0021).

Table 7. Univariate analysis of the effect of individual factors on glaucoma

Table 8 shows the MLR using only the factors significantly associated with glaucoma. The log odds of glaucoma cases when all other factors were controlled were: Est. 2.2110; CI: 1.2088, 3.2523; p<0.0001. Gender, marital status (married and widowed), religious affiliation (Muslim), family history of glaucoma, diabetes and previous eye surgery were significant for glaucoma when other factors were controlled. Gender was significant with a decreased log odds of glaucoma status in women compared with men (Est. 0.8828; CI: 1.2736, -0.4996; p<0.0001) when other factors were controlled. The log odds of married (Est. -0.4426; CI: 0.8662, -0.0234; p=0.0393) and widowed (Est. -1.3240; CI: -2.1362, -0.5509; p=0.0010) were lower than those of single people when other factors were controlled.

Table 8. Multiple logistic regression analysis of the effect of individual factors on glaucoma while controlling for other factors

Muslims exhibited reduced log odds of glaucoma status compared with Christians (Est. -1.2230, CI: -1.6658, -0.7969; p<0.0001) when other factors were controlled. The expected log odds of those without a family history of glaucoma were lower than in those with a family history of glaucoma (Est. -0.6947; CI: -1.0952, 0.3002; p=0.0006) when other factors were controlled.

Glaucoma was lower among those without diabetes (Est. 0.5928; CI: -0.9884, -0.2014; p=0.0031) when other factors were controlled. The expected log odds of those who had no previous eye surgery exhibited lower incidence of glaucoma compared with those who had previous eye surgery (Est. -1.1190; CI -1.8097, -0.4644; p=0.0011) when other factors were controlled. Hence, the logistic regression model is as follows:

Table 9 illustrates the goodness-of-fit test performed using the Hosmer-Lemeshow test. It is clear that p=0.2185 is greater than alpha level of 0.05, which means that there is no statistical significance, and therefore the model is good.

Table 9. Hosmer-Lemeshow test results

Discussion

We investigated risk factors for glaucoma in Abesim Health Center from 2019 to 2021. After identifying risk factors associated with glaucoma, we developed the best fitting model for the target population. As a public health threat, glaucoma is asymptomatic in the early stages and is usually detected when symptoms begin to appear. According to [14, 33], glaucoma can lead to irreversible blindness, although the risk variables that contribute to the cause of this blindness vary depending on the study.

We addressed this issue by identifying the risk factors that should not be ignored when testing for glaucoma in the study population using data from Abesim Health Center. Based on the collected data, glaucoma status was considered as the dependent variable, while all other variables were considered as independent variables. Chi-square test, t-test, and logistic regression model were the three statistical approaches applied to the dataset.

First, we performed the chi-square test to evaluate the association between risk variables and glaucoma status. The results of chi-square test showed that diabetes, family history of glaucoma, and previous eye surgery were the medical characteristics that were associated with glaucoma. Gender (male and female), marital status (single, married, divorced, and widowed), and religious affiliation (Christian, Muslim, and other religions) were the demographic characteristics associated with glaucoma. To evaluate the importance of glaucoma status and age, we also used t-test. The results of t-test confirmed that age did not have a statistically significant effect on glaucoma status, which is contrary to the findings in another study [8]. Thus, according to the study data, age alone may not be a strong predictor of glaucoma status.

To create a logistic regression model connecting the risk variables with the outcome (response variable), it was necessary to analyze the significance of the effects of different risk factors on glaucoma status. After single logistic regression analysis to evaluate the effect of each risk factor on glaucoma status, MLR was employed to evaluate the effect of each risk factor while accounting for additional risk variables. The results of the logistic regression model showed that glaucoma status was significantly lower in women, which was consistent with other studies [8, 34]. Our model once again demonstrated that glaucoma status was significantly lower in married and widowed people, Muslims, people with no family history of glaucoma, as well as individuals without diabetes and previous eye surgery.

The results of our study are consistent with the hypothesis that adults are more likely to develop glaucoma than children [25]. However, this study contradicts the findings of another study [35] that showed that neither gender nor diabetes mellitus are associated with glaucoma. Again, the lack of association between hypertension and glaucoma status in this study contributes to the debate about whether hypertension should be included as a risk factor for glaucoma and contradicts the results of the study by S.O. Baboolal et al. [18]. 

Limitations

Abesim Health Center was the only source of the study sample, which could reduce the applicability of our results. However, these findings highlight the importance of risk assessment in glaucoma screening by emphasizing specific factors that may be different for this demographic group. However, several correlations, particularly those related to hypertension, remain unclear and indicate the need for further research. This work could be expanded in the future by examining possible interactions between risk factors and using a larger, more diverse sample. Studies of this kind could improve screening procedures and direct targeted preventive measures for glaucoma in high-risk groups.

Conclusion

Our statistical analysis identified gender, marital status, religious affiliation, family history of glaucoma, diabetes, and previous eye surgery as the main risk factors for glaucoma in the study population. By providing information on specific preventive and intervention measures, these results enrich our understanding of the demographic and health-related factors that may influence glaucoma risk.

Availability of data and materials

We do not have permission to distribute the data.

Acknowledgments

We express our gratitude to the management of Abesim Health Center who contributed to the data collection for this study.

Conflict of interest

The authors declare no conflicts of interest.

Ethical approval

This study was approved by the Ethics Committee of Abesim Health Center, Sunyani, Ghana (No. 21 of 2022-01-14).

Declaration on the use of artificial intelligence tools

The authors declare that they did not use artificial intelligence (AI) tools in the preparation of this article.