The psychosis analysis in real-world on a cohort of large-scale patients with schizophrenia

The overall analytical framework is shown in Fig. 1. From the risk and treatment effect existing in the follow-up data, we extracted the focus groups with negative therapy effect and high risk. Age, gender, blood glucose and other indicators were as the closely monitored indicators. The characters of group were used to design the prediction model and feature clustering, which was used to analyze the group and quantify treatment effect and risk. The quantified treatment effect and risk were used afterwards for estimating the influence of different factors in schizophrenia.

The study population including 348,801 individuals with valid registered records who living in Guangdong province and were diagnosed with ICD code F20.* during the period from January 1, 2010 to May 30, 2019, of which 1755 were excluded due to missing or implausible data. The follow-up data was obtained from the Guangdong Mental Health center network medical System (GDMHS), which was initiated in 2000 and covered over 99% of schizophrenia patients in the province. Within the disease registration report system, patient profile, diagnosis and treatment information in hospitals were reported and recorded into the system via a computer network.

Validity of the data was confirmed by psychiatrist doctors and a chief physician in the hospital after data entry. Moreover, the Health Commission verified the data annually through sampling survey to ensure data reliability. Since the data from GDMHS did not contain individually identifiable information, and therefore individual informed consent was not required.

We conducted a descriptive analysis of the data based on age groups and disease course length. The evaluation models of therapeutic effect and risk were focusing on gender, economic status, medication adherence, education level, risk evaluation, region distribution, register type and age group. The data was also represented as points and density levels to project them geographically for presenting the location characteristics of patients. The location factor was taken into account in order to investigate the regional distribution of mental disorders.

The measures of main treatment outcome in this study were quantified treatment effect and risk. The calculation method of the efficiency rate is as Eq. (1). where Cured is the patients cured, Improved is the patient has improvement on disease, and Total is the total follow-up records of this research group.

During the evaluation of treatment risk, the assessment can be classified into six levels based on the differences of patients’ symptoms, as shown in Table 1.

In this study, we adopted various statistical analysis methods to explore the influence of factors on the incidence of outcomes to reveal the treatment characteristics of real-world schizophrenia at different stages of disease course as well as relevant factors affecting the treatment effect and risk of schizophrenia.

To assess the therapeutic characteristics of schizophrenia in different populations and stages of the disease, we grouped patients according to different disease courses and calculated proportion of each factor in different disease stages. The influence of each factor on the indicators was divided into five different stages: 0–6 months, 7–12 months, 13–24 months, consolidation period (2–5 years), and maintenance period (over 5 years). In addition, we sorted and grouped different attributes for each factor, such as households was grouped as urban and rural areas, economic conditions was grouped as poverty and non-poverty, etc., to explore the intra-group differences in the influence of different factors.

In the descriptive analysis, we conducted a statistical analysis towards the data. We grouped patients by age and gender, to observe the number of patients in different age groups and whether there existed differences between different gender groups. In the meantime, the course of disease was grouped according to the follow-up time points, and analyzed count and variation trends of male and female cases in different disease courses respectively. Finally, the regional distribution of disease was depicted based on the follow-up sites of mental patients to analyze the regional characteristics of the incidences of schizophrenia.

During the quantitative analysis of treatment and risk, the patient treatment data, such as treatment development status including cure, improvement, no-change or aggravation, was recorded. In order to calculate efficiency and risk scores, we grouped the data by course of disease in year, and then summarized the follow-up records by year and patient ID. According to a list of evaluation metrics, the total effective rate and risk of each patient per year were calculated. After that, the mean (95% confidence interval) was used to calculate the total effective rate and risk per year. We used the time length of the disease as an independent variable to investigate the variation trend of different factors on the time axis of the disease course towards efficiency and risk, and to compare the differences between different attributes of the same factor. In order to further analyze the trend of schizophrenia among different factors, the Cox-Stuart method was used to test the significance of the risk and efficacy trend of each group to explore the changing trend of risk and efficacy of different attributes in each factor group with the course of disease. In risk assessment, if a patient appeared with multiple different evaluation levels, level 0 to 2 was represented as 0 point and level 3 to 5 was represented 1 point. Then the amount of times appeared a year is counted as the count of corresponding points.

In addition, we investigated the characteristics of the regional distribution of treatment effects and risks, which were conducive to the early detection of problems from geographical distribution. A hotspot map was utilized to represent the distributions of relevant indicators of patients. First, we exploited a gradient of color levels to map the patient population density in different areas. Then, multiple groups of efficient intervals were divided and the interval values were indicated by the depth of color, with the deeper color representing the higher efficient value. Efficient mappings were represented in node color on the map. In order to map the risk assessment, we used risk scores as a variable and plotted them as the sizes of nodes, with the larger nodes representing higher risk score in the region. Meanwhile, with the goal of analyzing the demographic characteristics of a region, the average age and length of disease course of the patients in the region were counted to be the labels of the nodes.

Analyses were stratified to examine the epidemiology of psychosis in Guangdong province from different perspectives, with exploring the characteristics of psychosis population, and the differences and evolution trends of schizophrenia in urban and rural, region, economic status, education status and other aspects.

The statistical information of social demographic characteristics was shown in Table 2. We first analyzed the internal characteristics of the follow-up data in groups. Comparing the distribution of various factors at different stages, the count of patients at different stages of the disease presented an increasing trend. The number of patients in the maintenance period was the most, while patients at the onset of 0–6 months was the least. According to the follow-up age, young people and middle-aged and elderly people were the main population with the disease, representing that the disease age was mainly distributed between 18 and 65 years old. In terms of geographical category, the population groups of patients in Guangdong province were mainly distributed in the economically developed areas, e.g., Guangzhou, Foshan and Shenzhen, as well as the densely populated areas, e.g., Chaoshan, and Meizhou. In addition, most schizophrenic patients had a low level of education, and the poverty rural areas having more patients may also represent a lower level of education, compared with the non-poor urban areas.

Among the follow-up data, the number of cases at the age of initial onset (in units of 10) were grouped and summarized, and the distribution of onset age were shown in the Fig. 2. The highest incidence of schizophrenia in both men and women was aged between 20 and 24, and the count of patients between 15 to 29 years old was the majority. Also, there were more patients in men before the age of 30–34 years old than in women, and fewer patients after the age of 34. In view of such situation, it was appeared to be attributed to men may be related to work pressure and women may be related to menopause physiology. We also observed the variation in the number of patients with schizophrenia over course of disease (respectively observed in men and women). As shown in the Fig. 3, by dividing the course of disease into different intervals, most patients’ course of disease was concentrated between 5 and 10 years. The number of cases with course of disease increased from 0 to 10 years, and then presented a trend of decline. Meanwhile, the result demonstrated that men patients had more cases per course of disease than women.

We first computed the effectiveness and risk of each factor using a list of quantitative strategies aforementioned. According to the social demography of the data, the effects of various factors with the time intervals of disease course on the efficiency and risk were analyzed. Finally, we explored the efficiency and risk of trends of disease course in eight groups respectively.

Overall, the efficacy tended to worse with the course of the disease, and the risk score was higher with the course of the disease. Through statistical analysis of various factors, there was little difference in the effect of gender on efficiency. However, there was a significant difference in the effect of gender on risk score, demonstrating that the risk score of men was higher than that of women (Figs. 4 and 5). The effect of economic conditions for treatment efficacy was not prominent (Fig. 6). When the course of disease exceeded 1 year, the treatment efficiency of poor economic condition was relative higher. While the effect of economic conditions on the risk score was more obvious, reflecting in the fact that poorer economic conditions had higher risk scores at each time point in the disease course compared with better economic conditions, and the risk score existed a certain gap between different economic conditions (Fig. 7). Medication adherence exerted a significant effect on the therapeutic efficiency (Fig. 8). Three different grades of medication adherence indicated that better medication adherence represented higher therapeutic efficiency. In particular, there was a prominent difference between the effective rate of intermittent medication and non-medication, suggesting that drugs played a certain role in the improvement of psychiatric treatment. As for risk assessment, there was little difference between medication compliance levels and a certain degree of fluctuation with the development of disease course. In general, risk assessments for regular medication adherence were lower (Fig. 9). Dividing the age groups by six groups manifested that the effective rate of young people and middle-aged and elderly people was relatively higher, and the effective rate of youngsters shown a relatively rapid decline in the first 10 years of the course of disease, while young people had a higher risk score in the overall course of disease distribution (Figs. 10 and 11). In view of the distribution of efficiency and risk in urban and rural areas, the efficiency of urban environment was higher in the first 10 years of disease course, while the efficiency of rural was overpassed the urban when the disease course continues to increase. In terms of risk, the risk score of urban areas was lower than that of rural areas in each disease course, which existed some certainly consistent with the analysis of the corresponding economic status (Figs. 12 and 13). According to the classification of the education level of the population, there was little difference in the treatment efficiency among different education levels, but the patients who had received higher education tended to have the highest treatment efficiency at the beginning of the disease, while the illiterate and semi-literate patients had a lower treatment efficiency (Fig. 14). In terms of risk assessment, correspondingly, the patients who had received higher education had the lowest risk score in each course of disease, while the risk score of the patients who had been illiterate and semi-literate and primary education was relative higher (Fig. 15). The distribution of treatment efficiency and risk of the major 8 regions of Guangdong province was analyzed and visualized in (Fig. 16). The regions with better economic conditions, such as Guangzhou, Foshan and Shenzhen, had higher treatment efficiency at the initial stage of disease course, but shown the lower treatment effect in the overall course of disease. On the contrary, in the risk assessment, the better economic conditions reflected the lower risk score and the risk was generally higher in the remote areas of north Guangdong (Fig. 17). Correlated the effectiveness and risk, four different treatment effects were classified in the risk assessment according to the patients’ recovery, including cure, improvement, no change and aggravation. It could be seen that the risk assessment value represented by the aggravation of the patient’s illness was the highest and different with the other three indicators. In the analysis of treatment effect, the risk was divided into 6 different grades, and overall the lower risk degree corresponded to the better treatment effect. But there were still some cases where the risk was higher and the treatment effect was also higher. In the risk assessment, because of the incomplete follow-up records of some patients’ cure conditions, data in some disease courses was missed (Figs. 18 and 19).

Furthermore, statistical method (Cox-Stuart) was used to analyze the significance of the risk and therapeutic effect trend in each group. Among them, the intro-group factors whose risk increased significantly with the course of the disease were: urban and rural register factors, illiteracy and semi-illiteracy attributes in education level factors, poverty attributes in economic factors, young people and middle-aged people in age groups, females in gender factors, as well as remote mountainous areas in northern Guangdong in geographical categories. While in the effective rate analysis, the effective rate of each group mostly showed a significant decline with the course of disease, except for the following two groups: level 5 in risk factors and child attributes in age group. These results could also be derived from the corresponding curve trends.

Through the comparison between the different factors within the set of attributes to be efficient and risk difference, we identified that different degrees of medication adherence, different condition of economic, different ages and risk level had difference in treatment effect in general. In the risk assessment, there were significant differences among the attributes of the group in gender, economic status, age group, urban-rural and treatment efficiency.

A hotspot map was used to represent the distribution of effectiveness and risk assessment of the patients, where node colors and node sizes represent the treatment effect and risk assessment respectively, as shown in Fig. 20. By observing the color level mapping to the number of cases, the areas with a large number of cases tended to be distributed in areas with high population density, such as Guangzhou, Dongguan and Chaoshan area. By analyzing the mapping of node colors and sizes, it could be found that in the economically developed cities, such as Foshan, Guangzhou and the Pearl River delta region. The treatment effect was moderate but the risk assessment was generally low. Meanwhile, it could be seen that the treatment effect of schizophrenia was poor and the risk assessment was significantly higher in the remote mountainous area of the rural-urban junction area in north Guangdong, which is relatively backward in economic development, especially in Lianshan county and Fogang county. It worth to be noted that Chaoshan area (including Chaozhou and Shantou cities) was a special area with a large number of patients and a relatively long course of disease. Although this area possessed a certain degree of therapeutic effect, the risk assessment was generally high and the distribution was relatively dense, indicating that this might be an area that needed to be paid more attention to. In southern Guangdong province, some coastal cities, such as Kaiping and Yangjiang, had low population density, but poor treatment effect and high-risk assessment, indicating that these were also one of the regions with serious mental illness in current situation. While in southern coastal areas such as Maoming and Zhanjiang, antipsychotic treatment was more effective and less risky.

Through the above analysis of the distribution characteristics of schizophrenia in various regions of Guangdong province, it could be found that the efficacy and risk were correlated. Poverty areas with backward economic conditions such as northern Guangdong were often accompanied by poor curative effect and high-risk assessment. On the contrary, in areas with developed economy and good medical resources, such as Guangzhou, Foshan, Shenzhen, the risk assessment was lower, indicating that the disease was effectively controlled. Meanwhile, Chaoshan area deserved more attention for its special geographical location and disease distribution condition.