Late diagnosis of Human Immunodeficiency Virus infection and associated factors

Objective: to analyze the occurrence of late diagnosis of infection by the Human Immunodeficiency Virus and its associated factors. Method: this is an epidemiological, cross-sectional and analytical study, carried out with 369 people followed-up by Specialized Assistance Services, undergoing anti-retroviral treatment, and interviewed by means of a questionnaire. Univariate analysis was performed using Pearson’s chi-square test or Fisher’s exact test and Kruskall-Wallis test, and multivariate analysis using the ordinal logistic regression model of proportional odds. Results: the occurrence of 59.1% for late diagnosis of the infection was observed; the probability of later diagnosis is greater among people who have a steady partnership, when compared to those who do not; with increasing age, particularly above 35 years old; among those with lower schooling; for those who seek the health services to have an HIV test when they feel sick; and for those who test HIV less often or never do it after sex without a condom with a steady partner. Conclusion: the knowledge on the high proportion of late diagnosis and its associated factors verified in this study make the planning and implementation of new policies and strategies aimed at the timely diagnosis of the infection imperative.


Introduction
Since the beginning of the Human Immunodeficiency

Virus (HIV) infection pandemic in the 1980s, Brazil
has implemented a series of government and social measures to tackle the epidemic. In the last few decades, a significant decrease in AIDS morbidity and mortality has been observed in the country, due to the introduction of universal and free access to antiretroviral therapy (ART), the harm reduction policy, the implementation of combined prevention strategies, the recommendation of treatment as prevention, and the wide range of diagnostic tests (1)(2)(3) . These strategies have jointly contributed to the increase in the life expectancy and quality of life of people living with HIV (PLHIV), to the decrease in hospital admissions due to the reduction of opportunistic infections, and to the reduction of HIV transmission (4)(5)(6) .
Despite the efforts and integrated actions of the governments, of civil society, of social movements, and of non-governmental organizations in tackling the AIDS epidemic, the tendency for its morbidity and mortality to decline, the expanded access to ART, and the technological advances in case management, this condition remains at the top of public health problems, affecting the population's quality of life and impacting the economy and the social and family structures (7)(8) .
The delay in the diagnosis and the consequent late assistance to PLHIV are some of the main concerns in combating the epidemic (7,9) . The early diagnosis, together with the immediate initiation of the treatment, brings irrefutable health benefits to PLHIV, due to its greater effectiveness in maintaining the immune status and reducing morbidity and mortality. It also contributes to its prevention, since the spread of the infection is avoided in a phase marked by high viral loads and high infectious potential, which also results in greater investment of time and resources by the health systems (10)(11)(12) .
Clinical and laboratory monitoring of HIV infection is performed by counting CD4+T lymphocytes (LT-CD4+) and viral load (VL). According to data from the Ministry of Health (MoH), in 2018 27% of PLHIV arrived at the health service with an HIV infection late diagnosis (LD), considering the CD4 count criterion below 200 cells/mm 3(13) .
A research study carried out in Brazil, which used the criterion of late onset of clinical follow-up for asymptomatic patients with an LT-CD4+ count below 350 cells/mm 3 , in the 2003-2006 period, revealed that the prevalence of late onset was 58.6%, resulting in an increase of more than a third in the AIDS mortality rates. Another relevant conclusion of the study was that if all patients had started treatment in a timely manner, the decrease in AIDS mortality could have been 62.5% (against the 43.0% observed), between 1995 and 2006, increasing by 45.2% (14) the effectiveness of the program to deal with this epidemic. To perform the sample calculation, the 95% confidence interval (CI) was considered, the proportion of late diagnosis was 60% (18)(19), the maximum allowed error was 5% and the probability of sample loss was 10%, resulting in a sample of 369 people with HIV.
As the sampling was stratified, the sample calculation was performed for each stratum, that is, for each SAS, proportionally to the population. In order to ensure the minimum sample size in the period of one month of data collection and to select the research participants, a probabilistic draw of the days for the collection in each SAS was carried out.
The number of days to be drawn in each service was obtained by the ratio between the minimum sample size calculated for each SAS and the mean daily attendance for dispensing anti-retroviral drugs in each service. An additional day (backup) was drawn for collection in each service, assuming the occurrence of unforeseen events or difficulties in the dynamics of the services' operation.
Thus, single-stage cluster sampling, stratified by the health service, was used for the draw procedure, with the day being the primary sampling unit and the strata designed by the establishments.
During data collection, the planning of the days drawn was not carried out exactly, due to the service logistics in the services, which on some dates did not allow access to interview users. These days were replaced by other dates that made it possible to reach the sample size previously established. In addition, there was approximately 15% of refusals to participate in the study, a fact already expected due to the fear of the stigma experienced by many PLHIV. However, it is noteworthy that the interviews were conducted in places that guaranteed the privacy of the participants, which favored people's adherence to participate in the study, reaching the calculated sample.
Regarding the variables used in the study, the time of diagnosis was considered as the outcome, obtained by the LT-CD4+ count proxy variable at the time of diagnosis (10,12,14,19) . This response variable was categorized as follows: timely diagnosis, characterized by an LT-CD4+ count equal to or greater than 350 cells/mm 3 ; late diagnosis, defined by LT- In the multivariate analysis, to assess the factors associated with late diagnosis, as it is an ordinal categorical outcome (very late, late and timely diagnosis), the ordinal logistic regression model of proportional odds was used (20) . In this study, the Odds Ratio (OR) represents the chance of a patient having a later diagnosis.
During the modeling process, all the variables with a p-value below 0.20, according to the univariate analysis, were included in the multivariate model. simultaneously and the backward modeling process was performed again, with only the significant variables remaining at the 5% level.
After adjusting the final model, the OR values were estimated, with respective 95% confidence intervals (95% CIs). It should be noted that the final model presented a good fit, according to the Deviance statistic (p-value = 0.955), and the assumption of parallel lines proved to be valid (p-value = 0.590).
In all the analyses, the confidence intervals had a 95% confidence level and the p-value < 0.05 decided

Results
Most of the study participants were male (cisgender men) (55.0%) and were in adulthood, and aged The results also showed that 238 (64.5%) of the total participants had a high VL at the time of diagnosis, with values above 10,000 copies/ml. People with a steady partnership were diagnosed later, as were those of the Catholic religion. In addition, it was shown that the lower the schooling level (in years of study), the later the diagnosis. Table 2 shows the factors related to access to the health services and their association with HIV infection diagnosis.
The only variable that was statistically associated with a late diagnosis of HIV infection in relation to access to the health services, in the univariate analysis (p-value < 0.05), was the reason why they sought the health service to perform the HIV test (p-value < 0.001). It was verified that the diagnosis was later among those who sought the service because they felt sick. Table 3 shows the variables related to the sexuality of the interviewed PLHIV that were associated with HIV infection diagnosis, in the univariate analysis.
It was observed that the diagnosis was later among those who never tried to do the quick test after intercourse without a condom with a steady partner and who reported being ashamed to suggest the use of condoms to their partner (p-value < 0.05).
In the multivariate analysis, considering all the variables that remained in the final models of each block in a single model, the variables shown in Table 4 remained in the final model.

Discussion
The high occurrence of late or very late HIV infection diagnosis (59.1%) found portrays an alarming reality, with higher data than the national ones, in which the percentage of 42% of people diagnosed with a lower LT-CD4+ count was identified at 350 cells/mm 3 in 2015 (18) .
However, the percentage of delayed diagnosis of HIV infection was similar to results found in other developing countries, such as China, which corresponded to 72.6% in the period from 2009 to 2010 (9) , Ethiopia, with 68.8% of delayed diagnosis in 2014 (21) , and Mexico, with a late diagnosis prevalence of 49% in the period from 2008 to 2013 (22) . However, these surveys used heterogeneous criteria to define a delay in the diagnosis or late onset, which non-standardizes the results and hinders the comparison among the countries.
In Brazil, research studies on late onset for the care with HIV infection published from 2011 to 2016 estimated prevalence rates ranging from 52.5% to 69.8% (14,(23)(24)(25)(26) . These results are below what is expected for consolidated public health programs for PLHIV, such as that of Brazil (24) .
Like Brazil, most of the countries in Latin America face a concentrated epidemic, with a large number of people still undiagnosed and a high prevalence of late diagnosis and, consequently, late onset of ART (27) .
Late diagnosis is a continuous, worrying, and serious challenge for the control of the AIDS pandemic, as it is directly related to higher rates of morbidity and mortality (28)(29) and to the need for greater investment by the health systems (10,30) .  (23) . Several studies have described similar percentages of LD among men and women, and it is relevant to consider pregnancy as a possible confounding variable, since the request for HIV testing is a prenatal routine and favors the timely diagnosis in pregnant women (12,23) .  (25,29,(32)(33)(34) .
Marital status is also a determining factor for LD.  (38)(39) . Generally, people with a steady affective partnership, due to the low perception of risk (37) , use condoms less frequently during sexual intercourse, which makes them vulnerable to HIV (40) and to late diagnosis.
The increase in schooling associated with a lower chance for later diagnosis was also evidenced in other studies (26,29,38,(41)(42) . This direct relationship between low schooling and increased risk for LD is probably due to greater difficulty in the access to information on means of  (42) .
The motivation for HIV screening and the frequency of testing after sexual intercourse with a steady partner were associated with the occurrence of the infection LD.
Those individuals who were motivated to take the test for feeling sick, and those who took the test occasionally, almost never or never had a better chance of being diagnosed late. Periodic screening for the virus is related to people's perception of the health-disease process and of their vulnerability to HIV, as well as to their access to the health services that offer the test (39,42) .  (43) .
A number of research studies show that people do not consider having an HIV test because they are apparently healthy (38,42) . In Brazil, as the epidemic is concentrated, people are referred for HIV screening usually only when they have symptoms suggestive of infection or when they exhibit certain behaviors that increase the risk of transmitting the virus. Thus, screening moves away from a risk-focused strategy, as it disregards those individuals who do not openly express attitudes that are generally associated with an increased risk of contamination (26) . Therefore

Conclusion
The high occurrence of late diagnosis for this infection is predominantly associated with sociodemographic and sexual aspects, without disregarding factors related to access to the health services. The likelihood of a later diagnosis is greater as age increases, among people who have a steady partnership and lower schooling, in individuals who sought the health service to perform the HIV test because they felt sick, and among those who never or almost never performed the quick test after an unprotected sexual intercourse with a steady partner.
It should be noted that the analysis was carried out in a broad and in-depth manner with primary sources of investigation, addressing, in addition to sociodemographic variables, factors related to sexuality and access to the health services, which strengthens the reliability and relevance of the findings.