Background

Globally, 36.7 million people were living with Human Immunodeficiency Virus (HIV) in 2016 [1]. The United Nations Programme on HIV/AIDS (UNAIDS) 90–90-90 target states that 90% of people living with HIV should know their status, 90% of people living with HIV who know their status should be on treatment, and 90% of people on treatment should be virally suppressed. A record 19.5 million people were accessing antiretroviral therapy in 2017, and for the first time, more than half of all people living with HIV are on treatment [2].

In Ethiopia, Antiretroviral treatment (ART) began in 2003 and free ART was launched in January 2005. In 2016, only 67% of HIV positive people are believed to know their status. Additionally, the treatment coverage of all HIV positive people was 59% and only 51% of all HIV positive people on highly active antiretroviral therapy (HAART) had viral suppression [1]. In 2017, an estimated 738,976 Ethiopians were living with HIV and all of them were eligible for ART treatment. However, it was reported that only 426,000 were taking ARV in the same year [3].

The estimated 36-month retention from 2008 to 2013 was; 65% in Africa, 80% in Asia, and 64% in Latin America and the Caribbean [4]. In low and middle income countries HIV patient retention at 36 months on treatment averages 65–70% [4]. Patient retention in care is a challenge in many African countries [5, 6]. Similarly, in various regions of Ethiopia, during the pre-ART era, about one third of HIV positive people receiving clinical care were lost to follow up [7, 8]. Retention in care within 12 month of treatment initiation varies from 83 to 94% in Ethiopia [9].

Qualitative studies in Addis Ababa, Bahir Dar and Gondar, Ethiopia, revealed that fear of stigma, care dissatisfaction, use of holy water, fasting, and economic constraints discouraged retention in care. Whereas social support and restored health and functional ability motivated retention [10, 11].

Though Ethiopia has improved patient retention from 2005 to 2013 [12], still there are large number of patients that discontinue from treatment [13]. Unlike other regions of the world [4], there is paucity of evidence on the rate of attrition among HIV positive patient in Ethiopia. A systematic review and meta-analysis on ART treatment discontinuation of HIV patients in Ethiopia indicated various factors are responsible for treatment discontinuation [13]. However, the study was limited to only 9 papers and only three regions out of nine regions of the country. At the same time, this previous study did not include pooled incidence and prevalence of loss to follow up, death and transfer out [13]. In addition, the study does not show the magnitude of patient retention and attrition. In view of this, we decided to investigate the retention, attrition and its determinants at national level. Patient attrition includes loss to follow up (LTF), death (D) and transfer out (TO) which is the official transfer of the patient to another clinic. Therefore, this study aimed to determine the pooled magnitude of HIV patients’ clinical retention, attrition and identify factors associated with retention and attrition.

Methods

Study design

Systematic review and meta-analysis was done using studies conducted in English language in Ethiopia. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [14] was strictly followed during search of studies and analysis process.

Eligibility criteria

We included studies conducted on adult HIV positive individuals’ retention in HIV care in Ethiopia. Studies that reported lost to follow up, defaulters and treatment discontinuation, transfer out and death were included. Studies that used any length of follow up period were eligible for this review. There was no restriction based on study designs of primary studies except systematic review and meta-analysis. Both published and unpublished primary studies from the start of ART program in January 2005 in Ethiopia to last date of literature review on June 6, 2019 were included. The reason for restriction of study period was that the first article in Ethiopia about HIV patients’ retention in clinical care data were collected in 2005 and published in 2008 and the last date of literature review search was on Jun 6, 2019. Moreover, qualitative studies which didn’t report quantitative outcomes of interest were excluded.

Search strategy

Published research was identified through a systematic search of PubMed, Google Scholar and the African Journals Online (AJOL). Information were retrieved from online published data sources and unpublished thesis works. Online databases include PubMed, Google scholar and African Journals online (AJOL). We used search terms in PubMed as (retention) OR attrition) AND HIV) OR AIDS) AND patients) OR positive people) AND Ethiopia). Moreover, keywords for review included; retention, attrition, loss to follow up, dead, discontinuation, defaulters and treatment outcomes of ART in Ethiopia and were used to search additional articles. We audited the references of all articles which deemed important for our outcomes of interest. Moreover, we searched for unpublished works in the postgraduate library of Debre Markos University and other universities institutional repositories.

Study selection and quality appraisal

Studies were evaluated using title, abstract and full text. Studies that used retrospective follow up design were included only if full cohort assessment report was obtained. Quality was assessed using Critical Appraisal and Assessment of Methodological Quality of Studies for Systematic Review and Meta-analysis of Observational Epidemiological Studies Reporting Prevalence and Cumulative Incidence Data [15]. Two authors independently assessed the quality of the studies. Where disagreement occurred, studies were discussed until consensus was reached. Studies were assessed on nine dimensions and for each dimension the authors would indicate that the criteria was met or not met. In cases where there was insufficient information to determine the study could be graded “unclear.” The following criteria were used: 1. was the sample frame appropriate to address the target population? 2. Were study participants sampled in appropriate way? 3. Was the sample size adequate? 4. Were the study subjects and the setting described in detail? 5. Was the data analysis conducted with sufficient coverage of the identified sample? 6. Were valid methods used for the identification of the condition? 7. Was the condition measured in a standard, reliable way for all participants? 8. Was there an appropriate statistical analysis? 9. Was the response rate adequate, and if not, was the low response rate managed appropriately [15]. To be included in the meta-analysis, articles were required to meet at least 5 of the nine criteria. Quality appraisal table can be found in (Additional file 1).

Data extraction

Relevant data was extracted from eligible studies and entered into Microsoft Excel. Data collected included authors, year of publication, study period, study area, sample size, study design, our outcome of interest (retention, loss to follow up, transfer out and death), and length of follow up period.

Measurement of outcomes

Reported rates of retention in care, loss to follow-up, death and transfer out were collected from included studies. The rate of retention in care was defined as the rate of persons in the study who remain in care across periods of follow-up which was no missed visit of clinical schedule for more than three consecutive months. Study attrition was defined as the number of patients who were lost to follow-up, transferred out, or died [16]. While LTFU was defined as HIV positive patients who miss scheduled visits to the clinic for more than three consecutive months after the last visit [17]. Death and transfer out were defined as confirmed in the patient’s record by the clinicians who were in charge of care [18]. In order to facilitate analysis, all rates were changed into100 person years of observation.

Where available, statistically significant measures of association were collected for factors associated with retention in care or attrition from care identified by studies included in the study. Then pooled odds ratio was calculated using frequency values from the primary studies and identified factors were classified as being either socio-demographic or clinical factors associated with patient retention or attrition.

Heterogeneity and publication bias

Study heterogeneity was assessed by calculating I2 test statistics. Statistical significance was set at p < 0.05 I2 scores were classified into low, moderate and high inconsistency based on the I2 value of less than 25, 50 and 75% respectively [19]. Random effect analysis was carried out to determine the pooled estimates of patient retention and attrition. Publication bias was assessed using Egger’s test. For meta-analysis result with statistically significant publication bias, the Duval and Tweedie nonparametric trim and fill analysis using the random effect method of analysis was conducted [20].

Statistical methods and analysis

Data analysis was conducted using Stata version 14. Pooled prevalence of LTFU, death and transfer out were estimated using each study prevalence and standard error with 95% confidence interval (CI). For LTFU, we used both pooled prevalence and incidence density separately since most of the primary studies were follow up studies. Separate meta-analysis was conducted for the LTFU, death and transfer out outcomes. The result of meta-analysis was presented using forest plots. The meta-analysis was conducted using the random effects model of analysis since it minimizes heterogeneity of the included studies [21]. Subgroup analysis was conducted for the incidence rate of loss to follow up. Studies were stratified into subgroups based on length of follow up,

Results

A total of 3910 records were identified through online database. Of these, 1820 were identified through PubMed 900 were identified through Google Scholar, 980 were identified through Google Search and 200 from AJOL. An additional 2 articles were identified through unpublished sources. A total of 3912 were screened for inclusion. One hundred twenty were found to be duplicates and removed. Three thousands seven hundred ninety two were screened for inclusion and 3717 were excluded by title and abstract review. The full text of 75 articles were assessed for eligibility. Of these, only 45 articles passed the minimum quality score of 5 out of 9 points and were included in the meta-analysis. A quality appraisal table can be found in (Additional file 1).

Characteristics of included studies

Most (32) of the designs of the primary articles was retrospective follow up [9, 10, 16,17,18, 22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48], five case control [8, 49,50,51,52], six prospective cohort study [37, 53,54,55,56,57,58] and two cross sectional survey designs [59, 60]. Every region of Ethiopia was represented in the included studies. A total of 546,250 study participants were included from 45 articles Fig. 1. Moreover, study characteristics can be found in Table 1.

Fig. 1
figure 1

Studies included in systematic review and meta-analysis of HIV patients attrition and its determinants in Ethiopia

Full size image
Table 1 Characteristics of included articles (n = 45)
Full size table

Magnitude of retention in HIV care

Of the 45 original articles, 32 articles reported magnitude of patient retention in HIV care in Ethiopia [8,9,10, 17, 23,24,25,26,27,28, 30,31,32, 34, 36, 37, 41,42,43,44,45, 48, 52,53,54,55,56,57,58, 60, 62, 63]. The magnitude of retention in HIV care was 70.65% (95% CI, 68.19, 73.11) with I2 value of 99.9% at P-value < 0.001. The minimum retention was 32.50% (95% CI, 28.56 to 36.44) in a 9 years follow up period [24] and the maximum retention was 94% (95% CI, 90.1 to 97.7) in 1 year. The result is given in Fig. 2.

Fig. 2
figure 2

Magnitude of HIV patients retention of care in Ethiopia from 2005 to 2019

Full size image

The maximum and minimum follow up period was 14 years and 3 months respectively. Except one article that used 3 months of follow up [27], all original studies used a follow up period of above 1 year. Most (19) of the studies that reported a follow up period from 2 to 5 years had average retention in care of 71.09% (95% CI, 68.75,73.44). Furthermore, 9 articles that used more than 5 years follow up period that reported on average retention rate of 68.92% (95% CI, 65.02, and 72.81) Fig. 3.

Fig. 3
figure 3

Magnitude of HIV patients’ retention by years of follow up in Ethiopia from 2005 to 2019

Full size image

The magnitude of attrition was an aggregations of LTFU, transfer out and death. Accordingly, the magnitude of loss to follow up was 15.17% (95% CI: 11.86, 18.47). The lowest and highest rate of loss to follow up was 3.13% (95% CI: 2.90, 3.36) and 31.40% (95% CI: 31.37, 31.43) Fig. 4.

Fig. 4
figure 4

Magnitude of loss to follow up of HIV patients in Ethiopia from 2005 to 2019

Full size image

Magnitude of attrition rate of HIV/AIDS patients in Ethiopia

Incidence of loss to follow up was 13.79/100 person years of observation (95% CI, 9.66–17.93). The incidence rate was highest in the follow up years 2 through 5 years which was 16.69/100 person years of observation, (95% CI, 6.77–26.61). Subgroup analysis of incidence rate of loss to follow up was reported by 12 papers with resulted cumulative incidence density of 13.79/100 person years of observation (95% CI, 9.66–17.93). It is higher in the follow up years ranged from 2 to 5 years 16.69/100 person years of observation, (95% CI, 6.77–26.61) Fig. 5.

Fig. 5
figure 5

Incidence rate of loss to follow up among HIV patient on treatment in Ethiopia

Full size image

Most common contributing factors for patient attrition was loss to follow up (15.17%), transfer out (12%) and death (5.71%) respectively Table 2.

Table 2 Summery of retention and attrition rates among HIV patients attending clinical care in Ethiopia from 2005 to 2019
Full size table

Mortality of HIV patients in Ethiopia

Overall mortality of HIV patients in this meta-analysis was 6.75% (95% CI: 6.22, 7.27). Mortality rates were lowest in the first years of follow up (4.4% (95% CI: 0.72, 8.16)) and highest when follow up exceeded 5 years (7.65% (95% CI: 6.38, 8.92)). Pooled mortality rates are shown in Fig. 6.

Fig. 6
figure 6

Magnitude of HIV patient mortality in Ethiopia from 2005 to 2019

Full size image

HIV patients transfer out from one to other health facility

About 11.17% (95% CI: 7.12, 15.21) of all study subjects transferred out during the study period. The cumulative transfer out percentage was lowest in the earliest years of follow up, and highest by the end of the follow-up period. 5.18% (95% CI: 1.60, 8.76) and 18.35% (95% CI: 13.60, 23.11) Fig. 7.

Fig. 7
figure 7

Magnitude of transfer out of HIV patients in Ethiopia from 2005 to 2019

Full size image

Factors related to patient retention and attrition

Socio-demographic and behavioral factors

Five original papers stated that marital status was associated with patient retention in clinical care. Of which three articles reported that married people were less likely to be lost to follow up [8, 30, 51] and the probability of dropout for patients with separated marital status was about 16.82% higher than those patients with marital status divorced [41]. Another article [44] reported as single marital status as an independent predictor of death. Pooled odds ratio from 13 primary studies indicated that currently unmarried people are 1.52 times more likely not to be retained in clinical care (OR, 1.52, 95% CI, 1.15–2.01). Moreover, ten papers reported that male patients were less likely to be retained in HIV care in Ethiopia [8, 16, 23, 30, 33, 36, 53, 55, 59, 60]. Yet, one study in Ethiopia reported that females were less likely to be retained in clinical care [26]. However, it was not statically significant from pooled analysis of odds ratio from18 primary studies (OR, 1.12, 95% CI, 0.97–5.45). Similarly, eight papers [17, 22, 23, 28, 30, 36, 49, 54] reported that young age was an important predictor of loss to follow up but this was not statistically significant in the pooled analysis (OR, 0.88, 95% CI, 0.27–2.89). Four papers reported that educational status had an influence on retention in care [10, 30, 41, 54]. All except one [41] primary articles has reported that people with no or primary education were more likely to experience loss to follow up compared to those with secondary and above educational level. Two studies reported that urban residents were better in retention than their rural counterparts [22, 49]. Three papers indicated that economic constraints such as being dependent patients for source of food and daily laborers were risk factors for treatment interruption [10, 49, 51]. Being a merchant, farmer, daily labour and jobless had a greater risk of dropout [38, 41]. On the other hand, one study reported that merchants were at higher risk of death [45] However, except marital status, none of the socio-demographic factors were found statistically significant for patient retention in care (Table 3).

Table 3 Socio-demographic factors associated with HIV/AIDS patients’ attrition from care in Ethiopia (2005–2019)
Full size table

HIV patients who did not disclose their HIV status were 6.36 times more likely to experience attrition from clinical care (OR, 6.36, 95% CI, 3.58–11.29). Patients who were not substance users were 59% more likely to be retained in clinical care (OR, 0.41, 95% CI, 0.17–0.98). Mental health issues such as taking hard drugs (cocaine, cannabis and IV drugs) and excessive alcohol drinking were risk factors for treatment defaulting [52]. Any types of substance use was reported as risk for loss to follow up [38]. It was also reported that patients with higher score for stress were poor in ART adherence [57]. Moreover, patients with mental status being not at ease were at increased risk for loss to follow up [51] (Table 3).

Clinical related factors

Fourteen studies [10, 18, 23, 24, 31, 35, 38, 43, 44, 54, 57,58,59,60] indicated that poor functional status (ambulatory or bed ridden) was a factor for patient attrition (OR, 2.11, 95% CI 1.33–3.34). Twelve papers [17, 18, 22, 23, 30, 43,44,45, 49, 54, 58, 59] reported that patients with advanced WHO clinical stages III or IV were also at greatest risk of death and loss to follow up (OR, 1.85, 95% CI, 1.36–2.51). Seventeen papers [8, 10, 27, 29,30,31, 35, 38, 43,44,45, 49, 55, 57,58,59,60] reported that baseline CD4 count less than 200 cells/ μL and greater than 350 cells/ μL was reported as risk factor for attrition but it was not significant in the pooled odds ratio analysis (OR, 1.09, 95% CI,0.49–2.41). Four papers reported that lower level of hemoglobin is a risk factor for death [18, 39, 43, 44]. Four studies reported that lower weight and BMI is risk factor for patient death and loss to follow up [38, 44, 45, 57].

Treatment related factors such as baseline ART regimen of Zidovudine-Lamivudine-Nevirapine (AZT-3TC-NVP) was key factor for patient attrition [22, 33]. Six studies [18, 22, 24, 38, 43, 57] demonstrated that poor drug adherence was a risk factor for patient attrition of both loss to follow up and death (OR, 6.60, 95% CI 1.41–30.97). Three studies reported that adverse drug side effects are responsible for patient loss to follow up and death [18, 39, 57]. Six studies reported that opportunistic infection were the main responsible factors for patient death and loss to follow up includes [38, 39]. These included TB [39, 44,45,46] and diarrhea for more than 3 months led a risk of death [18]. Nine papers reported that provision of isoniazid preventive therapy (IPT) [17, 18, 35, 38, 39, 50, 54] and CPT [25, 29] prophylaxis indicted that optimize patient retention.

Two papers reported on quality of care indicated that patients whose next appointment weren’t recorded were at risk of loss to follow up [8, 16]. Three papers reported that loss to follow up was higher among HIV patients in hospitals compared to health centers [28, 37, 54] while, one study reported that loss to follow up was higher among patients attending health centers than those in hospitals [53]. On the other hand, two studies reported contrary result one as higher death rate in health centers than in hospitals and other is vice versa [37, 54]. The pooled odds ratio indicated that there is no difference on level of retention based on health facility (OR, 1.23, 95% CI, 0.58–2.62) Table 4.

Table 4 Clinical factors associated with HIV/AIDS patients’ attrition from care in Ethiopia (2005–2019)
Full size table

Discussion

This systematic review and meta-analysis investigated available evidence on the magnitude and associated factors of HIV patient retention and attrition in clinical care in Ethiopia.

The study indicated that the pooled prevalence of HIV patients’ clinical retention was 70.65%; and attritions was 15.17% loss to follow up, 6.75% death, and 11.17% transfer out. Furthermore, the incidence of loss to follow up was 13.79 person years of observation. Factors such as being currently unmarried, non-disclosed of HIV status, history of poor drug adherence, poor functional status, presence of opportunistic infections, lower BMI, substance use, lower hemoglobin and advanced WHO clinical stages were significantly associated with patient attrition.

Levels of HIV patient retention in antiretroviral treatment program in Ethiopia (70.65%) was comparable to the report from a study in India (70.7%) [64]. On the other hand, it was lower than patient retention in Asia (80%) [4], Sub Saharan Africa (77.5%) and Cape Town, South Africa 94% [65,66,67], Anambra, Nigeria (80%) [68], KwaZulu-Natal, South Africa (77.5%) [69], Khayelitsha, South Africa (85.9%) [5] and Southeastern United States (91.7%) [70]. This could be explained by the inclusion in our analysis of studies with longer follow-up periods. The longer the follow up period the lower the rates of clinical retention [69]. In addition, other possible reasons for the observed differences include the fact that countries have different levels of infrastructure, HIV burden, and HIV associated stigma that can cause loss to follow and transfer out to other health facilities [71].

On the other hand, the level of patient retention in this study is higher than that observed studies conducted among HIV infected patients from Africa (65%), Latin America and Caribbean (64%) [4], Asia-Pacific Region [72], Tanzania (25%) [73] and South Africa [74]. In Africa, Asia, and Latin America that the value ranged from 3.1 to 45.1% [75]. The reason might be the differences in cut off point for loss to follow up (LTFU) and treatment eligibility criteria of the latter two studies. Now a days, immediate initiation of ART following HIV positive test result may contributed for good patient retention [66, 74, 76,77,78]. This might be due to test and treat strategy is effective in patient retention since it decreases the chance of loss to follow up from determining treatment eligibility to adherence preparation [79, 80].

In this study, rate of HIV patients lost to follow up, death and transfer out is similar with study report form India [64] but higher than study from Australia and Asia [81]. Loss to follow up was the major cause of attrition, followed by death and transfer out. This is in line with a previous systematic review in sub-Saharan Africa [67].

On the other hand, rate of lost to follow up (13.79 per 100 person years of observation) was lower than observed from previous studies from Asia Pacific region (21.4 person years of observation) [72], Malawi (26 and 48 per 100 person years of observation for pre-ART and ART patients respectively), Guinea-Bissau (51.1 per 100 person-years of observation) [82] and study conducted among children in Ethiopia 29.7% [7]. The reason can be due to difference of study population and the current study was composite of both earlier and recent times study findings while the previous study was conducted at the beginning of the treatment. Moreover patient retention was better in Asia than Africa from previous study [61] that reflects country specific difference in terms of patients retention. Moreover, the pooled death rate was 6.75% which is lower than previous systematic review 5–40% of death in Ethiopia [83]. The variation may be due to small sample size in the previous study and different time of follow up.

Our review identified a number of socio-demographic and clinical factors that could be targets for future interventions. It is indicated that currently unmarried people (never married, separated, divorced and widowed) were 1.52 times more likely not to be retained in care. This is similar with a study in South Africa that reported women without regular partner were not retained in clinical care [69]. This might be because of married people get support from their spouse to adhere to their treatment. Likewise, those who did not disclose their HIV status to someone were 6.36 times more likely not to be retained in care. This can be explained by the fact that those who disclosed their sero-status can get support from someone who knew the HIV status and this facilitates regular treatment attendance. This might be due to people who had perceived HIV related stigma are more likely to experience treatment attrition [13] . Therefore, partner testing, HIV status disclosure and mutual support are important components of HIV intervention [3, 84]. In contrast to these findings, one study in Nigeria showed that married women were poor in clinical retention [68]. These differences may be due to difference by cultural practices and the attitude of people on the role of gender in both countries.

Substance abuse and mental distress have negative effects on HIV patient retention in clinical care. This finding is in line with a systematic review in developing countries [85]. This might be due to poor drug adherence as a result of distressed mental status. Retention in care was significantly higher in patients treated by mental health trained healthcare workers [23]. This can be explained by the fact that the decision making ability of patients abusing substances is altered and this facilitates poor treatment adherence [13]. Inline to this, the previous studies also reported that patients with no bereavement concern were less likely to experience loss to follow up while patients with mental status not at ease were at increased risk of defaulting from treatment [50, 51].

It was evidenced that advanced clinical stage such as poor baseline functional status, WHO clinical stage III or IV, suboptimal ART adherence, presence of one or more opportunistic infections and being underweight had higher risk of patient attrition. Such findings are similar to findings from studies in Ethiopia [83], Malawi ([86], Guinea-Bissau [82] and Asia Pacific region [72]. The reasons for patients to be non-adherent were forgetfulness, side effects, feeling sick and running out of medication [87].

Limitation of the study

Original articles used different length of follow up period that had variable outcomes of attrition. Moreover, studies conducted across different treatment guidelines revision may affect patient retention and treatment outcome.

Conclusion

About two third of HIV patients were retained in care. The most common cause of attrition among HIV positive people in Ethiopia were loss to follow up, transfer out and death respectively. It can be concluded that Ethiopia has long to walk to achieve the minimum targets of patient retention achieved by most of low income countries. There were increased attrition rate for longer period of follow up that need due attention from clinicians. Due to change of treatment eligibility criteria from time to time and varying definitions of loss to follow up, it is necessary to reconsider the definitions of LTFU. Moreover, tracing studies are necessary to rule out final destination of loss to follow up patients. Social support is necessary for all HIV patients and encouraging disclosure to family members. This also implied HIV patients require mental health interventions in addition to the medical model of treatment. Patients with poor drug adherence, advanced HIV disease stage, opportunistic infection, poor functional status (ambulatory or bedridden), underweight and anemic patients need special attention that includes adequate adherence preparation and nutritional intervention. In general it was evidenced that number of socio-demographic determinants prone patients for loss to follow up while the clinical factors leads to death. Hence, in order to achieve the desired level of patient retention in clinical care, a comprehensive interventions which are targeted at socio-demographic, clinical, laboratory and behavioral factors is necessary.