Incidence and predictors of loss to follow-up among adults receiving ART in Metekel Zone, Northwest Ethiopia: a retrospective cohort study

Abstract

Background

Loss to follow-up (LTFU) remains a significant global public health challenge, particularly among patients receiving antiretroviral therapy (ART). There is paucity of information on incidence and predictors of LTFU in the Metekel zone, which limits the ability to monitor the effectiveness intervention.

Objective

To assess the incidence and identify the predictors of loss to follow-up among adult people living with HIV at public health facilities in the Metekel Zone, Northwest Ethiopia.

Methods

A retrospective cohort study was conducted among 540 people living with HIV who enrolled in ART from June 28, 2017, to June 27, 2022, in the Metekel Zone. The study was conducted in one public hospital and eight public health centers. Patient records were randomly selected and extracted using a structured checklist. The Kaplan-Meier curves were compared using the log-rank test, and the Cox proportional hazard model was used to identify the independent predictors of LTFU. The strength of the association was measured using adjusted hazard ratio (AHR) along with 95% confidence intervals (CIs).

Results

Among 540 participants, 304 (56.3%) were females and the mean (SD) age participants was 35.7(12.41) years. The overall LTFU incidence rate was 15.12 (95% CI: 12.9–17.6) per 100 person-years. Tuberculosis infection at baseline (AHR: 2.12 95% CI: 1.33 − 3.38), underweight at ART initiation (AHR: 5.36, 95% CI: 3.46–8.31), poor ART adherence (AHR: 2.14, 95% CI: (1.32 − 3.46), and most recent viral load > 1000 copies/mL (AHR: 2.71, 95% CI: 1.55 − 4.72) were identified significant predictors of LTFU.

Conclusion

The observed LTFU incidence rate among adults receiving ART in Metekel Zone was notably higher than documented in prior studies. Hence, enhanced monitoring and support strategies should target patients with baseline tuberculosis infection, poor treatment adherence, underweight at baseline, and high viral load, particularly within the first year of treatment initiation. In addition, effective strategies to improve patient retention in HIV care should be designed considering the identified risk factors.

Introduction

Retention in HIV care is a critical component of successful antiretroviral therapy (ART) programs. Loss to follow-up (LTFU) in antiretroviral therapy (ART) is an administrative classification used when a patient has not been reported as dead or transferred out and has not attended a clinic or collected ART medications for three consecutive months or more after the last scheduled appointment [1]. Loss to follow-up remains a major challenge in human immunodeficiency virus (HIV) care, undermining efforts to improve treatment outcomes, reduce new infections, and address health inequities [2].

The rapid scale-up of ART has significantly improved the survival and quality of life among people living with HIV (PLWH) by reducing HIV-related mortality [3]. As ART coverage has expanded, particularly in resource-limited settings, the number of individuals receiving treatment has increased markedly [4]. However, retention in care remains a critical gap across the HIV treatment cascade, with attrition occurring at multiple stages of care [5, 6].

The burden of LTFU varies across geographic settings. Evidence from Asian countries reports LTFU levels of approximately 9% [6], while higher estimates have been observed in sub-Saharan Africa, with pooled proportions around 15.07% [7], and even higher levels reported in specific countries such as Nigeria (23.9%) [8]. In Ethiopia, incidence estimates show variation ranging from 3.7 [9] to 26.6 per 100 person-years [5]. Overall, these variations reflect substantial heterogeneity in LTFU across contexts, likely driven by variations in health system performance, patient characteristics, and service delivery models.

Loss to follow-up has serious clinical and public health consequences. It reduces the immunological and survival benefits of ART and increases the risk of AIDS-related morbidity, mortality, and hospitalization. It also associated with poor adherence, drug toxicity, treatment failure, and development of drug resistance [10, 11]. In addition, LTFU reverses the preventive benefits of ART by increasing the risk of HIV transmission and hindering progress toward the global goal of ending the HIV pandemic [12]. It also represents an inefficient use of limited health system resources [13, 14].

Studies have identified determinants of LTFU, including socio-demographic factors such as age, sex, educational status, marital status, occupational status, residence [5, 10, 15]; clinical and treatment-related factors such as cluster of differentiation (CD4) count, viral load, hemoglobin, disease stage, functional status, ART regimen, adherence, nutritional status, and opportunistic infections [13, 16–18]; and psychosocial and health system-related factors such as stigma and discrimination, non-disclosure of HIV status, lack of treatment supporters, long waiting times, and distance to health facilities [16–19]. However, these findings remain inconsistent across settings.

In Ethiopia, evidence on LTFU is highly variable, and most studies based on single-facility data, limiting generalizability [10–21]. In addition, contextual factors such as conflict, insecurity, and pastoralist mobility in regions like Benishangul Gumuz may further complicate retention in care by disrupting service delivery, limiting access to health facilities, and increasing patient movement [21]. These factors are likely to influence both the incidence and predictors of LTFU, yet they remain insufficiently explored in multi-facility studies.

Therefore, this study aimed to assess the incidence and predictors of LTFU among adult PLWH receiving ART at public health facilities in the Metekel Zone. By incorporating data from multiple facilities in a conflict-affected and pastoralist setting, the study provides more contextually relevant evidence on retention challenges. The findings are expected to inform policymakers and program implementers in designing targeted, context-specific, and cost-effective interventions to improve ART retention and treatment outcomes.

Methods

Study design, period and setting

A retrospective cohort study was conducted in public health facilities in the Metekel Zone of Benishangul-Gumuz Regional State, Ethiopia. Benishangul-Gumuz Regional State is administratively divided into three administrative zones, one of which is Metekel zone. The Metekel Zone shares its border with Kamashi to the south and southwest, Sudan to the west, and the Amhara region to the north and east. The zone’s administrative center, Gelgel Belles, located approximately 540 km northwest of Addis Ababa. Metekel zone is administratively structured into seven districts and one town administration. According to the 2007 Ethiopian national census, the zone had a total population of 484,118, of whom 256,030 were males. The majority of the populations (278,752) resides in rural areas and primarily depends on agriculture for their livelihood.

The Metekel zone has three hospitals (one general hospital and two primary hospitals), 21 health centers, and 168 health posts. ART services are delivered by one general hospital and eight health centers based on the revised 2018 national guidelines. At the time of study, 3,187 individuals were actively receiving ART services based on the report Zonal Health Office, with adults comprising 96% patients. Charts of adults registered in ART were abstracted between June 20, 2022, and July 5, 2022. The study focused on data from June 28, 2017, to June 27, 2022, a period during which the updated standardized formats, documentation, and recording systems were fully and consistently implemented.

To enhance retention in care, the ART program employs several patient tracing and re-engagement mechanisms, including telephone follow-up by case managers and adherence supporters, home visits by community adherence supporters and networks of PLWH, appointment calendar or reminder systems, and mother support groups for pregnant women. During ART initiation, patients’ contact information, including residential address, telephone number, and treatment supporter details, is routinely collected to facilitate tracing of missed appointments.

Population

The source population consisted of all PLWH aged 15 years and older who received HIV care at public health facilities in Metekel Zone. The study population comprised of a randomly selected PLWH aged 15 years and older who were on HIV care at selected public health facilities between June 28, 2017, and June 27, 2022.

Eligibility criteria

The study included all PLWH aged 15 years and older who had at least one month of follow-up after initiating ART between June 28, 2017, and June 27, 2022. Patients with incomplete baseline and/or follow-up data, particularly on socio-demographic variables such as the date of ART initiation and/or the date of the last visit were excluded. Similarly, patients transferred from other health facilities were excluded if their record lacked complete baseline and follow-up information.

Sample size determination and sampling procedures

The sample size was determined using STATA version 15, considering the following assumptions: a 5% level of significance, 80% statistical power, probability of an event of 0.116, an adjusted hazard ratio of 2.09 for WHO clinical stage IV from previous study [21], and a 10% for potential withdrawals. Based on these parameters, the final calculated sample size was 554.

At the time of the study, one hospital and eight health centers were providing HIV care services in Metekel Zone. Of these facilities, Pawi General Hospital was included purposively because of its high patient load, while three health centers, namely Debati Health Center, Gilgel Belles Health Center, and Manbuk Health Center were selected randomly.

A total of 716 PLWH were enrolled in ART care in the selected facilities during the study period, of which 471 were from Pawi General Hospital, 76 from Debati Health Center, 92 from Gilgel Belles Health Center, and 78 from Manbuk Health Center. A separate sampling frame was prepared for each selected health facility using the unique identification numbers of all PLWH enrolled in ART from June 28, 2017, to June 27, 2022. The final sample size of 554 was then allocated proportionally to each facility based on the number of PLWH enrolled in ART. Finally, study participants were selected from each facility using a simple random sampling technique in Microsoft Excel by applying the random number function (RAND). Accordingly, 364 participants were selected from Pawi General Hospital, 59 from Debati Health Center, 71 from Gilgel Belles Health Center, and 60 from Manbuk Health Center.

Data collection procedures

The data abstraction tool was developed by adapting content from standardized ART intake, registration, and follow-up forms currently in use at ART clinics [22], as well as through review of relevant literatures [5, 10, 20, 23]. Data sources included patient intake forms, follow-up cards, ART registers, and the electronic database. Additional clinical documents such as laboratory test results were used to extract the CD4 cell and viral load counts. Individual patient cards or electronic records were identified using District Health Information System 2 (DHIS2) identification numbers. Data collection was conducted by one health officer and three clinical nurses using standardized abstraction tool.

Study variables and operational definitions

The outcome variable of this study was LTFU, defined as a patient missing ART refills for three months or more from their last scheduled visit, without being recorded as ‘dead’ or ‘transferred out in their medical chart (1). Patients who died, transferred out, or remained in care at the end of the study period were considered censored. Participants were followed from ART initiation until the first occurrence of LTFU or a censoring event. Individuals who subsequently re-engaged in care after an LTFU episode were not followed further, as only the first LTFU event was considered in the survival analysis.

The predictor variables were baseline socio-demographic characteristics (age, sex, marital status, educational level, religious, place of residence, occupational, presence of a primary caregiver, distance from health facilities, and availability of registered phone number). Clinical baseline variables included WHO clinical stage, functional status, CD4 cell count, viral load, HIV disclosure status, active tuberculosis (TB). Follow-up clinical and treatment-related factors included presence of opportunistic infections (OIs), duration of follow-up, initiation of isoniazid (INH) and cotrimoxazole preventive therapy (CPT), body mass index (BMI), type of ART regimen, presence of anemia, and level of adherence to ART.

Functional status was classified according to national ART guidelines: working–able to carry out routine tasks both at home and outside; bedridden–unable to perform activities for a daily living; ambulatory–able to perform some daily tasks [24]. Self-reported ART adherence was measured by ART clinicians during a patients’ routine follow-up appointment based on missed doses and categorized as follows: good adherence (≥ 95%)–missing fewer than 3 doses out of 30 or fewer than 4 doses out of 60; fair adherence (85 – <95%)–missing 3–5 doses out of 30 or 4–9 out of 60; and poor adherence (< 85%)–missing more than 6 doses out of 30 or missing more than 9 out of 60 [25]. A patient was classified as having disclosure HIV status if he/she informed at least one person, other than healthcare providers, about his/her HIV status [4].

Data processing and analysis

Data were checked for completeness and consistency, entered into Epi Data Manager 4.4, and subsequently exported to STATA version 15 for analysis. Descriptive statistics summarized the data, using frequencies and percentages for categorical variables, and mean with standard deviation for normally distributed continuous variables or median with interquartile range for skewed continuous variables.

Kaplan-Meier (KM) curves were employed to estimate survival time and to compare survival probabilities across different variable categories. The Log-rank test (Mantel-Cox test) was used to determine statistically significant difference in survival rate between categories of variables. To identify the predictors of LTFU among patients receiving ART, a Cox proportional hazard model was performed. Variables with a p value of less than 0.25 in the bivariable Cox regression were included in the multivariable Cox regression analysis. The strength of the association was measured using adjusted hazard ratio (AHRs) along with 95% confidence intervals (CIs) and a p value less than 0.05 was considered statistically significant. The proportional hazards assumption of the model was assessed using the global Schoenfeld residual test and was satisfied (p = 0.273).

Data quality management

Data quality was maintained through careful development of the data abstraction tool, and selection of data collectors and supervisors based on their prior experience. A pretest was conducted on 5% of patient charts (27 charts) at Pawi general Hospital to evaluate the clarity and consistency of the tool, and the necessary modifications were made. Additionally, one-day training was provided for data collectors and supervisors on data collection tool and data collection procedures. Supervisors and the principal investigator monitored the data collection process daily.

Ethical consideration

The study was approved by the Institutional Review Board (IRB) of Jimma University Institute of Health (JUIH) with reference number IHRPEJ1060/22. Permission to conduct the study was obtained from the Metekel Zone Health Department and the heads of selected health facilities. Since the study used secondary data from patient medical charts, informed consent was waived. The study purpose was explained to coordinators and ART staff at health facilities prior to data collection. A unique identification number was assigned for each participant after data extraction to ensure confidentiality.

Results

Socio-demographic characteristics of the participants

A total of 554 records of PLWH receiving ART were reviewed, of which 540 met eligibility criteria, yielding a response rate of 97.5%. The mean age of the participants was 35.7 (SD = 12.41) years, with 29.3% falling within the 25–34 age group. Females accounted for 56.3% of the cohort, and 79.3% resided in rural areas. Concerning educational attainment, 227 (42%), had completed primary school, and more than half (57.6%) were married. Most of the patients were farmers (45.7%), and a significant majority (88.5%) had disclosed their HIV status. The majority (94%) of participants had mobile phone. Most of participants (68.7%) traveled over five kilometers the health facility, and 96.4% had caregiver (Table 1).

Table 1.

Baseline socio-demographic characteristics of PLWH on ART at public health facilities in Metekel Zone, Northwest Ethiopia, 2022

Variables	Categories	Frequency	Percent
Age	18–24 years	122	22.6
	25–34 years	158	29.7
	35–44 years	130	24.1
	≥ 45 years	130	24.1
Sex	Male	236	43.7
	Female	304	56.3
Residence	Urban	428	79.3
	Rural	112	20.7
Marital status	Single	152	28.1
	Married	311	57.6
	Divorced	59	10.9
	Widowed	7	1.3
	Separated	11	2
Religion	Orthodox	347	64.3
	Muslim	140	25.9
	Protestant	40	7.4
	Catholic	13	2.4
Level of education	No education	161	29.8
	Primary	227	42
	Secondary	134	24.8
	Tertiary	18	3.3
Distance of ART clinic	< 5 KM	169	31.3
	≥ 5 KM	371	68.7
HIV disclosure status	Disclosed	478	88.5
	Not disclosed	371	68.7
Occupational status	Driver	15	2.78
	Daily laborer	49	9.1
	Merchant	110	20.4
	Farmer	247	45.7
	Government	62	11.5
	Self employed	38	7
	Othersa	19	3.5
Partner HIV status	Known	475	87.96
	Unknown	65	12.04
Availability of care giver	Yes	521	96.4
	No	19	3.6
Telephone contact	Yes	508	94
	No	32	6

^ahouse wife, soldiers, and sex-workers

Baseline clinical and laboratory characteristics of PLWH on ART

Among participants, 315 (58.3%) were linked to ART care from voluntary counseling and testing (VCT). The majority (83.7%) started ART with a working functional status. One-quarter (25%) were categorized at WHO clinical stage III or IV. Regarding immunological status, 15.4% of participants had a most recent CD4 count below cells/mL. In terms of hemoglobin profile, 490 individuals (90.7%) initiated ART with hemoglobin levels above 10 g/dL. At baseline, 43 (7.9%) of participants had tuberculosis, and 35.4% had an opportunistic infections (OIs). Among those with OIs, pneumonia was most common (38.5%), followed by TB (35.5%) and herpes zoster (10.7%) (Table 2).

Table 2.

Baseline clinical and laboratory related characteristics of PLWH on ART at public health facilities in the Metekel Zone, Northwest Ethiopia, 2022

Variables	Categories	Frequency	Percent
ART linkage point	VCT	315	58.3
	OPD	117	21.7
	IPD	9	1.7
	MCH	36	6.7
	Others	63	11.7
Recent CD4 count	< 350 cells/ml	83	15.4
	≥ 350 cells/ml	457	84.6
Baseline functional status	Functional	452	83.7
	Ambulatory	78	14.4
	Bedridden	10	1.8
WHO clinical stage at baseline	Stage I	394	72.96
	Stage II	8	1.48
	Stage III	100	18.52
	Stage IV	38	7.04
Presence of opportunistic infection	Yes	191	35.4
	No	349	64.6
TB at baseline	Yes	43	7.96
	No	497	92.4
History of TB treatment	Yes	45	8.33
	No	495	91.67
Viral load count at 6 month	Not eligible/<6month	89	16.4
	≤ 1000 copies/mL	423	78.3
	> 1000 copies/mL	28	5.2
Baseline Hemoglobin	≤ 10 g/dL	50	9.3
	> 10 g/dL	490	90.7

Follow-up clinical and ART treatment-related characteristics of participants

Among the participants, 59.8% were on the TDF-3TC-EFV (1e) regimen, and 112 (20.7%) subsequently switched from their initial regimen. The main reasons for changing regimens were availability of new drugs, adverse drug reactions, and virologic failure. Regarding ART adherence, 23 participants (4.26%) had fair adherence, while 39 participants (7.22%) had poor adherence. At baseline, 243 participants (45%) were underweight (BMI < 18.5 kg/m2), and 60% had utilized the appointment spacing model (ASM) (Table 3).

Table 3.

Follow-up and ART treatment-related characteristics of PLWH on ART at public health facilities in the Metekel Zone, Northwest Ethiopia, 2022

Variables	Categories	Frequency	Percent
IPT treatment at baseline	Yes	330	61.1
	No	210	38.9
Body mass index	< 18.5 kg/m2	243	45
	≥ 18.5 kg/m2	297	55
Cotrimoxazole prophylaxis at baseline	Yes	338	62.6
	No	202	37.4
Initial ART regimen	TDF+3TC + EFV	323	59.81
	AZT+3TC + EFV	5	0.93
	TDF+3TC + DTG	212	39.6
Experience of regimen change	Yes	112	20.7
	No	428	79.2
Reason for change (n = 112)	Due to toxicity	12	10.7
	Virology failure	11	9.8
	New drug	89	79.4
Adherence to ART	Good	478	88.52
	Fair	23	4.26
	Poor	39	7.22
Appointment spacing model	Yes	324	60
	No	216	40

Incidence rate of loss to follow-up

A cohort of 540 PLWH receiving ART contributed a total of 1,037.0 person-years of follow-up, with a median follow-up period of 1.72 years (IQR = 0.87–2.84). At the end of the follow-up period, 157 patients (29.1%; 95% CI: 25.38–33.06) were classified as LTFU. The overall incidence density rate of LTFU was 15.12 per 100 person-years (95% CI: 12.76–17.49). Incidence rates of LTFU varied by duration of treatment, with highest rate observed during the first year: 17.76 per 100 person-years (95% CI: 14.22–22.14), followed by 16.90 (95% CI: 12.84–22.13) in the second year, 9.6 (95% CI: 5.92–15.50) in the third year, and 6.42 (95% CI: 2.62–15.14) in the fourth year. The rate increased in the fifth year, reaching 9.74 per 100 person-years (95% CI: 2.42–38.91).

The incidence of LTFU was slightly higher among females (15.72 per 100 person-years) compared to males (14.41 per 100 person-years). Similarly, a higher incidence was observed among rural residents compared to urban residents (18.57 vs. 14.25 per 100 person-years). However, there was no meaningful variation in the incidence of LTFU across age groups. Incidence of LTF was higher among patients with secondary educational status (17.80 per 100 person-years), without caregiver (34.20 100 person-years) and no telephone contact 25.50 per 100 person-years) (Table 4).

Table 4.

Incidence of loss to follow-up by selected sociodemographic charactersitics among PLWH enrolled to ART clinic at public health facilities in Metekel zone, Northwest Ethiopia, 2022

Variable	Category	Events (LTFU)	Person-years	Incidence rate per 100 person-years
Sex	Male	68	472.0	14.41
	Female	89	566.0	15.72
Age group	15–24	35	222.0	15.77
	25–34	49	310.0	15.81
	34–44	36	255.0	14.12
	≥ 45	37	251.0	14.74
Residence	Urban	118	828.0	14.25
	Rural	39	210.0	18.57
Education	No education	40	314.0	12.70
	Primary	71	443.0	16.00
	Secondary	44	247.0	17.80
	Tertiary	2	33.7	5.94
Care/support	Yes	147	1009.0	14.60
	No	10	29.3	34.20
Telephone contact	Yes	137	960.0	14.30
	No	20	78.5	25.50

Kaplan–Meir survival analysis showed that most of LTFU events occurred within the earlier years following ART initiation, with a marked declined in losses in later years (Fig. 1).

Fig. 1.

The Kaplan–Meier estimate of the loss to follow-up among PLWH enrolled to ART care at public health facilities in Metekel zone, Northwest Ethiopia from 2017–2022

The incidence of LTFU had shown differences with having TB at baseline. The incidence of LTFU among participants with TB at baseline was higher than the respondents with no TB at baseline (Fig. 2).

Fig. 2.

The Kaplan–Meier estimate of the loss to follow-up by TB at baseline among PLWH enrolled to ART clinic at public health facilities in Metekel zone, Northwest Ethiopia, 2022

Figure 3 presented the probability of LTFU among participants with sub-optimal status of fair and poor adherence was higher than the participants with good adherence (Fig. 3).

Fig. 3.

Kaplan–Meier estimates of loss to follow-up by adherence level among PLWH on attending the ART clinic at public health facilities in Metekel zone, Northwest Ethiopia, 2022

Predictors of loss to follow-up from ART

Both bivariable and multivariable Cox regression analyses were performed to identify predictors of LTFU. In bivariable analysis, variables with p-value < 0.25, including residence, disclosure status, partner’s HIV status, presence of caregiver, telephone contact, distance from health facility, baseline OIs and TB, history of TB treatment, taking of cotrimoxazole preventive therapy at baseline, treatment adherence, body mass index (BMI), and recent CD4 and viral load status were identified as candidate variables for inclusion the multivariable model. After adjusting for potential confounders, the predictors of LTFU were baseline tuberculosis (AHR = 2.12; 95% CI: 1.33 − 3.38), poor treatment adherence (AHR = 2.14; 95% CI: 1.32–3.46), being underweight at baseline (AHR = 5.36; 95% CI: 3.46–8.31), and elevated recent viral load (≥ 1000 copies/mL) (AHR = 2.71; 95% CI: 1.55 − 4.72) (Table 5).

Table 5.

Bivariable and multivariable Cox regression analysis of predictors of LTFU among PLWH receiving ART at public health facilities in the Metekel Zone, Northwest Ethiopia, 2022

Factors	Category	Outcome		cHR (95% CI)	aHR (95% CI)	P value
		LTFU, (157)	Censored, (383)
Residence	Urban	118	310	1	1
	Rural	39	73	1.30(0.90–1.87)	1.23(0.82–1.83)	0.30
Disclosure status	Disclosed	130	348	1	1
	Not disclosed	27	35	1.77(1.16–2.68)	1.40(0.85–2.30)	0.18
Partner HIV status	Known	132	343	1	1
	Unknown	25	40	1.59(1.03–2.44)	1.66(0.98–2.89)	0.06
Presence of care giver	Yes	147	374	1	1
	No	10	9	2.15(1.13–4.09)	1.04(0.51–2.13)	0.89
Telephone contact	Yes	137	364	1	1
	No	20	19	1.78(1.11–2.84)	1.15(0.67–1.99)	0.60
Distance from health facility	< 5 KM	35	134	1	1
	≥ 5KM	122	249	1.67(1.15–2.44)	1.14(0.76–1.70)	0.51
OIs at baseline	Yes	70	121	1.44(1.05–1.98)	1.40(0.95–2.05)	0.08
	No	87	262	1	1
TB at baseline	Yes	26	17	2.94(1.93–4.49)	2.12(1.33–3.38)	< 0.001*
	No	131	366	1	1
History of TB treatment	Yes	36	21	2.50(1.65–3.79)	1.44(0.91–2.29)	0.11
	No	121	362	1	1
Baseline CPT treatment	Yes	76	262	1	1
	No	81	121	2.15(1.57–2.95)	1.27(0.88–1.82)	0.19
Adherence status	Good	126	352	1	1
	Fair	8	15	1.64(0.80–3.35)	1.40(0.66–2.97)	0.38
	poor	23	16	3.05(1.95–4.77)	2.14(1.32–3.46)	0.02*
Body mass index (kg/m2)	< 18.5	129	114	7.40(4.90–11.1)	5.36(3.46–8.31)	< 0.001*
	>=18.5	28	269	1	1
Recent CD4 (cells/mm)	< 350	35	48	1.96(1.34–2.86)	1.26(0.84–1.89)	0.25
	≥ 350	122	335	1	1
Recent viral load (copies/mL)	< 1000	138	374	1	1
	≥ 1000	19	9	4.06(2.51–6.58)	2.71(1.55–4.72)	< 0.001*

1: reference category; ^* P value < 0.05

Discussion

Decentralized care and widespread of ART access greatly increased treatment availability, yet LTFU remains the major public health issue in HIV care. In this study, we assessed the incidence and predictors of LTFU among PLWH on ART at public health facilities in Metekel Zone, Northwest Ethiopia. The overall incidence of LTFU was 15.1 per 100 person-years. Presence of tuberculosis infection at baseline, poor treatment adherence, underweight, and elevated recent viral load (≥ 1000 copies/mL) were significant predictors of LTFU.

The observed incidence of LTFU in this study aligns with the prior studies conducted in Ethiopia and other settings [4, 10, 20, 26]. This high incidence may be attributed to the long distance from the health facility. As reported in Tables 5 and 122 (77.70%) of the LTFU participants travelled more than five kilometers to access healthcare services. Moreover, rural residence, lack of social support and absence of telephone contact may have contributed to the high incidence of the LTFU. In addition, contextual factors such as internal conflicts and pastoralist mobility may also have contributed for the observed high incidence. This finding is consistent with studies conducted in University of Gondar comprehensive specialized hospital (12.26 per 100 PY) [10], Pawi general hospital, Northwest Ethiopia (11.6 per 100 PY) [21], Bahir Dar city, Northern Ethiopia (15 per 100 PY) [27]. The high incidence underscores ongoing challenges related to patient retention despite decentralization of the service.

In addition, the finding revealed a pronounced pattern of LTFU, with the highest incidence occurring during the first year of ART initiation (17.8 per100 person-years). Early LTFU may reflect initial adverse effects, stigma encounter during the early disclosure or inadequate patient preparation at ART initiation. It highlights the need for context-specific interventions to prevent early LTFU.

On the other hand, the current incidence of LTFU is notably higher than studies conducted in Arba Minch general hospital (5.3 per 100 PY) [28], University of Gondar Comprehensive hospital (10.90 per 100 PY) [4], Debre Markos referral hospital 3.7 per 100 PY) [9], Woldia, Northeast Ethiopia [29] and Agaro town, Southwest Ethiopia (4.7 per 100 PY) [30]. This discrepancy might be due to several contextual factors. First, the Meketel Zone has been affected by repeated unrest and is considered a war prone-region. Ongoing insecurity and high levels of internal displacement likely disrupted health seeking-behavior, ART continuity, and the actual reporting of death versus LTFU. Second, it is possible that some patients transferred to other facilities without notifying the original care provider, leading to their misclassification as lost to follow-up. Third, as a predominantly pastoralist region, Metekel has considerable population movement and long travel distances to facilities, which are known predictors of higher LTFU rates [31, 32]. Finally, the disruption of public health services due to COVID-19 pandemic in the Metekel Zone during 2020–2021 may have contributed to the observed high incidence of attrition during this period.

In contrast, the current finding is lower than those reported in Jigjiga town, Eastern Ethiopia (26.6%) [5], sub-Saharan Africa (24.6%) [33], 26.7 per 100 person-years in Uganda, and 23.9% in Nigeria [23, 34]. This discrepancy may be attributed to the recent rapid scale-up and accessibility of effective antiretroviral therapy. Additionally, the difference in sample size, duration of follow-up, and the sociodemographic characteristics of the study participants might contribute for this variation. Furthermore, the low rate observed in this study might be influenced by the implementation of the “test and start” strategy introduced in Ethiopia in 2016 [25]. This finding highlights the critical need for close monitoring PLWH, particularly during the first six months following ART initiation.

In this study, patients who had TB infection at baseline had a 2.12-fold higher hazard LTFU compared to those without TB at baseline. This finding aligns with previous studies conducted in Axum, Ethiopia [35], and sub-Saharan Africa [36]. A possible explanation is that TB/HIV co-infected patients often face overlapping drug toxicities due to the use of multiple medications. Adverse reactions to antiretroviral therapy are common, and there are significant overlaps in the toxicity profiles of the first-line anti-tuberculosis and antiretroviral drugs [37, 38]. Additionally, TB/HIV co-infected patients face a high pill burden, which can challenges adherence to treatment. Conversely, a study from Indonesia reported no significant association between TB infection and LTFU after ART initiation [39], possible due to stronger implementation of the TB-HIV collaborative programs and more consistent monitoring of co-infected patients. This finding highlights the importance of baseline TB screening, close monitoring for drug-related side effects, and comprehensive counseling for patients starting ART.

This study also revealed that undernourished patients with a BMI below 18.5 kg/m² had a 5.36 times hazard of LTFU compared those with a BMI of 18.5 kg/m² or higher. This result is consistent with studies conducted in southern Ethiopia, Northwest Ethiopia and Malawi [10, 31, 40]. The increased risk of LFTU among undernourished patients may be attributed to the complex interplay between HIV infection and nutritional status. HIV significantly compromises nutritional status through multiple pathways: it suppresses appetite, impairs nutrient absorption, increased the body’s energy requirements, and increases vulnerability to opportunistic infections [41]. These effects can weaken patients physically and mentally, making it difficult for them to maintain regular clinic visits and adhere to lifelong treatment regimens. Additionally, malnutrition is often a proxy indicator for poverty and food insecurity, which may further contribute to poor retention in care due to transportation challenges or competing survival priorities. Moreover, the increased risk of undiagnosed tuberculosis and mortality may account for the greater loss to follow-up observed among individuals with low BMI, given that undernutrition is a recognized risk factor for both outcomes [42, 43]. This finding underscores the critical need for routine nutritional screening and support as an integral part of ART services. Implementing regular nutritional assessments, providing therapeutic and supplementary feeding programs, and linking patients to social support systems are essential strategies for improving retention and treatment outcomes among undernourished PLWH.

In line with previous studies [3, 40], 44– [29], patients with poor adherence had a 2.14-times higher hazards of LTFU compared to those with good adherence. Adherence to ART is a cornerstone of successful HIV treatment, as sustained adherence is essential for achieving viral suppression, preventing drug resistance, and maintaining long-term engagement in care [46, 47]. Patients with poor adherence may be more likely to miss clinic appointments and attend follow-up visits irregularly, thereby reducing opportunities to receive adherence counseling, treatment support, and reinforcement of positive health behaviors [48]. Consequently, they may be less likely to develop the treatment literacy and self-efficacy required for sustained treatment engagement. Evidences suggest that treatment literacy and self-management skills improve through continuous interaction with healthcare providers and accumulated treatment experience, which may explain why patients with limited treatment experience are particularly vulnerable to poor adherence and subsequent disengagement from care [49, 50].

Furthermore, inadequate adherence can result in insufficient drug exposure, viral replication, virologic failure, and disease progression, all of which may increase the risk of disengagement from care and eventual LTFU [46, 51]. These findings suggest that interventions aimed at improving retention should address both patient-level and health system-level barriers through strengthened adherence counseling, early identification of missed appointments, and continuous patient education. Given that sustained adherence is a primary determinant of viral suppression, the observed association between poor adherence and LTFU may also help explain the increased risk of LTFU among patients with viral non-suppression observed in this study.

The study revealed that viral suppression level was a significant predictor of loss to follow-up from HIV care. Patients with a recent viral load greater than 1000 copies/mL had a 2.71-fold higher hazard of loss to follow-up compared to those with viral load ≤ 1000 copies/mL, after adjusting for confounders. This finding is in line with studies conducted in Ethiopia [10, 29] and Asia [52] that reported a strong association between viral suppression and continued engagement in HIV care. It is well-known fact that viral suppression is a key indicator of treatment success and effective ART adherence, whereas viral non-suppression often reflects underlying challenges such as poor adherence, treatment fatigue, psychosocial barriers, or treatment failure [53, 54]. Patients with unsuppressed viral loads are at increased risk of immunologic decline, advanced HIV disease, and opportunistic infections, which may contribute to worsening health status, and eventual disengagement from care [55]. However, the association between non-suppressed viral loads and LTFU is most probably bi-directional. While viral non-suppression may increase vulnerability to treatment interruption through disease progression and poor clinical outcomes, interruptions in ART adherence and clinic attendance may lead to viral rebound and subsequent non-suppression [54, 56]. Thus, viral non-suppression may serve both as a consequence of suboptimal retention and as a predictor of future disengagement from care. This finding contributes to the growing body of evidence by highlighting the dynamic interplay between virologic outcomes and retention in care within decentralized ART programs. It underscores the importance of strengthening enhanced adherence support (EAS), timely viral load monitoring, and targeted retention strategies for patients with unsuppressed viral loads. Expanding access to viral load testing and ensuring prompt management of viral non-suppression may improve both treatment outcomes and long-term retention in HIV care.

Strengths and limitations

The strength of this study is that it was conducted after the implementation of the updated guidelines, which indicate the status of retention on ART after the guidelines. The study also included multicenter facilities that provide ART service for PLWH. Moreover, the study included viral load, which is a key indicator for ensuring adherence and viral suppression can be considered strength.

However, the results should be interpreted with caution in light of several limitations. The retrospective nature of the study required reliance on secondary data, and 14 records with incomplete data were excluded from the analysis, which may resulted in misclassification of LTFU status and introduced selection bias. Furthermore, some exposure variables may have changed over time, but only the recorded values available in the medical charts were analyzed, potentially leading to exposure misclassification. Adherence to ART was measured using self-reported pill counts, which may have led to biased estimates of adherence due to social desirability bias among patients. In addition, psychosocial factors such as HIV-related stigma and discrimination, depression, substance use, perceived health status and opting for traditional medicine were not assessed. Moreover, the study did not consider system-level factors such as geographic and transportation cost, treatment disruption, and satisfaction of patients with care provision.

Conclusion

The incidence of LTFU among PLWH receiving ART was considerably high, with most LTFU occurring within the first year of ART initiation. Patients at higher risk of LFTU include those with baseline TB infection, undernutrition, poor ART adherence, and elevated recent viral load.

These findings indicate that LTFU is a multifactorial problem driven by clinical, behavioral, and health-system factors, particularly during the early phase of treatment when patients are still adapting to long-term care. Therefore, program implementers and healthcare providers should prioritize early and targeted interventions for high-risk patients. Strengthening TB–HIV integrated care is essential to ensure early detection and coordinated management of co-infected patients. Nutritional assessment and support programs should be incorporated into routine HIV care to address undernutrition among vulnerable clients. In addition, adherence monitoring should go beyond self-reported measures by incorporating objective approaches such as pharmacy refill tracking, appointment-keeping records, and, where feasible, digital adherence tools. Differentiated service delivery models, including community ART delivery, multi-month dispensing, and flexible appointment spacing, should be scaled up to improve retention, particularly for stable patients. Special attention should also be given to rural residents and mobile populations through tailored follow-up mechanisms, outreach services, and strengthened patient tracking systems.

Policy makers should emphasize expanding decentralized HIV services and ensuring universal access to viral load testing and integrated chronic care, as these are critical to improving retention outcomes.

Future research should examine contextual and psychosocial determinants of LTFU, including social support, mental health, substance use, and mobility, and evaluate the effectiveness of differentiated service delivery models in similar settings. In addition, conducting prospective study is recommended to better establish temporal relationships and casual pathways.

Abbreviations

AIDS

Acquired immunodeficiency syndrome

aHR

Adjusted hazard ratio

ART

Antiretroviral therapy

CD4

Cluster of differentiation

CI

Confidence interval

cHR

Crude hazard ratio

CPT

Cotrimoxazole prophylaxis therapy

HIV

Human immunodeficiency virus

IPD

Inpatient department

LTFU

Loss to follow-up

OIs

opportunistic infections

OPD

outpatient department

PLWH

People living with HIV

TB

tuberculosis

VCT

Voluntary Counseling and Testing

WHO

World Health Organization

Author contributions

TD was involved in the conception, study design, execution and acquisition of data, analysis, and interpretation of data. CF and AL made significant contributions to the study design, analysis, and interpretation of the data. All authors made substantial contributions to reviewing the manuscript, agreed to submit to the current journal, and approved the manuscript.

Funding

This research did not receive any fund.

Data availability

The datasets used in this study are available from the corresponding author and are available upon request.

Declarations

Ethics approval and consent to participate

The study was approved by the Institutional Review Board (IRB) for health research at Jimma University with reference number Ref. No IHRPEJ1060/22.

Consent for publication

Not applicable.

Clinical Trial Number

Not applicable.

Data sharing statement

The datasets used in this study are available from the corresponding author and are available upon request.

Competing interests

The authors declare no competing interests.